Accepted Manuscript

Review papers

Identifying turbulent flow in carbonate aquifers

Stephen R.H. Worthington, Robert W.N. Soley

PII: S0022-1694(17)30448-1
DOI: http://dx.doi.org/10.1016/j.jhydrol.2017.06.045
Reference: HYDROL 22099

To appear in: Journal of Hydrology

Received Date: 19 March 2017

Revised Date: 24 June 2017
Accepted Date: 26 June 2017

Please cite this article as: Worthington, S.R.H., Soley, R.W.N., Identifying turbulent flow in carbonate aquifers,
Journal of Hydrology (2017), doi: http://dx.doi.org/10.1016/j.jhydrol.2017.06.045

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Identifying turbulent flow in carbonate aquifers

Stephen R.H. Worthington

Worthington Groundwater, 55 Mayfair Avenue, Dundas, Ontario L9H 3K9, Canada; (1) 905-627-1781;

sw@worthingtongroundwater.com

Robert W.N. Soley

Amec Foster Wheeler Environment & Infrastructure, Copper Beeches, St Kew, Bodmin, Cornwall, PL30

3HB, U.K.

Keywords

Turbulent flow; Dual porosity; Conceptual model; Carbonate aquifer; Groundwater flooding; Karst

Abstract

Turbulent flow has a different hydraulic response compared to laminar flow and so it is important to be

able to identify its occurrence in an aquifer, and to predict where it is likely to be found. Turbulent flow is

associated with large apertures and rapid velocities, and these occur most frequently in carbonate aquifers.

Methods for identifying turbulent flow include correlating spring discharge with head variation,

calculating Reynolds numbers from spring discharge and tracer velocity, and plotting the spatial variation

of head differences between high flow and low flow. The probability of turbulent flow increases as a

function of permeability and of spring discharge, and the probability increases in a downgradient

direction in an aquifer. Spring discharge is a key parameter for evaluating the presence of turbulent flow,

which is likely to occur where a spring with a discharge >1 L/s is fed by a single channel. Turbulent flow

appears to be a major contributing factor to the occurrence of groundwater flooding in carbonate aquifers.

1. Introduction

1
 Groundwater models have become an increasingly important part of hydrogeology in the last fifty

years. Traditionally, finite difference and finite element models have only simulated laminar flow, but in

recent years the capability to simulate turbulent flow had been added in programs such as CAVE (Liedl et

al., 2003; Hubinger et al., 2016), MODFLOW DCM (Sun et al., 2005), MODFLOW CFP (Shoemaker et

al., 2008a; Gallegos et al, 2013), KFM (Loper and Chicken, 2011), CTFC (Reimann et al., 2011), DisCo

(de Rooij et al., 2013), MODFLOW NLFP (Mayaud et al., 2015), and MODFLOW USG-Beta (GSI,

2017). The pipe flow model SWMM has also been used to simulate turbulent groundwater flow (Chen

and Goldscheider, 2014). This raises the question of where turbulent flow is likely to be found in aquifers.

The major consequence of the difference between laminar flow and turbulent flow is that the

head/discharge relationship is linear in the former case and non-linear in the latter case. This is relevant to

the calibration of numerical models and is especially relevant to head changes in transient models. In

addition, turbulent flow can mobilize sediment grains and transport them through aquifers, and such

suspended sediment impacts water quality.

The earliest testing of whether flow in an aquifer is laminar or turbulent was by Darcy (1856). He

considered whether flow to artesian wells and to springs better fitted his non-linear flow law (the Darcy-

Weisbach equation) or his linear flow law (Darcy's law). He concluded that flow to artesian wells fitted

his linear flow law in most cases and that flow to a large spring in a limestone aquifer fitted his non-linear

flow law (Darcy, 1856, p. 137-157, 182-183). The conclusions of Darcy were largely confirmed by later

studies. For instance, Hubbert (1940) showed that turbulent flow is rare in granular aquifers where grain

size is 1 mm or less, and Smith et al. (1976) showed that flow is generally turbulent in the major channels

in carbonate aquifers. The permeability of carbonate aquifers is higher on average than other bedrock

aquifers (Gleeson et al., 2011), so large apertures, rapid groundwater velocities, and thus turbulent flow

are more likely to occur than in other lithologies. Consequently, this paper will focus on carbonate

aquifers.

It has long been known that continuous, large-aperture channels are common in carbonate

aquifers. Caves represent the most striking manifestations of such channels. Some cave passages have

2
been mapped for distances exceeding 10 km, and tracer tests with velocities exceeding 1 km/d provide

evidence for channels over even longer distances (Worthington, 2015a). Turbulent flow in past conditions

has been deduced from bedform-erosional features and sediments in cave passages (Gale, 1984), and

turbulent flow under current conditions has been deduced from the combination of tracer velocity and

spring discharge (Atkinson, 1977), and from the combination of head measurements and spring discharge

(Bonacci, 2001; Jeannin, 2001; Sepúlveda 2009). Despite such studies, there has been no comprehensive

discussion of the range of different methods for identifying turbulent flow in aquifers, nor of the

likelihood of turbulent flow occurring in a given carbonate aquifer. These issues will be addressed in this

paper. Section 2 outlines the theory and methods used to identify turbulent flow, with examples in Section

3, practical applications in Section 4, and a discussion in Section 5.

2. Theory and methods

2.1 Laminar and turbulent flow

The transition between laminar flow and turbulent flow is determined by the dimensionless

Reynolds number R, which for pipes is defined as

where ρ is the density of water, v is groundwater velocity, d is pipe diameter, and µ is the dynamic

viscosity of water. In engineering, the critical Reynolds number that defines the transition from laminar to

turbulent flow is about 2100 for straight pipes and 1000 for openings with parallel walls (Street et al.,

1996, p. 233). In the latter case, d in Equation 1 refers to fracture aperture. In granular aquifers, mean

grain diameter is often used for d, and the critical Reynolds number ranges from 1 to 60 (Shoemaker et

al., 2008a). Natural channels in bedrock aquifers present an intermediate case between smooth, straight

manufactured pipes and the tortuous pathways in granular aquifers and so will have intermediate critical

Reynolds numbers. For instance, measurements have shown that turbulent flow in fractures and channels

3
can occur at Reynolds numbers well below 1000 (Qian et al., 2005; Shoemaker et al., 2008b; Chen and

Qian, 2009). Figure 1 shows how Reynolds numbers vary as a function of channel aperture and

groundwater velocity.

--- Approximate location of Figure 1 ---

Laminar flow in a porous medium is described by Darcy's law

where Q is discharge, K is hydraulic conductivity, i is hydraulic gradient, and A is the area of the cross-

section through which water flows. Laminar flow in a circular channel can be calculated from the Hagen-

Poiseuille equation (White, 1988)

where g is the acceleration due to gravity, and r is the radius of the channel. Equation 3 assumes that flow

is in a smooth pipe, and the calculated apertures are termed hydraulic apertures. Actual apertures will be

somewhat higher because the channel roughness is not factored into Equation 3. Turbulent flow in a

circular channel can be described by the Darcy-Weisbach equation (Ford and Williams, 2007)

where A is the cross-sectional area of the channel and f is the Darcy-Weisbach friction factor.

Straight sections of manufactured pipes typically have friction factors of 0.01 to 0.05, but in

addition there are substantial head losses at bends and changes in cross-section (Street et al., 1996).

Natural channels in carbonate aquifers have much higher friction factors than straight manufactured pipes

because the walls are not smooth and the channels are not straight, and results from several studies

suggest that the geometric mean friction factor is between 1 and 10 (Jeannin, 2001).

2.2 Methods for identifying turbulent flow

4
 Six methods are used here to identify situations where flow in channels is likely to be turbulent.

The first method calculates Reynolds numbers from channel diameter and velocity data (Equation 1). The

remaining five methods use inductive logic, and infer the presence of turbulent flow from water level,

spring discharge, groundwater velocity, cave stream, and suspended sediment data. The six methods are

not exhaustive. Any methods that yield estimates of channel diameter or groundwater velocity can help

define the extent of turbulence. For instance, the attenuation of a temperature signal can be used to

provide an independent estimate of channel diameter (Covington et al., 2012; Birk et al., 2014).

2.2.1 Using tracer tests and spring discharge to determine flow regime

The first way to determine whether channel flow is turbulent is where groundwater velocity has

been determined by a tracer test from a sinking stream to a spring and where spring discharge has also

been measured. Assuming flow in a circular channel, the diameter of the channel can be calculated from

the continuity equation

Then the Reynolds number can be calculated using Equation 1. The literature was searched for examples

in carbonate aquifers where both sink to spring tracer velocities and spring discharge have been measured,

and the results were compiled to determine the relationship between spring discharge and Reynolds

number.

2.2.2 Correlation with spring discharge with head in wells

The second way is to correlate spring discharge with the head difference between a well and a

spring to determine whether flow in the channel feeding the spring is laminar or turbulent. If flow in an

aquifer is laminar then it can be described by Darcy's law (Eq. 2) if flow is intergranular or by the Hagen-

Poiseuille equation (Eq. 3) if flow is in channels. In both cases there is a linear relationship between

discharge and head difference (Fig. 2a). However, if flow in the channels is turbulent then the head

difference is proportional to the square of the discharge (Eq. 4 and Fig. 2b). In most cases a monitoring

5
well will not lie directly on a major channel. The response in such an off-channel well will be linear if

flow in the major channel is laminar (Fig. 2c). However, if flow in the major channel is turbulent and flow

in the smaller channels or fractures between the well and the channel is laminar then the relationship

between head difference (h) and discharge is described by summing the head losses

where the terms on the right side of the equation refer to turbulent flow in the main channel or channels

and to laminar flow in the matrix, respectively (Fig. 2d). A more complicated situation occurs where flow

in the main channel is laminar in low-flow conditions and turbulent in high-flow conditions. In this case a

hybrid curve will result, with a break in slope at the point where flow becomes turbulent (Fig. 2e and 2f).

--- Approximate location of Figure 2 ---

The third way is possible where the location of major channels can be identified from

potentiometric surface maps that show troughs, where the troughs are inferred to coincide with major

channels (Worthington, 2015b). In this case, the discharge in the channels can be calculated from

where Qc is the discharge at any point along a major channel, Bc is the area of the groundwater basin

upgradient from that location, and I is the mean infiltration rate. After calculating discharge in the

channel, the Reynolds number can be calculated from the relationship between discharge and the

Reynolds number derived by the method in Section 2.2.1.

2.2.4 Areas of large head changes centred on potentiometric surface troughs

Discharge is proportional to hydraulic gradient where flow is laminar (Eq. 3), but is proportional

to the square root of hydraulic gradient where flow is turbulent (Eq. 4). This results in large increase in

head in high-flow conditions close to channels with turbulent flow, and provides the fourth way to

identify turbulent flow.

2.2.5 Cave streams

6
 The fifth way to recognize turbulent flow is where cave streams have been mapped and velocity

has been measured. This allows the Reynolds number to be determined using Equation 1. The large

diameters of channels with cave streams (typically 0.5 - 5 m) and the rapid velocities in these streams

(usually >1000 m/d) mean that flow is almost always turbulent (Fig. 1).

2.2.6 Suspended sediment in springs

The presence of high turbidity at springs provides a sixth indication of turbulent flow. Clastic

sediments become mobilized in both surface streams and in channels in carbonate aquifers when

velocities exceed ~0.2 m/s, although clay and silt may be transported at much lower velocities (Sundborg,

1956). Turbulent flow is a function not only of velocity but also of channel diameter (Equation 1 and Fig.

1), and large channels may have turbulent flow but have velocities that are too low to entrain sediment

(Fig. 1). Consequently, a spring may have low turbidity yet still have turbulent flow.

3. Examples of where turbulent flow can be identified

3.1 Determination of flow regime from tracer test data and spring discharge

Results from 229 tracer tests between sinking streams and springs in 13 countries were compiled

from the literature from studies where both spring discharge and tracer velocity were measured (see Table

S1 in the Supplementary Material). Studies were chosen to represent a wide range of conditions, from

short tracer tests to small springs to traces over tens of kilometres to some very large springs. It is

assumed that flow is through a circular channel with a tortuosity of 1.5, which is the mean measured

tortuosity from a sample of 85 flow paths in caves (Worthington, 2015a). In the case of 95 of the tracer

tests, part or all of the flow path is through a known cave. In all these caves, flow is through a single

channel, although in some cases there is distributary branching to more than one channel close to the

springs. The calculated Reynolds numbers apply to the channel in the downgradient part of the flow path.

7
There is substantially less flow in upgradient parts of the flow path, associated with smaller channels, so

that Reynolds numbers would be lower. However, discharge at a major sinkpoint was >50% of spring

discharge in 64 of the 229 cases, indicating that the Reynolds number in the upgradient part of the flow

path are in many cases somewhat less but still within the same order of magnitude as values in the

downstream part of the flow path.

Results show a strong correlation between the Reynolds number and spring discharge (Fig. 3),

with regression giving

(8)

In all cases the Reynolds number indicates flow in the turbulent regime. The regression suggests that flow

in the channel upgradient of a spring in a carbonate aquifer is likely to be turbulent where spring

discharge is >1 L/s, assuming that flow to the spring is through a single channel. In many cases, discharge

at the tracer input location was >1 L/s, so that flow was probably turbulent along the whole length of the

traced flow path.

--- Approximate location of Figure 3 ---

3. 2 Edwards aquifer, Texas, USA

Barton Springs are the principal springs that drain a 390 km2 area of the Edwards Aquifer to the

south of Austin, Texas (USA), and have a mean discharge of 1.8 m3/s (BSEACD, 2003; Scanlon et al.,

2003; Smith et al., 2005; Figs. 4 and 5). In addition, there is a contributing zone of 684 km2, and flow

from this area recharges the aquifer by several losing streams. The aquifer is in lower Cretaceous

limestone and dolostone, and tracer testing from the sinking streams has shown that typical groundwater

velocities exceed 1 km/d along the major channels connecting the sinks to Barton Springs (BSEACD,

2003; Smith et al., 2005). Head data from wells show that there is a trough in the potentiometric surface

to the south of Barton Springs, with groundwater flow paths converging on this trough and then flowing

north to the springs (Scanlon et al., 2003). Such troughs are common in carbonate aquifers, and represent

8
areas where focused flow has resulted in enhanced dissolution along preferential flow paths and hence

higher permeability (Worthington, 2009). Flow along this trough has been simulated by lines of high-

permeability cells in numerical models of the aquifer (Slade et al., 1985; Scanlon et al., 2003), and Sun et

al. (2005) used MODFLOW-DCM so that turbulent flow in channels could be simulated. The rapid flow

from sinking streams to Barton Springs results in highly variable water quality at Barton Springs (Mahler

and Lynch, 1999; Mahler and Bourgeais, 2013).

--- Approximate location of Figures 4 and 5 ---

Well YD-58-50-216 is 4 km upgradient from Barton Springs, and is close to the groundwater

trough that indicates high permeability in the aquifer (Fig. 4). The relationship between water level in the

well and spring discharge is non-linear (Fig. 6a). Assuming that flow in the main channel is turbulent, and

that flow between the well and the channel is laminar, regression using a second-order polynomial allows

the two components of head loss to be separated. This indicates that most of the head loss is due to

turbulent flow in the main channel (Fig. 6b). Turbulent flow in the main channel is confirmed by the

results from eight tracer tests where the flow is inferred to have passed close to the well along the main

channel leading to Barton Springs (Table 1 and Fig. 4). The channel cross-section and Reynolds number

were calculated using Equations 5 and 1, respectively. The same assumptions were made as for Table S1,

that flow is through a circular channel with a tortuosity of 1.5. Calculated Reynolds numbers in Table 1

are derived from discharges at Barton Springs and so represent conditions in the main channel close to the

springs, and assume that the spring discharge is from a single channel.

--- Approximate location of Figure 6 ---

Head differences between high flow and low flow represent a second indication of turbulent flow

(Section 2.2.4). The area with the largest head difference between high and low flows is aligned along the

major potentiometric surface trough in the aquifer (Figs. 4 and 5). This suggests that one or more

channels with turbulent flow are aligned along the trough that leads to Barton Springs. Turbulent flow is

also indicated by the combination of spring discharge and tracer velocity data (Table 1). Wells on either

side of the channel have a more muted head response, and this is likely to result from laminar flow

9
between the well and the channel (Fig. 2d). Head differences in the Barton Springs basin gradually drop

off away the main channel over a distance of several kilometres (Fig. 5). Very little of the area with a

seasonal head change >15 m is within the unconfined zone of the aquifer (Fig. 5). This is expected,

because an increase in flow in the unconfined zone will be buffered to some extent by storage in the

unsaturated matrix as head rises, so head changes will be less than in the confined zone.

Turbulent flow is also indicated by the presence of turbidity and suspended sediment in springs

(Section 2.2.6). Mahler and Lynch (1999) showed that there is a substantial increase in both discharge and

suspended sediment at Barton Springs following major rain events, with suspended sediment maxima

exceeding 20 mg/L.

3.3 Havant Springs, Hampshire, UK

Cretaceous Chalk outcrops over extensive areas in eastern and southern England, and accounts

for more than half the groundwater utilized in the country (Allen et al., 1997). The hydraulic conductivity

of the matrix (10-9 - 10-8 m/s) is much lower than that of the fractures (10-5 to 10-3 m/s), so the aquifer has

dual porosity characteristics, with almost all transport taking place in the network of interconnected

fractures and channels (Price et al., 1993).

A series of 28 springs at Havant in Hampshire have a mean discharge of 1.1 m3/s. The springs

drain an area of about 94 km2 of the unconfined Cretaceous Chalk outcrop of the South Downs. To the

south of the recharge area, the aquifer is confined below Paleogene sediments in the Chichester Syncline

(Fig. 7). There are no permanent surface streams on the Chalk outcrop, though there are a series of dry

valleys, some of which are subject to groundwater flooding when the water table rises to the surface in

high-flow periods. Head data from wells in low-flow conditions show that there are three troughs in the

potentiometric surface, with the troughs converging on the springs (Fig. 7). A tracer test from a doline to

the springs gave a velocity of 2.21 km/d over a distance of 5.75 km (Atkinson and Smith, 1974).

--- Approximate location of Figure 7 ---

10
 Correlation between spring discharge and heads in a well at Idsworth, 5.5 km upgradient from the

springs, shows a linear trend in low-flow conditions (Fig. 8). At higher water levels, the trend becomes

non-linear. This is similar to the situation in Figure 2f, and suggests that the well is not directly on the

main channels that have turbulent flow. It has been found that the highest permeability in the Chalk is

generally along dry valleys (Price et al., 1993). Consequently, one or more channels with turbulent flow

in high-flow conditions are likely to underlie the major dry valley that is about 900 m east of Idsworth

well (Fig. 7). The break in slope in the discharge-head plot suggests that flow in the channels from near

Idsworth Well to the springs is laminar in low-flow conditions, but becomes turbulent when discharge

exceeds ~0.85 m3/s (Fig. 8).

--- Approximate location of Figure 8 ---

Channel diameter can be calculated by rearranging Eq. 1 to

The velocity used in the calculation is from the tracer test described above, which took place when the

water level at Idsworth well was about 9.6 m above Havant Springs (Atkinson and Smith, 1974). This

corresponds approximately with the inferred transition from laminar to turbulent flow (Fig. 8). Assuming

that the groundwater temperature is 10° C, that the critical Reynolds number is 200-600 (see Section 2),

and that channel tortuosity is 1.5 (Worthington, 2015a), then calculated channel diameters are in the range

from 7 mm to 21 mm. These channel apertures are similar to those seen in downhole images in the Chalk

(Price et al., 1982; Schürch and Buckley, 2002; Maurice et al., 2012), and thus lend credence to the

calculations.

A well on Broadhalfpenny Down is also situated close to a groundwater trough (Fig. 7), but

provides a contrasting response. Correlation of weekly averages of water levels at the well with discharge

at the springs also shows a nonlinear response (Fig. 9). There were two anomalous data points during the

two weeks when levels rose >10 m, which may be due to a hysteresis effect during rapid water level

changes. The remainder of the data fit two markedly different trends, with the slope of best-fit lines being

11
much greater at high flow than at low flow (Fig. 9). The change in slope at the Broadhalfpenny Down

well (Fig. 9) is greater than at the Idsworth well (Fig 8).

--- Approximate location of Figure 9 ---

The simplest explanation for the changes in slope of the head-discharge plots at the Idsworth and

Broadhalfpenny Down wells is that they represent a change in the aquifer between the respective wells

and Havant Springs from laminar flow to turbulent flow. The transition at Idsworth occurs at a spring

discharge of 0.85 m3/s and at Broadhalfpenny Down it occurs at a spring discharge of 1.63 m3/s. This

suggests that transition to turbulent flow occurs at lower discharges in the downgradient parts of the basin

where channels are expected to be larger than in upgradient parts of the basin. However, the

potentiometric surface troughs associated with the two wells are largely independent (Fig. 7), and an

alternative explanation is that the eastern trough, which passes close to Idsworth well, has larger channels

than the central trough that passes close to Broadhalfpenny Down well. This would explain the transition

to turbulent flow at lower discharges in the eastern trough.

There is an increase in the turbidity of Havant Springs following heavy rain (Atkinson and Smith,

1974), and this provides further evidence for turbulent flow.

3.4 Turnhole Spring, Kentucky, USA

There have been many detailed studies of the Mississippian limestone aquifer in the Mammoth

Cave area, Kentucky (USA), where Turnhole Spring is one of the largest springs (Figure 10). A

combination of tracer testing, water level measurements in wells, and mapping of cave streams enabled

the major flow paths in the aquifer to be delineated (Quinlan and Ewers, 1989). It was found that all

major cave streams coincide with troughs in the potentiometric surface. Conversely, the troughs show

where there is preferential flow and probably large channels.

The recharge to the limestone area of central Kentucky was determined from the difference in

discharge at two locations along the Green River to be 19.5 L/s/km2 (Hess and White, 1989). Equations 7

12
and 8 were then applied to calculate the discharge and Reynolds numbers, respectively, along the main

flow paths (Fig. 10). Results show that there are two channels with a Reynolds number greater than

300,000, five channels with values between 100,000 and 300,000, and eleven channels with values

between 30,000 and 100,000. There are also likely to be large numbers of channels with smaller Reynolds

numbers, but still with turbulent flow. However, only a few of these are shown on the map because the

available head data is insufficiently detailed to enable the identification of minor troughs in the

potentiometric surface.

--- Approximate location of Figure 10 ---

3.5 Cave streams

The challenges of cave exploration mean that long cave streams have rarely been mapped. In most cases it

is only the segments that are at or above the water table and have an air space that have been explored,

because scuba diving in caves is arduous and dangerous. Logsdon River in Mammoth Cave is an example

of a long cave stream, where the 9.75 km section at the water table has been mapped (Fig 10). Flow

characteristics have been measured by Raeisi et al. (2007) and the channel incorporated in a MODFLOW

model of the aquifer (Worthington, 2009). Occasionally, scuba divers have mapped long underwater

caves, such as the cave system that discharges at Wakulla Springs, Florida, USA. The major passages

have been incorporated in a MODFLOW CFP model of the aquifer, with tracer testing giving additional

information on major flow paths (Loper et al., 2005; Gallegos et al., 2013). Hydraulic characterization of

the major channels feeding a large spring in Switzerland Cave was facilitated by cave mapping, combined

with head and discharge measurements (Jeannin, 2001). The case studies described above illustrate how

data collected in caves has enhanced understanding of flow characteristics in the major channels that are

often associated with large springs in carbonate aquifers.

3.6 Suspended sediment in springs

13
The presence of high turbidity or suspended sediment is common in carbonate springs, and has been

noted above at Barton Springs (Section 3.2) and at Havant Springs (Section 3.3). Suspended sediment

concentrations increase rapidly as a function of discharge, and during very high flows can reach 1000

mg/L (Newson, 1971; Herman et al., 2008). Turbidity can be measured continuously at springs, and is a

useful proxy for pathogenic bacteria, which is important for springs used as water supplies (Pronk et al.,

2007). However, as noted in Section 2.2, large channels may have turbulent flow but have velocities too

low to entrain sediment. Furthermore, there may be little particulate sediment in springs in pure

carbonates that have only percolation recharge. Consequently, low suspended sediment concentrations in

springs do not necessarily imply laminar flow.

4 Practical applications

4.1 Groundwater flooding

Flooding that occurs as a result of a rise in the water table to the ground surface is termed

groundwater flooding. It has attracted particular attention on the outcrop of the Cretaceous chalk aquifer

in southern England and northern France following extensive flooding in the winter of 2000-2001 (Habets

et al., 2010; Hughes et al., 2011), and also on the outcrop of the Carboniferous limestone in Ireland,

following extensive flooding in 2009 (Naughton et al., 2012, 2017).

In the groundwater catchment for Havant Springs, there is an area near Idsworth well that

experiences groundwater flooding in high-flow conditions (Fig. 7). The relationship between spring

discharge and heads at Idsworth well is linear at low flows (Fig. 8). If this linear relationship extended to

the highest flows then the head would be 22.1 m above spring elevation at the maximum measured

discharge of 1.956 m3/s. This relationship is shown by the dashed line in Figure 8. Instead, the head-

discharge relationship becomes non-linear at high flows and the maximum head is 31.0 m above spring

14
elevation. The increase in head of 8.9 m above the calculated linear value is explained by flow becoming

turbulent in the main channels in the aquifer, as described in Section 3.1.2, and provides an explanation

for the groundwater flooding in the Idsworth area.

Robins and Finch (2012) summarized the current interpretation of the processes that cause

groundwater floods, but then asked whether some additional processes might be at work. Turbulent flow

is one such process that does not appear to have been previously considered as an explanation for

groundwater flooding. In high-permeability aquifers dominated by fracture flow, of which carbonates are

the prime example, turbulent flow provides a straightforward and logical explanation for the large rises in

water levels at high flow that can result in groundwater flooding.

4.2 Groundwater models

Turbulent flow has been implemented in MODFLOW in two distinct ways (Shoemaker et al.,

2008a). Mode 1 of Conduit Flow Process (CFPM1) adds a series of linked cylindrical pipe elements that

are embedded within the porous medium model (e.g., Gallegos et al 2013). Mode 2 of Conduit Flow

Process (CFPM2) represents turbulent flow within a layer in the model (e.g., Shoemaker et al 2008b).

These two modes represent two contrasting ways in which turbulent flow occurs in carbonate aquifers.

The Mississippian limestone aquifer at Mammoth Cave represents the situation where single large

channels are associated with troughs in the potentiometric surface. This provides optimum conditions for

turbulent flow to occur, and as a result it is widespread in the Mammoth Cave aquifer (Fig. 10). This

could be incorporated in MODFLOW using CFPM1.

The Cretaceous Chalk aquifer at Havant (Fig. 7) represents a rather different situation, where

evidence from wells suggests that multiple channels of more modest dimensions are present (Price et al.,

1982; Schürch and Buckley, 2002, Maurice et al., 2012). The smaller size of the channels in the latter

case results in turbulent flow being less common in the Chalk aquifer than in the Mammoth Cave aquifer.

Nevertheless, head-discharge relationships show that turbulent flow is present at high flows in some areas

15
(Figs. 8 and 9), and may represent a major cause of groundwater flooding. Modelling of flow has shown

that transmissivity in the Chalk commonly varies over more than two orders of magnitude, with higher

values being found beneath valleys, close to large springs, and along the flow paths between losing or

sinking streams and springs (Rushton et al., 1989; Soley et al., 2012). These high-transmissivity areas

have the highest probability of turbulent flow. Televiewer images, flowmeter, and conductivity profiling

have shown that channels tend to be associated with prominent horizons such as the Chalk Rock and

Melbourn Rock, where they appear to form anastomosing preferential flow networks (Schürch and

Buckley, 2002; Maurice et al. 2012; Soley et al., 2012). Such flow could be incorporated in MODFLOW

using CFPM2.

5. Discussion

5.1 Head-discharge correlations

One of the main ways to identify turbulent flow is from plots of the head in wells against spring

discharge (Fig. 2). However, the methodology outlined in Section 2.2.2 makes a series of simplifying

assumptions. One is that the aquifer base at the well is below spring elevation, so that if flow at the spring

dropped to zero then the head at the well would be the same as the head at the spring and so the hydraulic

gradient would be zero. This is implicit in Figure 2, and the data in the Edwards aquifer (Fig. 6) and

Chalk aquifer (Fig. 8) provide close approximations to this situation. However, the low-flow heads at the

Broadhalfpenny Down well (Fig. 9) extrapolate to a zero-flow head that is 31.3 m above sea level, or

about 25 m above spring elevation. The well is in the Upper Chalk, the base of which is about 100 m

below sea level at this location (British Geological Survey, 1998). Hence the effective base of the aquifer

is more that 100 m above the base of the formation. This difference occurs because fracture aperture and

fracture frequency decrease with depth in the Chalk, and the effective aquifer is generally the upper 50-60

m of the saturated Chalk (Price et al., 1993). The measured water levels vary from 43 m to 84 m above

sea level (Fig. 9), giving an effective aquifer thickness that ranges from 10 m at low flow to 51 m at high

16
flow conditions. Thus the effective aquifer thickness at Broadhalfpenny Down accords with the general

situation in the Chalk.

A second simplifying assumption in Figures 2e and 2f is that there is an inflection point where

flow in the channel becomes turbulent along its whole length. That seems to be the case for the situation

shown in Figure 8, where the transition is at a spring discharge of about 0.85 m3/s. However, in other

cases there might be a progressive change to turbulent flow along the channel; in this case a plot of the

data would show more scatter than in Figure 8.

A third simplifying assumption is that head changes in the aquifer are synchronous. However,

sometimes head changes in the matrix lag head changes in channels (Quinlan and Ewers, 1989; Sauter,

1992), and this would increase the scatter in a head-discharge plot.

The simplest initial way to evaluate the correlation between spring discharge and head in wells is

to plot them as power functions. The exponent for laminar flow is 1 (Fig. 2a) and the exponent for

turbulent flow in 2 (Fig. 2b). An exponent between 1 and 2 indicates the presence of both laminar and

turbulent flow. For instance, the power function equation for the data in Figure 6a is

(10)

The exponent is close to 2 and so indicates that head loss is predominantly due to turbulent flow, and this

is confirmed by splitting the head loss into laminar and turbulent flow components (Fig. 6b). Figure 8

provides a contrasting situation. The power function exponent is 1.38 for the whole data set, indicating

that head loss is predominantly due to laminar flow, which is confirmed in Figure 8.

5.2 Variation in Reynolds numbers in channels

The Reynolds numbers shown for the Turnhole Spring basin (Fig. 10) represent average values.

There are local variations along cave streams, which may be associated with changes in cross-section,

hydraulic gradient, or whether the channel has an airspace or is below the water table (Lauber et al.,

2014). Furthermore, there may be substantial transient changes due to changes in discharge. For instance,

17
tracer velocity along a 6 km length of one of the larger channels in Figure 10 showed that velocity varied

from 200 m/d to 11 km/d (Quinlan and Ewers, 1988). Order of magnitude variation in Reynolds numbers

with discharge are also seen the Edwards aquifer (Table 1). Table S1 in the Supplementary Data also

provides a number of instances of the variation in Reynolds number with discharge, with 13 traces

between St Cuthbert’s Swallet and Wookey Hole (#31-43) and 13 traces between Lurgrotte and

Hammerbach Spring (#120-132) giving the best-documented examples.

6. Conclusions

We have presented six methods that may be used to assess the presence of turbulent flow.

However, in many carbonate aquifers there is a paucity of head data and there are no accessible cave

streams, limiting data collection to springs. Consequently, spring discharge is a key parameter for

evaluating the presence of turbulent flow. If discharge from an aquifer is by seepage at a seepage face

then it is probable that there will be no turbulent flow in the aquifer under natural gradient conditions.

Conversely, if an aquifer discharges via springs with flows > 1 L/s then turbulent flow close to the spring

is likely if flow is through a single channel (Fig. 9). However, flow may be through multiple channels, in

which case the transition to turbulent flow will occur at much larger spring discharges than 1 L/s (Section

3.3). Measurement of turbidity at springs is also a useful indicator of turbulent flow.

The probability of turbulent flow increases in a downgradient direction, and is most likely to

occur close to springs (Fig. 10). Its probability also increases as a function of spring discharge and of

permeability. Consequently, gauging of springs and identification of areas in an aquifer with high

permeability provides an indication of the distribution of turbulent flow in aquifers. Turbulent flow is

associated with much larger rises in head than laminar flow, and may be a major factor in groundwater

flooding.

Acknowledgments

18
We thank Alison Matthews of the Environment Agency for making available the data for Havant Springs

and Idsworth well, and for helpful comments on an earlier version of the manuscript. We also thank three

anonymous reviewers for their helpful suggestions. This research did not receive any specific grant from

funding agencies in the public, commercial, or not-for-profit sectors.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, in the online version, at ?????

References

Allen, D.J., Brewerton, L.J., Coleby, L.M., Gibbs, B.R., Lewis, M.A., MacDonald, A.M., Wagstaff, S.J.,

Williams, A.T., 1997. The physical properties of major aquifers in England and Wales. British

Geol Surv. Tech. Rep. WD/97/34, 312 p.

Atkinson, T.C., 1977, Diffuse flow and conduit flow in limestone terrain in the Mendip Hills, Somerset

(Great Britain). J. Hydrol., 35, 93-110.

Atkinson, T.C., Smith, D.I, 1974. Rapid groundwater flow in fissures in the Chalk - an example from

South Hampshire. Quart. J. Eng. Geol. 7, 197-205.

Birk, S., Wagner, T., Mayaud, C., 2014. Threshold behavior of karst aquifers: the example of the Lurbach

karst system (Austria). Environ. Earth Sci., 72, 1349-1356.

Bonacci, O., 2001. Analysis of the maximum discharge of karst springs. Hydrogeol. J., 9, 328-338.

British Geological Survey, 1998. Fareham, England and Wales Sheet 316, solid and drift geology,

1:50000.

BSEACD (Barton Springs/Edwards Aquifer Conservation District), 2003. Summary of groundwater dye

tracing studies (1996-2002), Barton Springs Segment of the Edwards Aquifer, Texas.

Chen, Z., Goldscheider, N., 2014, Modeling spatially and temporally varied hydraulic behavior of a

folded karst system with dominant conduit drainage at catchment scale, Hochifen–Gottesacker,

Alps. J. Hydrol., 514, 41-52.

19
Chen, Z., Qian, J., 2009, Experimental study of friction factor for groundwater flow in a single rough

fracture. J. Hydrodynamics, 21, 820-825.

Covington, M.D., Luhmann, A.J., Wicks, C.M., Saar, M.O., 2012. Process length scales and longitudinal

damping in karst conduits. J.Geophys. Res.: Earth Surface, 117, F01025.

Darcy, H., 1856. Les fontaines publiques de la ville de Dijon. Paris, Victor Dalmont.

Day, J.B.W., 1964. Infiltration into a groundwater catchment and the derivation of evaporation. Water

Supply Papers of the Geol. Surv. of G.B. Res. Rep. No. 2.

De Rooij, R., Perrochet, P., Graham, W., 2013. From rainfall to spring discharge: coupling conduit flow,

subsurface matrix flow and surface flow in a karst system using a discrete-continuum model.

Adv. Water Resour., 61, 9-41.

Environment Agency, 2014. Discharge data for Havant springs and for Idsworth well.

Environment Agency, 2017. Flood warning and alert areas. http://apps.environment-

agency.gov.uk/wiyby/37835.aspx, accessed 7 February 2017.

Ford, D.C., Williams, P.W., 2007. Karst hydrogeology and geomorphology. Chichester, England, Wiley,

562 p.

Gale, S.J., 1984. The hydraulics of conduit flow in carbonate aquifers. J. Hydrol., 70, 309-327.

Gallegos, J.J., Hu, B.X., Davis, H., 2013. Simulating flow in karst aquifers at laboratory and sub-regional

scales using MODFLOW-CFP. Hydrogeol. J., 21, 1749-1760.

Gleeson, T., Smith, L., Moosdorf, N., Hartman, J., Dürr, Manning, A.H., van Beek, P.H., and Jellinek,

A.M., 2011. Mapping permeability over the surface of the Earth. Geophys. Res. Lett., 46,

L02401, doi:10.1029/2010GL045565.

GSI, 2017. GSI modeling workshop: beyond MODFLOW-USG. http://www.gsi-

net.com/en/?option=com_k2&view=item&layout=item&id=466&Itemid=1127, accessed 13

March 2017.

Habets, F., Gascoin, S., Korkmaz, S., Thiéry, D., Zribi, M., Amraoui, N., Carli, M., Leblois, E., Leboux,

E., Martin, E., Noilhan, J., Ottlé, C., Viennot, P., 2010. Multi-model comparison of a major flood

20
 in the groundwater-fed basin of the Somme River (France). Hydrol. Earth System Sci., 14, 99-

117.

Herman, E.K., Toran, L., White, W.B., 2008. Threshold events in spring discharge: evidence from

sediment and continuous water level measurement. J. Hydrol., 351, 98-106.

Hess, J.W., White, W.B., 1989. Water budget and physical hydrology. In: Karst hydrology: concepts from

the Mammoth Cave area, (Eds. W.B. White and E.L. White), Van Nostrand Reinhold, New York,

p. 105-126.

Hubbert, M.K., 1940. The theory of groundwater motion. J. Geol. 48, 785-944.

Hubinger, B., Birk, S., Hergarten, S., 2016. A new equation solver for modeling turbulent flow in coupled

matrix-conduit flow models. Groundwater, 54, 596-602.

Hughes, A.G., Vounaki, T., Peach, D.W., Ireson, A.M., Jackson, C.R., Butler, A.P., Bloomfield, J.P.,

Finch, J., Wheater, H.S., 2011. Flood risk from groundwater: examples from a Chalk catchment

in southern England. J. Flood Risk Management, 4, 143-155.

Hunt, B.B., Smith, B.A., Beery, J., 2007. Potentiometric maps for low to high flow conditions, Barton

Springs segment of the Edwards Aquifer, central Texas. BSEACD Report of Investigations 2007-

1201. Barton Springs/Edwards Aquifer Conservation District, Austin, Texas.

Jeannin, P.Y., 2001. Modeling flow in phreatic and epiphreatic karst conduits in the Hölloch Cave

(Muotatal, Switzerland). Water Resour. Res., 37, 191-200.

Lauber, U., Ufrecht,W., Goldscheider, N., 2014. Spatially resolved information on karst conduit flow

from in-cave dye tracing, Hydrol. Earth System Sci., 18, 435-445.

Lee, L.J.E., Lawrence, D.S.L, Price, M., 2006. Analysis of water-level response to rainfall and

implications for recharge pathways in the Chalk aquifer, SE England. J. Hydrol., 330, 604-620.

Liedl, R., Sauter, M., Hückinghaus, D., Clemens, T., Teutsch, G., 2003. Simulation of the development of

karst aquifers using a coupled continuum pipe flow model. Water Resour. Res., 39, 1057,

doi:10.1029/2001WR001206

21
Loper, D.E., Werner, C.L., Chicken, E., Davies, G., Kinkaid, T., 2005. Carbonate coastal aquifer

sensitivity to tides. Eos, 86, 353-357.

Loper, D.E., Chicken, E., 2011. A leaky-conduit model of transient flow in karstic aquifers. Math.

Geosci., 43, 995-1009.

Mahler, B. J., Bourgeais, R., 2013. Dissolved oxygen fluctuations in karst spring flow and implications

for endemic species: Barton Springs, Edwards aquifer, Texas, USA. J. Hydrol., 505, 291-298.

Mahler, B.J., Lynch, F.L., 1999. Muddy waters: temporal variation in sediment discharging from a karst

spring. J. Hydrol., 214, 165-178.

Maurice, L.D., Atkinson, T.C., Barker, J.A., Williams, A.T., Gallagher, A.J, 2012. The nature and

distribution of flowing features in a weakly karstified porous limestone aquifer. J. Hydrol. 438-

439, 3-15.

Mayaud, C., Walker, P., Hergarten, S., Birk, S., 2015. Nonlinear Flow Process: a new package to

compute nonlinear flow in MODFLOW. Groundwater, 53, 645-650.

Naughton, O., Johnston, P.M., Gill, L.W., 2012. Groundwater flooding in Irish karst: the hydrological

characterisation of ephemeral lakes (turloughs). J. Hydrol., 470-471, 82-97.

Naughton, O., Johnston, P.M., McCormack, T., Gill, L.W., 2017. Groundwater flood risk mapping and

management: examples from a lowland karst catchment in Ireland. J. Flood Rick Management,

10, 53-64.

Newson, M.D., 1971. A model of subterranean limestone erosion in the British Isles based on hydrology.

Trans. Institute British Geographers, 54, 55-70.

Price, M., Downing, R.A., Edmunds, W.M., 1993. The Chalk as an aquifer, in: Downing, R.A., Price, M.,

Jones, G.P., (Eds.), The Hydrogeology of the Chalk of North-West Europe. Clarendon, Oxford,

pp. 33-58.

Price, M., Morris, B., Robertson, A., 1982. A study of intergranular and fissure permeability in Chalk and

Permian aquifers, using double packer injection testing. J. Hydrol., 54, 401-423.

22
Pronk, M., Goldscheider, N., Zopfi, J., 2007. Particle-size distribution as indicator for fecal bacteria

contamination of drinking water from karst springs. Env. Sci.Technol., 41, 8400-8405.

Qian, J., Zhan, H., Zhao, W., Sun, F., 2005. Experimental study of turbulent unconfined groundwater

flow in a single fracture. J. Hydrol., 311, 134-142.

Quinlan, J.F., Ewers, R.O., 1989. Subsurface drainage in the Mammoth Cave area, in White, W.B., and

White, E.L., eds., Karst hydrology: concepts from the Mammoth Cave area: New York, Van

Nostrand Reinhold, p. 65-103.

Raeisi, E., Groves, C., Meiman, J., 2007. Effects of partial and full pipe flow on hydrochemographs of

Logsdon river, Mammoth Cave, Kentucky, USA. J. Hydrol., 337, 1-10.

Reimann, T., Rehrl, C., Shoemaker, W.B., Geyer, T., Birk, S., 2011. The significance of turbulent flow

representation in single-continuum models. Water Resour. Res., 47, W09503,

doi:10.1029/2010WR010133.

Robins, N.S., and Finch, J.W., 2012, Groundwater flood or groundwater-induced flood. Quart. J. Eng.

Geol. Hydrogeol., 45, 119-122.

Rushton, K. R., Connorton, B. J., Tomlinson, L. M., 1989. Estimation of the groundwater resources of the

Berkshire Downs supported by mathematical modelling. Quart. J. Eng. Geol., 22, 329–241.

Sauter, M., 1992, Quantification and forecasting of regional groundwater flow and transport in a karst

aquifer (Gallusquelle, Malm, S.W. Germany), Tübinger Geowissenschaftliche Arbeiten, Reihe C,

No. 13, 150 p.

Scanlon, B.R., Mace, R.E., Barrett, M.E., Smith, B., 2003. Can we simulate flow in a karst system using

equivalent porous media models? Case study, Barton Springs Edwards Aquifer, USA. J. Hydrol.,

276, 137-158.

Schürch, M., Buckley, D., 2002. Integrating geophysical and hydrochemical borehole-log measurements

to characterize the Chalk aquifer, Berkshire, United Kingdom. Hydrogeol. J., 10, 610-627.

Sepúlveda, N., 2009. Analysis of methods to estimate spring flows in a karst aquifer. Ground Water, 47,

337-349.

23
Shoemaker, W.B., Kuniansky, E.L., Birk, S., Bauer, S., Swain, E.D., 2008a. Documentation of a conduit

flow process (CFP) for MODFLOW–2005: U.S. Geol. Surv. Techniques and Methods 6-A24.

Reston, Virginia.

Shoemaker, W.B., Cunningham, K.J., Kuniansky, E.L., Dixon, J., 2008b. Effects of turbulence on

hydraulic heads and parameter sensitivities in preferential groundwater flow layers, Water

Resour. Res. 44, DOI: 10.1029/2007WR006601.

Slade, R.M., Jr., Ruiz, L.M., and Slagle, D.L., 1985. Simulation of the flow system of Barton Springs and

associated Edwards aquifer in the Austin area, Texas. U.S. Geol. Surv. Water-Resources

Investigations Report 85-4299, 49 p.

Soley, R.W.N, Power, T., Mortimore, R.N., Shaw, P., Dottridge, J., Bryan, G, Colley, I., 2012. Modelling

the hydrogeology and managed aquifer system of the Chalk across southern England. London,

Geol. Soc. Spec. Publ. 234, 129-154.

Smith, D.I., Atkinson, T.C., Drew, D.P., 1976. The hydrology of limestone terrains. In: The science of

speleology, Eds: T.D. Ford and C.H.D. Cullingford. Academic Press, London, p. 179-212.

Smith, B., Hunt, B., Schindel, G., 2005. Groundwater flow in the Edwards Aquifer: comparison of

groundwater modeling and dye trace results. In: Beck, B.F. (Ed.) Sinkholes and the Engineering

and Environmental Impacts of Karst (2005). Amer. Soc. Civil Engineers, Geotech. Spec. Publ.,

No. 144, 131-141.

Street, R.L., Watters, G.Z., Vennard, J.K., 1996. Elementary fluid mechanics, Wiley, New York, 757 p.

Sun, A., Painter, S., Green, R., 2005. Modeling Barton Springs Segment of the Edwards Aquifer using

MODFLOW-DCM. In: Beck, B.F. (Ed.) Sinkholes and the Engineering and Environmental

Impacts of Karst (2005). Amer. Soc. Civil Engineers, Geotech. Spec. Publ., No. 144, 163-177.

Sundborg, Å., 1956, The River Klarälven: a study of fluvial processes. Geografiska Annaler, 125-237.

White, W.B., 1988. Geomorphology and hydrology of karst terrains. New York, Oxford University Press,

464 p.

24
Worthington, S.R.H., 2009. Diagnostic hydrogeologic characteristics of a karst aquifer (Kentucky, USA).

Hydrogeol. J.,17, 1665-1678.

Worthington, S.R.H., 2015a. Characteristics of channel networks in unconfined carbonate aquifers. Geol.

Soc. Amer. Bull. 125, 759-769.

Worthington, S.R.H., 2015b. Diagnostic tests for conceptualizing transport in bedrock aquifers. J.

Hydrol., 529, 365-372.

25
Figure captions

Figure 1. Groundwater velocity as a function of channel diameter, showing the range of laminar
and turbulent flow regimes, assuming transition from laminar to turbulent flow at Reynolds
numbers (R) between 100 and 1000 (modified after Sundborg, 1956).
Figure 2. Head (h) as a function of spring discharge (Q) for different well locations and different
combinations of laminar and turbulent flow.
Figure 3. Reynolds numbers for 229 channels between sinking streams and springs (see Table 1
in Supplementary data for details).

Figure 4. Low-flow heads and seasonal changes in heads in the aquifer that discharges at Barton
Springs, Texas, USA (modified from BSEACD, 2003, and Hunt et al., 2007)

Figure 5. West to east section in the Barton Springs catchment, along the line A-B shown on
Figure 3 (modified from Hunt et al., 2007)

Figure 6. Correlation between the head in well YD-58-50-216 and discharge at Barton Springs, 4
km downgradient from the well, showing data points (left) and split into two components (right).
Based on data from the U.S. Geological Survey and Texas Water Development Board

Figure 7. Catchment for Havant Springs, Hampshire, England, showing three major troughs in
the potentiometric surface (modified from Day, 1964; Atkinson and Smith, 1974; Environment
Agency, 2017)
Figure 8. Correlation between discharge at Havant Springs and water levels at the Idsworth well
(data from Environment Agency, 2014)
Figure 9. Correlation between average weekly discharge at Havant Springs and head in the well
at Broadhalfpenny Down between January 2000 and August 2002 (head data from Lee et al.,
2006; discharge data from Environment Agency, 2014)
Figure 10. Reynolds numbers for the major channels in the Turnhole Spring groundwater basin,
Kentucky, USA (adapted from potentiometric surface map by Quinlan and Ewers, 1989).

26
Table 1 - Details of eight tracer tests from sinking streams to Barton Springs (tracer and discharge
data from BSEACD, 2003).

Discharge at Distance to Groundwater

Tracer Reynolds
Tracer injection location Barton Barton velocity
number 3 number
Springs (m /s) Springs (m) (m/d)*
1 Marbridge Sink 0.74 15,800 370 105,000
2 Barber Falls 0.74 24,000 1500 213,000
3 Crooked Oak Cave 0.79 29,000 1260 202,000
4 Dahlstrom Cave 1.05 21,400 1020 209,000
5 Midnight Cave 2.35 13,400 1670 401,000
6 Brodie Sink 2.35 12,100 6030 762,000
7 Cripple Crawfish Cave 2.78 26,900 8970 1,010,000
8 Antioch Cave 2.80 22,300 2780 565,000
* Groundwater velocity is based on tracer arrival time

27
Highlights

 Six methods are described for identifying turbulent flow in aquifers

 Spring discharge is a key parameter for evaluating turbulent flow
 Turbulent flow occurs most frequently in carbonate aquifers
 Turbulent flow appears to be an important factor in groundwater flooding

28
 Green River Green River
Turnhole Spring Canada
L
USA
L study area
Kentucky
L
134 Mexico

Stream passage in cave (long segment)

Stream passage in cave (short segment)
14 Traced flow path
6
Stream sinkpoint
158 Spring
Water table contours (m)
1 Surface stream
17 83
1 5 Groundwater divide
19 207
L Logsdon River (cave passage)

219 Reynolds number

>300,000
100,000 - 300,000
30,000 - 100,000
0 5 km
10,000 - 30,000
 90
> 10 m rise
Q<1.63 m3/s h=0.0819Q - 79.9
Q>1.63 m3/s r2=0.76
80
Head at well (m above sea level)

70

60

h=0.0128Q + 31.3
r2=0.71
50

40

30
0.5 1 1.5 2
Discharge at Havant Springs (m3/s)
 35

30 h=4.53Q2+7.47Q
Head at well (m above spring elevation)

r2=0.93

25

20

15
h=11.3Q
r2=0.64
10 extrapolation of
linear trend
5

0
0 0.5 1 1.5 2
Discharge at springs (m3/s)
 0 2 km

122
107

91

50.95°N 76
61
BW
46
38
30
32
35

23

IW
15

Chic
hest
er S
50.85°N yncli
ne

area
of sp
rings
1.05°W 0.95°W

Low-permeability Paleogene strata

Area near Idsworth with groundwater flooding
Main flow lines in potentiometric surface troughs
Heads in the Chalk aquifer (m above sea level)
Groundwater divide
Dry valley
Tracer injection location
91 Groundwater elevation (m above sea level)
BW Broadhalfpenny Down well
IW Idsworth well
 25 25
Head (m above spring elevation)

Head (m above spring elevation)

20 20
h = 1.865Q2 + 0.574Qx
h = 1.865Q2 + 0.574Q R² = 0.93
R² = 0.93 total head loss
15 15

h = 1.865Qx2
turbulent flow
10 10 in channel

h = 0.574Q
laminar flow
5 5
off-channel

0 0
0 1 2 3 4 0 1 2 3 4
Discharge (m3/s) Discharge (m3/s)
 A B
recharge zone confined zone
Elevation (metres above sea level)

300 maximum
head difference

200

100
Edwards aquifer
high-flow heads
low-flow heads

0 0 5 km
 Well YD-58-50-216 Co
5 Tracer injection number in Table 1 lor
ad
Tracer injection location o
R.

4
Flow path traced to Barton Springs
140

13
Barton Springs
Low-flow heads (m asl)
>15 m seasonal change in head 189
>30 m seasonal change in head
Updip limit of recharge zone 165
Updip limit of confined zone
Updip limit of saline zone
Line of section AB shown in Fig. 4

0 4 km 30.2oN
5

A potentiometric
surface 1
165

poorly defined B
- few wells

4
165
30.1oN
3
21

2 8
9
18

7
3 97.9oW 97.8oW
 107

known channel
presumed channel

106
Reynolds number

105

104
R = 239,000 Q0.602
r2=0.89

103
0.001 0.01 0.1 1 10 100
Spring discharge (m3/s)
 a Laminar flow b Turbulent flow
Channel well Channel well
Head

Head
h=aQ

h=aQ2
0 0
0 Spring discharge 0 Spring discharge

c Laminar flow d Turbulent flow

Off-channel well Off-channel well
Head

Head

h=aQ
h=aQ2+bQ
0 0
0 Spring discharge 0 Spring discharge

e Hybrid regime f Hybrid regime

Channel well Off-channel well

h=aQ
Head

Head

h=aQ
h=aQ2 h=aQ2+bQ
0 0
0 Spring discharge 0 Spring discharge
 10000

lam turbulent
ina
Groundwater velocity (m/d)

1000 ro
rt
ur R=
10
bu 00
R=
len
laminar 10 t
0
100
R=
10

R=
1
10
0.1 1 10 100
Channel diameter (mm)