Measure Trials

Foreword
In response to concerns over the lack of information available on seepage from open
channel supply systems, in October, 1998, the Australian National Committee on
Irrigation and Drainage (ANCID) conducted a two-day workshop. The workshop was
held at Moama in southern New South Wales and had major support from the Murray
Darling Basin Commission, Land and Water Australia, the Commonwealth
Department of Primary Industries and Energy and 16 other industry organisations.
The workshop brought together 90 stakeholders and experts in the field of channel
seepage from throughout Australia.
The key outcomes from the workshop were a suite of recommendations seeking new
and extensive investigations aimed at improving the level of knowledge about channel
seepage.
In response to the recommendations, ANCID formed an industry Task Force to

advance the investigations. It has developed a three-stage project designed to
implement the recommendations.
Each stage of the project is briefly described as follows:
Stage 1 This project will investigate best practice, easy to use standards to be used in
identifying, measuring and quantifying channel seepage.
Stage 2 This project is aimed at providing best practice procedures and processes involved in
undertaking remedial work to seal channels suffering from seepage.
Stage 3 This project is designed to provide an easy to use User Support System needed to assist
industry in making decisions on whether or not to undertake what is often very
expensive remedial works on seeping channels.
This three-staged project is now well underway and will involve a total expenditure of
close to $2.5 million. Stage 1 has now been completed and Stages 2 and 3 are
scheduled for completion in December, 2003.
The major outcomes from each of the Stages of the project work will be in the form of
reports and Best Practice Guidelines Manuals. This report is one of the suite arising
from the project. It documents all of the field trials undertaken in Stage 1 of the
project, and provides the technical underpinning for the Stage 1 Best Practice
Guidelines Manual.
I would like to also acknowledge the significant support and funding provided to this
project by the Murray Darling Basin Commission, Land and Water Australia and
several rural water authorities and natural resource management agencies. Without
their valued support and interest, the project and this report would not have been
possible.
Stephen Mills
Chairman
ANCID
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE i

Publication Details
Published by Australian National Committee on Irrigation and Drainage (ANCID) c/-
Goulburn-Murray Water, PO Box 165, Tatura, Victoria, Australia, 3616.
This report is one in a series detailing the outcomes of a three-stage project investigating the
measurement, remediation and associated decision making for channel seepage.
Acknowledgements
This project would not have been possible without the generous investment of the Murray
Darling Commission through its Strategic Investigation and Education program and additional
significant investment by the following organisations:
Sunwater, Queensland
Goulburn-Murray Water
Land and Water Australia
Murray Irrigation
Murrumbidgee Irrigation
Southern Rural Water
Wimmera Mallee Water
There has also been wide interest in this study and significant input has been provided by a
wide and diversified range of interested people for which ANCID is very appreciative.
This document has been prepared on behalf of ANCID by Sinclair Knight Merz Pty Ltd.
Disclaimer
This report is published by ANCID on the basis that recipients of it would make their own
inquiries and obtain appropriate professional advice before relying on any information or any
expression of opinion or predication contained in this report. Neither ANCID nor any of the
contributing agencies is responsible for the results of any action taken on the basis of
information in this report nor for any errors or omissions in this report.
Copyright
This work is copyright. Photographs, cover artwork and logos are not to be reproduced, copied
or stored by any process without the written permission of the copyright holders or owners.
All commercial rights are reserved and no part of this publication covered by copyright may be
reproduced, copied or stored in any form or by any means for the purpose of acquiring profit or
generating monies through commercially exploiting (including but not limited to sales) any
part of or the whole of this publication except with the written permission of the copyright
holders.
However, the copyright holders permit any person to reproduce or copy the text and other
graphics in this publication or any part of it for the purposes of research, scientific
advancement, academic discussion, record-keeping, free distribution, educational use or for
any other public use or benefit provided that any such reproduction or copy (in part or in
whole) acknowledges the permission of the copyright holders and its source (Development of
Guidelines for the quantification and monitoring of seepage from earthen channels) is clearly
acknowledged.
To the extent permitted by law, the copyright holders (including its employees and consultants)
exclude all liability to any person for any consequences, including but not limited to all losses,
damages, costs, expenses and any other compensation, arising directly or indirectly from using
this report (in part or in whole) and any information or material contained in it.
The contents of this publication do not purport to represent the position of ANCID, Murray-
Darling Basin Commission or Land and Water Australia in any way and are presented for the
purpose of informing and stimulating discussion for improved management of the Basin's
natural resources.
WC01312:YR3_RPT_CH_1.3(FORE. EXEC SUMM, INTRO).DOC Final PAGE ii

Contents
Foreword
Executive Summary
1. Introduction...............................................................................................1
1.1 Background.........................................................................................1
1.2 Content and Scope of this Report ......................................................2
2. Overview of Trials Conducted..............................................................3
2.1 Technique Selection ...........................................................................3
2.2 Trial Program Summary Table ...........................................................4
2.3 Techniques Not Included in the Trials Program.................................4
2.3.1 Inflow-Outflow Tests .................................................................. 4
2.3.2 Mathematical Modelling.............................................................. 4
2.3.3 Hydrochemical Techniques and Tracing of Leakage Plume .......... 5
2.4 Assessment Methodology...................................................................7
3. Description of Trial Sites .......................................................................8
4. Non-Geophysical Techniques........................................................... 11
4.1 Pondage Tests..................................................................................11
4.1.1 Methodology............................................................................ 11
4.1.2 Results.................................................................................... 14
4.1.3 Discussion............................................................................... 14
4.1.4 Conclusions............................................................................. 16
4.2 Sub-surface Characterisation: Soil and Geological Profiling ...........19
4.2.1 Purpose of sub-surface profiling................................................ 19
4.2.2 Data collection ......................................................................... 19
4.2.3 Results.................................................................................... 21
4.2.4 Conclusions............................................................................. 24
4.3 Point Tests ........................................................................................25
4.3.1 Introduction ............................................................................. 25
4.3.2 Methodology............................................................................ 26
4.3.3 Ring Infiltrometer and Disc Permeameter – Toolondo Central ..... 26
4.3.4 Ring Infiltrometer and Disc Permeameter – Dahwilly Central....... 33
4.3.5 Idaho Seepage Meter - Dahwilly Central.................................... 37
4.3.5.1 Comparison with Pondage Test Results ............................... 38
4.3.5.2 Comparison with EM31 Results ........................................... 40
4.3.6 Ring Infiltrometer - Tabbita ...................................................... 41
4.3.7 Idaho Seepage Meter – Donald Main Channel ........................... 46
4.3.7.1 Comparison with Pondage Test Results ............................... 46
4.3.7.2 Comparison with EM31 Results ........................................... 48
4.3.8 Conclusions............................................................................. 49
4.4 Groundwater Techniques .................................................................51
4.4.1 Introduction ............................................................................. 51
4.4.2 Methodology............................................................................ 51
4.4.3 Donald Main Channel............................................................... 51
4.4.3.1 Monitoring Bore Set-Up ....................................................... 51
4.4.3.2 Groundwater Level Response.............................................. 51
4.4.3.3 Seepage Estimates ............................................................. 55
4.4.4 Tabbita.................................................................................... 58
4.4.4.1 Transect 1.......................................................................... 58
4.4.4.2 Transect 2.......................................................................... 62
4.4.4.3 Summary ............................................................................ 62
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE iii

4.4.5 Conclusions............................................................................. 63
4.5 Remote Sensing ...............................................................................64
4.5.1 Summary................................................................................. 65
5 Geophysical Techniques.................................................................... 66
5.1 Introduction .......................................................................................66
5.1.1 Theory of using geophysics to identify and quantify seepage ...... 66
5.1.2 Methodology............................................................................ 66
5.1.2.1 Years 1 and 2 Trials ............................................................ 67
5.1.2.2 Year 3 Trials ....................................................................... 68
5.1.3 Description of Trial Sites........................................................... 70
5.2 EM34 Trial Results ...........................................................................73
5.2.1 Introduction ............................................................................. 73
5.2.1.1 EM34 Systems ................................................................... 73
5.2.1.2 Methodology ....................................................................... 74
5.2.2 Results.................................................................................... 74
5.2.2.1 Rocklands – November 1999 and August 2001..................... 75
5.2.2.2 Donald Main – October 1999 and September 2001 ............... 77
5.2.2.3 Toolondo Central – August 2001 .......................................... 82
5.2.2.4 Dahwilly Central and Dahwilly East – February 2002. ............ 83
5.2.3 Conclusions............................................................................. 88
5.3 EM31 Trials .......................................................................................89
5.3.1 Introduction ............................................................................. 89
5.3.2 Background ............................................................................. 89
5.3.2.1 EM31 Systems ................................................................... 89
5.3.2.2 Methodology ....................................................................... 90
5.3.3 Year 1 and 2 – EM31 ............................................................... 91
5.3.3.1 Toolondo Central – December 2000, August 2001................. 92
5.3.3.2 Rocklands – August 2001 ...................................................101
5.3.3.3 Donald Main – September 2001 .......................................110
5.3.3.4 Dahwilly Central – June 2000..............................................117
5.3.3.5 Lake View – June 2000 ......................................................119
5.3.3.6 Tabbita – June 2000...........................................................123
5.3.3.7 Conclusions .......................................................................125
5.3.4 Regional Assessment of Key Relationships ..............................130
5.3.4.1 Multiple Linear Regression..................................................130
5.3.4.2 Simple Linear Regression...................................................136
5.3.5 Channel Specific EM31 Statistical Assessment and Extrapolation
of EM31 Relationships .......................................................................142
5.3.5.1 Introduction........................................................................142
5.3.5.2 Toolondo and Rocklands ....................................................142
5.3.5.3 Donald...............................................................................147
5.3.5.4 Dahwilly Channel ...............................................................148
5.3.5.5 Finley ................................................................................150
5.3.5.6 Lake View..........................................................................152
5.3.5.7 Summary ...........................................................................155
5.4 Resistivity Trials ..............................................................................156
5.4.1 Introduction ............................................................................156
5.4.2 Background ............................................................................156
5.4.2.1 Channel Specific Resistivity Systems ..................................156
5.4.2.2 Methodology ......................................................................156
5.4.3 Regional Assessment of Key Relationships ..............................158
5.4.3.1 Multiple Linear Regression..................................................158
5.4.3.2 Simple Linear Regression...................................................162
5.4.4 Channel Specific Resistivity Assessment .................................167
5.4.4.1 Toolondo West ..................................................................167
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE iv

5.4.4.2 Toolondo Central................................................................170
5.4.4.3 Toolondo East ...................................................................174
5.4.4.4 Dahwilly Central .................................................................177
5.4.4.5 Dahwilly East (Pretty Pine).................................................181
5.4.4.6 Finley ................................................................................183
5.4.4.7 Lake View Central ..............................................................185
5.4.4.8 Lake View West.................................................................189
5.4.5 Extrapolation at each Channel .................................................192
5.4.5.1 Toolondo...........................................................................192
5.4.5.2 Dahwilly.............................................................................194
5.4.5.3 Lake View..........................................................................195
5.5 Conclusions ....................................................................................197
5.5.1 Discussion and Conclusions ....................................................197
5.5.1.1 Seepage Detection Mechanisms.........................................197
5.5.1.2 Comparison of Trialed Geophysical Techniques ...................199
5.5.1.3 Critical Geophysical Survey Variables .................................202
5.5.1.4 Repeatability......................................................................203
5.5.1.5 Regional Assessment of Key Relationships .........................203
5.5.1.6 Confidence in Derived Relationships and Extrapolation of
Results .........................................................................................210
5.5.1.7 Preferred Methodology .......................................................211
5.5.1.8 International Developments in Geophysics and Channel
Seepage Measurement .......................................................................216
5.5.2 Summary of EM34 Results ......................................................217
5.5.4 Summary of Resistivity Results................................................229
6. Waranga Western Channel: Case Study......................................233

6.1 Study Objectives .............................................................................233
6.2 Outline of Work Undertaken ...........................................................233
6.3 Site Conditions ................................................................................234
6.3.1 Geology and Hydrogeology .....................................................234
6.3.2 Channel Conditions.................................................................234
6.4 EM31 Survey and Initial Drilling......................................................234
6.4.1 Description of EM31 and Drilling Program.................................234
6.4.2 Analysis .................................................................................237
6.4.3 Initial Identification of High Priority Areas ..................................238
6.5 Additional Drilling Program .............................................................239
6.5.1 Description.............................................................................239
6.5.2 Refinement of High Priority Areas ............................................239
6.6 Pondage Tests................................................................................242
6.6.1 Description.............................................................................242
6.6.2 Analysis .................................................................................243
6.6.3 Final Refinement of High Priority Areas ....................................247
6.7 Conclusions ....................................................................................248
7. Discussion of Results .......................................................................250
7.1 Overview .........................................................................................250
7.1.1 Assessment Methodology........................................................250
7.1.2 Description of Trial Sites..........................................................251
7.2 Pondage Tests................................................................................251
7.3 Sub-surface Characterisation.........................................................252
7.4 Point Tests ......................................................................................252
7.5 Groundwater Techniques ...............................................................254
7.6 Remote Sensing .............................................................................255
7.7 Geophysics .....................................................................................255
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE v

7.7.1 General Conclusions ...............................................................256
7.7.1.4 Repeatability......................................................................261
Results .........................................................................................268
7.8 Waranga Western Channel Case Study ........................................283
8. Conclusions.........................................................................................285
8.1 Overview .........................................................................................285
8.2 Pondage Tests................................................................................285
8.3 Sub-surface characterisation..........................................................285
8.4 Point Tests ......................................................................................286
8.5 Groundwater Techniques ...............................................................286
8.6 Remote Sensing .............................................................................286
8.7 Geophysics .....................................................................................287
8.7.1 General Conclusions ...............................................................287
8.7.1.4 Repeatability......................................................................289
Results .........................................................................................290
8.8 Waranga Western Channel Case Study ........................................296
9. Recommendations .............................................................................297
10. References ...........................................................................................299
Appendix A Geological Long Sections...........................................302
A.1 Wimmera Mallee Water Geological Long Sections........................303
A.2 Murrumbidgee Irrigation Geological Long Sections .......................308
A.3 Murray Irrigation Limited Geological Long Sections ......................312
Appendix B Pondage Test Results..................................................316
B.1 Wimmera Mallee Water Geological Long Sections........................316
st th
B.1.1 Donald Main Pondage Tests: 21 - 29 December, 2000..........316
th st
B.1.2 Toolondo Central Pondage Tests: 19 March – 1 April, 2001 ...317
th rd
B.1.3 Toolondo Central Pondage Test: 19 March – 3 April, 2002.....318
B.1.4 Toolondo East Pondage Tests: 20th March – 3rd April, 2002....319
th rd
B.1.5 Toolondo West Pondage Tests: 19 March – 3 April, 2002.....320
th th
B.1.6 Rocklands Pondage Tests: 15 - 29 March, 2001....................321
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE v i

B.2 Murrumbidgee Irrigation Pondage Test Results ............................322
th th
B.2.1 Tabbita Pondage Tests: 18 – 25 June, 2001 ..........................322
th nd
B.2.2 Lake View Central Pondage Tests: 25 June – 2 July, 2001....323
th th
B.2.3 Lake View Central Pondage Tests: 26 June – 7 July, 2002 ....324
B.2.4 Lake View West Pondage Tests: 26th June – 7th July, 2002......325
B.3 Murray Irrigation Limited Pondage Test Results............................326
th th
B.3.1 Dahwilly Pondage Tests: 4 – 18 June, 2001...........................326
th th
B.3.2 Finley Pondage Tests: 6 – 20 July, 2001................................327
B.3.3 Dahwilly Pondage Tests: 6th– 21st June, 2002 .........................328
th
B.3.4 Dahwilly East Pondage Tests: 30 May – 21st June, 2002 ........330
st th
B.3.5 Finley Pondage Tests: 31 May – 19 June, 2002 ....................331
B.4 Goulburn Murray Water Pondage Test Results .............................332
B.4.1 Waranga Western Channel......................................................332
Appendix C Bulk vertical hydraulic conductivity calculations 334

Appendix D Summary of Project Brief for Remote Sensing in the
Wimmera ............................................................................................335
Appendix E Statistical Output from Regional Assessment
Regression Analysis.................................................................................338
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE vii

Document History and Status
Rev. Date Reviewed By Approved By Revision Details
Distribution of copies:
Copy No. Quantity Issued To
Printed: 10 April, 2003

Last Saved: 6 February, 2003
File Name: I:\Wcms\Wc01312\060-Reports\REP00_01.10\Year 3 Reports\Yr3_Rpt_Ch_1-3(Fore, Exec Summ,
Intro).Doc
Author: Stephen Parsons
Project Manager: Paul Bolger
Name of Organisation: Murray Darling Basin Commission / Australian National Committee on Irrigation
and Drainage
Name of Project: Stage 1 - ANCID Channel Seepage Measurement Project
Name of Document: Final Year Channel Seepage Measurement Report - Chapters 1 to 4
Document Version: Final
Project Number: WC01312.060
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE viii

Executive Summary
E.1 Introduction
As the driest inhabited country in the world, Australia is dependent on its water
resources. One of the main mechanisms for the transport and delivery of water is via
earthen channels. Recent surveys have indicated that around 4% of the total water
supplied for rural use is lost due to channel seepage (ANCID, 2000b). Seepage from
earthen channels has therefore become an important issue in Australia for several
reasons, including the loss of an economically valuable resource and the contribution
of seepage water to land degradation issues such as salinity and waterlogging.
The Australian National Committee of Irrigation and Drainage (ANCID), in

conjunction with the Murray Darling Basin Commission (MDBC), initiated a three-
stage project to provide best practice information on channel seepage measurement
(Stage 1) and remediation (Stage 2) and to develop a suitable user support system
(Stage 3). An international literature survey on channel seepage measurement and an
Australia wide channel seepage survey of more than 40 rural water authorities have
been conducted as part of the Stage 1 investigation (ANCID, 2000a and ANCID,
2000b). This report documents the three years of field trials which were conducted as
part of the Stage 1 investigation.
Based on the Stage 1 trials, literature review and RWA survey, a guidelines manual
for channel seepage measurement has also been developed (ANCID, 2003). The
guidelines are intended to be for practical use in undertaking channel seepage
investigations across the Australian water industry. The guidelines are to be linked to
the channel seepage user support system (in progress) which provides a structured
management tool for channel managers.
Channel seepage measurement trials were conducted from early 2000 to mid 2002,
within Wimmera Mallee Water (WMW), Murray Irrigation Limited (MIL) and
Murrumbidgee Irrigation (MI). In addition, results from channel seepage
measurement investigations conducted on the Waranga Western Channel (by
Goulburn Murray Water) were incorporated into the final year of trials.
E.2 Overview of Trials Conducted

E.2.1 Technique Selection
The main channel seepage measurement techniques referred to in the literature, and
those discussed in the literature review (ANCID, 2000a) include:
q Pondage Tests
q Point measurement (channel full and empty)
q Geophysical Techniques
q Groundwater Techniques
q Soil Classification
q Remote Sensing
q Inflow - Outflow
q Mathematical Modelling
q Hydrochemical / Isotopic Mass Balance
q Tracing Leaking Plume
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE i

The most important criteria for selecting techniques suitable for channel seepage
measurement and identification are cost and accuracy. Significantly, RWAs rank cost
as the most significant factor in selecting seepage investigation techniques, with
technical accuracy of lesser importance (ANCID, 2000b). This finding was of
fundamental importance to the development of the trial program, and was the reason
why some techniques were not tried at all and why others became the focus of the
program.
Based on the outcomes of the literature review (ANCID, 2000a), the RWA survey
(ANCID, 2000b), and consideration of the primary objectives of the study, the trials
were focussed on the first six of the techniques listed above. The early trial program
covered all of these six techniques. The final year of the program was based on the
results from the first two years of trials. In order to maximise the usefulness of the
output of the trial program, the final year of trials was focussed on geophysics, which
demonstrated the greatest potential for meeting RWA requirements for channel
seepage assessment.
E.2.2 Trial Program Summary Table

Table E-1 summarises the trials conducted during the program. Pondage tests were
conducted at all sites, as they were the basis on which other techniques were assessed.
Drilling was also conducted at all sites in order to identify sub-surface conditions.
Remote sensing is included in the table even though trials were not undertaken as part
of the study. Available data was assessed but deemed unsuitable for use in the project.
E.2.3 Techniques Not Included in the Trial Program

The following techniques were not included in the trials:
q Inflow-Outflow Tests - These tests are not sufficiently accurate for measuring
losses over relatively short sections of channel (ie 1-2km). Over relatively long
lengths of channel this is an appropriate technique, and therefore the technique is
suitable for identifying and prioritising, at an Authority-wide level, channels
which have higher losses compared to others in the system.
q Mathematical Modelling - The intensity of data collection and level of specialist
input required does not make this method practical for use by RWAs for most
channel seepage investigations.
q Hydrochemical Techniques and Tracing of Leakage Plume - The high cost of such
trials means they are generally not practical solutions for RWAs.
For additional information on these techniques refer to the Literature Review

(ANCID, 2000a) and the Guidelines Manual (ANCID, 2003).
E.2.4 Assessment Methodology

In undertaking these channel seepage investigations, the basic approach adopted was:
q Identification of test site locations;
q Gathering available information on test sites;
q Measuring rates of seepage at test sites using direct measurement techniques -
pondage tests were used for this purpose;
q Comparison of the direct measurement technique with indirect techniques; and,
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE ii

q Extrapolation of results beyond the test zone to interpret seepage distribution - this
was applied for techniques which compared favourably with the direct technique.
It is well documented in the literature that while every channel seepage method has
certain disadvantages, almost universally pondage tests are regarded as the most
accurate method of quantifying seepage (ANCID, 2000a). Therefore the basic method
of assessment of the accuracy of each technique adopted in the trial program was by
comparison against pondage test data.
E.3 Description of Trial Sites

The seepage investigation sites all lie within the Murray Darling Basin (Figure E-1).
The channels investigated were main delivery channels, ranging in capacity from 80
ML/d (Tabbita) to 600 ML/d (Rocklands). With respect to lithology, sites ranged
from a clay profile, to a sand profile, as well as sites with rock at or near the surface.
Groundwater salinity ranged from moderately fresh to highly saline. Groundwater
depths ranged from very shallow (0.5 - 1.5m) to moderately deep (9-10m). Channel
dimensions were reasonably similar, with the depth of water at full supply level (FSL)
typically 1.5m and wetted perimeters of between 9-16m.
n Figure E-1 Trial Site Regional Location Map
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE iii

n Table E-1 ANCID Channel Seepage Measurement Project - Trials Summary Table
Technique
Date Conducted
Rural Water Channel
Pondage Tests Geophysics
Authority Sub-Surface Point Tests Groundwater Remote Sensing 1
Profiling Techniques
EM31 EM34 Resistivity
(all land based) (all based on-channel)
Toolondo March 01 (6 cells) Dec. 00 (land)

March 02 Dec. 00 Dec. 00 / Jan. 01 - Sept. 00 ?
(Central) March 02 (1 cells) Aug. 01 (land & boat) Aug. 01
June 02 (ring infiltrometer & disc
March 02 (land) permeameter)
Wimmera Toolondo (East) March 02 (4 cells) March 02 (land) - March 02 June 02 - - -

Mallee
Water Toolondo (West) March 02 (4 cells) March 02 (land) - March 02 June 02 - - -
Rocklands March 01 (6 cells) Aug. 01 (land & boat) Nov. 99 - Aug. 01 - - -

Aug. 01
Donald Main Dec. 00 (6 cells) Aug. 01 (land & boat) Oct. 99 - Sept. 01 Oct. 01 Dec. 00 – Aug 01 -
Sept. 01 (Idaho seepage meter)
Dahwilly June 01 (6 cells) June 99 (land) Feb. 02 Nov. 99 Aug. 00 Aug. 00 – Aug 01 -
March 02 (ring infiltrometer & disc
(Central) June 02 (7 cells) Feb. 02 (land & boat) May 02
permeameter)
Murray
Irrigation2 Feb. 01
(Idaho seepage meter)
Dahwilly (East) June 02 (3 cells) March 02 (land & - - - -

March 02 May 02
boat)
Finley July 01 (4 cells) July 00 (land) - July 00 - - -

March 02
June 02 (3 cells) Feb. 02 (land & boat) May 02
Lake View July 01 (6 cells) June 00 (land) Dec. 00 - Aug. 00 – Aug 02 -

- March 02
(Central) June 02 (4 cells) May 02
Murrumbidgee
Irrigation3 Lake View June 02 (4 cells) May 02 (land) - March 02 June 02 - - -
(West)
Tabbita June 01 (6 cells) July 00 (land) - - July 00 July 01 Aug. 00 – Aug 01 -
May 02 (ring infiltrometer)
Goulburn Murray Waranga May/June 02 (12 cells) Nov. 01 (land & boat) - - Nov 01 - - -
Water Western March 02
1. Available remote sensing data for the Wimmera was assessed but deemed not suitable for use in the project. A remote sensing trial was planned for the Wimmera but not conducted due to budget constraints.
The process of planning and preparing for this trial is discussed in the report.
2. Murray Irrigation: Deniboota was removed from the trial program (no works were conducted here) due to the remoteness of the site. The Retreat site (Mulwala Canal) was also dropped from the program
due to the size of the channel and associated cost of conducting pondage tests (an EM31 survey, soil surveying and bore installation was conducted at Retreat in June - August 2000).
3. Murrumbidgee Irrigation: Mirrool Creek Branch Canal was removed from the trial program (no works were conducted here)
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE iv

E.4 Pondage Tests
Pondage tests involve blocking a section of channel for a period and applying a water
balance to determine the seepage losses. They are widely considered the most
accurate means of channel seepage assessment and were the baseline technique against
which other techniques were assessed. Pondage tests were therefore conducted across
all sites, totalling 81 ponds. Seepage rates ranged from 0.1 mm/d to 48 mm/d. The
average and median seepage rate across all sites was 9.7 mm/d and 7.0 mm/d
respectively. Some sites anticipated to have high seepage rates actually contained low
rates, while others expected to have low rates were found to have a high rate of
seepage. Visible evidence of seepage was found to not necessarily imply high seepage
rates. At sites where pondage tests were repeated, a good degree of repeatability was
observed; the maximum difference between rates was 25%, with differences
attributed to changes in depth to watertable and channel bed properties.
E.5 Sub-surface Characterisation

Sub-surface characterisation was conducted to assist in general site characterisation as
well as to assist in geophysical interpretation. An attempt to estimate seepage based
on average soil permeability yielded no clear relationship between soil permeability
and seepage rate. The absence of a relationship was attributed to limitations inherent
in the method adopted (in particular the inadequate sampling density and the process
of assigning permeability to soil type), and the fact that in many of the channels
studied, factors apart from soil type are the primary control on seepage, including
bank dominated seepage and the influence of surface clogging layers. The density of
sampling and permeability testing required, in addition to the fact that soil type is not
always the factor controlling seepage, means that sub-surface characterisation is not
likely to be either an accurate or cost effective means of seepage quantification.
However, it remains a critical part of the site characterisation phase of a channel
seepage investigation.
E.6 Point Tests

Five point test trials were conducted during the investigation, using ring infiltrometers,
disc permeameters and Idaho seepage meters. These trials confirmed that point tests
are generally not reliable for directly quantifying seepage. Due to variable and
sometimes erratic values obtained in measurements, a large number of tests are
required to sufficiently determine the true seepage rate of a section of channel.
Therefore point tests are generally not considered reliable for absolute quantitative
purposes and should generally be limited to determining the distribution of seepage
losses ( ie, relative seepage). Even for this purpose a large number of tests are
recommended to minimise the effects of local variability. The Idaho seepage meter
appeared to provide the most reliable results of the three instruments, probably
reflecting the fact that the channel is full during the test and that truly saturated flow is
being measured.
E.7 Groundwater Techniques

Quantitative analysis of seepage rates was conducted on the Donald Main Channel
based on changes in groundwater level before and after channel filling. Qualitative
assessment only was conducted on the Tabbita site. Groundwater levels at the Donald
Main Channel were used to estimate seepage using analytical equations and seepage
estimates approximately equal to pondage test seepage were obtained, depending on
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE v

the input aquifer hydraulic conductivity used. Therefore, use of groundwater bores for
quantitative analysis of seepage is not considered accurate or cost effective for typical
RWA channel seepage investigations, due to the sensitivity of the mathematical
solution to hydraulic conductivity inputs and the cost of obtaining sufficiently reliable
estimates. In addition, bores are essentially a type of point test and as such do not
address the question of where the channel is seeping. A high density of bore transects
would be required for meaningful identification of local areas of seepage.
However, groundwater observation bores are a very valuable part of the site
characterisation phase of a channel seepage investigation. Further, groundwater bores
are a very useful post-remediation assessment tool, particularly for assessing the
effectiveness of remediation on reducing near channel land degradation. Where land
degradation issues are a significant driver in a channel seepage investigation,
groundwater bores are likely to form a key investigative tool, although as discussed
above should not be relied upon to provide an accurate quantitative analysis.
E.8 Remote Sensing

A remote sensing investigation was planned as part of the trials but was eventually not
undertaken due to budget constraints. Based on the literature review and preparation
of the brief for the proposed trials, it is concluded that remote sensing techniques:
q Are best suited to investigations where the primary aim is identification of land
degradation associated with channel seepage. It should not be used where the
seepage mechanism is predominantly vertical;
q Will be most useful where lateral seepage is predominant. For example, sites
with a high watertable, shallow impermeable layer or bank seepage are likely to
facilitate lateral seepage and cause seepage to have a surface expression;
q Should primarily be regarded as a seepage identification tool and not for seepage
quantification purposes;
q Require a suitable spatial resolution to allow definition of seepage zones. Ground
resolutions of less than 10 m are suggested;
q Are best conducted in the infra-red range of the electromagnetic spectrum, as this
area of the spectrum is strongly absorbed by water and will be able to most
clearly separate areas of varying soil moisture and plant water and growth status;
and,
q Are generally best collected during late summer and early autumn when
surrounding areas (apart from irrigation) will be distinctly drier.
E.9 Geophysics
E.9.1 General Conclusions
Seepage Detection Mechanisms
Geophysical techniques applied to seepage measurement primarily depend upon
measuring a contrast in terrain conductivity (or resistivity) in the sub surface profile
around the channel. They can be used in one of two ways:
1) Directly measuring the conductivity of the groundwater, and identifying the

conductivity contrast of fresher channel water as it seeps into and dilutes saltier
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE v i

native groundwater. Decreasing the salinity of the groundwater will cause a
decrease in electrical conductivity (or an increase in its inverse, resistivity).
2) Identifying contrasts in soil properties and inferring the likelihood of seepage
through more permeable materials in the zone above the watertable. Formations
more likely to allow seepage, such as sands, are naturally lower in conductivity
(higher in resistivity) due to lower porosity and lower cation exchange capacity
than tighter clay dominated formations. In addition the higher permeability of
such formations leads to better drainage and lower salt content, further reducing
conductivity. The magnitude of seepage is assumed to be related to unsaturated
zone soil properties beneath or adjacent to the channel.
Figure E-2 visually depicts how these two different approaches can be used to identify
or infer seepage.
n Figure E-2 Comparison of how geophysical techniques can be used to

identify channel seepage (LHS - inferred from soil property variations, RHS -
direct measurement of salinity impact on watertable)
2. Seepage inferred 1. Seepage measured directly

based on soil property in terms of salinity impact
variations (ie, above the on groundwater (ie, at &
watertable) into the watertable)
Clay
Sand
% Contribution to % Contribution to
response Watertable response
Fresher seepage plume
Saline groundwater
Technically the second method of ‘detection’ is not really detection, but the magnitude
of seepage is assumed to be related to unsaturated zone soil properties. In many cases
this is a reasonable assumption, supported by the fact that the inferred method of
detection was successful at most, but not all sites investigated in the trials. The
unsaturated zone is not necessarily the controlling influence on seepage, and
particularly in Australian conditions seepage is often controlled by a clogging (silt)
layer. Therefore, there is less risk in using the direct method of seepage detection.
The direct method of detection cannot be used in relatively non-saline groundwater
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE vii

environments, as the fresh seepage water will not contrast with the native
groundwater. As a guide it is recommended that groundwater salinity is at least three
to four times higher than the channel water salinity. Background variations in
groundwater salinity along the channel will affect the results of direct seepage
detection and will need to be allowed for during interpretation.
It is very important that the depth to watertable is known at the site before selecting a
geophysical technique. Based on this information a decision can be made as to
whether direct or inferred measurement will be undertaken and hence the technique
that will be adopted.
Comparison of Trialed Geophysical Techniques

The following have been identified as key criteria against which geophysical
techniques should be compared:
q Accuracy
q Cost and Speed
q Availability of Operators
q Data Processing
The three geophysical techniques trialed in this investigation (EM31, EM34 and
resistivity) are discussed in terms of each of these criteria.
Accuracy
The accuracy of a given geophysical technique will depend on whether inferred or
direct seepage detection is used. Generally direct measurement should be considered
more reliable than inferred measurement. For direct measurement the accuracy will
depend on how well the watertable is targeted. Therefore in theory on-channel
resistivity surveying should be the most accurate geophysical technique, as it is based
on direct seepage detection and can target the watertable independent of depth. At
most sites in the trials resistivity surveying results were comparable to EM31 and
EM34, and at three sites correlations with pondage tests were better than the EM
correlations.
The fundamental limitation with all EM surveys and other such fixed array type
geophysical surveys is that the result is averaged over a specific depth interval, which
may not be the critical interval of interest. Therefore (for direct detection) the
accuracy depends on how well the watertable is targeted by the particular EM
equipment, which in turn depends on the watertable depth. If the correct EM
equipment is selected to suit the watertable depth, in theory it should be close to the
accuracy of resistivity surveying.
The robustness of EM31, as demonstrated by the consistent results in the trials is due
to its relatively shallow depth focus (1-4m). For channels where there is a shallow
watertable (eg, surface to 3-4m), EM31 can be used for direct measurement of
seepage, which as discussed above is likely to be more reliable. When the watertable
is deep, EM31 infers seepage from near surface soil properties, which is suitably
accurate in most instances.
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE viii

Cost and Speed
EM31 surveys are the cheapest geophysical method, due to the speed of data
acquisition; EM34 is more expensive as two people are required for operation and the
equipment must be carried by hand. Resistivity surveying costs are difficult to
quantify given that the on-channel application of the technique is relatively new.
Costs are likely to come down as the technique is refined.
Availability of Operators
A number of commercial EM34 and EM31 contractors are in operation in South East
Australia. At present on-channel resistivity surveying is still in a development phase
and as such there are no commercially operating contractors who specialise in this
type of survey, but a number of geophysical exploration / surveying companies have
the capability to develop this type of equipment.
Data Processing
Data processing requirements for EM31 and EM34 surveying are minimal. By
comparison, data processing requirements for resistivity surveying are much higher,
due to the cost of inverting the data to produce a resistivity cross section.
Critical Geophysical Survey Variables

q Survey timing – If direct measurement of seepage is used, the survey must be
conducted while the channel is running (preferably for at least several weeks),
however if seepage is being inferred from soil properties then the timing of the
survey is not critical and can be conducted whether the channel is running or
empty.
On-channel versus on-land – Further work is required in this area, but overall in
the trials the most consistent results were returned on-land and this is considered
the safest option. Evidence collected in this investigation suggests on-channel
EM31 surveys should only be conducted where the geophysical technique can
penetrate into the watertable, and ideally target the top of the watertable.
Resistivity surveys however, can (and should) be conducted on-channel because
of their greater depth penetration capacity.
q Off-set distance and location for on-land surveys – The evidence collected in
these surveys indicates the best off-set distance for on-land surveys is immediately
adjacent the outside toe of the channel.
Repeatability
Generally a high degree of repeatability was observed between duplicate surveys. At
two sites where there was a significant difference in the results, changes in
groundwater conditions due to channel operation accounted for the observed
differences.
Regional Assessment of Key Relationships

For all of the sites used in the final year of analysis, multiple and simple linear
regression was undertaken to look for potential regional correlations between seepage
rates and geophysical response (for both EM31 and resistivity). The multi-variate
regression analysis indicated that, apart from the geophysical response, depth to
watertable was the next most significant explanatory variable.
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE ix

Based on distinct trends between sites with shallow and deeper watertables, the sites
were split into two data sets based on depth to watertable, in order to improve the
accuracy of the fitted regression model. This is most clearly illustrated in Figure E-3,
which shows EM31 data for all sites divided into two categories based on depth to
watertable. For sites with a deep watertable (5-10m below surface) the permeability
of the top 2m of the profile was shown to be an explanatory variable of secondary
importance.
Statistically the regional fitted regression models were generally moderate to good,
with correlation coefficients of around 0.5 – 0.6 and standard error of estimates of
around 50%. In some cases a higher correlation coefficient and relatively low
standard estimate of error was obtained, however this was for data sets with fewer data
points – greater number of points are required to improve confidence in these models.
Confidence intervals (80% and 90%) for the regression lines were generally fairly
broad, indicating that these regional equations can only be used to broadly classify
seepage rates (eg, into low, medium and high categories). Consequently it is
recommended that there is currently insufficient confidence in these regression
equations for their use to predict seepage at new sites without local calibration against
pondage tests.
In most instances the multi-variate analysis did not significantly improve the
regression model. The addition of the soil permeability parameter (for sites with a
deep watertable), while statistically significant, generally only resulted in marginal
improvements to the model. The cost of field tests to collect this data therefore
probably outweighs the benefits.
For the resistivity analysis, the ten metre depth slice was adopted as the variable for
use in the model. While a more accurate analysis could be conducted using the depth
at and just below the watertable, for the purpose of a consistent approach, this depth
slice was selected. It is likely the analysis could be significantly improved by using
resistivity data at and immediately below the watertable for each of the sites.
n Figure E-3 Regional EM31 Assessment: Pondage Test Seepage Versus

EM31 Conductivity with Sites Divided Based on Depth to Watertable
50
y = -0.690x + 71.961 Sites with Watertable 5-10m
2
R = 0.889
Sites with Watertable < 2m
40
Pondage Test Seepage (mm/d)
30
20
y = -0.129x + 12.788
2
R = 0.469
10
0
0 20 40 60 80 100 120
Average EM31 Conductivity (mS/m)
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE x

Confidence in Derived Relationships and Extrapolation of Results
Two key issues regarding relationships derived between channel seepage and
geophysical response need to be assessed:
1. What confidence is there that the derived relationship accurately describes
seepage within the area tested? - Confidence in the derived seepage-geophysical
relationship within the area tested can be assessed by a number of statistical
indicators, including: the correlation coefficient, standard error of estimate, and
prediction interval. The number of data points and seepage rate range
represented should also be considered.
2. How confidently can the relationship be used outside of the area tested in order
to predict seepage? - When extrapolating a geophysical-seepage relationship
outside of an area from which it was developed, firstly the strength of the
original relationship needs to be assessed (refer above). Secondly, the
representativeness of the new area in comparison to the conditions where the
relationship was derived should be evaluated.
Preferred Methodology
Based on the trials conducted in this investigation, and the methodology outlined in
the guidelines (ANCID, 2003) the following methodology for using geophysics to
identify and measure seepage is recommended:
1. Define project objective – The key issue that needs to be addressed is
identification of the primary reason the work is being undertaken.
2. Collate Site Data – Basic site information including depth to groundwater,
groundwater salinity, soil type and channel hydraulics should be collated at the
testing site and over the area the results are to be extrapolated.
3. Evaluate Site Data - This should be at a level to enable development of a first cut
conceptual model of the seepage mechanism, to detect where parameter changes
may impact on geophysical response, and to assist in technique selection.
4. Select Technique - The preferred geophysical seepage measurement technique is
one that directly detects the impact of seepage on the groundwater. To do this it
must have a depth focus on and immediately below the watertable. The
recommended technique for a given depth to watertable, for both direct and
inferred detection, is outlined below:
Direct Detection
q Shallow watertable (surface to approximately 5m): EM31 is recommended.
q Watertable deeper than 5m: EM34 (in vertical dipole mode, with the coil
spacing dependent on the depth to watertable) or on-channel resistivity can
be used. However, particularly for deeper watertables, it is easier to focus
on a given depth using resistivity.
Note that direct detection requires native groundwater salinity to be at least three
to four times more saline than channel water salinity.
Inferred ‘Detection’
q EM31 (vertical dipole) adjacent the channel can be used effectively in areas
with deeper watertables to infer seepage based on upper soil layer properties.
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE xi

A decision to use EM31 in an area with a deep watertable might be made due to
budget constraints, where a potentially slightly lower level of accuracy is
considered acceptable, or due to a lack of alternatives (eg, EM34 or resistivity
contractors not readily available). It this method is used however, it must be
made certain that seepage is controlled by the unsaturated zone and not surface
clogging processes. A more detailed description of preferred geophysical
techniques for seepage detection is presented in Table E-2.
5. Conduct Field Trials

5a. Conduct geophysical survey – Undertake geophysical survey in section of interest.
5b. Evaluate results – Plot survey results and overlay with known site conditions
(soils, hydrogeology, etc). Identify areas of suspected high, low and moderate
seepage.
5c. Conduct test drilling – Soil bores should be drilled at appropriate intervals along
the section to assist with interpretation of the geophysical survey. Bore locations
should be based on the geophysical survey results, and should cover a range of low,
moderate, and high conductivity / resistivity response;
5d. Conduct pondage tests – The number of pondage tests will depend on the length
of channel surveyed and the variability of conditions along the channel. Pondage tests
should be conducted across a range of low, moderate and high conductivity /
resistivity sites so as to establish a regression equation which represents the range of
geophysical response and should also cover the range of soil types. Individual cells
must be conducted over areas of like conductivity / resistivity.
5e. Develop and evaluate the relationship between seepage and geophysical response
– This involves plotting average geophysical response against pondage test seepage,
removal of outliers as appropriate, fitting of a regression line, statistical analysis to
determine the degree of confidence that can be placed in the derived relationship and
use of the derived relationship to predict seepage in new areas.
6. Evaluation – Evaluate whether investigation objectives have been met.
E.9.2 Summary of EM34 Results

Good to moderate relationships were obtained between average EM34 conductivity
and the corresponding pondage test seepage at most sites. For EM34 at a 10m coil
spacing in horizontal mode, the effective depth of penetration is around 6-7m, with a
shallow depth focus at around 1-3m. This meant that at sites where the watertable was
deeper than 5m, only a limited proportion of the response was caused by seepage
impacts in the saturated zone. Therefore at these sites the seepage detection
mechanism is predominantly via inference based on unsaturated zone soil properties.
The only site where no relationship was observed was at Dahwilly East, which was
largely due to the narrow seepage rate range. At the Toolondo Central site, where
conductivity measurement was entirely above the watertable, the unsaturated zone
lithology was a sufficiently accurate indicator of seepage and hence a reasonable trend
was observed (a fact reinforced by the success of EM31 at the site). Significantly, the
resistivity surveying showed improved correlations compared to the EM34, for the
depth slices focussed immediately below the watertable.
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE xii

n Table E-2 Recommended Geophysical Technique for Seepage Detection
and Measurement
Watertable Recommended Detection Method2 Approximate Depth
Depth Focus (m)4
Depth (m) Technique 1 of Penetration (m)3
Surface to 1.5 EM31 (horizontal dipole)5 Direct watertable impact 3 0-1
1.5 – 5 EM31 (vertical dipole)5 Direct watertable impact 6 1 – 3.5
EM34 – 10m coil spacing Direct watertable impact 15 3 - 10

(vertical dipole)6
OR
5 – 12
Resistivity 7, 9 Direct watertable impact NA 10 NA 11
OR
EM31 (vertical dipole)8 Soil property variations 6 1 - 3.5

(vertical dipole)6
OR
12 – 25
NA 10 NA 11
Resistivity 7, 9 Direct watertable impact
OR
Resistivity 9 Direct watertable impact NA 10 NA 11

> 25
OR
1. It is recommended EM techniques are conducted adjacent the channel (additional survey runs
can be conducted away from channel). Resistivity surveys should be conducted on-channel.
2. Direct detection of seepage impacts on the watertable is the recommended technique, but
inferred ‘detection’ based on soil property variations will often provide an adequate simulation
and may be more convenient for various reasons - refer to body of report for potential errors
associated with this method. Note that direct detection relies on a salinity contrast between the
channel water and the groundwater. It is recommended the groundwater should be at least 3 to
4 times more saline than the channel water, a condition usually met in Australian conditions.
3. Approximate detection of penetration: referred to in the Geonics manual (Mc Neil, 1980) as the
effective depth of exploration. This is the depth to which approx. 75% of response is attributed.
4. The ‘depth focus’ is a term used in this report to describe the depth (range) which is most
influential in terms of the relative contribution to the overall EM response (M c Neil,1980).
5. These can be conducted immediately adjacent to the channel or on-channel. Both are
recommended if budget allows. If on-channel is used for a watertable of 0-1.5m, the survey
should preferentially collect data in vertical dipole mode where the effects of channel water
will be less influential. For sites with a watertable 0-1.5m, EM31 on channel may be preferred
if significant land salinisation exists adjacent the channel.
6. Horizontal and Vertical Dipole: Note that as applied to EM34, vertical dipole does not refer to
the coil orientation with respect to the ground, and is in fact opposite to the coil orientation. In
vertical dipole mode the coils should be horizontal to the ground, which is a slower method
than horizontal mode where they are held perpendicular to the ground.
7. Resistivity is the preferred direct measurement technique for this depth to watertable but EM34
is provided as a potentially more accessible alternative.
8. This should be conducted immediately adjacent to the channel.
9. This should be conducted on-channel.
10. Penetration depth of resistivity depends on the particular set up (dipole spacing and length).
11. Resistivity surveys measures resistivity at a range of depths intervals within the profile (ie,
there is no fixed depth focus).
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE xiii

The Donald site survey was focussed on the saturated zone, however the EM31 survey
at the site demonstrated a slightly better relationship with pondage test seepage
compared to the EM34, but neither survey differentiated between the higher seeping
ponds. The improved correlation is probably attributable to the deeper depth focus of
the EM31 compared to the EM34 (10m, vertical dipole configuration).
At the Rocklands and Dahwilly sites, where the penetration depth (EM34 - 10m coil
separation, vertical dipole) was just sufficient to reach the watertable (but the focus
was above the watertable), the combination of measuring lithology changes in the
unsaturated zone and seepage impacts in the saturated zone worked to provide a
reasonable indicator of seepage. However it is significant that at Dahwilly, where
resistivity surveying was conducted, an improved relationship was obtained compared
to EM34 when the depth slice was focussed immediately below the watertable, where
seepage impacts are most discernible.
E.9.3 Summary of EM31 Results

Good relationships were obtained between average EM31 conductivity and the
corresponding pondage test seepage at most sites. For EM31 in vertical dipole mode,
the effective depth of penetration is around 6-7m, with a mid-range depth focus of
about 2 - 4.5m. This meant that at sites where the watertable was deeper than 5m,
only a limited proportion of the response was caused by seepage impacts in the
saturated zone. Therefore at these sites the seepage detection mechanism is largely via
inference based on soil properties in the unsaturated zone.
The only site where no relationship was observed was at Tabbita. A number of
possible causes for this were identified, but the predominant contributing factor is
unknown. At two sites (Rocklands and Lake View Central), the adjacent channel data
was used instead of all survey run data away from the channel. This was required to
obtain the best relationship, due to the interference effects of trees and rapid mixing of
seepage water away from the channel.
At the Toolondo Central site, where conductivity measurement was entirely above the
watertable, the unsaturated zone lithology was a sufficiently accurate indicator of
seepage and hence good trends were observed. The Donald and Lake View site
surveys were focussed on the saturated zone, and seepage was detected as it created a
conductivity low against higher background conductivity groundwater.
At the Rocklands and Dahwilly sites, where the penetration depth of the EM31 (in
vertical dipole) was just sufficient to reach the watertable, the combination of
measuring lithology changes in the unsaturated zone and seepage impacts in the
saturated zone combined to provide a reasonable indicator of seepage. However it is
significant to note that at Dahwilly, when the channel was not running, no relationship
was observed. This suggests seepage impacts in the watertable are the primary
detection mechanism at this site, a fact reinforced by the uniform nature of the
unsaturated zone lithology at the site. Seepage at Dahwilly is not controlled by the
unsaturated zone but by a clogging layer at the base of the channel. Techniques which
purely infer seepage from unsaturated zone soil properties will not work at such sites
(including remediated or lined channels).
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE xiv

At Waranga a reasonable relationship was observed, given the distance over which the
data forming the relationship was spread. Improvements might be expected using a
technique targeting the top of the watertable at this site.
E.9.4 Summary of Resistivity Results

Good relationships were obtained between average resistivity (from depth slices
immediately below the watertable) and the corresponding pondage test seepage at
most sites, and at three sites correlations were better than the EM results. The two
sites where there was no correlation was at Toolondo West and Lake View West. At
Toolondo West it appears that the type of sandstone at this site may be dominating the
response, however deeper drilling would be required to confirm this interpretation.
The lack of trend at the Lake View West site is probably due to the poor resolution of
the resistivity equipment at very shallow depth. This site contains the shallowest
watertable across all sites (0.5 - 1m). Improved resolution at shallow depth could
relatively easily be improved in future surveys. At Toolondo East also no trend was
observed, but this is solely attributed to the very narrow seepage rate range at this site.
E.10 Waranga Western Channel Case Study

It was proposed that the Waranga Western Channel (WWC), an open irrigation
channel maintained by Goulburn-Murray Water (G-MW), be upgraded in capacity
along approximately 50 km of the channel length. The channel has a well-
documented record of existing seepage problems. There was also concern that new
seepage paths may be opened up during the upgrading works program. Therefore
quantification of sections with existing seepage problems and identification and
quantification of sections where new seepage paths might be opened up was required.
To this end, geotechnical and geophysical investigations were carried out along the
channel, including an EM31 survey coupled with drilling of 128 shallow bores, further
geotechnical drilling, involving the drilling of an additional 107 bores, and the
conducting of twelve pondage tests.
Initially a combination of the EM31 results and a lithological classification devised for
the investigation (based on the amount of clay in the profile) was used to identify
sections of channel which were considered to represent ‘very high’ risk areas. It was
then recognised that, in addition to the drilling program, pondage tests were required
to quantify seepage rates and confirm interpretation of seepage rates based on the
geological and EM31 data. Based on the results of the pondage tests, the regression
relationship between EM31 and the pondage tests and the drilling program, areas
recommended for remediation were finalised. Given the broad confidence intervals in
the EM31 – seepage relationship, the EM31 predicted seepage was not used as the
sole means of assigning seepage risk but geological data and visual observations were
also integrated into the decision making process. The WWC seepage investigation is a
good example of the integration of geophysical, geological and pondage test data to
determine areas of highest seepage risk.
E.11 Recommendations
This study makes the following recommendations:
q Of the techniques trialed in this investigation, future channel seepage
measurement investigations should focus on geophysical techniques, as these have
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE xv

shown the most promise to cost-effectively and relatively accurately quantify
channel seepage. Remote sensing trials, however, were not conducted in these
investigations. This technique has the potential for rapid assessment of long
sections of channel where seepage has a surface expression, and as such deserves
carefully planned field trials in Australian conditions. The baseline data collected
in this report could be used to assist in calibration of such trials.
q Rural Water Authorities should adopt the preferred technique as outlined in this
report (and the Guidelines Manual; ANCID, 2003) for channel seepage
measurement investigations. This methodology relies on geophysics to identify
seepage, and pondage tests and soil bores to calibrate and interpret the geophysical
response.
q A national database be established to record all channel seepage measurement
geophysical trials.
q Further study into the best method of establishing a relationship between the
geophysical response and seepage rates is required. At present the bulking
process of averaging the geophysical response over the entire pondage test area
necessarily introduces errors into the geophysical - seepage relationship.
q Further experimental trials to improve the shallow depth resolution of the
resistivity equipment are recommended. Investigation into means of reducing
resistivity data processing time (and thus costs) are also suggested.
q Exploration of a method which detects seepage by measuring changes from
background conditions is recommended. A significant problem encountered in
these trials when attempting to extrapolate a relationship from one section of a
channel to another, was caused by the fact that the background conditions change
along the channel.
q Further testing of the relative merits of on-channel fixed array surveys compared
to adjacent channel fixed array surveys are required. The evidence collected in
this investigation suggests on-channel (fixed array) surveys should only be
conducted where the geophysical technique can penetrate into the watertable, and
ideally target the top of the watertable. However these conclusions are only based
on trials at three sites and further work is required to confirm this conclusion.
q A means of calibrating geophysical surveys where pondage tests cannot be
conducted needs to be explored.
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE x v i

1. Introduction
1.1 Background
Seepage from earthen channels has become an important issue in Australia for two
reasons:
i) The loss of an economically valuable resource; and,
ii) the contribution of seepage water to land degradation issues such as salinity and
waterlogging.
In response, the Australian Committee of Irrigation and Drainage (ANCID), in

conjunction with the Murray Darling Basin Commission (MDBC) initiated a project to
investigate channel seepage measurement. The main objectives of the study were to:
q Assess the current status of channel seepage identification, measurement and
quantification techniques;
q Trial and document a range of seepage identification, measurement and
quantification techniques; and
q Prepare and publish guidelines on the best practice techniques for identifying,
quantifying and monitoring channel seepage.
The first of these objectives was completed through the compilation of a channel
seepage literature survey and an Australia wide channel seepage survey of more than
40 rural water authorities (ANCID, 2000a and ANCID, 2000b). Seepage trials were
conducted from early 2000 to mid 2002, within the following Rural Water Authorities
(RWAs):
q Wimmera Mallee Water (WMW);
q Murray Irrigation Limited (MIL); and
q Murrumbidgee Irrigation (MI).
In addition, results from channel seepage measurement investigations conducted on

the Waranga Western Channel (by Goulburn-Murray Water) were incorporated into
the final year of trials.
This report summarises the results of the three years of trials. Channel seepage
measurement guidelines (ANCID, 2003) were prepared concurrently with this seepage
measurement trials report, and are based on the findings of these trials, the literature
review and the RWA survey.
The scope and design of the trials were planned by Sinclair Knight Merz, and the trials
were carried out by the RWAs. The work was either carried out in-house by the
RWAs or sub-contracted, depending on the nature of the trials.
Early in the trials program considerable attention was dedicated to determining how
and where the seepage trial programs would be run in each of the Authorities. The
importance of ensuring agreement among all key stakeholders as to how the programs
were to be designed was important to the success of the entire project. Site visits were
carried out at each Authority in February and March 2000. Based on these visits,
three sites were selected in each Authority to trial a range of channel seepage
techniques:
WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE 1

q Wimmera Mallee Water – Donald Main Channel, Toolondo Channel and
Rocklands Channel;
q Murray Irrigation – Dahwilly Main, Deniboota Main and Retreat (Mulwala
Canal); and,
q Murrumbidgee Irrigation – Tabbita Channel, Lake View Branch Canal and
Mirrool Creek branch canal.
Subsequently, Murray Irrigation removed Deniboota from the list of channels to be

trialed (ie. no works were conducted there) due to the remoteness of the site. Finley
channel was chosen to replace the Deniboota trial site. In addition, the Retreat site
(Mulwala Canal) was also dropped from the trials program due to the size of the
channel and associated cost of conducting pondage tests (an EM31 survey was
conducted at this site). Murrumbidgee Irrigation also removed Mirrool Creek branch
canal from its trials program, choosing to put more resources into assessing the two
remaining channels.
The Wimmera Mallee Water channel running season is approximately 6 months out of
sync with those of Murray and Murrumbidgee, whose channels generally close for
several months in the June-August period. Wimmera Mallee Water’s channels supply
domestic and stock water to irrigation supplies and are run in the cooler months of the
year to save seepage and evaporation losses. As a consequence its channels are often
shut down for several months over summer. This led to complications in coordinating
some aspects of the trials between Authorities (eg first year pondage tests).
1.2 Content and Scope of this Report

The contents of this report are described below:
q Section 2 – Overview of Trials Conducted: Presents a summary of the trial
program, describes the process by which the program was developed and the
methodology behind assessment of the trials;
q Section 3 – Trial Sites Description: Provides a summary description of key
factors affecting seepage at each trial site;
q Section 4 – Non-Geophysical Techniques: Presents the results of all non-
geophysical techniques trialed during the program, including:
q Pondage Tests
q Sub-Surface Characterisation
q Point Tests
q Remote Sensing
q Section 5 – Geophysical Techniques: Presents the results of all geophysical
techniques trialed during the program, including:
q EM34
q EM31
q Resistivity Profiling
q Section 6 – Waranga Western Channel Case Study: Presents the Waranga
Western Channel seepage investigation as a case study of developing the
methodology and techniques required in a channel seepage measurement
program; and,
q Section 7 – Conclusions and Recommendations: Summarises the main findings of
the trials and presents key recommendations arising out of the study.Overview of
Trials Conducted

2. Overview of Trials Conducted
2.1 Technique Selection
The main seepage measurement techniques referred to in the literature, and those
discussed in the literature review (ANCID, 2000a) include:
q Pondage Tests
q Point measurement (channel full and empty)
q Geophysical Techniques
q Soil Classification
q Remote Sensing
q Inflow – Outflow
q Mathematical Modelling
q Hydrochemical / Isotopic Mass Balance
q Tracing Leaking Plume
Based on the outcomes of the literature review (ANCID, 2000a), the RWA survey
(ANCID, 2000b), and consideration of the primary objectives of the study, the trials
were focussed on the first six of these techniques. The early trial program covered the
six techniques described above. The final year of the program was based on the
results from the first two years of trials. In order to maximise the usefulness of the
output of the trial program it was decided that the final year of trials should focus on
one technique, which demonstrated the greatest potential for meeting RWA
requirements for channel seepage measurement / identification. The technique
selected was geophysics. The rationale behind this decision is described in section 5.
The most important criteria for selecting techniques suitable for channel seepage
measurement and identification are cost and accuracy. Significantly, RWAs rank cost
as the most significant factor in selecting seepage investigation techniques, with
technical accuracy of lesser importance (ANCID, 2000b). This finding was of
fundamental importance to the development of the trial program, and was the reason
why some techniques (eg hydrochemical) were not tried at all in the program and why
others became the focus of the program. The trials focused on developing general
principles which could be applied to identification and measurement under the
operating conditions of the managing water authority.
Technique selection requires consideration of the difference between seepage

quantification and seepage identification. For the former, pondage tests are the
recognised standard. For identification, visual inspection (and to a lesser degree EM
techniques) seem to be the current standard used by RWAs (ANCID, 2000b). The
objective of the trial program was to develop other approaches (referring to a single
technique or a combination of techniques) that could deliver (a) quantification of
channel seepage in either a more cost effective or accurate way than pondage tests and
(b) identify seepage in a more accurate or cost effective way than visual inspection. In
all cases there is a trade off between cost and accuracy.

2.2 Trial Program Summary Table
Table 2-1 summarises the trials conducted during the program. This table shows that
pondage tests were conducted at all sites, as they were the basis on which other
techniques were assessed (refer section 2.4). Drilling was also conducted at all sites in
order to identify sub-surface conditions. The emphasis placed on geophysics is
readily apparent from this table. Remote sensing is included in the table even though
trials were not undertaken as part of the study. Available data was assessed but
deemed not suitable for use in the project. However due to the significant planning
and preparation for the remote sensing trial, sufficient background information was
collated and considered worthy of inclusion in this report (refer section 5). Channels
which were initially included in the program, but removed for various reasons, are
listed at the bottom of the table.
2.3 Techniques Not Included in the Trials Program

The following section briefly describes techniques which were not used in the trials
and the reason for their exclusion.
2.3.1 Inflow-Outflow Tests

Inflow-outflow tests were initially included on the list of techniques to be trialed,
however investigation into the level of accuracy obtainable using this method led to its
exclusion from the program. Measurements using a current meter are, at best,
accurate to 2%, i.e. 4% for the two measurements at either end of the section (Thiess,
pers. comm., 2000). Typically seepage is only 3 to 4% of flow in a given section, and
therefore seepage will not be able to be distinguished from the error bounds of the
measurement.
Over a relatively long length of channel this is an appropriate technique, due to the
greater volume of water lost to seepage. Therefore the technique is suitable for
identifying and prioritising, at an Authority-wide level, channels which have higher
losses compared to others in the system. It will not identify where within the section
the channel is seeping. The emphasis in this study, however was on relatively short
(1-2km) sections of channel and the isolation (including identification and/or
measurement) of seepage within those sections.
2.3.2 Mathematical Modelling

Mathematical models are based on the physics of groundwater flow and have been
found to yield reliable estimates of channel seepage, when the required field data such
as watertable elevations, soil and aquifer characteristics, and the hydraulic conditions
under which seepage occurs are collected. Groundwater modelling can simultaneous
incorporate a range of factors which affect seepage into the analysis. Modelling is
valuable if there is a need to understand the details of the flow mechanisms at
particular areas.
The flow system can be simulated and calibrated against variation of water levels in
the aquifer with time under changed hydraulic conditions in the channel. This enables
an understanding of the way seepage occurs, the factors which affect the seepage
entering the groundwater, and the potential consequences of seepage on local land
degradation. Similarly the impact of remedial works could be assessed using
modelling to test scenarios.

To be effective, there is a need to measure soil hydraulic conductivities, which
requires personnel competent in performing hydraulic conductivity tests, soil survey
crews, and experienced supervising personnel competent in hydrogeology. The
detailed field work required to characterise flow paths and hydrogeological conditions
in the vicinity of the channel is likely to involve considerable time and expense and
require significant expertise. The intensity of data collection to adequately
characterise a reach of channel for modelling purposes generally means that this
method is not practical for widespread use by RWAs.
However there are valid reasons for considering numerical modelling in investigations
requiring detailed studies. Models offer the potential for an understanding of the
mechanisms and rates related to the channel, but they can also take into account the
impact of regional land management factors such as irrigation or increased
groundwater recharge in areas surrounding the channels. Modelling can therefore be
very useful in identifying the benefits of channel management and remedial works
within the broader land management framework.
2.3.3 Hydrochemical Techniques and Tracing of Leakage Plume

Hydrochemical / isotopic mass balances and tracing of the leakage plume techniques
were on the original list of techniques to be trialed but were eliminated due to the
considerable expense involved in conducting these type of trials. The high cost of
such trials means they are not going to be practical solutions for RWAs to adopt.

n Table 2-1 ANCID Channel Seepage Measurement Project - Trials Summary Table
Technique
Date Conducted
Rural Water Channel
Pondage Tests Geophysics
Authority Sub-Surface Point Tests Groundwater Remote Sensing 1
Profiling Techniques
EM31 EM34 Resistivity
(all land based) (all based on-channel)
Toolondo March 01 (6 cells) Dec. 00 (land)

March 02 Dec. 00 Dec. 00 / Jan. 01 - Sept. 00 ?
(Central) March 02 (1 cells) Aug. 01 (land & boat) Aug. 01
June 02 (ring infiltrometer & disc
March 02 (land) permeameter)
Wimmera Toolondo (East) March 02 (4 cells) March 02 (land) - March 02 June 02 - - -

Mallee
Water Toolondo (West) March 02 (4 cells) March 02 (land) - March 02 June 02 - - -
Rocklands March 01 (6 cells) Aug. 01 (land & boat) Nov. 99 - Aug. 01 - - -

Aug. 01
Donald Main Dec. 00 (6 cells) Aug. 01 (land & boat) Oct. 99 - Sept. 01 Oct. 01 Dec. 00 – Aug 01 -
Sept. 01 (Idaho seepage meter)
Dahwilly June 01 (6 cells) June 99 (land) Feb. 02 Nov. 99 Aug. 00 Aug. 00 – Aug 01 -
March 02
(Central) June 02 (7 cells) Feb. 02 (land & boat) May 02 (ring infiltrometer & disc
permeameter)
Murray
Irrigation2 Feb. 01
(Idaho seepage meter)
Dahwilly (East) June 02 (3 cells) March 02 (land & - - - -

March 02 May 02
boat)
Finley July 01 (4 cells) July 00 (land) - July 00 - - -

March 02
June 02 (3 cells) Feb. 02 (land & boat) May 02
Lake View July 01 (6 cells) June 00 (land) Dec. 00 - Aug. 00 – Aug 02 -

- March 02
(Central) June 02 (4 cells) May 02
Murrumbidgee
Irrigation3 Lake View June 02 (4 cells) May 02 (land) - March 02 June 02 - - -
(West)
Tabbita June 01 (6 cells) July 00 (land) - - July 00 July 01 Aug. 00 – Aug 01 -
May 02 (ring infiltrometer)
Goulburn Murray Waranga May/June 02 (12 cells) Nov. 01 (land & boat) - - Nov 01 - - -
Water Western March 02
1. Available remote sensing data for the Wimmera was assessed but deemed not suitable for use in the project. A remote sensing trial was planned for the Wimmera but not conducted due to budget
constraints. The process of planning and preparing for this trial is discussed in the report.
2. Murray Irrigation: Deniboota was removed from the trial program (no works were conducted here) due to the remoteness of the site. The Retreat site (Mulwala Canal) was also dropped from the program
due to the size of the channel and associated cost of conducting pondage tests (an EM31 survey, soil surveying and bore installation was conducted at Retreat in June – August 2000).
3. Murrumbidgee Irrigation: Mirrool Creek Branch Canal was removed from the trial program (no works were conducted here)

2.4 Assessment Methodology
In undertaking these channel seepage investigation, the basic approach adopted was:
q Identification of test site locations;
q Measuring rates of seepage at test sites using direct measurement techniques –
pondage tests were used for this purpose (refer to following paragraph);
q Comparison of the direct measurement technique with indirect techniques; and,
q Extrapolation of results beyond the test zone to interpret seepage distribution –
this was applied for techniques which compared favourably with the direct
technique.
It is well documented in the literature that, while every channel seepage method has
certain disadvantages, almost universally pondage tests are regarded as the most
accurate method of quantifying seepage (ANCID, 2000a). Therefore the basic method
of assessment of the accuracy of each technique adopted in the trial program is by
comparison against pondage test data. The method by which the pondage test is used
for comparison differs for different forms of data (eg point test data verses essentially
continuous output from a geophysical survey) However the pondage test seepage rates
are the baseline means of assessment in this study. Therefore when terms such as
‘accurate’ or ‘successful / unsuccessful’ are used to describe a technique, this is in
comparison to pondage test data which are assumed to be accurate for the purpose of
this investigation.

3. Description of Trial Sites
Figure 3-1 presents the location of the channel seepage measurement trial sites. All of
the sites lie within the Murray Basin. With the exception of Rocklands and Toolondo
channels, which are traverse the flanks of the Grampians ranges, all of the sites are
located relatively down-catchment in fluvial and alluvial deposits of sands, silts and
clays. All of the channels investigated were main delivery channels, ranging in
capacity from 80 ML/d (Tabbita) to 600 ML/d (Rocklands).
Table 3-1, Table 3-2, and Table 3-3 summarises the key characteristics at each of the
sites, including soil and geological properties, and channel hydraulic characteristics.
Each table represents a separate RWA. Site characteristics for the Waranga Western
channel are presented in the case study on this channel, in Section 6.
These tables illustrates the range of conditions across the sites. With respect to
lithology, sites ranged from a clay profile (eg Finley), to a sand profile (eg Dahwilly
Central), as well as sites with rock at or near the surface (eg Toolondo and
Rocklands). Groundwater salinity ranged from moderately fresh (eg 2500 EC at
Dahwilly East) to highly saline (eg 30,000 EC at Donald). Groundwater depths
ranged from very shallow (eg 0.5 – 1.5m at Lake View) to moderately deep (eg 9-10m
at Toolondo). Channel dimensions were reasonably similar, with the depth of water at
full supply level (FSL) typically 1.5m and wetted perimeters of between 9-16m.
Figure 3-1 Trial Site Regional Location Map
Appendix A contains geological long sections for each of the sites, based on soil bores
drilled along the channel sections (and groundwater bores where available). These
sections also present pondage test bank locations and seepage rates and provide an

excellent summary of the site geology which is useful to reference while reading the
body of the report.
n Table 3-1 Wimmera Trial Sites: Summary of Key Characteristics
Characteristic Rocklands Toolondo Donald
Soil / Geology Has not been assessed but Silty sand to medium grained sand Generally over test section:
Surface channel casual observations suggest of 0.05 – 0.1m thickness overlying 0-0.1m: Sand – Silty Sand
soils clayey soils are predominant a generally stiff clay. 0.1 – 0.5m: predominantly Clayey
Sand (Sandy Clay in part)
Predominantly sandy clay in Toolondo Central: Predominantly
Soil / rock beneath
southern half of site overlying clay to sandy clay overlying Surface layer (approx 0.5 - 1m) of
& around channel
sandstone at depth. Sand at 1- sandstone varying across the site clayey sand overlying predomin-
2m rising to surface, also in between 3 – 9m depth. Fine to antly clay, but replaced by fine-med.
south. Sandstone rises to near medium grained sand 1-2m deep grained sand in southern half of trial
surface in northern part of site, located sporadically across site. site, starting near surface and
underlying shallow clay soils. dropping to 2-3m below surface in
Toolondo West: Surface layer the centre of the section
(0.5 - 1m) of weakly cemented fine
sand overlying 1-2m of sandy clay,
overlying medium clay to 7m, with
underlying sandstone
Toolondo East: Heavy – medium
clay to 3-4m, overlying shallow
sandstone
Hydraulic
Average depth of Approx. 1.6m Approx. 1.5m Approx. 1.4m
water at FSL
Exact depth not known. EM31 Exact depth to watertable not 1.5-3m, with 1-2m fluctuations
Average depth to drilling indicates >4m. known. EM31 drilling indicates adjacent channel due to channel
watertable and Potentiometric contours (SKM, >8m. Groundwater bores in area influence
groundwater 1995) indicate likely depth to indicate likely to be 9-10m range.
salinity potentiometric surface is in the
5-7m below NS range.
Toolondo Central: 14-16m Approx. 8 - 10m
Channel Wetted Approx. 15-17m range Toolondo West: approx. 14-16m
Perimeter at FSL Toolondo East: 13-15m
Channel Water 800 EC (µS/cm) 800 EC (µS/cm) 800 EC (µS/cm)

Salinity

n Table 3-2 Murray Irrigation Trial Sites: Summary of Key Characteristics
Characteristic Dahwilly Finley
Soil / Geology
Surface channel soils Predominantly silty clay, overlain by silty crust
(generally 0.05m thick)
Soil / rock beneath & Dahwilly Central: Medium to coarse grained Heavy clay to 8-9m, overlying sandy clay.
around channel sand to at least 10m depth. Sandy clay to clay
loam in top metre along most of channel length
Dahwilly East: Medium to coarse grained

sand to approx 6-8m, underlain by clayey
sand. Sandy clay to clay to 1-2 m at surface
Hydraulic
Typical depth of Approx. 1.25m Approx. 1.5m
channel water at FSL
Average depth to Approx. 4-5m Approx. 1.5m
watertable & Approx. 5000 EC at Central, 2500 EC at East Approx. 20,000 EC
groundwater salinity
Channel Wetted Approx. 8-9m Approx. 9m
Perimeter at FSL
Channel Water 75 EC (µS/cm) 75 EC (µS/cm)

Salinity
Table 3-3 Murrumbidgee Irrigation Trial Sites: Summary of Key Characteristics
Characteristic Tabbita Lake View
Soil / Geology
0.1m silty surface crust, underlain by silty grey
Surface channel soils clay, in turn underlain by a heavy grey clay.
Channel walls intersect approximately 0.5m
interval of clayey sand. Lake View Central - Sandy clay – sandy clay
loam overlying medium to heavy clay starting
Soil / rock beneath & Generally low permeability clays and sandy from between 2-6m below surface
around channel clays to 7-8m
Lake View West - Surface layer (approx 1m)
of sandy clay overlying 3-4m of gravelly clay,
overlying medium clay
Hydraulic
Average depth of Approx. 1.25m Approx. 1.5m
channel water at FSL
Average depth to Approx. 1 - 1.5m Approx. 1.5 m at Central, 0.5-1m at West
watertable & Approx. 8000-10000 EC Approx. 6000 EC
groundwater salinity
Channel Wetted Approx. 8-9m Approx. 10m
Perimeter at FSL
Channel Water 220 EC (µS/cm) 220 EC (µS/cm)

Salinity

4. Non-Geophysical Techniques
This section discusses all non-geophysical techniques assessed during the trials
program. The following type of trials are covered in this section:
q Pondage tests;
q Sub-surface profiling;
q Point Tests;
q Groundwater techniques;
q Point Tests; and,
q Remote Sensing.
4.1 Pondage Tests

4.1.1 Methodology
Principle
The pondage test method applies a water balance to an isolated reach of channel to
determine the seepage losses. Seepage losses constitute the drop in water level over
time in the channel after accounting for evaporation and rainfall (and any potential
diversions). They are widely considered the most accurate means of measuring
channel seepage and generally regarded as the best technique against which other
methods can be assessed. Therefore, as discussed in Section 3, pondage tests were
used as the baseline measurement technique for the trials, and were conducted at every
trial site.
Methodology
In these trials pondage tests were conducted by blocking a section of channel with
embankments at each end and filling the section with water up to, or slightly higher
than, the level at which it usually flows during operation. Most of the pondage tests
were conducted at the end of the irrigation season, and in this situation the banks were
generally constructed while the channel was full, as shown in Figure 4-1 (look for
better photo). To minimise the risk of seepage through the banks, (most) banks were
constructed with a plastic liner. All the pondage cells (except for some at Waranga
Western Channel) were constructed back to back, to minimise bank construction costs.
Water level decline in the channel was measured by a hook gauge which was read
twice daily (morning and afternoon), or daily during some tests. Evaporation was
measured using a class A evaporation pan (located on the channel bank) and rainfall
with a standard rainfall gauge.
Pondage tests were analysed using a spreadsheet developed for the project. The
method of analysis is briefly described below (refer ANCID, 2003 for further detail).
The basic equation for calculating seepage losses using the pondage test method is
presented below. The method of analysis assumes that the only inflow into the reach
is rainfall. Run-off was ignored in the assessment due to the difficulty in estimation.
If the channel is positioned to receive run off and there is sufficient rainfall the test
could be effected. However, for most tests there was no rainfall, and only a handful of
tests received sufficient rainfall to generate run-off.
WC01312:YR3_RPT_CH_4(NON_GEOPH_TECH).DOC Final PAGE 11

n Figure 4-1 Pondage embankment construction (a) and installation of plastic
liner (b)
a)
b)

The following equation assumes that the only outflows are due to evaporation and
seepage (the terms are also depicted in Figure 4-2):
W [(d − d 2 ) − E + R ]
S= 1
P(t 2 − t ) 1
Where:
S = Seepage rate [volume / area / time]
W = Average surface width between t1 and t2 [length]
d1 = Water level at t1 (averaged between u/s and d/s gauges) [length]
d2 = Water level at t2 (averaged between u/s and d/s gauges) [length]
E = Evaporation along reach between t1 and t2 [length]
R = Rainfall along reach between t1 and t2 [length]
P = Averaged wetted perimeter between t1 and t2 [length]
t1 = Time at first measurement of water levels [time]
t2 = Time at subsequent measurement of water levels [time]
Units must be consistent for all terms in the equation.
Pondage test seepage rates are reported in this project in mm/d. However it is very
important to note that this is equivalent to a volume per area per day (m3/m2/d ÷ 1000
which equals mm/d, which is equivalent to L/m2/d). The area in this equation
represents the channel wetted area and not the surface water area, as is sometimes
reported in channel seepage studies.
The pondage test results are presented in this report as the seepage rate between each
measurement interval during the tests. A typical graph of the pondage test results is
presented in Figure 4-2 below. This particular test shows an initially higher seepage
rate due to the wetting up of the soil profile and then levelling out to a reasonable
steady seepage rate after four to five days. The degree to which this higher rate at the
start occurs will be primarily due to the length of time the pond has been saturated
prior to the commencement of readings and the filling height prior to testing.
n Figure 4-2 Components of Pondage Test Water Balance
Qo
L
E
I
S
Water level at t
1
Water level at t
2
d1
d2

Figure 4-3 represents one of the highest cases of initial loss compared to steady state
loss observed during the pondage tests.
n Figure 4-3 Example Pondage Test Results
Seepage Rate Vs Time For Pondage Test One: Donald Main Channel
80
70
60
Seepage Rate (mm/day)
50
40
30
20
10
0
21/12/2000 22/12/2000 23/12/2000 24/12/2000 25/12/2000 26/12/2000 27/12/2000 28/12/2000 29/12/2000 30/12/2000
Time
This figure shows that as well as an overall trend, there are variations between
measurements. This variation is in part due to hook gauge reading error, evaporation
measurement error and may also be due to diurnal seepage variations [refer p7 of the
Literature Review ANCID (2000a) for further details]. The cause of these variations
was not be studied in this project. The figure of primary interest in this study is the
steady state seepage rate, ignoring daily variations. The actual seepage rate used in
the trials is the average seepage rate after any initially high rates or other non-
representative conditions (eg heavy rainfall) have subsided or been allowed for. For
example, the seepage rate number taken from Figure 4-3 was 32 mm/day. This is the
number which will be used for comparison with other techniques and which is
presented in the body of this report.
4.1.2 Results
The results of all pondage tests conducted during the trials is summarised in Table 4-1.
Graphs of all of the pondage test results are presented in Appendix B. Geological long
sections in Appendix A display pondage test bank locations and corresponding
seepage rates for each of the sites.
4.1.3 Discussion
Magnitude of Results
Seepage rates ranged in magnitude across the trials sites from 0.1 mm/d (essentially
zero) at the Toolondo East, up to 48 mm/d on the Donald Main Channel. Seepage
rates were generally lower than anticipated, particularly given that a number of the
trial sites were selected based on suspected high seepage rates.

The average seepage rate across all ponds was 9.7 mm/d (0.0097 m3/m2/d). However
the median rate of 7.0 mm/d (0.007 m3/m2/d) removes the effect of the high seepage
sites and is more representative of typical seepage rates across most of the trial areas.
As discussed in the literature review (ANCID 2000a), pondage tests may provide a
slight underestimate of seepage rates compared to flowing conditions due to increased
clogging effects under no flow conditions.
Some sites which were anticipated to have high seepage rates contained low rates,
while other sites which were not expected to have high rates (and were essentially
randomly selected from a seepage rate perspective) contained high seepage rates. For
example, ponds 1 and 2 at Dahwilly Central (2001 pondage test trials) were expected
to have high seepage rates due to the clean, thick sandy profile beneath the channel.
However as these ponds were located immediately upstream of a check structure, the
silt build up behind the structure caused seepage to be restricted to around 5 mm/d.
Further, on the Waranga Western Channel, a number of seepage sites which had been
mapped as ‘known seepage sites’, due to visible surface evidence of seepage, were
actually found to have very low seepage rates (< 3 mm/d) during pondage tests. This
demonstrates that high rates are not necessarily required to cause surface seepage
effects.
In contrast, the four ponds at Lake View West were not anticipated to have high
seepage rates, (there were no surface signs of seepage etc) and therefore the RWA was
surprised to find seepage rates of 20 – 25 mm/d at the site. Seepage through a loamy
surface material, predominantly through the banks of the channel, is the most likely
seepage mechanism at this site.
Visible evidence of seepage does not necessarily imply high seepage rates (this does
not necessarily mean that it is not of concern however) and conversely the absence of
visible signs of seepage does not mean that significant seepage is not occurring. Even
knowledge of soil types and the sub-surface profile does not always translate to an
accurate understanding of seepage rates or seepage processes. The pondage tests and
assessment of geological and other data help to develop an understanding of the
seepage mechanism. The combined data can help to determine if seepage is vertical
through the base of the channel or lateral through the banks.
It should also be noted that the seepage rate obtained from pondage test data is a bulk
value which represents the average seepage rate over several hundred metres of
channel. Obviously within the pond section the actual seepage rate will vary above
and below the section average. The degree of variability in the seepage rate will
depend on the variability of sub-surface conditions and channel characteristics
(particularly the presence / absence and thickness of a silt layer and bank material and
construction techniques).
Repeatability
At several sites pondage tests were repeated in the same location, (or at least
overlapping), as pondage tests from the previous year. The following comments relate
to the repeatability of pondage tests as observed at these sites.
q Toolondo Central – Pond 4 recorded a seepage rate of 10 mm/d during the March
2001 pondage tests at Toolondo Central. In March 2002 a pondage test was

conducted within this cell, but in a section slightly shorter than the original cell,
in order to straddle an area of suspected higher seepage (based on the EM31
results). The 2002 pondage cell recorded a seepage rate of 12 mm/d, which is
essentially what would be expected for this repeat test;
q Finley – Ponds 1-3 at Finley were conducted in exactly the same locations in July
2001 and July 2002. Rates of 5.2, 5.6 and 4.1 mm/d were recorded in the 2001
tests and 7.0, 5.3, and 5.4 mm/d in the 2002 tests, for Ponds 1, 2 and 3
respectively. The difference between the two years is: 1.8, -0.3 and 1.3 mm/d,
While these rates are comparable, except for pond 2, rates were slightly higher in
2002. The increase in rates could be explained by a deeper watertable at the site
in 2002, creating a steeper discharge gradient.
q Lake View – The four pondage cells conducted in June 2002, covered
approximately the same areas as ponds 3-6 in the July 2001 tests, but the bank
locations were not the same. The following points summarise the overlap (refer
to the Lake View Long Section in Appendix A for a visual presentation of the
location of these ponds) Suffixes refer to year of pondage tests:
q P1(02) = 7.1 mm/d coincided with P3(01) = 9.0 mm/d & P4(01) = 7.0 mm/d
q P2(02) = 5.8 mm/d & P3(02) = 4.3 mm/d coincided with P5(01) = 7.1 mm/d
Using a weighted average method, a comparison of the two areas is:
q 7.1 mm/d 02 vs 8.3 mm/d 01, and
q 5.5 mm/d 02 vs 7.1 mm/d 01,
This represents a decrease of 1.2 mm/d and 1.6 mm/d for the 2002 pondage tests,
or in term of percentage, 16-23%.
In summary, at the three sites where information is available to compare pondage test
results, a reasonable degree of repeatability was observed. Differences of up to 25%
in seepage rate were observed between pondage tests conducted in 2001 and 2002.
These differences are considered acceptable for the purposes of this investigation, and
are not necessarily attributable to error inherent in the pondage test method.
Differences of this magnitude could be attributable to changes in site characteristics.
The two most likely natural characteristics to change, include:
q Changes in depth to watertable - This will have greater effect at sites with
shallow depth to watertable. Apart from the background watertable depth, depth
to watertable can change due to seasonal fluctuations as well as channel related
fluctuations. Depth to watertable at the time of the pondage test will be related
to the length of the time the channel has been running prior to the test or since it
had been running prior to the test;
q Changes in channel bed properties – This is most likely to be caused by natural
siltation of the channel, which will generally increase over time. Man-made
changes such as de-silting of a channel can have dramatic effects on seepage
rates.
4.1.4 Conclusions
Pondage tests involve blocking a section of channel for a period and applying a water
balance to determine the seepage losses. They are widely considered the most
accurate means of channel seepage assessment and were the baseline technique against
which other techniques were assessed. Pondage tests were therefore conducted across
all sites, totalling 81 ponds.

Seepage rates ranged in magnitude across the sites from 0.1 mm/d (Toolondo East) to
48 mm/d (Donald). The average and median seepage rate across all sites was 9.7
mm/d and 7.0 mm/d respectively. Some sites anticipated to have high seepage rates
actually contained low rates (due to surface clogging layer), while others expected to
have low rates were found to have a high rate of seepage. Visible evidence of seepage
was found to not necessarily imply high seepage rates.
At three sites where pondage tests were repeated, a good degree of repeatability was
observed. The maximum difference between seepage rates was 25%. Differences in
pondage tests rates from one season to another are probably attributable to changes in
depth to watertable and channel bed properties. The differences are considered
acceptable for the purposes of this investigation, and not considered to be significantly
due to errors in the pondage test method.

n Table 4-1 Summary of Pondage Test Results
Date of Test Seepage Results by Pond

Rural Water Channel Rate Comment
(no. of cells, (mm/d)
Authority test duration) Range (Pond 1, Pond 2 ……)
(mm/d)
March 01 (6 cells, 13 days ) 1 – 11 (11, 11, 3.7, 10, 1.1, 2.8) High variation in seepage observed in some ponds due to run-off and delayed
Toolondo seepage associated with a relatively large rainfall event on 20th March (20mm)
(Central) March 02 (1 cell, 14 days) 12 (12) Largely centred over pond 4 in 2001 pondage tests; good repeatability.
Wimmera Toolondo (East) March 02 (4 cells, 14 days) 0–1 (0.1, 0.4, 0.4, 0.7) Very low rates
Mallee Water
Toolondo (West) March 02 (4 cells, 15 days ) 1–5 (1, 1.6, 3, 4.9)
Rocklands March 01 (6 cells, 15 days) 4 – 13 (8.7, 11, 13, 5.4, 4.3, 4.3) High variation in seepage observed in some ponds due to run-off and delayed
seepage associated with a relatively large rainfall event on 20th March (20mm)
Donald Main Dec. 00 (6 cells, 9 days) 9 – 48 (45, 35, 36, 37, 48, 9) Mid-range test values were used rather than steady state due to low initial water
levels (approx. two-thirds normal supply level)
June 01 (6 cells, 14 days) 4 – 16 (4.8, 4.4, 13, 7.6, 12, 16) First few days excluded from analysis due to erratic behaviour. Low rates in P1 &
Dahwilly P2 despite sandy profile, due to silt accumulation upstream of check
(Central) June 02 (7 cells, 15 days) 4 – 10 (4.2, 4.9, 9.5, 7.7, 7.8, 1.1, Only P6 & P7 overlap with June 01 tests. P6 & P7 (02) are P1 & P2 (01). These
2.8) two cells (underlined) were remediated in July 2001, reducing seepage from 4.8
Murray mm/d to 1.1 mm/d and from 4.4 mm/d to 2.8 mm/d.
Irrigation Very narrow seepage range
Dahwilly (East) June 02 (3 cells, 22 days) 9 – 10 (10, 10, 9)
July 01 (4 cells, 15 days) 4–6 (5.2, 5.6, 4.1, 3.9) Due to absence of survey data seepage rate calculated based on average gauge
Finley drop multiplied by estimate of wetted perimeter to surface water ratio of 1.05
June 02 (3 cells, 19 days) 5–7 (7, 5.3, 5.4) Ponds 1-3 for 01 and 02 correspond. There is an increase of 0.3-1.8 mm/d
compared to 01. Some of difference may be due to lack of survey data in 01 tests.
July 01 (6 cells, 7 days) 7–9 (9, 9.3, 9, 7, 7.1, 8.5) Very narrow seepage range
Lake View
(Central) June 02 (4 cells, 7 days) 4–7 (7.1, 5.8, 4.3, 5.2) Areas between 01 and 02 tests overlap, but banks were in different locations.
Murrumbidgee Approx. 1-2mm higher rates in 01 tests, possibly due to longer drying period?
Irrigation High rates given soil type. Probably due to seepage through top metre of the
Lake View June 02 (4 cells, 7 days) 20 – 25 (20, 23, 25, 25)
(West) profile
Tabbita June 01 (6 cells, 7 days) 6 – 10 (6.4, 6.0, 9.8, 8.5, 6.2, 6.7)
Goulburn Waranga May/June 02 (11 cells, 1 – 13 (6.1, 3.3, 1, 1.7, 1.4, 1.6 Pondage tests were conducted over a large area: P1 was approximately XX km
Murray Western 12 days) 7.7, 7.7, 7.3, 13, 4.4) from P11
Water

4.2 Sub-surface Characterisation: Soil and Geological
Profiling
4.2.1 Purpose of sub-surface profiling
Sub-surface profiling of soils and geological conditions can be conducted in a channel
seepage investigation for various reasons, including:
1) As part of site characterisation, to identify the distribution of zones of higher and
lower potential seepage (including preferred pathways);
2) To help define seepage mechanisms; and / or,
3) To assign seepage rates to soil types and hence determine seepage through
changes in soil type (this can be conducted at a regional scale, using available soil
maps and published data on seepage rates, or at a local scale involving local soil
mapping and seepage testing)
In this investigation sub-surface profiling was conducted for the first two reasons. Site
characterisation is an important component of any channel seepage investigation and
site stratigraphy is probably the most important aspect requiring characterisation.
Interpretation of test results of essentially all techniques will be underpinned by the
conceptual understanding of the site, and therefore it is important that this is
established as accurately as possible.
As the trial program developed and the focus of the investigation became more
strongly focussed on geophysics, it was important that the characterisation of the sub-
surface was particularly targeted to assist in geophysical interpretation. In addition to
information on lithology, geophysical interpretation is assisted by information on soil
moisture, depth to watertable and groundwater salinity.
4.2.2 Data collection

Information on the sub-surface was collected via drilling bores. The key issues
addressed in developing a drilling program for each site were:
q Where and how many bores to drill;
q What depth to drill to, and,
q What type of drilling to use;
q How to log the drilling.
All of these issues were tightly constrained by cost. The approach adopted in this
study is described below:
q Location and number of bores: Bores were drilled immediately adjacent the
outside toe of the channel banks (or as close as practical to the bank). This
coincided with the location of the EM31 and EM34 surveys adjacent the channel
bank. While drilling in the channel may also be of assistance to determine the
stratigraphy directly beneath the channel, this is generally not practical due to the
considerable expense of drilling from a barge or boat. Interpolation between
drilling results on either side of the channel, provided by stratigraphic
interpretation, is sufficiently accurate.
The greater variability at the site, the greater the number of bores required to
characterise the site. In the final year drilling program (year 3) approximately 12-

14 bores were located along each channel section (300m – 1000m). Bores were
usually located in order to cover the range of geophysical responses at the site, as
well as in locations representing changes in geophysical response, which
potentially represent changes in geology.
q Depth: The most influential depth on seepage rates is considered to be the upper
profile, particularly the top 2-3 metres. The depth of drilling may be limited by
the type of drill rig employed, which in turn may be controlled by cost
constraints. The minimum depth of drilling in this study was to 4-5m, which
coincides with the approximate penetration depth of EM31. When resistivity
surveys were used in the final year of the trials, the drilling was increased to 10m.
While the resistivity method employed in these trials ‘sees’ to approximately
20m, a depth limit of 10m was selected due to budget constraints. At some sites
drill refusal was reached before 10m. Approximately every third bore was drilled
to 10m, and remaining bores to approximately 7-8m.
q Type of drilling: Drilling was generally conducted using solid stem augers. This
drilling method causes some disturbance of the samples but even given the
sampling limitations, this method was considered suitable for the level of
accuracy required for this investigation.
The trailer mounted drill rig employed at most of the sites for the drilling is
shown in Figure 4-4.
q Logging: Ideally a geologist or soil scientist should be available to log the bores.
One of the most important aspect of logging is consistency which is best achieved
by logging to recognised standards (such as the United Soil Classification System
or an Australian standard such as the Northcote system). Logging in these trials
was based on the Northcote classification system.
Across the 11 sites, drilling was undertaken by the two EM31 contractors, who
each have significant drilling and logging experience. Soil samples at one metre
intervals were collected in the third year drilling program to provide a record of
drilling and a means of checking on the logging results.

n Figure 4-4 Trailer mounted drill rig used for soil bore drilling
4.2.3 Results
Results of the drilling program are presented in geological long sections of each of the
channels are presented (Appendix A). To determine whether averaging of soil
properties could assist in estimating seepage, average permeabilities were compared
with pondage test seepage rates.
Based on the geological long sections, a bulk vertical hydraulic conductivity was
calculated for each of the ponds. Hydraulic conductivity was based on text book
vertical hydraulic conductivities for the given soil texture. The vertical hydraulic
conductivity for each soil type was assigned using published rates based on the
Northcote classification system (Reference, 19XX). Layered heterogeneity was
accounted for using the equation outlined in Freeze and Cherry (1979, p34). For each
pond, four different vertical hydraulic conductivities were calculated. These are
described below and graphically presented in Figure 4-5:
i) Weighted according to typical EM31 response - most heavily weighted

around 1 – 3m depth and terminated at 7m;
ii) Representative of surface permeability and evenly weighting over the top
2m of the profile;
iii) An average across the top 10m of the profile but giving more weight to
upper surface layers and less to layers at depth; and,
iv) Evenly weighted across the top 10m of the profile.
Details of the calculations, and an example calculation are provided in Appendix C.

n Figure 4-5 Cumulative Weighting Applied to Different Soil Permeability
Explanatory Variables
Cumulative Response Curves for Different Soil Permeability Weightings
100
Cumulative Weighting Applied to Soil Permeability
80
Explanatory Variable (%)
60
40
EM31 Weighted
Evenly weighted to 2m
20
Surface Weighted to 10m
Evenly Weighted to 10m
0
0 2 4 6 8 10
Depth (m)
Figure 4-6 presents the vertical hydraulic conductivity (Kv) value for the upper two
metres and the 10m inversely weighted scenarios, for all of the sites investigated in the
year 3 program. The absence of a clear correlation between Kv and seepage indicates
n Figure 4-6 Pondage Test Seepage Versus Representative Vertical Hydraulic

Conductivity for the Pondage Section (for Year 3 Drilling Sites)
Soils (Kv) vs Seepage
50
Upper 2m
Inverse Depth
40
Weighted (10m)
30
20
10
0
0 1 2 3 4 5 6 7 8 9 10
Bulk Vertical Hydraulic Conductivity (Kv, m/d)

that this approach may not be valid, or alternatively that no clearly defined direct
relationship does exist.
Reasons why this approach may not be valid is due to errors inherent in the method.
These include:
q The bulking and averaging process (of both seepage and Kv values) over the pond
length;
q Interpolation between bores; and,
q The process of assigning Kv values to soil type (ie infiltration tests were not
conducted in each soil type and the true conductivity is assumed based on the
described texture).
However, if it is assumed that the errors inherent in the method are relatively minor,
and that Figure 4-6 is accurate, two main factors could cause the scatter observed in
the data. Two main types of variation from an ‘expected’ trend, can be observed in
Figure 4-6:
q Sites with low vertical conductivity but relatively high seepage rates – There are
three main reasons why this may occur:
i) Seepage may be isolated to very short sections of high seepage which may
have been missed by the drilling program. Many measurements are required
to obtain a reliable estimate of the mean. Seepage will tend to be
underestimated as local high seepage zones go undetected.
ii) The predominant seepage mechanism could be lateral and shallow, through
the channel banks or immediately beneath the channel base – therefore soils
data collected at the edge of the channel will not be highly related to seepage
rates. In this case seepage will be dependent on the bank hydraulic
properties which in turn will be dependent on bank construction techniques
and materials.
iii) The predominant seepage mechanism is not through primary soil porosity
but through secondary porosity, such as cracking, or fractures (in hard rock),
or macro invertebrate activity (eg yabbies). Assigning a seepage rate based
on soil texture will therefore not be an accurate indicator of the true
permeability of the soils.
q Sites with high vertical conductivity but relatively low seepage rates – This
occurs where natural lining or clogging of a channel via silting or biological
processes is the dominant factor controlling seepage, ie the surface layer is the
most restrictive layer in the profile. This effect will be most prominent at sites
with high permeability soils, as the silt layer is relatively more influential and
more likely to be the most restrictive layer.
Summary
This analysis suggests that using soil data alone to estimate seepage rates is not likely
to be successful. The drilling undertaken in this study was not of sufficient density to
adequately characterise the site to determine seepage rates. Therefore the density of
sampling required to accurately represent seepage rates will generally be prohibitive
for most RWA channel seepage assessment studies. In addition, for this method to be
satisfactorily accurate, field assessment of seepage rates in different soil types is likely
to be required, rather than estimation based on the soil texture and published data.

Finally, the effects of surface clogging can often dominate seepage rates to the extent
that the permeability of underlying layers can be irrelevant.
4.2.4 Conclusions
Sub-surface characterisation by soil and geological profiling was conducted to assist
in general site characterisation (provided information on soils, depth to groundwater
and groundwater salinity) as well as to assist in geophysical interpretation. Bores
were generally drilled adjacent to the channel, up to 10 m in depth.
An attempt was made to estimate seepage based on average soil permeabilities, using
different weightings to test the influence of the soil across a range of depths. The
upper 2m of the soil profile gave the best indication of some relationship between
permeability and seepage rate, however no clear relationship between soil
permeability and seepage rate was obtained. A combination of factors is likely to
contribute to the absence of a relationship. Two types of factor contribute to the
absence of a clearly defined relationship between seepage and soil Kv:
q Errors inherent in method – There was insufficient definition of changes in soil
type along channel (ie low sampling density). Further, the process of assigning
Kv to soil type is inaccurate. The hydraulic conductivity for the particular soil
type should be field tested rather than assigned from literature
q Factors apart from soil type are the primary control on seepage rate: The two
most common factor are:
q Bank dominated seepage (ie due to poor bank construction etc) and,
q Surface clogging layer.
These factors explain why sites like Finley and Dahwilly can have such similar
seepage rates, even though the underlying soil at Dahwilly has permeability many
orders of magnitude higher than the clay at Finley. Seepage rates at Dahwilly are
controlled by the clogging layer on the base of the channel while seepage rates at
Finley are controlled by lateral bank seepage.
Soil mapping at a regional level to identify sites of potentially high seepage is a useful
first cut approach for prioritising areas for further investigation. Sub-surface
characterisation is also a very important part of developing a conceptual model of the
site, which will be used in interpreting other test results. However at the density
required for sufficient accuracy to be obtained compared to other methods, local soil
mapping (and point seepage tests) to quantify seepage rates and extrapolate to other
areas will not be cost effective, aside from the potential errors in this technique
described above. These conclusions regarding the usefulness of soil surveying are
confirmed by the literature review (ANCID, 2000a).

4.3 Point Tests
4.3.1 Introduction
Point measurement refers to any technique which measures seepage at a given point
within a channel, usually involving the application of water to the surface or
constructed hole within the channel and measurement of the rate at which it drains
away. Point tests can be divided into techniques which can be used while the channel
is operating and those used when the channel is empty.
Some techniques provide a direct estimate of seepage (eg seepage meters), whereas
others estimate hydraulic conductivity, which can be used as a relative indicator of
seepage, or can be used to calculate seepage. The advantage that channel operating
techniques have over channel empty techniques is that measurements reflect real
operating conditions, particularly the seepage processes and hydrogeological
conditions.
Five point test trials were conducted during the investigation and are reported on in
this section, including:
q Ring Infiltrometer and Disc Permeameter tests (Toolondo Central and Dahwilly
Central);
q Ring Infiltrometer (Tabbita); and,
q Idaho Seepage Meter (Dahwilly Central and Donald Main Channel).
These three techniques are briefly described below (refer to ANCID, 2000a, for
additional information).
Idaho Seepage Meter

The most common form of point measurement conducted under channel operating
conditions is the seepage meter. Seepage meters are essentially cylindrical
infiltrometers modified for use under water. The seepage meter method involves the
use of a water tight bell housing embedded into the channel bed where the water lost
per unit area through the base of this bell is the seepage loss from the channel.
Seepage meters generally are based on either a variable or constant head. The Idaho
seepage meter is a type of constant head seepage meter, which operates using a
Mariotte siphon reservoir from which seepage rate can be calculated directly from the
rate of fall in the reservoir level. Seepage meters cannot be used in deep channels,
fast flowing channels, or in channels with hard or gravelly beds, or where there is
significant weed growth.
Disc Permeameter
The disc permeameter is an instrument used to measure the hydraulic conductivity of
soil at (or near to) saturation. A disc covered with a semipermeable membrane
(typically 0.2 m in diameter) is placed on a surface and the subsequent infiltration of
water allows calculation of the hydraulic conductivity of that surface. Water is
supplied to the disc at a constant head at or near to a surface matric potential of zero.
This measurement method can only be used in empty channels. The method can be
used on an undisturbed channel bed, in some cases a thin layer of contact material may
be used to provide a level surface but this does not affect the measurement. In this
study, unsaturated disc permeameters were used and a small negative head was

applied to the disc to avoid leakage of water. This may result in the measured values
of hydraulic conductivity being marginally (but not significantly for the purposes of
this study) lower than the saturated value. Soil samples were extracted before and
after use of the disc permeameter and analysed for volumetric soil water content.
Ring Infiltrometers
Single and double ring infiltrometers are devices for determining the rate of
infiltration into soil from a circular source. Single ring infiltrometers were used in
these trials. Ring infiltrometers are normally metal rings with a diameter of 20 – 100
cm and a height of about 20cm. The ring is driven into the ground about 5 – 10 cm,
water is applied inside the ring with a constant head device, and intake measurements
are recorded until a steady infiltration rate is observed. If a constant head device is not
available to add water to the ring, a constant head can be maintained by manually
adding water and recording the volume. In these trials both manual and automatic
methods were used to maintain a constant head.
4.3.2 Methodology
The approach adopted in all the trials was also applied to the points tests. That is, that
the results were tied back to the pondage tests to attempt to determine their accuracy.
Depending on the pond length, a certain number of point tests were conducted at
approximately even intervals (where possible) along the pond. Based on the length of
channel represented by each test, a weighted average was applied to determine the
representative seepage from that pond. (If the tests were exactly evenly distributed
along the pond the representative seepage would be equivalent to the test average).
Due to the effect of very high individual seepage rates on the overall average, in some
instances the median rate was also examined as a potentially more representative
number.
In addition to comparison with pondage tests, at some sites the point tests were
compared with the immediately surrounding EM31 values and potential relationships
between these two variables examined.
4.3.3 Ring Infiltrometer and Disc Permeameter – Toolondo Central

Ring infiltrometer and disc permeameter tests were conducted on the Toolondo
channel on December 17th - 19th December 2000 and 8th - 9th January 2001. Tests
were conducted at intervals ranging between 20m and 40m along the channel, in three
corresponding pondage sections:
q one of relatively high seepage (Pond 2; 10.6 mm/d); and,
q two of relatively low seepage (Pond 3; 3.7 mm/d and Pond 6; 2.8 mm/d).
Figure 4-7 shows the location of the tests. The ring infiltrometer test and the disc
permeameter test were conducted as close to each other as possible (typically 1-2m).
The results of the tests are shown in Table 4-2, with the average, median and lower
quartile of the results presented. Figure 4-8 shows the instruments in use at the
channel.

n Figure 4-7 Toolondo Point Test Locations and Results
n Figure 4-8 Ring Infiltrometer (left) and Disc Permeameter (right) in use at the
Toolondo Channel

Comparison to Pondage Tests
The most obvious fact from the data is the high values recorded during the tests.
These are very high compared to the pondage test seepage rates for the cells.
Different results to pondage test numbers are expected as these results are not seepage
rates but saturated hydraulic conductivity values. This is closely related to the seepage
rate, but will also depend on other factors. The most important is the head of water in
the channel and the watertable elevation relative to the channel. For the conditions at
the Toolondo channel (watertable approximately 10m below channel water level and a
wetted perimeter of approximately 15m), based on some simple assumptions
regarding the flow pattern beneath the channel flow net analysis indicates that that
seepage rates of approximately two-third the hydraulic conductivity are expected
Therefore, recognising that saturated hydraulic conductivity values are not directly
equal to actual seepage rates, but that they should generally be similar, (and that in this
case seepage rates should be approximately two-thirds of the hydraulic conductivity),
it is evident that these values are much higher than they should be, if they are a true
reflection of seepage rates. Even the lower quartile values of the data represent
seepage 30-70 times greater than pondage test seepage rates.
The reason for these results is due to the soil profile in the base of the channel. Over
most of the channel in this section, a medium to coarse grained sand layer of 0.1- 0.2m
thickness overlies a silty clay. It is the clay which acts as the limiting layer to channel
seepage, but the ring infiltrometer and disc permeameter tests (which were conducted
on top of the channel) are essentially measuring the hydraulic conductivity of this sand
layer. In a select few of the tests (eg Pond 2, test 7 and Pond 3, test 4) it appears that
this sand layer was quite thin and values approaching the true field saturated hydraulic
conductivity for the restricting layer may have been measured.

n Table 4-2 Toolondo Channel Ring Infiltrometer and Disc Permeameter
Results (Field Saturated Hydraulic Conductivity)
Ring Infiltrometer Disc Permeameter

Test No. (mm/day) (mm/day)
P2-T1 1620 2248
P2-T2 1340 866
P2-T3 720 512
P2-T4 516 407
P2-T5 1830 793
P2-T6 1067 749
P2-T7 42 71
P2-T8 349
P2-T9 1424 2327
P2-T10 2130 2201
Average 1299 1082
Lower Quartile 1067 449
Median 1424 771
P3-T1 2017 1938
P3-T2 2442 870
P3-T3 2374 699
P3-T4 35 26
P3-T5 1255 2378
P3-T6 173 212
P3-T7 112
P3-T8 152 226
Average 684 708
Median 163 226
P6-T1 237 51
P6-T2 118 103
P6-T3 104 295
P6-T4 860 274
P6-T5 2260 2310
P6-T6 2190 2303
Average 962 889

Median 549 285
The test methodology aims to ensure measurements are conducted under saturated conditions. In reality, field
saturated hydraulic conductivity is often half of that under saturated conditions. The relationship between the true
saturated hydraulic conductivity (Ks) and the field saturated hydraulic conductivity (Kfs) is approximately: Ks = Kfs x
1.2.
Given that there was little to be gained from direct comparison to the pondage tests
rates, analysis was conducted to compare in a relative sense the point tests results to
the pondage test results. Figure 4-9 presents these results, where the average, median
and lower quartile figures for each pondage section were plotted against the
corresponding pondage test seepage rate for the section.
The first thing to observe from this data is that the point tests did record lower seepage
rates in the lower seepage ponds than in the higher seepage ponds, which is
demonstrated by the positive gradient in the lines of best fit in Figure 4-9. This is
important, as it indicates that these particular points tests, despite providing no
quantitative indication of seepage rates, may still be able to be used to determine
sections of channel with higher rates relative to other sections. Meaningful statistical
interpretation however is not possible with only three data points. Nevertheless, the
lines of best fit and their correlation coefficient have been plotted, primarily to

compare: a) the two point source methods, and b) the median, average and lower
quartile values generated from the point source data as indicators of the overall
seepage from the pond.
Figure 4-9 indicates the best statistical indicator of seepage for the Toolondo point
tests is the lower quartile of the data. This suggests that the high seepage rate results
are unreasonably biasing the true seepage results, which is more realistically reflected
in the lower end of the range of the seepage results. The data offers no clear
indication as to which of the methods might be a more accurate predictor of seepage,
but marginally indicates more confidence in the disc permeameter results. For
example, pond 2 has a pondage test seepage rate approximately 3 times that of pond 3
and 6. The median and lower quartile results for pond 2 are approximately 2-3 times
that of pond 3 and 6, while for the ring infiltrometer pond 2 is 4-6 times higher.
In terms of field procedures, the disc permeameter was more user friendly for
operators. Some inaccuracies were introduced to the ring infiltrometer method by the
float valve mechanism used to maintain the constant head of water in the ring, which
would sometimes stick and require manual fixing, and therefore the required
consistent head was not always continually sustained.
n Figure 4-9 Toolondo Pondage Test Seepage vs Point Source Tests Figure
Pondage Test Seepage Rates vs Ring Infiltrometer Pondage Test Seepage Rates vs Disc Permeameter Seepage
n
1600 1600
Average Average
1400 First Quartile 1400 First Quartile
Ring Infiltrometer (Kfs, mm/d)
Ring Infiltrometer (Kfs. mm/d)
Median y = 138.03x - 47.498 Median y = 36.238x + 693.74

1200 R 2 = 0.8906
1200
R 2 = 0.7354
1000 1000
800 800
y = 121.61x - 223.26 y = 67.089x + 58.109
600 600
R 2 = 0.9967 R 2 = 0.9827
y = 61.096x + 645.17
400 400
R 2 = 0.7684
y = 35.564x + 73.312
200 200
R 2 = 0.9694
0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
Pondage Test Seepage (mm/d) Pondage Test Seepage (mm/d)
Comparison of the Two Techniques

Figure 4-10 (A and B) compares the two point test techniques. There is only a
moderate degree of correlation between the two methods of testing. While differing
absolute numbers might be expected, a stronger correlation between the two
techniques was anticipated. Figure 4-10 B in particular illustrates the high number of
tests for which there is significant discrepancies between the results, and this is
reflected in the correlation coefficient of 0.52. Such discrepancies are due either to
local differences in soil properties, inherent equipment error or operator error. Tracing
the source of such errors is beyond the scope of this project, however it highlights the
variability and resultant uncertainty in point test measurement techniques.

n Figure 4-10 Toolondo Ring Infiltrometer Results Plotted Against Disc
Permeameter Results
A
Toolondo Point Test Results (December 2000 / January 2001)
2500
Ring Infiltrometer
Field Saturated Hydraulic Conductivity (Kfs, mm/d)
Disc Permeameter
2000
1500
1000
500
0
4600 4800 5000 5200 5400 5600 5800 6000
Approximate Chainage (m)
B
Ring Infiltrometer vs Disc Permeameter Results (Toolondo)
2500
Disc Permeameter Field Saturated Hydraulic
2000
Conductivity (Kfs, mm/d)
1500
y = 0.7582x + 135.24
R2 = 0.5272
1000
500
0
0 500 1000 1500 2000 2500
Ring Infiltrometer Field Saturated Hydraulic Conductivity (Kfs, mm/d)
Conclusions
The following concluding remarks can be made regarding the Toolondo point tests:
q Both ring infiltrometer and disc permeameter results significant overestimated
pondage test seepage. Even the lower quartile values of the data represent
seepage 30-70 times greater than pondage test seepage rates. This was due to the
fact that the tests were measuring the conductivity of the sand layer in the
channel, rather than the more restrictive underlying layer.
q The points tests recorded lower seepage rates in the lower seepage ponds than in
the higher seepage pond. This result does not preclude the use of points tests for

determine relative seepage. The disc permeameter appeared to estimate relative
seepage slightly better than ring infiltrometer.
q The side by side disc permeameter and ring infiltrometer results only compared
moderately well with each other, with significant discrepancies particularly at the
higher seepage rate range.
A true evaluation of point test techniques at the Toolondo channel site should involve
repeating the tests, but after removal of the layer of sand covering the restricting layer.
However, enough data was collected in these trials to conclude that the variability in
the soil (and the associated large number of tests required to characterise a given
section), the technical expertise required to properly conduct and analyse results, and
the inherent limitations of the equipment, generally do not make channel empty point
tests cost efficient or technically accurate. Channel full point tests (eg Idaho seepage
meter) would probably have more chance of succeeding under the particular
conditions encountered in the channel bed at the Toolondo channel.

4.3.4 Ring Infiltrometer and Disc Permeameter – Dahwilly Central
Ring infiltrometer and disc permeameter tests were conducted on the Dahwilly
channel on 21st - 24th August 2000. The tests were conducted on three adjoining
pondage sections: Pond 3 (13 mm/d), Pond 4 (7.6 mm/d) and Pond 5 (11.6 mm/d).
The results of the point tests are presented in Table 4-3 and although the disc
permeameter results fall in a very tight band (within 2-3 mm).
Figure 4-11 displays the location of the tests. The ring infiltrometer test and the disc
permeameter tests were conducted as close to each other as possible (typically 1-2m)
at intervals generally between 30m - 40m.
This figure illustrates that not all of the tests were evenly spaced within the pondage
section, and in fact six tests lay outside the areas included in the pondage tests (this
was due to a misunderstanding on the part of the RWA regarding the location of the
pondage tests). Unfortunately this limits the analysis of the results, as pond 3 and 5
have only several tests points contained within them.
Table 4-3 and Figure 4-13 indicate that the disc permeameter produced more
consistent results than the ring infiltrometer. As Table 4-3 indicates, a number of the
ring infiltrometer results had to be discarded due to irregularities observed during field
measurement. Generally this was due to seepage appearing on the surface around the
outside of the ring, indicating that the assumed flow conditions for the test were
breached. These results are therefore likely to be substantially higher than they should
be. (A figure of 20 mm/d was used for analysis purposes in these instances, as shown
by the bracketed result in the table). Due to further uncertainty this introduces to the
results, the disc permeameter results are the more reliable data set for analysis
purposes.

n Table 4-3 Dahwilly Channel Ring Infiltrometer and Disc Permeameter
1
Results (Field Saturated Hydraulic Conductivity )
Ring Infiltrometer Disc Permeameter

Test No. (mm/day) (mm/day)
P3-T1 2.8 5.4
P3-T2 12 4
P3-T3 8.0 2.5
P3-T4 31 16.0
Average 13.5 7.0
Lower Quartile 6.7 3.6
Median 10.0 4.7
P4-T4 31 16.0
P4-T5 1.0 2.3
P4-T6 30 18.8
P4-T7 3.1 12.5
P4-T8 4.7 7.0
P4-T9 115 (20) 6.0
P4-T10 2.1 6.4
Average 13.1 9.8
Median 4.7 7.0
P5-T10 2.1 6.5
P5-T11 98 (20) 13.9
P5-T2 412 (20) 4.8
Average 14.0 8.4

Median 20.0 6.5
1. The test methodology aims to ensure measurements are conducted under saturated conditions. In reality, field
saturated hydraulic conductivity is often half of that under saturated conditions. The relationship between the true
saturated hydraulic conductivity (Ks) and the field saturated hydraulic conductivity (Kfs) is approximately: Ks =
Kfs x 1.2.
2. Bracketed figures represent the figure used in the actual analysis (average, lower quartile etc). The actual number
recorded was not used due to irregularities observed during the test (most commonly due to seepage appearing on
the surface around the outside of the ring), and is likely to be a significant over estimate of seepage.
Comparison to pondage tests

Analysis was conducted to compare the point tests results to the pondage test results
for each section. The first point to note is that, compared to the point tests at
Toolondo, these seepage rates are of the right order of magnitude. For a watertable
depth of 5m and wetted perimeter of 10m, we would expect pondage seepage (in
terms of volume/area/day) to be very approximately half the saturated hydraulic
conductivity (based on some simple assumptions regarding seepage flow patterns
beneath the channel). The results are therefore well within an order of magnitude of
the true seepage rate.
Figure 4-12 presents these results, where the average, median and lower quartile
figures for each pondage section are plotted against the corresponding pondage test
seepage rate for the section. The disc permeameter results interestingly suggest a
negative relationship between pondage test seepage and the point tests (ie pondage test
seepage increasing with decreasing disc permeameter seepage), although the disc
permeameter results fall in a very tight band (within 2-3 mm).

n Figure 4-11 Dahwilly Point Test Locations (August 2000)
n Figure 4-12 Dahwilly Pondage Test Seepage vs Point Source Tests

Dahwilly P o ndage Test Seepage Rates vs Ring Infiltro meter Kfs
Dahw illy Pondage Test Seepage Rates vs Disc Permeameter Kfs
14 14
Average
13 13
First Quartile
12 12 Median
11 11
10 10
9 9
Average
8 8
First Quartile
7 7
Median
6 6
0 5 10 15 20 0 5 10 15 20
Ring Infiltrom eter (Kfs, m m /d) Disc Perm eam eter (Kfs, m m /d)
This demonstrates the failure of the disc permeameter results to characterise the true
seepage rate of the pond. This is most likely due to the inadequate density of the
testing program. The ring infiltrometer results are generally also inconclusive. The
average rates lie in a very narrow band, and do not distinguish between ponds. The

median results suggest some type of trend between the variables, and the lower
quartile data produces the best line of fit. However the high degree of uncertainty in a
number of the infiltration tests means that there can be little confidence in this
relationship (eg, Pond 5 has only 3 tests points and only 1 reliable test, so figures such
as median and lower quartile begin to lose their meaning). Again the conclusion must
be that there is inadequate sampling density to sufficiently characterise the ponds.
Comparison of the Two Techniques

Figure 4-13 plots the results of the two techniques against each other. This figure
shows that the results of the two techniques are very poorly correlated. While
different numbers may be expected from each technique, it would be expected that
overall there would be a strong correlation between the results. The figure shows that
there are a high number of tests for which there are significant discrepancies, and this
is reflected in the very poor correlation coefficient of 0.11. Such discrepancies are due
either to local differences in soil properties, equipment malfunction or operator error.
As discussed above, it appears that the greater source of error is in the functioning of
the ring infiltrometer. In addition to breaches of assumed seepage paths, as discussed
under the Toolondo point test section, some inaccuracies are introduced to the ring
infiltrometer method by the float valve mechanism which is used to maintain the
constant head of water in the ring. Due to the reduced number of reliable results from
the ring infiltrometer, the disc permeameter results are considered the more reliable
data set for this analysis.
n Figure 4-13 Dahwilly Point Tests: Ring Infiltrometer Results plotted against
Disc Permeameter Results
Dahwilly Disc Per meamet er vs Ring Inf ilt r omet er Field Sat urat ed Hydr aulic Conduct ivit y
35
30
y = 0.256x + 7.3482
Ring Infiltrometer (Kfs)
25
R 2 = 0.1133
20
15
10
0
0 10 20 30 40 50 60
Disc Perm eam eter (Kfs)
Conclusions
The following concluding remarks can be made regarding the Dahwilly point tests:

q Point tests were not evenly distributed across the pondage sections which limited
the conclusions able to be drawn from the results (two of the three ponds had only
several points tests contained within them).
q The results correlated poorly with pondage test seepage rates. In fact, the disc
permeameter results (which in these tests were deemed to be more reliable than
the ring infiltrometer results) showed an inverse relationship to pondage test
seepage. This demonstrates the failure of the disc permeameter results to
characterise the true seepage rate of the pond at the density of the testing program
employed in the trial. This confirms findings of investigations documented in
the literature – that one of the greatest barriers to being able to reliably use point
test data is the variability in the soil, and many tests are required to adequately
characterise a given section using this method.
q Unlike the Toolondo point tests, both the ring infiltrometer and disc permeameter
results were of the same order of magnitude as the pondage test seepage results,
which demonstrates that the upper layer at this section of the Dahwilly channel is
the limiting layer (a point confirmed and discussed in section 4.2, Sub-Surface
Characterisation)
q The results of the two techniques were very poorly correlated. This is more likely
attributable to error in the seepage rings than the disc permeameter results.
4.3.5 Idaho Seepage Meter - Dahwilly Central

Idaho seepage meter tests were conducted on the Dahwilly channel in February 2001.
The tests were conducted by Yanco Agricultural Institute, with assistance provided by
Murray Irrigation. Twelve locations were tested in four pondage sections comprising
P3, P4, P5 and P6. Table 4-4 presents the results of the testing at Dahwilly and Figure
4-14 displays each test location. There were six individual tests undertaken at each
test location at right angles to the channel [Left (L), Left of Middle (LM1), Middle 1
(M1), Middle 2 (M2), Right of Middle (RM1) and Right (R)].
The median value of these six tests at each particular chainage is presented in Table
4.4 below as the representative seepage rate for each location. Using these figures a
representative seepage rate for each pondage section was calculated. This was
calculated by summing weighted seepage rates for each test. The value of each
weighted result is dependent on the length of channel the test represents as a fraction
of pond length (ie based on geographic location alone and not on soil type or EM31
response).

n Table 4-4 Dahwilly Idaho Seepage Meter Results: January, 2001.
Pond No. Test Median Seepage Chainage Representative Weighted
(Pondage test No. Rate (m) Distance Seepage
seepage) (mm/d) (m) (mm/d)
1 33 1345 25 4
2 32 1376 25 4
3
(13 mm/d) 3 49 1411 60 13
4 15 1511 110 7
Sum 28
5 81 1586 40 16
4 6 151 1611 40 30
(7.6 mm/d) 7 21 1711 120 13
Sum 59
8 37 1811 100 19
5
(11.6 mm/d) 9 5 1911 100 3
Sum 21
10 79 2011 60 24
6 11 38 2036 50 9
(16.3 mm/d) 12 36 2103 90 16
Sum 49
4.3.5.1 Comparison with Pondage Test Results

An attempt was made to compare these representative seepage rates with the pondage
test seepage rates. As Figure 4-15 indicates, the results are inconclusive. There
appears to be some type of linear trend between three of the points, but the lowest
pondage test result of 7.6 mm/d (Pond 4) returned a high Idaho meter seepage rate (as
calculated according to this method) of 59 mm/d which throws any sort of correlation
in the results.
Two main problems with this analysis are apparent. The first is that the number of
tests which lead to the representative seepage rate are too few, given the length of the
ponds (approximately 2-4 tests per 200m length). Ideally one test should represent no
more than 20-30m length of channel and previous studies indicate a much higher
density may be required [Smith and Turner (1982) found that for a section of channel
366 metres long, 110 measurements (almost one every three metres) were required for
a 20% error in the overall seepage rate]. The second problem was discussed in the
previous section: that is that statistically, four points are not ideal for determining
correlations between variables. However due to time and financial constraints, the
tests were limited to the four pondage cells only.

n Figure 4-14 Idaho Seepage Meter Test Locations (January, 2001)
A general conclusion that can be drawn is that the seepage rates as measured by the
Idaho seepage meter are consistently higher than the pondage test seepage rates.
Normally when point tests are conducted, values significantly lower than actual
seepage rates are detected. This is due to the relatively large contribution of ‘hot
spots’ to the overall seepage rate, and the inability to detect these spots with point
tests. The cause of a variation from this trend might be that the bed of the channel
actually seeps significantly more than the banks of the channel. Consideration of the
Dahwilly channel geological long-section (refer Appendix A) lends support to this
theory. This section shows that along the test section, the base of the channel
consistently intersects the top of the fine sand layer, which is overlain by a sandy clay.
It is therefore reasonable to presume that the base of the channel would seep at a
higher rate than the banks, which would lead to point tests in the base of the channel
overestimating seepage compared to pondage tests which measure seepage across the
entire wetted perimeter of the channel.
It is also acknowledged that pondage test seepage rates may be slightly lower than
actual seepage rates due to settling of suspend solids under still water conditions,
which will be significantly reduced while the channel is running. This alone would
not account, however for the large differences observed between the pondage test and
the Idaho meter seepage rates.

n Figure 4-15 Idaho Seepage Meter Results vs Pondage Test Seepage
Dahwilly Idaho Seepage Meter Results (January 2001) vs Pondage Test Seepage
70
60
Idaho Meter Seepage Rate (mm/d)
50
40
30
20
10
0
6 8 10 12 14 16 18 20
Pondage Test Seepage Rate (mm/d)
4.3.5.2 Comparison with EM31 Results

An attempt was also made to correlate the Idaho seepage meter tests and the EM31
2002 survey data. The four to five EM31 values immediately surrounding an
individual Idaho test were averaged to obtain the representative apparent conductivity
for that location. The results of this analysis are presented in Figure 4-16. There was
one apparent outlier in the results, as highlighted in the figure. Trend lines with both
the outlier included and excluded are presented.
This figure shows an ‘expected’ inverse correlation between the EM31 conductivity
and the Idaho seepage meter results, ie increasing seepage with decreasing EM31
conductivity. However the correlation even with the outlier excluded is poor. The
EM31 conductivity average did not distinguish between seepage rate sites between 5 –
40 mm/d (as measured with the Idaho meter), but did to identify the two highest
seepage sites, both of which were 80 mm/d.
Definitive conclusions cannot be drawn from these results as the conductivity range is
very narrow and the data set too small. It appears that the conductivity has
distinguished between moderate and high seepage zones. The limited point
methodology of the assessment may also contribute to variability in the results.
Improvements could probably be made by averaging seepage and conductivity
readings over 20-30m lengths of channel, rather than lines across the channel. This
bulking process could assist in smoothing out local variations and anomalies that may
occur in isolated sections of channel.

n Figure 4-16 Idaho Seepage Meter Results vs EM31 (2002) Apparent
Conductivity
Dahwilly: Idaho Seepage Rates Versus EM31 Conductivity
160
apparent oulier
140
Idaho Meter Seepage Rate (mm/d)
120
100
80
Fitted line with outlier included
y = -8.49x + 402.67
60
R2 = 0.09
y = -8.14x + 378.96
2
40 R = 0.25
Fitted line with outlier excluded
20
0
38 39 40 41 42 43 44 45
Av. Surrounding EM31 Conductivity (mS/m)
4.3.6 Ring Infiltrometer - Tabbita

Ring infiltrometer tests were conducted on the Tabbita channel over the period 18th -
20th July 2001 by a soils specialist sub-contractor. Tests were conducted at
approximately 25 m intervals. The tests were conducted in three corresponding
pondage sections; one of marginally higher seepage (Pond 4; 8.5 mm/d) than the other
two sections (Pond 2; 6.0 mm/d and Pond 5; 6.2 mm/d). In addition to the tests at set
intervals, in each pondage section two series of tests across the channel were
conducted (LHS, centre and RHS) in order to determine differences in seepage rates
across the channel profile.
The bed was saturated at the time of the tests and water was ponded along the centre
line of the channel over a considerable length of the channel. This made it impractical
to carry out tests along the centre line of the channel as originally designed. Therefore
modifications were made to the original methodology to suit the conditions, including:
q Adjustments to the standard distance of 25 m between sites in order to find a site
where the surface was free of water;
q Placement of the rings on the right hand side of the bed, just above the waterline
in the channel;
q The sites where the three tests were to be carried out across the bed were shifted to
locations where the whole bed was exposed. This was usually a point where the
silt layer was higher than elsewhere. As the width of the bed was too narrow to
place the rings in a line at right angles to the centre line, the rings were placed
diagonally at about 1.5 to 2 m spacing; and,

q Some tests could not be carried out as the bed was covered in water and these sites
were inaccessible with the equipment used.
The following observations were made regarding the channel soil profile during the
test. The batters of the channel are covered with a layer of brown sandy clay about 1
cm thick over the material used to make the banks. The lower parts of the batter are
into the parent material which is a heavy clay. The bed is not flat, but is more of a
shallow V shape which is covered by about 10 cm of silty material. There is a brown
surface crust, but below this the material is a loose grey silty clay. Underlying this
silty clay is the parent material, comprised of a grey heavy clay.
The basic test methodology consisted of placing the ring on the surface of the soil and
driving it in about 15 cm. (This was to reach firmer soil rather than it being located in
the upper uncompacted or disturbed A1 horizon). The ring was filled with water to a
depth of about 5 cm (so that the entire surface of the soil was covered) and
measurements of loss of water from the ring were made regularly. The readings were
taken at about 10 minute intervals initially until the soil profile had saturated and when
the final infiltration rate was reached after 30 to 45 minutes the readings were taken at
wider time intervals for about 2 hours. The amount of water needed to refill the ring
to its original level (mark on the side of the ring or a needle point) was recorded either
as the depth by which the water level has fallen since the last reading or as the volume
of water needed to bring the water level back to the level of the marker in the ring.
This method of maintaining the constant head varied slightly from previous ring
infiltrometer tests conducted during these trials, where a float device was used to
regulate a constant head. For low seepage rate soils such as in the Tabbita channel,
where the head only falls slightly over the test, this is a more accurate way of
determining the infiltration rate. To improve the accuracy of the results, rings were
left in place over night on two nights thus giving eight tests with longer duration
where the accuracy could be expected to be better because of the longer time for
seepage from the rings.
The results of the Tabbita ring infiltrometer tests are presented in Table 4.5. Figure 4-
4 displays the location of the tests. The results are presented both in terms of the
actual infiltration rate recorded during the test (ie volume/area) and the field saturated
hydraulic conductivity (Kfs) calculated from the infiltration rate (refer Section 3.3.3 for
discussion of the difference in the seepage rate and the hydraulic conductivity).
Significantly lower and more consistent Kfs values are reported compared to other ring
infiltrometer tests conducted in this study (Dahwilly and Toolondo). This is primarily
due to the more uniform and less permeable underlying parent material into which the
rings were driven. The average Kfs for each section is presented in two forms. Firstly
as the total average for all tests, and secondly as the average of all locations.
This second average is calculated using an average figure for the two sections in each
pondage where tests were done cross the channel, meaning that the three tests only
contribute one figure to the total average. This is a more accurate means of
determining the total average for the section. Using this method the average Kfs for
Ponds 2, 4 and 5 is 7.7, 4.0 and 4.5 mm/d respectively. The average for Pond 2 is
highly influenced by the one results of 23.8 mm/d. Perhaps a more representative
figure is the median Kfs for the ponds, which are 4.5, 4.6 and 4.2 mm/d for Ponds 2, 4
and 5 respectively. Essentially the figures indicate that there is no statistical

difference between the seepage rates in each pond as determined by the ring
infiltrometer tests.
The results do indicate a significant difference between the seepage rates along the
centre line of the channel compared to the sides of the channel base. Summing results
from all ponds, the average value (Kfs) for the centre line of the channel is 1.0 mm/d,
compared to 6.8 mm/d for the sides. The median value indicates just as significant a
difference of 0.3 mm/d for the centre and 4.2 mm/d for the channel sides.
n Table 4-5 Tabbita Channel Ring Infiltrometer and Disc Permeameter Results
Actual Field Saturated Average Length of Chainage and

Test No.
Infiltration Rate Hydraulic Kfs Test Position
(mm/day) Conductivity For Location (mins)
(Kfs, mm/d) (mm/d)
P2-T1 0 0.0 0.0 49 4913 m, RHS
P2-T2 8 3.8 3.8 62 4937 m, RHS
P2-T3 0 0.0 101 4967 m, RHS
P2-T4 0 0.0 4.5 91 4967 m, centre
P2-T5 25 13.6 143 4967 m, LHS
P2-T6 44 23.8 23.8 88 4991 m, RHS
P2-T7 24 12.9 78 5022 m, LHS
P2-T8 0 0.0 6.2 77 5022 m, centre
P2-T9 11 5.7 41 5022 m, RHS
Average 12.4 6.6 7.7
Lower Quartile 0.0 0.0 3.8
Median 8.0 3.8 4.5
P4-T1 2 0.8 0.8 1060 5313 m, RHS

P4-T2 6 3.0 3.0 1015 5338 m, RHS
P4-T3 22 11.6 985 5363 m, LHS
P4-T4 1 0.7 4.6 995 5363 m, centre
P4-T5 2 1.5 105 5363 m, RHS
P4-T6 0 0.0 0.0 74 5382 m, RHS
P4-T7 10 5.3 5.3 160 5463 m RHS
P4-T8 14 7.4 200 5487 m LHS
P4-T9 0 0.0 6.9 110 5487 m centre
P4-T10 25 13.4 185 5487 m RHS
Average 8.2 4.4 4.0
Median 4.0 2.2 4.6
P5-T1 34 18.5 18.5 155 5516 m, LHS
P5-T2 18 9.2 9.2 110 5538 m RHS
P5-T3 8 4.2 4.2 127 5558 m RHS
P5-T4 0 0.0 0.0 103 5584 m RHS
P5-T5 4 2.3 1003 5607 m, LHS
P5-T6 2 1.1 2.6 987 5607 m, centre
P5-T7 8 4.2 970 5607 m, RHS
P5-T8 6 3.3 958 5632 m, LHS
P5-T9 8 4.4 6.4 90 5632 m. centre
P5-T10 21 11.6 80 5632 m, RHS
Average 10.9 5.9 4.5
Median 8.0 4.2 4.2
These results confirm previous seepage investigations that channels are more prone to
seepage at the sides than in the centre. [eg Smith and Turner (1982) studied a number
of channels within the Goulburn Valley using the Idaho seepage meter and found that

on average seepage rates were between 5-20 times greater at the sides of the channel
base compared to the centre].
n Figure 4-17 Tabbita Channel Ring Infiltrometer Locations and Results

Finally, analysis was conducted to compare the point tests results to the pondage test
results for each section. Figure 4-17 presents these results, where the average, median
and lower quartile figures for each pondage section were plotted against the
corresponding pondage test seepage rate for the section. This figure confirms the
above conclusions that essentially there is no statistical difference between the
seepage rates in each pond as determined by the ring infiltrometer tests. The average
data shows a negative trend which is not possible and can be dismissed due to the one
high seepage result influencing the average in Pond 2. The median results indicated a
marginal trend, but is not statistically significant.
The main problem with assessing any seepage measurement technique at the Tabbita
site is that the spread of seepage rates as indicated in the pondage tests is very low,
and therefore due to potential small errors in both the pondage tests and the particular
measurement technique, it is difficult to conduct any assessment that is statistically
meaningful.
n Figure 4-18 Tabbita Pondage Test Seepage vs Point Source Tests
Tabbita Pondage Test Seepage Rates vs Ring Infiltrometer Kfs

10
Average
9
First Quartile
Ring Infiltrometer (Kfs, mm/d)
8
Median
7
6 y = -0.95x + 11.94
y = 0.07x + 3.94
5
3
y = -0.09x + 3.73
2
0
5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0
However it should be noted that this investigation does not rule out the use of ring
infiltrometers as a means of estimating channel seepage. Six tests is probably not
enough to adequately characterise a length of channel 200 m long, even given the
relatively uniform soil properties of the Tabbita. As was initially suspected, one of the
greatest barriers to reliably using point test data is the variability in the soil, and many
more tests are required to adequately characterise a given section using this method
(ANCID, 2000a). From an economic perspective this means that assessment of
channel seepage using ring infiltrometers is not likely to be a cost effective solution.
There was a much better correlation observed between the seepage values from the
ring infiltrometer and the pondage test results at this site compared to the Toolondo
channel. This is primarily due to the more uniform and less permeable underlying
parent material into which the rings were driven at the Tabbita channel. The clayey

nature of the soils assists with providing a good seal around the ring, however the key
difference is attributable to the surface layer at Toolondo which is not in fact the most
restricting layer. At the Tabbita site, the surface soils in the channel are essentially
uniform and form the seepage controlling layer, meaning that analysis using surface
based instruments is appropriate.
4.3.7 Idaho Seepage Meter – Donald Main Channel

Idaho seepage meter tests were conducted on the Donald Main channel in October
2001. The tests were conducted by Akbar from Yanco Agricultural Institute, with
assistance provided by Wimmera Mallee Water. Twenty-two locations were tested in
four pondage sections comprising P2, P3, P4 and P6. Figure 4-19 presents the test
locations and results. As this figure illustrates, four tests were conducted across the
wetted perimeter of the channel at right angles to the channel in each test location.
Within each of these four locations a number of tests were conducted to arrive at the
average rate for that location.
4.3.7.1 Comparison with Pondage Test Results

The Idaho Seepage meter results were compared with the pondage test seepage rates.
A representative Idaho seepage meter rate for each pondage section was calculated by
obtaining weighted seepage rates for each section. In addition to using all Idaho tests,
a weighted average using the two results from the middle of the channel only was
calculated. These results are presented in Figure 4-20.
The best result was obtained for Pond 2 which recorded almost the same weighted
Idaho seepage meter reading (35 mm/d) as pondage test result (35 mm/d). Pond 4 also
recorded similar results between techniques (33 mm/d for the Idaho meter compared
to 37 mm/d for the pondage test). Pond 6 and Pond 3 deviated most from this 1:1
linear trend. Using all of the data the results indicate a moderate to poor correlation
between the Idaho seepage meter results and the pondage test seepage rates (R2 =
0.33). Pond 3 (36 mm/d pondage test seepage) is the main cause of the poor
correlation, which recorded the same weighted Idaho seepage average as Pond 6 (9
mm/d 36 mm/d pondage test seepage).
Taking the weighted average from only the middle two Idaho seepage tests across the
wetted perimeter slightly improves Pond 3 with respect to representing the pondage
test seepage however, this further worsens Pond 6 and Pond 4, resulting in overall
lower correlation (R2 = 0.27).
Two main conclusions arise from this analysis:

q The Idaho seepage rates are comparable in magnitude to the pondage test seepage
rates;
q There is the suggestion of a linear trend between pondage test seepage and
weighted Idaho seepage rates at the Donald Main site, although the correlation is
moderate to poor. The limited number of pondage sections on which the trend is
based (4) and the limited number of Idaho tests within each pond (5-6) are the
main reason for not obtaining a better correlation. The lack of spread within the
pondage test data is also a limiting factor.

n Figure 4-19 Donald Main Channel Idaho Seepage Meter Test Locations
(January, 2001)

n Figure 4-20 Donald Main Channel Idaho Seepage Meter Results vs Pondage
Test Seepage
Idaho Seepage vs Pond Seepage
45
Average Idaho
40 Seepage
P4 P4
Average Idaho
35
Seepage of Central P3 P3
Tests P2
30
y = 0.9163x + 5.8254
25
R2 = 0.333
average of all tests y = 1.1907x - 1.9982
20 2
R = 0.2719
average of central tests
15
10
P6
5
0
0 5 10 15 20 25 30 35 40 45
Average Idaho Seepage Meter Result for Pond (mm/d)
4.3.7.2 Comparison with EM31 Results

An attempt was also made to correlate the Idaho seepage meter tests and 2002 the
EM31 survey data. The four to five EM31 values immediately surrounding an
individual Idaho test were averaged to obtain the representative apparent conductivity
for that location. The results of this analysis are presented in Figure 4-21.
This figure shows no correlation between the EM31 conductivity and the Idaho
seepage meter results. When the ‘outlying’ pond (Pond 3) is removed a weak inverse
correlation is observed (ie increasing seepage with decreasing EM31 conductivity).
The EM31 conductivity average clearly distinguished between seepage rates (as
measured with the Idaho meter) in pond 6 and pond 2, but no distinction at all between
pond 3 and 4 is observed. Further investigation would be required to determine the
reasons for this.
The limited nature of the point methodology of the assessment may also contribute to
variability in the results. Improvements could probably be made by averaging seepage
and conductivity readings over 20-30m lengths of channel, rather than lines across the
channel. This bulking process could assist in smoothing out local variations and
anomalies that may occur in isolated sections of channel.

n Figure 4-21 Donald Main Channel Idaho Seepage Meter Results vs EM31
(2002) Apparent Conductivity
Donald Main Channel Idaho Seepage Vs EM31
50
Average Seepage
Pond 6
y = -0.33x + 46.68
Idaho Seepage Meter Rate (mm/d)
40 Pond 2
Trend line R2 = 0.19 Pond 4
excluding Pond 3
Pond 3
30
20
y = -0.14x + 33.63
R2 = 0.04
10
Trend line for all data
0
10 20 30 40 50 60 70 80 90
Av Surrounding EM31 Conductivity (mS/m)
4.3.8 Conclusions
These trials have confirmed that point tests are generally not reliable for directly
quantifying seepage. Due to variable and sometimes erratic values obtained in
measurements, the trials have illustrated that a large number of tests is required to
sufficiently determine the true seepage rate of a section. Therefore they are generally
not considered reliable for absolute quantitative purposes and should generally be
limited to determining the distribution of seepage losses ( i.e., relative seepage). Even
for this use a large number of tests are recommended to minimise the effects of local
variability. These conclusions equate to the findings of the literature review.
In addition, it was apparent in a number of channels that the bed of the channel was
seeping at a different rate to the walls of the channel. This appeared to be occurring at
a number of the point test sites, as evidenced by higher seepage rates in the base of the
channel than the pondage test rates. This is in contrast to the normal phenomenon
with point tests where lower seepage rates than actual are often obtained (due to the
non-detection of ‘hot spots’). In these cases, even very high density point test
sampling in the bed of the channel can not determine the actual seepage rate.
In terms of choice of equipment for point testing:

q The Idaho seepage meter appeared to provide the most reliable results of the three
instruments. This concurs with the fact that the channel is full during the test and
that truly saturated flow is being measured. However there are very few
operators skilled in use of the equipment and therefore testing is limited by their
availability. The tests are also very expense, due to the fact two operators are
required, including one skilled in use of the meter.

q Definitive comments cannot be made regarding the accuracy of the disc
permeameter compared to the ring infiltrometer. Some trouble was encountered
however with the ring infiltrometer in terms of seepage outside of the ring. The
disc permeameter is simpler to use than the ring infiltrometer, both in terms of
operation and manual handling.

4.4 Groundwater Techniques
4.4.1 Introduction
Groundwater observation bores are often required as part of a channel seepage
investigation (for geological and hydrogeological characterisation etc). However
groundwater bores can also be used to quantify seepage rates based on groundwater
level fluctuations. Quantitative assessment of seepage rates was conducted at the
Donald Main Channel and a qualitative assessment was conducted for the Tabbita
channel.
4.4.2 Methodology
Observation of groundwater levels in a series of piezometers located at right angles to
the centre line of a channel provides data to determine the flow lines and equipotential
lines of seepage water. The amount of seepage can be estimated by studying the
variations in the watertable combined with variations in channel running level. The
best period of observation is during the rise in watertable when a channel is put back
into operation or during the fall of the watertable at the end of the channel run season.
This approach requires a minimum of two groundwater observation bores at right
angles on either side of the channel. An estimate of aquifer hydraulic conductivity is
also required. This can be estimated based on the textural properties of the material
identified during the drilling of the bores, but is preferably obtained by aquifer
pumping tests or slug tests.
4.4.3 Donald Main Channel

4.4.3.1 Monitoring Bore Set-Up
Four transects of groundwater bores have been installed at the Donald Main Channel,
and fortnightly to monthly monitoring of water levels has been ongoing since the
installation of the bores. The location of the bores is presented in Figure 4-22.
Transects A and B existed prior to this ANCID study. One additional up-gradient
bore was added to Transect A and B and the entire transect C and D were installed as
part of this investigation. This drilling was completed from 1st-3rd August 2000.
Table 4-6 summarises the bore construction and survey elevation data for the
groundwater bores at the Donald site. As labelled in this table, Bores SC3 & SC4 and
SC5 & SC6 are nested sites (ie a deep and shallow bore in the same location).
4.4.3.2 Groundwater Level Response

Figure 4-23 and Figure 4-24 present the groundwater hydrographs for Transects A
and B, and C and D respectively. Transects A and B have a longer record as they
were monitored before the commencement of the project. Plotted on each of these
graphs is the gauge level of the channel. Where the gauge line is flat, the channel was
not in operation. The groundwater response to the operation of the channel is clearly
seen in these graphs. The near channel bores rise 1-2m in a matter of weeks after the
filling of the channel (rate dependent on distance from the channel). The hydrograph
responses for bores greater than 50m from the channel generally display 0.5-1m rises
in groundwater level, suggesting some impact. Although in some cases and at
particular times of year there could be an additional variation in water level caused by
regional recharge to the aquifer, this is not evident from the hydrographs.

n Table 4-6 Donald Main Channel Groundwater Bore Construction Data
Total Top of Natural Screen Screen
Transect Bore Depth Casing Surface From To Dominant
(Nested) ID (m) (m, AHD) (m, AHD) (m) (m) Lithology
A SA0 8.5 118.17 117.58 6 8.5 Sandy Clay

SA1 6.5 117.71 116.81 3.9 5.9 Sandy Clay
SA2 5.7 116.07 115.11 3.0 5.0 Sandy Clay

SA3 4.0 114.64 113.62 1.3 3.3 Clay
SA4 4? 113.43 111.99 2? 4? Sandy Clay?
SA5 4.8 111.26 110.25 2.2 4.2 Sandy Clay
SA 6 5.0 110.87 109.89 2.4 4.4 Sand
B SB0 11 121.05 120.60 7.0 11.0 Clay

SB1 6.5 118.18 117.16 3.9 5.9 Sand
SB2 5.0 115.54 114.35 2.4 4.4 Sand
SB3 5.5 114.51 113.47 3.0 5.0 Sandy Clay
SB4 5.0 114.46 113.40 2.0 4.0 Sandy Clay

SB5 5.0 114.15 113.13 2.4 4.4 Sandy Clay
SB6 6.0 113.17 112.16 3.4 5.4 Sandy Clay
SB7 5.0 111.66 110.60 3.9 5.9 Clay
C SC0 11 119.56 119.08 8.5 11 Clay

SC1 8.5 118.14 117.75 6 8.5 Clayey Sand
SC2 7 117.32 116.96 4.5 7 Clayey Sand
SC3 3.7 116.93 116.45 1.2 3.7 Sandy Clay
Nested
SC4 8.5 116.83 116.45 6 8.5 Clayey Sand
SC5 2.5 115.24 114.97 0.5 2.5 Clayey Sand
Nested
SC6 7 115.42 114.98 4.5 7 Sandy Clay
SC7 7 114.69 114.22 4.5 7 Sandy Clay
SC8 7 113.24 112.73 4.5 7 Sandy Clay
SC9 7 113.17 112.69 4.5 7 Clay
D SD0 10 120.12 119.68 7.5 10 Sandy Clay

SD1 7 117.37 116.91 4.5 7 Sand
SD2 7 115.36 114.94 4.5 7 Sandy Clay
SD3 7 114.43 114.12 4.5 7 Sandy Clay
SD4 7 112.94 112.49 4.5 7 Sandy Clay

n Figure 4-22 Donald Main Channel Bore Locations

n Figure 4-23 Donald Main Channel Groundwater Bore Hydrographs,
Transects A & B
Donald Main Channel Groundwater Hydrographs - Transect A

116
Gauge
SA 1-new
115 SA 2-new
SA 3-new
114 SA 4-old
SA 5-new
113 SA 6-new
SA 0-new
Water Level - AHD
112
111
110
109
108
107
1-11-98 9-02-99 20-05-99 28-08-99 6-12-99 15-03-00 23-06-00 1-10-00 9-01-01 19-04-01 28-07-01
Date
Donald Main Channel Groundwater Hydrographs - Transect B

116
Gauge
Bore SB1
115 Bore SB2
Bore SB3
Bore SB4
114 Bore SB5
Bore SB6
113 Bore SB7
Water Level - AHD
112
111
110
109
108
1-11-98 9-02-99 20-05-99 28-08-99 6-12-99 15-03-00 23-06-00 1-10-00 9-01-01 19-04-01 28-07-01
Date

n Figure 4-24 Donald Main Channel Groundwater Bore Hydrographs,
Transects C & D
Donald Main Channel Groundwater Hydrographs - Transect C

116
Gauge
SC 0
SC 1
115 SC 2
SC 7
SC 8
SC 9
114
SC3
SC4
SC5
113 SC6
RWL (m)
112
111
110
109
108
23/06/2000 12/08/2000 1/10/2000 20/11/2000 9/01/2001 28/02/2001 19/04/2001 8/06/2001 28/07/2001
Date
Donald Main Channel Groundwater Hydrographs - Transect D

115
Gauge
SD 0
SD 1
114
SD 2
SD 3
SD 4
113
112
RWL (m)
111
110
109
108
23/06/2000 12/08/2000 1/10/2000 20/11/2000 9/01/2001 28/02/2001 19/04/2001 8/06/2001 28/07/2001
Date
4.4.3.3 Seepage Estimates

Groundwater levels at the Donald Main Channel were used to estimate seepage
beneath borelines C and D. The seepage rates at Donald involved calculating
groundwater flow using the Dupuit Forcheimer equation for steady flow in an
unconfined aquifer. The solution for one side of the flow from the channel is given
by:
q = - 0.5 K (h22 – h12)

L
Where h2 is the head in the channel and h1 is the head in the observations well.

The hydraulic gradient was calculated using the head in the channel and the most
distant bores from the channel at times of channel operation and non-operation for
Borelines C and D. These are shown in Figure 4-25 and Figure 4-26.
n Figure 4-25 Groundwater levels across Boreline C
Sheridan Site C
Channel
120
118
116
RWL (m)
Natural Surface
High groundwater 114

level
112
Low groundwater
level 110
108
-100 -50 0 50 100 150
Distance (m)
n Figure 4-26 Groundwater Levels Across Boreline D
Sheridan Site D
122
Channel
120
118
116
RWL (m)
High groundwater Natural Surface

level
114
112
Low groundwater 110

level
108
-150 -100 -50 0 50 100 150
Distance
The Dupuit Forcheimer equation was applied to a range of groundwater conditions

where groundwater levels vary. The hydraulic conductivity (K) was assumed to range
between 0.2 and 1 m/day which is typical of a clayey sand/sandy clay aquifer. The
results are summarised in Table 4-7 below. The calculation assumes that the base of
the flow system is 10m below the “normal” groundwater level.

n Table 4-7 Summary of Donald Main Channel Seepage Estimates from
Groundwater Data
Conditions
Transect C
Hydraulic Conductivity 0.2 m/day 0.5 m/day 1m/day
Elevation of Datum (base of Aquifer) 100 100 100
Head in the Channel (h2) 114.5 114.5 114.5
Head in the Observation Bore (h1) 111 111 111
Distance between edge of channel and obs bore 70 70 70
Discharge (m3/day) 0.1275 0.31875 0.6375
Estimated seepage into aquifer (mm/day per m length) 42.5 106.25 212.5
for half of 6m wetted perimeter
Conditions
Transect d
Hydraulic Conductivity 0.2 m/day 0.5 m/day 1m/day
Elevation of Datum (base of Aquifer) 100 100 100
Head in the Channel (h2) 114.5 114.5 114.5
Head in the Observation Bore (h1) 110 110 110
Distance between edge of channel and obs bore 100 100 100
Discharge (m3/day) 0.11025 0.27563 0.55125
Estimated seepage into aquifer (mm/day per m length) 36.75 91.875 183.75
for half of 6m wetted perimeter
The seepage estimates (42.5mm/d Transect C and 36.75 mm/day for Transect D)
based on analytical groundwater flow calculations results for the assumed hydraulic
conductivity of 0.2m/d are consistent with the pondage test results (45mm/day) for
Pond1 where these transects are located.
The estimated rates using higher hydraulic conductivity values are clearly much
higher than the observed results. Further investigation (including slug tests to
determine actual hydraulic conductivities) are required to provide greater confidence
in the estimates based on piezometric levels alone.

4.4.4 Tabbita
Two groundwater monitoring bore transects were installed at the Tabbita trial site in
August 2000. The bore locations are shown in Figure 4-27, with bores 1-8 on
Transect 1 and bores 9-13 on Transect 2. Drilling information is detailed in Table 4-8.
n Table 4-8 Tabbita Groundwater Bore Construction Details

Nested Bore Depth TOC Depth Screen Screen Dominant Hydraulic
Sites ID (m) (Bore below From To Lithology Conductivity
Stick Up) surface (m) (m) (m/d)
m
Bore 1 9.47 0.53 8.94 8.5 6.5 Clay 0.00055
1&2
nested
2 4.31 0.56 3.75 3.25 2.25 sandy clay -
Bore 3 9.38 0.58 8.8 8.3 6.3 Clay 0.058
3&4
nested
4 6.73 0.55 6.18 5.6 4.6 Clay -
Bore 5 9.86 0.48 9.38 8.8 6.8 Clay 0.26
5&6
nested
6 4.25 0.48 3.77 3.2 2.2 sandy clay -
7 9.27 0.62 8.65 8.1 6.1 Clay 0.00018
8 6.95 0.67 6.28 5.8 3.8 sandy clay 0.00013
Bore 9 9.03 0.75 8.28 7.8 5.8 Clay 0.0011
9&10
nested
10 4.97 0.58 4.39 3.9 2.9 Clay -
11 7.14 0.36 6.78 6.3 4.3 Clay 0.5
12 9.02 0.3 8.72 8.2 6.2 Clay -

13 8.53 0.8 7.73 7.2 5.2 sandy clay -
Water level monitoring commenced on these bores from 27th September 2000 and is
currently ongoing. Hydrographs have been prepared for each transect up to 7th August
2001. Figure 4-28 and Figure 4-29 displays the hydrographs for transects 1 and 2
respectively.
4.4.4.1 Transect 1
Essentially 3 types of response are observed in the hydrograph behaviour in
Transect 1:
iv) Bores adjacent the channel ( nested site 1 & 2 and nested site 5 & 6);
v) Bores 10-15m away from the channel (nested site 3 & 4 and bore 7); and,
vi) Bore 8 which is well away from the channel.
These three categories of response are briefly discussed below:
The hydrographs of the adjacent channel bores in Transect 1 (bores 1& 2 and bores 5
& 6) indicate relatively constant water levels between October 2000 and June 2001,
with relatively minor fluctuations only observed in this period. This corresponds with
the fact that the channel was running throughout this time, largely masking any other
influences on hydrograph groundwater behaviour. The channel ceased operation in

early June 2000, however water was held upstream and released for the pondage tests
which finished on 26th June 2001. The rapid decline in groundwater levels can be seen
in the bores adjacent the channel following this period (approximately 0.7m between
monitoring periods). A delayed response was observed in the bores on the west side
of the channel (nested bores 5 & 6) compared to the response observed on the east side
(nested bores 1 & 2).
Bores which are a moderate distance (10-15m) away from the channel (nested site
3&4 and bore 7) display greater fluctuation in water levels during the period of
channel operation. The gradual fall of these hydrographs from the start of monitoring
until end of February and then the gradual rise again is probably due to annual
seasonal variation in the groundwater levels. These bores also drop appreciably
following the shut down of the channel, but generally only by approximately 0.4m.
The decreased effect obviously due to the increased distance from the channel. Again
there is a delayed effect in the response to the removal of water in the channel in the
bores on the west side of the channel (bore 7) compared to the east side (nested bores
3 & 4). The delay is more significant in this instance, with water levels continuing to
rise for approximately 1 month (check) after channel shut down.
Bore 8 displayed a continually declining water level throughout the period of

monitoring. Given its distance from the channel there is insufficient period of
monitoring at this stage to assess whether there is any channel seepage related
response in the water levels. The inconsistent response between bore 8 and the
intermediate distance bores, may an indicate that either bore 8 is anomalous and the
results should be treated with caution, or that other factors are effecting the response
of the bores closer to the channel (other than the channel) such as irrigation of the
adjacent vines, and causing the variation observed in the groundwater levels. Further
information of activities around the site which might influence local groundwater
conditions and a greater period of monitoring are required to assess these influences.

n Figure 4-27 Tabbita Groundwater Monitoring Bore Locations

n Figure 4-28 Tabbita Groundwater Bore Hydrographs – Transect 1
Tabbita Canal Groundwater Monitroing Results
114.5
114
113.5
channel partially empty pondage tests

113
WL AHD (m)
channel running
112.5 channel empty
Bore 1 D
Bore 2 S
112
Bore 3 D
Bore 4 S
Bore 5 D
111.5
Bore 6 S
Bore 7 D
Bore 8 D
111
01/09/00 31/10/00 30/12/00 28/02/01 29/04/01 29/06/01 28/08/01
Date
n Figure 4-29 Tabbita Groundwater Bore Hydrographs – Transect 2

Tabbita Canal Groundwater Monitoring Results
114.5
Bore 9 D
Bore 10 S
Bore 11 D
114
Bore 12 D
Bore 13 D
113.5
Channel Partially Empty

113
WL AND (m)
Channel Running
112.5
Pondage Test Channel Empty
112
111.5
111
12/08/00 01/10/00 20/11/00 09/01/01 28/02/01 19/04/01 08/06/01 28/07/01 16/09/01
Date

4.4.4.2 Transect 2
Essentially 2 types of response are observed in the hydrograph behaviour in
Transect 2:
1) Bores 10-15m away from the channel (bores 9, 10 and 13); and,
2) Bores 11&12 which is well away from the channel.
These categories of response are briefly discussed below.
The hydrographs from the bores 10-15m from the channel Transect 2 (bores 9, 10 &
13) show that bore 13 had relatively constant water levels between October 2000 and
June 2001, with relatively minor fluctuations only observed in this period. This
corresponds with the fact that the channel was running throughout this time, largely
masking any other influences on groundwater behaviour.
Bores 9 and 10, a similar distance to the channel as bore 13 show a greater
fluctuations in groundwater levels during the monitoring period with a gradual fall in
groundwater levels from the start of monitoring until end of February and then the
gradual rise again. This may be attributed to annual seasonal variation in the levels
with groundwater levels from these bores following similar trends to that of bores 11
and 12.
The channel ceased operation (pondage test completed) on 26th June 2001, and
groundwater levels in the bores adjacent the channel, bores 9, 10 and 13, dropped
sharply, by approximately 0.3m between 25th June 2001 and 7th August 2001.
Bores 11&12 are a greater distance from the channel and appear to be influenced more
from regional groundwater trends rather than from the channel.
4.4.4.3 Summary
In summary, both transects 1 and 2 indicate significantly declining waterlevels in the
bores close to the channel during the channel closure period (June and August 2001).
This is almost certainly attributable to the removal of the groundwater recharge source
of the seeping channel. Bores at an intermediate distance (10-15m) from the channel
displayed a subdued response to seepage and bores at a significant distance from the
channel (>50 m) showed little response that may be attributed to channel seepage over
the period of monitoring available.
The cross sections displayed in Figure 4-30 and Figure 4-31 display high and low
water levels in the two Tabbita bore transects.

n Figure 4-30 Tabbita Channel, Bore Transect 1: Groundwater Level Cross-
Section
Tabbita Seepage 1-8
115.5
115
114.5
114
HighWL
113.5
LowWL
113
112.5 RL
112
111.5
111
-100 -80 -60 -40 -20 0 20 40 60 80 100
Distanace
n Figure 4-31 Tabbita Channel, Bore Transect 2: Groundwater Level Cross-

Section
Tabbita Seepage 9-13
115.5
115
114.5
114 High W L
RWL (m)
113.5 Low W L
113 RL
112.5
112
111.5
111
-120 -100 -80 -60 -40 -20 0 20
Distanace
4.4.5 Conclusions
The trials conducted in this investigation suggest that use of groundwater bores for
quantitative analysis of seepage rates is not considered an accurate or cost effective
means of determining seepage rates for typical Rural Water Authority investigations.
In order of increasing importance the method is not considered accurate due to:

• Sensitivity to hydraulic conductivity inputs (eg Tabbita – depending on input
hydraulic conductivity, seepage rates of varying orders of magnitude can be
obtained) and the cost of obtaining sufficiently reliable estimates.
• Relies on assumptions regarding pre-channel groundwater water levels. These can
be estimated from conditions before and after channel filling, but depending on
site hydrogeology this assumption may or may not be accurate.
• It is essentially a type of point test and does not answer the question of what area
of the channel is seeping. A high density of bore transects would be required for
meaningful identification of local areas of channel seepage.
• Relies on an assumption of aquifer thickness (this may be able to be calculated but
in a deep aquifer this may be very expensive).
4.5 Remote Sensing

A remote sensing trial was planned for the Wimmera, including the Rocklands,
Toolondo and Donald Main Channels as part of the Stage 1 project. However the trial
was not undertaken as it could not be accommodated into the RWA budget.
However, the project brief for this trial is contained in Appendix D. The brief contains
information regarding the type of issues that require consideration when developing a
methodology for using remote sensing to identify channel seepage.
Prior and separate to the ANCID Stage 1 project, Wimmera Mallee Water arranged for
the collection of airborne video multispectral imagery along the Rocklands and
Toolondo channels (conducted by Charles Stuart University, CSU). This data was
evaluated by Sinclair Knight Merz (with a view to potentially incorporating the results
into this project) but unfortunately the data was not in a form suitable for use in the
project. The reasons for the unsuitability of this data are briefly documented below:
q Imagery was collected along the Rocklands and Toolondo channels and included
data in the red, green, blue and near infrared bands.
q The image pixel size was 2m and the registration accuracy was found to be
variable compared to the WMW geographic information system (GIS) channel
data. The ground control data provided by WMW consisted of channel outlines
and registration accuracy in the channel areas was considered to be 10 – 15 m by
CSU.
q There were some concerns with registration of the base GIS data and the pixel
size of the imagery.
q The quality of the images was variable for different bands. This has been
attributed to the time of year of the imagery (June) which equates to reduced solar
irradiation and reflectance, and possibly a light haze above the Toolondo Channel
which would affect the shorter wavelengths (bands 1 – 3 rather than band 4,
which is longer) as observed.
q The mosaics of the Toolondo Channel were found to be co-registered along the
channel within the area of interest, however, the colour matching could be
improved. The recent mosaics, displayed in infrared mode appeared to show
some correlation with the results of the EM data and may have some potential for
seepage detection.
q The re-sampled 8 m mosaics appeared to be acceptable at the upper limit for
viewing the channel, its immediate surrounds and location/presence of trees.

The following conclusions were made regarding the available data:
q Further investigation of the registration of one of the bands was required.
q The mosaics need to be improved, particularly with respect to colour matching,
co-registration, gaps and duplications before they can be included in the GIS.
q There may be some potential of the infra-red data to contribute to the detection of
channel seepage, however, this requires further detailed GIS analysis with other
data sets to confirm.
q The use of the imagery for detecting vegetation affects of channel seepage is
limited due to the timing of the image acquisition as the optimum time would be
during drier periods (ie. late Spring to Autumn).
4.5.1 Summary
Remote sensing trials were not conducted in the Stage 1 investigation. However
conclusions regarding remote sensing techniques (are made below) based on
knowledge gained through the literature review and preparation of a brief for proposed
trials:
q Remote sensing techniques offer considerable potential for rapid identification of
seepage zones (but not quantification) of large lengths of a channel system.
Remote sensing should primarily be regarded as a seepage identification tool and
not a seepage quantification tool. Quantification using remote sensing was to be
trialed in this study, however there are currently no documented studies of remote
sensing being used to quantify channel seepage.
q The techniques are best suited to investigation where the primary aim is
identification of land degradation associated with channel seepage. Remote
sensing techniques rely on the detection of differences in soil / moisture
properties in the upper surface. Therefore it has significant potential if there are
known surface effects of channel seepage. Conversely, they should not be used if
it is known that the seepage mechanism is predominantly vertical, such as is
likely to occur at sites with a deep watertable.
q Remote sensing will be most useful in environments where lateral seepage is
predominant. For example sites with a high watertable, shallow impermeable
layer or bank seepage - these environments represent conditions most likely to
facilitate lateral seepage and cause the seepage to have a surface expression.
q It offers a promising means of providing a first-cut identification tool for targeting
potential seepage sites, although a drawback is that it assumes seepage will have
a surface expression as moist soil or associated vegetation adjacent the channel.
q For this method to be cost effective, it needs to be conducted at a suitably large
scale. Costs are likely to come down and resolution likely to improve as the
technology develops, and will therefore become an increasingly attractive option.
q If remote sensing is proposed it should be well thought out and planned , given the
high cost of the resources involved.

5 Geophysical Techniques
5.1 Introduction
Geophysical techniques were identified in the literature review (ANCID, 2000a) as
having potential for channel seepage identification and quantification. In addition,
geophysical techniques fit well with the type of technique RWAs require for channel
seepage assessment. The national RWA survey indicated that RWAs considered cost
and speed to be the most important criteria in selecting a channel seepage
measurement technique, with accuracy of lower importance (ANCID, 2000b). This
suggests that RWAs are looking for a relatively cheap technique that can provide a
reasonable estimate of seepage rates, with some margin for error in estimates
considered an acceptable trade-off for improvements in cost and speed. The project
therefore incorporated trials of geophysics in its program.
5.1.1 Theory of using geophysics to identify and quantify seepage

Seepage of water from channels causes local changes in the physical properties of the
areas surrounding the channels. These include an increase in saturation, decrease in
groundwater salinity, decrease in stored salts and a rise in the watertable. These
changes all have an effect on the terrain electrical conductivity (or its inverse,
resistivity) and therefore can be detected using electrical geophysical techniques,
which measure resistivity or conductivity.
The overall conductivity / resistivity response is dependent on both soil lithology and
the salt content of any contained water. In general it can be assumed that the clays
will be more conductive due to their chemical structure. These properties are in
contrast to the sands, which generally have a lower conductivity. When both
parameters (lithology and groundwater salinity) are varying, interpretation can be
difficult. However, in the case of channel seepage, higher permeability soils and low
salinity water in areas of high channel seepage will enhance each other to produce a
low conductivity / high resistivity response. Therefore conductivity can be used to
map areas of high permeability soils and low salinity water emanating from the
channel.
The two techniques used in the trials were electromagnetics (Geonics) and resistivity,
based on techniques identified as most likely to be successful in the literature review
(ANCID, 2000a). These each provide different depth sub-divisions. EM31 and EM34
essentially average the conductivity over a depth (to provide one number
representative across that depth) where as the resistivity technique used was multi-
channel, which provides a depth distinction (various numbers for different depth
intervals are provided).
5.1.2 Methodology
The basic methodology used to assess the accuracy of geophysical techniques was
comparison with pondage test seepage rates. Pondage test seepage was compared to
the average geophysical response (conductivity or resistivity) along the length of the
pond. Usually pondage tests were conducted back to back, to minimise the number of
banks required and hence reduce costs. Comparison of the average geophysical
response and pondage test seepage involved plotting the two variables against each
other, and analysis of the trends.
WC01312:YR_3_RPT_CH_5(GEOPH_TECH).DOC Final PAGE 66

As the average geophysical response for the section was selected, it was important that
the pondage cell covered a length of similar geophysical response. To achieve this the
geophysical survey was conducted prior to the pondage tests, and the pondage test
bank locations selected based on the results of the geophysical survey (in most cases).
Pondage lengths generally varied between 100m - 300m in length.
While the average geophysical response across the pond was used, and is considered
suitable for this study due to the method of selecting a length of like response, there is
potential for using geo-statistics to improve the effects of spatial variability on
geophysical response.
5.1.2.1 Years 1 and 2 Trials

The first two years of geophysical trials were conducted using Geonics electomagnetic
systems (EM31 and EM34). EM31 systems ‘see’ to about 6-7m depth, while EM34
systems ‘see’ between 7m – 60m, depending on the coil spacing and dipole orientation
used. The following variables were tested during the first two years of the trials (with
respect to their impact on the accuracy of the technique compared to pondage test
seepage rates):
q The location of the survey (eg adjacent the channel versus away from the
channel, or down-gradient side of the channel versus up-gradient side of
channel);
q On-land versus on-channel (EM31 only);
q Dipole orientation of EM31 and coil spacing of EM34 (which both effect the
depth focus of the instrument); and,
q Repeatability of surveys.
In general, the years 1 and 2 trials showed good correlations between pondage test
seepage and the average EM conductivity for the pond (statistical evaluation showed
EM conductivity values correlated against pondage seepage rates to give correlation
coefficients as high as 0.9 for sections of channel). In addition to this primary finding,
other key findings from the years 1 and 2 trials were:
q Mechanism - The application of the EM technique appears to primarily depend
upon a significant contrast in terrain conductivity being attained due to fresh
seeped water invading the more saline formations and particularly changing the
salinity of the groundwater. Geological effects appear to be of secondary
importance. Therefore it was concluded that a key part of conducting an EM
survey for detecting channel seepage is the penetration depth of the survey. The
survey should concentrate on the zone immediately above and several metres
below the natural watertable, in order to detect the displacement of the natural
groundwater with fresher channel water. (As discussed in later in this section,
‘detection’ above the watertable can also be used to infer where seepage is most
likely to occur).
q Channel operation – The above hypothesis regarding the primary mechanism by
which the geophysical survey ‘predicts’ seepage indicates that it is important that
the EM surveys should be carried out when the channel has been in operation for
a period long enough to raise the local water table and for seepage to penetrate
the surrounding formations. This may be a relatively short period in sandy
formations and longer in clays.

q Survey location - Best results were obtained close to the channel bank on the
down hydraulic gradient side. Further away from the channel the salinity of the
groundwater grades to the background groundwater quality and seepage becomes
more difficult to identify.
q On-channel surveys - (EM31 in both vertical and horizontal dipole) – Results of
variable success were returned. Evidence indicates that these on-channel (boat)
surveys use a different method of seepage detection to the land surveys which
primarily rely on the detection of the displacement of the natural groundwater
with fresher seeped water. (eg in the horizontal dipole the EM31 is essentially
monitoring the presence, thickness, and clay content of the retarding layer
immediately below the channel. In the vertical dipole mode the success seems to
be dependent on the degree to which the survey ‘sees’ below the fresh flushed
zone created uniformly beneath the channel.) In summary, the land based survey
appeared to be a more accurate means of assessing channel seepage because it
relies on the direct detection of the seepage plume, and is largely unaffected by
the seepage mechanism (ie through the walls or base of channel), and is not
influenced by the flushed zone immediately beneath the channel.
Trials conducted in the first two years of investigation indicated the potential of
geophysical techniques to rapidly identify and quantify sections of channel seepage
were superior (both technically and economically) to other techniques assessed in the
trials. Therefore geophysical techniques became the primary focus of the year three
trials, the methodology for which is described below.
5.1.2.2 Year 3 Trials

Based on the successful first two years of trials and the RWA needs identified in the
national survey, the focus of the year three trials was on further development of
geophysical techniques. The key conclusions drawn from the year two trials required
further testing to asses the effectiveness of geophysical techniques to quantify channel
seepage. The two key objectives of the third year program were to:
1. Identify transferability of correlations between seepage rates and conductivity at

different locations.
At most sites in the first two years of trials a reasonable relationship between
pondage test seepage and average EM(31 & 34) conductivity was obtained. For
such a technique to be useful at a broad scale, the applicability of these
relationships outside of the pondage test area must be known, ie is the
relationship transferable beyond the immediate test area? This involved
identifying the key factors which determine the validity of extrapolating
geophysical results to predict seepage and the development of general rules for
extrapolating results from pondage tests.
2. Trial a system which simultaneously measures conductivity (or resistivity) at a

range of depths.
Investigations in the first two year of trials using (single channel) electromagnetic
techniques (EM31 and EM34), both inside the channels and along the outside toe
of the banks, gave results that indicated that these geophysical techniques were
superior (technically and economically) to other techniques assessed in the trials.
The EM techniques:

q Indicated areas of low conductivity that were related to probable seepage;
q Correlated with pondage test seepage at most sites; and
q Were of high resolution and located narrow areas of possible seepage.
Depending upon local and temporal conditions of watertable depth, different EM

systems appeared to perform better at different sites and at different times. An
important conclusion was that the EM surveys appeared to work best where the
conductivity measurement was focused at the depth where addition of seepage
water was most significant in diluting groundwater salinity. The limitation of
these single channel electromagnetic systems is that the instruments measure a
single parameter equivalent to a single depth. Therefore investigation of a system
that simultaneously measures at a range of depths was recommended to ensure
penetration to the depth where seepage water and groundwater salinity contrasts
are greatest. The advantage of this approach would be that it could be applied
independent of knowledge of groundwater depths and salinities. Investigation of
multi-channel systems that measure at a range of depths to around 15m below
surface was recommended.
The following methodology was adopted for the third year trials in order to meet these
objectives:
Objective 1 - Identifying the transferability of correlations established at one site to

different locations, was addressed by undertaking geophysical trials at sites with a
range of different site conditions which might affect the geophysical survey response.
To this end the following tasks were undertaken:
a) Mapping of key factors effecting geophysical survey response – At one channel

within each RWA (where pondage tests had been undertaken) key factors likely to
affect conductivity / resistivity response along 10 - 20 kilometres of channel were
collated. These factors included:
• Soils and Geology; and,
• Groundwater Depth and Quality.
Field work was not undertaken in development of the maps, but the finest scale
and best quality data available was used.
b) Selection of new trial sites – Using these maps two new trial areas were selected
along each channel. One area was selected where conditions were similar to the
original trial area and the second where conditions differed from the original trial
area
c) At new and original sites undertake geophysical surveys and pondage tests:
• An EM31 survey was undertaken at the two new sites and the original site;
• A new geophysical technique (resistivity) was also trialed at the two new
sites and the original site; and,
• Pondage tests were conducted at each of the sites to determine the accuracy
of the geophysical surveys. They were relatively short pond lengths and the
bank locations were based on the geophysical survey results (covering areas
of like conductivity).
d) Drilling of soil bores –Soil bores were drilled to assist in interpretation of the
geophysical surveys.

e) Additional analysis of data using a statistical approach – A more rigorous
statistical evaluation of the data was conducted in the third year of the trials,
including conducting a multi-variate analysis to determine potential impacts of
other variables on the geophysical response and pondage test relationship.
Objective 2 - To meet the second objective of trialing a system which simultaneously

measures conductivity (or resistivity) at a range of depths, a multi-channel system
operating from within the channel was deemed to be the most appropriate set-up. This
would allow seepage zones to be tracked from the channel base into the groundwater.
Various systems and techniques to achieve this simultaneous range of depth
penetrations were explored and the most appropriate technique was considered to be a
resistivity array towed behind a boat. This system was chosen for the trials reported
here. The system was custom designed for the project by Zonge Engineering and
Research Organisation, in consultation with a Sinclair Knight Merz senior
geophysicist. A paper describing the application of a similar system for canal seepage
assessment in the US was recently published (Hotchkiss et al, 2001).
5.1.3 Description of Trial Sites

Figure 5.1 presents the location of all geophysical trials used in the final year’s
analysis. The trials were conducted within the areas controlled by Wimmera Mallee
Water, Murray Irrigation and Murrumbidgee Irrigation. Goulburn Murray Water were
also able to contribute useful data to the project, due to seepage investigation work
associated with the upgrade of the Waranga Western Channel. Data inputs from this
project included an EM31 survey along a significant length of the channel. The data
was in a suitable form to incorporate into this study. In addition to the Waranga
Western Channel, the Rocklands and Donald Main Channel in the Wimmera were also
used in the EM31 analysis, using data collected in year 1 and 2 of the trials. At
remaining sites both EM31 and resistivity survey data were collected.

n Figure 5.1 Year Three Geophysical Trials Regional Location Map

Table 5-1 Year Three Geophysical Channel Seepage Measurement Trial Sites
Background Groundwater No. pondage Geophysical

Rural Water Channel
Summary of Lithology Salinity (µS/cm, EC) / tests / Techniques
Authority And Site
Depth to Groundwater (m) seepage Trialed
range (mm/d)
Wimmera Toolondo – Predominantly clay to sandy clay q ≈ 5,000 EC 6 cells Resistivity &
Mallee Water Central overlying sandstone @ 3 – 9m q ≈9m EM31
depth. Fine to medium grained 1 – 12 mm/d
(original site)
sand 1-2m deep located
sporadically across site
Toolondo – Surface layer (0.5 - 1m) of q ≈ 5,400 EC 4 cells Resistivity &
West weakly cemented fine sand q ≈9m EM31
overlying 1-2m of sandy clay, 1 – 5 mm/d
overlying medium clay to 7m,
with underlying sandstone
Toolondo – Heavy – medium clay to 3-4m, q ≈ 4,200 EC 4 cells Resistivity &
East overlying shallow sandstone q ≈9m EM31
0 – 1 mm/d
Murrumbidgee Lake View – Sandy clay – sand clay loam q ≈ 6,700 EC 4 cells Resistivity &
Irrigation Central overlying medium to heavy clay q ≈ 1.5 m EM31
starting from 2-6m below surface 5 – 7 mm/d
(original site)
Lake View – Surface layer (approx 1m) of q ≈ 5,400 EC 4 cells Resistivity &
West sandy clay overlying 3-4m of q ≈ 0.5 m EM31
gravelly clay, overlying medium 20 – 25 mm/d
clay
Murray Dahwilly – Medium to coarse grained sand q ≈ 5000 EC 7 cells Resistivity &
Irrigation Central to at least 10m depth. Sandy q ≈5m EM31
clay to clay loam in top metre 1 – 10 mm/d
(original site)
along most of channel length
Dahwilly – Medium to coarse grained sand q ≈ 2,500 EC 3 cells Resistivity &
East to approx 6-8m depth, underlain q ≈5m EM31
by clayey sand. Sandy clay to 9 – 10 mm/d
(Pretty Pine)
clay to 1-2 m at surface
Finley Heavy clay to 8-9m, overlying q ≈ 17,500 EC 3 cells Resistivity &
sandy clay. q ≈ 1.5 m EM31
5 – 7 mm/d
Additional Sites Used in EM31 Analysis
Wimmera Donald Main Surface layer (approx 0.5-1m) of q ≈ 30,000 EC 6 cells EM31 only
Mallee Water Channel clayey sand overlying predomin- q ≈2m
antly clay, but replaced by fine- 9 – 48 mm/d
med. grained sand in southern
half of trial site, starting near
surface and dropping to 2-3m
below surface in section centre
Rocklands Predominantly sandy clay in q ≈ 11,000 EC 6 cells EM31 only
southern half of site overlying q ≈5m
sandstone at depth. Sand at 1- 4 – 13 mm/d
2m rising to surface, also in
south. Sandstone rises to near
surface in northern part of site,
underlying shallow clay soils.
Goulburn Waranga Predominantly clay to sandy clay q ≈ 25,000 EC 11 cells EM31 only
Murray Water Western overlying sandstone at varying q ≈8m
depths. Fine to medium grained 1 – 13 mm/d
Channel
sand intersect channel and at
depth at various intervals along
the channel
Table 5-1 provides an overview of the key characteristics of each of the year three trial
sites, including:
q A summary of lithology;
q Background groundwater salinity and depth to groundwater; and,
q The number of pondage tests conducted at the site and the range of seepage rates
obtained during the tests.

This table illustrates that the seepage sites encompass a wide range of site conditions,
in terms of lithology, groundwater depth and salinity, and seepage rates. Further
details of the lithology at each of the sites are contained in Appendix A, which presents
a geological long section of each of the sites.
5.2 EM34 Trial Results

5.2.1 Introduction
5.2.1.1 EM34 Systems
Frequency domain electromagnetic (FEM) systems can measure the electromagnetic
properties of the soil profile up to a depth of 100m, with the penetration depth
dependent on the frequency and coil spacing. These studies used the Geonics style
FEM units which utilise the concept of low induction numbers to give an output in
conductivity. For a given coil spacing, Geonics EM systems can be used in horizontal
dipole or vertical dipole mode. The dipole mode effects the relative contribution of
the profile at different depths to the overall response. In general, near surface features
tend to dominate in the horizontal mode while the vertical mode is more influenced by
the ‘mid’ part of its depth range (McNeil, 1980).
Geonics EM34 systems can be used at various intercoil spacings so as to vary the
effective depth of exploration, in contrast to Geonics EM31 systems which have a
fixed coil spacing. Table 5-2 presents maximum exploration depths for Geonics
EM34 systems at various coil spacings and dipole orientations. Figure 5-2 shows an
EM34 unit in operation. It is a slower technique than EM31, as the required coil
spacing means that the coils must be carried by hand.
n Table 5-2 Exploration Depths for EM34 at Various Intercoil Spacings (after
McNeil, 1980)
Exploration Depth (m)

Intercoil Spacing (m)
Horizontal Dipoles Vertical Dipoles
10 7.5 15
20 15 30
40 30 60

n Figure 5-2 EM34 In Operation
5.2.1.2 Methodology
A horizontal dipole orientation was used in all of the EM34 trials, usually at a coil
spacing of 10m, with 20m also trialed at some sites. Based on previous experience
with EM34 channel seepage assessment (SKM, 1998) and on results emerging from
the initial EM31 trials, only one survey line on each side of the channel was conducted
(compared to EM31 where 3-4 lines were conducted). Each line was located
immediately adjacent the outside toe of the channel. The surveys were conducted by
subcontractors with experience in EM34 operation.
As described above (refer Section 5.1.2), the basic assessment methodology consisted
of comparison of the pondage test results to the average EM34 conductivity for the
corresponding pond length.
5.2.2 Results
The following EM34 surveys were conducted during the three years of trials and are
reported on in this section:
q Rocklands – November 1999 and August 2001;
q Donald Main – October 1999 and September 2001;
q Toolondo Central – August 2001; and,
q Dahwilly Central – February 2002.

5.2.2.1 Rocklands – November 1999 and August 2001
November 1999
An EM34 conductivity survey was conducted from 1st – 4th November 1999 (by sub-
contractor Michael Murphy), along the down gradient outside channel toe along 30km
of the Rocklands channel. The survey was conducted using a coil spacing of 10m
(horizontal dipole) and the channel was running at the time of the survey. Depth to
watertable at the site is around 5m, therefore the technique measures the conductivity
of the profile above and several metres below the watertable. The section of interest
from the survey (ie where the pondage tests were conducted) and the associated
conductivities are shown in Figure 5-3.
The EM34 data were plotted against the pondage tests results (average down gradient
channel toe EM34 conductivity for the pondage section verses the steady state
pondage test seepage rate). This analysis shows a consistent relationship between
seepage and conductivity as shown in Figure 5-4, (R2 = 0.79) which indicates a
reasonable correlation between the two variables. Despite a fairly narrow range of
seepage rates (4 – 13 mm/d), there is a general pattern that the lower conductivity
results relate to higher seepage.
The most significant deviation from the line is Pond 4, which according to the line of
best fit, would be expected to record a higher conductivity than was actually detected
in the survey (33 mS/m). There is no obvious explanation for this divergence. The
Rocklands geological section (refer Appendix A) does not indicate anything
anomalous with the geology beneath Pond 4.
However, the deviation is not highly significant and the overall correlation is
reasonable. The relationship established at the Rocklands site between seepage and
EM34 conductivity, as indicated by the regression equation is:
Seepage (mm/d) = 19.5 – 0.34 EM34Cond. (mS/m)
It can be concluded that the average EM34 conductivity correlates reasonably well
with pondage test seepage rates at this Rocklands site. The analysis suggests that the
relationship established can be used to predict seepage rates along the Rocklands
channel, in areas of like geology and hydrogeology.
August 2001
To test the repeatability of the results obtained in the November 1999 EM34 survey,
the survey was repeated on 11th August 2001 along the pondage test sections of the
Rocklands Channel. The same contractor, coil separation, dipole orientation
(horizontal) and location (along the down gradient outside channel toe) were used, and
again the channel was in operation at the time of the survey. The mapped results of
this survey have not been presented in this report as they are virtually identical to
those obtained in the November 1999 survey (Figure 5-3). In addition to the
November 1999 data, Figure 5-4 also plots the August 2001 data. This figure shows
the high degree of similarity between the two surveys, with again a high correlation
coefficient of 0.75 obtained and a very similar line of best fit equation.

n Figure 5-3 Rocklands EM34 Survey – November, 1999
The EM34 surveys at Rocklands demonstrate the repeatability of the method. It also
indicates that conditions between the two surveys were reasonably constant
(particularly seepage rates and groundwater levels). Not only were the average
conductivity results across the pondage tests virtually identical, the individual results
along the channel compared very well to each other in terms of magnitude and the
coincidence of peaks and troughs in conductivity (graphs not presented in this report).

n Figure 5-4 Rocklands EM34 (Nov 1999) verses Pondage Test Seepage
Rocklands Pondage Test Seepage Results vs EM34 Conductivity (November 99 and August 01)
14
Nov-99
Aug-01
P3
12 Linear (Nov-99)
P2 Linear (Aug-01)
10
P1
8
6
P4
P6 P5
4
2
y = -0.34x + 19.46 y = -0.28x + 17.60
R2 = 0.79 R2 = 0.75
0
15 25 35 45 55
5.2.2.2 Donald Main – October 1999 and September 2001

October 1999
An EM34 survey (10m coil spacing, horizontal dipole) was conducted from 20th - 25th
October 1999, along the down gradient outside channel toe along a 43km section the
Donald Main channel, east of Lake Buloke. The survey coincided with the shut down
of the channel, which had been running for the previous six months (groundwater
levels next to the channel were therefore quite high).
The depth to watertable at the site was about 2m below surface adjacent to the channel
at the time of the survey and therefore the EM34 measured the conductivity of the 2m
unsaturated interval and up to 5m below the watertable. The EM34 survey results
from the section of interest (ie where the pondage tests were carried out) are presented
in Figure 5-5.
The EM34 conductivity data along this section were plotted against the pondage test
results as presented in Figure 5-6. The pondage test seepage rates indicate that except
for one pond (Pond 6 with 9 mm/d), the seepage rates are in a fairly tight band ranging
between 35 mm/d and 48 mm/d. The first half of the pondage test results were used in
preference to the steady state results (which was the method used in analysis of other
channels) due to the relatively low initial water levels in the Donald pondage tests.
This analysis produced a poor correlation as shown in Figure 5-6, with a correlation
coefficient of 0.43. However, the general relationship of higher seepage relating to
lower conductivity is clear. The bulk of the results are clustered around the five
higher seepage rate ponds (22-42 mS/m) with the low seepage pond containing a
higher average conductivity of 54 mS/m. The results indicate that the EM34 survey

has distinguished between areas of high (35-50 mm/d) and low seepage (< 10 mm/d).
The EM34 survey has not been able to distinguish between various degrees of high
seepage (eg 35 mm/d cf 48 mm/d). The Donald Main Channel Geological Long
Section, (refer Appendix A), confirms the above interpretation of the EM34 results.
As the long section shows, Pond 6 is the only area to display a completely clayey
profile on both sides of the channel. This concurs with the highest conductivity and
the lowest measured seepage rate. Ponds 2, 3, 4 and 5 have a large sand unit 1-2m
below the channel base on the down gradient side of the channel, and Ponds 1-3 have
a sand unit virtually intersecting the channel on the up-gradient side of the channel.
Ponds 4 and 5 contribute significantly to the poor correlation and display higher
conductivities than would be expected for the magnitude of seepage losses detected in
the pondage tests (compared to ponds 1 - 3). The possible cause of the higher
conductivity results in ponds 4 and 5 is discussed in the following section, under the
September 2001 survey results.
n Figure 5-5 Donald Main Channel EM34 Survey – October, 1999

n Figure 5-6 Donald Main Channel Pondage Test Seepage verses EM34
Conductivity (October 1999)
Donald Main Channel Pondage Test Seepage vs EM34 Conductivity (October 1999)
60
50
P5
P1
40 P4
P3
P2
30 y = -0.75x + 60.90
R2 = 0.43
20
10
P6
0
10 20 30 40 50 60
September 2001
Based on previous results at the Donald Main Channel which indicated that EM34
could be used to identify relatively low and high seepage rate sites, the pondage test
sections of the Donald Main Channel were re-surveyed using EM34 (10m coil
spacing, horizontal dipole) on 21st September 2001. The same contractor and
technique were employed in this follow up survey. The channel had been flowing at
levels below full supply level for a little over one month (commencing 16/8/01), prior
to which it had not operated since December 2000. Depth to watertable was
approximately 3m adjacent the channel, approximately 1 - 1.5m lower, and 0.5 - 1m
lower away from the channel than compared to the October 1999 survey.
The results are presented in Figure 5-7 and a graph of these results as well as the
October 1999 results are plotted against pondage test seepage in Figure 5-8. As per
the 1999 EM34 survey, this analysis produced a poor correlation (R2 of 0.50).
However, again the general relationship of higher seepage relating to lower
conductivity is clearly evident.
Figure 5-7 also plots tree locations along the channel. It is worth noting the potential
effects of the trees on the results. In the October 1999 EM34 survey (described
above), it was observed that the two main ponds deviating from expected trends were
ponds 4 and 5, and that the conductivities were higher than expected. In the
September 2001 survey the two main ponds deviating from ‘expected’ are Pond 1 and
5, which again recorded higher conductivities than expected. Ponds 4 and 5, and to a
lesser degree Pond 1, are adjacent significant tree plantations. The trees adjacent
Ponds 4 and 5 were planted in order to control visible seepage and waterlogging in the
area (SKM, 1999). It is possible that the trees have taken up much of the seeped
channel water along these sections, concentrating salt in the profile and resulting in
overall higher conductivities than the corresponding sections of Ponds 2 and 3.

Alternatively, and perhaps in addition to the above mechanism, the trees may be
increasing seepage above what would naturally occur in the absence of the trees.
Correcting these ponds by lowering seepage rates, or reducing conductivities would
improve the correlation for this section of channel. This brief analysis indicates that
there is a need to consider the potential effects of trees adjacent the channel on
interpreting geophysical survey results. The effect is only likely to be significant
when there are significant numbers of trees involved (very close to the channel), and
where the sub-surface profile is sufficiently permeable to allow significant up-take of
the seeped water by the trees.
n Figure 5-7 Donald Main Channel EM34 Survey – September 2001

n Figure 5-8 Donald Main Channel Pondage Test Seepage vs EM34
Conductivity (October 1999 & September 2001)
Donald Main Channel Pondage Test Seepage vs EM34 Conductivity (April 2000 & Sept 2001)
60
Sept 2001 Survey
50 October 1999 Survey

P5
P1
40
P4
P3
P2
30
y = -1.07x + 88.21
y = -0.75x + 60.90 2
2 R = 0.50
R = 0.43
20
10
P6
0
10 20 30 40 50 60 70
Table 5-3 compares the average conductivities in each pondage section between the
two surveys. The September 2001 results consistently returned higher conductivities
than the October 1999 survey, with an overall average increase of 15 mS/m. This is
primarily due to the elevated and fresh groundwater levels in the October 1999 survey
due to 6 months of prior channel operation. The channel had only been running four
weeks (at reduced capacity) at the time of the September 2001 survey, possibly only
enough to flush some of the accumulated surficial salts down into the profile. The
October 1999 survey was therefore conducted under a sub-surface environment
dominated by seeped water and a flushed (ie relatively salt free) profile, while the
September 2001 survey was conducted in a environment probably only just beginning
to flush salts through the profile (ie relatively salt rich). Although groundwater levels
were lower in the September 2001 survey, the channel had been running for sufficient
time to ensure that the unsaturated zone was sufficiently moist to be conductive.
n Table 5-3 Comparison of Donald Main Channel EM34 Results: October 1999
and September 2001
Pondage Av. Sept 2001 Av. October 1999 Difference between Percentage
Conductivity Conductivity Sept 01 & Oct 99 increase from Oct
(mS/m) (mS/m) (mS/m) 99 to Sept 01
P1 51 28 23 83%
P2 44 25 19 76%
P3 39 22 17 79%
P4 49 42 7 17%
P5 49 36 13 35%
P6 66 54 13 23%

5.2.2.3 Toolondo Central – August 2001
An EM34 survey was conducted on 11th August 2001, along the down gradient outside
channel toe along 4 km of the Toolondo channel. The survey was conducted using a
coil spacing of 10m (horizontal dipole) and the channel was running at the time of the
survey. The results, plotted against pondage test seepage, are presented in Figure 5-9.
Only a moderate correlation is produced (R2 = 0.50), however the EM34
conductivities have distinguished between low (1 – 5 mm/d) and moderate seepage
rates (10 – 12 mm/d). The relationship is largely deteriorated by the high conductivity
recorded in Pond 1.
n Figure 5-9 Toolondo Channel Pondage Test Seepage Versus EM34

Conductivity (August 2001)
Toolondo Channel PT Seepage vs EM34 Conductivity (August 2001)
12
P1 EM34 (10m coil sep) - 11th August 01
P2
10 EM31- 6th August 01
P4
y = -0.40x + 34.94 y = -0.29x + 27.80

4 R2 = 0.50 R2 = 0.80
P3
P6
2
P5
0
55 65 75 85 95
Average conductivity (mS/m)
Figure 5-9 also plots the adjacent channel toe EM31 survey results (vertical dipole),
which returned a good correlation coefficient for the adjacent channel toe data of 0.80
(refer Section 5.3.3.1 for further discussion). The EM31 survey was conducted only
five days before the EM34, so it can be assumed groundwater conditions were
essentially identical. Understanding the differences between the EM31 (vertical
dipole) and the EM34 (horizontal dipole, 10m coil spacing) surveys is important.
Figure 5-9 shows an interesting difference between the two surveys. The EM31
conductivities were consistently higher than the EM34 results for the low seepage
ponds, whereas the EM34 conductivities were higher than the EM31 in the three
higher seepage rate ponds. This is probably explained by the deeper penetration depth
of the EM34. Depth to groundwater at the site is around 8-10m, therefore only the
EM34 with an effective exploration depth of around 8m may just ‘see’ into the
capillary fringe and the groundwater. In the sandy sections (Ponds 1, 2 and 4) the
EM34 is ‘seeing’ into the natural groundwater (ie more salty) and therefore returning a
higher conductivity. However in the clayey sections, the lithology, not the

groundwater salinity effect is the dominant feature of the response. Therefore in these
sections, the EM31 is detecting the higher conductivity of the clays compared to the
underlying sandstone, and consequently the reversal of the relationship between the
EM31 and EM34 survey results is observed in the high seepage and low seepage
areas. A possible counter to this interpretation is that while the EM34 (horizontal
dipole) may ‘see’ slightly deeper than EM31 (vertical), its depth focus is shallower.
5.2.2.4 Dahwilly Central and Dahwilly East – February 2002.

Dahwilly Central
A 2km length EM34 survey (10m and 20m coil spacing, horizontal dipole) was
conducted in February 2002, adjacent the outside toe of both sides of the Dahwilly
channel. The 10m coil spacing results (EM34-10m) are presented in Figure 5-10
(along with the Dahwilly East results).
The depth to groundwater at the time of the survey was approximately 5m. The
EM34(10m) configuration measures the conductivity of the 5m of unsaturated interval
and up to a couple of metres below the watertable. However, recall from the
introductory comments to this section that in horizontal dipole mode the focus of the
measurements is from the upper part of the profile. Therefore the relative contribution
from below the watertable is less than from above. At a 20m coil spacing, the bulk of
the measurement will be from below the watertable, with an effective penetration
depth of approximately 15m.
The average EM34 response per pond, plotted against pondage test seepage rates, is
presented in Figure 5-11. The pondage test seepage rates range from 1 to 10 mm/d
(the two low seepage rate ponds represent two ponds lined in July 2001). The
following comments are made regarding the results:
q The EM34(20m) data displayed a consistently higher conductivity than the
EM34(10m) data. This is due to the deeper penetration depth of the 20m
configuration, and reflects the increased influence of (salty) groundwater at
depth.
q Both the 10m and 20m coil spacing conductivities lie in a narrow seepage rate
range: 18 – 25 mS/m for 10m, and 32 – 44 mS/m for the 20 m coil spacing, ie
there is little differentiation between the ponds in terms of conductivity.
Excluding the lined ponds, the conductivity range is even narrower;
q The most important comment is regarding the effect of the lined ponds, which
appear to be skewing the results away from the ‘expected’ trend. The dark line
for both the 10m and 20m data sets is the regression line when the lined ponds are
excluded from the analysis, and the thinner lines are the regression lines for all
points including the lined ponds.
Higher conductivities were ‘expected’ beneath the lined ponds - given that they
have relatively low seepage rates, it was anticipated that they would contain a
greater proportion of (salty) groundwater compared to the unlined ponds.
However, it is apparent from these results that there may not have been sufficient
time between the pond lining (July 2001) and the EM34 survey (February 2002)
for the seepage water (prior to pond lining) to have migrated from beneath these
ponds. Over time it is anticipated that conductivities beneath the lined ponds will
increase as the lower salinity water originating from pre-lining seepage is
removed via advection and diffusion processes.

Therefore, the two lined ponds have been removed from the regression analysis.
n Figure 5-10 Dahwilly Channel EM34 Survey – February 2002

q The regression line for the EM34(10m) data, for the unlined ponds, displays a
good correlation coefficient (0.79) for the inverse relationship between seepage
and EM34 conductivity. This fits with the expected model for EM response –
that is that low conductivity equates to high seepage and high conductivity to low
seepage. However the average conductivity range for each pond over which the
relationship is derived is very narrow (22-25 mS/m) and therefore confidence in
this relationship is lessened. Two potential causes of the very narrow range are
suggested:
i) The very sandy nature and high transmissivity of the aquifer could result in
substantial mixing of the fresh water to adjacent pond, thus blurring the
identity of the pond from which the water originated and creating a
relatively uniform EM response in the saturated zone. The problem with
this explanation is that it runs contrary to the rationale for leaving the lined
ponds out of the analysis – that there has been insufficient dilution of pre-
lining seepage water beneath lined ponds.
ii) The 10m coil spacing and horizontal dipole orientation are largely affected
by surface response and are not highly influenced by the unsaturated zone.
Therefore the observed response for the 10m coil spacing is largely
dominated by the unsaturated zone. The geological logging at the site
(refer Appendix A) indicates that the unsaturated zone is relatively uniform
at this site which explains the very consistent response and narrow range of
results in the EM34(10m) data. This is considered the more feasible
explanation of the very narrow range observed. (Detection of unsaturated
zone differences to measure the likelihood of seepage has been
demonstrated to be successful at sites where there are significant
differences in the unsaturated zone, eg Rocklands, Toolondo. However at
this site it is apparent that the clogging layer is probably the controlling
influence on seepage, and therefore seepage differences must be detected
in terms of their impact on the groundwater salinity and not on soil
differences within the unsaturated zone);
q The 20m coil spacing data displays no trend, as confirmed by the fitted regression
line and resulting correlation coefficient (0.04). This is largely attributable to the
very narrow range over which the EM34(20m) response is spread (39 – 44
mS/m). Given the comments in the above dot point (that the narrow range in the
EM34(10m) data was due to a lack of penetration into the saturated zone) it
would be anticipated that the deeper penetration provided by the EM34(20m)
configuration should detect differences between high and low seepage ponds, as
caused by the varying volumes of fresh seepage water causing dilution of the
salty groundwater. This does not appear to be observed however. The response
is relative uniform across the five ponds. It is apparent that the EM34(20m)
configuration may have penetrated too deeply into the saturated zone, beneath the
area at and immediately below the watertable where the effects of seepage on
groundwater salinity are most dominant.
The potential of going too deep and missing the ideal zone in which to detect
seepage is well illustrated by the resistivity surveying conducted at this site (refer
Section 5.4.4.4 and Figure 5.96 Correlations of average resistivity values of
sections under ponds for Dahwilly Central). Figure 5.96 plots seepage versus
resistivity for different depth slices, and clearly shows that the ideal zone for
detecting seepage differences at the Dahwilly Central site is at and just below the

watertable. Therefore the best correlations are observed for the 6m, 8m and 10m
depth slices, while below this depth the results become more uniform across
ponds and reflect background groundwater salinities.
In summary it is apparent that the EM34(10m) configuration has not penetrated to
sufficient depth to detect changes in the saturated zone and was largely measuring
differences in the unsaturated zone, of which there are very little at the Dahwilly site.
In contrast the EM34(20m) configuration appears to have penetrated too deeply,
below the critical zone of seepage influence, and therefore tends to reflect background
groundwater salinities and displays a correspondingly uniform response. The
resistivity surveying (refer Section 5.4.4.4) which collects depth slices through the
profile confirms the above interpretation of the EM34 results, and shows that the
critical zone in which to detect seepage effects is at and immediately below the
watertable (approximately 5-8m at this site).
n Figure 5-11 Dahwilly Central EM34 Versus Pondage Test Seepage

Dahwilly Central EM34 Versus Pondage Test Seepage
10
y = -1.72x + 47.48
R2 = 0.79
8
y = 0.24x - 3.37
R2 = 0.04
6 EM34 - All ponds
(10m coil spacing)
EM34 - All ponds

4 (20m coil spacing)
EM34 - Excl. lined

ponds (10m coil
2 Lined Ponds spacing)
EM34 - Excl. lined
ponds (20m coil
spacing
0
15 20 25 30 35 40 45 50 55
EM34 Conductivity (mS/m)
Dahwilly East
A 2km, EM34 survey (10m and 20m coil spacing, horizontal dipole) was conducted in
February 2002, adjacent the outside toe of both sides of the Dahwilly channel,
approximately 4 to 5km east of the original Dahwilly site (described above). The 10m
coil spacing results are presented in Figure 5-10.
Depth to groundwater at this site was estimated to be around 5-6m at the time of the
survey. The EM34(10m) configuration measures the conductivity of the 5m of
unsaturated interval and up to a couple of metres below the watertable. However,
recall from the introductory comments to this section that in horizontal dipole mode
the focus of the measurements is from the upper part of the profile. Therefore the

relative contribution from below the watertable is less than from above. At a 20m coil
spacing, the bulk of the measurement will be from below the watertable, with an
effective penetration depth of approximately 15m.
The EM34 plotted against the pondage test seepage are presented in Figure 5-12. As
this figure shows, only three pondage test cells were conducted at the Dahwilly East
site. The Dahwilly Central results are also plotted in this figure for comparison
purposes. A much wider range of conductivities is observed at the Dahwilly East site,
with 24 – 40 mS/m for 10m, and 47 – 71 mS/m for the 20 m coil spacing.
The pondage test seepage rates for the three ponds is very narrow (8.9 mm/d – 10.4
mm/d) which limits the statistical significance of the results. However, the highest
seepage rate pond did record the lowest conductivity, suggesting that at this site the
inverse relationship between conductivity and seepage (observed at most sites) is
applicable. The EM34(10m) trend line is very much flatter than the steep trend line
observed at the Dahwilly East site.
The geology of the site also suggests that the ‘normal’ inverse relationship could be
expected at this site. While the unsaturated zone is quite similar to Dahwilly Central
(high permeability sands), the saturated zone is comprised of a clayey sand (refer
Dahwilly East Geological Long Section, Appendix A). This would slow the lateral
mixing of seeped channel water within the upper part of the aquifer and increase the
potential for detecting seepage water from beneath the pond from which it was
sourced.
Additional pondage tests across a wider seepage rate range would be required to
confirm the observed relationships at this site between pondage tests and conductivity
at other sites.
n Figure 5-12 Dahwilly East EM34 Versus Pondage Test Seepage
Dahwilly and Dahwilly East EM34 Versus Pondage Test Seepage

14
12
y = -0.04x + 11.42 y = -0.03x + 11.68

2
R = 0.26 2
R = 0.24
10
6 10m - Dahwilly Central
20m - Dahwilly Central

4 10m - Dahwilly East
20m - Dahwilly East

2
15 25 35 45 55 65 75

5.2.3 Conclusions
mechanism is predominantly via inference based on soil properties in the unsaturated
zone. Key summary comments for each of the sites are listed below:
q Rocklands – A good relationship was recorded in both surveys. A high degree of

repeatability was demonstrated between the two EM34 surveys conducted.
q Donald Main – A moderate relationship was recorded in both surveys. Further
points are required in the mid-seepage range to appropriately test the relationship.
The technique distinguished between high and low seepage but not within the
high seepage results range. Possible interference by adjacent trees may have
effected results in some ponds. A generally consistent increase in conductivity
was observed between repeat surveys. The difference was caused by the higher
watertable and reduced channel running time prior to the survey.
q Toolondo Central – A moderate relationship was observed but largely skewed by
the result in one pond. The relationship distinguished between high and low
seepage rates.
q Dahwilly Central - Moderate relationship for 10m coil separation but a very low
range of conductivity response was recorded across the five ponds used in the
analysis. This is because the EM34(10m) configuration does not penetrate to
sufficient depth to significantly detect changes in the groundwater and was
therefore mainly measuring differences in the unsaturated zone, which is largely
uniform at the Dahwilly site. The EM34(20m) configuration penetrated too
deeply below the watertable and therefore a uniform response was observed
reflecting native groundwater conditions.
q Dahwilly East – No relationship was observed. The seepage rate range was too
narrow for a meaningful relationship to be derived.
In summary, the only site where no relationship was observed was at Dahwilly East,
which was largely due to the narrow seepage rate range. At the Toolondo central site,
where conductivity measurement was entirely above the watertable, the unsaturated
zone lithology was a sufficiently accurate indicator of seepage and hence a reasonable
trend was observed (a fact reinforced by the success of EM31 at the site).
Significantly, the resistivity surveying showed improved correlations compared to the
EM34, for the depth slices focussed immediately below the watertable.
compared to the EM34 (R2=0.73 compared to R2=0.50), but neither survey
differentiated between the higher seeping ponds. The improved correlation is
probably attributable to the deeper depth focus of the EM31 compared to the EM34
(10m, vertical dipole configuration).

resistivity surveying was conducted, an improved relationship was obtained when the
depth slice was focussed immediately below the watertable, where seepage impacts
are most discernible.
5.3 EM31 Trials

5.3.1 Introduction
This section is presented in the following sub-sections:
q Background – This section describes the functioning of EM31 systems (their
typical response curves etc) as well as the specific methodology as to how they
were used in the trials (eg location, dipole orientation etc);
q Year 1 and 2 Trials - This presents the results of the year 1 and 2 EM31 trials,
across seven sites;
q Regional Assessment of Key Relationships – For all of the year 3 sites (plus some
year 1 and 2 sites) an attempt was made to look for potential correlations between
seepage rates across all sites and EM31 response. This was conducted applying
multiple linear regression using a number of key explanatory variables and simple
linear regression, using EM31 as the only explanatory variable.
q Channel Specific Assessment EM31 Assessment and Extrapolation of
Relationships - This section examines the EM31 - seepage relationship at each
channel and assesses this relationship compared to the regional relationship. This
section also assesses the success of extrapolation of seepage-EM31 relationships
from the ‘original’ site to other sites along the channel.
5.3.2 Background
5.3.2.1 EM31 Systems
Frequency domain electromagnetic systems can measure the electromagnetic
properties of the soil profile up to a depth of 100m, with the penetration depth
dependent on the frequency and coil spacing. This investigation used the Geonics
style Frequency Domain Electromagnetics (FEM) units which utilise the concept of
low induction numbers to give an output in conductivity. For a given coil spacing,
Geonics EM systems can be used in horizontal dipole or vertical dipole mode. The
dipole mode effects the relative contribution of the profile at different depths to the
overall response.
An EM31 system has a fixed coil spacing of 3.66m. This coil spacing, in the vertical
dipole orientation, provides an effective penetration depth of around 6m with the
dominant effect in the 1 – 3m range. Figure 5-13 presents the cumulative response for
Geonics EM31 systems for vertical and horizontal dipole orientations. It can be seen
from these curves that near surface features will tend to dominate in the horizontal
mode while the vertical mode is more influenced by the ‘mid’ part of its depth range
(where the gradient of these curves is steepest indicates from where the greatest
proportion of response emanates). For example, in horizontal mode about 40% of the

response comes from the top 1m of the profile, whereas approximately only 10% of
response is attributable to the upper 1m in vertical dipole mode.
EM31 and EM34 systems were used in years 1 and 2 of the trials, and both with
reasonable success. The attraction of an EM31 over an EM34 system (and the reason
it was adopted as the preferred EM system in the year 3 trials) is that it can be
mounted on a four wheeled motorbike and therefore represents a more rapid and
cheaper method of assessment. On the four-wheeled motorbike, readings were taken
at approximately 5m intervals. Surveys were usually conducted while the channel was
in operation.
n Figure 5-13 Cumulative Response Versus Depth For EM31 Horizontal and
Vertical Dipoles (after McNeil, 1980)
Cumulative response versus depth for EM31 vertical & horizontal dipoles
100
80
Cumulative Response (%)
Vertical Dipole
60
40
Horizontal Dipole
20
0
0 1 2 3 4 5 6 7
Depth (m)
5.3.2.2 Methodology
Generally the EM31 survey was conducted in eight traverses, four on each side of the
channel, except where surface features prevented access. The inner line was located
immediately adjacent the channel bank and the outer line up to 50-60m from the
channel. In the second year of trials, EM31 was also conducted on the channel (in a
boat) but the land based surveys consistently returned better results (ie better
correlations with pondage tests). Therefore land based surveys adjacent the outside
toe of the channel were the preferred survey method in the third year trials. Figure
5.14 illustrates the EM31 set up, both as conventionally used on land and the on-
channel boat system.
As described in Section 5.1.2, the basic assessment methodology consisted of

comparison of the pondage test results to the average EM34 conductivity for the
corresponding pond length. In addition, multi-variate analysis was conducted on the

data collected in the third year trials, in order to identify other parameters which may
have a significant impact on seepage rate and / or EM conductivity.
n Figure 5.14 EM31 Set-up on Land and On-channel (Toolondo Channel

August 2002)
5.3.3 Year 1 and 2 – EM31
This section presents the results of the year 1 and 2, EM31 trials,involving:
q Toolondo Central – December 2000, August 2001;
q Rocklands – August 2001;
q Donald Main – August 2001;
q Dahwilly Central – June 2000;
q Finley – July 2000;
q Lake View – June 2000; and,
q Tabbita – July 2000.

5.3.3.1 Toolondo Central – December 2000, August 2001
December 2000
An EM31 resistivity survey (vertical dipole) was conducted by Ken Bates Soil
Surveying on 13th and 14th December 2000 along a 4km section of the Toolondo
channel. The survey was carried out soon after the channel ceased flowing. In total
nine traverses were conducted, four on each side of the channel up to 50m from each
channel bank, and one in the channel (where possible). The base of the channel was
wet in places and an in-channel run could only be conducted in part. Survey results
are presented in Figure 5-15. (For analysis purposes, values were interpolated for a
75m section on the right hand side of Pond 2 to correct for missing data in this
location). Depth to groundwater at the site is approximately 9-10m, therefore EM31
surveys are measuring unsaturated zone properties.
Figure 5-15 shows a dominant response in the higher conductivity end of the range
with most of the channel and surrounds returning responses above 60 mS/m. The
areas around Pond 1, Pond 2, Pond 4 and the far southern area surveyed displayed the
lowest conductivities. The conductivities displayed in Ponds 1 and 2 suggest that
seepage is predominantly to the south, but also to the east in Pond 2. Pond 4 appears
to be seeping to both the east and west. The conductivities of Ponds 5 and 6 would
suggest low seepage in these areas and Pond 3 would appear to be seeping at the
southern end of the pond only. The areas of lowest conductivity on the plan, generally
coincide with the ponds where highest seepage rates were recorded.
The assessment shown by mapping is supported by plots of pondage test seepage

against the average EM31 results (refer Figure 5-16 and Figure 5-17). Figure 5-16
shows three averages of the survey data: all traverses, traverses on the LHS of the
channel and traverses on the RHS of the channel. Figure 5-17 shows two averages:
the in-channel traverse only and the channel sides (ie both sides excluding the in-
channel traverse). For each pond the average conductivity of these different traverses
was plotted against the corresponding (pondage test) seepage rate.
All combinations of the traverses except for the in-channel traverse produced a strong
(inverse) correlation. That is, there is a strong relationship between the average
electrical conductivity in the top four to six metres adjacent the channel and (pondage
test) seepage rates. Data from all traverses, including the in-channel run, produced a
correlation co-efficient of 0.72.
The results on the up-gradient side of the channel (RHS), produced a stronger
correlation (R2 of 0.78) than the down-gradient (LHS) side (R2 of 0.33). This is
attributable to the fact that the Toolondo trial site does not slope steeply however, and
therefore there is not a strong gravity driven preferential flow direction. The nature of
the geology at the site has a stronger influence on seepage. It appears that water is
preferentially being lost via the sandy up-slope sides of Ponds 1 and 2, rather than the
clayey down slope sides (shown on cross section in Appendix A). It can be concluded
that the average EM31 response for all data along the channel provides a reasonable
correlation with the pondage test results, while detailed breakdown of data is less
conclusive. This is largely a reflection of the gross characteristics of the EM31
averaging process and pondage test procedures.

n Figure 5-15 Toolondo EM31 Land Survey – December 2000
Figure 5-17 shows that using only the traverses outside of the channel (both LHS and
RHS) a strong correlation is obtained, comparable to the coefficient for all traverses,
of R2 = 0.73. The line of best fit for the in-channel data is poor, (and therefore not
plotted) due to the high conductivities in the high seepage ponds (Ponds 1 and 2, and
to a lesser in extent Pond 3). These high conductivities are explained by the fact that
immediately beneath the channel all the pores contain fresh water, and are effectively
saturated, even beneath low seepage ponds. This will tend towards a uniform
conductivity response, dominated by seepage water rather than lithology. In contrast,
the outside channel toe results infer the likelihood of seepage based on unsaturated
zone lithology (watertable at 10m) but are not effected by seepage water.

n Figure 5-16 EM31 Survey Results (Dec 2000)
Toolondo Channel PT Seepage vs EM31 Conductivity (Dec 2000)
14
All data
12 Left Side
P1 Right Side
P2
10
Linear (All data)

P4
y = -0.24x + 21.55 Linear (Right Side)
R2 = 0.78
8 Linear (Left Side)
6
y = -0.20x + 20.84
y = -0.36x + 31.31
R2 = 0.33
R2 = 0.72
4
P3
P6
2
P5
0
30 40 50 60 70 80 90 100
Av EM31 Conductivity (mS/m)
n Figure 5-17 EM31 Survey Results cont’ (Dec 2000) – Toolondo Channel
Toolondo Channel PT Seepage vs EM31 Conductivity (Dec 2000)
14
12
P1
P2
10
P4
Channel Only
8 Channel Sides (to 50m)

y = -0.33x + 28.33
Linear (Channel Sides (to 50m))
R2 = 0.73
6
4
P3
P6
2
P5
0
30 40 50 60 70 80 90 100
Av EM31 Conductivity (mS/m)
permeable sections of channel (as indicated by the pondage test results) the fresh
water contained in the pore spaces from the recently running channel may have seeped
away quickly vertically (compared to the clayey sections). However, due to slower
lateral movement this fresh water was still present adjacent the channel and was
therefore able to be picked up in the traverse outside of the channel. (Evaporation and
subsequent salt build up of water ponded in the channel may also have caused higher

conductivities and led to inaccurate readings.) The significant aspect of these results
is the good correlations between the conductivity adjacent the channel and the channel
seepage.
August 2001
Land Based Survey
The repeatability of the good results obtained during the EM31 survey of the
Toolondo channel in December 2000 was tested by conducting another EM31 survey
in August 2001. In contrast to the December 2000 survey, the channel was flowing at
the time of the August 2001 EM31 survey. The same contractor was employed for the
task (Ken Bates Soil Surveying), and the same dipole orientation (vertical) was used.
The survey was undertaken on the 6th August 2001 along a 2km section of the
Toolondo channel, incorporating the pondage test sections. In total eight traverses
were conducted, four on each side of the channel up to 50m from each channel bank.
(As per the December 2000 survey, for analysis purposes values were interpolated for
a 75m section on the right hand side of Pondage 2 to correct for missing data in this
location). On-channel traverses were also undertaken in a boat and the results are
discussed in the following section.
A comparison between the EM31 (land-based) survey results of December 2000 and
August 2001 are presented in Figure 5-18 and a graphical presentation of the survey is
shown in Figure 5-19. Figure 5-18 shows that correlation between the two surveys
was strong (R2=0.96). There was a slight difference between the results of the surveys
n Figure 5-18 Comparison between December 2000 and August 2001 EM31
Surveys along Toolondo Channel
December 2000 vs August 2001 EM31 Conductivity along Toolondo Channel

90
85
y = 0.77x + 13.22
80
December 00 EM31 Conductivities
R2 = 0.96
75
70
65
60
55
P2 P1 P4 P6 P5 P3
50
50 55 60 65 70 75 80 85 90
August 01 EM31 Conductivities

n Figure 5-19 Toolondo EM31 Land Survey – August, 2001
but the difference was generally consistent, with the August survey returning higher
conductivities than the January survey (average of 3.4 mS/m higher for each pond).
This difference is not considered significant and demonstrates the repeatability of the
technique.
Figure 5-20 and Figure 5-21 present the results of the Toolondo August EM31 land
based survey compared to the pondage test results. These figures should be compared
with Figure 5-16 and Figure 5-17 which are the corresponding results for the
December survey. The differences are small, and it is concluded that the good
correlations between EM31 surveys along the outside of the channel and pondage tests
were successfully repeated and appear to be marginally better under channel flowing
conditions. As per the January results, significantly better correlations were obtained
on the right hand (up-gradient) side of the channel than on the left.

n Figure 5-20 EM31 Land Based Survey Results (August, 2001) – Toolondo
Channel
Toolondo PT Seepage vs EM31Land Based Conductivity (August 01 - Channel Flowing)
12
All data
P1 Right Side
P2 Left Side
10 Linear (All data)
P4
Linear (Left Side)
y = 0.94x + 10.9 Linear (Right Side)
8 R2 = 0.96
6
y = -0.30x + 28.19 y = -0.22x + 23.19
R2 = 0.81 R2 = 0.55
4
P3
P6
2
P5
0
40 50 60 70 80 90 100
n Figure 5-21 EM31 Land Based Survey Results cont’ (August, 2001) –
Toolondo Channel
Toolondo PT Seepage vs EM31Land Based Conductivity (August 2001 - Channel Flowing)
12
Channel Only
P1
P2 Channel Sides (to
10 50m)
P4
Linear (Channel
Only)
Linear (Channel
8 Sides (to 50m))
y = -0.30x + 28.44
R2 = 0.79
6
y = -0.29x + 27.80
R2 = 0.80
4
P3
P6
2
P5
0
40 50 60 70 80 90 100

On-Channel Survey
In addition to the land-based survey at Toolondo on 6th August 2001, an on-channel
survey was conducted on the same day using a small boat to house the EM31
equipment. The on-channel survey was conducted in both vertical and horizontal
dipole arrangements to provide different penetration depths beneath the channel. The
results of the surveys for the vertical and horizontal modes are presented in Figure
5-22 and Figure 5-23 respectively. The influence of the metre column of water above
the instrument will affect the results in a uniform manner, provided the water depth
along the section of interest is approximately constant.
It is worth comparing Figure 5-22 and 5-23 with Figure 5-19, the land based survey
taken on the same day as the boat survey. The low conductivity areas (usually
interpreted as higher seepage) generally match with each other in the two figures.
Within the pondage test area, however, there are two exceptions. The eastern half of
Pond 1 and the northern half of Pond 2 are displayed as high conductivity in the on-
channel (boat) results, but in the land-based survey are mapped as low conductivity on
the right hand side. The reason for these differences is explored below.
Table 5-4 presents the land (vertical) and boat (horizontal dipole) data and illustrates
the difference between the two surveys. As the final column illustrates, there is
generally a 15-20 mS/m increase in conductivity for the land based data over the same
pondage test section compared to the boat data (horizontal dipole). The reason for this
difference is that the on-channel based survey is measuring the conductivity of the
completely flushed profile immediately beneath the channel (flushed of salts etc),
whereas the land based survey adjacent the channel is measuring a non-flushed
profile, which still contains salts and therefore higher conductivities. The reason that
this difference in conductivity does not exist in Ponds 1 and 2, is due to their sandy
profile (refer to Toolondo long section in Appendix A), which results in a wider
flushed zone (evidently at least to the edge of the outside channel toe) compared to
other ponds and thus lower conductivities.
n Table 5-4 Comparison of Land and On-Channel EM31 Results, August 2001
Average Land
Average Boat Conductivity:
Pondage Conductivity: Vertical Land Data – Boat Data
Horizontal Dipole (mS/m)
Dipole (mS/m)
P1 62 65 -3
P2 55 63 -8
P3 89 67 23
P4 66 52 14
P5 85 66 19
P6 76 62 14
Average 72 62 10

n Figure 5-22 Toolondo On-Channel EM31 Survey, August 2001 – Vertical
Dipole
Figure 5-24 presents the results of the on-channel survey for both vertical and
horizontal dipole modes, against the pondage test seepage data. The following
comments are made regarding this graph:
q There is no trend at all in the vertical dipole data, and the correlation for the
horizontal dipole is also very poor. The reason for the uniformity of the
conductivities (with the exception of pond 4) and hence the lack of correlation, is
attributed to the flushing mechanism described above. Even for low seepage rate
ponds, the profile immediately beneath the channel is essentially saturated with

seeped water. This uniform saturation produces a uniform conductivity response.
It is apparent that the effect of the seepage saturation largely masks other
influences on response such as lithology.
q The horizontal dipole returns higher conductivities than the vertical
configuration, as it more influenced by the upper part of the profile. The upper
part of the profile consists of the channel water, a higher degree of saturation and
generally more clay than deeper in the profile, all of which will produce higher
conductivity results.
n Figure 5-23 Toolondo On-Channel EM31 Survey, August 2001 – Horizontal

Dipole

n Figure 5-24 EM31 On-Channel Survey Results (August 2000) – Toolondo
Channel
Toolondo Channel PT Seepage vs EM31 On-Channel Conductivity (August 2001)
12
P1
P2
10
P4
8
y = -0.38x + 30.49
y = -0.15x + 13.44 R2 = 0.21
R2 = 0.05
6
4 Horizontal Dipole
P3
Vertical Dipole
P6
2
P5
0
30 35 40 45 50 55 60 65 70
5.3.3.2 Rocklands – August 2001

A land based and on-channel EM31 survey was conducted by Ken Bates Soil
Surveying on 7th - 8th August 2001 along the pondage test sections of the Rocklands
channel. Approximately 20 mm of rain was received the night before the survey
(according to the local farmer), however given that the rain did not occur during the
survey, the most important prerequisite of uniform conditions throughout the test were
satisfied. The contractor reported that the ground was not wet during the survey. The
channel was also in operation at the time of the survey.
The groundwater depth at the site is approximately 5m, therefore the vertical dipole
EM31 survey would be just ‘penetrating’ into the top of the watertable.
Land-Based Survey
The land based survey results are presented in Figure 5-25. (The tree plantation
shown in this figure, west of Ponds 1 and 2, forced the western runs adjacent to Pond
1 to be squashed between the plantation and the channel and adjacent to Pond 2
prevented the traverses being conducted immediately adjacent to the pond). As this
figure illustrates, the most dominant response was in the middle part of the
conductivity range with most of the channel and surrounds returning responses of
between 30 - 45 mS/m. Ponds 2 and 3 displayed the lowest conductivity, while Pond
1 returned the highest response.
Figure 5-26 and Figure 5-27 plot the pondage test seepage results against the average
EM31 results. Figure 5-26 shows three combinations of data: all traverses, traverses

on the LHS of the channel and traverses on the RHS of the channel. Figure 5-27
shows two combinations: the traverses adjacent to the channel only and the channel
sides (ie both sides excluding the adjacent channel traverse). For the length of channel
over which each pondage test was conducted, for each of these combinations of
traverses, the average conductivity over the distance is plotted against the
corresponding (pondage test) seepage rate.
All combinations of the traverses produce a moderately strong inverse correlation

between EM31 conductivity and pondage test seepage rate. That is, there appears to
be a clear inverse relationship between the average electrical conductivity in the top
four to six metres adjacent to the channel and (pondage test) seepage rates. Using
data from all traverses, a correlation co-efficient of 0.33 was produced. The
correlation coefficient for traverses on the right hand side of the channel was stronger
at 0.56 and weaker on the left at 0.26. As clearly seen in Figure 5-26 the two points
weakening this relationship were Pond 1, which returned a higher than expected
conductivity, and Pond 4 which returned a lower than expected conductivity.

n Figure 5-25 Rocklands EM31 Survey – August, 2001
Pond 1 actually returned a much lower conductivity response in the EM34 survey of
around 35 mS/m compared to 51 mS/m, (LHS) in the EM31 survey. This points
towards the fact that the Pond 1 results for the August 2001 EM31 survey (LHS)
might be anomalous, as there was better correlation between the EM31 and EM34 for

the remaining ponds. The higher than expected response in Pond 1 may be due to the
increased seasonal influence of the adjacent tree plantation (LHS Pond 2) which also
extends on the outside of the survey results in Pond 1. The removal of water by the
trees would cause a higher conductivity response than normal due to the absence of
the fresh water which relatively acts as a resistor compared to the saltier soil and
groundwater.
The cause of the lower than expected response in Pond 4 is a matter requiring further
investigation (the low response is largely attributable to the low response on the LHS
of the channel). For example, the area on the LHS of the channel may sit in a
topographic low and therefore be receiving significantly more run-off, simulating
channel seepage.
The reason why the best trend is clearly produced by the line adjacent the channel (R2
= 0.82 in Figure 5-27) is largely due to the reduction in the conductivity immediately
adjacent Pond 1 compared to the average across the 50m away from the channel. As
discussed above, the increase in conductivity away from the channel (particularly
noticeable on the LHS) is apparently due to other non-channel influences (probably
trees). The more accurate traverse in terms of representing channel seepage, is
therefore that which is least effected by these external influences.
The geological cross section for the Rocklands channel (refer Appendix A) generally
concurs with the results of the EM31 survey. The clay and shallow sandstone
underlying ponds 4 to six corresponds with the generally high conductivity response in
these areas and the low measured seepage. The sandstone is apparently of fairly low
permeability, given that the base of pondage four appears to intercept sandstone along
most of its length and that the pondage test returned a relatively low seepage rate (5.4
mm/d).
The generally lower conductivity response in ponds 1 to 3 also agrees with relatively
more permeable sandy clay and fine sands underlying this area. Slightly higher
conductivities in Pond 1 are again consistent with sandy clays and rock, although there
is some sand beneath the channel centre line. The seepage rate is 8.7 mm/d, consistent
with a mixed lithology in the channel bed.
In summary, the best correlation between the EM31 survey and pondage seepage rates
are obtained with data from adjacent the toe of channel.

n Figure 5-26 EM31 Survey Results (August 2001) – Rocklands Channel
Rocklands Channel PT Seepage vs EM31 Conductivity (August 2001)

14
All data
Left Side
P3
12 Right Side
y = -0.28x + 18.68 Linear (Right Side)
P2 R2 = 0.33
Linear (Left Side)

10 y = -0.44x + 24.92
Linear (All data)
R2 = 0.56
P1
8
y = -0.16x + 14.02
2
6 R = 0.26
P4
P5 P6
4
2
25 30 35 40 45 50 55
n Figure 5-27 EM31 Survey Results cont’ (August 2001) – Rocklands Channel
Rocklands Channel PT Seepage vs EM31 Conductivity (August 2001)

14
Adjacent Channel Only
P3 Channel Sides - excl. adj.

12
Linear (Channel Sides - excl. adj.)
P2 Linear (Adjacent Channel Only)

y = -0.42x + 23.69
10 2
R = 0.82
P1
y = -0.20x + 15.32
8 2
R = 0.18
6
P4
P5 P6
4
2
25 30 35 40 45 50 55

The correlation decreases with distance from the channel, due in part to the mixing
effect of the seeped channel water with the native groundwater, and also due to the
interference effects of the trees adjacent Pond 1 and 2.
On-Channel Survey
In addition to the land-based survey at Rocklands in August 2001, an on-channel
survey was conducted on the same day. The survey was conducted in both vertical
and horizontal dipole arrangements to provide different penetration depths beneath the
channel. The results of the surveys for the vertical and horizontal modes are presented
in Figure 5-28 and Figure 5-29 respectively.
During the survey, the boat sits on about one metre of water, which will affect the
results. However the results will be affected in a uniform manner, provided the water
depth along the section of interest is approximately consistent. The difference
between the relatively high and low seeping areas is most clearly evident in the
horizontal dipole mode (Figure 5-29). The red sections of Ponds 1-3 are in sharp
contrast to the blue sections of Ponds 4-6.
Figure 5-30 presents the results of the on-channel (boat) EM31 survey at Rocklands
verses the pondage test seepage results. The following comments are applicable to
these results:
q The conductivities in the vertical mode are much lower than the horizontal mode
(all < 20 mS/m). This is predominantly due to the effect of the metre (or greater)
depth of water in the channel which is very much more influential on the
horizontal than the vertical dipole results. Of secondary importance is the
generally more clayey nature of the 1-2 metres immediately beneath the channel
base (due to clay infilling during channel construction and natural sedimentation
etc) which will cause a higher conductivity response. Therefore, apart from the
water conductivity, in the horizontal dipole mode, the EM31 is essentially
measuring the percentage of clay in the layer immediately beneath the channel
bed. This is an important difference compared to the land based surveys outside
the channel which are primarily detecting the seepage itself where it has
displaced the more saline groundwater.
q There is a good correlation of the horizontal dipole results with the pondage test
data. The contrast between the lower conductivity and higher seeping ponds (P1-
P3: 32-34 mS/m) and the high conductivity and lower seeping ponds (P4-P6: 43-
53 mS/m) is clearly evident. The mechanism for detecting this difference was
described above: in the horizontal dipole mode the main property which is
detected is the clay content in the metre or so below the channel bed. The good
correlation between the horizontal dipole mode at Rocklands suggests that the
efficiency of this clay ‘liner’ beneath the channel has a very important influence
on seepage at this site. It is important to note that this method will only work if
the depth of water in the channel is reasonably constant, as even fairly small
variations in depth could lead to false interpretation (eg an increase in depth of
water will appear as in increase in conductivity and will be interpreted as an
increased clay presence and therefore decrease seepage, and vice-versa).
q The Rocklands geological cross-section (refer Appendix A) assists with
interpretation of the results. This figure shows that Ponds 1 and 2 are the only
ponds to contain sand deposits in the top 1-2 metres beneath the channel base.
The geophysical data suggests that the sand layer beneath Pond 2 may well
extend into Pond 3. However there were no bores in the section to help confirm

this theory. In contrast, the upper profile of Ponds 4-6 is dominated by clay and
sandstone.
n Figure 5-28 Rocklands On-Channel EM31 Survey, August 2001 – Vertical

Dipole

n Figure 5-29 Rocklands On-Channel EM31 Survey, August 2001 – Horizontal
Dipole

q A trend line using a power function has also been fitted to the horizontal dipole
data, to suggest that, in the horizontal dipole mode, a linear relationship may not
be the most accurate formula for describing the conductivity – seepage
relationship.
q There is a poor correlation between vertical dipole data and seepage. This is
partly because of low conductivity readings. The vertical data is mostly less than
15 mS/m which is approaching the reading resolution limits of the meter. An
error of at least +/- 2 mS/m can be expected at low conductivities. In addition, at
low conductivities small changes in other physical properties such as porosity,
cation exchange capacity and salt storage will have a large influence on the total
reading. The very low conductivities immediately beneath the channel (below
one metre or so beneath the base) reflect a flushed zone. This zone appears to
have been flushed of all salt, regardless of lithology and primarily represents the
conductivity of the fresh water and the porosity of the sediment). The absence or
presence of clay at this depth would obviously not control seepage as much as
clay closer to the channel, which is why the correlation breaks down in this
instance.
In summary, the in-channel EM31 in the horizontal dipole mode produced good
correlations between pondage test seepage rates by identifying clay content
immediately beneath the channel. The EM31 in the vertical dipole mode produced
poor correlations due in part to the poor resolution of the meter at low conductivities,
and in part due to the flushed zone beneath the channel, where clays and sands alike
have been completed filled with fresh water.
n Figure 5-30 Rocklands On-Channel EM31 Survey Results (August 2001) vs

Pondage Test Seepage
Rocklands Channel PT Seepage vs EM31 On-Channel Conductivity (August 2001)

14
Horizontal Dipole
P3
12 Vertical Dipole
P2
10
y = -0.36x + 22.62
P1 2
R = 0.78
8
y = -0.31x + 11.00
2 -2.12
R = 0.15 y = 17406.15x
2
R = 0.88
6
P4
P5 P6
4
2
0 10 20 30 40 50 60

5.3.3.3 Donald Main – September 2001
An EM31 survey was conducted by Ken Bates Soil Surveying on 6th September 2001
along a 2 km section of the Donald Main Channel. The channel had recently
commenced operation at reduced capacity (16/8/01) at the time of the survey, after
being out of operation for approximately 9 months. Both a land based and an on-
channel based survey were conducted. Depth to watertable at the time of the survey
was approximately 1.5m on the down slope side of the channel and approximately 3m
on the up slope side.
Land-Based EM31 Survey

The land-based survey was conducted with the EM31 in the vertical dipole
orientation. In total eight traverses were conducted, four on each side of the channel
up to around 50 m from each channel bank. The land based survey results are
presented in Figure 5-30. As this figure illustrates, the most dominant response was in
the 40 - 60 mS/m conductivity range, with the exception of Pond 6 which clearly
returned the highest response in the 70 - 80 mS/m range. Close examination of this
figure (particularly in Ponds 1-3) indicates lower conductivity readings closer to the
channel, and increasing away from the channel.
Figure 5-32 and Figure 5-33 plot the pondage test seepage results against the EM31
results for various combinations of the average conductivities from the survey runs.
Figure 5-32 shows three combinations of averages from the EM31 data: all traverses,
traverses on the LHS of the channel and traverses on the RHS of the channel. Figure
5-33 shows two combinations of averages from the EM31 data: the traverses adjacent
the channel only and the channel sides (ie both sides excluding the adjacent channel
traverse). The following observations are made from these plots:
q There is an inverse relationship between the average EM31 conductivity adjacent
the channel and (pondage test) seepage rates. Using data from all traverses
produced a correlation coefficient of 0.73. The correlation coefficient for
traverses on the right hand side and left hand of the channel produced weaker
correlations of 0.66 and 0.49 respectively, reflecting the ‘averaging’ of seepage
across the length and width of the pond section.
q There was little distinction between the seepage rates in the 35-50 mm/d range.
The correlation would be more useful if there was pondage test data in the middle
and lower seepage rate ranges.
q Figure 5-33 confirms the visual observation in Figure 5-31, that lower
conductivity readings occur adjacent the channel and increase away from the
channel. Figure 5-33 shows lower average conductivities near the channel in
Ponds 1, 3, 4 and 6. This is attributable to the greater influence of the low
salinity seepage water near the channel which is increasingly diluted by the high
salinity native groundwater away from the channel.
q The relationship between the EM31 land survey and the pondage tests is better
than the EM34 relationships. This is attributed to the deeper depth focus of
EM31 in vertical dipole (2-4m), which targets the watertable, compared to the
shallow depth focus of EM34 which is largely affected by the unsaturated zone.
EM34 surveys have been more successful in the past at Donald however, due to
the fact that the channel had been running longer prior to survey. The watertable
was therefore more elevated and the profile dominated to a greater degree by
seepage water. This highlights the importance of the channel running time on the

correct selection of EM survey equipment. There is an optimum time after
commencement of channel operation for surveying with a particular EM
configuration, depending on the hydraulic conductivity of the surrounding
formation and depth to watertable. For example, an EM configuration with a
shallow depth focus might be appropriate if the channel has been in operation for
a period of several months or more. However an EM configuration with a deeper
depth focus would be required if conducted soon after the channel begins
operation to ensure targeting of the watertable. If the channel has not been in
operation for a long period then the EM response will reflect local geological/soil
type changes and should be interpreted as such for zones of potential seepage.
n Figure 5-31 Donald Main Channel EM31 Land Based Survey – September
2001

n Figure 5-32 Donald Main Channel Pondage Test Seepage vs EM31 Land
Based Conductivity (September 2001)
Donald Channel PT Seepage vs EM31 Conductivity (September 2001)
60
All data
Left Side
y = -0.90x + 82.46 Right Side
50
P5 R2 = 0.73 Linear (Right Side)
Linear (Left Side)
P1
Linear (All data)

40
P4
P3 y = -0.59x + 64.16
P2 R2 = 0.49
30
20
y = -0.94x + 88.50
R2 = 0.66
10
P6
0
20 30 40 50 60 70 80
n Figure 5-33 Donald Main Channel Pondage Test Seepage vs EM31 Land
Based Conductivity, Continued (September 2001)
Donald Channel PT Seepage vs EM31 Conductivity (September 2001)
60
50 Channel Sides - excl. adj.

P5
P1
40
P4
P3
P2
30
y = -0.84x + 79.92
y = -0.94x + 81.59 R2 = 0.69
2
R = 0.69
20
10
P6
0
20 30 40 50 60 70 80

On-Channel EM31 Survey
In addition to the land-based survey at Donald Main in September 2001, an on-
channel survey was conducted on the same day. The survey was conducted in both
vertical and horizontal dipole arrangements to provide different penetration depths
beneath the channel. The results of the surveys for the vertical and horizontal modes
are presented in Figure 5-34 and Figure 5-35 respectively. The influence of the metre
column of water above the instrument will affect the results in a uniform manner,
provided the water depth along the section of interest is fairly constant.
The difference between the relatively high and low seeping areas can be seen in both
the dipole modes. It is worth comparing Figure 5-34 and Figure 5-35 with Figure
5-31, the land based survey taken on the same day as the boat survey. The low
conductivity areas (usually interpreted as higher seepage) generally match with each
other in the two figures, however the horizontal dipole mode consistently records a
higher conductivity than the vertical dipole, as can be seen by the greater dominance
of blue and yellow. The one exception to this is Pond 6, where a consistently high
conductivity response (red in Figures) is recorded for both ponds.
Table 5-5 presents the difference between the average land and boat data and between
the average horizontal and vertical dipole on the boat. The last columns of the table
compare the two dipole results on the water with the land results. The horizontal
dipole on the boat and the land based survey returned results of similar magnitude,
with the land survey displaying conductivities approximately 20% higher on average.
The land vertical dipole however returned much lower conductivities, with the land
survey generally equal to or greater than 100% of the boat survey in the vertical
dipole. The exception to this was Pond 6 where the results were almost the same
magnitude for the land and the boat (vertical dipole). The V/H column illustrates the
relationship between the vertical and horizontal dipoles. With the exception of Pond
6, the average vertical dipole response for each pondage section ranged between 48-
70% of the horizontal dipole response for the same pond. For Pond 6, the vertical
dipole response was actually marginally higher than the horizontal response.
n Table 5-5 Donald Main Channel - Comparison of Average Land and On-
Channel EM31 Results, September 2001
Pondage Land Boat Boat Boat V / Land – Horiz / Land – Vert /
(Vertical (Horizontal (Vertical Boat H Horiz Vert
Dipole) Dipole) Dipole)
(mS/m) (mS/m) (mS/m)
1 43 37 20 54 % 16% 117%
2 49 41 24 60% 20% 100%
3 41 38 18 48% 9% 129%
4 56 39 22 56% 44% 159%
5 50 41 29 70% 22% 74%
6 77 68 73 107% 12% 5%
Average 53 44 31 66% 21% 97%

n Figure 5-34 Donald Main Channel EM31 On-Channel Survey Results –
Vertical Dipole
The reason for the lower conductivity response in the surveys directly over the
channel (boat) is explained by the increased volume of fresh pore water immediately
below the channel, causing a reduction in the conductivity of the profile, compared to
the saltier pore water in sediments away from the channel.
The difference between the horizontal and vertical dipole response of the boat survey
is primarily controlled by the water in the channel which will dominate the horizontal
response. However an important secondary governing factor is the soils. This is best
understood by examining the Donald Main Channel Geological Long-Section (refer
Appendix A). This figure illustrates that generally the upper two metres along the
channel is comprised of clay, before more sandy and permeable sediments are
encountered. The shallower penetration depth of the horizontal dipole

n Figure 5-35 Donald Main Channel EM31 On-Channel Survey Results –
Horizontal Dipole
arrangement primarily targets the clays near the surface and therefore returns a higher
conductivity than the vertical dipole which has a response predominantly from the
deeper, sandier and less conductive sediments. The exception is Pond 6, where as the
geological section illustrates, this part of the channel is underlain by clay to at least 4m
depth below surface. Saturation of sand will result in low conductivity at depth,
whereas saturation of clay will result in higher conductivity at depth due to higher
porosity, cation exchange capacity and salt storage. Therefore there is essentially no
difference observed between the average responses in Pond 6 for the two dipole
configurations.

Figure 5-36 plots the results of the EM31 boat survey against the pondage test seepage
results. The correlation coefficient for the linear line of best fit was similar for the
vertical and horizontal dipoles, at 0.77 and 0.83 respectively. Similarly to the land
based EM31 results at the site, this figure illustrates that the survey did not distinguish
between seepage rates at the higher end of the range. For example in the horizontal
dipole mode Ponds 1 – 5 were in a tight band between 37 – 41 mS/m. For this reason,
the power function line of best with its flatter gradient through the higher seepage
results fits the data more closely. As with the land based data, the low seepage rate
Pond 6 is clearly distinct in its high conductivity response.
In summary, the relationship between the EM31 boat survey and the pondage tests is
comparable than the EM31 land based survey, which was in turn better the EM34
survey. To properly test this relationship however, more data points in the middle and
lower seepage rate ranges are required. The current trend line is heavily dependent on
the single low seepage rate result of Pond 6.
To summarise the results of the on-channel surveys, they did not work at sites where
the watertable was beyond the range of the EM31 (Toolondo), did work at sites with a
shallow watertable (Donald) and were partially successful when the watertable was
located at the edge of the depth penetration capacity of the EM31 (Rocklands).
Further work is required in this area, but the evidence collected in this investigation
suggests on-channel surveys should only be conducted where the geophysical
technique can penetrate into the watertable, and ideally target the top of the watertable.
For EM31 systems this would preclude EM31 on-channel use when the watertable is
deeper than approximately 4-5m.
n Figure 5-36 Donald Main Channel EM31
Donald Channel PT Seepage vs EM31 On-Channel Conductivity (September 2001)
60
Horizontal Dipole
50 Vertical Dipole
y = -0.57x + 52.72
R2 = 0.77
40
y = -1.04x + 80.76
R2 = 0.83
30
20
y = 472753.66x-2.56
R2 = 0.92
10
P6 P2 P3 P4 P1 P5
0
0 10 20 30 40 50 60 70 80
Average EM31 conductivity (mS/m)

5.3.3.4 Dahwilly Central – June 2000
An EM31 survey (vertical dipole) was conducted by Ken Bates Soil Surveying on 13th
June 2000 along a 5 km section of the Dahwilly channel. The channel was not
running at the time of the survey (and had ceased to run several weeks prior to the
survey). Depth to watertable at the time of the survey was around 5 to 6m. In total
eight traverses were conducted, four on each side of the channel up to 50m from each
channel bank. The survey results are presented in Figure 5-37. This figure shows that
the dominant response is in the low conductivity range. Virtually all of the pondage
test sections display average conductivities below 26 mS/m.
The low conductivities are primarily due to the sandy profile along the channel. The
geological long section at the site (refer Dahwilly Central Geological Long Section A,
Appendix A) shows that along the entire extent of the pondage sections a medium to
coarse sand layer is present from 1m to at least 4m below surface (Geological Long
Section B at the Dahwilly site shows that the sand extends to at least 10m depth).
Approximately one metre of sandy clay overlies this sand layer, but as the geological
section depicts, the elevation of the channel bed is approximately one metre below
surface and close to the top of the sand layer.
This geological profile is consistent with the long history of seepage problems
reported on this section of the channel. The geophysical response is also relatively
uniform and with the exception of Pond 6, the conductivities fall within a fairly
narrow band between 14-17 mS/m (for all data). However, despite the relative
uniformity of high permeability materials along the section, seepage rates in the
channel are not uniformly high across all ponds. It is noted that the seepage rates in
ponds one and two are very low (< 5 mm/d). It is understood that seepage is low in
these ponds due to the deposition of sediment upstream of the check at bank 3 (as
discussed in the pondage test section, 4.1.3), so that a low permeability layer overlies
the more permeable sandy interval.
Within the narrow conductivity range, there is no apparent trend of conductivity with
seepage rate, and if any trend at all, there appears to be an opposite trend to that
observed at other sites (ie increasing conductivity relates to increasing seepage).
Several factors may explain these results:
q The channel was not running at the time of the survey and previously seeped
water (ie from the previous irrigation season) is likely to have thoroughly mixed
with native groundwater. Therefore there is no seepage plume to detect.
q The main depth focus of EM31 is in the unsaturated zone. At some sites the
measurement of unsaturated zone soil properties with EM31 successfully
correlates with seepage based on the inferred likelihood of seepage, as indicated
by lithology (eg Toolondo, Rocklands). At these sites, however, a significant
difference between unsaturated zone soil properties is observed between ponds.
At the Dahwilly site the unsaturated zone is very uniform, and seepage rates are
actually controlled by the clogging layer, not the unsaturated zone. Therefore
seepage detection mechanisms must target impacts on the groundwater, not the
unsaturated zone. This is supported by the good correlations with resistivity at
this site, where the watertable is targeted (refer 5.4.4.4).
q The very high permeability of the aquifer at this site may be contributing to the
lack of distinction between seepage rates. At a ‘normal’ site the fresh seeped
water is detectable beneath the pond from which it has originated. However at

this site the very high permeability of the aquifer may be causing mixing of the
fresh water adjacent the ponds (by advection and dispersion), thus blurring the
identity of the pond from which the water emanated, and creating a relatively
uniform response in the saturated zone. (This is considered the least likely
explanation and the first two points are considered more feasible explanations).
q The high conductivity in pond 6 is due entirely to the high conductivity on the
LHS of the channel and is considered somewhat anomalous. This may be due to
clayey sand on this side of the channel, and also the resulting retained moisture.
Given the very high permeability of the unsaturated zone, seepage is still
controlled by the clogging layer in the channel and therefore the clay in the
unsaturated zone does not effect seepage rates.
n Figure 5-37 Dahwilly EM31 Survey Results, June 2000

n Figure 5-38 Dahwilly Pondage Test Seepage vs EM31 (July 2000) Results
Dahwilly Channel PT Seepage vs EM31 Land Based Conductivity (July 2000)

20
y = 0.93x - 6.17
R2 = 0.47
P6
15
y = 0.51x + 1.12
R2 = 0.49
P3
P5
10
P4
All data
y = 1.02x - 8.03 Left Side

5
P1 R2 = 0.04 Right Side
P2 Linear (All data)
Linear (Right Side)
Linear (Left Side)
0
0 5 10 15 20 25 30 35
5.3.3.5 Lake View – June 2000

An EM31 survey (vertical dipole) was conducted by Lloyd Angove Soil Surveying
and Drilling Pty Ltd on 28th June 2000 along a 2 km section of the Lake View channel.
In total 8 traverses were conducted, 4 on each side of the channel, up to 50m from
each channel bank. The channel was running at full supply level at the time of the
survey and had been running for the previous 9 months. Depth to watertable at the
site is approximately 1.5m. The survey results are presented in Figure 5-39. The
conductivities at the site are relatively high (median conductivity is approximately 90
mS/m), which is consistent with the overall clayey geological profile and shallow
depth to (saline) groundwater.
The lowest conductivities shown in this plan are on the left hand side of the channel
(for ponds 2-6), suggesting that there could be a fresh water plume and the grading of
this plume into the more saline background groundwater on the left hand side of the
ponds. The generally high conductivities, coupled with the relatively uniform clayey
profile, are consistent with uniform seepage rates as shown by the narrow range of
measured seepage rates (7.1 – 9.3 mm/d). However it is worth noting that these
seepage rates are higher than those obtained at other sites with clayey profiles (eg
Toolondo East).
Figure 5-40 and Figure 5-41 plot the pondage test results at Lake View against
average EM31 conductivity for the corresponding pondage section. (Note that Pond 6

was removed from the analysis as the EM31 survey only covered half of this pondage
section). Conclusions which can be drawn from these results are limited by the very
narrow range of seepage rates over which the correlation has been established
(approximately 2 mm/day). Further confidence in the results could be obtained by
conducting pondage tests across a wider range of seepage rates. With this in mind the
following conclusions can be drawn from the plots:
q There is little distinction between the conductivities across the ponds using all
data. This is apparently due to the fairly rapid mixing of the seeped plume back
into the native groundwater once away from the channel.
q The correlation of the adjacent channel results in Figure 5-41 is reasonable
(R2=0.58). In terms of predicting channel seepage, this result is more useful than
the correlation of one particular side of the channel, because without the pondage
test results, selecting which side of the channel to use for predictive purposes will
not always be obvious. However the lines adjacent the channel have consistently
returned good correlations with the pondage test data across most of the site
examined in this study.
q Overall the EM results at Lake View are positive and confirm the potential of EM
to map seepage plumes with a reasonable degree of accuracy compared to
pondage test data. These results suggest that EM31 (vertical dipole) is the
correct arrangement for (this section) of the Lake View channel to provide the
depth penetration required to pick up the diffusion of the seepage plume into the
groundwater.
q Most importantly these results show that at sites where a seepage plume is
quickly mixed back into the native groundwater, it is important that the adjacent
channel EM data is used in preference to all of the data, in order to adequately
distinguish between relatively high and low seepage sites.

n Figure 5-39 Lake View EM31 Survey Results, June 2000

n Figure 5-40 Lake View Pondage Test Results vs EM31 (June 2000) Results
Lake View Channel PT Seepage vs EM31 Conductivity (June 2000)
10
All data
Left Side
P2 Right Side
Linear (Right Side)
P3
P1
Linear (All data)
y = 0.00x + 7.92
8 2
R = 0.00
y = -0.04x + 12.41
2
y = -0.09x + 16.88 R = 0.61
2
R = 0.34
P5
7
P4
6
70 80 90 100 110 120 130 140 150
n Figure 5-41 Lake View Pondage Test Results vs EM31 (June 2000) Results
(cont’)
Lake View Channel PT Seepage vs EM31 Conductivity (June 2000)
10
P2 Channel Sides - excl. adj.

P1
9
P3
y = -0.05x + 13.52
R2 = 0.58
8
y = -0.08x + 15.84
R2 = 0.18
P5
7
P4
6
70 80 90 100 110 120 130 140 150

5.3.3.6 Tabbita – June 2000
An EM31 survey (vertical dipole) was conducted by Lloyd Angove Soil Surveying
and Drilling Pty Ltd on 28th June 2000 along a 3 km section of the Tabbita channel.
The survey results are presented in Figure 5-42. The channel was running at full
supply level at the time of the survey. Groundwater levels at the time of the survey
were approximately 1-1.5m below the natural surface adjacent the channel. In total
eight traverses were conducted, four on each side of the channel up to 50m from each
channel bank. On the eastern side of the channel access was limited in part due to
adjacent vineyards and in some sections only 2-3 traverses were possible.
As Figure 5-42 illustrates, the dominant response was in the high conductivity end of
the range with most of the channel and surrounds returning responses above 65 mS/m.
The left hand side, (which is also the down-slope side of the channel), of Ponds 1 and
5 are the only sections where significantly lower conductivities were detected. It is
apparent that water has migrated down-gradient, and is the probable cause of the lower
conductivity responses on the left hand side of the channel. Overall, the high
conductivities are primarily due to the clayey soil profile and shallow saline watertable
at the site (refer Appendix A Tabbita Geological Long Section). This logging of the
soil indicates most of the soils adjacent the channel section are comprised of light –
medium clays and sandy clays.
Figure 5-43 presents the results of the pondage tests plotted against the average
conductivities for the corresponding pondage section and shows no particular trend.
Pond 3 is perhaps the best example of the difference in these results compared to the
inverse conductivity-seepage relationship observed at other sites. Despite having the
highest seepage rate at nearly 10 mm/day, the overall conductivity response was the
second highest at 80 mS/m.
Several possible reasons are proposed for the lack of a meaningful seepage – EM31
relationship at this site:
1) The very narrow range of the seepage rates (difference between maximum and
minimum) detected in the pondage tests (approximately 4mm) suggests it is
difficult to find statistical correlations between variables when the range on the
control variable is very low. Essentially on average the ponds are seeping at a
constant rate across the pondage test area, and therefore trying to detect
differences compared to another variable may not be possible, as there are no
practical differences in the seepage rates.
2) The EM31 in vertical dipole mode may be seeing a little too deeply into the
aquifer, ie into native groundwater rather than seepage affected groundwater.
EM31 in horizontal mode may be a more appropriate configuration.
3) With such a low seepage rate range, the method of averaging conductivity over
ponds may be inappropriate. It may be more appropriate to map local
conductivity lows. Seepage cannot be considered to be occurring uniformly
across each pondage section, ie when there is high seepage over part of the pond,
then it is enough to sufficiently bias the conductivity results. However if seepage
is low then there might be the same amount of seepage from small sections in
what appears to be (in the conductivity survey) an otherwise non-seeping channel
section. Better placement of pond banks may over come some of these issues.
4) The primary seepage mechanism at this section of the Tabbita channel may be
such that the majority of seeped water does not reach the groundwater. The

geological long section of the Tabbita Channel (refer Appendix A) shows half a
metre of clayey sand along the surface of the site, intersecting the walls of the
channel. This suggests that the primary seepage mechanism is lateral seepage via
the upper clayey sand layer, through the walls of the channel and discharge onto
the outer channel toe. Reported problems with water logging and salinisation
along this stretch of the Tabbita Channel support this seepage mechanism theory.
n Figure 5-42 Tabbita EM31 Survey Results, June 2000

n Figure 5-43 Tabbita Pondage Test Seepage Versus Average EM31
Conductivity
Tabbita Channel PT Seepage vs EM31 Conductivity (June 00)
10
P3 All data
Left Side
Right Side

Linear (All data)
P4 Linear (Right Side)
8 y = 0.03x + 5.08
2
R = 0.05
y = 0.01x + 6.24
2
R = 0.01
7
P6
P1
P5
y = -0.08x + 14.25
6 2
P2 R = 0.20
5
60 70 80 90 100
If this is correct, the majority of seeped water would not reach the watertable and the
key means by which seepage is detected using geophysics is not possible (ie a
relatively fresh water plume imposed on more saline native groundwater). However,
section 4.4.4 casts some doubt over this theory. This section details groundwater
response to seepage at the site, and suggests a reasonable connection between the
channel and the groundwater.
5.3.3.7 Conclusions
corresponding pondage test seepage at most sites. At only one site (Tabbita) was there
no significant relationship identified. For EM31 in vertical dipole mode, the effective
depth of penetration is around 6-7m, with a mid-range depth focus of about 2 – 4.5m.
This meant that at sites where the watertable was deeper than 5m, only a limited
proportion of the response was caused by seepage impacts in the saturated zone.
Therefore at these sites the seepage detection mechanism is largely via inference based
on soil properties in the unsaturated zone. Key summary comments for each of the
sites are listed below:
Toolondo
q Good relationships between EM31 conductivity and pondage tests seepage were
recorded in all three surveys at Toolondo Central. This indicates that seepage
was able to be successfully inferred based on unsaturated zone soil properties.
q A high degree of repeatability between the surveys was observed.

q In-channel (shortly after channel shut down) and on-channel EM31 surveys
returned poor results. This is attributed to the fact that an EM31 survey above the
watertable ‘works’ by inferring seepage based on soil properties. However
immediately beneath the channel, even for low seepage rate ponds the profile
beneath the channel is saturated (or near saturated) with seeped water. This
uniform saturation produces a uniform conductivity response, and tends to mask
changes in lithology resulting in little differentiation between low and high
seepage sites. Significantly however the on-channel resistivity survey recorded
good correlations between seepage and resistivity (10m and 12m depth slices).
The EM31 on-channel however could not 'see' into the watertable.
q Better results were obtained with the EM31 compared to the EM34(10m) at this
site, possibly due to the greater number of EM31 traverses conducted (ie away
from the channel).
q Three Toolondo Sites (Central, East and West) - The relationship established for
all sites was moderately strong. Local correlations at Toolondo Central and
Toolondo West were stronger than the combination of sites. The Toolondo East
site displayed an opposite correlation, but the very narrow range of seepage rates
& the flat regression line indicates this is not a meaningful trend. Confidence
bands for the overall regression relationship are wide but indicate that the
relationship can be used to differentiate between high and low seepage sites. The
data most contributing to the low R2 and wide confidence bands is the four ponds
with sandy banks at Toolondo Central. It is apparent the shallow depth of the
sand causing the seepage (largely through channel banks) is largely missed by the
EM31(vertical) with a depth focus of around 2 - 4.5m.
q If the Toolondo Central site had been used to predict seepage at Toolondo West,
predicted seepage would have been 2-3 times too high. At Toolondo East it
would have been essentially accurate (0 mm/d), except in one pond seepage
would have been predicted at 4 mm/d when actual seepage is practically zero.
Rocklands
q A good relationship was observed between EM31 response and pondage test
seepage at the Rocklands channel trial site (for the adjacent channel EM31 data).
This indicates that seepage was able to be successfully inferred based on
unsaturated zone soil properties. However, with a depth to watertable of around
five metres, the EM31 survey may also have been detecting some seepage
induced salinity changes in the watertable.
q A poor response was observed when all survey runs were used, largely due to the
effect of trees adjacent one pond. The adjacent channel run was less affected and
accordingly better results were returned.
q The on-channel results recorded mixed results. In vertical dipole mode no trend
was observed. The configuration is focussed on the flushed zone beneath the
channel where uniform saturation from seepage appears to be masking lithology
response. In horizontal dipole a reasonable correlation was observed, apparently
through identification of lithology changes (clay content) immediately beneath
the channel. This was the only case observed where on-channel measurement
above the watertable successfully correlated with seepage. At other sites the
uniform saturation appeared to dominate the response over changes in lithology,
however at this site it is apparent that the changes in lithology close to the
channel surface are sufficiently contrasting to distinguish between high and low
seepage areas.

Donald Main
q A good relationship was observed between EM31 conductivity and pondage test
seepage but there is a poor spread of seepage data at the site (1 point of low and 5
of high seepage). With a relatively shallow watertable (2m), the EM31 detects
seepage at this site in terms of its impacts on the watertable. The EM31 survey
did not distinguish between higher seepage ponds (35 - 48mm/d). Confidence
bands are fairly wide for the regression line, particularly at the high conductivity
range, but indicate that the relationship can differentiate between high and low
seepage sites. Additional data points are required to tighten confidence bands.
q A better relationship was established with EM31 (R2=0.71) adjacent the channel
compared to EM34 (R2=0.50) but there was still no differentiation observed
between the higher seeping ponds. The improved relationship is probably due to
the greater depth focus of EM31, particularly on the up-slope side of the channel,
allowing deeper penetration into the watertable
q Moderate to good relationships were also observed for the on-channel surveys in
both horizontal and vertical dipole. With a shallow depth to watertable the EM31
on-channel survey detects seepage as it impacts the watertable.
Dahwilly
q For a survey conducted when the channel was not running, no relationship was
observed between EM31 conductivity and pondage test seepage. The technique
failed because the channel was not running and previously seeped water was
therefore likely to have thoroughly mixed with native groundwater. Unsaturated
zone lithology is a good indicator of seepage at some sites. However at Dahwilly
it is not the unsaturated zone controlling seepage rates, but the clogging layer at
the channel surface and therefore seepage must be detected directly (ie in terms of
impact on watertable) which means the channel must be in operation.
q In a repeat survey conducted when the channel was operating, a good relationship
was observed (at Dahwilly Central), confirming the importance of identifying the
seepage plume as the primary seepage detection mechanism at this site.
q Two Dahwilly Sites (Central and East) - The relationship established for both sites
is moderately strong. Local correlation at the Central site is slightly stronger than
the two sites combined. The East site displays a very weak correlation, but this is
due to the very narrow seepage range and few data points at this site. Confidence
bands are relatively wide, suggesting the regression relationship for both sites can
only be used to broadly indicate the likelihood of low or moderate seepage. The
slightly deeper depth to watertable at the East site appears to have put the
watertable largely beyond the range of EM31 and hence very different results are
obtained at the East site. Using the Central site regression relationship to predict
seepage at the East site would have resulted in over prediction of 1.5 - 2 times
actual seepage.
q Better correlations at both sites were obtained using the resistivity compared to
EM31 due to better targeting of the top of the watertable.
Lake View
q A poor relationship between pondage test seepage (July 2001) and EM31
conductivity (June 2000) was obtained at Lake View Central for all data due to
rapid mixing of the seepage plume away from the channel. However for
adjacent channel data a significantly improved relationship (to moderate) was
observed as seepage impacts are less diluted. Interpretation is limited at this site

due to the very narrow seepage rate range. Seepage is detected at this site in
terms of its impact on watertable salinity.
q No sensible trend was observed at the Lake View Central site using the same
EM31 survey data (all lines) and the June 2002 pondage tests. It is anticipated
however that a better response could be obtained using the adjacent channel data,
as was the case for the July 2001 pondage tests. In addition, the 2002 pondage
tests may not have been properly placed over sections of like conductivity.
q Both Sites (Central and West) - The relationship established for both sites is
moderately strong with a high correlation coefficient but the two data sets
creating the regression line have small conductivity and seepage rate ranges. It is
desirable to obtain data in the mid range to improve confidence in the
relationship. The Central site could not have been used to predict seepage at the
West site. However using Central data from adjacent the channel is likely to
improve this correlation.
Tabbita
q No relationship was observed between EM31 conductivity and pondage test
seepage. Possible reasons for the failure of the technique at this site include:
i) Narrow range of seepage rates (little differentiation in rates along section
of interest);
ii) Seepage mechanism may be such that majority of seeped water does not
reach watertable but move laterally (evaporating and causing salinisation as
evidence adjacent the channel);
iii) EM31 vertical dipole orientation may penetrate too deeply into the native
groundwater, below the zone most effected by seepage; and,
iv) The method of averaging conductivity may not be appropriate at this site
(or the ponds may need to be placed more carefully, ie over shorter sections
of high/low conductivity)
Finley
q While a moderate correlation coefficient was obtained for the pondage test –
EM31 conductivity relationship at this site (and the highest seeping pond did
record the lowest conductivity), the statistics are not meaningful due to the fact
that only three data points make up the relationship. The width of the prediction
intervals indicate that the regression relationship cannot be used to predict
seepage at this site. Additional data points across a wider seepage range are
required to improve the relationship.
In summary, the only site where no relationship was observed was at Tabbita. A
number of possible causes for this were identified, but the predominant contributing
factor is not known. At two sites (Rocklands and Lake View Central), the adjacent
channel data was used instead of all survey run data. This was required to obtain the
best relationship, due to the interference effects of trees and rapid mixing of seepage
water away from the channel.
At the Toolondo central site, where conductivity measurement was entirely above the
seepage and hence good trends were observed.

The Donald and Lake View site surveys were focussed on the saturated zone, and
seepage was detected as it created a conductivity low against higher background
conductivity groundwater.
At Waranga a reasonable relationship was observed, considering the distance over

which the data forming the relationship was spread. Improvements might be expected
using a technique targeting the top of the watertable at this site.
On-channel surveys did not work at sites where the watertable was beyond the range
of the EM31 (Toolondo), did work at sites with a shallow watertable (Donald) and
were partially successful when the watertable was located at the edge of the depth
penetration capacity of the EM31 (Rocklands). Further work is required in this area,
but the evidence collected in this investigation suggests on-channel surveys should
only be conducted where the geophysical technique can penetrate into the watertable,
and ideally target the top of the watertable. For EM31 systems this would preclude
EM31 on-channel use when the watertable is deeper than approximately 4-5m.

5.3.4 Regional Assessment of Key Relationships
For all of the sites described in Section 5.1.3 (site’s used in final year analysis), an
attempt was made to look for potential correlations between seepage rates across all
sites and EM31 response. This was conducted in two ways:
q Multiple linear regression, ie looking for correlation between seepage rate and
average EM31 response and other important variables which might effect seepage
(or EM31 response); and,
q Simple linear regression, ie looking for a direct correlation between seepage rate
and average EM31 response.
5.3.4.1 Multiple Linear Regression

Introduction
Multiple linear regression is a statistical technique that allows one response (or
dependent) variable to be predicted from a number of explanatory (or independent)
variables. In this study the response variable is seepage, as measured in pondage tests.
The explanatory variables which were included in the multiple linear regression
analysis included:
q EM31 conductivity;
q Soil permeability;
q Depth to watertable; and,
q Groundwater salinity.
The most important independent variable is EM31 conductivity. Previous work (years
1 and 2 trials) demonstrated that this is the most easily collected variable with a
generally strong correlation to seepage. The main aim of conducting multiple
regression analysis was therefore to determine which other variables could account for
variations in seepage apart from EM31. Groundwater salinity and depth to watertable
were included as independent variables because they will affect EM31 response.
The coefficients of the prediction equation are determined using a method of least
squares, that is, the coefficients are selected so that the sum of the squared residuals
are minimised. A residual is the difference between an observation and its predicted
value.
Data Sources
The response variable is observed seepage as measured in pondage tests (averaged
over the life of pondage tests, excluding outliers). This is the variable we want to be
able to explain. The explanatory variables and their method of collection is described
below:
q EM31 - Spatially this is the most reliable data source, with a high density of
readings over the pond length. EM31 response is measured at approximately 5m
intervals (using the EM31 set-up described in 5.3.2) and averaged over the length
of the pond.
q Soil Permeability - Average soil permeability was determined based on
geological cross sections constructed for each site (refer Appendix A). Four
different vertical hydraulic conductivities for each pondage cell were calculated,
and each tested in the multi-variate analysis as potential explanatory variables.

These are described below and graphically presented in Figure 4-4 (refer Section
4.2 - Sub-surface Characterisation):
i) Weighted according to typical EM31 (vertical dipole) response - most
heavily weighted around 1 – 3m depth and terminated at 7m [generally the
effective depth of EM31 (vertical) penetration];
ii) Representative of surface permeability and evenly weighted over the top
2m of the profile;
iii) An average across the entire profile but giving more weight to upper
surface layers and less to layers at depth (to 10m); and,
iv) Evenly weighted across 10m.
For each of the four scenarios, hydraulic conductivity was based on text book
vertical hydraulic conductivities for the given soil texture. The average across
multiple layers was calculated based on the method described in Freeze and
Cherry (1979). Details of these calculations, and an example to clarify the
process used, are provided in Appendix C.
q Groundwater salinities were obtained from regional hydrogeological maps
(MDBC 1:250,000 scale maps), or in the case of Dahwilly Channel at Murray
Irrigation using more detailed groundwater salinity information supplied by the
RWA. It is important to note that groundwater salinities based on adjacent
channel sampling (which are affected by mixing with fresher channel water) were
not used in this analysis, but rather groundwater salinities away from the effects
of channel seepage were adopted;
q Depth to watertable was obtained from near channel groundwater bores drilled
specifically for the project, or based on moisture and groundwater observations
recorded during drilling of the soil bores. At sites where soil bores did not
intersect groundwater, nearby groundwater bores were used to estimate the depth
to watertable.
Methodology
All of the variables described above were available for inclusion in the regression.
Stepwise multiple regression was used to select appropriate variables (using the
software program: SYSTAT, 1998). The process was interactive, allowing variables to
be added or removed from the model one at a time. Foremost the variables were
selected based on an understanding of the important seepage processes and the F
statistic. The F statistic is a measure of the amount of remaining variation (ie the
variation not explained by variables already in the model) explained by the variables.
Variables with an F-statistic less than 4 were not included in the model. Multi-
collinearly, the correlation of independent variables, was avoided.
Two measures were used to assess the accuracy, or ‘goodness’of fit of the regression.
These were the coefficient of determination (R²) and the standard error (SEE). The R²
measures the proportion of the total variation explained by the model. A large R² is
associated with a good model or prediction equation. The standard error is a measure
of the degree of scatter of the observed data points around the regression line. Hence
a small standard error is associated with a good model. The standard error has been
expressed as a percentage of the mean of the observed dependent variable (ie pondage
test seepage).

Removal of Outliers
Meaningful regression modelling involves the removal of any significant outliers.
Figure 5.44 presents average EM31 conductivity plotted against pondage test seepage
for all sites where trials were conducted, representing a total of 57 ponds. Each of the
sites is plotted with a different symbol to assist with identification of possible trends
between sites. Several outliers appear obvious in this data, including Donald Main
Channel, the Lake View West site and Finley channel.
The three Finley ponds, with conductivities three to five times higher than the next
highest response, are extreme outliers on the x-axis. Observation of the Finley
geological long section (refer Appendix A) indicates a heavy clayey profile to about
9m below the channel. In addition the groundwater at the site is highly saline
(approximately 17,000 EC) and the watertable at the site is quite shallow. All of these
factors combine to produce a very high conductivity response. The dominantly clay
profile suggests very low permeability and that the seepage mechanism is other than
vertical leakage through the profile. The seepage mechanism is most likely horizontal
bank seepage, associated with poor bank construction / compaction.
On this basis the Finley data was excluded from the overall analysis. It is apparent
that correlations specific to these type of environments would need to be established.
In this study there was insufficient data (three data points) to determine a meaningful
correlation in such a conductivity range. However, even within these three points the
lowest conductivity pond also had the highest seepage rate (ie the trend was in the
‘right’, or expected direction).
n Figure 5.44 EM31 Conductivity Versus Pondage Test Seepage at All Sites
50
Waranga Western
Donald Main
Toolondo
Rocklands
40
Dahwilly
Lake View (Original)
Lake View (West)

Finley
30
20
10
0
0 100 200 300 400 500

Results and Discussion
Multi-variate regression analysis was initially conducted on the entire data set
(excluding Finley), (refer Appendix E for statistical details). This indicated that the
following variables were significant explanatory variables:
q Average EM31 conductivity;
q Depth to watertable; and
q Upper Kv (permeability of the top 2m of the profile).
However, the standard error of estimate for the regression was very high (82%). In an
attempt to improve the accuracy of the fitted regression model, the sites were split into
two data sets on the basis of one of the key explanatory variables, depth to watertable,
as described below:
The data (with Finley removed) are plotted in Figure 5.45. This plot shows three
potentially ‘outlying sites’ (Donald Main, Lake View West and Lake View Original).
All of these sites are characterised by watertables close to the surface (0.5-2m). The
remaining sites have watertables of between 5 – 10m below surface. Given that the
dominant EM31 response (in vertical mode) is concentrated in the 1-3m range, the
elevated watertable appears to significantly affects the results, and produces two
distinctly different trends. This is made more clear in Figure 5.46, which divides the
data into two categories: watertable less than 2m and watertable 5 – 10m.
The difference in EM31 response between a high and low watertable site is due to the
shallow and saline groundwater. For example, the high seepage sites at Donald
(seepage > 30 mm/d) range in EM31 conductivity between 40-60 mS/m. Based on the
lower trend line (greater than 5 m depth to watertable) it would be expected that ponds
of such high seepage would have extremely low conductivities. However the shallow,
saline groundwater has a strong influence on the EM31 response, which is
concentrated in the 1-3m range, where the groundwater is absent at the sites producing
the lower trend line. Unsaturated conditions will clearly return lower conductivities
than in a saturated environment.
It might be expected that for high seepage sites such as Donald and Lake View West,
the high seepage rates would result in greater volumes of fresh water mixing with the
native groundwater, reducing overall conductivity. While this effect is occurring, it is
apparent that this diluting effect is much less significant than the effect of an elevated
(saline) watertable, and that the average salinity of groundwater in the 1-3m range is
more affected by the native groundwater salinity than the seepage plume.
This effect should be more pronounced at sites with high natural groundwater
salinities, such as at the Donald Main Channel. However, the data does not appear to
indicate a significant difference between the high salinity Donald site (30,000 EC) and
the moderately saline Lake View site (5,000 – 7,000 EC). This is most likely due to
other factors that contribute to the overall conductivity response, such as the more
clayey profile at Lake View.
The deep watertable sites in Figure 5.46 all have groundwater depths between 5-10m
below surface. Note that there were no sites in the study with watertables in the
intermediate (2-5m) range. It is considered likely that sites with a watertable in the 2-
5m range would lie between the two trend lines.

n Figure 5.45 EM 31 Conductivity Versus Pondage Test Seepage at All Sites
(Finley removed)
50
Waranga Western
Donald Main
Toolondo
40 Rocklands
Dahwilly

Lake View (West)
30
20
10
0
0 20 40 60 80 100 120
n Figure 5.46 EM 31 Conductivity Versus Pondage Test Seepage: Categorised

by Depth to Watertable
50
R2 = 0.889
40
30
20
y = -0.129x + 12.788
R2 = 0.469
10
0
0 20 40 60 80 100 120

The key difference between the two categories can be understood when considering
the EM31 focus and depth range. For the 0-2m sites EM31 is directly measuring
seepage in terms of its impact on the watertable zone. In contrast, for the 5-10m
category, the watertable is almost out of range of the EM31, and is therefore largely
measuring unsaturated zone properties. In this case, seepage is not directly being
detected, but rather the likelihood of seepage is being inferred based on unsaturated
zone properties. (Sites with a watertable at 5-6m are just in the range of EM31, but
outside of the focal part of the range).
In summary, it was found that division of the data into two sets based on depth to
watertable improved the regression characteristics of the two separate data sets
compared to the entire data set. Therefore the following discussion is divided into two
categories: watertable less than 2m and watertable 5 to 10m. Once the data was split
into these two categories, depth to watertable was not found to be a significant
explanatory variable, as would be expected.
Watertable Depth Five to Ten Metres

For sites where the watertable is 5-10m below surface, the equation found to provide
the best prediction of channel seepage was:
Seepage = 11.6 – 0.12 EM31 + 4.4 UKv (Equation 1)
Where, Seepage = Channel seepage (mm/d)

EM31 = EM31 conductivity adjacent each side of channel (mS/m)
UKv = Vertical hydraulic conductivity of top 2m of profile (m/day)
The equation was established with 40 data points. The coefficient of determination for
the relationship was R2 = 0.55 and the standard error of estimate was 48% of the
mean observed seepage rate. This standard error indicates that on average estimated
seepage is approximately 50% above or below the actual seepage rate.
Figure 5.47 provides a visual presentation of the degree of scatter in the prediction
equation. The equation appears to over-estimate seepage for low seepage rates (less
than 5 mm/d) and underestimate for high seepage rates.
EM31 was found to be the dominant explanatory variable with soil hydraulic
conductivity of secondary importance. This is probably more a reflection of the
accuracy of the collection technique of the variable, than the actual contribution of the
variable to seepage. If soil permeability could be accurately assessed along the length
of the channel via very high density and accurate testing, soil permeability would
probably be a more significant variable than this analysis indicates. However, the
method of estimating hydraulic conductivity and the density of sampling conducted in
these investigations is typical of RWA priorities. Therefore for the purposes of this
study soil hydraulic conductivity as an explanatory variable of seepage is considered
to be of less importance that EM31 conductivity.
Groundwater salinity and depth to groundwater were not found to be significant

explanatory variables in the analysis. The fact that depth to groundwater was not
significant is not surprising, as the data set had already been divided based on depth to
groundwater, largely removing the influence of this variable on the relationship.

n Figure 5.47 Predicted Versus Observed Seepage for Multiple Linear
Regression, Based on EM31 Conductivity and Upper Soil Profile Hydraulic
Conductivity (Sites with watertable 5- 10m)
14
12
10
Predicted Seepage (mm/d)
0
0 2 4 6 8 10 12 14
Observed Seepage (mm/d)
Watertable Less Than Two Metres

For sites with watertables within two metres of the surface (Lake View, Lake View
West and Donald), multiple linear regression analysis did not find any other variables
that were significant explanatory variables beyond EM31. Therefore multiple linear
regression analysis was not possible. The best fitting equation that can be derived
based on available data is the linear equation with EM31 as the only explanatory
variable. This analysis is presented in section 5.3.4.2.
5.3.4.2 Simple Linear Regression

Simple linear regression, using EM31 only as the explanatory variable, was conducted
to determine how much of an improvement the multiple linear regression actually
represents. As for the multiple regression analysis, the data was divided based on
depth to watertable.

For the data category representing sites with a watertable between 5 – 10m below
surface, established with 40 points (refer Figure 5.46), the best fitting linear regression
equation is:
Seepage = 12.8 – 0.13 EM31 (Equation 2)

This relationship is proposed as the equation on which prediction of the likely seepage
for an average EM31 measurement can be based (for depth to watertable 5-10m). The

coefficient of determination for the relationship was R2 = 0.47 and the standard error
of estimate was 51% of the mean observed seepage rate. This standard error indicates
that on average estimated seepage is approximately 50% above or below the actual
seepage rate.
Figure 5.48 provides prediction bands for estimating seepage based on EM31 response
when the watertable depth is 5-10m. Confidence intervals at 80% and 90%
confidence are presented. For example, for the 80% confidence intervals, these bands
indicate that if EM31 was used to predict seepage rates in a certain stretch of channel
(covering an area of similar EM31 response), we would be 80% certain that the
seepage rate in that section would lie within the confidence bands at that average
EM31 value. The 80% intervals are narrower than the 90% intervals, as there is less
certainty in pinning the value down to its ‘true’ seepage rate. This figure shows that
the prediction equation is accompanied by quite broad prediction intervals.
n Figure 5.48 Regional Prediction Intervals: Predicting Seepage from EM31

(watertable 5-10m)
20
Predicted
Upper_90
Lower_90
Measured
15 Upper_80
Lower_80
Predicted regional
regression line for
10 EM31 - seepage
relationship for
watertable 5-10m
0
10 20 30 40 50 60 70 80 90 100 110
Using this prediction equation, an EM31 survey with an average response (within a
like section of channel) of 30 mS/m, would suggest a seepage value of 10 mm/d. The
prediction bands indicate we would be 80% confident that the real seepage value in
that section would be between 5 – 13 mm/d and 90% confident that that it was
between 4 – 14 mm/d. These bands are quite wide, however they indicate that the
prediction equation can be usefully applied to broadly classify seepage rates (eg into
low, medium and high categories).
Figure 5.49 graphs the actual seepage versus the predicted seepage (using the
regression equation) and provides a visual presentation of the degree of scatter in the
prediction equation. The equation appears to over-estimate seepage for low seepage

rates (less than 5 mm/d) and under-estimate for high seepage rates, implying a non-
linear equation may provide a better fit. An exponential regression equation was
applied to the data and is shown along with the linear regression line in Figure 5-50.
The predicted versus observed seepage for the exponential fit is shown in Figure 5.54.
The statistics for these equations are shown in the figure. They reveal that while there
is a only a marginal improvement in the correlation coefficient for the exponential fit
(from R2 = 0.47 to 0.48), the standard error actual worsens (from 51 % to 62%).
Therefore, overall a less accurate is obtained using the non-linear equation, even
though it may visually appear to fit the data better.
n Figure 5.49 Predicted Versus Observed Seepage for Linear Regression

(watertable 5 –10 m)
14
12
10
0
0 2 4 6 8 10 12 14

n Figure 5-50 Linear Compared to Exponential Prediction Equations
(watertable 5-10m)
Pondage Test Seepage vs EM31 (Sites with WT: 5-10m)
15
Linear
y = -0.13x + 12.79
R2 = 0.47
SE = 51% of mean
10
Exponential
-0.036x
y = 27.372e
2
R = 0.476
SE = 62% of mean
0
0 20 40 60 80 100 120
n Figure 5-51 Predicted Versus Observed for Exponential Regression

(watertable 5 – 10m)
14
12
10
0
0 2 4 6 8 10 12 14

However there are several advantages to the exponential fit over the linear fit. The
first is that there is less of a pattern displayed in the observed versus predicted seepage
plots (ie Figure 5-51 compared to Figure 5.49), which suggests that the more realistic
model may in fact be the exponential model. The second is that the linear model
places a maximum limit on the seepage of about 12 mm/d, whereas the exponential
model appears to be more realistic, allowing for higher seepage rates in the very low
conductivity range (up to around 20-25 mm/d).
Comparing the linear regression to the multiple regression, the statistics indicate that
only a marginal improvement is made to the accuracy of the regression fit in the
multiple linear regression analysis (Equation 1), compared to the simple regression fit
(Equation 2). The R2 for Equation 1 was 0.55 and the standard error of estimate was
48%. Therefore a relatively modest improvement of 0.08 in the correlation coefficient
and 3% in the standard error of estimate is the only improvement gained in adding soil
permeability to the equation. The predicted versus observed seepage plots for the two
equations are also very similar.

The best fitting linear equation for sites with a watertable less than 2m (based on only
14 data points) is:
Seepage = 72 – 0.69 EM31 (Equation 3)

The coefficient of determination for the relationship was R2 = 0.89 and the standard
error of estimate was 23% of the mean observed seepage rate. This standard error
indicates that on average observed seepage was approximately 23% above or below
the predicted seepage rate. Figure 5.52 plots actual seepage versus predicted seepage
(using the regression equation) and provides a visual presentation of the degree of
scatter in the prediction equation.
The high R2 value and the relatively low standard estimate of error suggest a good
correlation for the variables. However the results should be tempered by the fact that
relatively few data points have been used to form the relationship. This is best
illustrated by the example of a data set comprising only two points. This will return a
correlation coefficient of 1 and a standard estimate of error of 0, but obviously there
would be little confidence in the accuracy of the resulting regression equation. To
improve confidence in the regression equation for the watertable less than two metres
scenario, additional points are required in the data set.
Figure 5.53 presents prediction bands (80% and 90%) for estimating seepage based on
EM31 response when the watertable is less than 2m, based on Equation 3. While the
prediction intervals are broader in magnitude (approximately 20 mS/m and 15 mS/m
for the 90% and 80% intervals respectively) than for the prediction bands for the
deeper watertable scenario, as a percentage of the overall seepage range covered by
each of the equations, they are narrower.

n Figure 5.52 Predicted Versus Observed Seepage for Linear Regression
(sites with watertable less than 2m)
60
50
40
Predicted Seepage
30
20
10
0
0 10 20 30 40 50 60
Observed Seepage
n Figure 5.53 Prediction Intervals: Predicting Seepage from EM31 (watertable

less than 2m)
70
Predicted
Upper_90
60 Lower_90
Measured
Upper_80
50
Lower_80
Predicted regional
40 regression line for
EM31 - seepage
relationship for
30 watertable < 2m
20
10
0
30 40 50 60 70 80 90 100 110

5.3.5 Channel Specific EM31 Statistical Assessment and Extrapolation
of EM31 Relationships
5.3.5.1 Introduction
In contrast to the regional assessment, the aim of examining the EM31 - seepage
relationship at each channel was to find if better relationships existed between EM31
and seepage at a local level. In particular, for those channels specifically targeted in
the year 3 program to conduct pondage tests and geophysical surveys in similar and
dissimilar environments, this section discusses the success of the extrapolation from
the ‘original’ site (refer section 5.1.2.2 for discussion of methodology). The channels
where this was conducted included:
q Toolondo;
q Dahwilly; and,
q Lake View.
Note that the Rocklands site was included in the Toolondo analysis as the site is
relatively close to the Toolondo channel and site characteristics are similar.
5.3.5.2 Toolondo and Rocklands

Figure 5.54 presents EM31 results against pondage test seepage for the Rocklands and
Toolondo Channels. The Toolondo results are comprised of three sites as depicted in
the figure: Toolondo Central (original site), Toolondo East and Toolondo West.
The EM31 surveys were conducted at these sites in March 2002.
n Figure 5.54 Toolondo and Rocklands Pondage Test Seepage Versus EM31
Conductivity
Toolondo and Rocklands Channels

20
Predicted
18 Upper_90
Predicted regression
Lower_90
line for 4 sites
16 Measured
y = -0.16x + 14.22
2 Upper_80
R = 0.47
14 Lower_80
Rocklands
12 Toolondo West
Rocklands Toolondo East
10 2
R = 0.33 Toolondo Central
8 Toolondo Central
2
R = 0.78
6
4 Toolondo West
R2 = 0.85
2
Toolondo East
2
R = 0.49
0
25 35 45 55 65 75 85 95 105

The Toolondo Central site is the original site where pondage tests and EM31 surveys
were conducted in years 1 and 2 (refer section 5.3.3.1). The Toolondo East and West
sites are several kilometres from the Central (original) sites and represent the sites
which were selected to be different and similar to the original site. The Toolondo
West site is similar to the Central site, while the East site differs in terms of lithology,
with very heavy clay to approximately 3.5 - 4m depth. Rocklands data collected in
August 2001 was included in the analysis as the site is relatively close to the Toolondo
channel, and the site characteristics are reasonably similar to the Toolondo Central
site.
Figure 5.55, Figure 5.56, and Figure 5.57 present the EM31 results at each of the
Toolondo sites. (Figure 5-25 presents the results at the Rocklands sites). Note that
different scales are used in each of these figures for presenting the EM31 conductivity.
A reasonable correlation coefficient for the fitted regression line of 0.47 was obtained.
The standard estimate of error for the data is moderate to poor, at 61%. Confidence
bands are wide but indicate the equation can differentiate (at 80% confidence)
between high and low seepage sites.
The data decreasing the correlation coefficient and increasing the width of the
confidence bands are the four ponds of high seepage at the Toolondo Central site,
which contain relatively very high proportions of fine to medium grained sand in the
upper 1-2m of the profile (as shown in the Toolondo Geological Long Section in
Appendix A). It is apparent that while the sand is causing relatively high seepage
rates, the EM31 conductivity is not as low as expected adjacent these ponds compared
to the remainder of the channel. A potential cause of this is that the shallow depth of
the sand layer through which most of this seepage occurs (via channel walls) is not in
the primary focus range of the EM31 in vertical dipole mode, and therefore does not
influence the results as much as a sand layer immediately below the channel base (eg
at 2-3 metres depth).
Extrapolation from Toolondo Central Site

The regression line based on the Toolondo Central ponds is presented in Figure 5.54.
This trend line has almost double the gradient of the combined Rocklands/Toolondo
sites regression line. If the Toolondo Central regression relationship had been used to
predict seepage at:
q Toolondo West (‘similar’ site): Predicted seepage rates would have been 2-3
times greater than actual rates due to the steeper Toolondo Central regression line
q Toolondo East (‘different’ site): Predicted seepage rates would have been
accurate (essentially predicting 0 mm/d) except at Pond 2, where predicted
seepage would have been approximately 4 mm/d (actual 0.4 mm/d). However the
Rocklands/Toolondo trend line would also over predict the rate at this pond at
approximately 3 mm/d.
Regression lines were fitted through each of the sites individually (Toolondo Central
line only is plotted in Figure 5.54). The correlation coefficients for each of the sites
when plotted individually are:
q Toolondo Central: 0.78
q Toolondo West: 0.86
q Toolondo East: NA – slight trend in reverse direction
q Rocklands: 0.33 (adjacent the channel: 0.82)

These results concur with soil types at the sites. The Toolondo East site was selected
due to its change in soil type from a sandy clay to a very heavy clay profile and no
trend was discernible within this site. However, as a data group, the site still fits
within the broader regression trend, even though at a site specific level the trend line is
reversed (the slope of the reverse trend line is almost flat compared to slopes of the
other Toolondo and Rocklands sites, and therefore not highly significant).
These results, along with examination of the channel specific data points in Figure
5.54 suggest that within the individual pondage test sites, better correlations can be
obtained, with R2 values of around 0.8 at Toolondo Central and West.
As would be suspected, this analysis shows that local correlations are more accurate
for local predictions. However the regression line (‘Predicted’) presented in Figure
5.54 is a more useful tool for water managers as it allows prediction over longer
sections of channel and across varying conditions.

n Figure 5.55 Toolondo Central EM31 Survey, March 2002
n Figure 5.56 Toolondo East EM31 Survey, March 2002

n Figure 5.57 Toolondo West EM31 Survey, March 2002

5.3.5.3 Donald
Figure 5.58 presents EM31 results (conducted September 2001) against pondage test
seepage at the Donald Main channel. This survey is reported on in detail in Section
5.3.3.3 and the results are graphically presented in this section in Figure 5-31. This
site is now briefly re-visited to apply a more rigorous statistical evaluation.
A correlation coefficient of 0.73 was obtained and the standard estimate of error was
also good at 23% of the mean. Confidence bands are wide but indicate the equation
can differentiate between high and low seepage sites. For example, if the EM31
survey result is 45 mS/m, with 80% confidence it can be concluded that seepage is
between 30-55 mm/d. While this range is still wide, it indicates with a reasonable
degree of confidence that seepage at the site is high. The bands widen at the higher
conductivity range, which is a reflection of the small number of data points
contributing to the construction of the regression equation. If a 75 mS/m (or greater)
conductivity is recorded, it can be concluded with 80% confidence that seepage is less
than 30mm/d.
n Figure 5.58 Donald Main Channel Pondage Test Seepage Versus EM31
Conductivity
Donald Main Channel
80
Predicted
70 Upper_90
y = -0.90x + 82.46
Lower_90
R2 = 0.73
60 Measured
Upper_80
50 Lower_80
40
30
20
10
0
30 40 50 60 70 80 90

5.3.5.4 Dahwilly Channel
Figure 5.59 presents EM31 results against pondage test seepage at the Dahwilly
channel, comprising surveys at the Dahwilly Central and Dahwilly East (Pretty Pine)
sites. The two sites are approximately four to five kilometres apart. The Dahwilly
East site represents the site which was selected with similarities to the original site.
Pondage tests at the ‘different’ sites selected along the Dahwilly channel were not
conducted due to operational constraints, and thus analysis could not be undertaken.
The geology at the Dahwilly Central and Dahwilly East sites are reasonably similar
(refer Appendix A for geological sections of both sites). Figure 5.60 and Figure 5.61
show maps of the EM31 results at Dahwilly and Dahwilly East (Pretty Pine). The
EM31 surveys were conducted in February 2002.
It should be noted that the Dahwilly Central site examined here is slightly different in
location to the Dahwilly site referred to earlier in the report (refer section 5.3.3.4).
The pondage tests conducted here (2002) bordered the six pondage tests conducted in
2001. In fact, ponds 1 and 2 from the June 2001 pondage tests were pond 6 and 7 in
the June 2002 pondage tests. [Theses two cells were remediated in July 2001. The
two lined cells were removed from the analysis however, due to insufficient time
between the pond lining and the EM31 survey (February 2002)]. It is important to
note the significant difference between these results and the June 2000 EM31 survey.
In the 2000 survey, no trend was detected in the data as the channel was not running,
and therefore there was no seepage impacts on the watertable to detect. Further, the
unsaturated zone at this site is relatively uniform and therefore unlike at other sites
seepage cannot be inferred from unsaturated zone properties. At this site seepage is
controlled by a clogging layer in the channel and not the unsaturated zone.
n Figure 5.59 Dahwilly Pondage Test Seepage Versus EM31 Conductivity
Dahwilly Channel
Predicted
20
Trend line based on Upper_90
both Dahwilly sites
18 ly = -0.12x + 12.64 Lower_90
Dahwilly Central
2
R = 0.58 trend line Measured
16
y = -0.49x + 31.42
Upper_80
2
R = 0.64
14 Lower_80
12 Dahwilly East
Dahwilly Central
10
Li (D h ill
8
6
y = -0.05x + 11.03
R2 = 0.15
4
Dahwilly East
trend line
2
0
10 20 30 40 50 60

n Figure 5.60 Dahwilly Central EM31 Survey, February 2002
n Figure 5.61 Dahwillly East (Pretty Pine) EM31 Survey, March 2002

The fitted regression line for the Dahwilly site had a correlation coefficient of 0.58
and a relatively low standard estimate of error of 21% of the mean. The 80%
prediction interval bands are 5-6 mm/d wide, which are quite wide as a percentage of
the range covered by the regression equation. This suggests the equation can only be
confidently used in a broad sense to indicate the likelihood of low (< 5 mm/d) or
moderate seepage (> 5 mm/d).
Extrapolation from Dahwilly Central Site

The regression line fitted through the Dahwilly central site (plotted in Figure 5.59 as a
lighter colour line) has a slightly improved correlation coefficient compared to the line
for both sites, but most notably is much steeper than the overall trend line. (A trend
line was not fitted through the Dahwilly East site as there are insufficient points for
this to be meaningful). If the Dahwilly Central trend line had been used to predict
seepage at the Dahwilly East site, predicted seepage rates would have been 1.5 to 2
times greater than the actual rates recorded in the pondage tests.
The cause of the lower conductivity response at the Dahwilly East site (and thus the
reason that the Dahwilly East EM31 conductivities do not plot on the Dahwilly
Central trend line) compared to the Dahwilly Central site may be due to the fresher
salinity native groundwater (approximately 2,500 EC compared to the 5,000 EC).
However a more likely explanation is that the watertable is probably slightly deeper at
this site, which would effectively remove the watertable from the penetration depth of
the EM31 and lower conductivities. The resistivity survey conducted at Dahwilly East
(refer section 5.4.4) suggests the watertable is around 6-7m. Bores at the Dahwilly
central site indicate the watertable is approximately 5m below surface.
This analysis confirms what was illustrated from the extrapolations at the Toolondo
site: local correlations will generally produce more accurate predictions. Even though
the Dahwilly East site is only 4-5 km from the Central site, the slightly deeper
watertable (and possibly change in groundwater salinity) is sufficient to produce a
different trend line. However the regional prediction equation is a more useful tool as
it allows prediction over longer sections of channel and across varying conditions.
5.3.5.5 Finley
Figure 5.62 presents pondage test seepage against EM31 conductivity for the Finley
channel. The fact that only three ponds were used to establish the regression line
means that the correlation coefficient (0.47) and the standard estimate of error (18% of
the mean) are essentially meaningless. It is a reminder of the danger of relying on
statistics outside of an understanding of the data and the context in which they are
used. The prediction interval bands reveal the inadequacies of the data to produce
any meaningful conclusions, with intervals of 14 mm across the displayed range (300
– 500 mS/m). Additional data points are required to improve the prediction interval.
The Finley site is an extremely clayey site (refer to the Finley Geological Long
Section in Appendix A), with a high and salty watertable which produces the very high
EM31 conductivities recorded at this site. There were no other sites in the
investigation approaching conductivities of this magnitude. Given the moderate
seepage rates and yet the very clayey profile, it was concluded in the regional
assessment that the probable mechanism of seepage at this site is not vertical seepage
through the base of the channel (ie due to in-situ soil permeability) but lateral seepage
through the channel banks due to poor bank construction techniques or materials.

n Figure 5.62 Finley Pondage Test Seepage Versus EM31 Conductivity
Finley
16
Predicted
14 Upper_90
Lower_90
12
Measured
Upper_80
10
Lower_80
y = -0.008x + 9.146
8
R2 = 0.467
0
300 320 340 360 380 400 420 440 460 480 500
n Figure 5.63 Finley EM31 Survey, February 2002
Figure 5.63 presents a map of the EM31 results at the Finley site. Notwithstanding the
discussion above, (that the data cannot be used to make meaningful statistical
conclusions), this figure does highlight that the fact that the technique appears to have
detected an area of relatively higher seepage in the eastern half of pond one, which is
supported by the fact that this pond has the highest seepage rate.

5.3.5.6 Lake View
Figure 5.64 presents pondage test seepage plotted against EM31 conductivity at the
Lake View channel, for both the Lake View Original and the Lake View West site.
Figure 5.65 and Figure 5.66 show maps of the EM31 results at the two sites. The
Lake View central EM31 data was collected in June 2000, whereas the Lake View
West data was collected in May 2002. In both cases data collected from pondage tests
conducted in 2002 are used in the analysis.
A high correlation coefficient of 0.94 was obtained for the relationship between EM31
and seepage between the two sites and the standard estimate of error was also good, at
17% of the mean. Confidence bands are moderately wide at about 8 mm/d (at 80%
confidence level) and 10 mm/d (90% confidence level). However they suggest that the
equation is suitable for differentiating between low, medium and high seepage rate
sites.
Some warning is necessary however regarding the spread of the data. Both of the sites
are clustered around a relatively small seepage rate and average conductivity range –
the danger of this is best illustrated again by the example that only two data points will
return a correlation coefficient of 1, but obviously there would be little confidence in
the accuracy of the resulting regression equation. To improve confidence in the
regression equation at this site, data is required in the mid conductivity range.
n Figure 5.64 Lake View Pondage Test Seepage Versus EM31 Conductivity
Lake View Channel

30
y = -0.64x + 68.69 Predicted
2
R = 0.54
Regression line for Upper_90
25 Lake View West Lower_90
Measured
20
Upper_80
Lower 80
15
Regression line
for both sites
y = -0.65x + 68.70 y = -0.09x + 16.88

10 2 2
R = 0.94 R = 0.34
Regression line
5 for Lake View
Central
y = 0.24x - 17.42
2
R = 0.46
0
60 65 70 75 80 85 90 95 100 105 110

n Figure 5.65 Lake View Central EM31 Survey, June 2000
Extrapolation from Lake View Central Site

The Lake View West site was selected to test the ability to extrapolate from one site to
another. In Figure 5.64 the line of best fit has been added through the Lake View
Central data. This regression line actually has the opposite gradient to that
theoretically expected (and to that observed in most of the sites examined in this
study). This illustrates that local trends cannot necessarily be assumed to be reliable,
particularly when the range of the data is quite small. In this case the pondage test
range was only 4.3 – 7.1 mm/d, and the (average) conductivity range very small at 96

- 100 mS/m. However, it should be noted that good correlations were obtained using
this same EM31 data but the old pond locations (refer section 5.3.3.5). In particular
improved correlations were obtained using the survey line immediately adjacent the
channel, rather than all survey lines, due to the mixing that occurred away from the
channel. This suggests that improved results could be obtained with the new pondage
test locations if the adjacent channel data was used in this analysis. However, for
consistency purposes (ie, for comparison with other channels) all the data was used.
n Figure 5.66 Lake View West EM31 Survey, May 2002

5.3.5.7 Summary
Table 5-6 summarises the main conclusions and findings of the EM31 analysis of each
channel.
n Table 5-6 Summary of Channel Specific EM31 Surveys to Pondage Test

Relationships
Region Number Correlation Standard
(channels of coefficient Estimate of Comment
comprising Ponds (R2) Error (as %
region) of mean)
Rocklands / 21 0.47 63% q Improved correlations are obtained at each of the
Toolondo individual sites (R2 approx 0.7-0.8)
(Rocklands, q Using the Central site to predict seepage at the West
Toolondo East, site would have caused over prediction of 2-3 times
Central & West) actual seepage rate (due to high surface sands content
at Toolondo central)
q Prediction interval range of 8 mm/d (80% confidence)
fairly wide but considered acceptable given data spans
4 diff. Sites
Donald 6 0.73 23% q Good R2 & SEE but based on small no. data points,
especially at high cond. / low seepage end of range
q Prediction interval range of 25 mm/d (80% confidence)
but still useful for low risk – high risk assessment.
Additional data points likely to tighten confidence
intervals
Waranga 11 0.40 61% q R2 improved to 0.60 when two outliers removed
q High SEE reflects high scatter and fairly broad
confidence intervals (8mm/d at 80%). Typical of what
would be expected given the spatial range of the
pondage tests
q Useful for high / low risk seepage assessment
Dahwilly 8 0.58 21% q Reasonable R2, SEE
(Dahwilly Central q Prediction intervals relatively wide (5-6 mm)
& East) q Using the central site to predict seepage at the east site
would have caused over prediction of 1.5 -2 times actual
seepage, due to differences in gw salinities (and
potentially depth to watertable) between the two sites
Finley 3 0.47 18% q Statistics not meaningful due to low no. data points
q Confidence intervals indicate this relationship cannot be
used to predict seepage at site
q Extremely high conductivity site – very clayey profile
suggests bank leakage rather than soil seepage is the
dominant seepage mechanism
Lake View 8 0.94 17% q Good R2, SEE & prediction interval but 2 data sets
(Lake View creating the regression line have small conductivity and
Central & East) seepage rate ranges – Desirable to obtain some data in
mid range to improve confidence
q No sensible trend at the Lake View original site – could
not have been used to predict Lake View West seepage.
However improvement likely if adjacent channel only
data used (due to rapid mixing), as per yr 1 & 2 analysis
From examination of the above table it is apparent that the more data points that are
added to generate the regression line, the greater the scatter about the regression line.
This can be seen in the high standard error of estimates for the Rocklands/Toolondo
and Waranga sites. This is a reflection of the fact that the ponds in these sites covered
a wide area and range of sub-surface conditions. However, while there is more scatter
about these lines, these relationships are probably more useful to RWAs as they
encompass a wider spatial range and wider range of conditions, and can therefore be
used with greater confidence over a broader area.

5.4 Resistivity Trials
5.4.1 Introduction
This section is divided into the following sub-sections:
q Background – This section describes the functioning of resistivity systems as well
as the specific methodology as to how they were used in the trials;
q Regional Assessment of Key Relationships – For all of the year three sites an
attempt was made to look for potential correlations between seepage rates across
all sites and resistivity response. This was conducted applying multiple linear
regression using a number of key explanatory variables and simple linear
regression, where resistivity was the only explanatory variable.
q Channel Specific Resistivity Assessment - This section examines the resistivity -
seepage relationship at each channel. Variations with resistivity and depth are
discussed and various combinations of seepage relationships explored for
different depth slices within the profile.
5.4.2 Background
5.4.2.1 Channel Specific Resistivity Systems
Resistivity can be measured using grounded (or immersed) current electrodes to
impress an applied voltage across a section of the ground. Differences in voltage
distribution can be used to calculate apparent ground resistivity. The method depends
on good electrode connection and hence can be slow where extensive electrode
preparation is necessary. Thus rates of acquisition are usually only around 5 km/
depending upon conditions. The exception to this is when the electrodes can be
immersed in water thus overcoming the need for electrode preparation. Systems also
can be linked with a recording device and GPS positioning for rapid survey procedure.
In such circumstances continuous recording can be achieved at rates of greater than 5
kph or 40km/d. Hotchkiss et.al. (2001) employed such a device for measuring
seepage from irrigation channels in Nebraska, USA. Similar devices are commonly
used down bore holes to measure formation resistivity.
The advantage of resistivity systems is that a single transmitting dipole can be used
with a number of receiving dipoles. These dipoles positioned at increasing distance
from the transmitting electrodes can be used to calculate the depth and conductivity
relationships of the sub-surface. This allows a conductivity profile to be established,
as opposed to conventional frequency domain EM systems which provide only a
single average conductivity for the profile.
5.4.2.2 Methodology
For the year 3 trials, a multi-electrode array was built in what is commonly referred to
as a dipole-dipole configuration. A pair of current electrodes separated by a distance x
are followed by a series of receiver electrodes all separated by the same distance. The
closest receiver electrodes sample the resistivity in the near surface (around one third
to half x) and the more distant electrodes ‘see’ deeper into the ground. Using an array
of receiver dipoles allows the possibility for a resistivity section to be created.
Such arrays have been in common use in mineral exploration for the past 50 years.
However data is normally collected while the whole array is stationary, partly because
good contact between electrodes and the ground must be obtained. This may be

difficult on dry land. In the present study however the array was immersed (floating)
on the channel. Contact of electrodes in the water and thus to the underlying ground
was good and the array could be towed at speeds of around 5 to 8 kph while data was
collected. This allowed data collection of 2 km sections in around 20 minutes. Figure
5.67 shows the resistivity array being used at the Toolondo East site. All resistivity
surveys were conducted in mid-March 2002.
In these trials the sites were surveyed in all cases with a five metre dipole array of 6
dipoles. The same section of channel was then re-surveyed with a 10m dipole array in
all sites except Toolondo Central which was the first site and Toolondo West where
water levels were very low. Time did not allow analysis of the 10m dipole data in this
study.
n Figure 5.67 Resistivity Array Deployed in Year 3 Trials (in operation at the
Toolondo East site)

5.4.3 Regional Assessment of Key Relationships
As was conducted in Section 5.3.4 for the EM31 analysis, this section explores
potential correlations between seepage rates across all sites and resistivity response, at
sites where resistivity surveying was conducted (refer section 5.1.3). In comparison to
the EM31 assessment, this excluded Rocklands, Donald and Waranga channels from
the analysis, and reduced the data set to 23 ponds, across eight sites. This regional
assessment was conducted in two ways:
q Multiple linear regression, ie looking for correlation between seepage rate and
average resistivity response and other important variables which might affect
seepage (or the resistivity response); and,
q Simple linear regression, ie looking for a direct correlation between seepage rate
and average resistivity response.
5.4.3.1 Multiple Linear Regression

Introduction
As described in Section 5.3.4, multiple linear regression is a statistical technique that
allows one response (or dependent) variable to be predicted from a number of
explanatory (or independent) variables. In this study the response variable is seepage,
as measured in pondage tests. The explanatory variables which were included in the
multiple linear regression analysis included:
q Resistivity;
q Soil permeability;
The most important independent variable is resistivity. The main aim of conducting
multiple regression analysis was therefore to determine which other variables could
account for variations in seepage apart from resistivity. Groundwater salinity and
depth to watertable were included as independent variables because they will affect
the resistivity response.
Data Sources
The response variable is observed seepage as measured in pondage tests (averaged
over the life of pondage tests, excluding outliers). The explanatory variables of soil
permeability, depth to watertable and groundwater salinity and their method of
collection is described in Section 5.3.4.1. Resistivity is considered to be the most
reliable data source, as spatially this contains the highest density of readings over the
pond length. Resistivity was measured using the on-channel set-up described in
section 5.4.2 and averaged over the length of the pond.
A further consideration with the resistivity analysis compared to the EM analysis is

that data is collected at a number of depths within the profile for the resistivity survey.
The whole profile average may not be the most appropriate or the best fitting data to
use in the regression analysis. Therefore the most appropriate resistivity depth slice
must be selected. Figure 5.68 shows seepage rates versus average resistivity at
various depths for sites with watertable between 5 – 10 metres (Toolondo and
Dahwilly). This figure shows the progressively improving correlation coefficient
obtained for the deeper sections within the profile, with 10 m returning the strongest

correlation. The 8 m average was also relatively strong, but the 12 m relationship is
weaker, indicating that around 10 m appears to be the optimum measurement depth for
this data set. Based on this analysis, the linear and multiple linear regression analysis
uses the ten metre section resistivity data for regression modelling.
Using the one depth slice across all sites has limitations. As the channel specific
resistivity assessment indicates (refer section 5.4.4), often the depth interval
correlating best with seepage is at and immediately below the watertable. Therefore
this approach may be appropriate for sites with a watertable in the 5 - 10m range, but
may not be best for sites with a shallower watertable. However, for the purpose of
applying a consistent approach a single depth slice was used across all sites.
n Figure 5.68 Pondage Test Seepage Versus Resistivity Average For Various
Depths at Toolondo and Dahwilly Channels
Resistivity vs Seepage (Toolondo and Dahwilly)
14
12
y = 0.011x + 4.224 y = 0.005x + 4.696
2
R = 0.137 y = 0.010x + 4.193
R2 = 0.087
2
R = 0.160
10 y = 0.007x + 4.335
R2 = 0.150 y = 0.004x + 4.651 y = 0.002x + 4.919
2
R2 = 0.102 R = 0.064
8
At 8m
At 6m
6 At 10m
All
At 4m
4 At 12m
Linear (At 8m)
Linear (At 6m)
Linear (At 10m)
2 Linear (All)
Linear (At 4m)
Linear (At 12m)
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Resistivity (ohm m)
Methodology
The methodology used in the analysis was the same as that undertaken for the EM31
multiple regression analysis and is also described in Section 5.3.4.1. Two measures
were used to assess the accuracy, or ‘goodness’ of the regression. These were the
coefficient of determination (R²) and the standard error (SEE). The R² measures the
proportion of the total variation explained by the model. A large R² is associated with
a good model or prediction equation. The standard error is a measure of the degree of
scatter of the observed data points around the regression line. Hence a small standard
error is associated with a good model. The standard error has been expressed as a
percentage of the mean of the observed dependent variable (ie pondage test seepage).

Results and Discussion
Figure 5.69 presents the resistivity at 10m depth within the profile at each pondage
cell versus pondage test seepage. At first glance there does not appear to be any
relationship between the two variables, with the Dahwilly channel returning
resistivities many times higher than all other sites. The first cut statistical analysis
indicated that depth to watertable was a significant variable across the sites and
therefore to improve correlations and allow a more meaningful regional analysis to be
conducted, the data was divided based on watertable depth (as was conducted for the
EM31 analysis).
n Figure 5.69 Resistivity (Whole Profile Average) Versus Pondage Test

Seepage at All Sites
Resistivity vs Seepage at All Sites
30
Toolondo
Dahwilly
25 Finley
Lake View (West)

20
15
10
0
0 100 200 300 400 500 600 700 800 900 1000
Resistivity ohm m (average across profile)

For sites where the watertable was 5 to 10 m below surface (Toolondo and Dahwilly),
resistivity and the hydraulic conductivity of the upper two metres of the soil profile
were found to be significant variables. Groundwater salinity was not found to be
significant. Depth to watertable was also found not to be significant, however this was
expected as the original data set was split into two based on depth to watertable. The
regression equation found to provide the best prediction of channel seepage was:
Seepage = 3 + 0.01 Resistivity10m + 7.46 UKv (Equation 4)

Resistivity10m = Average resistivity at 10m depth recorded on channel (ohm m)

mean observed seepage rate.
Various transforms were examined to improve the accuracy of the regression. It was
found that raising the seepage to the power of 0.2 improved the model with respect to
the standard error of estimate, which was reduced to 19% of the mean observed
seepage rate (to the power of 0.2). A marginal reduction in the correlation coefficient
was observed, decreasing to R2 = 0.42.
The non-linear multiple regression equation, for predicting channel seepage at sites
where the watertable is between 5 - 10m is:
5
Seepage = [ 1.12 + 0.000825 Resistivity10m + 0.47 UKv ] (Equation 5)

Resistivity10m = Resistivity at 10m depth recorded on channel (ohm m)
Figure 5.70 provides a visual presentation of the degree of scatter in the prediction
equation (Equation 5).
n Figure 5.70 Predicted Versus Observed Seepage for Multiple Non-Linear

Regression, Based on Resistivity and Upper Soil Profile Hydraulic
Conductivity (Sites with Watertable 5-10m)
14
12
10
0
0 2 4 6 8 10 12 14

For sites with watertables within two metres of the surface (Lake View and Lake View
West), multiple linear regression analysis did not find any other variables that were
significant explanatory variables beyond the resistivity data. This was also found to
be the case for the EM31 regional analysis. It is logical that soils would be a
significant variable for the 5-10m watertable sites, but not for the less than 2m
watertable sites, as at the shallow groundwater sites the groundwater will tend to mask
the impact of soil type on the response. Therefore multiple linear regression analysis
was not possible. The best fitting equation that can be derived based on the available
data is presented below in Section 5.4.3.2 - Simple Linear Regression.
5.4.3.2 Simple Linear Regression

Watertable Five to Ten Metres
Figure 5.71 graphs pondage test seepage with resistivity (to 10 m) for sites where the
watertable is 5 to 10 metres below surface. For the resistivity analysis the number of
channels fitting this category was reduced to two (3 sites at Toolondo and 2 sites on
the Dahwilly channel). Four points within this data set appear to be outliers. They are
the four high seepage ponds at Dahwilly Central. However they were not be removed
from the analysis as there was no obvious grounds for their removal.
Therefore the linear regression equation (using resistivity only as an explanatory

variable) predicting channel seepage at sites where the watertable is greater than 2m
is:
Seepage = 4.2 + 0.01 Resistivity10m (Equation 6)

mean observed seepage rate. These statistics indicate that the accuracy of the
regression is very poor, in large part due to the four high seepage rate ‘outliers’ at the
Toolondo Central site. With these outliers excluded the correlation coefficient
improves dramatically to R2 = 0.63. However as discussed above there was no
obvious basis for their removal, and thus this equation is not presented.
seepage rate (to the power of 0.2). A marginal improvement in the correlation
coefficient was observed, increasing to R2 = 0.21.

n Figure 5.71 Pondage Test Seepage Versus Resistivity (av at 10m) for Sites
with Depth to Watertable 5 – 10m
Resistivity (10m) vs Seepage at Toolondo & Dahwilly (Sites with DTWT > 5-10m)
14
12
y = 0.010x + 4.193
R2 = 0.160
10
6
Toolondo
4
Dahwilly
2
Sites with DTWT
5-10m
0
0 100 200 300 400 500 600
Resistivity (at 10m) ohm m
The non-linear regression equation, using resistivity only as an explanatory variable,

for predicting channel seepage at sites where the watertable is 5-10m below surface is:
5
Seepage = [ 1.19 + 0.0008 Resistivity10m] (Equation 7)

Figure 5-72 presents actual seepage rates versus predicted seepage rates for the
resistivity prediction equations, both for the linear prediction equation and the non-
linear equation. This figure shows that the standard estimate of error has been
improved for the non-linear fit by reducing the predicted seepage at the lower end of
the seepage range. However, the non-linear prediction equation increases the error of
prediction at the higher end of the seepage range, causing the four outliers to increase
in distance from the prediction line. In summary, neither the linear or non-linear
simple regression equations are satisfactory predictors of seepage. The very wide
prediction bands for the non-linear prediction equation (Equation 7) presented in
Figure 5.73 confirms this conclusion.

n Figure 5.72 Predicted vs Observed Seepage for Simple Linear and Non-
Linear Regression (Sites with watertable > 2m: Toolondo and Dahwilly)
14
Non-Linear Predicted Eqn
12
Linear Prediction Eqn
10
0
0 2 4 6 8 10 12 14
n Figure 5.73 Prediction Intervals: Predicting Seepage from Resistivity

(average at 10m) For Watertable 5 - 10m
2.5
2
Predicted Seepage0.2 (mm/d)
1.5
1
Predicted
Upper_90
Lower_90
0.5
Measured
Upper_80
Lower_80
0
0 100 200 300 400 500
Resistivity (at 10m), (ohm m)

Figure 5.74 presents pondage test seepage versus average resistivity (at ten metres) for
sites where the watertable is less than two metres. For the resistivity analysis this data
set is comprised of only the Lake View channel (four ponds at the original site and
four ponds at the western site). Finley was removed from the data as an outlier, based
on the reasoning described in section 5.3.4.2.
The linear equation (using resistivity only as an explanatory variable) predicting

channel seepage at sites where the watertable is less than 2m is:
Seepage = 1.7 Resistivity10m - 0.66 (Equation 8)

the relationship was R2 = 0.62 and the standard error of estimate was 6.3, which is
27% of the mean observed seepage rate. These statistics indicate that the accuracy of
the regression is reasonable. However, the results must be interpreted based on the
fact that only a relatively few number of data points have been used to form the
relationship, and all data points were collected on the same channel. Further testing to
add different environments to this data set is necessary before a reasonable degree of
confidence can be placed in this prediction equation (outside of the immediate area
where the relationship was established).
n Figure 5.74 Pondage Test Seepage Versus Resistivity (at 10m) for Sites
with Depth to Watertable Less than Two Metres (Lake View Channel)
Resistivity vs Seepage at Lake View Original and Lake View West (DTWT < 2m)
30
25
20
y = 1.7129x - 0.6654
2
R = 0.619
15
10
0
0 2 4 6 8 10 12 14 16 18 20
Resistivity ohm m (average at 10m)
Figure 5.75 presents prediction intervals for Equation 8. The large width of the
prediction bands is caused by the small number of data points which have been used to
generate the regression equation. For example, for a resistivity reading of 8 ohm m, at
90% confidence levels the actual seepage rate could be anywhere in the range of 0 –
25 mm/d, which is not a very helpful guide. The width of these bands does not give a

great deal of confidence in using this equation for prediction, reinforcing the above
comments regarding the need for additional data points to improve confidence in the
prediction equation. Again it should be noted that the 10m depth slice is probably not
the most appropriate depth focus for sites with a shallow watertable. This is also
contributing to the uncertainty in the regression equation.
n Figure 5.75 Prediction Intervals: Predicting Seepage from Resistivity

(average to 10m) For Watertable Less than 2m (Lake View Channel)
30
25
20
15
Predicted
10
Upper_90
Lower_90
5 Measured
Upper_80
Lower_80
0
0 2 4 6 8 10 12 14 16 18 20
Resistivity (ohm m)
The following summary comments can be made regarding the linear and multiple
regression analysis:
q The multiple regression analysis has significantly improved the accuracy of the
regression equation compared to the simple regression with only one variable,
increasing the coefficient of determination from R2 = 0.21 to R2 = 0.42. This
improvement can be observed in a comparison between Figure 5-72 and Figure 5-
70. However, as was the case for the non-linear simple regression equation, the
multiple regression equation still failed to account for the four high seepage rate
pond outliers at Toolondo Central.
q As was the case in the EM31 multi-variate analysis, the variable which was found
to be most significant in the regression equation was the vertical hydraulic
conductivity in the upper two metres (upper Kv). Again this confirms that the
upper soil profile is by far the most significant part of the profile controlling
seepage (for sites with deep watertable). In terms of significance in the equation
(ie the importance of the variable in describing variation) the resistivity and upper
Kv variables were both essentially equal in influence as explanatory variables.
q While this analysis indicates a reasonably fitting regression equation, it could not
be used with the same degree of confidence as the EM31 based equation due to:
q The lower correlation coefficient;
q The few number of data points and small range of environments
represented by the data points; and,

q The unexplained outliers in the resistivity analysis that were not present
in the EM31 regression equation.
5.4.4 Channel Specific Resistivity Assessment

5.4.4.1 Toolondo West
Resistivity surveys were conducted along 2 kilometres of channel at Toolondo West.
The water level in the channel was low and the boat had to be dragged at slow speed
along the channel bed. This did not appear to affect results.
Figure 5-76 presents the change in resistivity with depth at the site (over the pondage
test areas). The average resistivity values are much higher than seen at Toolondo
Central site some 3 km to the east. This figure shows increasing resistivity with depth,
up to the watertable, across all ponds. The watertable appears to be located between
12m to 14m – this is based on the distinct decrease in resistivity (attributed to the
saline groundwater) observed between these depths. The sharp increase in resistivity
in pond 1 and to a lesser degree in pond 2 at 10-12m is probably attributable to the
sandstone at this depth. Sandstone is likely to have a reduced porosity compared to
the overlying sediments which will show as an increase in resistivity.
Figure 5.77 presents average resistivity values for the ponds at the Toolondo West site
plotted against pondage test seepage rates for each of the ponds. For depths of 8m and
below the sandstone appears to be dominating the response. It is apparent that the
sandstone is of low permeability and therefore high resistivity may equate to low
seepage at this site. Typically (ie, observed at all other sites) high resistivity indicates
seepage of fresher water into a saline water table. However at this site it is apparent
that the influence of the sandstone resistivity is dominating the response, even below
the watertable and effectively masking the effect of the fresher seeped water in the
saline groundwater. Either the sandstone is absent beneath ponds 3 & 4 (which the
long section in Appendix A suggests it may be), or it is less cemented and therefore
has a lower resistivity.
n Figure 5-76 Resistivity Versus Depth At Toolondo West
Toolondo West
Resistivity (ohm-m)
0 20 40 60 80 100 120 140
0
Pond 1
2
Pond 2
4
Pond 3
Depth Below Surface (m)
6 Pond 4
10
12
14
16
18
20

n Figure 5.77 Resistivity Against Pondage Test Seepage at Toolondo West
Toolondo West
6
At 4m
y = 0.32x - 3.30
R2 = 0.80
At 8m
At 10m
5 At 12m
At 16m
At 2m
4 Linear (At 4m)
Linear (At 8m)
Linear (At 10m)
Linear (At 12m)
3 Linear (At 16m)
Linear (At 2m)
y = -0.03x + 4.80
2 R2 = 0.62
1
y = 0.91x - 7.62 y = -0.25x + 9.45 y = -0.11x + 7.62
y = -0.05x + 5.60
R2 = 0.34 R2 = 0.93 R2 = 0.92
R2 = 0.72
0
0 20 40 60 80 100 120 140
Resistivity (ohm-m)
The reasonable correlation observed at shallow depth (2 and 4m) is probably due to
difference in clay content immediately below the channel. This method of detection is
the same as that of the EM31 (when the watertable is deep). It is not the actual impact
of the seepage plume in the groundwater which is detected, but the lithology below the
channel. In many cases, the resistivity / conductivity response in the near surface will
provide a good indication of the likelihood of seepage, as it has in this case. In some
instances this method may break down, for example due to the effect of a channel liner
(artificial or natural). In such circumstances the geology in the sub-surface will not
provide an indication of the rate of seepage, as seepage is controlled by a thin surface
layer.
Figure 5.77 shows that the differences in resistivity are minor in the near surface but
the correlation becomes strongly negative at greater depth. It appears that reduced
porosity (which will show as an increase in resistivity) may dominate the responses.
Porosity will be inversely related to permeability and thus areas with highest
resistivity (low porosity) at depth have lowest seepage.
Figure 5.78 shows the relatively low resistivity values for the southern end of the line
at Toolondo west. Towards the north the background resistivity values are much
higher which possibly reflects sandstone basement rock close to surface but may be
due to increased seepage. Individual spikes in the data reflect either better cemented
sandstones with low porosity and/or major seepage areas where seepage of the
irrigation channel water has significantly diluted the more saline groundwater. This
effect appears greatest in the top of the saturated zone between 10 and 14m.
Figure 5.79 shows the resistivity at 18m below surface for the section of channel
surveyed at Toolondo West. This reinforces the analysis of Figure 5.78: the northern
section shows higher background resistivity compared to the remainder of the
surveyed channel. This higher background level may be related to shallower

n Figure 5.78 Plot of resistivity values for Toolondo West against distance
along the channel [ Resistivity (ohm-m) on vertical axis and distance (m) on
horizontal axis ]
12000
10000 2m
8000 6m
6000 14m
4000 18m
2000 10m
0
0 500 1000 1500 2000 2500
n Figure 5.79 Depth slice from 18m below surface showing lower resistivity
(yellow to red) in southern section of surveyed line compared with higher
resistivity (blue) in northern section.
sandstone with low porosity and or increased seepage in this section. The evidence
from pondage tests at the site suggests sandstone rather than seepage is the cause,
however the depth of drilling at the site is insufficient to confirm this interpretation.
Deeper drilling would be required for definitive interpretation of the data.

From the geophysical results it appears the low seepage observed at Toolondo West is
occurring over most of the ponds into the unsaturated surface. This seepage plus the
presence of sandstone is apparently resulting in elevated resistivities at depth in Ponds
1 and 2. Below the watertable there are narrow zones with low permeability which
reduces seepage in some sections. The resistivity of these zones shows a significant
increase over background levels. Above the water able there is little difference in
resistivity that can be related to seepage. However at 2m and 4m a positive correlation
is observed between seepage and attributable to differences in clay content in the
upper surface, as EM31 works in the unsaturated zone when the watertable is deep.
Below the watertable there is a negative correlation with resistivity (to that expected
and observed at other sites) most likely related to the effect of the very resistive nature
of sandstone on the response.
5.4.4.2 Toolondo Central

Resistivity traverses were conducted over a 2 km section of the Toolondo Channel
from the south (599370mE, 5916042mN) to north (598138mE, 5916840mN). In
general the resistivities at this site are much lower than that detected at Toolondo
West. This is most likely attributable to differences in lithology, possibly related to
variations in the degree of cementing and hence porosity of the sandstone.
Figure 5-80 presents the change in resistivity with depth at the site, over the pondage
test areas. This figure appears to show a fresher zone at 2m due to seepage effects
immediately beneath the channel, underlain by unsaturated material increasing in
resistivity with depth (interpreted as increasingly fresh sandstone) up until the
watertable. The watertable appears to be located around 10m to 11m, based on the
distinct decrease in resistivity observed between these depths, caused by the relatively
high salinity of the groundwater.
n Figure 5-80 Resistivity Versus Depth At Toolondo Central
Resistivity (ohm-m) Toolondo Central
0 5 10 15 20 25 30 35
0
4
6
Pond 1
8
Pond 2
10 Pond 3
Pond 4
12
Pond 5
14
Pond 6
16

Figure 5.81 shows a series of correlations for different depth slices between seepage
rate and average resistivity for the Toolondo Central section. This plot shows the
improvement in the correlation using depth slices immediately below the watertable.
At 2m there is virtually no correlation and only a poor one at 6m. At 10m and 12m
the best correlation is obtained with an R2 of around 0.6. This fits with the theory that
seepage is best detected as it impacts the watertable and creates a fresher plume
compared to background conditions.
n Figure 5.81 Resistivity Against Pondage Test Seepage at Toolondo Central
Toolondo Central
14
Resis. @ 2m
Resis. @ 6m
12
Resis. @ 8m
10 Resis. @ 10m
Resis. @ 12m
8
y = 0.14x + 4.55 Resis. @ 14m
R2 = 0.03
6 Linear (Resis. @
y = 0.70x - 3.00 2m)
R2 = 0.26 Linear (Resis. @
y = 0.69x - 5.22 6m)
4
R2 = 0.51 Linear (Resis. @
y = 0.57x - 4.28 8m)
2
2
R = 0.59 Linear (Resis. @
10m)
y = 0.55x - 2.57
2 Linear (Resis. @
R = 0.56 12m)
0
5 10 15 20 25 30 35 40 45
Resistivity (ohm-m)
Figure 5.82 shows a plot of the resistivity at 14m below surface. Peaks represent
zones of higher resistivity, which are interpreted as major seepage of fresher channel
water into the more saline groundwater.
Figure 5.83 shows the resistivity long section beneath Pond 1 (from 2002 pondage
tests). The section shows, a conductive surface layer underlain by a more resistive
layer and a conductive layer at depth. These interpretations are confirmed by
comparison with the geological long section for the site (Appendix A), which shows a
sand in the top metre over much of the section, underlain by a sandy clay to medium
clay to a depth of between 5 and 10m underlain by sandstone with a more saline water
table below 10m. The sandstone layer appears irregular in thickness and/or resistivity
and the inversion software has not resolved a well-defined layer.
The section (
Figure 5.83) shows a high resistivity zone extending from surface to depth centred in
the single 2002 pondage test cell. This pond recorded a relatively high seepage rate,
compared to other ponds at the site, of 12 mm/d. A similar feature is present on the

section under pond 1 and 2 from the 2001 pondage tests results, which also recorded
high seepage rates (> 10 mm/d).
12000
Resistivity - ohm.metres
10000
8000
6000
4000
2000
0
0 500 1000 1500 2000 2500
Distance - metres
n Figure 5.82 Plot of resistivity at 14m below surface for section along
Toolondo Channel.
n Figure 5.83 Resistivity section under Pond 1 (2002 Pondage Tests) at

Toolondo Central (Blue = high; red = low). Arrow shows probable seepage
zone under pond.
S N
Results from the 6m depth slice are shown in Figure 5-84. These results show a
similar pattern of conductivity distribution to the results obtained by the on-channel
EM-31 survey conducted in August 2001 (refer 5.3.3.1). The short section of channel

sectioned off to form the pond for the 2002 channel seepage test, straddles an area of
very high resistivity in the resistivity surveys. This was also detected as a low
conductivity feature by the on-channel EM-31 survey in August, 2001 (refer 5.3.3.1).
The average resistivity over this section of channel at 14m below surface is 52 ohm
metres compared to 11 ohm metres for the whole of the resistivity traverse and 12
ohm metres for the whole traverse section between 6 and 18m depth. Figure 5-85
shows how the resistivity peaks at around 10m below surface which is the top of the
saturated zone.
n Figure 5-84 Depth slice at 6 metres below surface (Toolondo Central)

n Figure 5-85 Resistivity values at various depths beneath the pond created at
Toolondo Central in 2002.
Toolondo Central 2002 Pond Resistivity
100
6 metre
8 mete
10 metre
80
12 metre
14 metre
Resistivity (ohm m)
16 metre
60 18 metre
40
20
0
5916400 5916425 5916450 5916475 5916500
Northing (m)
5.4.4.3 Toolondo East

Figure 5-86 presents the change in resistivity with depth at the Toolondo East site,
over the pondage test areas. The sandstone which is intersected at 4-5m at this site is
actually detected as having a lower resistivity than the overlying clayey sediments.
The cause of the low resistivity in the sandstone is not known. The watertable at this
site is thought to be greater than ten metres below surface (as per Toolondo West and
Central), and therefore cannot explain this decrease. There is very little difference in
the resistivity response between the ponds, except for Pond 2, which between 6-10 m
is somewhat higher than the other ponds.
Toolondo East site has very low seepage levels and a very narrow range of resistivity
values. The differences in both seepage and resistivity values are within the possible
error range for both data sets. As such no meaningful correlations were observed
between the resistivity and seepage data, as shown in Figure 5.87.

n Figure 5-86 Resistivity Versus Depth at Toolondo East
Toolondo East
Resistivity (ohm-m)
0 5 10 15 20
0
4
10
Pond 1
12
Pond 2
14 Pond 3
Pond 4
16
Figure 5.87, Figure 5.88 and Figure 5.89 show the resistivity sections under the
channel length corresponding to the pondage sections for Toolondo East. There are no
locations that can be immediately interpreted as major seepage zones, such as seen in
Figure 5.81 at Toolondo Central. Seepage is minimal at this site, as confirmed by the
pondage test results, and from the resistivity sections appears to be relatively evenly
distributed within ponds.

n Figure 5.87 Correlation of resistivity and seepage results for Toolondo east.
Toolondo East
1.0
Average
<5m
0.8 <10m
0.6
0.4 2 2
R = 0.0066 R = 0.0138 2
R = 0.0265
0.2
0.0
10 11 12 13 14 15 16 17 18
Resistivity (ohm m)
n Figure 5.88 Resistivity section for pond 1, Toolondo East.
n Figure 5.89 Resistivity section for ponds 2 and 3 Toolondo east

n Figure 5.90 Resistivity section for pond 3 and 4 Toolondo East
5.4.4.4 Dahwilly Central

Resistivity surveys were conducted along 2 km of the Dahwilly Channel (Central site)
and pondage tests conducted over approximately the western half of the surveyed
channel. Figure 5.92 presents the change in resistivity with depth at the site, over the
pondage test areas. This figure shows a highly resistive unsaturated zone beneath the
channel, with a sharp drop in resistivity at the watertable, which is located around 5m.
With increasing depth below the watertable the resistivity continues to decline as the
fresher channel water is increasingly diluted. At around 16m it appears that natural
background resistivity conditions are reached (approximately 140 ohm-m). Note that
the resistivity values are significantly higher than at the Wimmera sites. This is
attributable to the more coarse and ‘clean’ (ie relatively clay free) sands at the site.
These have a much lower conductivity than the clayey sediments at Toolondo.
n Figure 5-91 Resistivity Versus Depth at Dahwilly Central
Dahwilly (Central)
Resistivity (ohm-m)
0 500 1000 1500 2000 2500
0
4
8
Pond 1
10
Pond 2
12 Pond 3
14 Pond 4
Pond 5
16
Pond 7
18
The Dahwilly Central site includes two sections of channel which were remediated;
one section with a plastic liner and the other with a rubber compound. The liners were
tested in two separate pondage tests, 2.6 and 2.7. Figure 5.92 shows the resistivity
sections from these ponds.

The liner in pond 2.6 is more resistive to electrical current flow. The shape of the
resistivity response seen in the section is governed by the shape of the pond liner, with
apparently high resistance underlying a thin conductive layer, as all the current is
concentrated in the channel water. The ends of the liner area are characterised by
conductivity anomalies extending from the surface. These anomalies are caused by
the fact that the current is confined to the surface above the liner and by effects related
to the asymmetry of the resistivity array. This pond was therefore removed from the
correlation analysis because of these effects on the resistivity.
In pond 2.7 the liner appears relatively transparent to electrical current and shows a
more saline water table at depths of 8 to 10 metres below surface which corresponds
with groundwater reported from bores of 6 to 7 metres. The section under pond 2.7
possibly represents more ‘natural’ (ie, without channel) conditions compared to other
parts of the section. Only minor seepage was reported from both the lined ponds (1.1
mm/d for pond 6 and 2.8 mm/d for pond 7). Possible seepage zones are marked by ‘S’
on the section. These appear to be minor compared to the effects further to east along
this channel.
n Figure 5.92 Resistivity Sections Below Lined Ponds at Dahwilly Central.

Sections over ponds 2.7 and pond 2.6 show different effects due to the pond
liners. (The liner beneath pond 6 interfered with the survey and therefore the
results were discarded)
Figure 5.93 shows that the resistivity under ponds 2.5 and 2.4 at Dahwilly Central are
significantly higher than under the lined, low-seepage pond 2.7. (This is also clearly
illustrated in Figure 5-91). The change in the profile at depth is most obvious in
comparing Figure 5.92 and Figure 5.93; note the (pink) low resistance results at depth
beneath pond 7 compared to the (yellow-brown) higher resistance results beneath
ponds 1-5. This is due to dilution of the more saline groundwater by the seeped

n Figure 5.93 Resistivity sections under ponds 2.5 and 2.4 at Dahwilly Central.
Refer Figure 5-92 for resistivity legend
(fresher) channel water which is more resistive. It appears that the seepage
mechanism is by relatively continuous diffusion along the channel, in contrast to the
more isolated seepage paths observed on the Toolondo channel. This seepage
mechanism is also suggested by the lithology at the site, which indicates the entire
length of channel surveyed is underlain by approximately 10m of medium to coarse
grained sand, and is more likely to result in uniform seepage rather than seepage
‘hotspots’. Ponds 1-5 have seepage rates around 4 times that of pond 2.7, with an
average resistivity of around 275 +/- 15 ohm metres, compared to 145 ohm metres for
pond 2.7.
n Figure 5.94 Resistivity sections under Dahwilly Central ponds 2.4 and part
of 2.3. See Figure 5.92 for resistivity legend
S2 S3 S4 S1
Figure 5.94 shows possible localised seepage zones, as indicated in the resistivity
sections under ponds 2.4 and part of 2.3. This site should be compared with pond 7
(Figure 5.92) where there is very low seepage. Overall, resistivity is higher for these
ponds. In particular, under pond 2.3 there is a short section where resistivity is
significantly higher than the rest of the section (S1). This high localised resistivity
zone may reflect the higher seepage for this section. Further minor local seepage may
be located at S2, S3 and S4. The surface and deepest sections are relatively uniform
and the greatest deviation is at S1 where the shallow higher resistivity water appears to
change the resistivity of the underlying section to a greater depth into the unsaturated
zone.

n Figure 5.95 Resistivity section under ponds 2.1 to 2.3 at Dahwilly Central.
See figure 4.1 for legend of resistivity values.
Figure 5.95 shows a relatively uniform resistivity section stretching from pond 2.1 to
the eastern part of pond 2.3. Ponds 2.1 and 2.2 have very similar seepage rates (4-5
mm/d), whereas pond 2.3, with the high resistivity zone suggests significantly higher
seepage, as confirmed by the pondage test result of 9.5 mm/d.
Plots of pondage test seepage rate against resistivity at various depths within the
profile are shown in Figure 5.96. The increasingly improved correlation with depth,
peaking immediately below the watertable can be seen at the 8m and 10m depth slices
where the seeped water has diluted the saltier background water. At greater depth
within the profile (16m), conditions trend towards background groundwater conditions
due to dilution and mixing. Therefore there is less distinction between different
seepage rate ponds at these depths.
n Figure 5.96 Correlations of average resistivity values of sections under

ponds for Dahwilly Central
Dahwilly Central
12
y = 0.03x - 2.27 y = 0.02x - 2.81

R2 = 0.62 R2 = 0.86
10 At 2m
y = 0.01x - 2.27 y = 0.00x + 2.06
R2 = 0.72 R2 = 0.46 At 6m
8 At 8m
At10m
At 16m
6
Linear (At 2m)
Linear (At 6m)

4
Linear (At 8m)
Linear (At10m)
2
y = 0.02x - 2.16 Linear (At 16m)
R2 = 0.85
0
0 500 1000 1500 2000 2500
Resistivity (ohm-m)

5.4.4.5 Dahwilly East (Pretty Pine)
Resistivity surveys were conducted along 1.2 km of the Dahwilly Channel at Pretty
Pine, about 5km to the east of the Dahwilly central site. Pondage tests were conducted
on the western half of the surveyed channel.
Figure 5-97 presents the change in resistivity with depth at the site, over the pondage
test areas. This figure shows an increasingly resistive unsaturated zone beneath the
channel, with a sharp drop in resistivity at the watertable, which is located around 6m.
As at other sites, with increasing depth below the watertable the resistivity continues
to decline as the fresher channel water is increasingly diluted. Natural background
resistivity conditions may not have been reached at 14m as indicated by the still
increasing gradient of the curves between 12m to 14m.
Note that the resistivities are significantly lower than at the Dahwilly central site. This
is attributable to the finer grain size and more clayey sands at this site compared to the
coarser and ‘clean’ (ie relatively clay free) sands at the Dahwilly central site. At depth
below the watertable however, resistivity values between the two sites are comparable.
n Figure 5-97 Resistivity Versus Depth at Dahwilly East

Dahwilly East (Pretty Pine)
Resistivity (ohm-m)
0 50 100 150 200 250 300 350 400 450
0
4
10
Pond 1
12
Pond 2
14
Pond 3
16
Figure 5.98 shows the resistivity section across the surveyed section is relatively
uniform. The section shows a thin conductive surface layer overlying a thick (8-10m)
unsaturated zone of higher resistivity above a more conductive layer at depth (due to
more saline groundwater). This interpretation is confirmed by the geological long
section of the site (refer Appendix A) which shows approximately 1.5m of clay to sand
clay, overlying sand to clayey sand, to at least 10m. The resistivity section also
appears to detect the presence of the clayey sand profile (compared to clean sand) in
the western end of pond 1, as indicated by the absence of the high resistance upper
layer at this location in Figure 5.98. The section shows a zone of probable higher

seepage on pond 3 and areas of probable lower seepage showing as low resistivity on
pond 1.
n Figure 5.98 resistivity section from Pretty Pine section of Dahwilly Channel
n Figure 5.99 Correlation of seepage and resistivity from ponds at Pretty Pine
Dahwilly East (Pretty Pine)
11
y = 0.01x + 8.27
R2 = 0.85 y = 0.01x + 8.09
y = 0.01x + 7.59
R2 = 0.90
R2 = 0.68
10.5
y = 0.01x + 7.58
R2 = 0.91
10 At 2m
At 4m
At 6m
9.5 At 10m
At 14m
Linear (At 2m)
9 Linear (At 4m)
Linear (At 6m)
y = 0.01x + 7.26
R2 = 0.80
Linear (At 10m)
Linear (At 14m)
8.5
0 50 100 150 200 250 300 350 400 450
Resistivity (ohm-m)

Figure 5.99 shows a good correlation of seepage versus average resistivity for the
three ponds at Pretty Pine. However, there is only a narrow range of values for
seepage. The best correlation is obtained using only resistivity values from the 6m
depth slice. The correlation is less favourable using data from deeper than 10m.
At Pretty Pine the influence of seepage is best seen in the resistivity data at the
watertable or within the top few metres of the watertable. Resistivity data from deeper
in the section appears to be less affected by seepage. This may be due to the fact that
at around 8m, the profile becomes slightly clayey (clayey sand) which may limit
vertical migration of seepage into the aquifer. Limited emphasis should be placed on
these correlations due to the small number of ponds at the site.
5.4.4.6 Finley
A short section of channel at Finley was surveyed using resistivity profiling covering
the pondage test lengths. A plot of resistivity versus depth for the site is presented in
Figure 5.101. The upper two metres contains a layer of higher resistivity, most likely
caused by shallow seepage into the shallow watertable. There appears to be relatively
little seepage impact below 2m with resistivity rapidly dropping at 4m. The
resistivity values of between 5-10 ohm-m within the 4 – 10m depth range are among
the lowest readings recorded across all of the sites and are due to the very heavy clays
and highly saline groundwater at the site (comparable values also observed at Lake
View). Between 8 and 10m the resistivity begins to increase again which reflects a
change in lithology at this depth from a heavy clay to a sandy clay (refer to Finley
Long Section in Appendix A).
n Figure 5-100 Resistivity Versus Depth at Finley

Finley
Resistivity (ohm-m)
0 10 20 30 40 50 60
0
4
10
12 Pond 1
Pond 2
14
Pond 3
16

The resistivity section in Figure 5.101 also shows that the resistivity at the Finley site
is much lower than all other sites surveyed. Possible impact on the groundwater is
only evident along the surface and at depth on the west end of pond 1. There is also a
narrow range of seepage values recorded across the three ponds. The results suggest
that the high clay levels at the site are preventing seepage from diffusing into the
groundwater except in more permeable narrow zones such as the west end of Pond 1.
Most of the seepage is in lateral surface flow.
n Figure 5.101 Resistivity section at Finley
The average of the inverted resistivity sections is plotted against pondage seepage tests
and presented in Figure 5.102. This figure shows that the best correlations were
obtained in the 4m to 6m range. The shallowest resistivity values at 2m have a reverse
correlation. It might have been expected that the better correlations would have been
obtained at 2m, given that seepage at this site is predominantly shallow and migrates
laterally. However this result may be a reflection of the data processing and the fact
that the 5m array did not resolve the surface resistivity very well. The fact that a
reasonable correlation was obtained at 4m suggests some seepage impact into the
watertable.
However, the very narrow spread of seepage values (5.5 – 7 mm/d) at this site limits
the conclusions that can be drawn from this analysis. Note that only resistivity values
for the western 140m of pond 1 were collected (35m of data was missed on the eastern
end of the pond). It was assumed that this did not affect the analysis.

n Figure 5.102 Pondage Test Seepage Versus Resistivity at Finley
Finley
7.5
At 2m
At 4m
7 At 6m
y = 0.52x + 2.43 At 8m
R2 = 0.93 At 10m
6.5 At 14m
Linear (At 2m)
y = -0.02x + 6.06 Linear (At 4m)
R2 = 0.00 Linear (At 8m)
6
Linear (At 6m)
Linear (At 14m)
Linear (At 10m)
5.5
5 y = -0.12x + 7.23 y = -0.06x + 8.15

y = -0.08x + 7.69
y = 0.44x + 1.89 R2 = 0.36 R2 = 0.89
2 R2 = 0.61
R = 0.92
4.5
4
0 10 20 30 40 50 60
Resistivity (ohm-m)
5.4.4.7 Lake View Central
A resistivity survey was conducted along a 2 km section of the Lake View (Central)
channel site.
Figure 5.105 presents the change in resistivity with depth at the site, over the pondage
test areas. As per the Finley site, the upper two metres contains a layer of higher
resistivity, due to seepage into the shallow watertable (approximately 1.5m). The
curves suggest that seepage has impacted up to 6m within the profile. Below 6m the
resistivity remains reasonably constant, suggesting little seepage impact below this
depth and values probably reflect saturated background resistivity conditions.
Figure 5.104 shows the resistivity results plotted against pondage seepage at the Lake
View (Central) channel site. The best relationships between resistivity and seepage
are achieved using resistivity data from 6m to 8m depth. There is essentially no
relationship observed for the 2m and 4m data which might be expected given the
shallow watertable at this site. As stated in the Finley section this may well be a
reflection of the data processing and subsequent poor resolution of near surface
resistivities.
The fact that reasonable correlations were observed for 6 – 8m may also be explained
by the site lithology, as much as seepage impacts on the groundwater. This site has a
heavy clay starting between 3-8m below surface. Beneath Pond 1 (which contains the
highest seepage) the medium to heavy clay starts at 8m depth, compared to Ponds 2-4,
where the heavy clay occurs between 2m and 3m from the surface (refer Lake View
geological long section, Appendix A). Therefore seepage appears to be greatest in

ponds with higher resistivity within clayey sediments, possibly corresponding to
sandier sections. The resistivity sections for the site (
Figure 5.105, Figure 5.106 and Figure 5.107) show that there are probable localised
seepage paths under pond 1 and to a lesser extent under pond 4, extending to depth. A
less well defined potential path under pond 2 is also indicated. For localised seepage
paths within large ponds the results are less likely to correlate well, due to the bulk
averaging of the resistivity across the entire pond.
n Figure 5-103 Resistivity Versus Depth At Lake View (central) Channel
Resistivity (ohm-m) Lake View
0 2 4 6 8 10 12 14
0
4
10
Pond 1
12 Pond 2
Pond 3
14
Pond 4
16

n Figure 5.104 Correlation of resistivity with pondage leakage tests for the
Lake View (original) test site.
Lake View
7.5
7 y = 0.85x + 0.90
R2 = 0.58 At 2m
y = 0.38x + 3.37
6.5 At 4m
R2 = 0.33
y = 0.36x + 1.16
R2 = 0.04 At 6m
6
At 8m
5.5 At 14m
Linear (At 2m)

y = 0.52x + 3.05 y = -0.25x + 7.66
5
R2 = 0.60 R2 = 0.11 Linear (At 6m)
Linear (At 4m)

4.5
Linear (At 8m)
4 Linear (At 14m)
3.5
0 2 4 6 8 10 12 14
Resistivity (ohm-m)
n Figure 5.105 Resistivity section under Pond 1 (Possible seepage paths at

points marked X; Pond is 270m long)
x x x

n Figure 5.106 Resistivity section for Pond 2 at Lake View (original). (Possible
seepage paths at points marked X; pond is 280m long)
n Figure 5.107 Resistivity section for ponds 3 and 4 at Lake View (Original)
(Possible seepage paths at points marked X; ponds are 80m long)
x x

5.4.4.8 Lake View West
A resistivity survey was conducted along a 2km section of the Lake View West
channel. Figure 5-108 presents the change in resistivity with depth at the site, over the
pondage test areas. The response is reasonably similar to the Lake View (central) site.
Depth to watertable at the site is shallow at around 0.5 - 1.5m. As at Lake View
(central), the upper part of the profile contains a layer of higher resistivity, due to
seepage into the shallow watertable. The resistivities are higher in this upper zone
compared to Lake View (central) probably due to the higher seepage rates at Lake
View west and greater dilution of the salty groundwater. The curves suggest that
seepage has impacted up to 6m within the profile.
For Ponds 3 and Pond 4 the resistivity remains reasonably constant (below 6-8m),
suggesting little seepage impact below this depth and values probably reflecting
saturated background resistivity conditions. The resistivity below 8m in Pond 1
declines rapidly, back to background levels similar to those observed in Pond 3 and
Pond 4. In pond 2 the resistivity significantly increases below 10m, with a value of 37
ohm-m at 16m, compared to around 10 ohm-m for the remaining three ponds. This
response is most likely related to a geological anomaly beneath this pond. Drilling at
the site was only conducted to 10m however and this cannot be confirmed. At this site
lithology changes at this depth are unlikely to impact on seepage in any case.
n Figure 5-108 Resistivity Versus Depth at Lake View West
Resistivity (ohm-m) Lake View West
0 5 10 15 20 25 30 35 40
0
Pond 3
2
Pond 4
4
Pond 1
6
Pond 2
8
10
12
14
16
18

n Figure 5.109 Correlation of resistivity and seepage rates for Lake View West
Site
Lake View West

27
26 y = -0.76x + 32.47
R 2 = 0.90
25
y = 0.0116x + 23.105
y = -0.36x + 27.19 R 2 = 0.0046
24 R 2 = 0.77
23
At 2m
22 At 4m
At 6m
At 10m
21
At 16m
Linear (At 16m)
20
y = -0.52x + 30.04 y = -0.75x + 35.59 Linear (At 2m)
R 2 = 0.96 R 2 = 0.94 Linear (At 4m)
19 Linear (At 6m)
Linear (At 10m)
18
0 5 10 15 20 25 30 35 40
Resistivity (ohm-m)
Figure 5.109 shows the resistivity results plotted against pondage seepage at the Lake
View West site. All ponds at this site returned high seepage rates. Figure 5.109
shows negative correlations (ie high resistivity equating to low seepage) compared to
those detected at other sites. This is largely attributable however to the high resistivity
values at Pond 1. The anomalous behaviour of Pond 1 (compared to remaining ponds)
is clearly illustrated in Figure 5-108, displaying opposite trends with depth. It would
be expected that the best and positive trends would be observed in the 2-4m range,
based on the fresh seepage into the shallow watertable aquifer and the fact that it
appears that most of the seepage is confined to the upper part of this aquifer. The fact
that such a trend was not observed in the resistivity data, but was observed in the
EM31 survey conducted at the site (refer Figure 5.64) again suggests that for shallow
readings the resistivity resolution may have been unsatisfactory. The wide spread of
resistivities at depth is caused by the anomalous readings at depth in Pond 2.
In summary, the lack of a meaningful relationship between resistivity and seepage at

this site is attributed to:
q Poor resolution of near surface data by the resistivity array (this is supported by
the fact that the EM31 recorded a sensible correlation at this site);
q The ‘anomalous’ result in pond 1. Something at depth (presumably a dramatic
change in lithology, although this is not detected in the logs) is causing elevated
resistivities relative to other ponds; and,
q The narrow spread of seepage data (20 – 25 mm/d).

It is not certain the relative contribution each of these factors makes to the overall poor
relationship.
Figure 5.110 shows a depth slice of the resistivity at 18m below surface and Figure
5.111 shows the resistivity at 6m below surface. The difference between the two
depths is particularly clear for ponds 2 and 4 which both display an increase in
resistivity at depth (pond 3 also does but not to the same extent). This suggests a layer
of higher permeability may be located at depth. Seepage from the base of the channel
appears to be moving laterally in the upper part of the profile but not penetrating to
depth. The high resistivity in the near surface and lower seepage rate observed at
Pond 1 may be due to a near surface layer of low permeability and porosity, although
this is not clearly demonstrated from the logging.
n Figure 5.110 Resistivity depth slice at 18m below surface (red is low, yellow
moderate and blue high resistivity)
Pond 1
Pond 2
Pond 3
Pond 4

n Figure 5.111 Resistivity at depth of 6m below the channel (red is low,
yellow - moderate and blue high resistivity)
Pond 1
Pond 2
Pond 3
Pond 4
5.4.5 Extrapolation at each Channel
For each of the channels where resistivity surveys were conducted several kilometres
apart (Toolondo, Dahwilly and Lake View) this section looks at the accuracy of
extrapolation from one site to another, which was one of the key aims of the year three
trials.
5.4.5.1 Toolondo
Three sites were trialed at Toolondo channel, all within about 13km of each other:
q Toolondo Central – The original site on which previous years trials had been
conducted.
q Toolondo West – Similar to Toolondo Central in terms of geology and
hydrogeology.
q Toolondo East – Change in surface soil type compared to Toolondo Central and
West: heavy to medium grey cracking clays, compared to weathered sandstone
profiles at Toolondo Central and West. Similar geology at depth, however with
sandstone at 3-4m.

n Figure 5-112 Pondage Test Seepage Versus Resistivity at All Toolondo
Sites
Toolondo (all 3 sites)
12
Toolondo West (Av 12&14m)
Toolondo Central (10m)

10
y = 0.55x - 5.29
Toolondo East (Av 10,12 & 14m)
R2 = 0.62
8 Toolondo West (2m)
0
0 20 40 60 80 100 120
Resistivity (ohm-m)
Figure 5-112 plots all of the regression trend lines for the three Toolondo sites. A
depth slice or average of depth slices has been selected from just below the watertable,
which is between 10-14m across all of the sites. The following observations are made
regarding this plot:
q It can immediately be seen that Toolondo East is measured across such a narrow
seepage and resistivity range that no meaningful trend can be obtained from it.
However, as a collection of points it does lie within the expected range for the
Toolondo Central site;
q The Toolondo West trend line does not fit within the Toolondo Central and
Toolondo East relationship, and displays an inverse relationship between seepage
and resistivity. The possible reasons for this are discussed in Section 5.4.4.1. It is
apparent that the effects of seepage on the watertable (which would cause a
increase in resistivity) are being masked by the particular lithology of the site. A
possible explanation is that the sandstone is much more cemented at this site and
this reduced porosity (which will show as an increase in resistivity) may dominate
the response. Porosity may be directly related to permeability and thus areas with
highest resistivity at depth have lowest seepage.
q A plot of the Toolondo West trend line for the 2m depth slice is shown. This fits
reasonably well with the Toolondo Central and East data. In fact the heavy line in
this plot is the trend line for the Toolondo Central, East and West (2m depth slice
data) sites which results in a reasonable correlation coefficient of 0.62. Section
5.4.4.1 explains that the observed good correlation at 2m is probably due to
detection of changes in clay content in the near surface, and does not rely on
mapping the seepage plume itself. In practice, without the pondage tests, it could
not be known that the upper profile was a better depth to concentrate on than the

section immediately below the watertable, which makes this observation
somewhat academic.
The overall conclusion regarding extrapolation from these sites is that if the Toolondo
Central data had been used to predict seepage rates at the Toolondo West site, seepage
would have been greatly overestimated. This is due to the fact that seepage does not
appear to be having an impact on groundwater salinity below the watertable, possibly
due to differences in the nature of the sandstone at the Toolondo West site (greater
cementing equals reduced porosity which equals higher resistivity but lower seepage),
which masks any seepage impacts. This highlights the dangers of extrapolation, even
in environments which on the surface and according to available information appeared
geologically similar and are quite close to each other (within 5 km). This suggests that
interpolation rather than extrapolation is a safer means of using pondage tests in
investigations along large reaches of channel.
5.4.5.2 Dahwilly
Two sites were trialed on the Dahwilly channel, approximately 6-7 km apart:
q Dahwilly Central – The original site on which previous years trials had been
conducted.
q Dahwilly East – Similar to Dahwilly central in terms of geology and
hydrogeology.
n Figure 5-113 Pondage Test Seepage Versus Resistivity at Dahwilly Sites
Dahwilly (both sites)

12
10
Dahwilly Central
4 (8m)
Dahwilly East,
(Pretty Pine) Av
2 6&8m
Finley (4m)
0
0 100 200 300 400 500 600 700 800
Resistivity (ohm-m)
Figure 5-113 plots the regression trend lines for two Dahwilly sites. Finley (the third
site tested within Murray Irrigation) is also included for comparison purposes. For the

two Dahwilly sites, resistivity depth slices were selected from just below the
watertable, which is around 5-6m depth. The following observations are made
regarding this plot:
q Two very different trend lines for the two Dahwilly sites are observed. The
regression trend line for the combined data set (light line) essentially has a
correlation coefficient of zero, indicating no trend between the two sites. Within
each site however a reasonable correlation is observed and the trend lines display
similar gradients. These results suggest different background conditions at the
two sites, despite the apparent similarities of the sub-surface conditions. For the
seepage rates observed at Dahwilly East, for the data to fit with the Dahwilly
Central site, much higher resistivities would need to be recorded. The lower than
‘expected’ resistivities at Dahwilly East are most likely explained by the finer
sands at this site and the more clayey profile, particularly below the watertable.
The sands at the Dahwilly Central Site are very coarse and relatively clay free.
The other possible explanation is that the background groundwater at the Dahwilly
East site is of higher salinity. The available background information suggests this
is not the case (refer Table 5-1), but may not be based on bores close to the site.
q In terms of extrapolating relationships from one site to another, this confirms the
outcomes from the Toolondo site – that relationships between resistivity and
seepage can change over relatively short distances, even when available
information suggests the environments are reasonably similar, and therefore
extrapolation without pondage tests is dangerous. The overall conclusion
regarding extrapolation from these sites is that if the Toolondo Central resistivity
data had been used to predict seepage rates at the Toolondo West site, seepage
would have been greatly underestimated at less than 4 mm/d, compared to actual
rates of around 10 mm/d. Again, the use of interpolation rather than extrapolation
would appear to be the preferred approach. In this case actual on site bore data at
Dahwilly East would assist is comparing background groundwater salinities.
q It is apparent that Finley is located in quite a different environment, as suggested
by the very low resistivities. The trend line between the data points is almost
vertical in comparison to the two trend lines of the two Dahwilly sites. It is
therefore not appropriate to include this data set in the Dahwilly analysis.
5.4.5.3 Lake View

Two sites were trialed on the Lake View Channel, approximately 8 km apart:
q Lake View Central – The original site on which previous years trials had been
conducted.
q Lake View West – Similar to Lake View Central in terms of geology and
hydrogeology.
Figure 5-114 plots the regression trend lines for the two Lake View sites. Depth
slices from 6m (solid symbols) and from 2m (hollow symbols) are presented. The
watertable at both sites is shallow, at between 0.5 - 1.5 m below surface. The
following observations are made regarding this plot:
q Two very different trend lines for the individual (6m depth slice) Lake View sites
are observed. The regression trend line for the combined 6m depth slice data set
(dark line) has a poor coefficient of 0.23, indicating little trend between the two
sites. A slightly improved relationship is obtained for the 2m slice (dashed line)
although the trend is still only moderate at best. As discussed in Section 5.4.4.8

the inverse correlation at the Lake View West site is largely caused by a possibly
anomalous result in pond 1. This same ‘outlier’ appears to be contributing to the
poor overall trend. Further testing would be required for definitive conclusions to
be drawn regarding the validity of a relationship between the two sites. In
particular the collecting of additional data at seepage rates between the two sites is
necessary, as both are clustered about a narrow seepage range.
q As per the previous two sites (Toolondo and Dahwilly) these results suggest
different background conditions exist at the two sites. For the seepage rates
observed at Lake View West to fit with the Lake View Central site data, much
higher resistivities would need to be recorded. The lower than ‘expected’
resistivities at Lake View West are most likely explained by the gravelly clay
which extends to 5m, and probably acts as a better conductor than the sandy clay
loams and sandy clays at the Lake View Central Site. If the Lake View Central
data had been used to predict seepage rates at the Lake View West site, seepage
would have been greatly underestimated (5-10 mm/d instead of 20-25 mm/d).
q Also as per the two previous sites, in terms of extrapolating relationships from one
site to another, this confirms that relationships between resistivity and seepage can
change over relatively short distances, and therefore extrapolation without
pondage tests is dangerous.
n Figure 5-114 Pondage Test Seepage Versus Resistivity at Lake View Sites
Lake View (both sites)
30
y = -0.36x + 27.19 Lake View

R2 = 0.77 2m Central (6m)
25
6m
Lake View
West (6m)
20
y = 0.97x + 6.57
R2 = 0.23
15
y = 1.98x - 13.82
R2 = 0.37
10
5
6m y = 0.85x + 0.90 2m
R2 = 0.58
0
0 5 10 15 20 25
Resistivity (ohm-m)

5.5 Conclusions
The geophysics conclusions are set out in the following manner:

q Discussion and Conclusions
q Summary of EM31 Results
q Summary of Resistivity Results
The general conclusions draw together all of the results and include a comparison of
the three techniques and a discussion of: seepage detection mechanisms, confidence in
derived relationships, extrapolation of results and comparison of this investigation to
recent international studies using geophysics for seepage measurement and detection.
The three sections summarising each of the techniques are based around a summary
table for each technique which condenses the key outcomes from each trial into one
table, with an emphasis on comparing the results of each trial with the corresponding
pondage tests in that section.
5.5.1 Discussion and Conclusions

5.5.1.1 Seepage Detection Mechanisms
Geophysical techniques applied to seepage measurement primarily depend upon
measuring a contrast in terrain conductivity (or resistivity) in the sub surface profile
around the channel. They can be used in one of two ways:
1) Directly measuring the conductivity of the groundwater, and identifying the
conductivity contrast of fresher channel water as it seeps into and dilutes saltier
native groundwater. Decreasing the salinity of the groundwater will cause a
decrease in electrical conductivity (or an increase in its inverse, resistivity).
2) Identifying contrasts in soil properties and inferring the likelihood of seepage
through more permeable materials in the zone above the watertable. Formations
more likely to allow seepage, such as sands, are naturally lower in conductivity
(higher in resistivity) due to lower porosity and lower cation exchange capacity
Figure 5-115 visually depicts how these two different approaches can be used to
identify or infer seepage.

n Figure 5-115 Comparison of how geophysical techniques can be used to
identify channel seepage (LHS – inferred from soil property variations, RHS –
Seepage inferred Seepage measured directly

Clay
Sand
Saline groundwater
of seepage is assumed to be related to unsaturated zone soil properties beneath or
adjacent to the channel. In many cases this is a reasonable assumption, however the
unsaturated zone is not necessarily the controlling influence on seepage. For example,
over time most Australian channels tend to silt up and the resulting surface clogging
layer is often more restrictive than the unsaturated zone. Therefore unsaturated zone
lithology may not be related to seepage rates, as seepage is controlled by the thickness
and conductance of the clogging layer.
Nevertheless, it was found that the inferred method of identifying contrasts in soil
properties (ie, where the watertable was deeper than the penetration depth of the
geophysical equipment) was successful at most sites conducted during the trials
(ANCID, 2003). There is less risk however in using the direct method of seepage
detection, because as the name implies it is not inferred, but direct. An example of
where the ‘inferred’ method of detection did not work was at Dahwilly Central where
an EM31 survey was conducted while the channel was not running. The survey was
therefore measuring changes in the unsaturated zone and not in the groundwater. At
this site the silt layer in the channel, not the unsaturated zone is the restrictive layer
and therefore no correlation was observed. When the survey was repeated with the
channel running, a good correlation was obtained.
Some possible limitation of the direct method of seepage detection are listed below:
(However, it is still considered the preferred technique over inferring seepage based
on soil property variations).

q In relatively non-saline groundwater environments, the fresh seepage water will
not contrast with the native groundwater. This is not expected to be a problem in
most Australian conditions;
q In environments where the channel seepage water might be rapidly mixed with
native groundwater, such as sites with high groundwater gradients or highly
transmissive environments, the salinity impact on the groundwater may not be as
significant. This can largely be overcome by using survey traverses close to (or
on) the channel; and,
q Groundwater salinity variations along the channel will affect the results and this
needs to be allowed for in the interpretation.
In summary, it is very important that the depth to watertable is known at the site
before selecting a technique. Based on this information a decision can be made as to
5.5.1.2 Comparison of Trialed Geophysical Techniques

q Accuracy
q Cost and Speed
q Data Processing
The three techniques trialed in this investigation (EM31, EM34 and resistivity) are
discussed in terms of each of these criteria.
Accuracy
In theory on-channel resistivity surveying should be the most accurate of the
geophysical techniques trialed, as it is based on a direct method of seepage detection
(refer section 5.5.1.1). As the technique allows definition of changes in resistivity /
conductivity through the profile, the depth where seepage impacts will be most
evident (at and below the watertable) can be targeted. At most sites resistivity
surveying results were comparable to EM31 and EM34, and at three sites (Dahwilly
Central, Dahwilly East and Finley) the resistivity correlations with pondage tests were
better than the EM31 and EM34 correlations. The Dahwilly site demonstrates the
benefits of targeting the watertable for seepage detection in an environment where
seepage is not controlled by the unsaturated zone, but by a surface clogging layer in
the channel.
Resistivity did not prove to be quite as ‘accurate’ as EM31 in environments with a

shallow depth to watertable. This was largely attributed to poor near surface
resolution of the particular resistivity equipment used in the survey, and not inherent
in the resistivity method itself. The resistivity surveying was conducted using
equipment that was used for the first time in these trials and as such was largely
experimental. The designers of the equipment indicate that improved resolution at
shallow depth could relatively easily be achieved in future surveys by slight design
modifications (Allen, pers. comm. 1/11/02).

EM34 at 10m coil separation in the horizontal mode provides a similar depth
penetration to EM31 (vertical mode) and therefore is similarly accurate (but slower to
use). EM34 at a 20m coil separation provides a deeper penetration and focus. At one
trial site, the depth focus was apparently too far below the watertable and the critical
zone was missed. This is a fundamental limitation with all Geonics EM surveys and
other such fixed array type geophysical surveys – the result is averaged over a specific
depth interval, which may not be the critical interval of interest.
However, the robustness of the EM31, as demonstrated by the consistent results in the
trials is due to its relatively shallow depth focus (1-4m). For channels where there is a
shallow watertable (eg, surface to 3-4m), EM31 can be used for direct measurement of
is deep, EM31 infers seepage from near surface soil properties. This works in most
instances but may break down where clogging processes rather than unsaturated zone
lithology control seepage.
The significant advantage of resistivity surveying is that the final output is a two
dimensional profile of resistivity beneath the channel. Not only does this allow easier
interpretation of the results but it can also provide an indication of seepage
mechanisms. For example, at the Toolondo central site the resistivity profile shows
isolated sections of high resistivity (low seepage) emanating from the channel. This
is in contrast to the Dahwilly channel where the profile suggested seepage by
relatively continuous diffusion along the channel. This seepage mechanism is
supported by the lithology at the Dahwilly site, which indicates the entire length of
channel is underlain by approximately 10m of medium to coarse grained sand, and
hence is more likely to result in uniform seepage rather than seepage ‘hotspots’.
Cost and Speed

Approximate costs for the three techniques trialed are given below, based on
geophysical surveys undertaken in these trials. However, it is important to note that
for geophysical surveys often a significant proportion of the costs are overhead costs
(mobilisation, equipment set up etc) and therefore the unit cost per kilometre will
usually be substantially less for long sections of channel. Costs will obviously vary
depending on site specific condition (eg, on land – fences, other obstacles, on channel
– bridges, checks, fences etc).
q EM31 Surveys:
q Wimmera Mallee Water: For 6 kms, on-land including 4 traverses on
each side of channel (over 3 sites): $400/km (includes mobilisation,
data processing and mapping).
q Murray Irrigation: For 8 kms, on-land including 4 traverses on each
side of channel (over 4 sites): $340/km (includes mobilisation, data
processing and mapping).
q Murrumbidgee Irrigation: On-land including 4 traverses on each side
of channel, on each side of channel, the unit cost ranged from
$650/km (3km section) to $800/km (1 km section). (includes
mobilisation, data processing and mapping). On-channel survey
cost was $330/km for a 3 km section.

EM31 is currently the cheapest of the geophysical methods due to the speed of
data acquisition.
q EM34 Surveys:
q Wimmera Mallee Water: For 4 kms over 2 sites: $250/km (1 traverse
only on one side of the channel) ie, $500/km for both sides of
channel (excludes mobilisation).
q Murray Irrigation: For 6 kms (on each side of channel) over 3 sites:
$435/km (includes mobilisation).
EM34 is more expensive than EM31 as two people (on foot) are required to
operate the equipment.
q Multi-electrode Resistivity Surveying – The follow costs were for resistivity

surveying across 11 sites (approximately 2km each in length) in the Wimmera,
Murray and Murrumbidgee Irrigation areas:
q Resistivity towed array surveys: $900/km [Includes mobilisation
(from Adelaide), travel between sites, production and all equipment
costs]
q Data processing costs: $220/km.
Note that resistivity surveying costs are difficult to quantify given that the
technique is relatively new. Costs are likely to come down as the technique is
refined, the equipment becomes commercially available and subsequently
competition is introduced.
A number of commercially operating EM34 and EM31 contractors are in operation in
south east Australia, sufficient to ensure reasonable competition and prices. At
present on-channel resistivity surveying is still in a development phase and as such
there are no commercially operating contractors who specialise in this type of survey.
However, a number of geophysical exploration / surveying companies have the
capability to develop this type of equipment (such as the company who conducted
these trials) and should the demand for such surveying increase, it is expected other
companies could also develop this capability. At present however this may be a
constraint on resistivity surveying.
Data Processing
Data processing requirements for EM31 and EM34 surveying are minimal. In
comparison data processing requirements for resistivity surveying are considerable,
due to the cost of inverting the data to produce a resistivity cross section. This is not
really a constraint of the technique, but adds to the overall cost of resistivity
surveying. It should be ensured that the contractor undertaking the resistivity
surveying also has the capability to undertake the data processing. Approximate costs
for data processing are provided above.

5.5.1.3 Critical Geophysical Survey Variables
q Survey timing – The timing of the geophysical survey will depend on the method
of seepage detection being used. If seepage is being inferred from soil properties
then the timing of the survey is not critical and can be conducted whether the
channel is running or empty. However if direct measurement of seepage is used,
the survey must be conducted while the channel is running, and preferably after it
has been running for a least one month (depending on depth to watertable and
vertical hydraulic conductivity), to ensure seepage has impacted the groundwater.
q On-channel versus on-land – During the trials, on-channel (ie, in a boat) EM31
surveys:
• Did not work at one site where the watertable was beyond the range of the EM31
and returned similar (reasonable) results to the on-land survey at another site
(Waranga).
• Did work at sites with a shallow watertable; and,
• Were partially successful when the watertable was located at the edge of the depth
penetration capacity of the EM31.
Further work is required in this area, but the evidence collected in this
investigation suggests on-channel EM31 surveys should only be conducted where
the geophysical technique can penetrate into the watertable, and ideally target the
top of the watertable. In other words, the method of inferred seepage based on
unsaturated zone soil properties does not appear to work on-channel. It is
apparent that the flushing effect immediately beneath the channel is dominating
changes in lithology. For EM31 systems this would preclude EM31 on-channel
use when the watertable is deeper than approximately 3-4m.
However, there is some conflicting evidence, as demonstrated by the trial results

summarised above. Overall however, the most consistent results were returned
on-land and this is considered the safest option. A possible limitation of on-land
surveying, may be at sites which contain significant land salinisation immediately
adjacent to the channel. Therefore, if the budget allows, it is recommended that
both on-land and on-channel surveys be conducted. Resistivity surveys can be
conducted on-channel because of their greater depth penetration capacity.
these surveys indicates the best off-set distance for on-land surveys is
immediately adjacent the outside toe of the down slope side of the channel. For
either method of seepage detection this is recommended. For inferred seepage
‘detection’ the soil type next to the channel is most likely to be representative of
the soil type beneath the channel. For direct measurement, immediately adjacent
the channel will be the zone of greatest seepage impact on the watertable. Away
from the channel this impact will be diluted. However at sites without a steep
gradient or high transmissivity, an average of survey traverses up to 50m on each
side of the channel was found to improve the correlation between seepage and the
geophysical survey at most sites.

Traverses on either side of the channel are recommended, however if budget is a
significant constraint, a traverse on the down-slope side of the channel should be
preferred.
q Other variables:
• Trees – In two surveys (Rocklands and Donald), tree plantations adjacent
the channel appeared to interfere with the survey results. The postulated
mechanism is that the trees are consuming the seeped water and therefore the
observed impact (in the geophysical survey) on the native groundwater is
lessened.
• Rain – Rainfall did not interfere with the surveys conducted in these trials.
However it is possible that surveys conducted after heavy rainfall on light to
moderate soil types (ie which allowed significant infiltration) could interfere
with the conductivity / geophysical response and therefore should be
avoided. Surveys inferring seepage based on shallow soil properties or
direct measurement in shallow watertable environments would be most
effected.
5.5.1.4 Repeatability
two sites where there was a significant different in the results, changes in groundwater
conditions due to channel operation accounted for the difference. These sites are
described below:
• Donald - A generally consistent increase (approximately 15 mS/m) was
observed across the surveyed area between the October 1999 survey and the
September 2001 survey. This increase was caused by the more saline
conditions at the time of the 2001 survey. The channel had been running for
six months prior to the 1999 survey, creating a sub-surface environment
dominated by fresh water and a flushed profile. The reduced channel
running time prior to the 2001 survey meant a relatively more saline profile
and hence higher conductivity.
• Dahwilly - The average EM31 conductivity for a survey conducted when the
channel was not running was less than half the conductivity recorded while
the channel was running. This is different to what was observed at Donald,
due to the different depth of groundwater at the two sites. When the channel
is not running at Dahwilly, the watertable is largely out of reach of EM31
detection and the response is a reflection of the coarse and low conductivity
sands in the unsaturated zone. When the channel is running, the watertable
is elevated into the range of the EM31 detection and hence conductivities
increase significantly.
5.5.1.5 Regional Assessment of Key Relationships

For all of the sites used in the final year of analysis an attempt was made to look for
potential correlations between seepage rates across all sites and geophysical response
(EM31 and resistivity). This was conducted using multiple linear regression and
simple linear regression. The additional explanatory variables included in the multiple
linear regression analysis included:

q Soil permeability (vertical hydraulic conductivity): Averages across 4 different
depths were used;
EM31 Multiple Linear Regression

Multi-variate regression analysis was initially conducted on the entire data set
(excluding Finley which was removed as an outlier). This indicated that the following
were significant explanatory variables:
q Upper Kv (vertical permeability of the top 2m of the profile).
However, the standard error of estimate (expressed as a percentage of the mean

pondage test seepage) for the regression was high (82%). A plot of conductivity
versus seepage dividing the data into two categories based on depth to watertable is
n Figure 5-116 Regional EM31 Assessment: Pondage Test Seepage Versus

EM31 Conductivity with Site Divided Based on Depth to Watertable
50
2
R = 0.889
40
30
20
y = -0.129x + 12.788
2
R = 0.469
10
0
0 20 40 60 80 100 120
Based on this clear division between sites with a shallow and deeper watertable (in an
attempt to improve the accuracy of the fitted regression model) the sites were split into
two data sets based on these two categories of depth to watertable. The difference in
EM31 response between a deep and shallow watertable site is explained by the effect
of the shallow and saline groundwater which for a given seepage rate causes a much
high conductivity response.
Once the data was split into these two categories, depth to watertable was not found to
be a significant explanatory variable, as would be expected. The findings for the two
categories are summarised below:

q Watertable Depth Five to Ten Metres - For sites where the watertable is 5-10m
below surface, the equation found to provide the best prediction of channel
seepage was:
Seepage = 11.6 – 0.12 EM31 + 4.4 UKv (Equation 1)

The equation:
• Was established with 40 data points;
• Has a correlation coefficient of 0.55;
• Has a standard error of estimate of 48% (of mean observed seepage rate)
conductivity of secondary importance. Groundwater salinity and depth to
groundwater were not found to be significant explanatory variables in the
analysis.
q Watertable Less Than Two Metres - For sites with a watertable within two metres
of the surface multiple linear regression analysis did not find any other variables
that were significant explanatory variables beyond EM31. The fact that soil data
was not significant is expected as the groundwater near the surface is likely to
dominate the response. It was somewhat surprising that groundwater salinity was
not found to be a significant variable for this data set, and is probably a reflection
of the limited number of sites (three) that make up the data set.
EM31 - Simple Linear Regression

represents. The data was again divided based on depth to watertable.

For the sites with a watertable 5 – 10m below surface, the best fitting linear regression
equation was found to be:

The equation:
Confidence intervals (for both 80% and 90%) for this relationship were established
and showed that the prediction equation is accompanied by quite broad prediction
bands. This probably limits the use of this regional equation to broadly classifying

seepage rates (eg into low, medium and high categories), which is not a surprising
outcome given the wide range of sites represented by the equation.
A plot of actual seepage versus predicted seepage shows that the equation tends to
overestimate seepage for low seepage rates (less than 5 mm/d) and underestimate for
high seepage rates, implying a non-linear equation may provide a better fit. An
exponential regression equation was applied to the data. The fitted exponential curve
showed that while there is only a marginal improvement in the correlation coefficient
for the exponential fit (from R2 = 0.47 to 0.48), the standard error actual worsens
(from 51 % to 62%). Therefore, overall a less accurate fit is obtained using the non-
linear equation, even though it may visually appear to fit the data better.
However the advantage of the exponential fit over the linear fit is that there is less of
a pattern displayed in the observed versus predicted seepage plots (which suggests that
the more realistic model may in fact be the exponential model). Further, the linear
model places a maximum limit on the seepage of about 12 mm/d, whereas the
exponential model appears to be more realistic, allowing for higher seepage rates in
the very low conductivity range (up to around 20-25 mm/d).
Comparing the linear regression to the multiple regression, the statistics indicate that
only a marginal improvement is made to the accuracy of the regression fit in the
multiple linear regression analysis (Equation 1), compared to the simple regression fit
(Equation 2). The R2 for Equation 1 was 0.55 and the standard error of estimate was
permeability to the regression equation.

The best fitting linear equation for sites with a watertable less than 2m is:

The equation:
The high correlation coefficient value and the relatively low standard estimate of error
suggest a good correlation for the variables. However the results should be tempered
by the fact that relatively few data points were used to form the relationship. To
improve confidence in the regression equation for the watertable less than two metres
scenario, additional points are required in the data set.
Prediction bands (80% and 90%) for estimating seepage based on EM31 response
when the watertable is less than 2m, indicate that while the prediction intervals are
broader in magnitude than for the prediction bands for the deeper watertable scenario,
as a percentage of the seepage range covered by each of the equations, they are
narrower.

In summary, multi-variate analysis did not significantly improve the regression model.
The permeability of the upper part of the soil profile was found to be a significant
parameter, but the improvements to the model with this parameter included were
marginal. The benefits of conducting field tests to collect this data are therefore
probably outweighed by the costs.
The linear regression equation for sites with a watertable five to ten metres below
surface is a reasonable fitting equation, given the range of sites on which it is based.
However it should not be relied upon to accurately predict seepage, and should be
limited to assigning seepage to low, medium and high categories. The same
comments are applicable to the regression equation developed for sites with a shallow
watertable (less than two metres). The better statistics for these sites is attributable to
the fewer data points and smaller range of environments represented.
Resistivity Multiple Linear Regression

As for the EM31 analysis, the same approach of a regional assessment using multiple
and linear regression analysis was conducted for the resistivity results. The same
additional variables were used. For the resistivity the ten metre depth slice was
adopted. While a more accurate analysis would use the depth at and just below the
watertable, for the purpose of a consistent approach, this depth slice was selected. The
first cut analysis indicated that depth to watertable was an important variable.
Therefore the analysis was again based on division of the data into sites of shallow
and deep watertable.

For sites where the watertable was five to ten metres below surface (Toolondo and
Dahwilly) the equation found to provide the best prediction of channel seepage was:

The equation:
seepage rate (to the power of 0.2). A marginal reduction in the correlation coefficient
was observed, decreasing to R2 = 0.42.
The non-linear equation, using resistivity only as an explanatory variable, for

predicting channel seepage at sites where the watertable is greater than 2m is:
5


For sites with a watertable within two metres of the surface (Lake View and Lake
View West), multiple linear regression analysis did not find any other variables that
were significant explanatory variables beyond the resistivity data.
Resistivity Simple Linear Regression

For the resistivity analysis the number of channels fitting this category was reduced to
two (3 sites at Toolondo and 2 sites on the Dahwilly Channel). Four points within this
data set appeared to be outliers. They were the four high seepage ponds at Dahwilly
Central. However they were retained in the analysis as there was no obvious grounds
for their removal.
Therefore the linear equation (using resistivity only as an explanatory variable)

predicting channel seepage at sites where the watertable is greater than 2m was found
to be:

The equation:
These statistics indicate that the accuracy of the regression is very poor, in large part
due to the four high seepage rate ‘outliers’ at the Toolondo central site. With these
outliers excluded the correlation coefficient improves dramatically to R2 = 0.63.
However as discussed above there was no obvious basis for their removal.

5

In summary, neither the linear or non-linear simple regression equations were found
to be satisfactory predictors of seepage. The very wide prediction bands for the non-
linear prediction equation confirm this conclusion.

The linear equation predicting channel seepage at sites where the watertable is less
than 2m (data set comprised of only the Lake View channel) is:
Seepage = 1.7 Resistivity10m - 0.66 (Equation 8)

The equation:
These statistics indicate that the accuracy of the regression is reasonable. However,
the results must be interpreted in light of the fact that only a few data points have been
used to form the relationship, and all data points were collected on the same channel.
Further testing to add different environments to this data set is necessary before a
reasonable degree of confidence can be placed in this prediction equation.
The following summary comments are made regarding the linear and multiple
regression analysis for the resistivity:
q The multiple regression analysis significantly improved the accuracy of the
regression equation compared to the simple regression with only one variable,
increasing the coefficient of determination from R2 = 0.21 to R2 = 0.42.
q It is likely the analysis could be significantly improved by using resistivity data at
and immediately below the watertable for each of the sites.
q As was the case in the EM31 multi-variate analysis, the variable which was found
seepage.
q While this analysis indicates a moderately fitting regression equation, it could not
q The few number of data points and small range of environments
represented by the data points;
q The unexplained outliers in the resistivity analysis that were not present
in the EM31 regression equation.

5.5.1.6 Confidence in Derived Relationships and Extrapolation of Results
geophysical response are:
1) What confidence is there that the derived relationship accurately describes seepage
within the area tested?; and
2) How confidently can the relationship be used outside of the area tested to predict
seepage?
Based on the findings of these investigations, these two issues are summarised below.
Confidence in Derived Relationships

Confidence in the derived seepage-geophysical relationship within the area tested can
be assessed by a number of factors, including:
2
q Correlation coefficient (R ) – This explains the amount of variation explained by
the regression equation. Most geophysical – seepage relationships derived in this
investigation had correlation coefficients of between 0.5 to 0.9, and typically
were around 0.75, meaning that 75% of the seepage variation could be explained
by the geophysical response.
q Standard estimate of error – This is a measure of the degree of scatter about the
regression line. A data set may have an R2 of one but a high degree of scatter.
For the regional simple linear regression (EM31) the standard error of estimate
(of mean seepage rate) was around 50% for sites with watertable 5-10m below
surface and was around 25% for sites with watertable less that 2m. For individual
channels this was generally lower at about 20%.
q Prediction interval - Prediction bands for most seepage - conductivity / resistivity
regression lines were generally quite broad. These bands suggest that often the
regression line can only be used to classify areas into high, moderate and low
seepage. The uncertainty is partially due to the data handling processes, which
are based on averaging pondage test seepage and geophysical response over long
sections of channel. There is also error inherent in the method in that given the
large number of variables that simultaneous impact on channel seepage, it is not
possible to tightly characterise seepage based on geophysical response. The
prediction intervals were generally tighter for sites with greater ranges of
seepage. Prediction intervals are likely to be improved by greater number of
pondage tests across the broadest possible seepage range.
Knowing that seepage is probably within a certain range (to a given level of
certainty), even if the range is fairly broad, is still considered an improvement on
the existing seepage knowledge base of many Rural Water Authorities, and can
only lead to more informed decision making.
q Number of data points – The number of data points on which the relationship is
derived is also a very important consideration. For instance, some sites with
relatively few data points returned very good statistics (high R2, low SEE and
relatively tight prediction bands). These results need to be taken with caution
because of the few number of points contributing to the relationship (eg, only two
data points will have perfect statistics but the results have no meaning).
Generally the more data points contributing to the relationship, the greater the
confidence in the relationship.

q Seepage range – The seepage range over which relationship is established is
important, as it improves confidence in the robustness of the relationship. A
limitation at a number of sites in this study was the narrow range of seepage rates.
The trends observed across a very tight seepage range can often be meaningless,
and at a number of sites where tests were conducted several kilometres away on
the same channel, a more realistic relationship for that channel was derived.
Confidence in Extrapolation
The following points need to be considered when extrapolating a geophysical –
seepage relationship outside of the area in which it was developed:
q Was the relationship strong in the area tested – This is the first test. If the
relationship was not strong in the area in which it was derived (refer to above
discussion) then there will be little confidence in extrapolating such a
relationship.
q How representative is the area in which to extrapolate of the tested area – The area
in which the relationship was developed must encompass the range of conditions
over which the extrapolation is to occur. This may be quite difficult to determine.
While soil, geological, hydrogeological maps and even test drilling provide an
indication of changes along the channel, the results from these trials suggest that
these are generally not at a sufficient scale to detect how they will impact on
geophysical response. For example, at the Dahwilly sites which showed
reasonably similar characteristics (depth to watertable, groundwater salinity and
lithology) the EM31 conductivity response was very different and thus
extrapolation from the Dahwilly Central site to the Dahwilly East site would have
resulted in significant errors in seepage estimate.
In fact, the evidence coming out of the trials in terms of extrapolation, even to
sites that were apparently similar and usually only several kilometres along the
same channel, was that the derived relationship was not suitable to predict
seepage at the new site. The key outcome of this is that unless intensive data
collection is conducted to ensure continuity of site conditions to the area of
extrapolation, interpolation rather than extrapolation must be used. This is
explained in further detail in the following section, but essentially means that
pondage tests should be conducted at regular intervals along the entire section of
interest, to ensure that the full range of site conditions is accounted for in the
derived regression relationship.
The more data points collected from different sections along the channel that are
added to generate the regression line, may increase scatter about the regression
line. This can be seen in the high standard error of estimates for the
Rocklands/Toolondo and Waranga regression equations. This is a reflection of
the fact that the ponds in these sites covered a wide area and range of sub-surface
conditions. However, while there is more scatter about these lines, these is more
confidence in using these relationships for seepage prediction as they encompass
a wider spatial range and wider range of conditions, and can therefore be used
with greater confidence over a broader area, although at the expense of some
degree of accuracy.
5.5.1.7 Preferred Methodology

Based on the trials conducted in this investigation, the following methodology for
using geophysics to identify and measure seepage is recommended: (Note that this

methodology assumes some prior knowledge as to which channels or sections of
channel require investigation. This knowledge may have been acquired based on flow
records, visual inspection or a regional investigation using a technique like remote
sensing).
1. Define project objective – why is the work being undertaken? The types of
questions to be asked include:
q Is the primary objective to identify relatively high seepage points or to measure
the volume of seepage?
q Is it necessary to establish the rate of seepage? – either the actual or relative rate.
q What degree of confidence in the results is required?
q Over what length of channel is the information required?
q What is the available budget?
At the end of this process there will be a clear definition of the reasons for undertaking
the seepage investigation (eg asset management), budget considerations, scale of the
operation (eg whole channel, specific channel lengths etc), need for accuracy, or
relativity. This process will effect all future decision making.
2. Collate and Evaluate Site Data - It is important that information on depth to

groundwater, background groundwater salinity, soil type and channel hydraulics
are known or gathered), both at the site where the testing is conducted, and over
the area the results are to be extrapolated. This does not have to be at a detailed
level, but should be sufficient to be able to propose a conceptual model of the
seepage mechanism, to detect where changes in these parameters may impact on
the geophysical response, and to assist in technique selection. Channel hydraulic
information is required to help determine potential channel seepage mechanisms.

one that has a depth focus on and immediately below the watertable. Whether this
is achieved using EM or resistivity is not highly important. However, generally it
is easier to focus on a given depth with resistivity (EM provides an average across
a range) and this can be achieved independent of knowledge of groundwater
depth. The advantages and disadvantages of each of the techniques (refer above)
need to be assessed in light of the specific project objectives. The recommended
technique for a given depth to watertable is outlined below:
Preferred Geophysical Techniques
q The preferred technique for geophysical channel seepage assessment is directly

detecting the impact of seepage on the groundwater. This means that the
instrument must focus on the zone immediately above and several metres below
the watertable:
• For a shallow watertable (surface to approximately 5m) EM31 is suitable for

direct seepage detection.
• For watertables deeper than 5m, EM34 (in vertical dipole mode) or
resistivity can be used. However, particularly for deeper watertables, it is
easier to focus on a given depth with resistivity and this can be achieved
independent of knowledge of groundwater depth. The significant advantage

of resistivity is that it provides a profile of the resistivity beneath the
channel. The disadvantage is that resistivity technology for channel seepage
assessment is relatively new and therefore more expensive.
q EM31 (vertical dipole) adjacent the channel can be used effectively in areas with
deeper watertables, although it does not directly measure the seepage impact on
the watertable. This is due to fact that the upper soil layers are the most influential
on channel seepage and the relatively shallow depth focus of EM31 measures
these upper soil layer properties. The method infers zones of likely channel
seepage by identifying materials in the unsaturated zone most susceptible to
seepage. A decision to use EM31 in deeper watertable area might be based on:
• Cost and required accuracy – If a potentially slightly lower level of accuracy

is considered acceptable then EM31 represents a cheaper alternative than
EM34 or resistivity; or,
• Lack of alternatives – EM34 or resistivity contractors are not readily
available.
It this method is used however, it must be made certain that seepage is controlled by
the unsaturated zone and not surface clogging processes. Otherwise errors will
potentially be introduced to the assessment process.
The preferred geophysical techniques for seepage detection are summarised in Table
5-7.

n Table 5-7 Recommended Geophysical Technique for Seepage Detection
and Measurement

Depth Focus (m)4

(vertical dipole)6
OR
5 – 12
Resistivity7 Direct watertable impact NA10 NA11
OR

(vertical dipole)6
OR
12 – 25
NA10 NA11
Resistivity7 Direct watertable impact
OR

> 25
OR
1. It is recommended EM techniques are conducted adjacent the channel (additional survey

runs can be conducted away from the channel). Resistivity surveys should be conducted
on -channel.
2. Direct detection of seepage impacts on the watertable is the recommended technique, but
inferred ‘detection’ based on soil property variations will often provide an adequate
simulation and may be more convenient for various reasons (refer to body of report for
potential errors associated with this method)
3. Approximate detection of penetration is referred to in the Geonics manual (McNeil, 1980)
as the effective depth of exploration. This is the depth to which approximately 75% of the
response is attributed.
4. The ‘depth focus’ is a term used in this report to describe the depth (range) which is most
influential in terms of the relative contribution to the overall EM response (McNeil,1980).
5. These can be conducted immediately adjacent to the channel or on-channel. Both are
recommended if budget allows. If on-channel is used for a watertable of 0-1.5m, the
survey should preferentially collect data in vertical dipole mode where the effects of
channel water will be less influential. For sites with a watertable 0-1.5m, EM31 on channel
may be preferred if significant land salinisation exists adjacent the channel.
6. Horizontal and Vertical Dipole: Note that as applied to EM34, vertical dipole does not
refer to the coil orientation with respect to the ground, and is in fact opposite to the coil
orientation. In vertical dipole mode the coils should be horizontal to the ground, which is
a slower method than horizontal mode where they are held perpendicular to the ground.
7. Resistivity is the preferred direct measurement technique for this depth to watertable but
EM34 is provided as a potentially more accessible alternative.
10. The penetration depth of resistivity depends of the particular system set up (dipole spacing
and length).
11. Resistivity surveys measures resistivity at a range of depths intervals within the profile (ie,
there is no depth focus).

4. Conduct geophysical survey – Undertake the geophysical survey over the section
of interest, giving due consideration to factors such as appropriate timing of the
survey and other important variables (refer section 5.5.1.3).
5. Evaluate results – Plot geophysical survey results along the section and overlay
with known site conditions (soils, geology, hydrogeology and channel hydraulic
data). Based on these plots identify areas of suspected high, low and moderate
seepage, assuming low conductivity / high resistivity equates to higher seepage.
6. Conduct test drilling – Soil bores should be drilled at appropriate intervals along
the length of the geophysical survey. The primary aim of the drilling is assist with
interpretation of the geophysical survey. Some key principles of the drilling
program are described below:
q Based on the geophysical survey results, conduct drilling across a range of
low, moderate and high conductivity / resistivity sites;
q Drill at least some bores into the watertable, and construct some as
permanent observation bores;
q Generally drilling should be conducted on the outside toe of the channel;
q Logging and sampling of the bores should ideally be undertaken by someone
trained in soil / geological classification and a consistent classification
system should be followed.
q Depending on the density of data collected, presenting the results in a
geological long section should be considered.
7. Conduct pondage tests - At appropriate intervals over the entire test area pondage
tests are to be conducted. The number of tests will depend on the length of
channel surveyed and the variability of conditions along the channel. The
following guidelines are suggested:
q Pondage tests should be conducted across a range of low, moderate and high
conductivity / resistivity sites so as to establish a regression equation which
represents the range of geophysical response across the area.
q Similarly, based on the soil drilling results, the pondage tests should be
based on a range of different soil types and / or groundwater conditions.
q Pondage tests must be conducted over areas of like conductivity / resistivity.
That is they should not staddle areas of (significantly) different geophysical
response, as this will complicate interpretation of the results and
development of the regression equation.
q Due to the cost of conducting pondage tests, it is recommended that at least
two cells back to back should be conducted at each site for efficiency
purposes. Using available structures should also be considered to minimise
bank construction costs.
q The pond length can be variable, but as a guide they should generally not be
more than 400-500m and not less than 50m.
By conducting pondage tests in this manner across the area of the geophysical
survey, prediction of seepage rates outside of pondage test areas will be based on
interpolation rather than extrapolation, which improves confidence in the
predicted seepage. While pondage tests are expensive, they are a critical part of
the interpretation process.

8. Develop and evaluate relationship between seepage and geophysical response –
The following key steps should be conducted:
q Plot geophysical response against pondage test seepage.
q Outliers in particular should be assessed in light of all available information,
including the conceptual seepage mechanism, test drilling results, channel
hydraulics etc. If there are legitimate grounds for excluding outliers they
should be removed.
q If from this data two or more different trends can be observed due to
identifiable differences in sub-surface conditions, then two different
regression equations should be generated.
q Fit a regression line through the data.
q Statistical analysis should be conducted to determine the degree of
confidence that can be placed in the derived relationship
q Using the derived relationship the channel length should be divided into
seepage categories of various seepage rates based on geophysical response,
with accompanying error estimates.
9. Evaluation - Evaluate whether investigation objectives have been met. One of

the key questions to address is whether there is sufficient confidence in the
derived relationship. In addition to the particular statistics of the regression line,
this will largely depend on the project objectives. Further pondage tests or other
testing may be required to further improve confidence in the relationship.
5.5.1.8 International Developments in Geophysics and Channel Seepage

Measurement
Since the writing of the Literature Review conducted as part of this project (ANCID,
2000a), several papers have been published relating to international developments in
channel seepage measurement using geophysics. Two key papers are briefly
summarised below. The important point relating to this work is that it is focussed in
the same direction as the geophysical investigations in these trials: developing
geophysical techniques that can be compared to some form of direct seepage
measurement, derivation of a relationship between the two and then extrapolation /
interpolation to new areas.
Determining Irrigation Canal Seepage with Electrical Resistivity (Hotchkiss et al,

2001)
Summary of Abstract: Procedures were developed and tested for quantifying seepage
losses in unlined irrigation channels reaches on the order of 30m, in the Central
Nebraska Public Power and Irrigation District canal system. The procedure uses
electrical resistivity (ER) measurements while canals are in service to determine the
resistivity of the underlying clay layer. ER data were correlated to canal depth and
then to seepage rate. Seepage rates were determined using seepage meters. Accuracy
is approximately ± 20%, comparable to that achieved using stream gauges methods (±
5% error of total canal discharge). The ER approach, however can easily pinpoint

seepage zones more precisely, allowing a reduction in the length of canal lining
projects.
Comparisons to this investigation:

q The method used in the Hotchkiss et al (2001) study relies on measuring
unsaturated zone soil properties and not seepage directly, ie it is a method which
infers the likelihood of seepage. It is based on the theory that ‘as resistivity
increases, seepage rates will also increase because more sandy or less clayey
materials will have higher resistivities’ (Hotchkiss et al, 2001). This was also
shown to be successful in the ANCID investigations project (eg EM31 at deep
watertable sites). This is a valid method but potentially has limitations at some
sites.
q Seepage meters rather than pondage tests were used as the direct seepage
measurement technique (possibly limited by operational requirements, ie could
not shut channel down). Confirming the general unreliability of seepage meter
readings, 50% of sites where readings were conducted were discarded as
unreliable. This highlights the advantage of pondage tests over seepage meters.
The only advantage of the seepage meter is that smaller length areas can be
selected; 30m used in Hotchkiss et al (2001) study. (It is not recommended
pondage tests be conducted over less than 50m of channel).
q Best correlations were found at 4m depth – similar to this project. Soil properties
from the upper part of the profile were found to be most influential on seepage
rates.
q Due to considerable depth and variation in channel water, the data first had to be
correlated to canal depth and then to seepage rate, whereas in this ANCID study,
the depth of channel water was reasonably consistent and this step was not
required.
5.5.2 Summary of EM34 Results
Table 5-8 presents a summary of all EM34 trials conducted in the program compared
to pondage test seepage. All trials were conducted in horizontal dipole mode, at a
10m coil spacing and along the outside toe of the down slope side of the channel. The
exception was the two Dahwilly sites where both 10m and 20m coil spacings were
used, and survey runs along the outside toe of both sides of the channel were
conducted.


q Donald Main – A moderate relationship was recorded in both surveys. Further
points are required in the mid-seepage range to appropriately test the relationship.
The technique distinguished between high and low seepage but not within the
high seepage results range. Possible interference by adjacent trees may have
effected results in some ponds. A generally consistent increase in conductivity
q Toolondo Central – A moderate relationship was observed but largely skewed
by the result in one pond. The relationship distinguished between high and low
seepage rates.
which was largely due to the narrow seepage rate range. At the Toolondo central site,
where conductivity measurement was entirely above the watertable, the unsaturated
zone lithology was a sufficiently accurate indicator of seepage and hence a reasonable
compared to the EM34 (R2=0.73 compared to R2=0.50), but neither survey

n Table 5-8 Summary of EM34 Trials Compared to Pondage Test Seepage
Channel Survey Survey Survey Channel No. of Range of Correlation Water- Seepage Summary on Seepage - Geophysical Use of Repeatability Comparison with other Reported
Type Date Location Operating Pondage seepage Coefficient table Measurement Relationship relationship geophysical techniques in Section
Cells results Depth Mechanism for
(approx.) extrapolation
Rocklands EM34, 10m Nov. 1999 Down Yes 6 4 - 13 mm/d 0.79 5m Inferred from unsat. Good Relationship: A good spread of - Refer below (Aug. Adjacent channel results 5.2.2.1
coil gradient zone soil properties & seepage results, relatively low scatter about 2001 survey) very similar to EM31
separation, outside dilution of g'water the regression line and a good correlation correlations
horizontal channel toe with seeped water coefficient.
dipole
Rocklands EM34, 10m Aug. 2001 Down Yes 6 4 - 13 mm/d 0.75 5m Inferred from unsat. Good Relationship: A good spread of - High degree of - 5.2.2.1
coil gradient zone soil properties & seepage results, relatively low scatter about repeatability
separation, outside dilution of g'water the regression line and a good correlation demonstrated
horizontal channel toe with seeped water coefficient.
dipole
Donald EM34, 10m Oct. 1999 Down Yes 6 9 - 45 mm/d 0.43 2m Primarily dilution of Moderate Relationship: Good range of - Refer below (Sept. A slightly better 5.2.2.2
Main coil gradient (commenced groundwater from seepage results but moderate to poor 2001 survey) relationship was established
separation, outside operation 4 fresher seepage water correlation coefficient. Further points with EM31(v) adjacent the
horizontal channel toe weeks prior to required in mid seepage range. channel (R2=0.73), but still
dipole survey) Distinguished between high and low no differentiation between
seepage but not within the high seepage the higher seeping ponds
results. Possible interference by trees
Donald EM34, 10m Sept. 2001 Down Yes 6 9 - 45 mm/d 0.50 3m Primarily dilution of As above for Oct. 1999 survey - A generally consistent - 5.2.2.2
Main coil gradient (commenced groundwater from increase (15-20
separation, outside operation 4 fresher seepage water mS/m) observed
horizontal channel toe weeks prior to compared to Oct.
dipole survey) 1999 survey.
Regression line
shifted vertically (but
gradient similar).
Diff. attributed to
higher WT & reduced
channel running time
Toolondo EM34, 10m Aug. 2001 Down Yes 6 1 - 11 mm/d 0.50 10m Primarily inferred Moderate Relationship: Good range of - - An improved relationship 5.2.2.3
Central coil gradient from differences in seepage results but moderate correlation obtained with EM31,
separation, outside unsat. zone soil coefficient - largely skewed by the result in indicating method of
horizontal channel toe properties one pond. Distinguished between high and detecting lithology changes
dipole low seepage rates. in the unsaturated zone has
worked at this site. For
resistivity improved co-
efficients were obtained for
10m & 12m slices, due to
better penetration into top
of saturated zone
Dahwilly EM34, 10m Feb. 2002 Adjacent both Yes 7 1.1 - 9.5 0.79 5m Primarily inferred Moderate Relationship: Good correlation This - Similar strength relationship 5.2.2.4
Central coil outside (only 5 mm/d (4.2 - from unsat. zone soil coefficient - but very low range of relationship observed for EM31, with
separation, channel toes used in 9.5 mm/d: properties & conductivity response across five ponds would not have better spread of
horizontal analysis) used in secondarily dilution used in the analysis. The EM34(10m) successfully conductivity results.
dipole analysis) of g'water configuration does not appear to have quantified rates Stronger resistivity results
penetrated to sufficient depth to at Dahwilly East obtained at 8m & 10m
significantly detect changes in the saturated - would have depth slices due to targeting
zone and was therefore mainly measuring resulted in of most effected area of
differences in the unsaturated zone, which significant WT.
is largely uniform at the Dahwilly site. underestimation
Dahwilly EM34, 20m Feb. 2002 Adjacent both Yes 7 1.1 - 9.5 0.04 5m Primarily dilution of No relationship: The EM34(20m) NA - No - - 5.2.2.4
Central coil outside (only 5 mm/d (4.2 - groundwater from configuration appears to have penetrated relationship
separation, channel toes used in 9.5 mm/d: fresher seepage water too deeply into the profile, below the upper observed
horizontal analysis) used in zone in the watertable aquifer where
dipole analysis) seepage effects are most prominent. The
uniform response observed is therefore
essentially a reflection of the salty native
groundwater

Channel Survey Survey Survey Channel No. of Range of Correlation Water- Seepage Summary on Seepage - Geophysical Use of Repeatability Comparison with other Reported
Type Date Location Operating Pondage seepage Coefficient table Measurement Relationship relationship geophysical techniques in Section
Cells results Depth Mechanism for
(approx.) extrapolation
Dahwilly EM34, 10m Feb. 2002 Adjacent both Yes 3 8.9 - 10.4 0.26 5-6m Inferred from unsat. Poor relationship: Seepage rate range is too - - No relationship observed 5.2.2.4
East (Pretty coil outside mm/d zone soil properties & narrow to draw meaningful conclusions for EM31. A sensible
Pine) separation, channel toes dilution of g'water relationship was observed
horizontal with seeped water for the resistivity 6m depth
dipole slice.
Dahwilly EM34, 20m Feb. 2002 Adjacent both Yes 3 8.9 - 10.4 0.24 5-6m Primarily dilution of Poor relationship: Seepage rate range is too - - - 5.2.2.4
East (Pretty coil outside mm/d groundwater from narrow to draw meaningful conclusions
Pine) separation, channel toes seepage
horizontal
dipole

Table 5-9 presents a summary of all EM31 trials undertaken in the program compared
to pondage test seepage. Trials on land were conducted in vertical dipole mode and
generally included 3-4 runs along the outside of the channel, with the outer runs
approximately 50m from the channel. The exception was the Waranga Western
Channel where due to the length of the survey only one line on the outside toe of each
bank was conducted. Trials conducted on-channel were undertaken in both vertical
and horizontal dipole.
depth of penetration is around 6-7m, with a mid-range depth focus of about 2 – 4.5m
(refer 5.3.2.1). This meant that at sites where the watertable was deeper than 5m, only
a limited proportion of the response was caused by seepage impacts in the saturated
zone. Therefore at these sites the seepage detection mechanism is largely via
inference based on soil properties in the unsaturated zone. Key summary comments
for each of the sites are listed below:
Toolondo
G Good relationships between EM31 conductivity and pondage tests seepage were
G A high degree of repeatability between the surveys was observed.
G In-channel (shortly after channel shut down) and on-channel EM31 surveys
returned poor results. This is attributed to the fact that an EM31 survey above the
watertable ‘works’ by inferring seepage based on soil properties. However
immediately beneath the channel, even for low seepage rate ponds the profile
uniform saturation produces a uniform conductivity response, and tends to mask
changes in lithology resulting in little differentiation between low and high
G Better results were obtained with the EM31 compared to the EM34(10m) at this
from the channel).
G Three Toolondo Sites (Central, East and West) - The relationship established for
all sites was moderately strong. Local correlations at Toolondo Central and
Toolondo West were stronger than the combination of sites. The Toolondo East
site displayed an opposite correlation, but the very narrow range of seepage rates
& the flat regression line indicates this is not a meaningful trend. Confidence
bands for the overall regression relationship are wide but indicate that the

G If the Toolondo Central site had been used to predict seepage at Toolondo West,
would have been essentially accurate (0 mm/d), except in one pond seepage
would have been predicted at 4 mm/d when actual seepage is practically zero.
Rocklands
G A good relationship was observed between EM31 response and pondage test
seepage at the Rocklands channel trial site (for the adjacent channel EM31 data).
This indicates that seepage was able to be successfully inferred based on
unsaturated zone soil properties. However, with a depth to watertable of around
G A poor response was observed when all survey runs were used, largely due to the
G The on-channel results recorded mixed results. In vertical dipole mode no trend
was observed. The configuration is focussed on the flushed zone beneath the
channel where uniform saturation from seepage appears to be masking lithology
response. In horizontal dipole a reasonable correlation was observed, apparently
seepage areas.
Donald Main
G A good relationship was observed between EM31 conductivity and pondage test
seepage but there is a poor spread of seepage data at the site (1 point of low and 5
of high seepage). With a relatively shallow watertable (2m), the EM31 detects
seepage at this site in terms of its impacts on the watertable. The EM31 survey
did not distinguish between higher seepage ponds (35 - 48mm/d). Confidence
bands are fairly wide for the regression line, particularly at the high conductivity
range, but indicate that the relationship can differentiate between high and low
G A better relationship was established with EM31 (R2=0.71) adjacent the channel
G Moderate to good relationships were also observed for the on-channel surveys in

Dahwilly
G For a survey conducted when the channel was not running, no relationship was
observed between EM31 conductivity and pondage test seepage. The technique
failed because the channel was not running and previously seeped water was
therefore likely to have thoroughly mixed with native groundwater. Unsaturated
zone lithology is a good indicator of seepage at some sites. However at Dahwilly
it is not the unsaturated zone controlling seepage rates, but the clogging layer at
the channel surface and therefore seepage must be detected directly (ie in terms of
impact on watertable) which means the channel must be in operation.
G In a repeat survey conducted when the channel was operating, a good relationship
G Two Dahwilly Sites (Central and East) - The relationship established for both
sites is moderately strong. Local correlation at the Central site is slightly stronger
than the two sites combined. The East site displays a very weak correlation, but
this is due to the very narrow seepage range and few data points at this site.
Confidence bands are relatively wide, suggesting the regression relationship for
both sites can only be used to broadly indicate the likelihood of low or moderate
seepage. The slightly deeper depth to watertable at the East site appears to have
put the watertable largely beyond the range of EM31 and hence very different
results are obtained at the East site. Using the Central site regression relationship
to predict seepage at the East site would have resulted in over prediction of 1.5 - 2
times actual seepage.
G Better correlations at both sites were obtained using the resistivity compared to
EM31 due to better targeting of the top of the watertable.
Lake View
G A poor relationship between pondage test seepage (July 2001) and EM31
adjacent channel data a significantly improved relationship (to moderate) was
observed as seepage impacts are less diluted. Interpretation is limited at this site
due to the very narrow seepage rate range. Seepage is detected at this site in
terms of its impact on watertable salinity.
G No sensible trend was observed at the Lake View Central site using the same
G Both Sites (Central and West) - The relationship established for both sites is
moderately strong with a high correlation coefficient but the two data sets
creating the regression line have small conductivity and seepage rate ranges. It is
desirable to obtain data in the mid range to improve confidence in the
West site. However using Central data from adjacent the channel is likely to

Tabbita
G No relationship was observed between EM31 conductivity and pondage test
seepage. Possible reasons for the failure of the technique at this site include:
v) Narrow range of seepage rates (little differentiation in rates along
section of interest);
vi) Seepage mechanism may be such that majority of seeped water does not
vii) EM31 vertical dipole orientation may penetrate too deeply into the native
viii) The method of averaging conductivity may not be appropriate at this site
(or the ponds may need to be placed more carefully, ie over shorter sections
of high/low conductivity)
Finley
G While a moderate correlation coefficient was obtained for the pondage test –
EM31 conductivity relationship at this site (and the highest seeping pond did
record the lowest conductivity), the statistics are not meaningful due to the fact
that only three data points make up the relationship. The width of the prediction
Waranga Western Channel

G A moderate to poor relationship was recorded between EM31 conductivity and
pondage test seepage. However the results should be considered in light of the
fact that they represent seepage sites more than 20 km apart (significantly further
than other sites). The watertable is generally beyond the penetration depth of the
EM31 along the survey reach and therefore the likelihood of seepage is inferred
based on soil properties beneath the channel.
G Some of the scatter in the results can be explained by incorrect pond placement
(ie not straddling areas of like conductivity) as well as some geological
anomalies. When points of high variance and leverage are removed, the
correlation coefficient improves to 0.62 (from 0.40). The prediction interval
bands suggest the relationship can be used to distinguish between sites of low and
high seepage, but is limited in interpreting mid-range seepage.
G An improvement would be expected if the top of the watertable was targeted,
rather than inferring seepage from unsaturated zone soil properties. Given that
the ponds are significantly spaced apart, this relationship can be used for
interpolation, bearing in mind the associated broad prediction intervals associated
with the regression line.
In summary, the only site where no relationship was observed was at Tabbita. A
number of possible causes for this were identified, but the predominant contributing
factor is not known. At two sites (Rocklands and Lake View Central), the adjacent
channel data was used instead of all survey run data. This was required to obtain the

At the Toolondo central site, where conductivity measurement was entirely above the


n Table 5-9 Summary of EM31 Trials Compared to Pondage Test Seepage
Channel Survey Survey Survey Channel No. of Range of Correlation Standard Water- Seepage Summary on Seepage - Geophysical Use of Repeatability Comparison with other Reported
Type Date Location Operating Pondage seepage Coefficient Error of table Measurement Relationship relationship for geophysical techniques in Section
Cells results Estimate Depth Mechanism extrapolation
(approx.)
Toolondo EM31, Dec. 4 traverses No (1-2 6 1 - 11 All data: 0.72; - 10m Inferred from All Data / Channel sides: Good Relationship – - Refer below Better results obtained than 5.3.3.1
(Central) vertical 2000 each side weeks after mm/d In-channel unsat. zone soil good spread of seepage results, moderate scatter (Aug. 2001 the EM34(10m), possibly
dipole of channel channel data: 0.04; properties about regression line and a good correlation survey) due to the greater number of
(up to stopped Channel sides: coefficient. In-channel data: No relationship. traverses conducted (ie
50m) & flowing) 0.73 Immediately beneath channel soils are uniformly away from the channel).
One in saturated (even for low seepage ponds) which in Similar results achieved
channel turn produces a uniform conductivity response. using resistivity (10m/12m
(Outside channel results predominantly measure depth slices) but this detects
unsat. zone soil properties) actual seepage impacts in
the watertable
Toolondo EM31, Aug. 4 traverses Yes 6 1 - 11 All data: 0.81 - 10m Inferred from All Data: Good Relationship - good spread of - High degree of Refer above 5.3.3.1
(Central) vertical 2001 each side mm/d Adj. channel unsat. zone soil seepage results, moderate scatter about regression repeatability of
dipole of channel only: 0.80 properties line and a good correlation coefficient. results
(up to compared to
50m) Dec. 2000.
Correlation
coefficient
between surveys
of 0.96 for
average
pondage cell
response
Toolondo EM31, Aug. On- Yes 6 1 - 11 Horizontal - 10m Inferred from No relationship / Very poor relationship: Even for - - Significantly, the on- 5.3.3.1
(Central) vertical 2001 channel mm/d dipole: 0.21 unsat. zone soil low seepage rate ponds the profile beneath the channel resistivity recorded
dipole Vertical and water channel is saturated with seeped water. This much better correlations
& Dipole: 0.05 content uniform saturation produces a uniform (below the watertable) than
horizont properties conductivity response. Lithology effects appear the on-channel EM31. The
al dipole to be largely masked by seepage saturation at this EM31 could not 'see' below
site the watertable whereas the
resistivity could target this
zone
Rocklands EM31, Aug. 4 traverses Yes 6 4 - 13 All data: 0.33 - 5m Inferred from Poor relationship for 'All Data' largely due to - - Adjacent channel results 5.3.3.2
vertical 2001 each side mm/d Adj. channel unsat. zone soil effect of trees adjacent one pond. Therefore, best very similar to EM34
dipole of channel only: 0.82 properties & relationship observed for 'Adjacent Channel' data correlations
(up to dilution of (most distant from trees): Good spread of
50m) g'water with seepage values, moderate scatter about regression
seeped water line and good correlation coefficient
Rocklands EM31, Aug. On- Yes 6 4 - 13 Horizontal - 5m Inferred from Horizontal dipole: Reasonable correlation - - - 5.3.3.2
vertical 2001 channel mm/d dipole: 0.78 unsat. zone soil observed through identification of lithology
dipole Vertical properties & changes (clay content) immediately beneath
& Dipole: 0.15 dilution of channel. Did not distinguish between high rates
horizont g'water with however. Vertical Dipole: No trend observed.
al dipole seeped water Configuration is focussed on flushed zone
beneath the channel (immediately above
watertable) where uniform saturation from
seepage (in clays and sands) appears to be
masking lithology response. (Also poor
resolution of meter at low conductivities)
Donald EM31, Sept. 4 traverses Yes 6 9 - 48 All data: 0.73 23% 1.5m Primarily Moderate to good relationship: Good correlation - - A slightly better 5.3.3.3
Main vertical 2001 each side (commence mm/d Adj. channel (downslope) dilution of coefficient, acceptable scatter about regression relationship established and
dipole of channel d operation only: 0.69 ; 3m groundwater line (SEE = 23%) but poor spread of seepage data with EM31adj. the channel
(up to 3 weeks (upslope) from fresher (1 point of low & 5 of high seepage). Did not compared to EM34 5.3.5.3
50m) prior to seepage water distinguish between higher seepage ponds (35 - (R2=0.50) but still no
survey) 48mm/d). Confidence bands wide, particularly at differentiation between the
high conductivity range, but can differentiate higher seeping ponds.
between high & low seepage sites. Additional Slightly improved
data points required to tighten confidence bands. relationship probably due to
greater depth focus of
EM31 (particularly on

(approx.)
upslope side)
Donald EM31, Sept. On- Yes 6 9 - 48 Horizontal - 1.5m Inferred from Moderate to good relationship for both horizontal - - 5.3.3.3
Main vertical 2001 channel (commence mm/d dipole: 0.83 (downslope) unsat. zone soil & vertical dipole: Good correlation coefficient
dipole d operation Vertical ; 3m properties & but some scatter about regression line & poor
& 3 weeks Dipole: 0.77 (upslope) dilution of spread of seepage data (1 point of low seepage
horizont prior to g'water with and 5 of high seepage). Did not distinguish
al dipole survey) seeped water between higher seepage ponds (35 - 48mm/d).
Data suggests exponential or other non-linear
trend line may provide better fit
Dahwilly EM31, June. 4 traverses No (Ceased 6 4 - 16 0.49 - 5-6m Inferred from No relationship observed. Technique failed at - - EM34 'worked' because the 5.3.3.4
Central vertical 2000 each side to operate mm/d (opposite unsat. zone soil this site because the channel was not running at channel was running
dipole of channel several correlation) properties the time (previously seeped water therefore likely
(up to weeks prior to have mixed in with native groundwater) and it
50m) to survey) primarily targeted the unsaturated zone in any
case. While unsaturated zone lithology is a good
indicator of seepage at some sites, at Dahwilly it
is not the unsaturated zone controlling seepage
rates, but the clogging layer at the channel
surface.
Lake View EM31, June. 4 traverses Yes 5 7 - 9 All data: 0.34 - 1.5m Primarily Poor relationship for All Data due to rapid mixing - - Similar R2 to Lake View 5.3.3.5
Central vertical 2000 each side mm/d Adj. channel dilution of of seepage plume away from channel, however Resistivity, March
dipole of channel only: 0.58 groundwater for Adjacent Channel data this improves to 2002(Res) v July 2001(PT):
(up to from fresher moderate as seepage impacts are more 0.58 & 0.60.
50m) seepage water discernible. Interpretation limited at this site due
to the very narrow seepage rate range (7-9
mm/d).
Tabbita EM31, June. 4 traverses Yes 6 6 - 10 All data: 0.01 - 1m Primarily No relationship observed. Possible reasons for - - - 5.3.3.6
vertical 2000 each side mm/d dilution of technique failure at this site: i) Narrow range of
dipole of channel groundwater seepage rates (little differentiation in rates along
(up to from fresher section of interest), ii) Seepage mechanism may
50m) seepage water be such that majority of seeped water does not
reach watertable but move laterally (evaporating
and causing salinisation), iii) EM31 vertical
dipole may penetrate too deeply into native gw,
below diluted zone, iv) method of average
conductivity ,may not be appropriate at this site
(or ponds need to be placed more carefully - over
shorter sections of high/low conductivity)
Toolondo EM31, Rockl'ds 4 traverses Yes Rockl'ds: 0 - 12 Rockl'ds: 0.33 All Sites: Rockl'ds: Inferred from The relationship established for all sites is If the Toolondo - West: Correlations of the 5.3.5.2
(Central, vertical : Aug. each side 6 mm/d (0.84 adj.) 61% 5m unsat. zone soil moderately strong. Local correlations at Central site had 2m & 4m resistivity depth
East & dipole 01 of channel Toolondo: Toolondo: Toolondo properties (& at Rocklands (adj.), Tool. Central, & West are all been used to slices with PT seepage
West) and Toolond (up to Central - Central - 0.78, (Central, Rocklands some stronger than the combination of sites. The Tool. predict seepage at: match with EM31
Rocklands o 50m) 7, East - 4 East - 0.49 East & dilution of East site displays an opposite correlation, but the Tool. West - correlations, which are
(Central, and West (opposite) and West): 10- g'water with very narrow range of seepage rates & the flat predicted seepage focussed in the 1-4m depth
East & - 4 Total - West - 0.86 12m seeped water) regression line indicates this is not a meaningful would have been range East: no meaningful
West): 21 All Sites - trend. Confidence bands are wide but indicate 2-3 times too correlations obtained with
March 0.47 that the regression relationship can be used to high, Tool. East - resistivity (too narrow
2002 differentiate between high & low seepage sites. essentially seepage range) Central:
The data most influential on the low R2 and wide accurate (0 mm/d) See above
confidence bands is the 4 sandy ponds at except one pond
Toolondo Central - it is apparent the shallow would have
depth of the sand causing the seepage is largely predicted 4mm/d
missed by EM31(vertical) when actual is 0
mm/d
Dahwilly EM31, Feb. 4 traverses Yes Central - 5 4 - 10 Central: 0.64 All Sites: Central: 5m Inferred from The relationship established for combined sites is Using Central site EM31 was Better correlations at both 5.3.5.4
(Central & vertical 2002 each side and East - mm/d & East: 0.15; 21% & East: 6m unsat. zone soil moderately strong. Local correlation at Central to predict seepage conducted in sites obtained using the
East) dipole of channel 3; Total - All Sites: 0.58 properties & site is slightly stronger than the 2 sites combined. at East site: over 2000 adjacent to resistivity compared to
(up to 8 dilution of The East site displays a very weak correlation, pred'n of 1.5 -2 the Central site. EM31(vert) due to better
50m) g'water with but this is due to the very narrow seepage range times actual No trends were targeting of the top of the
seeped water and few data points at this site. Confidence bands seepage, due to observed as the watertable. Similar results
are relatively wide, suggesting regression slight diff. in channel was not obtained for EM34, but

(approx.)
relationship for both sites can only be used to depth to WT running. In this EM34 displays wider
broadly indicate the likelihood of low or (caused EM31 to survey (2002) conductivity range (due to
moderate seepage. Slightly deeper depth to largely miss with the channel deeper focus)
watertable at East site appears to have put seepage impact on running, the
watertable largely beyond EM31(vertical) range gw at East site) EM31 (vertical)
and hence very diff. results obtained at East site. and gw salinities detected
between the two seepage impacts
sites. on the
watertable and
hence trends
and correlations
were observed.
Finley EM31, Feb. 4 traverses Yes 3 5 - 7 0.47 18% 1.5m Primarily Statistics not meaningful due to low no. data - - Better correlations obtained 5.3.5.5
vertical 2002 each side mm/d dilution of points, (did detect area of higher seepage at 4m using resistivity than
dipole of channel groundwater however as conductivity low) the EM31(vert) due to
(up to from fresher Confidence intervals indicate regression better targeting of the top of
50m) seepage water relationship cannot be used to predict seepage at the watertable
this site
Extremely high conductivity site – very clayey
profile suggests bank leakage rather than soil
seepage is the dominant seepage mechanism
Additional data points and wider seepage
range required to improve relationship
Lake View EM31, Central: 4 traverses Yes Central - 4 4 - 25 Central: 0.46 17% Central: Primarily Good R2, SEE & prediction interval but 2 data LV Central site - West: The fact that a 5.3.5.6
(Central & vertical June each side & West - mm/d (opposite) & 1.5m & dilution of sets creating the regression line have small could not have sensible correlation was not
West) dipole 2000 of channel 4; Total - West: 0.54; West: 0.5m groundwater conductivity and seepage rate ranges – Desirable been used to observed for shallow
West: (up to 8 All Sites: 0.94 from fresher to obtain some data in mid range to improve predict LV West resistivity data but was for
May 50m) seepage water confidence seepage. However EM31 suggests poor
2002 No sensible trend at LV Central site – however LV Central resolution of surface data by
important to note that the same EM31 data at LV adjacent data the resistivity array at this
Central when applied to 2001 PTs returned a likely to improve site. Central: Refer above
sensible inverse correlation - this correlation was correlation (refer
further improved when the 2 survey lines adj. 'Summary on
channel were used (refer above). It is anticipated Seepage…..'
a better response could be obtained with the column). LV
current PTs using the adj. channel data (this has West data
not be done however for consistency) reasonably
predicts Central
seepage
Waranga EM31, Nov. Traverse Yes 11 1 - 13 0.40 63% Variable Primarily Moderate to poor R2, SEE & prediction Given that the - - 6
vertical 2001 on each mm/d (approx. inferred from interval - however these results should be ponds are
dipole side of 8m) unsat. zone soil considered in the light of the fact that they significantly
channel properties represent seepage sites more than 20 km apart. spaced from each
(outer toe) Some of the scatter in the results can be other, this
& on the explained by incorrect pond placement (ie not relationship can
channel straddling areas of like conductivity) as well as be used for
some geological anomalies. When points of high interpolation,
variance & leverage removed, R2 improves to bearing in mind
0.62. Prediction interval bands suggest the associated
relationship can be used to distinguish b'tw.n sites broad prediction
of low & high seepage, but is limited in intervals
interpreting mid-range seepage sites associated with
Improvement expected if top of WT targeted the regression
line.

5.5.4 Summary of Resistivity Results
Table 5-10 presents a summary of all resistivity data collected in the program
compared to pondage test seepage. The depth at which the best correlation was
recorded is presented. Good relationships were obtained between average resistivity
(from depth slices immediately below the watertable) and the corresponding pondage
test seepage at most sites. Key summary comments for each of the sites are listed
below:
Toolondo
G Central – At around 10-12m the best correlation was obtained (R2=0.6), which is
the zone immediately below the watertable and fits with the expected mechanism
of seepage detection (ie, in the depth interval of groundwater most effected by
seepage). Within individual ponds, the resistivity cross sections show sub-
sections of localised higher seepage. There is very little correlation in the
unsaturated zone.
G East - The very narrow range of seepage rates and resistivity values meant that no
meaningful correlations were observed at the Toolondo East site.
G West - At and below the watertable the expected inverse trend between high
resistivity - low seepage and low resistivity - high seepage was not observed. It is
apparent that the sandstone at this site may be dominating the response. Low
permeability sandstone may be causing a high resistivity response (normally
associated with high seepage), and masking the effect of seepage on the saline
groundwater. Deeper drilling would be required to confirm this interpretation.
The reasonable correlations obtained at shallow depth (2m and 4m depth slices)
are most likely due to changes in clay content beneath the channel, and
corresponds with observed EM31 relationships.
G All Sites – The Toolondo East data (av. 10/12/14m depth slices) lies within the
expected extrapolated range for the Central (10m) site. The West (av 10/12m
depth slices) data does not fit within the Central and East relationship (possibly
due to changes in lithology masking seepage impacts). The West 2m depth slice
does fit within the relationship, however without pondage tests it would not have
been known that this was the better depth on which to focus.
Dahwilly
G Central - Good correlations were observed at the 6m, 8m and 10m depth slices,
which fits with the expected mode of seepage detection below the watertable.
Correlation coefficients worsen in the unsaturated zone. Resistivity long sections
indicate that seepage is generally diffuse across the surveyed area, in contrast to
Toolondo where localised seepage is evident from the resistivity data.
G East – A strong correlation was observed at 6m which corresponds with the top of
the watertable. However, the very narrow range of seepage rates and small
number of data points limits the significance of these correlations.
G Both Sites - Two vertically offset regression lines for the Dahwilly sites are
recorded and there is no observed trend for the combined regression line. This
suggests different background conditions (despite apparent similarities between
sites), including finer and more clayey sands at Dahwilly East and possibly higher
background salinity groundwater.

Finley
G The best correlation was observed at 4m and 6m which corresponds with the zone
below the top of the watertable (1.5m). A misleading opposite (inverse)
correlation was observed at 2m depth, when a stronger, direct correlation was
expected based on depth to watertable. This may be a reflection of the poor
surface resolution of the resistivity equipment. The narrow seepage range and
small number of data points also limits the significance of the correlations.
Lake View
G Central - The best correlation was observed at 6m and 8m which corresponds
with the zone several metres below the top of the watertable (1.5m). A very weak
trend was observed at 2m and an (inverse) correlation at 4m depth. It was
expected based on the depth to watertable that the best correlations would be
observed at around 2-4m depth. This may be a reflection of the poor surface
resolution of the resistivity equipment. Site lithology below the watertable may
also be significantly contributing to the response. The narrow seepage range also
limits the significance of the correlations.
G West – There was no meaningful relationship observed between resistivity and
seepage immediately below the watertable. This is attributed to: i) Poor
resolution of near surface data (this theory is supported by the reasonable
correlation observed between EM31 and seepage at this site) and/or, ii) The
'anomalous' result in pond 1. Something is causing elevated resistivities relative
to other ponds at the site (possibly lithology, faulty data collection etc), although
nothing obvious was detected in the drilling.
G Both Sites - Neither the 2m or 6m depth slice from Lake View Central could have
been used to accurately predict seepage at Lake View West. However a slightly
better fitting trend line is obtained for the 2m depth slice. The apparently
anomalous result at Pond 1 (Lake View West) is largely skewing these
relationships. Using the Lake View Central site to predict seepage at the Lake
View West site would have caused significant under estimation of actual seepage.
In summary, most sites displayed a good correlation between seepage and the
resistivity at and immediately below the watertable. The two sites that did not were
Toolondo West and Lake View West. At Toolondo West it appears that the type of
sandstone at this site may be dominating the response, however deeper drilling would
be required to confirm this interpretation. A reasonable trend was obtained at shallow
depth, but without the information supplied by the pondage tests this could not have
been known. The lack of trend at the Lake View West site is probably due to the poor
resolution of the resistivity equipment at very shallow depth. This site contains the
shallowest watertable across all sites (0.5 – 1m). Improved resolution at shallow
depth could relatively easily be improved in future surveys (Allen, pers. comm.
1/11/02). At Toolondo East also no trend was observed but this is solely attributed to
the very narrow range of seepage rates at this site.

n Table 5-10 Summary of Resistivity Trials Compared to Pondage Test Seepage
Channel Survey Survey Survey Channel No. of Range of Watertable Depth of Correlation Seepage Summary on Seepage - Geophysical Use of Comparison with other Reported
Type Date Location Operating Pondage seepage Depth Best Coefficient Measurement Relationship relationship for geophysical techniques in Section
Cells results (approx.) Correlation Mechanism extrapolation
Toolondo Resistivity March. On- Yes 4 1-5 12m 2m 0.80 Inferred from At & below the WT the expected inverse Toolondo East data The inverse relationship 5.4.4.1
West 2002 channel mm/d unsat. zone soil trend betw'n high resis./low seepage & low (Av. 10/12/14m) lies between EM31 conductivity
properties resis./high seep was not observed. It is within expected and seepage at this site
apparent that the s'stone at this site may be extrapolated range concurs with the direct
dominating the response. Low permeability for Central (10m) relationship between shallow
s'stone may be causing a high resistivity site. The West (Av resistivity (2 & 4m) and
response (normally associated with high 10/12m) data does seepage at this site.
seepage), and masking the effect of seepage not fit within Central
on the saline groundwater. The reasonable & East relationship
correlations obtained at shallow depth (possibly due to
(2&4m) are most likely due to changes in changes in lithology
clay content beneath the channel. Deeper masking seepage
drilling would be required to confirm this impacts). The West
interpretation 2m depth slice does
Toolondo Resistivity March. On- Yes 6 1 - 11 10-11m 10m / 12m 0.59 / 0.56 Dilution of Around 10-12m the best correlation is fit within Better result obtained than the 5.4.4.2
Central 2002 channel mm/d groundwater obtained (R2=0.6), which fits with the relationship, however EM34(10m) due to better
from fresher expected mechanism of seepage detection (ie without PTs it would targeting of the top of the
seepage water in the depth interval of g'water most effected not have been known watertable. Slightly better
by seepage). Within ponds cross sections that this was a better EM31 surveys results (outside
show sub-sections of higher seepage. Very depth to focus on. channel) were obtained (R2:
little correlation in the unsaturated zone. 0.7-0.8) but this relies on a
different mechanism
Toolondo Resistivity March. On- Yes 4 0-1 10m No correlation No correlation Dilution of The very narrow range of seepage rates and No correlations found with 5.4.4.3
East 2002 channel mm/d groundwater resistivity values meant that no meaningful other techniques either due to
from fresher correlations were observed at the Toolondo narrow range of variables.
seepage water East site.
Dahwilly Resistivity March. On- Yes 6 1 - 10 5m 8m / 10m 0.86 / 0.85 Dilution of Good correlations were observed at 8m & Two (vertically) Better correlations obtained at 5.4.4.4
Central 2002 channel mm/d groundwater 10m, which fits with the expected mode of offset regression 8m and 10m than the
from fresher seepage detection (ie targeting of the depth lines for the two EM34(10m & 20m) and
seepage water interval where g'water is most effected by Dahwilly sites are EM31(vert) due to better
seepage). The fit at 6m is also good recorded. There is no targeting of the top of the
(R2=0.72), but correlation coefficients observed trend watertable.
worsen in the unsaturated zone. Cross between the sites.
sections indicate that seepage is generally This suggests
diffuse (ie uniform) across the surveyed area. different background
Dahwilly Resistivity March. On- Yes 3 9 - 10 5-6m 6m 0.91 Dilution of The best (and strong) correlation was conditions (despite Better correlations obtained at 5.4.4.5
East 2002 channel mm/d groundwater observed at 6m which corresponds with the apparent similarities 6m than the EM34(10m &
from fresher top of the watertable. However, the very betw'n sites), 20m) and EM31(vert) due to
seepage water narrow range of seepage rates and small no. including finer & better targeting of the top of
of data points (only 3) limits the significance more clayey sands at the watertable.
of these correlations. Dahwilly East (&
possibly higher
background salinity
gw).
Finley Resistivity March. On- Yes 3 5-7 1.5m 4m / 6m 0.92 / 0.93 Dilution of The best correlation was observed at 4m & - Better correlations obtained at 5.4.4.6
2002 channel mm/d groundwater 6m which corresponds with the zone below 4m than the EM31(vert) due to
from fresher the top of the watertable (1.5m). A better targeting of the top of
seepage water misleading opposite (inverse) correlation was the watertable
observed at 2m depth, when a stronger,
meaningful correlation was expected based
on the depth to WT. This may be a reflection
of the poor surface resolution by the
resistivity equipment. The narrow seepage
range & small no. of data points (3) also

Channel Survey Survey Survey Channel No. of Range of Watertable Depth of Correlation Seepage Summary on Seepage - Geophysical Use of Comparison with other Reported
Type Date Location Operating Pondage seepage Depth Best Coefficient Measurement Relationship relationship for geophysical techniques in Section
Cells results (approx.) Correlation Mechanism extrapolation
Lake View Resistivity March. On- Yes 4 4-7 1.5m 6m / 8m 0.58 / 0.60 Dilution of The best correlation was observed at 6m & Neither the 2m or 6m Similar R2 to Lake View 5.4.4.7
Central 2002 channel mm/d groundwater 8m which corresponds with the zone several depth slice from EM31, June 2000(EM31) v
from fresher metres below the top of the watertable Lake View Central July 2001(PT): 0.58.
seepage water (1.5m). A very weak trend was observed at could have been used
(& possibly 2m and an (inverse) correlation at 4m depth. to accurately predict
lithology below It was expected based on the depth to WT seepage at Lake
WT) that the best correlations would be observed View West.
at 2-4m depth. This may be a reflection of However a slightly
the poor surface resolution by the resistivity better fitting trend
equipment. Site lithology below the line is obtained for
watertable may also be significantly the 2m depth slice.
contributing to the response. The narrow The apparently
seepage range also limits the significance of anomalous result at
the correlations. Pond 1 (LV West) is
Lake View Resistivity March. On- Yes 4 20 - 25 0.5 - 1m No correlation No correlation Dilution of Lack of meaningful relationship between largely skewing these A sensible (inverse) 5.4.4.8
West 2002 channel mm/d groundwater resistivity - seepage immediately below the relationships. Using correlation was obtained at this
from fresher watertable attributed to: i) Poor resolution of the LV Central site to site between EM31 and PT
seepage water near surface data (supported by reasonable predict seepage at the (R2=0.54). The fact that such
correlation between EM31 & PT at this site) West site would have a correlation was not observed
and/or, ii) 'Anomalous' result in pond 1. caused sig. under for shallow resistivity data
Something (lithology?) causing elevated estimation of actual suggests poor resolution of
resistivities relative to other ponds, although seepage. surface data by the resistivity
nothing obvious detected in drilling. array.

6. Waranga Western Channel: Case Study
This section presents a case study of channel seepage aspects of a channel capacity
upgrade of the Waranga Western Channel (WWC). The WWC is an open irrigation
channel maintained by Goulburn-Murray Water (G-MW), in north-west Victoria. The
location of the channel is shown in section 5, Figure 5-1.
6.1 Study Objectives

channel maintained by Goulburn-Murray Water (G-MW), be upgraded in capacity.
The area proposed for upgrade was from the Loddon Weir to west of Boort (north-
west Victoria), approximately 50km in total length. The increase in capacity is
required primarily to supply additional irrigation water to a new horticultural
development near the Boort area, as well as to meet the needs of existing customers in
the region.
The channel has a well-documented record of existing seepage problems. The extent
of channel seepage in the Boort West of Loddon Salinity Management Plan area has
been a concern to local landholders for a number of years. The Channel Seepage
subcommittee of the Boort West of Loddon Community Working Group initiated a
Channel Seepage Program in 1993 (McConachy, 1993). However previous
investigations were unable to satisfactorily identify priority sites for remedial works
(G-MW, 2000). In addition, there was concern that new seepage paths may be opened
up during the upgrading works program.
Therefore G-MW required quantification of sections in the WWC channel with

existing seepage problems and identification and quantification of sections where new
seepage paths might be opened up. To this end, geotechnical and geophysical
investigations were carried out along the channel.
6.2 Outline of Work Undertaken

The following investigative works were undertaken on the WWC:
q EM31 survey – November 2001: A 46km EM31 survey was conducted on-
channel and on-land on each side of the channel. This was coupled with drilling
of 128 bores adjacent the channel (to 4m depth) to ground truth the survey;
q Additional geotechnical drilling – March 2002: An additional 107 bores were
drilled and 34 piezometers installed. Bores were generally drilled to a depth of at
least 6m, and some up to 10m. Work was conducted in accordance with AS1726-
1993 Geotechnical Site Investigations;
q Pondage tests – May/June 2002: 12 pondage tests were conducted at various
locations along the length of channel under investigation.
The chapter is based on the chronological reporting of the above three stages of
investigation, with an initial section describing site conditions.
WC01312:YR_3_RPT_CH_6(WWC CASE STUDY).DOC Final PAGE 233

6.3 Site Conditions
6.3.1 Geology and Hydrogeology
The WWC traverses through Shepparton Formation surface sediments that are
widespread throughout the southern Murray Basin. These consist predominantly of
clays to sandy clays, with shoe string sand deposits (fine to medium grained) which
intersect the channel at various locations along the study length. Underlying the
Shepparton Formation along most of the channel length is extremely weathered to
moderately weathered sandstone. The sandstone represents the indurated zone of the
top of the Parilla Sands aquifer. In sections along the channel this indurated zone rises
very close to the surface and in part intersects the channel.
Depth to groundwater also varies along the channel however typically the depth of the
watertable is between 6 – 10 metres, but has been recorded within several metres of
the surface in some locations immediately adjacent the channel. According to the 1:
250,000 scale St Arnaud Hydrogeological mapsheet, the regional groundwater is
typically highly saline, between 3,000 and 35,000 mg/L TDS. The salinity generally
increases in the direction of groundwater flow, towards the north-west. However, in
the proximity of the Loddon River, the groundwater freshens significantly to between
1,000 and 3,000 mg/L TDS.
6.3.2 Channel Conditions

q Channel capacity: from 270 Ml/d to 1200 Ml/d
q Regular operation procedures - None
q The channel has been in operation for more than 70 years.
q Channel maintenance - Weed spraying once a year. No de-silting has been
conducted in the last 25 years. Over this time, the channel has deposited a
significant silt deposit on the bed of the channel. Longitudinal surveys conducted
as part of the investigation indicate that in some sections of channel the silt layer
is up to 0.4m thick, but generally it is between 0.05 - 0.15m in thickness. This
layer will have a very significant impact on seepage rates, and it is likely that in
most sections of the channel this clogging layer will be the controlling influence
on seepage rates, rather than underlying soil types and geology.
6.4 EM31 Survey and Initial Drilling

6.4.1 Description of EM31 and Drilling Program
An EM31 survey (vertical dipole) was conducted by Ken Bates Soil Surveying in
November 2001. The survey included a traverse on each side of the channel,
immediately adjacent the channel toe, as well as a survey on the channel. The on-
channel survey (and the pondage test results) is depicted in Figure 6-1. With a depth
to watertable of approximately 6-10m the EM31 is largely measuring unsaturated zone
soil / lithology properties. (Note that due to high density of data collection, and small
figure size, this plot overlaps many data points and may give a slightly misleading
representation of conductivities, depending on which layer was plotted last).

To assist with interpretation of the EM31 survey, a drilling program of 128 boreholes
(to 4m depth) was conducted. Bores were drilled adjacent to the channel, over the
length of the proposed upgrade. This was later supplemented by a total of 12 test-pits,
excavated by backhoe, at selected locations adjacent to the channel. The EM31
contractor developed a seepage risk map, dividing the channel into low, medium and
high seepage risk categories based on EM value (check). The high risk section of the
channel totalled approximately one-third (15km) of the length of the channel.
n Figure 6-1 EM31 Survey on the Waranga Western Channel

6.4.2 Analysis
A correlation was sought between the results of the EM31 survey and the ground
conditions encountered during the drilling. The soil logs and the EM31 survey data
were examined in order to review the interpretations of the risk of channel seepage as
shown on the ‘risk’ maps produced by the EM31 contractor. The EM31 survey
results, together with the textural soil descriptions of the bore logs determine whether
the survey results are reasonable with respect to the nominated seepage risk.
The textural descriptions of the soils (130 soil bores in total) comply with Northcote
(1979) classification system and for each texture a permeability description was
provided (Table 6-1). Sandstone, interpreted as being representative of the upper
Parilla Sands formation, was regularly encountered in the soil bores, and textures were
assigned to this unit according to the degree of weathering.
n Table 6-1 Distribution of Impermeability Grades
Textural Description Permeability Impermeability

Description Classification
Medium to heavy clay Very low 1.00
Light clay Low 0.85
Silty clay, sandy clay, fine sandy clay, fine Medium 0.70
sandy clay loam, silty clay loam, clay loam,
sandy clay loam
Silty loam Medium to High 0.55
Loam, light sandy clay loam, sandy loam, High 0.40
clayey sand, loamy sand
Fine sand High to Very High 0.25
Sand, coarse sand, fine gravel, gravel Very High 0.10
An 'impermeability classification' was assigned to each texture; 0.10 being

representative of the most permeable and 1.0 representative of the least permeable
soils (Table 1). Subsequently, for each bore, an impermeability grade was assigned to
the individual textural intervals. A bulk impermeability grade (%) was derived for
each bore according to the thickness and impermeability classification of each textural
interval in the soil profile.
The distribution of the EM31 survey results with respect to the bulk impermeability
grade is presented in Figure 6-2. The results are separated into two classifications, no
rock present in the profile and rock present in the profile. Variation permeabilities
were assigned to the rock intervals, and therefore the rock in the profile does not
appear to affect the relationship between the EM31 survey results and permeability.
The data suggests there is a coarse relationship between the EM31 survey results and
the permeability of the soils, with lower EM31 readings (lower conductivity), in some
instances, associated with soils of higher permeability. Accordingly, it was considered
reasonable that the EM31 survey results be used to broadly assess the potential degree
of seepage risk.
The data suggests there is a coarse relationship between EM31 conductivity and the
permeability of the soils, with low EM31 conductivity, in some instances, associated
with soils of higher permeability. Accordingly, it was considered reasonable that the
EM31 survey results be used to broadly assess the potential degree of seepage risk.

n Figure 6-2 Bulk Impermeability Grade vs EM31 Conductivity (Initial Drilling
Program
100
90
80
Bulk Impermeability Grade (%)
70
60
No rock present in profile

50
40 Rock present in the

profile
30
20
10
0
0 50 100 150 200 250
EM31 (ECa) mS/m
Importantly, the EM31 data are highly scattered (with respect to the bulk
impermeability grade), and therefore, the potential for seepage cannot be definitively
inferred from the EM results. For example, according to the data analysis, a low
EM31 reading does not necessarily imply a high soil permeability. On the basis of the
highly dispersive relationship, the highest seepage rates may not necessarily be
associated with the lowest EM 31 results. Further, it was noted that caution should be
exercised, in using the data solely to identify seepage risk on the basis that a range of
factors may affect the EM31 survey results. The results could not be used to quantify
seepage rates. The absence of pondage tests was seen as a deficiency in the program
methodology.
6.4.3 Initial Identification of High Priority Areas

The high risk section of the channel as categorised on the EM31 contractor supplied
map totalled approximately one-third (15km) of the length of the channel.
A combination of the EM31 results and the impermeability grade was used to identify
sections of channel which were considered to represent ‘very high’ risk areas (as
opposed to the ‘high’ risk categorisation based solely on the EM31 survey). These
were defined as zones where the EM31 was less than or equal to 50 mS/m and the
bulk impermeability was less than or equal to 65%. This category is marked on Figure
6-2 by the dashed lines. Four significant lengths of channel were identified as fitting
these criteria (ignoring short lengths of channel, ie less than 200m). The details of
these reaches are presented in Table 6-2. This table also presents the overlap with
known (ie visible) seepage sites.

n Table 6-2 Very High Risk Sections [EM31 (mS/m) < 50 and Impermeability
Grade < 65%] as identified from EM31 survey and original drilling program
Site Chainage (m) Length (m) Overlap With Known
Seepage Sites
1 196,340 – 198,500 2,160 No
2 201,250 – 205,000 3,750 202,090 – 202,590 (500m)
3 213,930 – 214,710 780 214,010-214,620 (610m)
4 217,060 – 217,660 600 No
Total 7,290m 1,110m
6.5 Additional Drilling Program

The total area identified for seepage control works (very high risk areas as defined
above) and rock excavation (also required as part of the upgrade) exceeded initial cost
estimates for the upgrading project and were greater than the available funding. These
initial estimates were based on preliminary costings prepared before geotechnical
investigations were carried out. It was apparent to G-MW that further investigation
was required to further refine the ‘very high risk’ areas, and potentially reduce the 7.2
km recommended for lining.
6.5.1 Description
The additional phase of the geotechnical investigation involved the drilling and
logging of an additional 107 boreholes, and was carried out in March 2002. A total of
34 piezometers were also installed in selected boreholes. The initial geotechnical
investigation carried out in association with the EM31 survey was done relatively
quickly and cheaply. The drill-rig used was a trailer mounted custom made rig that
could not penetrate the rock or cemented sandstone layers. Relatively simple borelogs
were prepared. The subsequent geotechnical investigation was completed in
accordance with AS1726-1993, Geotechnical Site Investigations. The rig used had the
ability to penetrate quite hard rock layers. The bores were generally drilled to 6 metres
depth. The locations of boreholes for the second geotechnical investigation were
selected in order to further refine the extent of:
q ‘Very high risk’ seepage zones; and
q Zones where rock material intersects the channel profile.
6.5.2 Refinement of High Priority Areas

Methodology
The status of the classification of the original ‘very high’ risk sections in light of the
additional bore information was assessed, in addition to potential new areas of very
high risk outside the originally defined areas. Using the new bore information, the
following tasks were conducted:
q Calculation of impermeability grade of new bores (as conducted previously -
refer previous section for methodology);
q Plot of impermeability grade verses EM31 value;
q Construction of geological cross sections, using divisions of high, moderate and
low permeability; and,
q Identification of bores / sections logged with high moisture content / saturation.
For sections outside of the very high risk zones the review involved:

q Identification of bores with greater than 50% content of gravel, sand, clayey sand
etc (ie potentially high permeability).
Results
Figure 6-3 plots the results of the impermeability rating for each of the new bores
drilled within the very high risk zones against the corresponding EM31 value. This
figure shows that, in general, lower EM31 readings (lower conductivity), are generally
associated with soils of higher permeability (lower impermeability), as suggested by
the bore data (as was previously observed). It was therefore considered reasonable
that the EM31 survey results (extrapolated from bores) be used to broadly assess the
potential degree of seepage risk.
n Figure 6-3 - EM31 Versus Impermeability Grade: Additional Bores Within

Very High Risk Areas
EM31 Results Versus Impermeability Grade
100
90
80
70
Impermeability (%)
60
50
40
30
20
10
0
0 10 20 30 40 50 60
EM31 value (ECa)
However this plot does display some scatter, particularly around the middle part of the
graph, which indicates that caution needs to be applied in using the EM31 versus soil
permeability relationship as the definitive tool for identifying potential seepage areas.
The uncertainty in the results highlighted the need for pondage tests to be undertaken.
The advantage of obtaining a relationship between EM response and pondage tests is
that it represents a direct relationship between seepage and the EM data. EM versus
soil permeability is one step removed and requires additional interpretation to account
for the fact that soil permeability is not a direct measure of seepage.
Based on the cross sections constructed for the four areas of ‘very high risk’, together
with the EM31 results recommendations were made as to whether the ‘very high’ risk
rating should be maintained in the light of the new data. An example of the process
that was used to assess each section is presented in Table 6-3.

n Table 6-3 Review of Very High Risk Rating: Section 1
Chainage Length Very High
(m) Reccom- Risk Sites (m) Discussion / Justification Moisture
(Site Ref. endation Original Revised Content
Name)
196,340 – Leave as 560 560 BH168 – BH170 identified significant sand layers The moisture
196,900 very high at depth and coinciding with new design bed in content in all these
risk BH168. While BH169 – 170 only have permeable bores was low,
1A layers below the design bed, the gravel in BH170 indicating that
is very close to the design bed surface and the current seepage
GrSC in BH169 could provide a seepage path. rates may not be
BH53 indicates sand intersecting the channel, as high as
however it is on the upgradient side, and the EM31 suggested by the
conductivity does not indicate the same very low EM31 and bulk
response recorded around BH168. Therefore impermeability
whether the very high risk classification should analysis.
extend to, say CH197,150, should be determined However, the
by drilling on the down gradient side of the channel channel
in this location remodelling works,
including de-
196,900 – Down- 700 0 This section has been down-graded to high risk, silting, will re-
197,600 grade to primarily based on the information from BH171, establish
high risk which indicated Sandy Clays for most of the connection
1B profile. The decrease in EM31around Ch197,600 between the
indicates the potential for seepage is increasing at channel and the
this location higher
permeability
197,600 – Leave as 400 400 This section should maintain its very high risk layers. This may
198,000 very high classification due to the extremely sandy profile create seepage
risk intersecting the channel (BH50). While this bore is pathways
1C on the upslope side (LHS), the extent of the sand suggested by the
profile and the very low EM31 response (RHS) in geological cross
this section indicates potential high seepage sections, but
currently blocked
198,000 – Down- 250 0 The predominantly clayey profile of BH60 warrants by the silt bed.
198,250 grade to down-grading this section to high risk
high risk
1D
198,250 – Leave as 250 450 High sand content in BH173 & BH51 indicates this
198,700 very high section should stay in the very high risk section.
risk Although the sand in BH173 is at depth, sand on
1E + the LHS (BH51) intersects the channel indicating
Upgrade the presence of a significant seepage pathway.
200m to
very high
risk
Total 2,160 1,410
Bores Outside of Very High Risk Sections

Areas outside of the original very high risk regions (as identified in the previous
review) were assessed. This was undertaken by identifying bores which contained a
greater than 50% content of high permeability material. Two potentially significant
sections of channel susceptible to seepage were identified. Other shorter sections with
seepage potential were identified, but available evidence suggests these are quite
isolated occurrences, ie less than 100m.
Summary
Following the review of the additional drilling the areas classified as very high
seepage risk actually increased by 990m (from 7290m to 8280m). This included some
areas being removed and some added to the very high risk category. To assist with
prioritisation of these sites it was recommended that the profiles with significant sand
layers intersecting the channel be considered the highest priority (as de-silting of the

channel will expose these layers and allow potentially very high seepage). Table 6-4
lists these priority sites.
n Table 6-4 Prioritisation of Very High Risk Sites

Priority Section Chainage Length (m)
1 1A 196,340 – 196,900 560
2 2A 199,900 – 201,000 1,100
3 2C 201,600 – 203,300 1,700
4 3A 213,840 – 214,760 920
5 4A 216,850 – 217,800 950
Total 5230
G-MW recognised that in addition to the drilling program, pondage tests were
required to:
q Quantify seepage rates (and potentially identify a relationship between EM31 and
pondage test seepage rates);
q Confirm interpretation of seepage rates based on geology and EM31 data.
6.6 Pondage Tests

6.6.1 Description
Pondage tests on their own will provide additional point data on certain reaches of the
channel. This data will be useful for identifying likely seepage rates for the type of
sub-surface conditions encountered at that site, and other similar sites based on soil
bore information.
However, while a large number of bores have been drilled along the channel, the
EM31 remains the only continuous data available. The assessment methodology
described in Section 6.4 and 6.5 rely significantly on the EM31 results and an attempt
was required to confirm the appropriateness of this reliance. It is therefore important
to attempt to identify the relationship between EM31 response and pondage test
seepage, particularly given the spread of results identified in the EM31 versus
impermeability relationship. If a reasonable relationship can be identified, greater
confidence can be placed in using the EM31 data for determining seepage control
locations.
Some concern was expressed by G-MW on the necessity of conducting pondage tests
when the degree of siltation in the channel means that pondage tests conducted now
are unlikely to be unrepresentative of seepage rates post de-silting and construction.
While post-silting rates are likely to be much higher than existing rates, it is
reasonable to assume that the areas of highest seepage now, will be the areas of
highest seepage post-construction and the low areas now will still relatively be the low
seepage areas post de-silting, or at worst this will be the case after several years of
channel operation.
Therefore the primary reason for conducting pondage tests was to confirm the
interpretation of the EM31 (ie that the lowest conductivity does in fact equal the
highest seepage), upon which the entire selection process of the sections to be clay
lined was based. The secondary reason for conducting the pondage tests was to

identify existing seepage rates along the channel. These figures will allow a
reasonable estimate of the water savings resulting from channel lining.
These figures will also assist in improving the decision making process as to the cut-
off criteria to be applied in determining which areas of channel should be remediated.
The current approach (which identified 8.3km of very high priority channel sections)
has a some what arbitrary cut-off. Being able to assign approximate seepage rates to
conductivity response will assist in refining this process.
G-MW proposed pondage tests locations and produced justification of these sites. The
basis of the site selection was on selection of different types of (potentially) high
seepage sites (eg rock sites, high risk EM31, ‘wet’ sites, previous core trenching,
historically known seepage etc). These are reasonable criteria on which to select sites,
as seepage across a range of these different types of environments needs to be known.
However it should not be the only selection criteria, and the important task of
attempting to link the EM31 to the pondage tests also needs to be considered. It was
recommended that pondage tests be conducted in the following manner:
q Cover a range of EM31 response, not just high risk but also low risk areas. This
means conducting tests where seepage rates are low. (Sites initially selected by
G-MW sites sufficiently covered the low conductivity range, however further
tests were recommended in the higher EM31 conductivity response area);
q 2-3 cells in 3-4 different areas along the channel (ie 6-12 cells total);
q Relatively short length cells (150-300m);
q Each individual cell should target a section of like EM31 response - If the area of
the pondage banks crosses over significantly different EM31 conductivities,
meaningful interpretation of the EM31 results in relation to the seepage rates is
difficult, as the pondage test provides no indication as to where the seepage is
occurring. If each cell covers an area of similar EM31 response, meaningful
assessment can be made of the seepage versus EM31 relationship; and,
q Tests should be a minimum duration of one week.
6.6.2 Analysis
Figure 6.6 summarises the relationship between the pondage tests and EM31 results
(land based EM31 data). The pondage test seepage (mm/d) is plotted against the
average EM31 conductivity (mS/m) over the corresponding pond length. The
pondage test seepage rate used in this analysis excludes the results of the first day of
the tests, as they were not representative of the longer term seepage rate. The EM31
data is the average of the land based EM31 results on each side of the channel (ie
excludes the on-channel results taken from the boat). Each point in the figure is also
labelled with the pond number and a very brief summary of the dominant geology for
the section, based on the nearest bore information.
This figure shows that a moderate to poor linear relationship exists between the two
variables, with a correlation coefficient of 0.40. (A similar correlation coefficient was
obtained for the boat EM31 data versus the pondage tests). The standard estimate of
error was 63% of the mean, which is indicative of the fairly high degree of scatter
about the fitted line. Given that these results represent ponds up to 20 km apart this
relationship was considered reasonable. The likely cause of some of the scatter in the
results is discussed below.

n Figure 6-4 Pondage Test Seepage Versus EM31 Conductivity
Pondage Test Seepage Rate vs EM31 Conductivity : Waranga Western Channel

110
P5 Clay: 0-5 m
LHS & RHS
100 Linear (LHS & RHS)
P3 Clay-SCly:0-6m
S. Clay: 0- 1 m
90 Clay: 1-3 m
Wthrd Rck(SC/CS): 2-6m
80 P11 Clay: 1-1.5 m

P2 Rock(SGrC):1.5-6m
S. Clay: 0-2.5 P1
m
70 Clay: 2.5-6m
Gr Clay: 0-1 m
S. Clay: 1-3m
60 W.Rck(SC):3-6m y = -4.19x + 84.53
P6 2
R = 0.40
P4
50 Clay: 0-0.5 m S.Clay: 0-2 m
Clay: 0-3.25 m
Gr. Snd: 0.5-2m Rock(C.Snd): 2-6 m
Gr Snd-Gr: 3.25-6m P10
S. Cly :2-5m P7
40
P9
30 S. Clay: 0-1.5 m P8
Gr Snd-Rck in S. Clay: 0-1 m
part: 1.5-6m Rock (S.Cly): 1-2.5m
W.Rck (Snd):2.5-6m
20
0 2 4 6 8 10 12 14 16
Seepage Inferred from Unsaturated Zone Properties: The depth to watertable along
the WWC channel is generally 8-10m below surface and therefore usually out of the
range of the EM31 equipment. Therefore seepage is not directly measured, but
inferred from the lithology under / beneath the channel. An example of the problems
this may cause is highlighted in pond 4, where EM31 response belies the real seepage
rate.
Pond 4 contains a clay layer to about 3m which is underlain by gravel and gravelly
sand. The EM31 (in vertical dipole) is still strongly influenced by material at depth
and therefore this very permeable material below 3m is significantly contributing to
the relatively low conductivity (52 mS/m). However, the upper 3m of clay is
providing a more than adequate buffer to seepage and thus pondage seepage rates are
low (approx 2 mm/d). This highlights the limitations of using geophysical techniques
to ‘detect’ seepage based on unsaturated zone soil properties.
Further, the WWC is a channel with a significant silt bed. Generally the silt layer
rather than soil properties beneath the channel limit the seepage rate. This will
contribute to misleading results when comparing unsaturated zone lithology to
seepage rates, and is certainly the cause of some of the scatter in the WWC regression
relationship (pond 6 appears to be a good example of this). Direct measurement of
seepage impacts in the watertable are likely to improve this relationship and the
statistics of the regression equation. This improvement was observed in a similar
channel in this ANCID study where the silt layer was controlling seepage rate where
the unsaturated zone was of high permeability (in the Dahwilly channel - refer section
5).

Pond Placement: The placement of the ponds may also explain some the deviations of
the data points from the trend line. The ponds were designed to be located over
sections of like EM31 conductivity. However due to field restraints imposed on the
locations of some of the ponds and also potentially due to error in locating the pond
banks in the field, the pondage tests did not always cover areas of like conductivity.
As noted previously, it is important that this is the case, because this method assumes
that a particular conductivity will correlate to a particular seepage rate. Therefore if a
ponded area contains a wide range of conductivities the assumptions behind the
analysis are compromised.
A good measure of the variability of the data is the standard deviation. Table 6-5
displays the results of the pondage tests, the average EM31 conductivity for the pond,
the standard deviation of the EM31 conductivity for each pond and the standard
deviation as a percentage of the average EM31 conductivity for the section. Results
for both the land based and boat based results are presented. For the land based data,
Pond 1 (49%), Pond 8 (33%) and Pond 10 (32%) have the highest standard deviations
as a percentage of the average result and therefore display the greatest variance about
the mean. This indicates that these ponds are not well located over sections of like
conductivity, and are therefore less likely to plot well on the regression line in
Figure6-4.
n Table 6-5 Standard Deviation of EM31 Conductivity Data
Av EM 31 Cond. Stand. Dev. / Av EM31

Pond No. PT Seepage Stand. Dev. of Cond.
(mS/m) Cond.
mm/d Land Boat Land Boat Land Boat
1 6.1 78 37 38 13 49% 35%
2 3.3 75 57 13 6 17% 11%
3 1 93 72 13 9 14% 13%
4 1.7 52 64 10 13 19% 20%
5 1.4 106 132 21 18 20% 14%
6 1.6 58 42 14 7 24% 17%
7 7.7 46 35 13 2 28% 6%
8 7.7 30 28 10 7 33% 25%
9 7.3 34 32 5 3 15% 9%
10 13.1 44 33 14 5 32% 15%
11 4.4 82 79 17 17 21% 22%
Figure 6-5 plots EM31 conductivity versus pondage test seepage with the 3 points of
high variance and pond 4 removed. This improves the fit considerably, with an R2 of
0.62, however most of this improvement is due to the removal of the pond 4 data point
(refer above discussion) which is perhaps not justifiable. Therefore Figure 6-4 is used
in the analysis.

n Figure 6-5 Pondage Test Seepage Vs EM31 Conductivity, With Removal of
High Variance Points & P4
Pondage Test Seepage Rate vs EM31 Conductivity (high variance points removed) : Waranga Western Channel
110
P5
LHS & RHS
100 Linear (LHS & RHS)
P3
90
80 P11
P2
70
y = -7.30x + 98.40
R2 = 0.62
60
P6
50
P7
40
P9
30
20
0 2 4 6 8 10 12 14 16
Implications of Correlation for Use of EM31 Data

The most important aspect of the pondage test data assessment was to consider the
implications of the observed correlation (between pondage test and EM31), in light of
how the EM31 data have been used to date in the investigation. This is important
given that the priority areas for lining have relied significantly on the EM31 results. In
other words, was the right thing done in relying on the EM31 results as an accurate
predictor of seepage locations?
The variability of the data in Figure 6-4 indicates that a high degree of confidence
cannot be placed in the EM31 data for determining exact seepage rates. However in
broad terms, EM31 appears to have been accurate in predicting between high and low
seepage rates. This ‘broad brush’ use of the regression equation is confirmed by the
wide prediction interval bands which have been plotted in Figure 6.6. 80% and 90%
prediction interval bands have been plotted and indicate that:
1) There can be 80% confident that when EM31 conductivity is greater than 70
mS/m, seepage rates are less than 8 mm/d; and,
2) When EM31 conductivity is 30 mS/m or less, we can be 80% confident that

seepage rates are greater than 4 mm/d.
With the pondage tests locations containing high EM31 variability removed, the
confidence intervals contract significantly. This statistical analysis indicates that the
interpretation of areas of high EM31 conductivity as low seepage risk and areas of low
EM31 as areas of higher seepage risk is reasonable but limited to broad classification.

The confidence bands indicate that there is little confidence in EM31 to distinguish
between areas of ‘moderate’ seepage (P7, 8 & 9: 7-8 mm/d) and ‘high’ seepage (P10:
13 mm/d).
However given that the only data point in the ‘high’ range (Pond 10) displayed high
variance, a more accurate conclusion would be that there is no reliable data in the
‘high’ seepage rate range in order to conclude one way or another whether EM31 can
distinguish between the high and moderate seepage rate areas. Therefore areas of low
EM conductivity need to be assessed in conjunction with drilling data. This approach
was followed as much as possible in the lining selection review.
n Figure 6.6 Waranga Western Channel Pondage Test Seepage Versus EM31
Conductivity – Prediction Interval Bands

20
Predicted
18 Upper_90
Lower_90
16
Measured
Predicted Seepage (m3/m2/d)
14 Upper_80
Lower_80
12
y = -0.096x + 11.089
10
R2 = 0.400
8
0
20 30 40 50 60 70 80 90 100 110
6.6.3 Final Refinement of High Priority Areas

Based on the results of the pondage test areas and relationship with EM31 and the
drilling program, the areas recommended for remediation were finalised. Sites were
defined as either priority one or priority two seepage risk sites, depending on the
degree of perceived seepage risk. Priority one sites were considered to require
remediating as part of the upgrade, while priority two sites were to be monitored
closely for seepage following the upgrade. EM31 predicted seepage was not used as
the sole means of assigning seepage risk, given the uncertainty in the EM31 – seepage
relationship. Geological data and visual observations were also integrated into the
decision making.

6.7 Conclusions
documented record of existing seepage problems. The extent of seepage in this
section of the channel had been a concern to local land holders for a number of years.
In addition, there was concern that new seepage paths may be opened up during the
upgrading works program. Therefore G-MW required quantification of sections of
existing seepage problems in the WWC channel and identification and quantification
of sections where new seepage paths might be opened up. To this end, geotechnical
and geophysical investigations were carried out along the channel, including the
following investigative works (in chronological order):
q EM31 survey – November 2001: A 46 km EM31 survey was conducted on-
drilled and 34 piezometers installed. Bores were generally drilled to a minimum
depth of 6m, and some up to 10m;
The high risk section of the channel, as categorised by the EM31 contractor based on
EM31 magnitude, totalled approximately one-third (15km) of the length of the
investigated section of the channel. To further refine this area, a combination of the
EM31 results and impermeability grade (a lithological classification devised for the
investigation based on the amount of clay in the profile) was used to identify sections
of channel which were considered to represent ‘very high’ risk areas. Using this
system, four significant lengths of channel were identified as very high risk. The total
length of ‘very high risk’ area was 7.2 km. It was apparent to G-MW that further
investigation was required before committing to the significant expense of lining such
a length of channel.
A further 107 bores were drilled (to greater depth than the original drilling program).
The above process using impermeability grade (incorporating results from the new
bores) and EM31 response was again conducted. Following the review of the
additional drilling, the areas classified as very high seepage risk actually increased by
about 1 km (from 7.3 km to 8.3 km). This included some areas being removed and
some added to the very high risk category.
It was recognised that in addition to the drilling program, pondage tests were required
to quantify seepage rates (and potentially identify a relationship between EM31 and
pondage test seepage rates) and confirm interpretation of seepage rates based on
geology and EM31 data. Therefore 12 pondage tests were conducted at various
intervals along the channel, covering a range of environments and areas of different
A moderate to poor relationship was recorded between EM31 conductivity and

pondage test seepage. However given that these represent seepage sites more than 20

km apart (significantly further than other sites tested in the trials), this was not
unexpected. The watertable along the WWC is generally beyond the penetration
depth of the EM31 and therefore the likelihood of seepage is inferred based on soil
properties beneath the channel. This will contribute to misleading results when
comparing unsaturated zone lithology to seepage rates, and is the cause of some of the
scatter in the WWC regression relationship. For example, the WWC contains a
significant silt bed which generally controls the seepage rate rather than soil properties
beneath the channel. Some of the scatter in the results could also be explained by
incorrect pond placement (ie not straddling areas of like conductivity). When points
of high variance and leverage were removed, the correlation coefficient improved to
0.62 (from 0.40).
The prediction interval bands suggested the relationship could be used to distinguish
between sites of low and high seepage, but is limited in interpreting mid-range
seepage. An improvement would be expected if the top of the watertable was
targeted, rather than inferring seepage from unsaturated zone soil properties, as much
of the seepage in the WWC is controlled by the silt layer in the channel and not the
unsaturated zone permeability. Given that the ponds were significantly spaced apart,
it was concluded that this relationship can be used for interpolation, bearing in mind
the associated broad prediction intervals associated with the regression line.
Based on the results of the pondage tests and relationship with EM31 and the drilling
program, areas recommended for remediation were finalised. Sites were defined as
either priority one or priority two seepage risk sites, depending on the degree of
perceived seepage risk. Priority one sites were considered to require remediating as
part of the upgrade, while priority two sites were to be monitored closely for seepage
following the upgrade. Given the uncertainty in the EM31 – seepage relationship, the
EM31 predicted seepage was not used as the sole means of assigning seepage risk but
geological data and visual observations were also integrated into the decision making.
The Waranga Western Channel seepage investigation is a good example of the

integration of geophysical, geological and pondage test data to determine areas of
highest seepage risk. In the end the required objectives were achieved, however the
investigation could have been improved by following the process outlined in this
report (refer to conclusions - geophysics). That is, a geophysical survey targeting the
top of the watertable, followed by test drilling to a suitable depth (based on the
geophysical survey results) and then followed up by pondage tests, also based on the
results of the geophysical survey and the test drilling. By clearly establishing this
process from the outset, this would have avoided the need for two rounds of drilling,
and probably could have provided GM-W with a more rapid answer.

7. Discussion of Results
7.1 Overview
In response to increasing concern regarding channel seepage issues, the Australian n Project initiated by
Committee of Irrigation and Drainage (ANCID) representing Australian rural water ANCID and MDBC
authorities (RWAs), in conjunction with the Murray Darling Basin Commission
(MDBC) initiated a project to investigate channel seepage measurement. One of the
main objectives of the study was to trial and document a range of seepage
identification, measurement and quantification techniques.
Channel seepage measurement trials were conducted from early 2000 to mid 2002, by n Trials were conducted
Wimmera Mallee Water (WMW), Murray Irrigation Limited (MIL) and between 2000-2002 in
Murrumbidgee Irrigation (MI). In addition, results from channel seepage four RWAs
measurement investigations conducted on the Waranga Western Channel (by
Goulburn-Murray Water) were incorporated into the final year of the trials.
Based on the outcomes of other components of the project [the Literature Review
(ANCID, 2000a), the RWA survey (ANCID, 2000b)], and consideration of the
primary objectives of the study, the trials focussed on the following techniques:
q Pondage tests n The trial program
q Point measurement (channel full and empty), developed was based
on early components
q Geophysical techniques,
of the project
q Groundwater techniques,
q Soil classification; and,
q Remote sensing.
Pondage tests were conducted at all sites, as they were the basis on which other n Pondage tests and
techniques were assessed. Drilling was also conducted at all sites in order to identify drilling were
sub-surface conditions. The final year of trials focussed on geophysics, which in the conducted at all sites.
first two years of the trial program had demonstrated the greatest potential for meeting The final year of trials
focussed on
RWA requirements for rapid and low cost channel seepage assessment.
geophysics

q Inflow-Outflow Tests: Not sufficiently accurate for measuring losses over
relatively short sections of channel (ie 1-2km). Over relatively long lengths of
channel this is an appropriate technique, and therefore the technique is suitable
for identifying and prioritising, at an RWA wide level, channels which have
higher losses compared to others in the system.
input required does not make this method practical for use by RWAs for most
channel seepage investigations.
q Hydrochemical Techniques and Tracing of Leakage Plume - The high cost and
expertise required of such trials means they are generally not practical solutions
for RWAs.
7.1.1 Assessment Methodology

In undertaking these channel seepage investigations, the basic approach adopted was:
WC1312:YR3_RPT_CH_7-10(DISC, CONC, REC, REFS).DOC Final PAGE 250

q Identification of test site locations; n A common
methodology was
adopted at each site,
q Measuring rates of seepage at test sites using direct measurement techniques – including assessment
pondage tests were used for this purpose (refer to section 7.2); of seepage with direct
q Comparison of the direct measurement technique with indirect techniques; and, techniques and
comparison with
q Extrapolation of results beyond the test zone to interpret seepage distribution – indirect techniques
this was applied for techniques which compared favourably with the direct
technique.
7.1.2 Description of Trial Sites

The seepage investigation sites all lie within the Murray Darling Basin. The channels n All trial sites were
investigated were main delivery channels, ranging in capacity from 80 ML/d (Tabbita) within the Murray
to 600 ML/d (Rocklands). With respect to lithology, sites ranged from a clay profile, Darling Basin. Sites
to a sand profile, as well as sites with rock at or near the surface. Groundwater varied in terms of
channel capacity,
salinity ranged from moderately fresh to highly saline. Groundwater depths ranged
lithology and
from very shallow (0.5 – 1.5m) to moderately deep (9-10m). Channel dimensions groundwater salinity
were reasonably similar, with the depth of water at full supply level (FSL) typically and depth
1.5m and wetted perimeters of between 9-16m.
7.2 Pondage Tests

Pondage tests involve blocking a section of channel for a period and applying a water n Pondage tests are
balance to determine the seepage losses. They are widely considered the most considered the most
accurate means of channel seepage measurement and were adopted as the baseline accurate means of
technique against which other techniques were assessed. Pondage tests were therefore measurement. They
were adopted as the
conducted across all sites, totalling 81 ponds.
baseline technique in
the trials
Seepage rates ranged in magnitude across the sites from 0.1 mm/d (Toolondo East) to
48 mm/d (Donald). The average and median seepage rate across all sites was 9.7 n Seepage rates across
all sites ranged from
mm/d and 7.0 mm/d respectively. Some sites anticipated to have high seepage rates
0.1-48 mm/d, with a
actually contained low rates (due to surface clogging layer), while others expected to median rate of 7.0
have low rates were found to have a high rate of seepage. Visible evidence of seepage mm/d
was found to not necessarily imply high seepage rates.
At three sites where pondage tests were repeated, a good degree of repeatability was n Where pondage tests
observed. The maximum difference between seepage rates was 25%. Differences in were duplicated, a good
pondage test rates from one season to another are probably attributable to changes in degree of repeatabiility
depth to watertable and channel bed properties. The differences are considered was observed
acceptable for the purposes of this investigation, and not considered to be significantly
due to errors in the pondage test method.

7.3 Sub-surface Characterisation
Sub-surface characterisation by soil and geological profiling was conducted to assist n Sub-surface
in general site characterisation (provided information on soils, depth to groundwater characterisation was
and groundwater salinity) as well as to assist in geophysical interpretation. Bores undertaken to assist in
were generally drilled adjacent to the channel, up to 10 m in depth. site characterisation
and to assist in
geophysical
An attempt was made to estimate seepage based on average soil permeabilities, using
interpretation
different weightings to test the influence of the soil across a range of depths. The
upper 2m of the soil profile gave the best indication of some relationship between
permeability and seepage rate, however no clear relationship between soil
permeability and seepage rate was obtained. A combination of factors is likely to
contribute to the absence of a relationship. Two types of factor contribute to the n No clear relationship
absence of a clearly defined relationship between seepage and soil permeability (Kv): between average soil
permeability and
q Limitations inherent in method – There was insufficient definition of changes in seepage rate was
soil type along channel (ie low sampling density). Further, the process of obtained
assigning Kv to soil type is inaccurate. The hydraulic conductivity for the
particular soil type should be field tested rather than assigned from literature.
q Factors apart from soil type are the primary control on seepage rate: The two
most common factor are:
q Bank dominated seepage (ie due to poor bank construction etc) and,
q Surface clogging layer.
These factors explain why sites like Finley and Dahwilly can have such similar
seepage rates, even though the underlying soil at Dahwilly has permeability many
orders of magnitude higher than the clay at Finley. Seepage rates at Dahwilly are
controlled by the clogging layer on the base of the channel while seepage rates at
Finley are controlled by lateral bank seepage.
The density of sampling required and the cost of seepage rate tests in specific soil n Sub-surface
types, and in addition to the fact that soil type is not always the factor controlling characterisation for
seepage, means that it is not likely to be an accurate or cost effective means of seepage quantifying channel
quantification. However it remains a critical part of the data gathering and site seepage is unlikely to
be accurate or cost-
characterisation phase of a channel seepage investigation.
effective
7.4 Point Tests

Five point test trials were conducted during the investigation, using ring infiltrometers, n Five point test trials
disc permeameters and Idaho seepage meters. The trials were conducted at Toolondo were conducted
(Central), Dahwilly (Central), Tabbita, and the Donald Main Channel. during the
investigation using
ring infiltrometers,
i) Dahwilly (Central): Ring Infiltrometer and Disc Permeameter – Both the ring
disc permeameters
infiltrometer and the disc permeameter failed to characterise the true seepage and Idaho seepage
rates of the ponds. This is most likely due to the inadequate sampling density meters
of the testing program. However, results were of the same order of magnitude
as the pondage tests. The results of the two techniques were very poorly
correlated against each other. This is probably due to error in the infiltration
rings.
ii) Toolondo (Central): Ring Infiltrometer and Disc Permeameter - Seepage rates
obtained from the disc permeameter and ring infiltrometer tests on the
Toolondo channel were several orders of magnitude higher than actual

seepage rates (from the pondage tests). Tests measured the hydraulic
conductivity of a sand layer in the channel, rather than the underlying more
restricting clay and silt layer. In a relative sense the tests distinguished
between the higher seeping pond and the lower seeping pond. Best results
were obtained for the lower quartile data indicating that the higher results
were unreasonably biasing the true seepage rates.
iii) Dahwilly (Central): Idaho Seepage Meter – For three of the ponds there was a
distinct linear trend between the Idaho seepage meter results and the pondage
test results. However the lowest seepage pond (8 mm/d) returned the highest
weighted Idaho meter reading of 59 mm/d. This eliminated any overall
correlation in the results. It is concluded that too few Idaho tests were
conducted to adequately characterise the channel. Idaho seepage meter rates
were generally two to three times higher than the pondage test rate, but in the
‘outlying’ pond was eight times higher. It is surmised that the base of the
channel is seeping at a higher rate than the walls, a theory supported by site
geology.
An attempt was made to correlate the Idaho seepage meter results with the
EM31 survey data. The resulting correlation was inverse, as expected, but the
fit was quite poor. The EM31 conductivity did not distinguish between Idaho
seepage meter rates in the 5 – 40 mm/d range, but did differentiate the two
highest seepage sites (80 mm/d). Definitive conclusions could not be drawn
from these results as the conductivity range is very narrow (39 –45 mS/m) and
the data set small.
iv) Tabbita: Ring Infiltrometer – Ring infiltrometer tests were conducted in three
ponds on the Tabbita Channel. The ring infiltrometer results did not
distinguish between the pondage test seepage rates of the ponds. Median
seepage rates for the three ponds were all between 4-5 mm/d, compared to
pondage tests rates of 6-8.5 mm/d. The very narrow spread of the pondage
tests range at the site limits the statistical significance of these results. The
lower ring infiltrometer seepage compared to the pondage test seepage is
either due to the low sampling density of the testing program (therefore
missing ‘hotpots’) or to the fact that walls of the channel are seeping more
than the channel bed.
v) Donald Main Channel: Idaho Seepage Meter – Idaho seepage meter tests
were conducted in four ponds on the Donald Main Channel. The seepage
meter results were comparable in magnitude to the pondage test results. The
correlation between the pondage test results and the Idaho results was
moderate to poor. The limited number of pondage sections on which the trend
is based (4) and the limited number of Idaho tests within each pond (5-6) are
contribute to the poor correlation.
An attempt was made to correlate the Idaho seepage meter results with the n These trials confirmed
EM31 survey data. No correlation was observed. The EM31 conductivity that point tests are
clearly distinguished between seepage rates in two of the ponds, but no generally not reliable
distinction was made in the remaining ponds. The reason for this is unclear. for directly
quantifying seepage,
due to the
These trials have confirmed that point tests are generally not reliable for directly
impractically large
quantifying seepage. Due to variable and sometimes erratic values obtained in number of tests
measurements, the trials have illustrated that a large number of tests is required to required
sufficiently determine the true seepage rate of a section of channel. Therefore they are

generally not considered reliable for absolute quantitative purposes and should
generally be limited to determining the distribution of seepage losses ( i.e., relative
seepage). Even for this use a large number of tests are recommended to minimise the
effects of local variability. These conclusions equate to the findings of the literature
review (ANCID, 2000a).
In addition, it was apparent in a number of channels that the bed of the channel was n Beds of the channels
seeping at a different rate to the walls of the channel. This appeared to be occurring at appeared to be
a number of the point test sites, as evidenced by higher seepage rates in the base of the seeping at a different
channel than the pondage test rates. This is in contrast to the normal phenomenon rate to the walls
with point tests where lower seepage rates than actual are often obtained (due to the
non-detection of seepage ‘hot spots’). In these cases, even very high density point test
sampling in the bed of the channel cannot determine the actual seepage rate.
In terms of choice of equipment for point testing:

q The Idaho seepage meter appeared to provide the most reliable results of the three n The Idaho Seepage
instruments. This concurs with the fact that the channel is full during the test and meter appeared to
that truly saturated flow is being measured. However there are very few provide the most
operators skilled in the use of the equipment and therefore testing is limited by reliable results of the
three instruments
their availability. The tests are also very expensive, due to the fact two operators
are required, including one skilled in the use of the meter.
q Definitive comments cannot be made regarding the accuracy of the disc
permeameter compared to the ring infiltrometer. Some trouble was encountered
however with the ring infiltrometer in terms of seepage outside of the ring. The
disc permeameter is simpler to use than the ring infiltrometer, both in terms of
operation and manual handling.

Groundwater observation bores are often required as part of the site characterisation n Relatively rapid
phase of a channel seepage investigation. Quantitative analysis of seepage rates was responses in ground-
conducted on the Donald Main Channel based on changes in groundwater level before water levels were
and after channel filling. Groundwater response to the operation of the channel is was observed adjacent to
Donald Main Channel
observed in adjacent channel bores, with rises of more than two metres within weeks
in response to channel
of channel commencement. Bores at 50m distance from the channel displayed rises filling/emptying
between 0.5 m and 1 m.
Groundwater levels at the Donald Main Channel were used to estimate seepage n Groundwater levels
adjacent two bore lines using the Dupuit Forcheimer equation for flow in an were used to estimate
unconfined aquifer. Assuming an aquifer thickness of 10m, seepage estimates seepage rates
approximately equal to pondage test seepage were obtained for an assumed hydraulic
conductivity of 0.2 m/d. Qualitative assessment only was conducted on the Tabbita
site. Very clear response to channel shutdown was observed in groundwater
hydrographs at this site. n Use of groundwater
bores for quantitative
Use of groundwater bores for quantitative analysis of seepage rates is not considered analysis is not
an accurate or cost effective means for typical Rural Water Authority channel seepage considered accurate
investigations. In order of increasing importance the method is not considered or cost effective for
accurate due to: most RWA purposes

q Sensitivity to hydraulic conductivity inputs (eg Tabbita – depending on input
hydraulic conductivity, seepage rates of varying orders of magnitude can be
obtained) and the cost of obtaining sufficiently reliable estimates.
q Relies on assumptions regarding pre-channel groundwater levels. These can be
estimated from conditions before and after channel filling, but depending on site
hydrogeology this assumption may or may not be accurate.
q It is essentially a type of point test and does not answer the question of what area
of the channel is seeping. A high density of bore transects would be required for
meaningful identification of local areas of channel seepage.
q Relies on an assumption of aquifer thickness (this may be able to be calculated
but in a deep aquifer this may be very expensive).
However, groundwater observation bores are a very valuable part of the site n Groundwater bores
characterisation phase of a channel seepage investigation. Further, groundwater bores are a very useful post-
are a very useful post-remediation assessment tool, particularly for assessing the remediation tool
7.6 Remote Sensing

A remote sensing trial was planned in the Wimmera. However the trial was not n A remote sensing trial
undertaken due to RWA budget constraints. Previously collected data (ie. prior to was planned but
ANCID trials) was evaluated with a view to incorporating the results into the project eventually not under-
but unfortunately the data was not in a form suitable for inclusion in the project. taken due to budget
constraints
While remote sensing trials were not conducted, based on the literature review and
preparation of the brief for the proposed trials, it is concluded that remote sensing
techniques:
q Are best suited to investigations where the primary aim is identification of land n Remote sensing
degradation associated with channel seepage. It should not be used if it is known techniques are best
that the seepage mechanism is predominantly vertical, such as likely to occur at suited to investigations
sites with a deep watertable; concerned with land
degradation associated
q Will be most useful in environments where lateral seepage is predominant. For with channel seepage
example, sites with a high watertable, shallow impermeable layer or bank seepage and/or where lateral
- these environments represent conditions most likely to facilitate lateral seepage seepage is predominant
and cause the seepage to have a surface expression; and,
q Remote sensing should primarily be regarded as a seepage identification tool and
not for seepage quantification purposes.
7.7 Geophysics
The geophysics conclusions are set out in the following manner:
q General Conclusions
q Summary of Resistivity Results

7.7.1 General Conclusions
Geophysical techniques applied to seepage measurement primarily depend upon n Geophysical
measuring a contrast in terrain conductivity (or resistivity) in the sub surface profile techniques identify or
around the channel. They can be used in one of two ways: measure seepage by
detecting contrasts in
1) Directly measuring the conductivity of the groundwater, and identifying the terrain conductivity by
conductivity contrast of fresher channel water as it seeps into and dilutes more either:
saline native groundwater. Decreasing the salinity of the groundwater will cause a n Directly measuring
decrease in electrical conductivity (or an increase in its inverse, resistivity). seepage induced
changes in ground-
2) Identifying contrasts in soil properties and inferring the likelihood of seepage water conductivity, or
through more permeable materials in the zone above the watertable. Formations n Identifying contrasts
more likely to allow seepage, such as sands, are naturally lower in conductivity in soil properties
(higher in resistivity) due to lower porosity and lower cation exchange capacity above the watertable
Figure 7-1 visually depicts how these two different approaches can be used to identify
or infer seepage.
n Figure 7-1 Comparison of how geophysical techniques can be used to

identify channel seepage (LHS – inferred from soil property variations, RHS –
Seepage inferred Seepage measured directly

Clay
Sand
Saline groundwater

Technically the second method of ‘detection’ is not really detection, but the magnitude n The inferred method
of seepage is assumed to be related to unsaturated zone soil properties beneath or of ‘detection’ assumes
adjacent to the channel. In many cases this is a reasonable assumption. However the the magnitude of
unsaturated zone is not necessarily the controlling influence on seepage. For example, seepage is related to
unsaturated zone soil
over time most Australian channels tend to silt up and the resulting surface clogging
properties
layer is often more restrictive than the unsaturated zone. Therefore unsaturated zone
lithology may not be related to seepage rates, as seepage is controlled by the thickness
and conductance of the clogging layer.
Nevertheless, it was found that the inferred method of identifying contrasts in soil n The inferred method
properties (ie, where the watertable was deeper than the penetration depth of the ‘detection’ was
geophysical equipment) was successful at most sites conducted during the trials successful at most
(ANCID, 2003). There is less risk however in using the direct method of seepage sites but not all.
There is less risk in
detection, because as the name implies it is not inferred, but direct. An example of
the direct method of
where the ‘inferred’ method of detection did not work was at Dahwilly Central where detection.
an EM31 survey was conducted while the channel was not running. The survey was
therefore measuring changes in the unsaturated zone and not in the groundwater. At
this site the silt layer in the channel, not the unsaturated zone is the restrictive layer
and therefore no correlation was observed. When the survey was repeated with the
channel running, a good correlation was obtained.
Some possible limitations of the direct method of seepage detection are listed below: n Possible limitations of
(However, it is still considered the preferred technique over inferring seepage based the direct method of
on soil property variations). detection are in non-
q In relatively non-saline groundwater environments, the fresh seepage water will saline groundwater
environments, sites
not contrast with the native groundwater. As a guide it is recommended that the
with steep ground-
groundwater salinity is at least 3 to 4 times higher than the channel water salinity. water gradients or
This is not expected to be a problem in most Australian conditions; high transmissivities
q In environments where the channel seepage water might be rapidly mixed with and sites with highly
native groundwater, such as sites with high groundwater gradients or highly variable groundwater
transmissive environments, the salinity impact on the groundwater may not be as salinity
significant. This can largely be overcome by using survey traverses close to (or
on) the channel; and,
q Groundwater salinity variations along the channel will affect the results and this
needs to be allowed for in the interpretation.
In summary, it is very important that the depth to watertable is known at the site n Knowledge of depth to
before selecting a technique. Based on this information a decision can be made as to watertable is
whether direct or inferred measurement will be undertaken and hence the technique important before
that will be adopted. technique selection

q Accuracy
q Cost and Speed
q Data Processing

Accuracy
In theory on-channel resistivity surveying should be the most accurate of the n Theoretically resistivity
geophysical techniques trialed, as it is based on a direct method of seepage detection surveying should be the
(refer Section 7.7.1.1). As the technique allows definition of changes in resistivity / most accurate
conductivity through the profile, the depth where seepage impacts will be most technique. At most sites
resistivity was
evident (at and below the watertable) can be targeted. At most sites resistivity
comparable to EM and
surveying results were comparable to EM31 and EM34, and at three sites (Dahwilly at three sites the results
Central, Dahwilly East and Finley) the resistivity correlations with pondage tests were were more accurate
better than the EM31 and EM34 correlations. The Dahwilly site demonstrates the
benefits of targeting the watertable for seepage detection in an environment where
seepage is not controlled by the unsaturated zone, but by a surface clogging layer in
the channel.
Resistivity did not prove to be quite as ‘accurate’ as EM31 in environments with a

shallow depth to watertable. This was largely attributed to poor near surface
resolution of the particular resistivity equipment used in the survey, and not inherent
in the resistivity method itself. The resistivity surveying was conducted using
equipment that was used for the first time in these trials and as such was largely
experimental. The designers of the equipment indicate that improved resolution at
shallow depth could relatively easily be achieved in future surveys, by using
exponentially rather than linearly spaced arrays (Allen, pers. comm. 31/10/02).
EM34 at 10m coil separation in the horizontal mode provides a similar depth n Fixed array
penetration to EM31 (vertical mode) and therefore is similarly accurate (but slower to geophysical surveys
use). EM34 at a 20m coil separation provides a deeper penetration and focus. At one are limited in that the
trial site, the depth focus was apparently too far below the watertable and the critical result is averaged
over a specific depth
zone was missed. This is a fundamental limitation with all Geonics EM surveys and
interval
other such fixed array type geophysical surveys – the result is averaged over a specific
depth interval, which may not be the critical interval of interest.
However, the robustness of the EM31, as demonstrated by the consistent results in the n EM31 was generally
trials is due to its relatively shallow depth focus (1-4m). For channels where there is a demonstrated to be a
shallow watertable (eg, surface to 3-4m), EM31 can be used for direct measurement of robust technique at
seepage, which as discussed above is likely to be more reliable. When the watertable both deep and shallow
watertable sites
is deep, EM31 infers seepage from near surface soil properties. This works in most
instances but may break down where clogging processes rather than unsaturated zone
lithology control seepage.
The significant advantage of resistivity surveying is that the final output is a two n Resistivity surveying
dimensional profile of resistivity beneath the channel. Not only does this allow easier has advantage of
interpretation of the results but it can also provide an indication of seepage providing a profile of
mechanisms. For example, at the Toolondo Central site the resistivity profile shows resistivity beneath the
channel
isolated sections of high resistivity (low seepage) emanating from the channel. This
is in contrast to the Dahwilly channel where the profile suggested seepage by
relatively continuous diffusion along the channel. This seepage mechanism is
supported by the lithology at the Dahwilly site, which indicates the entire length of
channel is underlain by approximately 10m of medium to coarse grained sand, and
hence is more likely to result in uniform seepage rather than seepage ‘hotspots’.

Cost and Speed
Approximate costs for the three techniques trialed are given below, based on n The unit cost per
geophysical surveys undertaken in these trials. However, it is important to note that kilometre will reduce
for geophysical surveys often a significant proportion of the costs are overhead costs as the survey length
(mobilisation, equipment set up etc) and therefore the unit cost per kilometre will increases
usually be substantially less for long sections of channel. Costs will obviously vary
depending on site specific conditions (eg, on land – fences, other obstacles, on channel
– bridges, checks, fences etc).
q EM31 Surveys:
q Wimmera Mallee Water: For 6 kms, on-land including 4 traverses on each n EM31 is currently the
side of channel (over 3 sites): $400/km (includes mobilisation, data cheapest geophysical
processing and mapping). method due to the
speed of data
q Murray Irrigation: For 8 kms, on-land including 4 traverses on each side of acquisition
channel (over 4 sites): $340/km (includes mobilisation, data processing and
mapping).
q Murrumbidgee Irrigation: On-land including 4 traverses on each side of
channel, on each side of channel, the unit cost ranged from $650/km (3km
section) to $800/km (1 km section). (includes mobilisation, data processing
and mapping). On-channel survey cost was $330/km for a 3 km section.
EM31 is currently the cheapest of the geophysical methods due to the speed of
data acquisition.
q EM34 Surveys:
q Wimmera Mallee Water: For 4 kms over 2 sites: $250/km (1 traverse only n EM34 is more
on one side of the channel) ie, $500/km for both sides of channel (excludes expensive than EM31
mobilisation). as two people are
required for operation
q Murray Irrigation: For 6 kms (on each side of channel) over 3 sites: $435/km
(includes mobilisation).
EM34 is more expensive than EM31 as two people (on foot) are required to
operate the equipment.
q Multi-electrode Resistivity Surveying – The follow costs were for resistivity n Resistivity surveying
surveying across 11 sites (approximately 2km each in length) in the Wimmera, costs for seepage
Murray and Murrumbidgee Irrigation areas: assessment are
difficult to quantify as
q Resistivity towed array surveys: $900/km [Includes mobilisation (from the technique is new –
Adelaide), travel between sites, production and all equipment costs] costs are likely to
q Data processing costs: $220/km. come down over time
Note that resistivity surveying costs are difficult to quantify given that the on-
channel application of the technique is relatively new. Costs are likely to come
down as the technique is refined, the equipment becomes commercially available
and subsequently competition is introduced.

Australia, sufficient to ensure reasonable competition and prices. At present on-
channel resistivity surveying is still in a development phase and as such there are no
commercially operating contractors who specialise in this type of survey. However, a
number of geophysical exploration / surveying companies have the capability to
develop this type of equipment (such as the company who conducted these trials) and
should the demand for such surveying increase, it is expected other companies could
also develop this capability. However at present this may be a constraint on resistivity
surveying.
Data Processing
comparison, data processing requirements for resistivity surveying are considerable,
due to the cost of inverting the data to produce a resistivity cross section. This is not
really a constraint of the technique, but adds to the overall cost of resistivity
surveying. Care needs to be taken to ensure that the contractor undertaking the
resistivity surveying also has the capability to undertake the data processing.
Approximate costs for data processing are provided above.

q Survey timing – The timing of the geophysical survey will depend on the method
of seepage detection being used. If seepage is being inferred from soil properties
then the timing of the survey is not critical and can be conducted whether the
channel is running or empty. However if direct measurement of seepage is used,
the survey must be conducted while the channel is running, and preferably after it
has been running for a least one month (depending on depth to watertable and
vertical hydraulic conductivity), to ensure seepage has impacted the groundwater.
q On-channel versus on-land – During the trials, on-channel (ie, in a boat) EM31
surveys:
q Did not work at one site where the watertable was beyond the range of the
EM31 and returned similar (reasonable) results to the on-land survey at
another site (Waranga);
q Did work at sites with a shallow watertable; and,
q Were partially successful when the watertable was located at the edge of the
depth penetration capacity of the EM31.
Further work is required in this area, but the evidence collected in this n On-channel EM31
investigation suggests on-channel EM31 surveys should only be conducted where surveys should
the geophysical technique can penetrate into the watertable, and ideally target the generally only be
top of the watertable. In other words, the method of inferred seepage based on conducted when depth
unsaturated zone soil properties does not appear to work on-channel. It is to watertable is 3-4m
or less
apparent that the flushing effect immediately beneath the channel is dominating
changes in lithology. For EM31 systems this would preclude their on-channel
use when the watertable is deeper than approximately 3-4m.
n More consistent EM
results were returned
However, there is some conflicting evidence, as demonstrated by the trial results on-land than on-
summarised above. Overall however, the most consistent (EM) results were channel.
returned on-land and this is considered the safest option. A possible limitation of
on-land surveying, may be at sites which contain significant land salinisation

immediately adjacent to the channel. Therefore, if the budget allows, it is
recommended that both on-land and on-channel surveys be conducted.
Resistivity surveys can be conducted on-channel because of their variable and
greater depth penetration capacity.
n On-land surveys
q Off-set distance and location for on-land surveys – The evidence collected in should be conducted
these surveys indicates the best off-set distance for on-land surveys is adjacent to the outside
immediately adjacent the outside toe of the down slope side of the channel. For toe of the channel
either method of seepage detection this is recommended. For inferred seepage
‘detection’ the soil type next to the channel is most likely to be representative of
the soil type beneath the channel. For direct measurement, immediately adjacent
tothe channel will be the zone of greatest seepage impact on the watertable.
Away from the channel this impact will be diluted. However at sites without a
steep gradient or high transmissivity, an average of survey traverses up to 50m on
each side of the channel was found to improve the correlation between seepage
and the geophysical survey at most sites.
Traverses on either side of the channel are recommended. However if the budget
is a significant constraint, a traverse on the down-slope side of the channel should
be the priority.
q Other variables:
q Trees – In two surveys (Rocklands and Donald), tree plantations adjacent to n Other variables such
the channel appeared to interfere with the survey results. The postulated as trees and rain may
mechanism is that the trees are consuming the seeped water and therefore the interfere with survey
observed impact (in the geophysical survey) on the native groundwater is results
lessened.
q Rain – Rainfall did not interfere with the surveys conducted in these trials.
However it is possible that surveys conducted after heavy rainfall on light to
moderate soil types (ie which allowed significant infiltration) could interfere
with the conductivity / geophysical response and therefore should be
avoided. Surveys inferring seepage based on shallow soil properties or
direct measurement in shallow watertable environments would be most
affected.
Generally a high degree of repeatability was observed between duplicate surveys. At n Change in ground-
two sites where there was a significant difference in the results, changes in water elevation
groundwater conditions due to channel operation accounted for the difference. These accounted for any
sites are described below: differences in
duplicate surveys
q Donald - A generally consistent increase (approximately 15 mS/m) was observed
across the surveyed area between the October 1999 survey and the September
2001 survey. This increase was caused by the more saline conditions at the time
of the 2001 survey. The channel had been running for six months prior to the
1999 survey, creating a sub-surface environment dominated by fresh water and a
flushed profile. The reduced channel running time prior to the 2001 survey
meant a relatively more saline profile and hence higher conductivity.
q Dahwilly - The average EM31 conductivity for a survey conducted when the
channel was not running was less than half the conductivity recorded while the
channel was running. This is different to what was observed at Donald, due to

the different depth of groundwater at the two sites. When the channel is not
running at Dahwilly, the watertable is largely out of reach of EM31 detection and
the response is a reflection of the coarse and low conductivity sands in the
unsaturated zone. When the channel is running, the watertable is elevated into
the range of the EM31 detection and hence conductivities increase significantly.
For all of the sites used in the final year of analysis an attempt was made to look for n An attempt was made
potential correlations between seepage rates across all sites and geophysical response to look for potential
(EM31 and resistivity). This was conducted using multiple linear regression and correlations between
simple linear regression. The additional explanatory variables included in the multiple seepage rates,
geophysical response
linear regression analysis included:
and other key
q Soil permeability (vertical hydraulic conductivity): Averages across 4 different
variables
depths were used;
EM31 Multiple Linear Regression

Multi-variate regression analysis was initially conducted on the entire data set n For the EM31 multi-
(excluding Finley which was removed as an outlier). This indicated that the following variate regression
were significant explanatory variables: analysis, average
EM31 conductivity,
depth to watertable
and the permeability
q Upper Kv (vertical permeability of the top 2m of the profile).
of the upper 2m were
significant
However, the standard error of estimate (expressed as a percentage of the mean explanatory variables
pondage test seepage) for the regression was high (82%). A plot of conductivity
versus seepage dividing the data into two categories based on depth to watertable is
n Figure 7-2 Regional EM31 Assessment: Pondage Test Seepage Versus

EM31 Conductivity with Sites Divided Based on Depth to Watertable
50
2
R = 0.889
40
30
20
y = -0.129x + 12.788
2
R = 0.469
10
0
0 20 40 60 80 100 120

Based on this clear division between sites with shallow and deeper watertables (in an n A clear division
attempt to improve the accuracy of the fitted regression model) the sites were split into between sites with a
two data sets based on these two categories of depth to watertable. The difference in shallow and deep
EM31 response between a deep and shallow watertable site is explained by the effect watertable was
observed. The data
of the shallow and saline groundwater which for a given seepage rate causes a much
sets were split along
higher conductivity response. these lines
Once the data was split into these two categories, depth to watertable was not found to
be a significant explanatory variable, as would be expected. The findings for the two
categories are summarised below:
q Watertable Depth Five to Ten Metres - For sites where the watertable is 5-10m n For sites with a water-
below surface, the equation found to provide the best prediction of channel table 5-10m below
seepage was: surface, EM31 was the
dominant explanatory
Seepage = 11.6 – 0.12 EM31 + 4.4 UKv (Equation 1) variable. Soil
permeability was of
secondary importance
The equation:
conductivity of secondary importance. Groundwater salinity and depth to
groundwater were not found to be significant explanatory variables in the
analysis.
q Watertable Less Than Two Metres - For sites with a watertable within two metres n For sites with a water-
of the surface multiple linear regression analysis did not find any other variables table within 2m of the
that were significant explanatory variables beyond EM31. The fact that soil data surface EM31 was the
was not significant is expected as the groundwater near the surface is likely to only significant
dominate the response. It was somewhat surprising that groundwater salinity was explanatory variable
not found to be a significant variable for this data set, and is probably a reflection
of the limited number of sites (three) that make up the data set.
EM31 - Simple Linear Regression

represents. The data was again divided based on depth to watertable.

For the sites with a watertable 5 – 10m below surface, the best fitting linear regression
equation was found to be:

The equation:
• Was established with 40 data points; n Confidence intervals
for the regional simple
• Has a correlation coefficient of 0.47; linear regression
• Has a standard error of estimate of 51% (of mean observed seepage rate) equation indicates the
equation is useful for
Confidence intervals (for both 80% and 90%) for this relationship were established broadly classifying
and showed that the prediction equation is accompanied by quite broad prediction seepage rates
bands. This probably limits the use of this regional equation to broadly classifying
seepage rates (eg into low, medium and high categories), which is not a surprising
outcome given the wide range of sites represented by the equation.
A plot of actual seepage versus predicted seepage shows that the equation tends to n An exponential
over-estimate seepage for low seepage rates (less than 5 mm/d) and under-estimate for regression equation
high seepage rates, implying a non-linear equation may provide a better fit. An displayed a slightly
exponential regression equation was applied to the data. The fitted exponential curve less accurate
statistical fit but
showed that while there is only a marginal improvement in the correlation coefficient visually appears more
for the exponential fit (from R2 = 0.47 to 0.48), the standard error actual worsens appropriate
(from 51 % to 62%). Therefore, overall a less accurate fit is obtained using the non-
linear equation, even though it may visually appear to fit the data better.
However the advantage of the exponential fit over the linear fit is that there is less of
a pattern displayed in the observed versus predicted seepage plots (which suggests that
the more realistic model may in fact be the exponential model). Further, the linear
model places a maximum limit on the seepage of about 12 mm/d, whereas the
exponential model appears to be more realistic, allowing for higher seepage rates in
the very low conductivity range (up to around 20-25 mm/d).
Comparing the linear regression to the multiple regression, the statistics indicate that n Only a marginal
only a marginal improvement is made to the accuracy of the regression fit in the improvement is
multiple linear regression analysis (Equation 1), compared to the simple regression fit observed in the
(Equation 2). The R2 for Equation 1 was 0.55 and the standard error of estimate was multiple regression
compared to the liner
regression
permeability to the regression equation.

The best fitting linear equation for sites with a watertable less than 2m is:

The equation:

The high correlation coefficient value and the relatively low standard estimate of error n For linear regression
suggest a good correlation for the variables. However the results should be tempered at sites with a shallow
by the fact that relatively few data points were used to form the relationship. To watertable a good
improve confidence in the regression equation for the watertable less than two metres statistical fit was
obtained, but
scenario, additional points are required in the data set. relatively few data
points were used to
Prediction bands (80% and 90%) for estimating seepage based on EM31 response form the relationship
when the watertable is less than 2m, indicate that while the prediction intervals are
broader in magnitude than for the prediction bands for the deeper watertable scenario,
as a percentage of the seepage range covered by each of the equations, they are
narrower.
In summary, multi-variate analysis did not significantly improve the regression model. n Multi-variable
The permeability of the upper part of the soil profile was found to be a significant analysis did not
parameter, but the improvements to the model with this parameter included were significantly improve
marginal. The cost of conducting field tests to collect this data therefore probably the regression model
outweighs the benefits.
The linear regression equation for sites with a watertable five to ten metres below
surface is a reasonable fitting equation, given the range of sites on which it is based.
However it should not be relied upon to accurately predict seepage, and should be
limited to assigning seepage to low, medium and high categories. The same
comments are applicable to the regression equation developed for sites with a shallow
watertable (less than two metres). The better statistics for these sites are attributed to
the fewer data points and smaller range of environments represented.
Resistivity Multiple Linear Regression

As for the EM31 analysis, the same approach of a regional assessment using multiple
and linear regression analysis was conducted for the resistivity results. The same
additional variables were used. For the resistivity the ten metre depth slice was
adopted. While a more accurate analysis would use the depth at and just below the
watertable, for the purpose of a consistent approach, this depth slice was selected. The
first cut analysis indicated that depth to watertable was an important variable.
Therefore the analysis was again based on division of the data into sites of shallow
and deep watertable.

For sites where the watertable was five to ten metres below surface (Toolondo and
Dahwilly) the equation found to provide the best prediction of channel seepage was:

The equation:

Various transforms were examined to improve the accuracy of the regression. It was n For sites with a water-
found that raising the seepage to the power of 0.2 improved the model with respect to table 5-10m below
the standard error of estimate, which was reduced to 19% of the mean observed surface, resistivity and
seepage rate (to the power of 0.2). A marginal reduction in the correlation coefficient the permeability of the
upper 2m were
was observed, decreasing to R2 = 0.42. significant explanatory
variables. A power
The non-linear equation, using resistivity only as an explanatory variable, for transform slightly
predicting channel seepage at sites where the watertable is greater than 2m is: improved the regression
statistics
5


For sites with a watertable within two metres of the surface (Lake View and Lake n For sites with a water
View West), multiple linear regression analysis did not find any other variables that table within 2m of the
were significant explanatory variables beyond the resistivity data. surface, resistivity was
the only significant
explanatory variable
Resistivity Simple Linear Regression

For the resistivity analysis the number of channels fitting this category was reduced to
two (3 sites at Toolondo and 2 sites on the Dahwilly channel). Four points within this
data set appeared to be outliers. They were the four high seepage ponds at Dahwilly
Central. However they were retained in the analysis as there was no obvious grounds
for their removal.
Therefore the linear equation (using resistivity only as an explanatory variable)

predicting channel seepage at sites where the watertable is greater than 2m was found
to be:

The equation: n For simple linear

• Was established with 23 data points; regression at sites
with a watertable 5-
• Has a correlation coefficient of 0.16; 10m below surface, a
• Has a standard error of estimate of 68% (of mean observed seepage rate) very poor statistical fit
was observed between
These statistics indicate that the accuracy of the regression is very poor, in large part resistivity and seepage
due to the four high seepage rate ‘outliers’ at the Toolondo Central site. With these rate, largely due to
outliers excluded the correlation coefficient improves dramatically to R2 = 0.63. several outliers
However as discussed above there was no obvious basis for their removal.
n A power transform did
Various transforms were examined to improve the accuracy of the regression. It was not significantly
found that raising the seepage to the power of 0.2 improved the model with respect to improve the
the standard error of estimate, which was reduced to 20% of the mean observed regression statistics


5

In summary, neither the linear nor the non-linear simple regression equations were
found to be satisfactory predictors of seepage. The very wide prediction bands for the
non-linear prediction equation confirm this conclusion.

The linear equation predicting channel seepage at sites where the watertable is less n For linear regression
than 2m (data set comprised of only the Lake View channel) is: at sites with a shallow
watertable, a good
Seepage = 1.7 Resistivity10m - 0.66 (Equation 8) statistical fit was
obtained but the
relationship is based
on relatively few data
pointsl
The equation:
These statistics indicate that the accuracy of the regression is reasonable. However,
the results must be interpreted in light of the fact that only a few data points have been
used to form the relationship, and all data points were collected on the same channel.
Further testing to add different environments to this data set is necessary before a
reasonable degree of confidence can be placed in this prediction equation.
The following summary comments are made regarding the linear and multiple
regression analysis for the resistivity:
q The multiple regression analysis significantly improved the accuracy of the n It is likely the
regression equation compared to the simple regression with only one variable, resistivity regression
increasing the coefficient of determination from R2 = 0.21 to R2 = 0.42. analysis could be
significantly improved
q It is likely the analysis could be significantly improved by using resistivity data at by using resistivity
and immediately below the watertable for each of the sites. data at and
immediately below the
q As was the case in the EM31 multi-variate analysis, the variable which was found watertable
seepage.
q While this analysis indicates a moderately fitting regression equation, it could not

q The few number of data points and small range of environments represented
by the data points;
q The unexplained outliers in the resistivity analysis that were not present in
the EM31 regression equation.
geophysical response are:
1. What confidence is there that the derived relationship accurately describes seepage
within the area tested?; and,
2. How confidently can the relationship be used outside of the area tested to predict
seepage?
Based on the findings of these investigations, these two issues are summarised below.
Confidence in Derived Relationships

Confidence in the derived seepage-geophysical relationship within the area tested can
be assessed by a number of factors, including:
q Correlation coefficient (R2) – This explains the amount of variation explained by n Confidence in the
the regression equation. Most geophysical – seepage relationships derived in this derived seepage
investigation had correlation coefficients of between 0.5 to 0.9, and typically geophysical
were around 0.75, meaning that 75% of the seepage variation could be explained relationship within an
area can be assessed
by the geophysical response.
by the correlation
q Standard estimate of error – This is a measure of the degree of scatter about the coefficient, standard
regression line. A data set may have an R2 of one but a high degree of scatter. estimate of error,
For the regional simple linear regression (EM31) the standard error of estimate prediction interval,
(of mean seepage rate) was around 50% for sites with the watertable 5-10m number of data points
and seepage rate
below surface and was around 25% for sites with the watertable less that 2m. For
range
individual channels this was generally lower at about 20%.
q Prediction interval - Prediction bands for most seepage - conductivity / resistivity
regression lines were generally quite broad. These bands suggest that often the
regression relationship can only be used to classify areas into high, moderate and
low seepage. The uncertainty is partially due to the data handling processes,
which are based on averaging pondage test seepage and geophysical response
over long sections of channel. There is also error inherent in the method in that
given the large number of variables that simultaneous impact on channel seepage,
it is not possible to tightly characterise seepage based on geophysical response.
The prediction intervals were generally tighter for sites with greater ranges of
seepage. Prediction intervals are likely to be improved by a greater number of
pondage tests across the broadest possible seepage range.
Knowing that seepage is probably within a certain range (to a given level of
certainty), even if the range is fairly broad, is still considered an improvement on
the existing seepage knowledge base of many Rural Water Authorities, and can
only lead to more informed decision making.

q Number of data points – The number of data points on which the relationship is
derived is also very important. For instance, some sites with relatively few data
points returned very good statistics (high R2, low SEE and relatively tight
prediction bands). These results need to be viewed with caution because of the
low number of points contributing to the relationship (eg, only two data points
will have perfect statistics but the results have no meaning). Generally the more
data points contributing to the relationship, the greater the confidence in the
relationship.
q Seepage range – The seepage range over which relationship is established is
important, as it improves confidence in the robustness of the relationship. A
limitation at a number of sites in this study was the narrow range of seepage rates.
The trends observed across a very tight seepage range can often be meaningless,
and at a number of sites where tests were conducted several kilometres away on
the same channel, a more realistic relationship for that channel was derived.
Confidence in Extrapolation
The following points need to be considered when extrapolating a geophysical –
seepage relationship outside of the area in which it was developed:
q Was the relationship strong in the area tested – This is the first test. If the n When considering the
relationship was not strong in the area in which it was derived (refer to above validity of extrapolat-
discussion) then there will be little confidence in extrapolating such a ing a geophysical-
relationship. seepage relationship
to a new area, the
q How representative is the area in which the extrapolation is to occur, of the area strength of the
in which the relationship was developed. The area in which the relationship was original relationship
developed must encompass the range of conditions over which the extrapolation needs to be assessed,
is to occur. This may be quite difficult to determine. While soil, geological, and the representat-
hydrogeological maps and even test drilling provide an indication of changes iveness of the new
area to conditions
along the channel, the results from these trials suggest that these are generally not
where the relationship
at a sufficient scale to detect how they will impact on geophysical response. For was derived
example, at the Dahwilly sites which showed reasonably similar characteristics
(depth to watertable, groundwater salinity and lithology), the EM31 conductivity
response was very different and thus extrapolation from the Dahwilly Central site
to the Dahwilly East site would have resulted in significant errors in seepage
estimate.
In fact, the evidence coming out of the trials in terms of extrapolation, even to
sites that were apparently similar and usually only several kilometres along the
same channel, was that the derived relationship was not suitable to predict
seepage at the new site. The key outcome of this is that unless intensive data
collection is conducted to ensure continuity of site conditions to the area of
extrapolation, interpolation rather than extrapolation must be used. This is
explained in further detail in the following section, but essentially means that
pondage tests should be conducted at regular intervals along the entire section of
interest, to ensure that the full range of site conditions is accounted for in the
derived regression relationship.
The more data points collected from different sections along the channel that are
added to generate the regression line, may increase scatter about the regression
line. This can be seen in the high standard error of estimates for the
Rocklands/Toolondo and Waranga regression equations. This is a reflection of

the fact that the ponds in these sites covered a wide area and range of sub-surface
conditions. However, while there is more scatter about these lines, there is more
confidence in using these relationships for seepage prediction as they encompass
a wider spatial range and wider range of conditions, and can therefore be used
with greater confidence over a broader area, although at the expense of some
degree of accuracy.

identify and measure seepage is recommended: (Note that this methodology assumes
some prior knowledge as to which channels or sections of channel require
investigation. This knowledge may have been acquired based on flow records, visual
inspection or a regional investigation using a technique like remote sensing).
1. Define project objective – why is the work being undertaken? The types of n The first step in an
questions to be asked include: investigation is to
q Is the primary objective to identify relatively high seepage points or to define the project
measure the volume of seepage? objective
q Is it necessary to establish the rate of seepage? – either the actual or relative

rate.
q What degree of confidence in the results is required?
q Over what length of channel is the information required?
q What is the available budget?
At the end of this process there will be a clear definition of the reasons for undertaking
the seepage investigation (eg asset management), budget considerations, scale of the
operation (eg whole channel, specific channel lengths etc), need for accuracy, or
relativity. This process will effect all future decision making.
2. Collate Site Data It is assumed that if decisions for action have been made, there n Collation and
is already some knowledge of site conditions. In the event that there are no details evaluation of site data
key data should be collected. It is important that information on depth to will assist in
groundwater, background groundwater salinity, soil type and channel hydraulics development of a
seepage mechanism
are known or gathered, both at the site where the testing is conducted, and over the
conceptual model and
area the results are to be extrapolated. in technique selection
3. Evaluate Site Data - It is possible that the process of evaluating the data will
have already been performed, formally or intuitively, to identify the need for
action at the site. Evaluation does not have to be at a detailed level, but should be
sufficient to be able to propose a conceptual model of the seepage mechanism, to
detect where changes in these parameters may impact on the geophysical
response, and to assist in technique selection. Channel hydraulic information is
required to help determine potential channel seepage mechanisms.

one that has a depth focus on and immediately below the watertable. Whether this
is achieved using EM or resistivity is not highly important. However, generally it

is easier to focus on a given depth with resistivity (EM provides an average across
a range) and this can be achieved independent of knowledge of groundwater
depth. The advantages and disadvantages of each of the techniques (refer above)
need to be assessed in light of the specific project objectives. The recommended
technique for a given depth to watertable is outlined below:
Preferred Geophysical Techniques n The preferred

q The preferred technique for geophysical channel seepage assessment is directly geophysical technique
is one that is focussed
detecting the impact of seepage on the groundwater. This means that the
on and immediately
instrument must focus on the zone immediately above and several metres below below the watertable
the watertable:
q For a shallow watertable (surface to approximately 5m) EM31 is suitable for n The recommended
direct seepage detection. technique for sites
with a shallow
q For watertables deeper than 5m, EM34 (in vertical dipole mode, with the watertable is EM31
coil spacing dependent on the watertable depth) or resistivity can be used.
However, particularly for deeper watertables, it is easier to focus on a given n For sites with a water
depth with resistivity and this can be achieved independent of knowledge of table deeper than 5m,
groundwater depth. The significant advantage of resistivity is that it EM34 or resistivity
provides a profile of the resistivity beneath the channel. The disadvantage is are recommended
that resistivity technology for channel seepage assessment is relatively new
and therefore more expensive.
q EM31 (vertical dipole) adjacent the channel can be used effectively in areas with n EM31 can be used for
deeper watertables, although it does not directly measure the seepage impact on sites with deeper
the watertable. This is due to fact that the upper soil layers are the most influential watertables but relies
on channel seepage and the relatively shallow depth focus of EM31 measures on inferring rather
than directly
these upper soil layer properties. The method infers zones of likely channel
measuring seepage
seepage by identifying materials in the unsaturated zone most susceptible to
seepage. A decision to use EM31 in an area with a deep watertable might be
based on:
q Cost and required accuracy – If a potentially slightly lower level of accuracy

is considered acceptable then EM31 represents a cheaper alternative than
EM34 or resistivity; or,
q Lack of alternatives – EM34 or resistivity contractors are not readily
available.
It this method is used however, it must be made certain that seepage is controlled by
the unsaturated zone and not surface clogging processes. Otherwise errors will
potentially be introduced to the assessment process.
The preferred geophysical techniques for seepage detection are summarised in Table
7-1.

n Table 7-1 Recommended Geophysical Technique for Seepage Detection
and Measurement

Depth Focus (m)4

(vertical dipole)6
OR
5 – 12
Resistivity7, 9 Direct watertable impact NA10 NA11
OR

(vertical dipole)6
OR
12 – 25
NA10 NA11
Resistivity7, 9 Direct watertable impact
OR

> 25
OR
1. It is recommended EM techniques are conducted adjacent the channel (additional survey runs can be
conducted away from the channel). Resistivity surveys should be conducted on-channel.
2. Direct detection of seepage impacts on the watertable is the recommended technique, but inferred
‘detection’ based on soil property variations will often provide an adequate simulation and may be more
convenient for various reasons (refer to body of report for potential errors associated with this method).
Note that direct detection relies on a salinity contrast between the channel water and the groundwater. It is
recommended the groundwater should be at least 3 to 4 times more saline than the channel water, a
condition that will be met in the vast majority of Australian conditions.
3. Approximate detection of penetration is referred to in the Geonics manual (McNeil, 1980) as the effective
depth of exploration. This is the depth to which approximately 75% of the response is attributed.
4. The ‘depth focus’ is a term used in this report to describe the depth (range) which is most influential in
terms of the relative contribution to the overall EM response (McNeil,1980).
5. These can be conducted immediately adjacent to the channel or on-channel. Both are recommended if
budget allows. If on-channel is used for a watertable of 0-1.5m, the survey should preferentially collect
data in vertical dipole mode where the effects of channel water will be less influential. For sites with a
watertable 0-1.5m, EM31 on channel may be preferred if significant land salinisation exists adjacent the
channel.
6. Horizontal and Vertical Dipole: Note that as applied to EM34, vertical dipole does not refer to the coil
orientation with respect to the ground, and is in fact opposite to the coil orientation. In vertical dipole
mode the coils should be horizontal to the ground, which is a slower method than horizontal mode where
they are held perpendicular to the ground.
7. Resistivity is the preferred direct measurement technique for this depth to watertable but EM34 is provided
as a potentially more accessible alternative.
10. The penetration depth of resistivity depends of the particular system set up (dipole spacing and length).
11. Resistivity surveys measures resistivity at a range of depths intervals within the profile (ie, there is no
fixed depth focus).

5a. Conduct geophysical survey – Undertake the geophysical survey over the section n Conduct geophysical
of interest, giving due consideration to factors such as appropriate timing of the survey survey
and other important variables (refer Section 7.7.1.3).
5b. Evaluate results – Plot geophysical survey results along the section and overlay n Evaluate results and
with known site conditions (soils, geology, hydrogeology and channel hydraulic data). identify areas of
Based on these plots identify areas of suspected high, low and moderate seepage, suspected high, low
assuming low conductivity / high resistivity equates to higher seepage. and moderate seepage
5c. Conduct test drilling – Soil bores should be drilled at appropriate intervals along n Conduct test drilling
the length of the geophysical survey. The primary aim of the drilling is assist with to assist with geo-
interpretation of the geophysical survey. Some key principles of the drilling program physical interpretation
are described below:
q Based on the geophysical survey results, conduct drilling across a range of
low, moderate and high conductivity / resistivity sites;
q Drill at least some bores into the watertable, and construct some as
permanent observation bores;
q Generally drilling should be conducted on the outside toe of the channel;
q Logging and sampling of the bores should ideally be undertaken by someone
trained in soil / geological classification and a consistent classification
system should be followed.
q Depending on the density of data collected, presenting the results in a
geological long section should be considered.
5d. Conduct pondage tests - At appropriate intervals over the entire test area pondage n Conduct pondage tests
tests are to be conducted. The number of tests will depend on the length of channel across a range of
surveyed and the variability of conditions along the channel. The following guidelines geophysical survey
are suggested: response
q Pondage tests should be conducted across a range of low, moderate and high
conductivity / resistivity sites so as to establish a regression equation which
represents the range of geophysical response across the area.
q Similarly, based on the soil drilling results, the pondage tests should be
based on a range of different soil types and / or groundwater conditions.
q Pondage tests must be conducted over areas of like conductivity / resistivity.
That is they should not staddle areas of (significantly) different geophysical
response, as this will complicate interpretation of the results and
development of the regression equation.
q Due to the cost of conducting pondage tests, it is recommended that at least
two cells back to back should be conducted at each site for efficiency
purposes. Using available structures should also be considered to minimise
bank construction costs.
q The pond length can be variable, but as a guide they should generally not be
more than 400-500m and not less than 50m.
By conducting pondage tests in this manner across the area of the geophysical
survey, prediction of seepage rates outside of pondage test areas will be based on
interpolation rather than extrapolation, which improves confidence in the

predicted seepage. While pondage tests are expensive, they are a critical part of
the interpretation process.
5e. Develop and evaluate relationship between seepage and geophysical response –
The following key steps should be conducted:
q Plot geophysical response against pondage test seepage. n Develop and evaluate
relationship between
q Outliers in particular should be assessed in light of all available information, seepage and
including the conceptual seepage mechanism, test drilling results, channel geophysical response,
hydraulics etc. If there are legitimate grounds for excluding outliers they including regression
should be removed. analysis and
statistical significance
q If from this data two or more different trends can be observed due to of the relationship
identifiable differences in sub-surface conditions, then two different
regression equations should be generated.
q Fit a regression line through the data.
q Statistical analysis should be conducted to determine the degree of
confidence that can be placed in the derived relationship
q Using the derived relationship the channel length should be divided into
seepage categories of various seepage rates based on geophysical response,
with accompanying error estimates.
6. Evaluation - Evaluate whether investigation objectives have been met. One of n Evaluate whether
the key questions to address is whether there is sufficient confidence in the original investigation
derived relationship. In addition to the particular statistics of the regression line, objectives have been
this will largely depend on the project objectives. Further pondage tests or other met
testing may be required to further improve confidence in the relationship.

Measurement
2000a), a paper was published relating to international developments in channel
seepage measurement using geophysics. The paper is briefly summarised below. The
important point relating to this work is that it is focussed in the same direction as the
geophysical investigations in these trials: developing geophysical techniques that can
be compared to some form of direct seepage measurement, derivation of a relationship
between the two and then extrapolation to new areas.
Determining Irrigation Canal Seepage with Electrical Resistivity (Hotchkiss et al,

2001)
Summary of Abstract: Procedures were developed and tested for quantifying seepage
losses in unlined irrigation channel reaches in the order of 30m in length, in the
Central Nebraska Public Power and Irrigation District canal system. The procedure
uses electrical resistivity (ER) measurements while canals are in service to determine
the resistivity of the underlying clay layer. ER data were correlated to canal depth and
then to seepage rate. Seepage rates were determined using seepage meters. Accuracy
is approximately ± 20%, comparable to that achieved using stream gauges methods (±
5% error of total canal discharge). The ER approach, however can easily pinpoint
seepage zones more precisely, allowing a reduction in the length of canal lining
projects.

Comparisons to this investigation:
q The method used in the Hotchkiss et al (2001) study relies on measuring n The method used in
unsaturated zone soil properties and not seepage directly, ie it is a method which the US study relied on
infers the likelihood of seepage. It is based on the theory that ‘as resistivity inferred rather than
increases, seepage rates will also increase because more sandy or less clayey direct seepage
detection
materials will have higher resistivities’ (Hotchkiss et al, 2001). This was also
shown to be successful in the ANCID investigations project (eg EM31 at deep
watertable sites). This is a valid method but potentially has limitations at some
sites (refer section 7.7.1.1).
q Seepage meters rather than pondage tests were used as the direct seepage
measurement technique (possibly limited by operational requirements, ie could
not shut channel down). Confirming the general unreliability of seepage meter
readings, 50% of sites where readings were conducted were discarded as
unreliable. This highlights the advantage of pondage tests over seepage meters.
The only advantage of the seepage meter is that smaller length areas can be
selected; 30m used in Hotchkiss et al (2001) study. (It is not recommended
pondage tests be conducted over less than 50m of channel).
q Best correlations were found at 4m depth – similar to this project. Soil properties
from the upper part of the profile were found to be most influential on seepage
rates.
q Due to considerable depth and variation in channel water, the data first had to be
correlated to canal depth and then to seepage rate, whereas in this ANCID study,
the depth of channel water was reasonably consistent and this step was not
required.

Good to moderate relationships were obtained between average EM34 conductivity n Good to moderate
and the corresponding pondage test seepage at most sites. For EM34 at a 10m coil relationships were
spacing in horizontal mode, the effective depth of penetration is around 6-7m, with a obtained between
shallow depth focus at around 1-3m. This meant that at sites where the watertable was EM34 and pondage
test seepage at most
sites

q Donald Main – A moderate relationship was recorded in both surveys. Further n The technique
points are required in the mid-seepage range to appropriately test the relationship. distinguished between
The technique distinguished between high and low seepage but not within the high and low seepage
high seepage results range. Possible interference by adjacent trees may have but not within the high
seepage range
affected results in some ponds. A generally consistent increase in conductivity
q Toolondo Central – A moderate relationship was observed but largely skewed
by the result in one pond. The relationship distinguished between high and low
seepage rates.

In summary, the only site where no relationship was observed was at Dahwilly East, n The only site where no
which was largely due to the narrow seepage rate range. At the Toolondo Central site, relationship was
where conductivity measurement was entirely above the watertable, the unsaturated observed between
zone lithology was a sufficiently accurate indicator of seepage and hence a reasonable seepage and EM34
was due to the narrow
seepage rate range
The Donald site survey was focussed on the saturated zone. However the EM31
survey at the site demonstrated a slightly better relationship with pondage test
seepage compared to the EM34 (R2=0.73 compared to R2=0.50), but neither survey

This meant that at sites where the watertable was deeper than 5m, only a limited
proportion of the response was caused by seepage impacts in the saturated zone.
Therefore at these sites the seepage detection mechanism is largely via inference based
on soil properties in the unsaturated zone. Key summary comments for each of the
sites are listed below:
Toolondo
q Good relationships between EM31 conductivity and pondage tests seepage were

q A high degree of repeatability between the surveys was observed.
q In-channel (shortly after channel shut down) and on-channel EM31 surveys n Good relationships were
returned poor results. This is attributed to the fact that an EM31 survey above the obtained between EM31
watertable ‘works’ by inferring seepage based on soil properties. However and pondage tests in all
immediately beneath the channel, even for low seepage rate ponds the profile three surveys at
Toolondo Central and a
high degree of
uniform saturation produces a uniform conductivity response, and tends to mask repeatability between
changes in lithology resulting in little differentiation between low and high surveys were observed.
q Better results were obtained with the EM31 compared to the EM34(10m) at this
from the channel).
q Three Toolondo Sites (Central, East and West) - The relationship established for n The relationships
all sites was moderately strong. Local correlations at Toolondo Central and established across the
Toolondo West were stronger than the combination of sites. The Toolondo East three Toolondo sites
site displayed an opposite correlation, but the very narrow range of seepage rates were moderately strong
but local correlations at
and the flat regression line indicates this is not a meaningful trend. Confidence each of the sties were
bands for the overall regression relationship are wide but indicate that the stronger
q If the Toolondo Central site had been used to predict seepage at Toolondo West,
would have been essentially accurate (0 mm/d), except in one pond where
seepage was predicted at 4 mm/d when actual seepage is practically zero.
Rocklands
q A good relationship was observed between EM31 response and pondage test n A good relationship
seepage at the Rocklands channel trial site (for the adjacent channel EM31 data). was obtained between
This indicates that seepage was able to be successfully inferred based on EM31 adjacent
unsaturated zone soil properties. However, with a depth to watertable of around channel data and
pondage test seepage
q A poor response was observed when all survey runs were used, largely due to the
q The on-channel results recorded mixed results. In vertical dipole mode no trend n The on channel results
was observed. The configuration is focussed on the flushed zone beneath the were mixed: no trend in
channel where uniform saturation from seepage appears to be masking lithology vertical dipole mode
response. In horizontal dipole a reasonable correlation was observed, apparently and a reasonable
correlation in horizontal
dipole mode

seepage areas.
Donald Main
q A good relationship was observed between EM31 conductivity and pondage test n The technique
seepage but there is a poor spread of seepage data at the site (1 point of low and 5 distinguished between
of high seepage). With a relatively shallow watertable (2m), the EM31 detects high and low seepage
seepage at this site in terms of its impacts on the watertable. The EM31 survey but not within the high
did not distinguish between higher seepage ponds (35 - 48mm/d). Confidence seepage range. There
was a poor spread of
bands are fairly wide for the regression line, particularly at the high conductivity seepage data at the
range, but indicate that the relationship can differentiate between high and low site
q A better relationship was established with EM31 (R2=0.71) adjacent the channel
q Moderate to good relationships were also observed for the on-channel surveys in
Dahwilly
q For a survey conducted when the channel was not running, no relationship was n No relationship was
observed between EM31 conductivity and pondage test seepage. The technique observed for the
failed because the channel was not running and previously seeped water was survey as the channel
was not running
therefore likely to have thoroughly mixed with native groundwater. Unsaturated during the survey. A
zone lithology is a good indicator of seepage at some sites. However at Dahwilly repeat survey when
it is not the unsaturated zone controlling seepage rates, but the clogging layer at the channel was
the channel surface and therefore seepage must be detected directly (ie in terms of operating obtained a
impact on watertable) which means the channel must be in operation. good relationship
q In a repeat survey conducted when the channel was operating, a good relationship
q Two Dahwilly Sites (Central and East) - The relationship established for both n The slightly deeper
sites is moderately strong. Local correlation at the Central site is slightly stronger watertable at the East
than the two sites combined. The East site displays a very weak correlation, but site contributed to the
this is due to the very narrow seepage range and few data points at this site. poor results and
different response at
Confidence bands are relatively wide, suggesting the regression relationship for
this site
both sites can only be used to broadly indicate the likelihood of low or moderate
seepage. The slightly deeper watertable at the East site appears to have put the
watertable largely beyond the range of EM31 and hence very different results are
obtained at the East site. Using the Central site regression relationship to predict
seepage at the East site would have resulted in over prediction of 1.5 - 2 times
actual seepage.
q Better correlations at both sites were obtained using the resistivity compared to n Resistivity results
EM31 due to better targeting of the top of the watertable. were better than
EM31 due to better
Lake View targeting of the water
q A poor relationship between pondage test seepage (July 2001) and EM31 table

adjacent channel data a significantly improved relationship (to moderate) was n A poor relationship
observed as seepage impacts are less diluted. Interpretation is limited at this site was obtained using all
due to the very narrow seepage rate range. Seepage is detected at this site in data but a much
terms of its impact on watertable salinity. improved relationship
was observed using
q No sensible trend was observed at the Lake View Central site using the same adjacent channel data
q Both Sites (Central and West) - The relationship established for both sites is n A moderately strong
moderately strong with a high correlation coefficient but the two data sets relationship was
creating the regression line have small conductivity and seepage rate ranges. It is obtained for both
desirable to obtain data in the mid range to improve confidence in the sites, but the two data
sets have narrow
conductivity and
West site. However using Central data from adjacent the channel is likely to seepage rate ranges
Tabbita
q No relationship was observed between EM31 conductivity and pondage test n No relationship
seepage. Possible reasons for the failure of the technique at this site include: between EM31 and
pondage test seepage
i) Narrow range of seepage rates (little differentiation in rates along section of was obtained at
interest); Tabbita
ii) Seepage mechanism may be such that majority of seeped water does not
iii) EM31 vertical dipole orientation may penetrate too deeply into the native
iv) The method of averaging conductivity may not be appropriate at this site (or
the ponds may need to be placed more carefully, ie over shorter sections of
high/low conductivity)
Finley
q While a moderate correlation coefficient was obtained for the pondage test – n Additional data points
EM31 conductivity relationship at this site (and the highest seeping pond did are required at Finley
record the lowest conductivity), the statistics are not meaningful due to the fact to enable meaningful
that only three data points make up the relationship. The width of the prediction interpretation
q A moderate to poor relationship was recorded between EM31 conductivity and n A moderate to poor
pondage test seepage. However the results should be considered in light of the relationship was
fact that they represent seepage sites more than 20 km apart (significantly further between EM31 and
than other sites). The watertable is generally beyond the penetration depth of the pondage test seepage
was recorded at
EM31 along the survey reach and therefore the likelihood of seepage is inferred
Waranga. The
based on soil properties beneath the channel. relationship can be
q Some of the scatter in the results can be explained by incorrect pond placement used however to
(ie not straddling areas of like conductivity) as well as some geological distinguish between
anomalies. When points of high variance and leverage are removed, the low and high seepage
correlation coefficient improves to 0.62 (from 0.40). The prediction interval

bands suggest the relationship can be used to distinguish between sites of low and
high seepage, but is limited in interpreting mid-range seepage.
q An improvement would be expected if the top of the watertable was targeted, n An improvement in the
rather than inferring seepage from unsaturated zone soil properties. Given that seepage geophysical
the ponds are significantly spaced apart, this relationship can be used for relationship is likely if
interpolation, bearing in mind the associated broad prediction intervals associated the watertable was
targeted
with the regression line.
In summary, the only site where no relationship was observed was at Tabbita. A n At only one site was
number of possible causes for this were identified, but the predominant contributing no relationship
factor is not known. At two sites (Rocklands and Lake View Central), the adjacent observed between
channel data was used instead of all survey run data. This was required to obtain the seepage and EM31
At the Toolondo Central site, where conductivity measurement was entirely above the
At the Rocklands and Dahwilly sites, where the penetration depth of the EM31 (in n Seepage at the
vertical dipole) was just sufficient to reach the watertable, the combination of Dahwilly site is
measuring lithology changes in the unsaturated zone and seepage impacts in the controlled by a
saturated zone combined to provide a reasonable indicator of seepage. However it is surface clogging layer
– therefore techniques
which infer seepage
was observed. This suggests seepage impacts in the watertable are the primary based on unsaturated
detection mechanism at this site, a fact reinforced by the uniform nature of the zone soil properties
unsaturated zone lithology at the site. Seepage at Dahwilly is not controlled by the will not work

On-channel surveys did not work at sites where the watertable was beyond the range n Evidence suggests on-
of the EM31 (Toolondo), did work at sites with a shallow watertable (Donald) and channel EM31 surveys
were partially successful when the watertable was located at the edge of the depth should only be
penetration capacity of the EM31 (Rocklands). Further work is required in this area, conducted where the
watertable can be
penetrated

most sites. Key summary comments for each of the sites are listed below:
Toolondo
q Central – At around 10-12m the best correlation was obtained (R2=0.6), which is n Best correlations were
the zone immediately below the watertable and fits with the expected mechanism obtained at 10-12m,
of seepage detection (ie, in the depth interval of groundwater most effected by the zone immediately
seepage). Within individual ponds, the resistivity cross sections show sub- below the watertable
sections of localised higher seepage. There is very little correlation between
seepage rates and resistivity in the unsaturated zone.
q East - The very narrow range of seepage rates and resistivity values meant that no
meaningful correlations were observed at the Toolondo East site.
q West - At and below the watertable the expected inverse trend between high n Expected correlations
resistivity - low seepage and low resistivity - high seepage was not observed. It is were not observed at
apparent that the sandstone at this site may be dominating the response. Low Toolondo West,
permeability sandstone may be causing a high resistivity response (normally possibly due to effects
of low permeability
associated with high seepage), and masking the effect of seepage on the saline
sandstone
groundwater. Deeper drilling would be required to confirm this interpretation.
The reasonable correlations obtained at shallow depth (2m and 4m depth slices)
are most likely due to changes in clay content beneath the channel, and
corresponds with observed EM31 relationships.
q All Sites – The Toolondo East data (averaging 10/12/14m depth slices) lies
within the expected extrapolated range for the Central (10m) site. The West
(averaging 10/12m depth slices) data does not fit within the Central and East
relationship (possibly due to changes in lithology masking seepage impacts). The
West 2m depth slice does fit within the relationship. However without pondage
tests it would not have been known that this was the better depth on which to
focus.
Dahwilly
q Central - Good correlations were observed at the 6m, 8m and 10m depth slices, n Good correlations
which fits with the expected mode of seepage detection below the watertable. observed at and below
Correlation coefficients worsen in the unsaturated zone. Resistivity long sections the watertable.
indicate that seepage is generally diffuse across the surveyed area, in contrast to Seepage was
Toolondo where localised seepage is evident from the resistivity data. generally diffuse.
q East – A strong correlation was observed at 6m which corresponds with the top of
the watertable. However, the very narrow range of seepage rates and small
number of data points limits the significance of these correlations.
q Both Sites - Two parallel but vertically offset regression lines for the Dahwilly n Different background
sites are recorded and there is no observed trend for the combined regression line. conditions were the
This suggests different background conditions (despite apparent similarities probable cause of lack
between sites), including finer and more clayey sands and possibly higher of trend in combined
regression equation
background salinity groundwater at Dahwilly East.
Finley
q The best correlation was observed at 4m and 6m which corresponds with the zone
below the top of the watertable (1.5m). A misleading opposite (inverse)
correlation was observed at 2m depth, when a stronger, direct correlation was

expected based on depth to watertable. This may be a reflection of the poor
surface resolution of the resistivity equipment. The narrow seepage range and
small number of data points also limits the significance of the correlations.
Lake View
q Central - The best correlation was observed at 6m and 8m which corresponds n Best correlations
with the zone several metres below the top of the watertable (1.5m). A very weak were observed at 6-
trend was observed at 2m and an (inverse) correlation at 4m depth. It was 8m. Poor surface
expected based on the depth to watertable that the best correlations would be resolution of
resistivity equipment
observed at around 2-4m depth. This may be a reflection of the poor surface
may have caused the
resolution of the resistivity equipment. Site lithology below the watertable may weak trend at 2-4m
also be significantly contributing to the response. The narrow seepage range also
q West – There was no meaningful relationship observed between resistivity and
seepage immediately below the watertable. This is attributed to:
i) Poor resolution of near surface data (this theory is supported by the
reasonable correlation observed between EM31 and seepage at this site)
and/or
ii) The 'anomalous' result in pond 1.
Something is causing elevated resistivities relative to other ponds at the site
(possibly lithology, faulty data collection etc), although nothing obvious was
detected in the drilling.
q Both Sites - Neither the 2m or 6m depth slice from Lake View Central could have n Seepage at Lake View
been used to accurately predict seepage at Lake View West. However a slightly West was under-
better fitting trend line is obtained for the 2m depth slice. The apparently estimated using the
anomalous result at Pond 1 (Lake View West) is largely skewing these Lake View Central
regression equation
relationships. Using the Lake View Central site to predict seepage at the Lake
View West site would have caused significant under-estimation of actual seepage.
In summary, most sites displayed a good correlation between seepage and the n Most sites displayed a
resistivity at and immediately below the watertable. The two sites that did not were good correlation
Toolondo West and Lake View West. At Toolondo West it appears that the type of between seepage and
resistivity at and
sandstone at this site may be dominating the response. However deeper drilling would
immediately below the
be required to confirm this interpretation. Further investigation into the potential
watertable
variations in resistivity in sandstone is required (eg, potential effects of iron content in
rock or amount of clay in the cementing material). A reasonable trend was obtained at
shallow depth, but without the information supplied by the pondage tests this could
not have been known.
The lack of trend at the Lake View West site is probably due to the poor resolution of
the resistivity equipment at very shallow depth. This site contains the shallowest
watertable across all sites (0.5 – 1m). Improved resolution at shallow depth could
relatively easily be improved in future surveys by using exponentially rather than
linearly spaced arrays (Allen, pers. comm. 31/10/02). At Toolondo East also no
trend was observed but this is solely attributed to the very narrow range of seepage
rates at this site.

7.8 Waranga Western Channel Case Study
It was proposed that the Waranga Western Channel (WWC), an open irrigation n The upgrading of the
channel maintained by Goulburn-Murray Water (G-MW), be upgraded in capacity Waranga Western
along approximately 50 km of the channel length. The channel has a well- Channel, in addition
documented record of existing seepage problems. The extent of seepage in this to existing seepage
concerns, initiated an
section of the channel had been a concern to local landholders for a number of years.
investigation to
In addition, there was concern that new seepage paths may be opened up during an identify (and quantify)
upgrading works program. Therefore G-MW required quantification of sections of high risk channel
existing seepage problems in the WWC with identification and quantification of sections
sections where new seepage paths might be opened up. To this end, geotechnical and
geophysical investigations were carried out along the channel, including the following
investigative works (in chronological order):
q EM31 survey – November 2001: A 46 km EM31 survey was conducted on-
drilled and 34 piezometers installed. Bores were generally drilled to a depth of at
least 6m, and some up to 10m;
The high risk section of the channel, as categorised by the EM31 contractor based on n The initial EM31
EM magnitude, totalled approximately one-third (15km) of the length of the survey identified 15km
investigated portion of the channel. To further refine this area, a combination of the of high seepage risk
EM31 results and impermeability grade (a lithological classification devised for the channel. Combined
with lithological data
investigation based on the amount of clay in the profile) was used to identify sections 7.2km of this section
of channel which were considered to represent ‘very high’ risk areas. Using this was identified as very
system, four significant lengths of channel were identified as very high risk. The total high risk
length of the ‘very high risk’ area was 7.2 km. It was apparent to G-MW that further
investigation was required before committing to the significant expense of lining such
a length of channel.
A further 107 bores were drilled (to greater depth than the original drilling program).
The above process using impermeability grade (incorporating results from the new
bores) and EM31 response was again conducted. Following the review of the
additional drilling, the areas classified as very high seepage risk actually increased by
about 1 km (from 7.3 km to 8.3 km). This included some areas being removed and
some being added to the very high risk category.
It was then recognised that, in addition to the drilling program, pondage tests were n Pondage tests were
required to quantify seepage rates (and potentially identify a relationship between conducted to quantify
EM31 and pondage test seepage rates) and confirm interpretation of seepage rates seepage rates and
based on geology and EM31 data. Therefore 12 pondage tests were conducted at confirm geophysical
interpretation
various intervals along the channel, covering a range of environments and areas of
different geophysical response.

pondage test seepage. However when considered in light of the fact that they

represent seepage sites more than 20 km apart (significantly further than other sites n A moderate to poor
tested in the trials), this was not unexpected. The watertable along the Waranga EM31 seepage
channel is generally beyond the penetration depth of the EM31 and therefore the relationship was
likelihood of seepage is inferred based on soil properties beneath the channel. Some observed, but
represents sites more
of the scatter in the results could be explained by incorrect pond placement (ie not
than 20km apart
straddling areas of like conductivity) as well as some geological anomalies. When
points of high variance and leverage were removed, the correlation coefficient
improved to 0.62 (from 0.40).
The prediction interval bands suggested the relationship could be used to distinguish n The EM31 seepage
between sites of low and high seepage, but is limited in interpreting mid-range relationship could be
seepage. An improvement would be expected if the top of the watertable was used to distinguish
targeted, rather than inferring seepage from unsaturated zone soil properties, as much between sites of high
and low seepage.
of the seepage in the Waranga Western Channel is probably controlled by the silt layer
in the channel and not the unsaturated zone. Given that the ponds were significantly
spaced apart, it was concluded that this relationship can be used for interpolation,
bearing in mind the associated broad prediction intervals associated with the
regression line.
Based on the results of the pondage test areas, the relationship with EM31 and the n Based on the EM31 -
drilling program, the areas recommended for remediation were finalised. Sites were seepage relationship,
defined as either priority one or priority two seepage risk sites, depending on the pondage tests and
degree of perceived seepage risk. Priority one sites were considered to require drilling, sites were
defined in terms of
remediating as part of the upgrade, while priority two sites were to be monitored
seepage priority
closely for seepage following the upgrade. Given the uncertainty in the EM31 –
seepage relationship, the EM31 predicted seepage was not used as the sole means of
assigning seepage risk but geological data and visual observations were also integrated
into the decision making.
The Waranga Western Channel seepage investigation is a good example of the

integration of geophysical, geological and pondage test data to determine areas of
highest seepage risk. In the end the required objectives were achieved. However the
investigation could have been improved by following the process outlined in the
geophysical conclusions section of this report. That is, a geophysical survey, followed
by test drilling to a suitable depth (based on the geophysical survey results) and then
followed up by pondage tests, also based on the results of the geophysical survey and
the test drilling. By clearly establishing this process from the outset, this would have
avoided the need for two rounds of drilling, and probably could have provided GM-W
with a more rapid answer.

8. Conclusions
8.1 Overview
In response to increasing concern regarding channel seepage issues, ANCID
representing Australian RWAs, in conjunction with the MDBC initiated a project to
investigate channel seepage measurement. Trials were conducted in four RWAs from
2000 to mid 2002. They were focussed on the following techniques:
q Pondage tests
q Point measurement (channel full and empty),
q Geophysical techniques,
q Groundwater techniques,
q Soil classification; and,
q Remote sensing.

q Inflow-Outflow Tests: These were deemed not sufficiently accurate for
measuring losses over relatively short sections of channel (ie 1-2km).
input required means this method is not practical for most RWA investigations.
q Hydrochemical Techniques and Tracing of Leakage Plume - The high cost and
expertise required means they are generally not practical solutions for RWAs.
8.2 Pondage Tests

Pondage tests were conducted across all sites (totalling 81 ponds), as they were the
basis on which other techniques were assessed. Seepage rates ranged from 0.1 mm/d
to 48 mm/d. The average and median seepage rate across all sites was 9.7 mm/d and
7.0 mm/d respectively. Some sites anticipated to have high seepage rates actually
contained low rates, while others expected to have low rates were found to have a high
rate of seepage. Visible evidence of seepage was found to not necessarily imply high
seepage rates. At sites where pondage tests were repeated, a good degree of
repeatability was observed; the maximum difference between rates was 25%, with
differences attributed to changes in depth to watertable and channel bed properties.
8.3 Sub-surface Characterisation

Sub-surface characterisation was conducted to assist in general site characterisation as
well as to assist in geophysical interpretation. An attempt to estimate seepage based
on average soil permeability yielded no clear relationship between soil permeability
and seepage rate. The absence of a relationship was attributed to limitations inherent
in the method adopted (in particular the inadequate sampling density and the process
of assigning permeability to soil type), and the fact that in many of the channels
studied, factors apart from soil type are the primary control on seepage, including
bank dominated seepage and the influence of surface clogging layers. The density of
sampling and permeability testing required, in addition to the fact that soil type is not
always the factor controlling seepage, means that sub-surface characterisation is not
likely to be either an accurate or cost effective means of seepage quantification.
However, it remains a critical part of the site characterisation phase of a channel
seepage investigation.

8.4 Point Tests
Five point test trials were conducted during the investigation, using ring infiltrometers,
disc permeameters and Idaho seepage meters. These trials confirmed that point tests
are generally not reliable for directly quantifying seepage. Due to variable and
sometimes erratic values obtained in measurements, a large number of tests is required
to sufficiently determine the true seepage rate of a section of channel. Therefore point
tests are generally not considered reliable for absolute quantitative purposes and
should generally be limited to determining the distribution of seepage losses ( ie,
relative seepage). Even for this purpose a large number of tests are recommended to
minimise the effects of local variability. The Idaho seepage meter appeared to provide
the most reliable results of the three instruments, probably a reflection of the fact that
the channel is full during the test and that truly saturated flow is being measured.

Quantitative analysis of seepage rates was conducted on the Donald Main Channel
based on changes in groundwater level before and after channel filling. Qualitative
assessment only was conducted on the Tabbita site. Groundwater levels at the Donald
Main Channel were used to estimate seepage using the Dupuit Forcheimer equation
and seepage estimates approximately equal to pondage test seepage were obtained,
depending on the input aquifer hydraulic conductivity used. Therefore, use of
groundwater bores for quantitative analysis of seepage is not considered accurate or
cost effective for typical RWA channel seepage investigations, due to the sensitivity
of the solution to hydraulic conductivity inputs and the cost of obtaining sufficiently
reliable estimates. In addition, bores are essentially a type of point test and as such do
not address the question of where the channel is seeping. A high density of bore
transects would be required for meaningful identification of local areas of seepage.
However, groundwater observation bores are a very valuable part of the site
characterisation phase of a channel seepage investigation. Further, groundwater bores
are a very useful post-remediation assessment tool, particularly for assessing the
8.6 Remote Sensing

A remote sensing investigation was planned as part of the trials but was eventually not
undertaken due to budget constraints. Based on the literature review and preparation
of the brief for the proposed trials, it is concluded that remote sensing techniques:
q Are best suited to investigations where the primary aim is identification of land
degradation associated with channel seepage. It should not be used where the
seepage mechanism is predominantly vertical;
q Will be most useful where lateral seepage is predominant. For example, sites
with a high watertable, shallow impermeable layer or bank seepage are likely to
facilitate lateral seepage and cause seepage to have a surface expression; and,
q Should primarily be regarded as a seepage identification tool and not for seepage
quantification purposes.

8.7 Geophysics
8.7.1 General Conclusions
Geophysical techniques identify or measure channel seepage by detecting contrasts in
terrain conductivity below the channel in one of two ways:
1) Directly measuring seepage induced changes in groundwater conductivity; or,
2) Identifying contrasts in soil properties above the watertable and inferring the
likelihood of seepage.
of seepage is assumed to be related to unsaturated zone soil properties. In many cases
this is a reasonable assumption, supported by the fact that the inferred method of
detection was successful at most, but not all sites investigated in the trials. The
unsaturated zone is not necessarily the controlling influence on seepage, and
particularly in Australian conditions seepage is often controlled by a clogging (silt)
layer. Therefore, there is less risk in using the direct method of seepage detection.
The direct method of detection cannot be used in relatively non-saline groundwater
environments, as the fresh seepage water will not contrast with the native
groundwater. As a guide it is recommended that groundwater salinity is at least three
to four times higher than the channel water salinity.
It is very important that the depth to watertable is known at the site before selecting a
geophysical technique. Based on this information a decision can be made as to

q Accuracy
q Cost and Speed
q Data Processing
Accuracy
The accuracy of a given geophysical technique will depend on whether inferred or
direct seepage detection is used. Generally direct measurement should be considered
more reliable than inferred measurement. For direct measurement the accuracy will
depend on how well the watertable is targeted. Therefore in theory on-channel
resistivity surveying should be the most accurate geophysical technique, as it is based
on direct seepage detection and can target the watertable independent of depth. At
most sites in the trials resistivity surveying results were comparable to EM31 and
EM34, and at three sites correlations with pondage tests were better than the EM
correlations. The other significant advantage of resistivity surveying is that the final

output is a two dimensional profile of resistivity beneath the channel. This allows
easier interpretation of the results and provides an indication of seepage mechanisms.
The fundamental limitation with all EM surveys and other such fixed array type
geophysical surveys is that the result is averaged over a specific depth interval, which
may not be the critical interval of interest. Therefore (for direct detection) the
accuracy depends on how well the watertable is targeted by the particular EM
equipment, which in turn depends on the watertable depth. If the correct EM
equipment is selected to suit the watertable depth, in theory it should be close to the
accuracy of resistivity surveying.
The robustness of EM31, as demonstrated by the consistent results in the trials is due
to its relatively shallow depth focus (1-4m). For channels where there is a shallow
watertable (eg, surface to 3-4m), EM31 can be used for direct measurement of
is deep, EM31 infers seepage from near surface soil properties, which is suitably
accurate in most instances.
Cost and Speed

EM31 surveys are the cheapest geophysical method, due to the speed of data
acquisition; EM34 is more expensive as two people are required for operation and the
equipment must be carried by hand. Resistivity surveying costs are difficult to
quantify given that the on-channel application of the technique is relatively new.
Costs are likely to come down as the technique is refined, and the equipment becomes
commercially available.
Australia. At present on-channel resistivity surveying is still in a development phase
and as such there are no commercially operating contractors who specialise in this
type of survey, but a number of geophysical exploration / surveying companies have
the capability to develop this type of equipment.
Data Processing
comparison, data processing requirements for resistivity surveying are much higher,
due to the cost of inverting the data to produce a resistivity cross section.

q Survey timing – If direct measurement of seepage is used, the survey must be
conducted while the channel is running (preferably for at least several weeks),
however if seepage is being inferred from soil properties then the timing of the
survey is not critical and can be conducted whether the channel is running or
empty.
q On-channel versus on-land – Further work is required in this area, but overall in
the trials the most consistent results were returned on-land and this is considered
the safest option. Evidence collected in this investigation suggests on-channel
EM31 surveys should only be conducted where the geophysical technique can
penetrate into the watertable, and ideally target the top of the watertable. In other
words, the method of inferred seepage based on unsaturated zone soil properties

does not appear to work on-channel. For EM31 systems this would preclude their
use on-channel when the watertable is deeper than approximately 3-4m.
If budget allows, it is recommended that both on-land and on-channel surveys be

conducted. Resistivity surveys can (and should) be conducted on-channel
because of their greater depth penetration capacity.
these surveys indicates the best off-set distance for on-land surveys is
immediately adjacent the outside toe of the channel. At sites without a steep
gradient or high transmissivity, an average of survey traverses up to 50m on each
side of the channel was found to improve the correlation between seepage and the
geophysical survey at most sites. Traverses on either side of the channel are
recommended, but if the budget is a significant constraint, a traverse on the
down-slope side of the channel should be the priority.
two sites where there was a significant difference in the results, changes in
groundwater conditions due to channel operation accounted for the difference.

For all of the sites used in the final year of analysis, multiple and simple linear
regression was undertaken to look for potential regional correlations between seepage
rates and geophysical response (for both EM31 and resistivity). The multi-variate
regression analysis indicated that, apart from the geophysical response, depth to
watertable was the next most significant explanatory variable.
Based on distinct trends between sites with shallow and deeper watertables, the sites
were split into two data sets based on depth to watertable, in order to improve the
accuracy of the fitted regression model. For sites with a deep watertable (5-10m
below surface) the permeability of the top 2m of the profile was shown to be an
explanatory variable of secondary importance.
Statistically the regional fitted regression models were generally moderate to good,
with correlation coefficients of around 0.5 – 0.6 and standard error of estimates of
around 50%. In some cases a higher correlation coefficient and relatively low
standard estimate of error was obtained, however this was for data sets with fewer data
points – greater number of points are required to improve confidence in these models.
Confidence intervals (80% and 90%) for the regression lines were generally fairly
broad, indicating that these regional equations can only be used to broadly classify
seepage rates (eg, into low, medium and high categories). Consequently it is
recommended that there is currently insufficient confidence in these regression
equations for their use to predict seepage at new sites without local calibration against
pondage tests.
In most instances the multi-variate analysis did not significantly improve the
regression model. The addition of the soil permeability parameter (for sites with a
deep watertable), while statistically significant, generally only resulted in marginal
improvements to the model. The cost of conducting field tests to collect this data
therefore probably outweighs the benefits.

For the resistivity analysis, the ten metre depth slice was adopted as the variable for
use in the model. While a more accurate analysis could be conducted using the depth
at and just below the watertable, for the purpose of a consistent approach, this depth
slice was selected. (There also appeared to be some inaccuracies in the near surface
resistivity data). It is likely the analysis could be significantly improved by using
resistivity data at and immediately below the watertable for each of the sites. In
addition, the fewer data points and smaller range of environments used in the
resistivity analysis contributed to the poorer fit of the model. This adds greater
uncertainty to use of the model, compared to the EM31 regression equations.

geophysical response need to be assessed:
1. What confidence is there that the derived relationship accurately describes
seepage within the area tested? - Confidence in the derived seepage-geophysical
relationship within the area tested can be assessed by a number of statistical
indicators, including: the correlation coefficient, standard error of estimate, and
prediction interval. The number of data points and seepage rate range
represented should also be considered.
2. How confidently can the relationship be used outside of the area tested in order
to predict seepage? - When extrapolating a geophysical-seepage relationship
outside of an area from which it was developed, firstly the strength of the
original relationship needs to be assessed (refer above). Secondly, the
representativeness of the new area in comparison to the conditions where the
relationship was derived should be evaluated.

identify and measure seepage is recommended:
1. Define project objective – The key issue that needs to be addressed is

identification of the primary reason the work is being undertaken.
2. Collate Site Data – Basic site information including depth to groundwater,
groundwater salinity, soil type and channel hydraulics should be collated at the
testing site and over the area the results are to be extrapolated.
3. Evaluate Site Data - This should be at a level to enable development of a first cut
conceptual model of the seepage mechanism, to detect where parameter changes
may impact on geophysical response, and to assist in technique selection.
one that directly detects the impact of seepage on the groundwater. To do this it
must have a depth focus on and immediately below the watertable. The
recommended technique for a given depth to watertable is outlined below:
Direct Detection
q Shallow watertable (surface to approximately 5m): EM31 is recommended.
q Watertable deeper than 5m: EM34 (in vertical dipole mode, with the coil
spacing dependent on the depth to watertable) or on-channel resistivity can

be used. However, particularly for deeper watertables, it is easier to focus
on a given depth using resistivity.
Note that direct detection requires native groundwater salinity to be at least three to
four times more saline than channel water salinity.
Inferred ‘Detection’
q EM31 (vertical dipole) adjacent the channel can be used effectively in areas
with deeper watertables to infer seepage based on upper soil layer properties.
A decision to use EM31 in an area with a deep watertable might be made due to
budget constraints, where a potentially slightly lower level of accuracy is considered
acceptable, or due to a lack of alternatives (eg, EM34 or resistivity contractors not
readily available). It this method is used however, it must be made certain that
seepage is controlled by the unsaturated zone and not surface clogging processes.
5a. Conduct geophysical survey – Undertake geophysical survey in section of interest.

5b. Evaluate results – Plot survey results and overlay with known site conditions
(soils, hydrogeology, etc). Identify areas of suspected high, low and moderate seepage.
5c. Conduct test drilling – Soil bores should be drilled at appropriate intervals along
the section to assist with interpretation of the geophysical survey. Bore locations
should be based on the geophysical survey results, and should cover a range of low,
moderate and high conductivity / resistivity;
5d.. Conduct pondage tests – The number of pondage tests will depend on the length
of channel surveyed and the variability of conditions along the channel. Pondage tests
should be conducted across a range of low, moderate and high conductivity /
resistivity sites so as to establish a regression equation which represents the range of
geophysical response and should also cover the range of soil types. Individual cells
must be conducted over areas of like conductivity / resistivity.
5e. Develop and evaluate the relationship between seepage and geophysical response
– This involves plotting average geophysical response against pondage test seepage,
removal of outliers as appropriate, fitting of a regression line, statistical analysis to
determine the degree of confidence that can be placed in the derived relationship and
use of the derived relationship to predict seepage in new areas.
6. Evaluation – Evaluate whether investigation objectives have been met.

Measurement
2000a), a paper from the US (Hotchkiss et al, 2001) was published describing a
similar use of geophysics to measure channel seepage as that adopted in these trials.
The method relied on inferred rather than direct measurement of seepage. The
important point relating to this work is that it is focussed in the same direction as the
geophysical investigations in these trials. That is, developing geophysical techniques
that can be compared to some form of direct seepage measurement, derivation of a
relationship between the two and then extrapolation to new areas.

and the corresponding pondage test seepage at most sites. EM34 with a 10m coil
spacing in horizontal mode was the main EM34 set up used. For this particular
configuration, the effective depth of penetration is around 6-7m, with a shallow depth
focus of around 1-3m. Therefore at sites where the watertable was deeper than 5m,
only a limited proportion of the response is caused by seepage impacts in the saturated
zone. Therefore at these sites the seepage detection mechanism is predominantly via
inference based on unsaturated zone soil properties.
Rocklands – A good relationship was recorded in both surveys. A high degree of

Donald Main – A moderate relationship was recorded in both surveys. The technique
distinguished between high and low seepage but not within the high seepage rate
range. A consistent increase in conductivity was observed between repeat surveys,
caused by a change in watertable elevation.
Toolondo Central – A moderate relationship was observed, largely skewed by one
pond. The relationship distinguished between high and low seepage rates.
Dahwilly Central – A moderate relationship was observed but this was over a narrow
conductivity range, reflecting the site’s relatively uniform unsaturated zone. The
EM34(20m) configuration penetrated too deeply below the watertable and the uniform
response reflected native groundwater conditions. Significantly, the resistivity
surveying improved the relationship, as the zone immediately below the watertable
could be targeted, where seepage impacts are most distinct.
Dahwilly East – No relationship was observed. The seepage rate range was too
which was largely due to the narrow seepage rate range. At the Toolondo Central site,
where conductivity measurement was entirely above the watertable (ie, inferred), the
unsaturated zone lithology was a sufficiently accurate indicator of seepage and hence a
reasonable trend was observed. At the Rocklands and Dahwilly sites, where the
penetration depth was just sufficient to reach the watertable (but the focus was above
the watertable), the combination of measuring lithology changes in the unsaturated
zone and seepage impacts in the saturated zone worked to provide a reasonable
indicator of seepage.

Therefore where the watertable is deeper than 5m, only a limited proportion of the
response is caused by seepage impacts in the saturated zone. At these sites the
seepage detection mechanism is largely via inference based on soil properties in the
unsaturated zone. Summary comments for each site are listed below:
Toolondo - Good relationships between EM31 and pondage test seepage were
recorded in all three surveys at Toolondo Central and a high degree of repeatability

between surveys was observed. Poor results were obtained for on-channel surveys,
supporting the theory that on-channel EM31 surveys are not suitable when the
watertable is beyond the EM31 range (ie, 6m). The relationship established across the
three Toolondo sites was moderately strong but local correlations at each of the sites
were stronger. The Toolondo East site displayed an opposite correlation, but the very
narrow range of seepage rates indicates this is not a meaningful trend.
Rocklands - A good relationship was observed between EM31 response adjacent the
channel and pondage test seepage. A poor response was observed when all survey
runs were used, largely due to the effect of trees adjacent one pond. On-channel
results were varied. No trend was observed in vertical dipole mode and a reasonable
correlation was observed in horizontal dipole mode.
Donald Main - A good relationship was observed between EM31 and pondage test
seepage. The technique distinguished between high and low seepage but not within
the high seepage range. Moderate to good relationships were also observed for the on-
channel surveys in both dipole modes, supporting the theory that EM31 on-channel
surveys are suitable if the depth to watertable is within the EM31 range (ie, 6m).
Dahwilly - For a survey conducted when the channel was not running, no relationship
was observed between EM31 and seepage. In a repeat survey conducted when the
channel was operating, a good relationship was observed. The initial survey failed
because water seeped from the previous season had mixed with the native
groundwater, and no salinity contrast remained. While unsaturated zone lithology is a
good indicator of seepage at some sites, at Dahwilly it is not the unsaturated zone
controlling seepage, but the clogging layer at the channel surface. Therefore seepage
must be detected in terms of its direct impact on the watertable.
The relationship established for the two Dahwilly sites was moderately strong, despite
the fact that the East site on its own displays a very weak correlation, due to a very
narrow seepage rate range. The slightly deeper watertable at the East site appears to
have put the watertable largely beyond the range of the EM31 and hence very different
results are obtained compared to the Central site. Better correlations at both sites were
obtained using the resistivity due to better targeting of the watertable.
Lake View – A poor relationship was obtained at Lake View central using all data,
but a much improved relationship (to moderate) was observed using adjacent channel
data. Interpretation is limited at this site due to the very narrow seepage rate range.
The relationship established for both sites was moderately strong with a high
correlation coefficient but the two data sets creating the regression line have small
conductivity and seepage rate ranges.
Tabbita - No relationship was observed between EM31 and seepage at this site.
Possible reasons for the failure of the technique at this site include the narrow range of
seepage rates, the seepage mechanism may be such that majority of seeped water does
not reach watertable, EM31 vertical dipole orientation may penetrate too deeply into
the native groundwater, and the method of averaging conductivity may not be
appropriate at this site.

Finley - A moderate correlation coefficient was obtained for the pondage test – EM31
relationship but the statistics are not meaningful as only three data points make up the
relationship.
Waranga Western Channel - A moderate to poor relationship was recorded,

however the results represent seepage sites more than 20 km apart. The watertable is
generally beyond the penetration depth of the EM31 along the survey reach and
therefore seepage is inferred based on soil properties. Some of the scatter in the
results can be explained by incorrect pond placement as well as some geological
anomalies. When points of high variance and leverage are removed, the correlation
coefficient improves to 0.62 (from 0.40). The prediction interval bands suggest the
relationship can be used to distinguish between sites of low and high seepage.
In summary of the EM31 results, the only site where no relationship was observed was
at Tabbita. A number of possible causes for this were identified, but the predominant
contributing factor is not known. At two sites the adjacent channel data was used
instead of all survey run data. This was required to obtain the best relationship, due to
the interference effects of trees and rapid mixing of seepage water away from the
channel. At the Toolondo Central site, where conductivity measurement was entirely
above the watertable, the unsaturated zone lithology was a sufficiently accurate
indicator of seepage and hence good trends were observed. The Donald and Lake
View site surveys were focussed on the saturated zone, and seepage was detected as it
created a conductivity low against higher background conductivity groundwater. At
Waranga, a reasonable (to poor) relationship was observed, however improvements
might be expected using a technique targeting the top of the watertable at this site.
significant to note that when the channel was not running, no relationship was
observed. This suggests seepage impacts in the watertable are the primary detection
mechanism at this site. Seepage at Dahwilly is not controlled by the unsaturated zone
but by a clogging layer at the base of the channel. Techniques which purely infer
seepage from unsaturated zone soil properties will not work at such sites.

most sites. Key summary comments for each of the sites are listed below:

Toolondo - At the Toolondo Central site the best correlations between seepage and
resistivity were obtained around 10-12m (R2=0.6), which is the zone immediately
below the watertable. There was very little correlation in the unsaturated zone. At
Toolondo East the very narrow range of seepage rates and resistivity values meant that
no meaningful correlations were observed. At Toolondo West the expected inverse
trend between low resistivity and high seepage at and below the watertable was not
observed. It was apparent that the low permeability sandstone may have been causing
a high resistivity response (normally associated with high seepage), and masking the
effect of seepage on the saline groundwater. The reasonable correlations obtained at
shallow depth are most likely due to changes in clay content beneath the channel, and
correspond with observed EM31 – seepage relationships at the site.
Based on the regression relationship established at the Toolondo Central site, the
Toolondo East data lies within the expected extrapolated range for the Central site.
The Toolondo West data does not fit within the Central and East relationship however,
possibly due to changes in lithology described above.
Dahwilly - Good correlations were observed at Dahwilly Central (for the 6m to 10m
depth slices), which fits with the expected mode of seepage detection below the
watertable. At Dahwilly East a strong correlation was observed at 6m which
corresponds with the top of the watertable. However, this is based on a very narrow
range of seepage rates and small number of data points.
The two regression lines for the Dahwilly sites are parallel but vertically offset and
there is no observed trend for the combined regression line. This suggests different
background conditions, including more clayey sands and possibly higher background
salinity groundwater at Dahwilly East.
Finley - The best correlation was observed at 4 - 6m which corresponds with the zone
below the top of the watertable (1.5m). An inverse correlation observed at 2m depth
is probably a reflection of the poor surface resolution of the resistivity equipment.
Lake View - At Lake View Central the best correlation was observed at 6m and 8m,
which corresponds with the zone several metres below the top of the watertable
(1.5m). A very weak trend was observed at 2m and an inverse correlation at 4m
depth. As previously noted, this may be a reflection of the poor near surface
resolution of the resistivity equipment. Site lithology below the watertable may also
be significantly contributing to the response. At the Lake View West site there was no
meaningful relationship immediately below the watertable. This is attributed to either
poor near surface resolution and/or an 'anomalous' resistivity result in one pond.
Neither the relationship derived from the 2m or 6m depth slice from Lake View
Central could have been used to accurately predict seepage at Lake View West. Using
the Central site to predict seepage at the Lake View West site would have caused
significant under-estimation of actual seepage.
In summary, most sites displayed a good correlation between seepage and the
resistivity at and immediately below the watertable. The two sites that did not were
Toolondo West and Lake View West. At Toolondo West it appears that the type of
sandstone at this site may be dominating the response. However deeper drilling would
be required to confirm this interpretation.

The lack of a correlation at the Lake View West site is probably due to the poor
resolution of the resistivity equipment at very shallow depth. This site contains the
shallowest watertable across all sites (0.5 – 1m). Improved resolution at shallow
depth could relatively easily be improved in future surveys by using exponentially
rather than linearly spaced arrays (Allen, pers. comm. 31/10/02). At Toolondo East
also no trend was observed but this is solely attributed to the very narrow range of
seepage rates at this site.
8.8 Waranga Western Channel Case Study

documented record of existing seepage problems. There was also concern that new
seepage paths may be opened up during the upgrading works program. Therefore G-
MW required quantification of sections with existing seepage problems and
identification and quantification of sections where new seepage paths might be opened
up. To this end, geotechnical and geophysical investigations were carried out along
the channel, including an EM31 survey (November 2001) coupled with drilling of 128
shallow bores, additional geotechnical drilling (March 2002) including the drilling of
an additional 107 bores, and twelve pondage tests (May/June 2002) conducted at
various locations along the channel.
Initially a combination of the EM31 results and a lithological classification devised for
the investigation based on the amount of clay in the profile was used to identify
sections of channel which were considered to represent ‘very high’ risk areas. It was
then recognised that, in addition to the drilling program, pondage tests were required
to quantify seepage rates and confirm interpretation of seepage rates based on geology
and EM31 data. Therefore 12 pondage tests were conducted at various intervals along
the channel, covering a range of environments and geophysical response.

pondage test seepage. The prediction interval bands accompanying the relationship
suggested it could be used to distinguish between sites of low and high seepage, but is
limited in interpreting mid-range seepage. It was concluded that this relationship
could be used for prediction, bearing in mind the associated broad prediction intervals
associated with the regression equation.
Based on the results of the pondage tests, the regression relationship between EM31
and the pondage tests and the drilling program, the areas recommended for
remediation were finalised. Sites were defined as either priority one or priority two
seepage risk sites, depending on the degree of perceived seepage risk. Given the
broad confidence intervals in the EM31 – seepage relationship, the EM31 predicted
seepage was not used as the sole means of assigning seepage risk but geological data
and visual observations were also integrated into the decision making process. The
WWC seepage investigation is a good example of the integration of geophysical,
geological and pondage test data to determine areas of highest seepage risk.

9. Recommendations
This study makes the following recommendations:
q Of the techniques trialed in this investigation, future channel seepage
measurement investigations should focus on geophysical techniques, as these
have shown the most promise to cost-effectively and relatively accurately
quantify channel seepage. Remote sensing trials, however, were not conducted in
these investigations. This technique has the potential for rapid assessment of long
sections of channel where seepage has a surface expression, and as such deserves
carefully planned field trials in Australian conditions. The baseline data collected
in this report could be used to assist in calibration of such trials.
q Rural Water Authorities should adopt the preferred technique as outlined in the
conclusions of this report (and the Guidelines Manual; ANCID, 2003) for channel
seepage measurement investigations. This methodology relies on geophysics
(preferably using direct detection of seepage impacts in the groundwater) to
identify seepage, and pondage tests and soil bores to calibrate and interpret the
q A considerable amount of geophysical data and interpretation was conducted in
this study. However the conclusions drawn were still limited by the size of the
data sets, and the limited range of environments over which the data was
collected. For example, more surveys are required at channels with a depth to
watertable of two to five metres. It is therefore recommended that a national
database be established to record all channel seepage measurement geophysical
trials. Surveys entered into the database must include a minimum level of site
information, including direct measurement of seepage rates, depth to watertable,
groundwater salinity, description of soil type and geology and channel hydraulic
information.
q Further study into the best method of establishing a relationship between the
geophysical response and seepage rates is required. At present the bulking
process of averaging the geophysical response over the entire pondage test area
necessarily introduces errors into the geophysical – seepage relationship. Further
investigation could focus on improved statistical methods for processing the
geophysical data (eg Geostatistics). Alternatively it could focus in the direction
of improving field techniques to directly measure seepage for comparison with
the geophysical response. For example, if a cheap method of bank construction
could be devised (eg manually inserted barriers) greater numbers of pondage tests
could be conducted which would significantly improve the regression equations
resulting from these relationships.
q The resistivity surveys undertaken in the final year of the trials were the
culmination of the investigation. These surveys showed significant promise, as
they allowed targeting of the seepage impacts on the watertable and visualisation
of seepage processes. However problems were encountered with the shallow
depth resolution of the equipment, which affected the accuracy of the results at
sites with a shallow depth to watertable. Further experimental trials to overcome
these teething problems are recommended. Investigation into means of reducing
resistivity data processing time (and thus costs) are also recommended.
q Exploration of a method which detects seepage by measuring changes from
background conditions is recommended. A significant problem encountered in
these trials when attempting to extrapolate a relationship from one section of a

channel to another, was caused by the fact that the background conditions change
along the channel. This means that the relationship between seepage and
geophysical response changes. A possible means around this is to conduct a
survey adjacent to the channel but away from channel seepage influence (eg 50-
100m away). These results would then be subtracted from the adjacent channel
(or on-channel) survey results. In theory the remaining difference should be
representative of channel seepage induced changes and not lithological changes.
Such a technique may allow improved transferability of correlations on or
between channels. This could even be undertaken with existing data sets of
EM31 surveys conducted in these trials using the 50m traverse line, provided
there are no seepage impacts at this distance. Possible limitations with this
method may include the effects of salinisation away from the channel and
different land uses adjacent to the channel.
q Further testing of the relative merits of on-channel fixed array surveys compared
to adjacent channel fixed array surveys are required. The evidence collected in
this investigation suggests on-channel (fixed array) surveys should only be
conducted where the geophysical technique can penetrate into the watertable, and
ideally target the top of the watertable. However these conclusions are only
based on evidence from three sites and further work is required to confirm this
conclusion.
q A means of calibrating geophysical surveys where pondage tests cannot be
conducted needs to be explored. On small to medium channels where pondage
tests cannot be conducted for operational reasons, points tests using the Idaho
meter could be used, although this could be expensive to ensure proper
calibration. For very large channels were pondage tests cannot be conducted due
to the very high expense of bank construction, alternative means of calibration
need to be devised.

10. References
Australian National Committee on Irrigation and Drainage (ANCID), (2000a): Open
Channel Seepage and Control – Vol 1.1 Literature Review of Channel Seepage
Identification and Measurement, Prepared by Sinclair Knight Merz, August
2000.
Australian National Committee on Irrigation and Drainage (ANCID), 2000b: Open

Channel Seepage and Control – Vol 1.2 Current Knowledge of Channel
Seepage Issues and Measurement in the Australian Rural Water Industry,
Prepared by Sinclair Knight Merz, August 2000.
Australian National Committee on Irrigation and Drainage (ANCID), 2002: Open

Channel Seepage and Control – Best Practice Guidelines for Channel Seepage
Identification and Measurement, Prepared by Sinclair Knight Merz, December
2002.
Fetter, C.W., 1988. Applied Hydrogeology. 2nd Edition. Macmillan Publishing

Company.
Freeze, AR and Cherry, AC, 1979. Groundwater. Prentice Hall, NJ.
Goulburn Murray Water (G-MW), 2000: Evaluation of EM31 to identify areas subject
to channel seepage – Waranga Western Channel – Boort. (Unpublished report),
December 2000.
Hotchkiss, R.H., Wingert, C.B. and Kelly, W.E., 2001. Determining Irrigation Canal
Seepage with Electrical Resistivity. Journal of Irrigation and Drainage
Engineering, January / February 2001.
McConachy, F., 1993. Channel Seepage Study. Rural Water Corporation, Salinity
Unit, Kerang, (unpublished report).
McNeil, J.D., 1980. Electromagnetic Terrain Conductivity Measurement at Low

Induction Numbers. Geonics Limited, Technical Note TN-6, October 1980.
Sinclair Knight Merz, 1995. Rocklands-Toolondo Channel Seepage in the Telangatuk

East Area, undertaken for Wimmera Mallee Water (unpublished report),
September 1995.
Sinclair Knight Merz, 1998. Investigation of channel seepage at Ralstons and

Sheridans, Donald Main Channel Seepage Investigation undertaken for
Wimmera Mallee Water (unpublished report), September 1998, p 95-104.
Smith, R.J., and Turner, A.K., 1982. Measurements of Seepage from Earthen
Irrigation Channels. Civ. Eng. Trans., Inst. Eng. Aust., CE24(4): 338-345.
Systat, version 9, Copyright 1998. (Statistical software package used in analysis).

Appendix A Geological Long Sections
WC01312:YR3_RPT(APPENDICES).DOC Final PAGE 302

A.1 Wimmera Mallee Water Geological Long Sections
Figure A-1 Donald Main Channel Geological Long Section

Figure A-2 Rocklands Channel Geological Long Section

Figure A-3 Toolondo Central Channel Geological Long Section

Figure A-4 Toolondo West Geological Long Section

Figure A-5 Toolondo East Geological Long Section

A.2 Murrumbidgee Irrigation Geological Long Sections
Figure A-6 Tabbita Geological Long Section

Figure A-7 Lake View Central Geological Long Section (based on 2000 drilling)

Figure A-8 Lake View Central Geological Long Section (based on 2000 & 2002 drilling)

Figure A-9 Lake View West Geological Long Section

A.3 Murray Irrigation Limited Geological Long Sections
Figure A-10 Dahwilly Central Geological Long Section (Based on 2000 drilling)

Figure A-11 Dahwilly Central Geological Long Section (Based on 2002 drilling)
(Note: Pond 6 & 7 in Figure A-11 are the same ponds as Pond 1 & 2 respectively in Figure A-10. Pond 6 & 7 were lined in July 2001.)

Figure A-12 Dahwilly East (Pretty Pine) Geological Long Section

Figure A-13 Finley Geological Long Section

Appendix B Pondage Test Results
B.1 Wimmera Mallee Water Geological Long Sections
B.1.1 Donald Main Pondage Tests: 21st - 29th December, 2000
Seepage Rate Vs Time For Pondage Test One: Donald Main Channel Seepage Rate Vs Time For Pondage Test Two: Donald Main Channel
80 80
70 70
Seepage Rate (m3/m2/day; mm/day)

60 60
50 50
40 40
30 30
20 20
10 10
0 0
21/12/00 22/12/00 23/12/00 24/12/00 25/12/00 26/12/00 27/12/00 28/12/00 29/12/00 30/12/00 21/12/00 22/12/00 23/12/00 24/12/00 25/12/00 26/12/00 27/12/00 28/12/00 29/12/00
Time Time
Seepage Rate Vs Time For Pondage Test Three: Donald Main Channel Seepage Rate Vs Time For Pondage Test Four: Donald Main Channel
80 80
70 70
Seepage Rate (m 3/m2/day; mm/day)
60 60
50 50
40 40
30 30
20 20
10 10
0 0
21/12/00 22/12/00 23/12/00 24/12/00 25/12/00 26/12/00 27/12/00 28/12/00 29/12/00 21/12/00 22/12/00 23/12/00 24/12/00 25/12/00 26/12/00 27/12/00 28/12/00 29/12/00 30/12/00
Time Time
Seepage Rate Vs Time For Pondage Test Five: Donald Main Channel Seepage Rate Vs Time For Pondage Test Six: Donald Main Channel
80 80
70 70
60 60
50 50
40 40
30 30
20 20
10 10
0 0
21/12/00 22/12/00 23/12/00 24/12/00 25/12/00 26/12/00 27/12/00 28/12/00 29/12/00 30/12/00 21/12/00 22/12/00 23/12/00 24/12/00 25/12/00 26/12/00 27/12/00 28/12/00 29/12/00 30/12/00
Time Time

B.1.2 Toolondo Central Pondage Tests: 19th March – 1st April, 2001
Seepage Rate Vs Time For Pondage Test One: Toolondo Channel Seepage Rate Vs Time For Pondage Test Two: Toolondo Channel
25 25

20 20
15 15
10 10
5 5
0 0
18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01 30/03/01 1/04/01 18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01 30/03/01 1/04/01
Time Time
Seepage Rate Vs Time For Pondage Test Three: Toolondo Channel Seepage Rate Vs Time For Pondage Test Four: Toolondo Channel
25 25
20 20
15 15
10 10
5 5
0 0
18/03/2001 20/03/2001 22/03/2001 24/03/2001 26/03/2001 28/03/2001 30/03/2001 1/04/2001 18/03/2001 20/03/2001 22/03/2001 24/03/2001 26/03/2001 28/03/2001 30/03/2001 1/04/2001
Time Time
Seepage Rate Vs Time For Pondage Test Five: Toolondo Channel Seepage Rate Vs Time For Pondage Test Six: Toolondo Channel
25 25
20 20
15 15
10 10
5 5
0 0
18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01 30/03/01 1/04/01 18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01 30/03/01 1/04/01
Time Time

B.1.3 Toolondo Central Pondage Test: 19th March – 3rd April, 2002
Seepage Rate Vs Time For Pondage Test One: Toolondo

45
40
35
30
25
20
15
10
0
19/03/2002 21/03/2002 23/03/2002 25/03/2002 27/03/2002 29/03/2002 31/03/2002 2/04/2002
Time

B.1.4 Toolondo East Pondage Tests: 20th March – 3rd April, 2002
Seepage Rate Vs Time For Pondage Test One: Toolondo (East) Seepage Rate Vs Time For Pondage Test Two: Toolondo (East)
14 5
12 4
3
10
2
8

1
6
0
4 19/03/2002 21/03/2002 23/03/2002 25/03/2002 27/03/2002 29/03/2002 31/03/2002 2/04/2002
-1
2
-2
0
-3
19/03/2002 21/03/2002 23/03/2002 25/03/2002 27/03/2002 29/03/2002 31/03/2002 2/04/2002
-2
-4
-4 -5
-6 -6
Time Time
Seepage Rate Vs Time For Pondage Test Three: Toolondo (East) Seepage Rate Vs Time For Pondage Test Four: Toolondo (East)
5 10
4
8
3
6
2
1 4
0
19/03/2002 21/03/2002 23/03/2002 25/03/2002 27/03/2002 29/03/2002 31/03/2002 2/04/2002 2
-1
-2 0
19/03/2002 21/03/2002 23/03/2002 25/03/2002 27/03/2002 29/03/2002 31/03/2002 2/04/2002
-3
-2
-4
-4
-5
-6 -6
Time Time

B.1.5 Toolondo West Pondage Tests: 19th March – 3rd April, 2002
Seepage Rate Vs Time For Pondage Test One: Toolondo West Seepage Rate Vs Time For Pondage Test Two: Toolondo West
7 10
6
8
5
6
4

3 4
2
2
1
0 0
19/03/2002 21/03/2002 23/03/2002 25/03/2002 27/03/2002 29/03/2002 31/03/2002 2/04/2002 19/03/2002 21/03/2002 23/03/2002 25/03/2002 27/03/2002 29/03/2002 31/03/2002 2/04/2002
-1
-2
-2
-4
-3
-4 -6
Time Time
Seepage Rate Vs Time For Pondage Test Three: Toolondo West Seepage Rate Vs Time For Pondage Test Four: Toolondo West
25 25
20
20
15
15
10
10
0
19/03/2002 21/03/2002 23/03/2002 25/03/2002 27/03/2002 29/03/2002 31/03/2002 2/04/2002
0
-5 19/03/2002 21/03/2002 23/03/2002 25/03/2002 27/03/2002 29/03/2002 31/03/2002 2/04/2002
Time Time

B.1.6 Rocklands Pondage Tests: 15th - 29th March, 2001
Seepage Rate Vs Time For Pondage Test One: Rocklands Channel Seepage Rate Vs Time For Pondage Test Two: Rocklands Channel
30 30
25 25

20 20
15 15
10 10
5 5
0 0
14/03/01 16/03/01 18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01 14/03/01 16/03/01 18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01
Time Time
Seepage Rate Vs Time For Pondage Test Three: Rocklands Channel Seepage Rate Vs Time For Pondage Test Four: Rocklands Channel
30 30
25 25
20 20
15 15
10 10
5 5
0 0
14/03/01 16/03/01 18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01 14/03/01 16/03/01 18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01
Time Time
Seepage Rate Vs Time For Pondage Test Five: Rocklands Channel Seepage Rate Vs Time For Pondage Test Six: Rocklands Channel
30 30
25 25
20 20
15 15
10 10
5 5
0 0
14/03/01 16/03/01 18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01 14/03/01 16/03/01 18/03/01 20/03/01 22/03/01 24/03/01 26/03/01 28/03/01
Time Time

B.2 Murrumbidgee Irrigation Pondage Test Results
B.2.1 Tabbita Pondage Tests: 18th– 25th June, 2001
Seepage Rate Vs Time For Pondage Test One: Tabbita Seepage Rate Vs Time For Pondage Test Two: Tabbita
16 16
14 14

12 12
10 10
8 8
6 6
4 4
2 2
0 0
18/06/01 19/06/01 20/06/01 21/06/01 22/06/01 23/06/01 24/06/01 25/06/01 26/06/01 18/06/01 19/06/01 20/06/01 21/06/01 22/06/01 23/06/01 24/06/01 25/06/01 26/06/01
Time Time
Seepage Rate Vs Time For Pondage Test Three: Tabbita Seepage Rate Vs Time For Pondage Test Four: Tabbita
16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
18/06/01 19/06/01 20/06/01 21/06/01 22/06/01 23/06/01 24/06/01 25/06/01 26/06/01 18/06/01 19/06/01 20/06/01 21/06/01 22/06/01 23/06/01 24/06/01 25/06/01 26/06/01
Time Time
Seepage Rate Vs Time For Pondage Test Five: Tabbita Seepage Rate Vs Time For Pondage Test Six: Tabbita
16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
18/06/01 19/06/01 20/06/01 21/06/01 22/06/01 23/06/01 24/06/01 25/06/01 26/06/01 18/06/01 19/06/01 20/06/01 21/06/01 22/06/01 23/06/01 24/06/01 25/06/01 26/06/01
Time Time

B.2.2 Lake View Central Pondage Tests: 25th June – 2nd July, 2001
Seepage Rate Vs Time For Pondage Test One: Lake View Branch Canal Seepage Rate Vs Time For Pondage Test Two: Lake View Branch Canal
16 16
14 14

12 12
10 10
8 8
6 6
4 4
2 2
0 0
25/06/01 26/06/01 27/06/01 28/06/01 29/06/01 30/06/01 1/07/01 2/07/01 3/07/01 25/06/01 26/06/01 27/06/01 28/06/01 29/06/01 30/06/01 1/07/01 2/07/01 3/07/01
Time Time
Seepage Rate Vs Time For Pondage Test Three: Lake View Branch Canal Seepage Rate Vs Time For Pondage Test Four: Lake View Branch Canal
16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
25/06/01 26/06/01 27/06/01 28/06/01 29/06/01 30/06/01 1/07/01 2/07/01 3/07/01 25/06/01 26/06/01 27/06/01 28/06/01 29/06/01 30/06/01 1/07/01 2/07/01 3/07/01
Time Time
Seepage Rate Vs Time For Pondage Test Five: Lake View Branch Canal Seepage Rate Vs Time For Pondage Test Six: Lake View Branch Canal
16 16
14 14
12 12
10 10
8 8
6 6
4 4
2 2
0 0
25/06/01 26/06/01 27/06/01 28/06/01 29/06/01 30/06/01 1/07/01 2/07/01 3/07/01 25/06/01 26/06/01 27/06/01 28/06/01 29/06/01 30/06/01 1/07/01 2/07/01 3/07/01
Time Time

B.2.3 Lake View Central Pondage Tests: 26th June – 7th July, 2002
Seepage Rate Vs Time For Pondage Test One: Lake View Branch Canal Seepage Rate Vs Time For Pondage Test Two: Lake View Branch Canal
10 10
9 9
8 8
7

6 6
5 5
4
4
3
3
2
2
1
1
0
0
26/06/2002 27/06/2002 28/06/2002 29/06/2002 30/06/2002 1/07/2002 2/07/2002 3/07/2002 4/07/2002
Time 26/06/2002 27/06/2002 28/06/2002 29/06/2002 30/06/2002 1/07/2002 2/07/2002 3/07/2002 4/07/2002
Time
Seepage Rate Vs Time For Pondage Test Three: Lake View Branch Canal Seepage Rate Vs Time For Pondage Test Four: Lake View Branch Canal
10 10
9 9
8 8
7 7
6 6
5 5
4 4
3 3
2 2
1 1
0 0
26/06/2002 27/06/2002 28/06/2002 29/06/2002 30/06/2002 1/07/2002 2/07/2002 3/07/2002 4/07/2002 26/06/2002 27/06/2002 28/06/2002 29/06/2002 30/06/2002 1/07/2002 2/07/2002 3/07/2002 4/07/2002
Time Time

B.2.4 Lake View West Pondage Tests: 26th June – 7th July, 2002
Seepage Rate Vs Time For Pondage Test One: Lake View West Seepage Rate Vs Time For Pondage Test Two: Lake View West
35 35
30 30
25 25

20 20
15 15
10 10
5 5
0 0
26/06/2002 27/06/2002 28/06/2002 29/06/2002 30/06/2002 1/07/2002 2/07/2002 3/07/2002 4/07/2002 26/06/2002 27/06/2002 28/06/2002 29/06/2002 30/06/2002 1/07/2002 2/07/2002 3/07/2002 4/07/2002
Time Time
Seepage Rate Vs Time For Pondage Test Three: Lake View West Seepage Rate Vs Time For Pondage Test Four: Lake View West
50 40
45
35
40
30
35
25
30
25 20
20
15
15
10
10
5
5
0 0
26/06/2002 27/06/2002 28/06/2002 29/06/2002 30/06/2002 1/07/2002 2/07/2002 3/07/2002 4/07/2002 26/06/2002 27/06/2002 28/06/2002 29/06/2002 30/06/2002 1/07/2002 2/07/2002 3/07/2002 4/07/2002
Time Time

B.3 Murray Irrigation Limited Pondage Test Results
B.3.1 Dahwilly Pondage Tests: 4th– 18th June, 2001
Seepage Rate Vs Time For Pondage Test One: Dahwilly Channel Seepage Rate Vs Time For Pondage Test Two: Dahwilly Channel
60 60
50 50

40 40
30 30
20 20
10 10
0 0
2/06/2001 4/06/2001 6/06/2001 8/06/2001 10/06/2001 12/06/2001 14/06/2001 16/06/2001 18/06/2001 20/06/2001 2/06/2001 4/06/2001 6/06/2001 8/06/2001 10/06/2001 12/06/2001 14/06/2001 16/06/2001 18/06/2001 20/06/2001
Time Time
Seepage Rate Vs Time For Pondage Test Three: Dahwilly Channel Seepage Rate Vs Time For Pondage Test Four: Dahwilly Channel
60 60
50 50
40 40
30 30
20 20
10 10
0 0
2/06/2001 4/06/2001 6/06/2001 8/06/2001 10/06/2001 12/06/2001 14/06/2001 16/06/2001 18/06/2001 20/06/2001 2/06/2001 4/06/2001 6/06/2001 8/06/2001 10/06/2001 12/06/2001 14/06/2001 16/06/2001 18/06/2001 20/06/2001
Time Time
Seepage Rate Vs Time For Pondage Test Five: Dahwilly Channel Seepage Rate Vs Time For Pondage Test Six: Dahwilly Channel
60 60
50 50
40 40
30 30
20 20
10 10
0 0
2/06/2001 4/06/2001 6/06/2001 8/06/2001 10/06/2001 12/06/2001 14/06/2001 16/06/2001 18/06/2001 20/06/2001 2/06/2001 4/06/2001 6/06/2001 8/06/2001 10/06/2001 12/06/2001 14/06/2001 16/06/2001 18/06/2001 20/06/2001
Time Time

B.3.2 Finley Pondage Tests: 6th– 20th July, 2001
Finley Pondage Test 1 (July 2001) Finley Pondage Test 2 (July 2001)
20 20
Pond 1 Pond 2
15 Average Seepage Rate 15 Average Seepage

(11/7 - 20/7) Rate (11/7 - 20/7)
Guage Drop - Evap + Rainfall (mm)

10 10
5 5
0 0
06/07/01 08/07/01 10/07/01 12/07/01 14/07/01 16/07/01 18/07/01 20/07/01 22/07/01 06/07/01 08/07/01 10/07/01 12/07/01 14/07/01 16/07/01 18/07/01 20/07/01 22/07/01
-5 -5
Time Time
Finley Pondage Test 3 (July 2001) Finley Pondage Test 4 (July 2001)
20 20
Pond 3 Pond 4
Average Gauge Drop - Evap + Rainfall (mm)
15 Average Seepage Rate 15 Average Seepage Rate

(11/7 - 20/7) (11/7 - 20/7)
10 10
5 5
0 0
06/07/01 08/07/01 10/07/01 12/07/01 14/07/01 16/07/01 18/07/01 20/07/01 22/07/01 06/07/01 08/07/01 10/07/01 12/07/01 14/07/01 16/07/01 18/07/01 20/07/01 22/07/01
-5 -5
Time Time

B.3.3 Dahwilly Pondage Tests: 6th– 21st June, 2002
Seepage Rate Vs Time For Pondage Test One: Dahwilly Central (Ponds 1-5) Seepage Rate Vs Time For Pondage Test Two: Dahwilly Central (Ponds 1-5)
8 9
7 8
7
6

6
5
5
4
4
3
3
2
2
1 1
0 0
5/06/2002 7/06/2002 9/06/2002 11/06/2002 13/06/2002 15/06/2002 17/06/2002 19/06/2002 21/06/2002 23/06/2002 5/06/2002 7/06/2002 9/06/2002 11/06/2002 13/06/2002 15/06/2002 17/06/2002 19/06/2002 21/06/2002 23/06/2002
Time Time
Seepage Rate Vs Time For Pondage Test Four: Dahwilly Central (Ponds 1-5)
Seepage Rate Vs Time For Pondage Test Three: Dahwilly Central (Ponds 1-
5) 14
18
16 12
14
m m 10
m/ 12 m/
da da
y) y)
Se 8
Se 10
ep ep
ag 8 ag
6
e e
Ra Ra
6
te te
4
( (
4
2 2
0
0
5/06/2002 7/06/2002 9/06/2002 11/06/2002 13/06/2002 15/06/2002 17/06/2002 19/06/2002 21/06/2002 23/06/2002
5/06/2002 7/06/2002 9/06/2002 11/06/2002 13/06/2002 15/06/2002 17/06/2002 19/06/2002 21/06/2002 23/06/2002
Time
Time
Seepage Rate Vs Time For Pondage Test Five: Dahwilly Central (Ponds 1-5)
14
12
10
0
5/06/2002 7/06/2002 9/06/2002 11/06/2002 13/06/2002 15/06/2002 17/06/2002 19/06/2002 21/06/2002 23/06/2002
Time

Seepage Rate Vs Time For Pondage Test Two (7): Dahwilly Channel Central
8 - Remediated Ponds
6
0
5/06/2002 7/06/2002 9/06/2002 11/06/2002 13/06/2002 15/06/2002 17/06/2002 19/06/2002 21/06/2002 23/06/2002
-1
Time
Seepage Rate Vs Time For Pondage Test One (6): Dahwilly Channel Central -
6 Remediated Ponds
4
0
5/06/2002 7/06/2002 9/06/2002 11/06/2002 13/06/2002 15/06/2002 17/06/2002 19/06/2002 21/06/2002 23/06/2002
-1
Time

B.3.4 Dahwilly East Pondage Tests: 30th May – 21st June, 2002
Seepage Rate Vs Time For Pondage Test One: Dahwilly East (Pretty Pine)
18
16
14
12
10
0
29/05/2002 3/06/2002 8/06/2002 13/06/2002 18/06/2002 23/06/2002
Time
Seepage Rate Vs Time For Pondage Test Two: Dahwilly East (Pretty Pine)
16
14
12
10
0
29/05/2002 3/06/2002 8/06/2002 13/06/2002 18/06/2002 23/06/2002
Time
Seepage Rate Vs Time For Pondage Test Three: Dahwilly East (Pretty Pine)
16
14
12
10
0
29/05/2002 3/06/2002 8/06/2002 13/06/2002 18/06/2002 23/06/2002
Time

B.3.5 Finley Pondage Tests: 31st May – 19th June, 2002
Seepage Rate Vs Time For Pondage Test One: Finley Main

12
10
0
31/05/2002 3/06/2002 6/06/2002 9/06/2002 12/06/2002 15/06/2002 18/06/2002 21/06/2002
Time
Seepage Rate Vs Time For Pondage Test Two: Finley Main

16
14
12
10
0
31/05/2002 3/06/2002 6/06/2002 9/06/2002 12/06/2002 15/06/2002 18/06/2002 21/06/2002
Time
Seepage Rate Vs Time For Pondage Test Three: Finley Main

16
14
12
10
0
31/05/2002 3/06/2002 6/06/2002 9/06/2002 12/06/2002 15/06/2002 18/06/2002 21/06/2002
Time

B.4 Goulburn Murray Water Pondage Test Results
B.4.1 Waranga Western Channel
Pondage Test 1 - 8-mile regulator Pondage Test 2 - Tonkin 1
Estimated Water Level Supply Level Start Level Estimated Water Level Supply Level Start Level
105.74 105.71
105.70
105.72
105.69
Water Level (m AHD)
seepage loss
Water Level (m AHD)

105.7 105.68
seepage loss 105.67
105.68
105.66
105.66 105.65
105.64 105.64
105.63
105.62
105.62
105.6 105.61
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
Date Date
Pondage Test 3 - Tonkin 2

Pondage Test 4 - Wilson 1
Estimated Water Level Supply Level Start Level
105.71
104.34
105.70
seepage loss seepage loss
105.69 104.32
Water Level (m AHD)
Water Level (m AHD)
105.68
104.30
105.67
104.28
105.66
105.65 104.26
105.64
104.24
105.63
105.62 104.22
105.61 104.20
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
Date Date
Pondage Test 5 - Wilson 2 Pondage Test 6 - Gierisch
104.34 104.30
104.29
104.32 seepage loss
seepage loss
104.28
Water Level (m AHD)
Water Level (m AHD)
104.30
104.27
104.28 104.26
104.26 104.25
104.24
104.24
104.23
104.22
104.22
104.20 104.21
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
Date Date

Pondage Test 7 - 24-mile regulator Pondage Test 8 - Pinks Road
104.30 104.30
104.20 104.20
seepage loss
Water Level (m AHD)

Water Level (m AHD)
104.10 seepage loss

104.10
104.00
104.00
103.90
103.90
103.80
103.80
103.70
103.60 103.70
29/5/02
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
29/5/02
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
Date Date
Pondage Test 9 - Lanyons Pondage Test 10 - Parkers 1
104.15 103.95
104.10
seepage loss 103.90
Water Level (m AHD)
Water Level (m AHD)
104.05
104.00
103.85
103.95 seepage loss
103.90
103.80
103.85
103.80 103.75
103.75
103.70 103.70
29/5/02
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
29/5/02
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
Date Date
Pondage Test 11 - Parkers 2
103.79
103.78
Water Level (m AHD)
103.77
103.76 seepage loss
103.75
103.74
103.73
103.72
29/5/02
30/5/02
31/5/02
1/6/02
2/6/02
3/6/02
4/6/02
5/6/02
6/6/02
7/6/02
8/6/02
9/6/02
10/6/02
11/6/02
Date

Appendix C Bulk Vertical Hydraulic Conductivity Calculations
Figure C-1 Example spreadsheet of bulk vertical hydraulic conductivity calculation for each pondage cell
Donald Representative Whole Profile - Evenly Weighted

Lithology Hydraulic Conductivity EM31 Response Near Surface (0-2m) Whole Profile - Surface Weighted
Number (Resistivity)
(P1-P3)
Depth Pond Pond Pond Effevtive Pond Effevtive Pond Effevtive Pond Pond Effevtive Pond Pond
POND 1 POND 2 POND 3 POND 1 POND 2 POND 3 Thickness Pond 2 Pond 3 Thickness Pond 1 Pond 2 Thickness Pond 3 Thickness Pond 2
Interval 1 2 3 1 3 1 2 1 3
(Weighting) (Weighting) (Weighting) (Weighting)
1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3
0 - 0.5 ClyS ClyS ClyS ClyS ClyS ClyS ClyS ClyS ClyS 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.50 0.50 0.50 0.48 0.96 0.96 0.96 2.5 5.00 5.00 5.00 0.90 1.80 1.80 1.80 0.5 1.00 1.00 1.00
0.5 - 1 FS FS FS MC MC MC FS MC MC 2 2 2 1E-04 1E-04 1E-04 2 1E-04 1E-04 2.00 0.00 0.67 1.07 0.53 10681 1.60 2.5 1.25 25000 3.75 0.90 0.45 9000 1.35 0.5 0.25 5000 0.75
1 - 1.5 FS FS FS MC FS FS FS FS MC 2 2 2 1E-04 2 2 2 2 1E-04 2.00 1.33 1.33 1.30 0.65 0.98 0.98 2.5 1.25 1.87 1.87 0.90 0.45 0.67 0.67 0.5 0.25 0.37 0.37
1.5 - 2 FS FS FS FS FS FS FS FS FS 2 2 2 2 2 2 2 2 2 2.00 2.00 2.00 1.32 0.66 0.66 0.66 2.5 1.25 1.25 1.25 0.90 0.45 0.45 0.45 0.5 0.25 0.25 0.25
2 - 2.5 FS FS FS FS FS FS FS FS FS 2 2 2 2 2 2 2 2 2 2.00 2.00 2.00 1.20 0.60 0.60 0.60 0.80 0.40 0.40 0.40 0.5 0.25 0.25 0.25
2.5 - 3 FS FS FS FS FS FS SC FS FS 2 2 2 2 2 2 0.07 2 2 2.00 2.00 1.36 1.02 0.51 0.51 0.75 0.80 0.40 0.40 0.59 0.5 0.25 0.25 0.37
3 - 3.5 FS FS FS FS FS FS SC FS FS 2 2 2 2 2 2 0.07 2 2 2.00 2.00 1.36 0.82 0.41 0.41 0.60 0.70 0.35 0.35 0.52 0.5 0.25 0.25 0.37
3.5 - 4 FS FS FS SC FS FS SC FS FS 2 2 2 0.07 2 2 0.07 2 2 2.00 1.36 1.36 0.64 0.32 0.47 0.47 0.70 0.35 0.52 0.52 0.5 0.25 0.37 0.37
4 - 4.5 SC SC SC SC SC SC SC FS FS 0.07 0.07 0.07 0.07 0.07 0.07 0.07 2 2 0.07 0.07 1.36 0.51 7.30 7.30 0.38 0.60 8.57 8.57 0.44 0.5 7.14 7.14 0.37
4.5 - 5 SC SC SC SC SC SC FS FS FS 0.07 0.07 0.07 0.07 0.07 0.07 2 2 2 0.07 0.07 2.00 0.43 6.11 6.11 0.21 0.60 8.57 8.57 0.30 0.5 7.14 7.14 0.25
5 - 5.5 SC SC SC SC SC SC FS FS FS 0.07 0.07 0.07 0.07 0.07 0.07 2 2 2 0.07 0.07 2.00 0.39 5.50 5.50 0.19 0.40 5.71 5.71 0.20 0.5 7.14 7.14 0.25
5.5 - 6 SC SC SC SC SC SC SC SC SC 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.36 5.10 5.10 5.10 0.40 5.71 5.71 5.71 0.5 7.14 7.14 7.14
6 - 6.5 SC SC SC SC SC SC SC SC SC 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.30 4.34 4.34 4.34 0.30 4.29 4.29 4.29 0.5 7.14 7.14 7.14
6.5 - 7 SC SC SC SC SC SC SC SC SC 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.17 2.40 2.40 2.40 0.30 4.29 4.29 4.29 0.5 7.14 7.14 7.14
7 - 7.5 SC SC SC SC SC SC SC SC SC 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.20 2.86 2.86 2.86 0.5 7.14 7.14 7.14
7.5 - 8 SC SC SC SC SC SC SC SC SC 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.20 2.86 2.86 2.86 0.5 7.14 7.14 7.14
8.5 - 9 SC SC SC SC SC SC SC SC SC 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.10 1.43 1.43 1.43 0.5 7.14 7.14 7.14
9.5 -10 SC SC SC SC SC SC SC SC SC 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.07 0.10 1.43 1.43 1.43 0.5 7.14 7.14 7.14
10 0.282 0.001 0.519 10 1.143 0.000 0.842 10 0.188 0.001 0.303 10 0.113 0.002 0.145
The vertical hydraulic conductivity for each pondage cell was calculated based on the following formula (after Freeze and Cherry, 1979). For each layer ‘d’ (as shown in the figure below) where different soil types were
encountered within the one pondage cell, the average of the different soil types was used, as shown in the above example.
d1 k1
d2 k2
d
dn kn
d
Kz = n
di
å
i =1 K i

Appendix D Summary of Project Brief for
Remote Sensing in the Wimmera
Background
The brief presented in this Appendix is the brief prepared for a planned remote sensing
trial in the Wimera. This was not carried out due to budget constraints (refer Section
4.5). The brief is presented here as an indication of the issues requiring consideration
when conducting remote sensing as part of a channel seepage investigation. It is not a
prescriptive method for using this technique. The brief outlined the requirements for
conducting an investigation into the feasibility of using remote sensing techniques for
channel seepage identification and quantification. Remote sensing techniques can
offer an efficient method to detect and locate seepage without interfering with channel
operations. Remote sensing refers to any kind of data recording by a sensor which
measures energy emitted or reflected by objects located at some distance from the
sensor and includes aerial photography and satellite imagery.
All remote sensing techniques for channel seepage detection assume that channel
seepage has a surface expression adjacent to the channel. This may be detected as
increased soil moisture and / or vegetation vigour and water status. These techniques
are limited to detecting seepage that migrates laterally through the channel banks,
and/or re-surfaces near the channel toe. Remote sensing cannot account for seepage
that moves vertically to the groundwater and does not re-surface. A key aspect of
remotely sensed data is that it must be at a suitable resolution to allow definition of
seepage zones. Typical seepage zones may be 10 – 20 m in width adjacent to a 10 –
20 m channel. Therefore ground resolutions of less than 10 m are required.
The proposed project was to be inter-related with the EM surveys and soil survey
assessments. It was proposed that the results be brought into the proposed project
using GIS.
Figure E-1 shows the major regions of the electromagnetic spectrum that are used in
remote sensing. The regions most useful for channel seepage detection include
visible, reflected (near) infrared and thermal infrared.
n Figure E-1 Major regions of the electromagnetic spectrum pertinent to

remote sensing
UV Visible Reflected IR Thermal (emitted) IR Microwave
0
0.2µm 0.5 1 5 10 20 100µm 0.1 cm 1 cm 1m
Wavelength(not to scale)
Human vision Imaging radar
Thermal IR scanners
Photographic cameras Passive microwave
Electro-optical sensors

Project Description
The project aims are to investigate and evaluate remote sensing techniques and spatial
data analysis for the identification and quantification of channel seepage. The project
should be conducted on selected pilot areas within the WMW channel system and in
an area where seepage sites have been detected and quantified using other methods.
Task One – Identification of Channel Seepage

This task was to involve the identification of channel seepage using remotely sensed
data analysis techniques to identify potential seepage sites. This would be followed by
on ground verification of these sites and an evaluation of the accuracy of the
technique/s implemented. The aim of the task was to evaluate the ability of remotely
sensed imagery to detect known channel seepage sites.
Task Two – Quantification of Seepage

The second task was to involve the investigation of the potential of an integrated
spatial data analysis approach for quantifying channel seepage. The spatial data
analysis would combine results from Task One with those from the EM and Soil
surveys, and pondage test data using GIS (Geographical Information Systems). The
aim of this task was to determine the ability of integrated spatial data analysis to
quantify channel seepage.
Methodology
Task One – Identification of Channel Seepage
Data source review and image acquisition
A review of available remotely sensed data was to be undertaken culminating in a
comparison of spatial and spectral resolutions, and costs of acquisition and analysis.
Previous data collected by WMW was to be assessed and published literature
investigated.
The nature of channel seepage suggests that the source data should have high spatial
resolution (10 m or less) and that it is multispectral (ie. has data collected from more
than one distinct region of the electromagnetic spectrum). Distinct data from the
infra-red region is expected to be the most beneficial as this area of the spectrum is
strongly absorbed by water and will be able to most distinctly separate areas of
varying soil moisture and plant water and growth status.
The review was also to investigate the optimum data collection time. It is expected
that increased surface moisture and vegetation growth due to channel seepage would
be particularly evident during late summer and early autumn when surrounding areas
(apart from irrigation) would be distinctly drier. In addition, imagery from more than
one date would be useful to remove the effects of crop irrigation and other seasonal
variations. Thus the temporal dimension of the imagery was also to be investigated.
Remotely sensed image data sources may include:

q Digital infra-red aerial photography;
q Airborne high resolution sensor data; and,
q Satellite imagery.

Image analysis
Using high-level image analysis software (eg. ER MAPPER or ERDAS IMAGINE)
multidate imagery was to be combined and analysed for the selected study area along
with ground data. Automated image analysis and GIS techniques were to be
developed where possible to identify and map potential channel seepage sites.
Accuracy assessment and evaluation

The accuracy of the developed remote sensing techniques was to be assessed by
comparison to field surveys, with respect to:
q the percentage of sites identified as potential locations of channel seepage
that are actually channel seepage locations (the converse of this is the error
of commission, or over estimation);
q the percentage of channel seepage locations identified using the techniques
(the converse of this is the error of omission, or under estimation); and
q The key characteristics of inaccurately detected sites were to be identified to
allow the development of recommendations that may reduce these errors.
Task Two – Quantification of seepage
Spatial data from a number of sources, including Task One, was to be combined and
analysed using GIS. Data sources were to include:
q airborne radiometric and electromagnetic data (EM data)
q soil survey assessments.
q channel flow and width
q pondage test data
The extent to which the input data could quantify seepage at known locations would
be assessed by comparison to the pondage test data.
Trial Requirements
Budget
The budget for the pilot study was expected to be $15,000 to $20,000, allowing for
approximately $5,000 to $10,000 in image costs.
Information Technology
Software requirements include high-end image analysis software such as ER
MAPPER and / or ERDAS IMAGINE and advanced GIS software such as ArcInfo
and Arcview including support of raster data formats. The imagery and GIS data
required to be collated may need approx . 1 to 5 gigabytes.
Skills and experience

The required skills and experience to undertake this type of task would include:
q Geophysical interpretation (EM31 & EM34), soil data interpretation and
analysis;
q Knowledge / familiarity with GIS and spatial analysis;
q Remote sensing data processing and analysis;
q Accuracy assessment and evaluation.

Appendix E Statistical Output from Regional
Assessment Regression Analysis
Task 1 – Multiple Linear Regression With DTWT 5-10m
(All site but Finley, Lake View, Lake View West and Donald)
Dep Var: SEEPAGE N: 40 Multiple R: 0.743 Squared multiple R: 0.552
Adjusted squared multiple R: 0.528 Standard error of estimate: 2.739 (0.48%

of mean)
Effect Coefficient Std Error Std Coef Tolerance t P(2

Tail)
CONSTANT 11.623 1.294 0.000 . 8.985

0.000
EM31 -0.118 0.021 -0.626 0.960 -5.573
0.000
UPPER_KV 4.352 1.666 0.294 0.960 2.613
0.013
Analysis of Variance
Source Sum-of-Squares df Mean-Square F-ratio P
Regression 341.936 2 170.968 22.783 0.000

Residual 277.660 37 7.504
-------------------------------------------------------------------------------
*** WARNING ***

Case 32 has large leverage (Leverage = 0.406)
Durbin-Watson D Statistic 1.450

First Order Autocorrelation 0.247

Plot of Residuals against Predicted Values
10
5
RESIDUAL
-5
-5 0 5 10 15
ESTIMATE
Regression on I:\W CMS\W c01312\060-Reports\D A T 00_01.10\Statistics (calibration)

Residual s vs. Seepage Residuals vs. EM31
7 7
6 6
5 5
4 4
3 3
Residual
Residual
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
-4 -4
0 2 4 6 8 10 12 20 30 40 50 60 70 80 90 100 110
Seepage EM31
Residuals vs. Upper_kv
7
6
5
4
3
Residual
2
1
0
-1
-2
-3
-4
0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 1.1 1.2
Upper_kv
Note: Transformations of Upper_kv did not greatly improve the fit of the
model and so were not followed any further.

Task 2 – Multiple Linear Regression With DTWT < 2m
(Lake View, Lake View West and Donald)
Note – no other variables were significant.

Adjusted squared multiple R: 0.880 Standard error of estimate: 5.306 (0.23%

of mean)

Tail)
CONSTANT 71.961 5.178 0.000 . 13.897

0.000
EM31 -0.690 0.070 -0.943 1.000 -9.797
0.000
Regression 2702.405 1 2702.405 95.978 0.000

Residual 337.879 12 28.157
-------------------------------------------------------------------------------
*** WARNING ***

Case 5 is an outlier (Studentized Residual = 2.626)

First Order Autocorrelation -0.384
15
10
RESIDUAL
-5
-10
0 10 20 30 40 50
ESTIMATE

Regression on temp.sdy (calibration)
Residual s v s. seepage Residuals vs. EM31

10 10
8 8
6 6
4 4
Residual
Residual
2 2
0 0
-2 -2
-4 -4
-6 -6
-8 -8
5 10 15 20 25 30 35 40 45 50 60 70 80 90 100 110
seepage EM31
Note: No transformations required.
Task 3 –Regression of only EM31 With DTWT 5-10m

Adjusted squared multiple R: 0.455 Standard error of estimate: 2.942 (51% of

mean)

Tail)
CONSTANT 12.788 1.304 0.000 . 9.807

0.000
EM31 -0.129 0.022 -0.685 1.000 -5.796
0.000
Regression 290.708 1 290.708 33.589 0.000

Residual 328.888 38 8.655
-------------------------------------------------------------------------------


4
RESIDUAL
-2
-4
-6
-5 0 5 10 15
ESTIMATE
Task 4 –Prediction bands for equations with only EM31

Work saved in other excel spreadsheets.
Task 5 – Multiple Linear Regression with Resistivity at 10m

All sites were used initially and Resistivity was not found to be significant,
only DTWT. Therefore used the two groups used to predict with EM31.
DTWT 5-10m
17 case(s) deleted due to missing data.

mean)

Tail)
CONSTANT 3.004 0.964 0.000 . 3.117

0.005
UPPER_KV 7.460 2.394 0.524 1.000 3.116
0.005
RESISTIVITY 0.010 0.004 0.408 1.000 2.430
0.025
Regression 166.164 2 83.082 7.705 0.003

Residual 215.646 20 10.782
-------------------------------------------------------------------------------
*** WARNING ***

Case 32 has large leverage (Leverage = 0.597)

10
5
RESIDUAL
-5
0 5 10 15
ESTIMATE

Residual s vs. Seepage Residuals vs. Upper_kv
8 8
7 7
6 6
5 5
4 4
Resi dual
Resi dual
3 3
2 2
1 1
0 0
-1 -1
-2 -2
-3 -3
4 5 6 7 8 9 10 11 12 13 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 1.1 1.2
Seepage Upper_kv
Residuals vs. Resistivity
8
7
6
5
4
Residual
3
2
1
0
-1
-2
-3
50 100 150 200 250 300 350 400 450 500 550
Resi stivity
Looked at transforms. Raised Seepage to the power of 0.2 and greatly

improved the standard error. The new results are provided below:
REGRESSION MODEL RESULTS

~~~~~~~~~~~~~~~~~~~~~~~~
Dependent var: Seepage N: 23 Mult. R: 0.647 Mult. R^2:

0.419
Std error est.: 0.243 (19% of
mean)
Variable Power Coefficient Std Error t-Stat

Prob(t)
Depndnt Var 0.2000
Constant 1.0000 1.1152 0.0714 15.62916
0.00000
Upper_kv 1.0000 0.4716 0.1776 2.65607
0.01516
Resistivity 1.0000 0.8250E-03 0.0003 2.75541
0.01220
Source Sum-of-Squares df Mean-Square F-ratio
Regression 0.854 2 0.427 7.217

Residual 1.183 20 0.059
Total 2.037 22

Residual s vs. Seepage Residuals vs. Upper_kv
.40 .40
.30 .30
.20 .20
.10 .10
Resi dual
Resi dual
0 0
-.10 -.10
-.20 -.20
-.30 -.30
-.40 -.40
1.20 1.30 1.40 1.50 1.60 1.70 1.80 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 1.1 1.2

Seepage Upper_kv
Residuals vs. Resistivity
.40
.30
.20
.10
Residual
0
-.10
-.20
-.30
-.40
50 100 150 200 250 300 350 400 450 500 550
Resi stivity
DTWT < 2m
(Only resistivity used because once any other variable was added it was no
longer significant).

mean)

Tail)
CONSTANT -0.662 5.278 0.000 . -0.125

0.904
RESISTIVITY 1.714 0.549 0.787 1.000 3.124
0.020
Regression 384.634 1 384.634 9.761 0.020

Residual 236.436 6 39.406
-------------------------------------------------------------------------------


10
5
RESIDUAL
-5
-10
0 10 20 30
ESTIMATE
Looked at transforms but not enough data to see any trends.
Task 6 –Regression with only Resistivity at 10m for DTWT 5-10m


mean)

Tail)
CONSTANT 4.201 1.051 0.000 . 3.996

0.001
RESISTIVITY 0.010 0.005 0.401 1.000 2.007
0.058
Regression 61.452 1 61.452 4.028 0.058

Residual 320.357 21 15.255
-------------------------------------------------------------------------------


10
5
RESIDUAL
-5
4 5 6 7 8 9 10
ESTIMATE
I looked at transforms and found that raising seepage to the power of 0.2
greatly improved the model. The results are below:
REGRESSION MODEL RESULTS

~~~~~~~~~~~~~~~~~~~~~~~~
Dependent var: Seepage N: 23 Mult. R: 0.463 Mult. R^2:

0.214
Std error est.:
0.276
Variable Power Coefficient Std Error t-Stat

Prob(t)
Depndnt Var 0.2000
Constant 1.0000 1.1906 0.0743 16.02543
0.00000
Resistivity 1.0000 0.8133E-03 0.0003 2.39344
0.02611
Source Sum-of-Squares df Mean-Square F-ratio
Regression 0.436 1 0.436 5.729

Residual 1.600 21 0.076
Total 2.037 22

Residual s vs. Seepage Resi duals vs. Resistivity

.4 .4
.3 .3
.2 .2
.1 .1
Residual
Residual
0 0
-.1 -.1
-.2 -.2
-.3 -.3
-.4 -.4
-.5 -.5
1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70 1.75 50 100 150 200 250 300 350 400 450 500 550
Seepage Resisti vi ty

Measure Trials

Uploaded by

Copyright:

Available Formats

Measure Trials

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Measure Trials

Uploaded by

Copyright:

Available Formats

Foreword

In response to the recommendations, ANCID formed an industry Task Force to

Each stage of the project is briefly described as follows:

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE i

WC01312:YR3_RPT_CH_1.3(FORE. EXEC SUMM, INTRO).DOC Final PAGE ii

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE iii

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE iv

6. Waranga Western Channel: Case Study......................................233

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE v

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE v i

Appendix C Bulk vertical hydraulic conductivity calculations 334

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE vii

Printed: 10 April, 2003

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE viii

The Australian National Committee of Irrigation and Drainage (ANCID), in

E.2 Overview of Trials Conducted

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE i

E.2.2 Trial Program Summary Table

E.2.3 Techniques Not Included in the Trial Program

For additional information on these techniques refer to the Literature Review

E.2.4 Assessment Methodology

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE ii

E.3 Description of Trial Sites

n Figure E-1 Trial Site Regional Location Map

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE iii

Toolondo March 01 (6 cells) Dec. 00 (land)

Wimmera Toolondo (East) March 02 (4 cells) March 02 (land) - March 02 June 02 - - -

Rocklands March 01 (6 cells) Aug. 01 (land & boat) Nov. 99 - Aug. 01 - - -

Dahwilly (East) June 02 (3 cells) March 02 (land & - - - -

Finley July 01 (4 cells) July 00 (land) - July 00 - - -

Lake View July 01 (6 cells) June 00 (land) Dec. 00 - Aug. 00 – Aug 02 -

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE iv

E.5 Sub-surface Characterisation

E.6 Point Tests

E.7 Groundwater Techniques

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE v

E.8 Remote Sensing

1) Directly measuring the conductivity of the groundwater, and identifying the

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE v i

n Figure E-2 Comparison of how geophysical techniques can be used to

2. Seepage inferred 1. Seepage measured directly

Fresher seepage plume

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE vii

Comparison of Trialed Geophysical Techniques

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE viii

Critical Geophysical Survey Variables

Regional Assessment of Key Relationships

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE ix

n Figure E-3 Regional EM31 Assessment: Pondage Test Seepage Versus

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE x

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE xi

5. Conduct Field Trials

6. Evaluation – Evaluate whether investigation objectives have been met.

E.9.2 Summary of EM34 Results

WC01312: YR3_RPT_CH_1-3(FORE, EXEC SUMM, INTRO).DOC Final PAGE xii

Surface to 1.5 EM31 (horizontal dipole)5 Direct watertable impact 3 0-1

1.5 – 5 EM31 (vertical dipole)5 Direct watertable impact 6 1 – 3.5

EM34 – 10m coil spacing Direct watertable impact 15 3 - 10

EM34 – 20m coil spacing Direct watertable impact 30 6 - 20