SPATIAL ANALYSIS

OF LAND USE/LAND COVER CHANGE

DYNAMICS USING REMOTE SENSING AND
GEOGRAPHIC INFORMATION SYTEMS:
A Case Study in the down stream and surroundings of
the Ci Tarum watershed

Asep Karsidi

A thesis submitted in fulfilment of the requirements

for the degree of Doctor of Philosophy

Department of Geographical and Environmental Studies

The University of Adelaide
South Australia

May 2004
 Abstract

This study is concerned with land use/land cover change detection, identification,

analysis and prediction using remote sensing and GIS techniques in the downstream of
'West
the Citarum watershed and its surroundings in Java, Indonesia. Supervised

Maximum Likelihood classification of PCA and NDVI transformed images are used to

classiff and identiff land use/land cover categories. A post-classification comparison

approach was used to detect land use/land cover changes, and a Markov Cellular

automata model is then used to predict possible future land use/land cover patterns in

the study area."Leafon" and "leaf off'phenomena of the broad leaf vegetation cover

have been recognised related to dry and wet season as well as rice field (planted) and

rice field (unplanted) related to growing season in the study area. Forest and plantation

area were extensive in wet season and less in dry season. Rice field þlanted) area was

large in harvesting time and less in planting time. Settlement has increased

continuously and is not influenced by season or weather. Overall' the KIA of the

classification was 0.89. Settlement and rice field are the main land use/land cover types

that have been changed and this is related to factors such as proximity to roads and to

urban and semi-urban centres. There is an indication that land use/land cover in the

study area was converted from intensive agriculture land such as rice field to

settlement, rather than from less intensive uses such as open/dry land, plantation or

forest. Discriminant analysis as well as overþ and simple linear analysis support

factors such as proximity to roads, urban and semi-trban centres, as well as slope, as

being most influential in land use/land cover change in the study area. The Markov

Cellular automata model affords a powerful descriptive and predictive model for land

11
use/land cover change and for future land use/land cover distribution in the study area,

but it needs some adjustment in order to obtain suitable results. Markov transition, as

well as suitabilit¡ maps of each land use/land cover category are created'

ll1
 Abbreviations and Glossary
BAKOSURTA}IAL The Coordinating Agency for National Survey and
Mapping
BAPPEDA Regional Development Planning Boards
BAPPENAS The National Development Planning Agency
BPN The National Land Agency
BPPT The Agency for Assessment and Application of
Technology

CA Cellular Automata
'Water
DGWRD Directorate General of Resources Development

FAO Food and Agriculture Organization

GTZ Gesellchaft fuer Technische Zusammenarbeit.
GIS Geo graphic Information Systems

IGBP International Geosphere and Biosphere Programme

ITC International Institute for Aero space
JABOTABEK Jakarta, Bogor, Tanggerang and Bekasi

Kabupaten Regency or District

Kecamatan Sub-district
KIA Kappa Index Agreement

LUCC Land use/cover Change

LUPAM Land Use Planning and Mapping
LREP Land Resources Evaluation and Planning

MCE Multi Criteria Evaluation

MOA Ministry of Agriculture
MOF Ministry of Forestry
MOLA Multi Objective Land Allocation
MPV/ Ministry of Public Work
MREP Marine Resources Evaluation and Planning

NDVI Normalized Difference Vegetation Index

PCA Principal Co mponent Analysis

RCSA The Research Centre of Soil and Agro-climate

RePPProt Regional Physical Planning Program for Transmigration

SARCS Southeast Asia Research Centre for STAR

lV
 Gontents

ii
Abstract
iv
Abbreviations and Glo ssarY
vlll
List of Figures xi
List of Tables
xiv
Declaration xv
Acknowledgement

CIIAPTER ONE I
l.l Introduction
7
1.2 Aims and Objectives 9
1.3 The Context ofthe StudY
18
1.4 Research Framework
2t
1.5 Thesis Outline

CHAPTER TWO 22
2.1 Introduction 22
2.2 Defining Land use/land cover change 25
2.3 The dynamic of Land use/land cover change
change 30
2.4 Some aspects related to Land use/land cover
pressur 30
2. .LLanduse and Demo graphic 32
Z.4.ZLanduse conflict and Regional food security
35
2.4.3 Landuse, Industry and Manufactruing 37
2.4.4 Landuse and Agriculture 39
2.4.5 SpatiaÍdimension of Land use/land cover change
detection
z.+.0 rrre important of Land use/land cover change
42
in sPatial Planning
43
2.4.TLanduseevolutionanddemographicpressrrreinlndonesia
planning in
2.+.t yalduse/land cover change dãtection and spatial 47
Indonesia
50
2.5 Conclusion

v
IIAPTER THREE
3.1 Introduction 52
3.2 Selection of Study area 53
3.3 Socio-EconomicCharacteristics 55
3.3.1 PoPulation and demograPhY 56
3.3.2 Employment 57
3.3.3 The Land Tenure 58
3.3.4 Existing Land use 6l
3.3.5 Infrastructure 63
3.4 BiophysicalCharacteristics 65
3.4.1 The Physiography and Soil type 66
3.4.3 Climate, Irrigation and Growing season 68
3.5 The availability of spatial data 7l
3.5.1 Existing sPatial data 7l
3.5.2 Satellite imagery 74
3.6 Conclusion 75

CTIAPTER F'OUR
4.1 Introduction 77
4.2 Land use/land cover change detection techniques 78
4.2.1Digital change detection of land use/land cover 83
4.2.1.1 Pre-classification 88
4.2.1.2 Po st-Classification Comparison 96
4.2.1.3 Hybrid 98
4.2.2 Selection of the appropriate detection and identification
Techniques 99
4.3 Method 101
4.3.1 Digital change detection techniques 103
4.3.1.1 Image Pre-Processing 103
4.3.1.2 Image Enhancement 104
4.3 .2 Land use/land cover Classification 106
4.3.3 Landuse/land cover change detection r07
4.3.4 Accuracy Assessment 108
4.4 Results digital change detection and identification in the study area 108
4. 4 . I lrlrttal reco gnitio n
108
4.4.2 Image Enhancement tt2
4.4.3 Larà use/land cover Classification 11s
4.4.4Theland use/land cover change detection t2l
4.4.4.1Annual Land use/land cover change from 1989 to 1993 122
4.4.4.2land use/land cover change in 1993 to 1995 and
1995 to 1997 125
4.4.4.3 Land use/land cover change 8- year time interval
1989-1997 t26
4.4.5 Landuse/land cover change identification and re-classification t28
4.5 Discussion 131
4.6 Conclusion t34

V1
CHAPTER FIVE
5.1 Introduction 137

5.2 Driver factors of Land use/land cover Change 138

139
5.2. I Discriminant AnalYsis
5.3 Method t42
5.3.1 Overlay analysis of possible driver factors r42
5.3.2 Simple linear and Discriminant Analysis r45
5.4 Result 147
t47
5.4.1 OverlaY analYsis
5.4.11Static drivers r47
151
5.4.1.2 DYnamic drivers
158
5.4.2 Quantitative analysis ; Discriminat function analysis
166
5.5 Discussion
170
5.6 Conclusion

CIIAPTER SIX
6.1 Introduction t7l
6.2 Land use/land cover change prediction mode 172
174
6.2.1 Markov chain model
177
6.2.2 Cellular automata
6.2.3 Prediction based on Markov-Cellular automata
181
Model within GIS
6.3 Method 183
184
6.3 . 1 Markov transition probability calculation
6.3.2 Multi criteria evaluation approach to create land suitability 185
6.3.3 Prediction process based on Markov-Cellular automata model 186

6.4 The Selecting Parameters and test of the Model 188

188
6.4. I Transition ProbabilitY
190
6.4.2LandSuitability of each land use/land cover category
6.4.3 The result of simulation or prediction process
t93
6.5 The Future of Land use/land cover in the Study area 198
199
6.5.1 Land use/land cover simulation 2007 (1O-year)
207
6.5 .2 Land use/land cover simulation 20 17 (20 -yeat)
6.5.3 Analysis spatial of Land use/land cover 1989, 1997,2007
and2077 in the StudY are 214
6.6 Discussion 224
228
6.7 Conclusion

CHAPTER SEVEN

7.1 Summary and Findings 231

240
7.2 Implications
240
7.2.1 Policy
241
7.2.2 New Research
243
REFERENCES
APENDICES
v11
 List of Figures

CHAPTER ONE 7
1.1 The Study area of the Citarum Watershed
11
1.2 Province ofthe Republic of Indonesia
Indonesian Country t2
1.3 Percentage of Forest cover in the main Island of
Bekasi and
l.A Housingãnd Industrial estate development area in Bogor,
I7
Karawang
20
1.5 Research framework

CHAPTER TWO 30
2.1 The mutual interaction of land use change dynamic
37
2,.2 Housing and Industrial estate development area in JABOTABEK
38
2.3 Land use in West Java1993
2.4 Overall model spatial planning structure related to
43
Land use
44
2.5 The conc Indonesia
45
2.6 Land use ounding West Java
49
2.7 Integrated Planning System for Regional Development

CHAPTER THREE 52
3.1 Study area as part of the Citarum'Watershed
54
3.2 The irve Kabupaten and Jakarta within study area
in Jakarta and
3.3 The new rown and Industrial estate development
54
Surrounding area
55
3.4 iand rrse tõq¡ in five Kabupaten within study area and Jakarta 62
3.5 Main Land use in West Java: 1989-2000
63
3.6 Roads Networks in the studY area
66
3.7 The Relief of West Java
67
3.8 The Physiography in five Kabupaten of study area
68
3.9 Soiltype itfive Kabupatenwithin study area
69
3.10 The Rainfall Regimes in West Java
70
3.11 The monthly rainfall regimes in the study area
7l
3.12 The Citarum river and inigation in the study area

CHAPTER FOUR
analysis of
4.1 The integrated model of detection, identification and
102
Land use/land cover change
4.2 post-classification approach of land use/land cover
"o-p*irot, r07
Change detection and identification
109
4.3a TM Composite 432 t989
110
4.3b TM Composite 4321990
110
4.3c TM Composíte 4321995
111
4.3d TM Composire 4321997
v111
4.4a The Spectral features of the Principal Component 5 (PC5) tt4
The Spectral features of Band Ratio (NDVD 115
4.4b
The Final classified image of land use/land cover 1989 116
4.5a
4.sb The Final classified image of land use/land cover 1990 l17
4.5c The Final classified image of land use/land cover 1991 t17
The Final classified image of land use/land cover 1992 118
4.5d
The Final classified image of land use/land cover 1993 118
4.5e
4.sf The Final classified image of land use/land cover 1995 t19
119
4.5g The final classified image of land use/land cover 1997
4.6 The Annual land use/land cover change in the study area l2l
4.7 Land use/land cover in the study area 1989 and 1997 r29
130
4.8 Land use/land cover change in the study area

CHAPTER FIVE
5.1a The Classification of population density in the study area r52
153
5.lb The Settlement distribution inthe study area
5.2 The increasing of Settlement and decreasing Rice field within
urban and semi-wban area 155
158
5.3a
159
5.3b
159
5.3c
s.3d
160
161
5.3e
161
s.3f
5.3g
t62

CHAPTER SIX
6.1 The Neighbourhood of the (ij) cell is formed by the (Ij) cell itself
And theiight adjacent cells 182
6.2a Suitability-map for settlement without proximity to Jakarta city 191

6.2b Suitabitity -up for settlement with proximity to Jakarta city 192
6.3a prediction t; óf g-iteration,3-kernel filter, suitability map of settlement
Without proximþ to Jakarta cþ 194
6.3b Original^Classifred Inage 1997 195

6.3c preáiction 2; of 8-iteration,3-kernel filter, suitability map of settlement

with proximity to Jakarta citY 194
6.4 The rèsuh of Land use/land cover change of lg-year simulation
Based on Markov calculation from 1989-1997 (Simulation 2007) 203
6.5 The result of Land use/land cover change of lO-year simulation
Based on Markov scenario (Scenariol 2007) 203
6.6 The result of Land use/land cover change of lO-year simulation
Based on Markov scenario (Scenario2 2007) 204
6.7 Classification Image 1997 (originat) 204
6.8 l0-years changes characteristic between simulation and scenario 206
6.9 Thé result of iand use/land cover 2}-year simulation (Simulation 2017)209
6.10 The result of Land use/land cover 2}-year simulation (Scenariol 2017) 210
6.11 The result of Land use/land cover 2}-year simulation (Scenario2 2017) 210
lX
 and scenario2 212
6.12 213
6.13a 2t3
6'13b
use/land nn
rr ^-r ,,^o/lo-¡t co km of
6.14 The trend of Land 2t6
the main roads 1989 to 2017 2t7
roads 1989 and 1997
6.15a Land use/hnl.orr", within 10 km of the main 218
6.15b Land use/land within 10 km of the main roads 2007 and20l7
'orr". cover within
6.16a Graphic of Je"r"arirrg and increasing land use/lanrl 220
10 kmurban and semi urban arefls; Bekasi
andcileungsi
cover within
6.16b Graphic of à".."u.i"g and increasing land use/land 221
10 km urban and semi urban
6.16c GraPhic of decreasing 222
l0 kmurban and semi 223
semiurban centre
6.t7 Land use/land cover within 10 km of urban and

x
 List of Tables

CHAPTER ONE
I .1 Distribution of Indonesian Population and Growth Rate I 980-2000 1 0
1.2 Forest cover and Population Density in the main Island in Indonesia
Year 1996 13
1.3 The Change of Agriculture area 1981 to 1999 (change fromrice fie1d
to other land use) in Java andBali 14
1.4 Area of rice field and settlement within Kabupaten in the study area 15
1.5 Decreasing rice field and increasing settlement in V/est Java 1989-2000 16
1.6 Number ofNational lndustrial and trading companies 16
1.7 Number and area of Industrial Estate in Kabupaten Bekasi year 2000 18

CHAPTER TWO
2.1 Driving force behind land use change 27
2.2 Change in Agricultural Land between 1981-1999 in Indonesian
Provinces 28
2.3 Changes in Agricultural land in Indonesia7983-1993 55
2.4 The average rice productivity in Java and offJava (Ton/halyr) 34
2.5 Conversion from rice field to settlement and industrial area in the
main island of Indonesia 1981-1999 34
2.6 Number of large and medium manufacturing 1997-2000 35
2.7 The land use classification system in Indonesia 40
2.8 Land use/land cover classification level and their source 4t
2.9 Scale map at different level of Spatial planning application 48

CHAPTER THREE
3.1 Population and Demographic Indicator 56
3.2 Labour Force and Occupation in five kabupaten and Kodya Bogor 58
3.3 Changes in Agricultural Land and Households 1983-1993 59
3.4 The change of Agricultural land to other land use year 1981-1999 60
3.5 Existing Land use in the study area (ha) 62
3.6 Infrastructures in five Kabupaten within the study area 64
3.7 The Growing season schedule 70
3.8 Partial list of Land Resources Inventory and Mapping Agencies in
Indonesia 73
3.9 Landsat TM data 74

CHAPTER FOUR
4.I The Comparison of the Terrestrial, Air photos and Remote-
Sensing techniques 80
4.2 The various techniques to identiS and monitor land use/land cover
change 82
4.3 Comparison of digital detection techniques categorised by
Independentl¡ Simulation and Hybrid 86

x1
4.4 The type of Image enhancement and function 105
4.5 1989
P principal component analysis transformation TM 113
4.6 Land use/land cover classificationscheme 116
4.7 assessment
Summery of Kappa index of agreement (KIA) 120
4.8a The Summary major change in land use/land cover category 198911990122
4.8b The Summary major change in land use/land cover category l990ll99l 123
4.8c The Summary major change in land use/land cover category 199211993 124
4.8d The Summary major change in land use/land cover category l99lll992l24
4.9a The Summary major change in land use/land cover category 199311995125
4.9b The Summary major change in land use/land cover category 199511997 126
4.10 The Summary major change in land use/land cover category 198911997 I27
4.11 Land use/land cover in the study areayear 1989-1997 (ha,) 130

CHAPTER FIVE
5.1 The Method to create area of possible influence or driver factors 144
5.2 Variable used in the discriminant analysis r46
5.3 Summery of land use/land cover change versus slope (1989-1997) 148
5.4 Summery of land use/land cover change versus Physiography
(te9e-tee7) t49
5.5 Summery land use/land cover change versus Land system
(1989-tee7) 150
5.6 Land use/land cover and Population density Classes 152
5.7 Land use/land cover change within 10 km of urban area (ha) 154
5.8 Land use/land cover change within 10 km of semi-urban area (ha) r54
5.9 Percentage of decreasing and increasing rice field and settlement
within urban and semi-urban in the study area r54
5.10a Land use/land cover in 1 km buffer of To11 rcad/way 157
5.10b Land use/land cover in 1 km buffer of State road 157
5.10c Land use/land cover in I km bufÊer of other road 158
5.11 R2 of Linear regression between population and land use/land cover 160
5.12 CovarianceMatrices t63
5.13 Eigenvalues 163
5 .l 4a Standardized Canonical Discriminant Function Coefficients 164
5.14b Structure Matrix 164
5.15 Classifrcation function Coeffi.cients 16s

CHAPTER SD(
6.1 Land suitability factors of each category 185
6.2 The Markov transition of Land use/land cover 1989-1997 189
6.3 The weight value of suitability factors 190
6.4a The Kappa Index of Agreement (KIA) value between
original 1997 -1997 predicted I and original 1997 -1997 predicted 2 t96
6.4b The Kappa Index of Agreement (KIA) value between
original 1989 - original 1997 and original 1989- predictedl997 196
6.5 Summary of probability expected to stay as same category l0- year
Scenario 200
6.6a The Markov transition of land use/land cover for lO-years prediction
(1997-2007), based on 1989-1997 images 20r

xl1
6.6b The Markov transition of land use/land cover for lO-years
(scenariol 2007) assumed that change will low 202
6.6c The Markov transition of land use/land cover for l0-years
(scenario2 2007) assumed that change will very low 20r
6.7 The lO-years change characteristic between Simulation and Scenario 206
6.8a Markov transition of land use/land cover for 2O-year (Scenario 2017)
based on scenariol 208
6.8b Markov trarsition of land use/land cover for 2û-year (Scenario 2017)
based on scenario2 208
6.9 The 20-years change cha¡acteristic between Simulation and Scenario 2lt
6.10 Land use/land cover within 10 km of main roads tn 1989,1997 ,
Scenario2 2007 and Scem¡io2 2017 2ls

xl11
Declaration

This work contains no material which has been accepted for the award of any other

degree or diploma in any university or any other tertiary institution, and to the best of

my knowledge and belief, contains no material previously published or written by

another person, except where due reference has been made in the text.

I give consent to this copy of my thesis, when deposited in the University Library,

being available for photocopyng and loan.

Dl:.
a

SI DATE:

(Asep Karsidi)

x1v
 Acknowledgement
and Environmental
This thesis was carried out in the Department of Geographical

Studies at The University of Adelaide'

Hugo, Dr Kym Nicolson and Dr'

I am indebted to my supervisors, Professor Graeme

and encouragement' I would like

Brett Bryan, for their supervisior¡ guidance' support
patience and endless enthusiasm'
to express my sincere gratitude for their inexhaustible

Thank to Dr. Indroyono soesilo soedarman

for his initial encouragement and advice to

continue my study in Adelaide and for his

irrepressible spirit' Thanks also to associate

Professor Lesley Potter for her encouragement and enthusiasm in her role as
Financial support for this study was provided by
the
postgraduate coordinator.
scholarship Schema, and financial
University of Adelaide,s international postgraduate
was provide by Deputy of TPSA BPPT'
support for the research presented in this thesis
Enol Bamford and all staff in
I would particularly like to thank Dr. Kevin Harris, Mr'
many yeaÏs of friendship and
GISCA for their constant support provided tliloughout

I also thank all colleagues and staff at P3TISDA-BPPT and

professional interaction
for many valuable
the laboratory of remote sensing and GIS P3TISDA-BPPT
and processing'
discussions and much advice during data collecting

am very grateful to the most important

people in my life' who have
Last but not least, I
years before' To my wife' Truly' and
helped me throughout my PhD and over the many

and support you give me over the

my son, Askar Reza Karsidi, thanks for all the love

years. I couldn't have done this daunting task without

you' "Papa syei"' the most

xv
 encouraging and truly a reminder
common word coming from my son, is very much

thatlmustcompletemystudyandgainthebesteverinmylife.

xv1
 Chapter One

INTRODUCTION

1.1 lntroduction

Accelerating changes to the earth's environment are being driven by growth in human

population, the increasing level of resource consumption by human societies and by

changes in technology and socio-political organnation. Changes in land use and land

cover arc part of this transformation. Land use and land cover are closely interrelated

but are not the srrme. The term "land use" denotes the human uttltzation of the land,

while "land cover" denotes the physical state of the land. Baulies (1997) indicates that

Iand use and land cover change dynamícally due to demographic pressure, road

construction, fire, nutrient deposition, and many other factors. Land use/land cover

change may be the result of natural processes or human activities (Turner II et al.,

1994, Tucker et al., 1991, Sage, 1994). Furthermore, technological advances and

expanding population have put increasing pressure on scarce resources and have

created a variety of complex land use dilemmas that affect persons at all levels of

society (Sommers, 1981).

In the last decade, land use/land cover change has become a major issue at the global

and regional level. International attention has focussed on it because ofconcern over

issues such as global warming and climate change.

1
Many studies have been conducted regarding the land use/land cover change issue

(Turner, 199I; Turner, I994;Mayer et al., 1994; Baulis, 1997; Thomas, 1997; Heilig'

1997; Rindfuss, 1998; Koning, 1999; Mc Connell, 2000; Solecki, 2001; Bicik, 2001).

Some of those studies focus on understanding the process of land use/land cover

change as well as developing predictive models of change at the global level (Turner,

I99l;Tumef, I994;Mayer et al., I994;Baulis, 1997). There have also been numerous

studies conducted at regional or country level looking at land use/land cover change

(Thomas, 1997; Heilig, 1997; Koning, 1999; Mc Connell, 2000; solecki, 2001; Bicik,

2001, Reid, 200 1 ; Schneider, 2001 ; Pontius.Jr, 200 I ).

Land use/land cover change is also a major issue in Indonesia. Demographic pressure

with an annual rate of population growth between 1990 - 2000 at the national level of

I.49 % (Central Bureau of Statistics, 2000) and an uneven population distribution has

had an effect on the land use pattern in Indonesia. The 2000 National Census placed

Indonesia's population at206 million (Central Bureau of Statistics, 2000). Of this total

populatior¡ 58 % live on the island of Java, and rural areas accommodated 57 Yo of

Java's population (Central Bureau of Statistics, 2000). However between the 1990 and

2000 census, the population in rural areas slightly decreased while the urban
population increased at the rate of 5.4%o per year over the period of 1990'2000. There

were 55.4 million urban residents in Indonesia in 1990 and 85.4 million in 2000
(Central Bureau of Statistics, 2000). Approximately one-fourth of this urban
population is concentrated in the Jakarta, Bogor, Tanggerang and Bekasi

(JABOTABEK) extended metropolitan region (Firman, 2002). The increase of urban

population has had an impact on land use. Other land has been converted from rural to

2
urban use, from less intensive to more intensive agriculture, or from agricultural to

non-agricultural use. Nasution (1999) reported that about 20 000 ha per year of

agricultural land, has been converted into non-agricultural land mainly in the areas

adjoining large cities on Java Island (Nasutior¡ 1999).

In terms of the land use and population growth relationship, Sandy (1977) produced a

land use evolution model for Indonesia. People will start to cultivate an area at 25 m

above sea levef which has a gentle slope and is free from flooding. When population

increases, they will extend to cultivate the areas to the upper slopes and with more

population growtt¡ the area that could be conserved (such as forest in the steep slopes

and wetlands or swamp forests near the coast) will be cultivated, and as a result this

marginal land will deteriorate. This simplifred model shows that population pressure

can create ecological stress and land resources crises. Many other studies (Clark, 1977;

WolmarU 1993; Haub, 1996) have revealed that where there has been a long history of

population growth, there have been substantial effects on land and environment change

(Wolmar¡ 1993). The evidence suggests that the combination of a large population

base, a relatively rapid rate of population growtl¡ and rapid rates of technological

change has an effect of increasing the rate of change on land and in the environment.

One ofthe major impacts is associated with the structural economic changes producing

a shift of people from rural to urban areas, leading to a greater proportion of the

population living in urban centres and a greater share of land use being in urbanized

areas (Hugo, 1982).

J
There have been numerous studies on the land use issue in Indonesia (Sandy, 1975,

Sandy, t977;FirmarU I997;Tampubolon"l997;Nasutior¡ 1997; Silalahi, 1999). These

studies have mostly focussed on making an inventory of land use conditions and the

causal relationship between social and biophysical factors. Very few studies have

sought to understand the dynamics of land use/land cover change over time and space

and to predict future changes. In this context, knowledge on the causes and eflects,

magnitude, spatial and temporal distribution of land use/land cover change dynamics,

as well as change prediction in Indonesia is far from complete. One of the problems is

the limited availability of spatial information regarding land use/land cover conditions.

There are a range of techniques for collecting information on land use/land cover

conditions, including field surveys and mapping from aerial photography and satellite

remote sensing. These techniques have various advantages and disadvantages. Field

surveys can provide extremely accurate data but are very time consuming and usually

impractical for large areas. Accurate mapping can be achieved from aeÅaI

photograph¡ but it depends upon the availability of photography at an appropriate

scale (Lunetta et at., 1999). Satellite remote sensing with various spatial resolutions

can be effectively used over large areas.

An additional advantage of satellite remote sensing is its application to multi-temporal

analysis (Foster, 1935). The high frequency of satellite overpass makes it relatively

easy to detect changes in land cover such as those resulting from land use changes.

Various studies have been conducted regarding land use/land cover identification using

remote sensing techniques (GordofL 1986; Jadhav, 1993; RanU 1993; Dobson, 1996;

4
 Karsidi et al., 1997; Henderson, 1997 Foody, 1999; Morisette, 1999; and V/ard;

2000).

The availability of accurate spatial information of land use/land cover, integrated with

vector data in a GIS, can facilitate frirther spatial analysis and model development.

Spatial analyses of land use/land cover change, such as where and what type of land

use/land cover has changed, and to what extent this change relates to social and

biophysical factors including why, is important for spatial planning purposes and to

ensgre scarce high qualþ agricultural land is not taken out of production prematurely.

Geographic Information Systems (GIS) ate a powerful tool to capture, manage,

manipulate, analyse and dispþ spatialty referenced data (Burrough, 1986). A common

approach of GIS is map overþ, which is capable of integrating many kinds of social

and biophysical data þers. This approach has been widely used in many spatial

analysis studies (Pathan et al., 1993; Schmid et al., 1995; Zeidle4 1997; Li, and Yeh"

1998; Cox, 1998; Debinski, 1999; and by Frederick et a1.,2001) but has beenrarely

done in Indonesia. However, this traditional GIS technique cannot effectively handle

the dynamic processes involved in land use/land cover change. Other techniques such

as Cellular Automata (CA) are a potential solution to handle dynamic process when

coupled with GIS techniques (Shi and Matthew, 2000). Cellular Automata are capable

of representing the spatial interaction among different systems in the real world and

generating a realistic prediction of a complex spatial pattern based on simple rules. It is

in these areas that this study seeks to make a contribution.

5
 This study explores land use/land cover change dynamics in an Indonesian region

using detection and spatial analysis via remote sensing and GIS techniques. It uses this

as a basis to predict future land use/land cover using a GIS based Cellular Automata
'West
model in the downstream areas of the Ci Tarum watershed, Java, Indonesia

(Figgre 1.1). The area includes the large-scale, multipurpose Jatiluhur reservoir. The

construction of the reservoir has created a dynamic response in the land use/land cover

change. Equalty important, this area has been heavily influenced by the expansion of

the mega-city of Jakarta and surroundings (Jakarta, Bogor, Tanggerang and Bekasi or

JABOTABEK). The development of new housing and other urban land use associated

with urbanization is a powerful force of change in the study area. Onthe eastern side of

Jakarta, where the study area is located, there has been dynamic e4pansion of housing

and industrial estate development. This is especially evident in the Bekasi kabupater¡

which separates the study area from Jakarta where there are many new housing estates

adjacent to industrial areas such as the Tambun industrial zone, Cikarang industrial

zone and Karawang integrated industrial estate (Goldblutrl 2001).

6
 INDONESIA
c

F
Ð gtæ¡
Ë

gF î

A
WEST JAVA

Serang

Bogor

Garut Tasik

Kabupaten boundary
Study area 30 0 30 60 Kllomelers
Wateshed

Figure 1.1. The Study area of the Ci Tarum'Watershed

1.2 Aims and Objectives

The main purpose of this study is to provide an understanding of, and to predict, land

use/land cover change using remote sensing and GIS techniques in Indonesia through a

case study in the downstream area, and surroundings of the Ci Tarum watershed in

West Java.

7
 The study consists of,

o detection of land use/land cover change over specific time intervals

o identtrtcafion of what change occurred and where

. analysrs of the causes and implications ofthe change

o prediction of land use/land cover change in the future.

It is hoped that by understanding more about the spatial phenomenon of land use - land

cover change dynamics, and developing and applying techniques to detect and predict

land use/land cover change using remote sensing and GIS, it will be possible to reduce

the misallocation of resources and mismanagement of land use in this area. Finall¡ it

is anticipated that this understanding will be of some assistance in the formulation of

policies that will improve land use planning and land resources mrinagement in

Indonesia more generally.

The more detailed objectives of this study are as follows:

o To detect and determine land use - land cover change dynamics using

remote sensing techniques.

o To understand the inter-annual dynamics of land use - land cover changes.

o To analyse the spatial dimension of land use/land cover change dynamics

associated with demographic pressure, economic and physical environment.

o To develop and apply methods that can be used to detect and predict land

use-land cover change dynamics using remote sensing and GIS.

o To predict future land use/land cover.

8
 1.3 The context of the studY

The Indonesian archipelago, as one of world's most ecologically diverse ecosystems

(Tomascik et al., lggT), has experienced significant land use/land cover change. The

issue of change becomes more critical, as the population increases, leading to the

increase in the demand for space. At the 2000 Census the population of Indonesia was

206,265,000 but it was unevenly distributed between the nation's 31 provinces (Table

l.l and Figure 1.2). Of the population about 60 Yo are located in Java and Bali, which

together have about 9 Vo of Indonesia's land. The average annual population growth

rate was I.98 % 1980 - 1990 and 1.49 % between 1990-2000.

Land use/land cover change, including from deforestatior¡ has been of significant

magnitude in the last 50 years (Fredeick et al., 2001). Forest cover has decreased and

this is related to increasing population density. In Java and Nusa Tenggara, which

have high population densities, forest cover was less than25 Yo of the total area, while

in Kalimantan and lrian Jaya where population density is lower, forest covers more

th;rrn75 Yo of the total area (Figure 1.3 and Table 1.2).

Recent studies (World Bank, 2001) indicate that because of land utilization policies,

coupled with the impact of climate variability, a much larger magnitude of change in

forest cover has been experienced in Indonesia than elsewhere. According to a remote

sensing study, estimates showed that between the years 1985-1997 the rate of forest

loss was 19,700,500 hectares or 1.5 million hectare every year. It was 119,700,500

hectares in 1985 and 100,000,000 hectares n 1997 (World Bank, 2001).

9
Table l.l:Distribution of Indonesian Population and Growth Rate, 1980 - 2000

Population Average Growth

No Province ('000) Rate (% per annum)
Census Gensus Census 1980- 1990-
1980 1990 2000 1990 2000
1 Aceh 2611 3416 3931 2.72 1.46
2 North Sumatra 8361 10256 r 1650 2.0€ 1.32
3 West Sumatra 3407 4000 4249 1.62 0.63
4 Riau 2169 3304 4958 4.30 4.35
5 Jambi 1446 2021 2414 3.40 1.84
6 South Sumatra 463C 6313 6990 3.15 2.39
7 Bengkulu 768 1179 1567 4.38 2.97
I Lampung 4625 6018 6741 2.67 1.17
o Kep.Bangka Belitung*) 900 0.97
10 Jakarta 6503 8259 8389 2.42 o.17
11 West Java 27455 35384 35730 2.57 2.03
12 Central Java 25373 28521 31228 1.18 0.94
13 Jogyaka 2751 2913 3122 0.57 0.72
14 East Java 29r 89 32503 34784 1.08 0.7c
15 Banten*) 8199 3.21
16 Bali 2470 2778 31 51 1.18 1.32
17 West Nusa Tenqqara 2725 3370 4009 2.15 1.82
18 East Nusa Tenqqara 2737 3269 3952 1.79 1.64
19 West Kalimantan 2486 3229 4034 2.65 2.29
20 Central Kalimantan 954 1396 1857 3.88 2.99
21 South Kalimantan 2065 2598 2985 2.32 1.45
22 East Kalimantan 1218 1877 2455 4.42 2.81
23 North Sulawesi 2115 2478 2012 1.60 1.33
24 Central Sulawesi 129C, 17'11 2218 2.87 2.57
25 South Sulawesi 6062 6982 8060 1.42 1.49
26 South East Sulawesi 942 1 350 1821 3.66 3.15
27 Gorontalo") 835 1.59
29 Maluku 1411 1858 1206 2.79 0.08
30 Maluku Utara*) 785 0.48
31 lrian Java (Paoua) 1174 1649 2220 3.46 3.22
INDONESIA 147490 179379 206265 1.98 1.49
*) New Province established in 1999 and 2000
Source: Central Bureau of Statistic, 2000

10
 UTAEA
Ê
\ Ê
KEPULÂUAH Êtau
t
)t' U-ALUKU ÜTNNA
.r
T &luß
\ KALIMAI¡TAI¡ JÂYÂ S
BA.RÂT
u
AJi
s4-t|G Êt"r q
TE¡{GAH
.l
\ O,ffi.
BEt¡G untu
uþ \
S€LATÂN *{ 'ql
A
ßAH 9ULASESI
çl'- SELAT¡H ¡t
Ê
?
8A¡fT EH æ '-o . ói {
10Êf
JAJflATT¡Uß
.ty'-

5dl 0

Figure 1.2. Province of the Republoc of Indonesia

Source: BAKOSURTANAL, 2000

11
Sumatera

I
Sulawesi

F
ryv'fl*"
Ë

ava &.lvladura
"
þ

Bali-
oy'o Fo¡ect covet
À
fl10 0
o/o Non Forest

Figurel.3. Percentage of forest cover inthe main island of Indone

Sources: Department of Forestry (Santoso, 1999)

t2
 Table 1.2: Forest cover and Population density in the main island of Indonesia
Year 1996

lsland TotalArea Population density Forest cover %

1000 ha person/sqkm**) 1000 ha
25067 Et
llsumatera 47074 8€
2Ûavaand Madura*) 1s21el 868 2916 22

3lKalimantan 530281 1C 38562 73

¿lsulawesi 't84121 7Z 11851 64
slMaluxu 78oel 27 6603 85
E 36032 88
6llrian Java 4O75el
zlgali-Nusa Tenqqara 88131 115 2494 28
Total 1 891 141 123524 65
Source: Department of Forestry (Harry Santoso, 1999)
*) From statisticaldata of Perum Perhutani
**) Galculated from lntercensal 1995 data

In the more developed islands of Java and Sumatra, land use/land cover changes

because of land use conversion have been especially pronounced. Nasution (1999)

estimated that between 1985 and 1995, the rate of land use conversion from rice field

to non-rice field in these islands reached 20,000 hectare per year. In the islands of Java

and Bali alone, more than 223,000 hectares of rice fields were converted to other uses

between 1981 and 1999 (Table 1.3), with nearly 35%o of thts involved in conversion to

settlement, industrial estate or other non-agricultural uses (Department of Agriculture

and National Land Agency, 2000). Other estimates indicate that in the areas
surrounding the National Capital of Jakarta, the conversion of prime rice fields to

housing estates amounted to 2,000 out of 23,000 hectares in 1986 alone. Between 1989

and 2000, settlement in kabupaten Bekasi and Karawang increased, while rice fields

decreased (Table 1.4). The increasing density of settlement in this area includes an

increasing area of industrial estates.

13
Table 1.3: The of the Iture area 1981 to1999 rice field to other land in

I West Java 15847.0C I 1 6952. 5. 55. 124

2 Sentral Java 19739.0C 1 0. 0. 13

logyakarta 2666.45 1423 0 2

4 East Java 20872.87 14177 51577 0. 30 631

E Bali 2626.81 677 1 69. 0

röta¡ 61752.13 50628.

Percentage 27.62 7 27

Source: Tabutation from Directorate Extensivication of Agricufture area Dep. Of Agriculture and National Land Ag

t4
 Table 1.4: Area of rice field and settlement within kabupatens in the study area

Ktbupaten Rice field (hal

Bekasi 72677 5524t 27t61 4t709

Karawang 1001s1 8979r 22493 30726
Purwakarta 16813 t543'l 11533 13077

Subans 8732r 78942 t8275 25741

Total 276962 2394tr 79462 Ill253
Source: Statisticql office ofWest Jøva, year 1989 and 2000.

Land use/land cover in the study area changed dramatically not just because of the

influence of the expansion of the mega city of Jakarta but also the influence of the

construction of Jatiluhur reservoir. The construction of the reservotr n 1962 has

created a dynamic response in the land use/land cover change pattern. The flood

control function of the reservoir has made more land located in the downstream area

being freed from annual floods. It makes areas suitable for permanent use such as

settlement. The inigation. function of the reservoir has made water more readily

available all year round, making many of the downstream areas more suitable for

intensive agriculture. Since Jatiluhur was constructed, rice fields with technical

irrigation have increased due to the efÊect of the reservoir. The increase in rice fields

with technical irrigation continued until 1994, but it has been decreasing slightly since

then. The other categories of rice field such as rain fed, semi-technical and non-

technical irrigation, have decreased since 1989 (Table 1.5). Hence the decrease of rain

fed, semi-technical and non-technical irrigation are partly due to conversion into

technical irrigation rice field, but since 1995 alltypes of rice field including technical

irrigation have decreased in area due to conversion into settlement. Settlement has

increased since the Jakarta-Cikampek toll way opened in 1989.

l5
 Table 1.5: Decreasing rice field and increasing settlement in V/est Java 1989-2000

Year Rice field (ha) Settlement (ha)

Technical Semitechnical Non-technical Rain fed
irriqation lrrigation irriqation
2000 458240 127766 105457 502540
225043
1998 463397 123408 1 05357 501321
231220
1996 467414 128966 107207 496638
232525
1995 474492 126863 3081 10 235317 483737
1994 482787 123118 310544 249533 458869
1989 441948 135277 322807 277994 416038
Source: Centrat Bureau of SfaÍrsfic; Compiled from West Java in Figures, 1989, 1994
1 99 5, I 996, 1 998, and 2000

The decrease in technical irrigation rice fields since 1994 is related to increasing

industry and housing development. This is clearly seen in the kabupatens within the

study area as shown in Table 1.6. The installation of hydropower plants has provided

electricity supply for industrial development along the watershed. The increase of

industrial development is related to an increase in the number of industry and trading

companies. Table 1.6 shows that in four kabupaten within the study area, the number

of national industrial and trading companies between 1989 and 2000 increased, with

the highest increase in Kabupaten Bekasi (Table 1.6).

Table 1.6: Number ofNational Industrial and trading companies

Subang 7260 t2428

Purwakarta 1995 4087
Karawang 477',| 83 51

Bekasi s699 20634

Total t9731 4s500
Source : Pr ovincial Industrial Service of West Jøva,
year 1989 and 2000.

I6
Industrial development has also involved establishing industrial estates such as those in

Cikampek, Karawang, Cikarang, and Bekasi. From Figure 1.4 it can be seen that new

developing housing and industrial estates were concentrated between Bekasi, Cikarang

and Karawang along the main road of Jakarta-Purwakarta a¡d toll way Jakarta-

Cikampek. The Karawang integrated industrial estate alone covered 2000 ha, and6234

ha of 18 industrial estate companies are located in the Bekasi integrated industrial

estate (Figure 1.4 and Table 1.7).

HsJN(tnm r(^¡.8E(^tl KAE ìKAR|¡TA¡TO

I t
I I
I I
l(mh¡l¡ûl
ldra
I
I llúl
I rcrhô-l
I I .CErrìf lryt¡
!trl,6.tætr

thiln Fl( I

¡¡Ëffinl.d
llmnH*l ¿ffi1¡
låitr?É¡

hr

@ New Housing
o Indushial Estate
N

À
@ Industrial zone Iln Scaled
New Town/industrial estate

Figurel.4. Housing and Industrial estate developmertt area in Bogor, Bekasi and
Karawang (Adoptedfrom Jayadinata, 1999 pp- 185)

17
 Table 1.7: Number and area of Industrial Estates in Kabupaten Bekasi year 2000

No Name Area (ha)

1 PT. Jababeka 1100
2 PT.Puradelta Lestari 1000
3 PT. Meqapolis Manuqgal lndustral Estate 805
4 PT.Bekasi Faiar lndustrial Estate 800
5 PT.Bekasi Matra lndustrial Estate 500
6 PT.Lioo Citv Development 472
7 PT.East Jakarta Industrial Park (EJIP) 320,4
I PT. Gerbano Teknolooi Cikaranq 240
I PT. lndocargomas Persada 230
10 PT.Cikaranq Hijau lndah 230
11 PT.Hyundai lnti Development 200
12 PT.Alinda Tama Sakti 200
13 PT. Permata Kirana Sakti 125
14 PT.Rawa lntan 100
15 PT.Gobel Dharma Nusantara 54
16 PT.Adito Mula Sakti 50
17 PT.YKK lndonesia Zipper Co.Ltd 20
18 PT.Kawasan Dharma lndustri 18
Total 6464.4
Source: Statistical 2000

1.4 Research Framework

The various information needs associated with land use/land cover change detection,

identificatior¡ analysis and prediction include :

l) gathering, collecting, processing, documenting and describing land use/land

cover change dynamics;

2) detecting and identiffing the land use/land cover change; and

3) analysing the land use/land cover change and developing an appropriate

method to detect and predict land use/land cover change dynamics using

remote sensing and GIS.

l8
This study includes three main elements: remote sensing, existing data collections and

geographical information systems to detect, analyse and predict land use/land cover

change (Figure 1.5). Figwe 1.5 presents a general overview of the conceptual

framework employed in this study. Remote sensing techniques are a way to detect land

use/land cover change on a digital basis. The integration of remote sensing and GIS

combined with existing spatial data is used to analyse land use/land cover change and

the possible driver factors as well as to operationalise a land use/land cover change

prediction model.

Identification and change detection of land use/land cover was based on Landsat TM

image analysis. The Maximum Likelihood Classification approach was adopted to

identify and make post-classification comparisons to detect changes in land use/land

cover type. Landsat TM images were available from 1989 to 1997 anrrually. These

images enable us to detect land use/land cover change on an annual basis, and the

influence of weather or season variability within this time interval can be recognised.

The spatial analysis of land use/land covers change, and its relationship with the

possible driver factors, was based on examining both static and dynamic driver factors.

Static factors such as slope, elevation and physiography, and dynamic factors such as

population densit¡ proximity to urban and semi-urban centres, proximity to roads and

proximity to toll \¡/ay, were the main focus of this analysis. The relationship of these

driver factors with land use/land cover was assessed in order to recognise the factors

that have a strong influence on land use/land cover change in the study area.

l9
Land use/land cover prediction is another focus in this study. Future land use/land

cover information is important since it can be used to identiff areas that require
priority attention or to anticipate mismanagement or misarrangement of land resources.

The Markov Cellular Automata (MCA) model was selected to predict future land

use/land cover in the study area. This model was based on a combination of cellular

automata and Markov chain models that can represent the spatial distribution of land

use/land cover change in the study area.

Remote sensing Existing Data

The ¿n't anil sc¡unLte af'collecting Mcr¡ss ttrttl tùrtct c¿illec:Íed by ather
b),¿t ¡levice, which is not in conÍücl i n¿{ ivi ¿lt rct I or agenc i es

with the ob¡uct

Data source:
Data source: National and local Govetnrnent's
Landsat TM 1989-19931 199-í. 1997 Offices

(irolnetric c orec titrn D¿rta ctillected

PCA nld NDW ( Populntion,

[nruge Procæsittg
Transibmration i¡hastrucfure" stlil.
eústirgl ltrlcl ust,
M¿rx.Likelihtxlrl LL:L(; claxficutíon slope)
Su1rcn'il;til clnssifìctttion

Post Chatgu Digitise base urnp mcl

clnssilìcntiurr detcctíott thenratic rnaps
c0rû)¿ìrHQn

Geo grnphic Informntion System

i n tu grü ri on, rnani pu Ia Í i o n, an a lys i s' ctnd d i sp I cty aJ' spa ti a I d ata

LULC change LULC possilrlc driver LULC chnnge

tletection f¡ctors anal;-sis Sinrulaton

a qualitative ¿pl)'sis
Overla¡' and (or,erla1,, buffer &
cross tabulatiorr intersectioÐ Markov basecl
a quantitative analy'sis Cellular Automata
(discriminant
frrnction)
Figure 1.5. Research framework

20
1.5 Thesis Outline

This thesis consists ofseven chapters. A general review ofland use/land cover change,

as well as defuritions and some aspects related to land use/land cover change
dynamics, are presented and discussed in Chapter 2. Clnpter 3 consists of a description

of the socio-economic and the biophysical characteristics of the study area and the

availability of spatial data for it. This chapter provide a description of socio-economic

and the physical characteristics within the fives kabupatens that cover the study area.

Firstly there is a description of population growtt¡ occupation, and land tenure as well

as infrastructure, secondly, the physical characteristics of slope, physisography, soit

climate and hydrology are discussed. Chapter 4 discusses digital change detection

using remote sensing and GIS. Change detection techniques are intensively reviewed

in this chapter-especially detection techniques that are related to digital change

detection from the satellite imagery. The dynamic change of land use/land cover

related to the season and growing cycle in the study area is discussed. Chapter 5

identifies possible driver factors in land use/land cover change. Static and dynamic

drivers using overlays as well as statistical analysis are discussed in this chapter.

Chapter 6 is concerned with prediction. The assessment parameters of the Markov-

Cellular Automata model that are available in ldris32 were conducted to get a suitable

prediction result. Future land use/land cover change in the study area was predicted or

simulated 10 and 20 years ahead using this model. Finally Chapter 7 presents the

conclusions of the study. It includes a sunmary as well as some major findings and

some implications for policy makers and planners in land use panning. Some
recommendations for future research are put forward in the last part of this chapter.

2I
 Chapter Two

LAND USE/LAND COVER CHANGE DYNAMICS

2.1 lntroduction
Human factors such as population growth and distribution, economic growth, and

physical factors such as topography, slope, soil type, climate and others strongly

influence land use/land cover changes (Skole and Tucke4 1993). Land use change is a

matter of historical process as relating to how people use the land. Nowadays, human

attachment to land is not just as habitat and living space, but also involves more

complex purposes such as industry and tourism (Mather, 1986). Land use/land cover

change is intricately linked to the dynamics of human activities. The dynamics of land

use change is interrelated between demand and supply structure of land uses. Land use

demand increases with population and elements of economic growth such as income

per capita, GNP and industrialisation. On the other side, supply of land as space to be

used remains static.

This chapter presents an overview of land use/land cover change dynamics, providing

definition and reviews some aspects that relate to land use/land cover change.

2.2 Defining Land use/land cover change

Until recently there was no agreement worldwide, or even at national levels, on

precisely what constitutes land use or land cover, or how to define them (McConnell

22
and Emilio, 2000). There are many definitions and descriptions of land use and land

cover, depending on the purpose ofthe application and the context ofthe study. It is

hence necessary to define the definitions and descriptions of land use, land cover, and

land use/land cover change terms that are used in this study.

In general land cover is the biophysical state of the earth's surface (Turner et al.,

1995). It may be defined as the observed physical (including the vegetatior¡ natural or

planted) and human, constructions which cover the earth's surface (Baulies, 1997).

McConnell and Emilio (2000) add that water, ice, bare rock or sand and salt flats or

similar un-vegetated surface, although strictly speaking part of the land (and water)

itself and not its coveÍ, are for practical reason often included in land cover. Moser

(1996) notes that the term "land cover" originally referred to the type ofvegetation that

covered the land surface, but has broadened subsequently to include human structures,

such as buildings or pavement, and other aspects of the physical environment, such as

soils, surfaces and gtoundwater.

Land zse refers to both the manner in which the biophysical attributes of the land are

manipulated, and the intent underlying that manipulation or the purpose of the land

used (Turner et al., 1995). Moreover, Iand use involves considerations of human

behaviour, with particularly crucial roles played by decision-makers, institutions, and

the interlevel integration of processes at one level with those at other levels of

aggregation. Therefore, land use generally represents human-induced intervention or

manipulation of the land to provide food, building materials, buildings site, firewood,

and clothing.

23
A second definition of land use is the purpose for which the land is being used (Young,

1994). This definition includes the management of land and human activities which are

directly related to the land. In addition, land use constitutes a series of activities

undertaken to produce one or more goods or services. Related to the previous

definition, ITC (International Institute for Aerospace) and FAO (Food Agriculture

Organzation) adopted the following definition of land use: a series of operations on

land, carried out by humans, with the intention to obtain products andlor benefits

through using land resources (deBie et al., 1996: p 5)

The term land use/land cover has commonly been used in association with land use,

particularþ in the field of remotely sensed analysis. Land use changes aÍe a major

determinant of land cover change. Land use/land cover inventories tend to be a mixture

of land use and land cover. One category of land use may correspond fairþ well to one

class of land cover. On the other hand, a single class of land cover may consist of

multiple uses. There are two categories of land use/land cover change. conversions

from one land use/land cover into another, and transformations within a given land

use/land cover type. Lambin et al. (1999'. p.37) refer to the process by which land

use/land cover is modified/converted as including two main components:

. The activities (or operation) and inputs that are undertaken (or

restricted) on a piece of land with significant land cover consequences;

and

. The goals/intentions motivating these operations, including both the

outputs (goods or services) that are expected, and theforces that cause

Iqnd uses lo occur in a certainway, at a certain time, in a certain place.

24
In terms of detection or identification, land use/land cover is the observed physical

cover at a given location and time, as might be seen on the ground or from remote

sensing. Furthermore, land use/land cover may be determined by direct observation,

whereas information on land use requires a statement of purpose from the person who

controls or carries out the land use. Remotely sensed data (e.g. from satellite images)

can often be used to map land uselland cover, for example, by identifying multi-

spectral signatures characteristic of certain land cover types. Land use, in turn, is often

related to land cover, so that land cover may be used to infer land use (Adger, 1994).

2.3 The dynamics of land use/land cover change

Land use/land cover can change due to natural or human factors. Natural factors such

as flooding, drought, forest fire, and volcanic eruption result in a change of land

use/land cover. Human factors such as demographic pressure, level of poverty, and the

economic and institutional structures of resource use are an indirect cause of land

use/land cover change. Moreover, with an increase in population and economic

growth, human activities have occupied more land (Mather, 1986). Changes in the way

which populations live their lives will change the function of the land, and the human

factors such as population and their activities are a driving force in changes in land

use/land cover. Sage and Grubler in Mayer (1994) argue that the main driving forces of

land use - land cover change are population, income and technology. Robinson (1994)

argues that the initial driving forces to be considered are population, income,

technology and price. Hence population and their activities, such as income and

technology, are forces shaping land use. But recognition of the link between human

25
activities and land use/land cover type is not simple. Single or multiple factors may

simultaneously or independently atrect changes of land use/land cover.

Land use type changes from forest to agriculture or from agriculture to settlement or

from agricultural to industrial or cornmercial, for example, are the result of changes in

human activities. On the other hand, the nature of soil such as mineral content, texture,

wetness and dryness as well as topography are factors that affect land use/land cover

change. Therefore, the interaction of human and biophysical factors influences the

dynamics of land use/land cover change. This interaction is complex and needs to be

simplified and categorized in order to identify the most influential human and

biophysical causes of land use/land cover change. McNeil (1994) concludes that there

are four major driving forces - political, economic, demographic and environment

(Table 2.1). Within these broad categories, he identifies specific attributes that strongly

influence land use/land cover patterns. Land uses for settlement, for example, may be

strongly influenced by demographic pressure such as population growth, education

levels and income per-capita or may act together with other factors such as economic

growth, policy or technology. In Indonesia, land use change is related to all these

factors. The policy of privatisation of forest concession companies for example forced

forest conversion in the main islands such as Kalimantan, Irian and Sumatra (World

Bank, 2001). In the islands with a high-density of population such as Java, land has

been converted from agricultural to settlement and industrial use. From Table 2.2 it can

be seen that a high rate of conversion from agricultural to settlement and industrial

areas has occurred in West Java, Central Java, East Java and Bali. This is evidence that

26
population or demographic pressure is one of many factors that influence land use

change in Indonesia.

Table 2.1: Driving forces behind land use change

1.Political:
Decision-making process of public participation nitary of federal structure
Local pressure, special ed, centralized/ umber of special-interest
interest and corruption
State capacity s ector pressure/influence sector expenditure/

land area

2.Economic:
Vulnerability to external vs. closed economy
pressure(economic/p olitical) concentration
sector dependence sector export/
export
of exchange rate management (Population -
active)
total
exchange rate
service ratio

Market allocation mechanism State controlled, market driven subsidies

sector expenditure/
P
Technological intensity low intensity of GDP
Level and division of wealth ealth/poverty-induce sector/GDP
(asset inequality)
agriculture/PEA total
absolute poor in
population

3.Demography:
Population pressure on High-low land
the land in population

agriculture/PEA total

4. Environment:
Natural resource quality Scarcity lr,o.u, yield, flow
Source: Adapted from McNeil ( 1 99a.p. 60)

27
 Table 2.2: Change of agriculture area to other land use between 1981-1999 in Indone

Nc Province To SettlemendTo lndustry o Drv l¡ndTo Plantationlto snrudTo open La

1 Aceh o.ool 0.oo o.od 0. o.o0l o
2 North Sumatera 0.0d 0.00 0.od o.ool o.ool o
a West Sumatera 4e5.s7l 22.21 12.0d 3e2.ool o.ool o
4 Riau 0.001 0.00 o,od 0.001 0.ool o
5 Jambi 0.001 o.o0 o.ool o.ool o.ool o.
6 Benqkulu 0.001 o.oc o.od o.ool o.ool o.
7 South Sumatera 61.2S1 oOC 0 0 105 11
I Lampunq 4221.001 1e7.0C o.od o.ool o o.

11 fooyakarta 2666.45 174 1423. I 338 8.

12 3ali 2626.81 677.001 2169. e62.6ol o.ool o

14 i/Vest Nusa Tenggara 13.72 0.001 o.od o.ool o.ool o
E
1 iast Nusa Tenqoara 0.00 o.ool o.od o,ool o.ool
1 € ullest Kalimantan 0.0c o.ool o.ool o.ool o.ool 0
11 last Kalimantan 0.0c o.ool o.od o.ool o.ool o.
1 I South Kalimantan 4315.0C 1620.00l O.od 13.001 0.001 o
1S Sentral Kalimantan 0.0c 0.001 0 o.ool o.ool o
2C South Sulawesi 2123.74 14s.3ol o.od 36s.251 oool o.
21 \ortheast Sulawesi 0.0c o.ool o.od o.ool o.ool o
22 Sentral Sulawesi 11.6e o.ool o.od o.ool o.ool 1 .
22 tlorth Sulawesi 0.0c o.ool o.od o.ool o.ool o
24 rian Java 0.0c 0ool 0.0d o.ool 0.001 0
2a \4aluku 0.0c 0.001 o.od o.ool o.ool o

Source:Tabulation from Directorate Extensivication of Agriculture area Dep. Of Agriculture and National L

28
The forces driving land use/land cover change can be distinguished into two categories

- direct and indirect. Direct drivers are associated with activities which directly interact

with and modify the physical environment, such as deforestation, urbanization, and

agricultural expansion (Turner and Meyer, 1994). The indirect drivers, or root causes,

influence how individuals or groups interact with, and change land use/land cover

@laikie et al., 1994). These are generally more complex as they are built into the

human system underþing a land use activity (Adger and Brown,1994; Krummer and

Tunner, 1994). An illustration of driving forces related to human systems can be seen

in the demand and supply relationship of a land use change model @igure 2 l). The

demand structure of land uses is very dynamic; it means that the demands for using

land for human activities increases from time to time under the influence of population

growth, community structure and economy. On the supply side, structure is relatively

static, the surface area is constant, and elasticity from this supply side is on the aspect

of physical characteristic and function of the land. Figure 2.1 shows the mutual

interaction between socio-economic activity structures, biophysical factors and the

dynamics of land use change. This flowchart explains the increase of land use demand

as an effect of an increase in socio-economic and welfare growth, increasing the

demand for goods and services, which in turn affects the increased demand for

productive land use. The flowchart also explains the opportunity of increasing on the

supply side. Institutions, policy, technology and spatial distribution/variation in the

natural quality of land (soil type, slope, altitude and climate) are the factors that could

lead to an increase on the supply side. Institutions and policy at national, local and

community level have an influence on technology and land use decisions. The

29
technology level will affect the capability to cultivate and employ the land

(Saetulhakim, 1999).

Danographic pressure;
Population Crrowth Institution and
and shift Policy

on Change of
ofGoods socio-economic
and and
ptrowth
Increase
demand

Dynamics of
Land use
Increase of supply;
A¡ea constant
Elasticity factors
supplies are:
in land
and
and spatial
landscaoe
land

Biophysical Technology
Characteristics Achievement
Change ofsupply
Elasticity of land use

Figure 2.1. The mutual interaction of land use change dynamic

Modified from Saetulhakim (1999)

2.4 Some aspects related to land use/land cover

change

2.4.1 Land use and Demographic pressure

Demographic pressure such as population growth and shift are a driving force of land

use/land cover change which is unique and can be quantified (Meyer, 1992). However,

population growth is not the main driving force as its importance is relative to the other

30
forces generating land use/land cover change such as technology and social advances

that improve the conditions of life. Young et al.(1991) did not find a significant

correlation between the land use/land cover changes with increases in population

density at national scale across the world. Hence increase in population density may be

sufficient to cause land use change but it is not necessary. Significant correlations

between population and land use/land cover change have been found when

investigation is restricted to regions possessing similar socio-environmental

characteristics. Comparative studies offer statistical evidence supporting the claim that

population growth drives or strongly contributes to forest clearance (Meyer, 1992).

A study that analyses the interactions of population growth and land cover change

@ilborrow, 1992) concludes that population growth is an important factor, but one

significantly modified by natural and institutional context Land use/land cover change

is closely associated with agriculture, but not clearly linked to population (Meyer,

7992'. pp. 54). Empirical evidence higtrlights the complexity of factors concerning

population pressure and its links with agricultural extensification and intensification.

These would appear to imply that whilst the initial response to increases in population

pressure may be agricultural extensification, when land becomes more difücult to

access, or only very marginal areas remain, then intensification occurs. Turner et al.

(197'1) argue that population pressure increases agricultural intensity: "increases in

population pressure causes increases in agriculture intensity"(Turner et ø1., 7977

p.38a). Hence land use/land cover change is usually an outcome of population growth.

Moreover, population change has a direct association with change in the way people

lead their lives and their activities, including agriculture activities. Population growth

31
is argued by some to have exceeded the capacity of the biosphere, as managed by

society, to sustain it @hrlich. I., 1990 and Ehrlich. I. 1988 cited by Meyer, 1992:

pp 52). Growth in population and economic activity require land not only to fulfil the

basic needs but also to ensure needs in social life and leisure activities are met. Thus,

the more people, the more land needed for agriculture to produce enough food and for

non-agriculture sector to build their housing, manufacturing and other activities.

2.4.2 Land use conflict and reg¡onal food secur¡ty

Land as space remains static and demand to use the land rapidly increases as a result of

population growth. The more people the more space that is needed for their living and

other activities such as agricultural and commercial activities. In many cases

population growth can bring land use conflict. Land use conflict frequently occurs

between agricultural and non-agricultural uses. This conflict occurs because of the

basic principle that each piece of land should be devoted to the use in which it would

yield the highest rent (see Hoover and Giarratani, 1999: pp.7). Therefore demand from

the non-agriculture sector is usuaþ much stronger than from agriculture due to the

much higher incomes which this sector can obtain from the same area of land. If this

condition occurs in highly productive areas for agriculture it can have an effect on

overall food supply. There is a significant correlation between the amount of cultivated

land and food supply (Li and Hong, 2000). The conversions of agriculture area into

non-agriculture area will other have an effect on regional food supply. This is because

urban areas lead to be located in the most productive agriculture area. This situation is

occurring in Indonesia. Indonesia was selÊsufücient in rice production in 1984. After

this time the rice fields began to decrease. In 1983, the area of agricultural land in Java
32
was about 5.42 million ha, but after 10 years decreased to 4.41 million ha or decreased

about l.l millionha (Table2.3).

Table 2.3 Changes in Agricultural Land in Indonesia 1983-1993

Area (Million Hectare) Area (Million Hectare) o/

/û
Java 5.42 4.41 -r.01
Bali &Nusa Tenggara r.22 1.07 -0.15
Sumatra 5.66 5.42 -0.24
Sulawesi t.64 1.78 0.14
Kalimantan 2.23 2.r9 -0.04
Maluku 0.38 o4 0.02
Irian Jaya 0.r7 0.18 0.01
Total 16.72 15.45 -1.01
Source: Central Bureau of Statistics

From 4.41 million ha agricultural area, Java, supported about 63Yo of domestic supply

of rice, the remaining 38Yo supported by 11.04 million ha from out of Java. The
decreasing of rice field in Java has an effect on decreasing rice production and rice

supply at the national and regional level (Table 2.4). Therefore, to compensate the
decreasing rice field in Java, the government of Indonesia has a long-term policy to

open one million ha rice field in peat area in Central Kalimantan under President Decry

No. 8. 26 December 1995 (Sudrajat et a1.,2001).

JJ
 Table 2.4. The average rice productivity in Java and offJava (Ton/halyear)

Bali & Nusa Tenggara 40.88 40.91 40.48 40,76

Sumatra 38.08 38.26 37.r2 37.82
Sulawesi 43.38 43.48 38.8 43.38
Kalimantan 26.36 26.38 24.r 2s.6r
Maluku and Irian Jaya 25.31 2s.4t 24.87 25.2
Source: Central Bureau ofStatistics Year Book oflndonesia 1996-1998

A major factor in the decline of agricultural land in Java is the increasing

industrialisation and urbanisation occuring in the island (Table 2.5). From Table 2.5 it

can be seen that the highest conversion of land from rice field to settlement and

industrial areas was in Java. It was 59125.32 ha change from rice field to settlement

and 15600.87 ha to industrial area. While in other islands such as in Irian Jaya and

Maluku there was no conversion from rice field to either settlement or industrial areas.

Table 2.5: Conversion from rice field to Settlement and

Industrial areas in the main island of Indonesia 1981-1999(ha)
No Province Settlement lndustry
1 Sumatera 4777.85 2',t9.21
¿ Java 59125.32 15600.87
3 Bali 2626.81 677.0C
4 Nusa Tenggara 13.72 0.0c
q (alimantan 4315.0C 1620.0C
7 Sulawesi 2135.42 145.3C
6 rian Java 0.0c 0.0c
7 Maluku 0.0c 0.0c

Source: Dep. Of Agricufture and National Land Agency

34
2.4.3 Land use, lndustry and Manufacturing

The demand for land by industry and other commercial purposes has strongly

expanded because of the rapidity of industrial development. Many factories have been

constructed and spread into the countryside in Java (Chapin, 1972; Benstein, 1994;

Hidding, 2002). The number of large and medium manufacturers in Java has increased

with about 80o/o located in Java (Table 2.6). Often industry requires more land than

needed at the time to provide for subsequent growth. In effect, a lot of agriculture or

rural areas have been converted into industrial or commercial area.

Table 2.6: Number of large and medium manufacturers 1997-2000

Iava 18024 t7?j,6l nszs 17995

t\/
/a\ 80.s1 so.46l $.22 81.15
Outside java 4362 4ls7l 4145 4t79
o,/
/o 19.49 le.54l 18.7s 18.85
Total 22386 2t42j,l 22070 22174
o/
,/o 100 lool loo 100
Source: Statistical ffice of West Jqva, year 1997,1998,199 and 2000

Tandy (1973) categorizes to 6 categories of industrial land use.

l. Offices and commercial building

2. Factories in urban situations

3. Factories in rural situations

4. Ancillary land use, e.g. cooling ponds, storage land, recreation

5. Special types of industry, e.g. power plant, dams or reservoirs

6. Extractive industries.

35
Each category has a different interrelation with land use/land cover change. Office and

commercial building areas usually planned in an urban situation, therefore land

allocation for this category has an effect on urban land use change (Chapin, 1972;

Rhind et al., 1980; Dawson, 1984; Cloke, 1989). However, in practice, there are

difÊculties because of competition with other purposes of land use. It is clear,

nowadays, that industries look at fresh sites in the countryside to build a factory. This

condition has an effect on land use/land cover change, because construction of new

factories needs new roads, communications and other facilities. In Indonesia, for

example, new towns are often built adjacent to industrial estates @igure 2.2). Lippo

Karawaci in Tanggerang or Lippo Cikarang in Bekasi is new towns surrounding

Jakarta adjacent to industrial estates and service-oriented facilities such as hotels and

ofüces. Development has extended beyond the administrative boundary of Jakarta

Special Region, transforming the adjoining district (JABOTABEK: Jakarta, Bogor,

Tanggerang and Bekasi). The development trend becomes clear when the land area

used for some new town projects is considered. For instance, to the west of Jakarta in

the Tanggerang district, the new towns of Serpong has taken ó000 ha and Tigaraksa

3100 ha. To the east of Jakarta in the Bekasi district, Cikarang Baru adjoining

Cikarang stretches over 5400 ha, and Bekasi'2000 an area of 2000 ha. These new

towns are adjacent to the industrial areas such as Bekasi integrated industrial estate,

Tambun industrial zone, Cikarang industrial zone and Karawang integrated industrial

estate (Djoko, 1996; Jayadinata, 1999; Goldblun¡ 2001).

36
 I(AÀTA¡FERA¡E DrqJÆqwn XA¡.E(ASI r(A¡.u( Rr$ro
liler!Í¡r{t
ft!fiì¡,lffir
8h OF¿Ctt ¿IEit
Tbg{rÌ{. l.ùOln ,fn¡tÉrtl
r#h*t0
hlaooh¡
O.FJI¡.ÈÍu
n¡ry*n 0.H,6.¡F0lr
Tripr¡
$.@lf,
l{ñ FK

tþpoûreíR
anFllo?!0hr I
t l¡tÉt3Hl
I l(mùÉll t.ff!r
I EddbfÊ-Ú
I !¡fhJaÊ
lllrtna
s.ml'
eüOS4ù¡ füÇrLrôtf
fLa 'ñe, i.5@lÐ l(ntrglrDl.t¡
2mfr

@u.m;r
New Housing
@
a Industrial Estate

@ Industrial Zone A
New Town/Industrial Estate Un Scaled

Figwe2.2. Housing and Industrial estates development area in JABOTABEK and

Karawang (Adopted from Jayadinata, I 999: p' I 8Ð

2.4.4 Land use and Agriculture

As previously mentioned, agricultural intensity is associated with population pressure;

therefore the behavour of the agricultural sector with respect to demand for land from

non-agricultural sectors depends on a variety of factors related to its demographic

situation, such as population growth, structure of farm populatior¡ technology

adaptation and others. Agricultural land use involves growth of food crops and other

economically valuables crops. Food crops in Indonesia consist of paddy/rice, maizø

cassav4 sweet potato, peanut and soybean. Other economically valuables crops consist

of tea, coffee, oil paln¡ tobacco and others. The extension of this agricultural land use

is constrained by climate and soil factors and also depends on the agricultural market

37
situation, changed production conditions, and government intervention such as

incentives and regulation (OECD, 1976: p. 3l). In the humid tropics, such as in

Indonesia, intensive cropping systems occupy most of the resource-rich land, fertile

soils, low slopes, and areas with adequate rainfall or irrigation. For example rice fields

are mostly located in the flat areas with good irrigation and fertile soil (Figure 2.3).

bon

Land use
üi;;i.; Mangrove
1ffi Forest
fl:!.î.Ti Seüement
Plantation estate
iliî.Wetland
+ÉÉË RicE field 2 cycles
üå+ Rice iield 1 rycle
¡p64 Water/reseruoir A
ffi
'l:i::;l+
Fishpond
30 g m Kilûm eåers
fren/dry land
i;::.;,:t frerúdry land

Figure 2.3.Land use in \ù/est Java 1993

Source: SAN ProiectBPPT, 1998.

The other economic value crops such as coffee or tea are located in areas upland of

paddy fields. These crops mainly occupy fertile soil with gentle slopes at an altitude

within 500 m above sea level. In many cases areas with poor soils and water resources

have received far less attention in agricultural development, but increasing population

and development pressures are forcing conversion of these areas into more productive

38
and intensive land uses. These pressures continue especially in areas where there is

indu strial development.

2.4.5 Spatial dimension of land use/land cover change

The issue of spatial scale is important in land use/land cover change analysis. Land

use/land cover change analysis at the world level, for example, is quite diflerent from

analysis at the national or local levels. Moreoveq the analysis of land use/land cover

change is essentially performed at the level of the land use/land cover classification

systems. These systems are tied usually to a particular spatial scale. For example, at the

world level, FAO distinguishes four or five major land use/land cover types

(Briassoulis, 2000). At the level of nations, the number of land use/land cover types

increases to around ten. At this level, each type was divided into a more detailed land

use/land cover typology- Agricultural land is further subdivided according to the type

of product (annual, perennial), pastures are distinguished according to ownership status

into public or private.

At the national level of Indonesia, for example, the National Land Agency of Indonesia

divides land use classification system into three levels: level I, II and III (Table 2.7).

The level of this classification system relates also to the source of data, acquisition

techniques and spatial resolution or map scale. Anderson ef al. (1976) summarised the

level of land use/land cover classification based on data source and spatial resolution

into five levels (Table 2.8). Table 2.8 shows that Landsat MSS is capable of producing

land use/land cover classification to level I with map scale of 1: 250 000 and smaller.

Classification level I consists of the main land use/land cover

39
 Table 2.7 . The land use classification system in Indonesia

Level I Level ll Level

Scale l: 200 000 Scale 1: 100000 Scale l:25 0
and 1: 5(l 00O 1:125
1 Settlement 1a Residential 1a1 Residential 6a1
2 R¡ce f¡lecl 'tb Cemetery )1Real Cemeterv 6b'l
Drvland Aor¡cultur. Plantat¡on 1c EmÞlacement ljn real Cemeterv 6b2
Mixe oarden 2a Rice field 2 time harvest 1c1 Permanent Emolacement 6c1
4 Forest 2b Rice field I time harvæt+horticulture 1c2 Temporary Emplacement 6c2
Eareland Rice field I time harvest+horticulture 2a1 Ricefield3xatavear &t1
Þ Lake and wetland 2d Rice field with can, tobacco, rosela 2a2 Rice field 2x atayear 7a1
7 Others horticultur 2b1 Rice field 2 x + hoÉiculture 7a2
3a 2b2 Rice field 1 x + horticulture
3b Drvland 2c1 Rainfad rice field 1 x a year 7b1
3c Veoetable 2c2 SwamriceF¡eldlxayear 7b2
3d Flower 2d1 Rice field with Suqarcan 7b2
4a Rubber 2d2 Rice field w¡th Tobaco 8a1
4b Coffee 2d3 Rice field with Rosela 8a2
4c else a1 'eoalan with certain croo 8bl
5a Mixe plantation ¡'l rvland O-l vearwith certain croo ab2
5b Fruit croÞ 3b2 Druland 1-3 vearwith certain croo 8b3
6a Dense forest k1 Veoetable with certa¡n croo 9a1
Secondarv forest 3d1 Flower with certain croo gaz
6c Homooen forest 4a1 Productive rubber 9a3
5d Swam forest 4a2 Pre-oroductive rubber 9b1
7a Marcinal land 4b1 Productive coffee 9b2
7b Badland 4b2 Pre-oroductive coffee 9c1
8a Grasslard 4c1 else 9d
8b Srhub 5a1 M¡xe Dlantation oroduct¡ve 1C
9a Pond 5a2 Mixe olantation ore-oroductive
9b Lake 5b1 Fruit olantation oroductive
9c Swam 5c2 Fru¡t Dlantation Þre-Þroductive
9d Dam/Reseruoir
10 lrrioation

40
category such as built-up area, agricultural area, open/bare land and water body or

wetland.

Table 2.8:Land use/land cover classification level and their sources

Classification Data source Map output scale

Level
Landsat MSS and similar 1:250000-1:lMillion
TI Fligh altitude (I 2500m) photography; I : 80000 and smaller
reconnaissance survey
il Medium altitude (3 000- 1 2 500m); I : 20000 - l: 80000
Field survey
tV Low altitude (below 3000m); 1:2500-1:20000
PhotosraDhy: ground survey
V fndividual building or sub-building l:100-1:1000
Source: Anderson et al. 1976

Recently, satellites can produce land use/land cover maps with more detail or higher

classification level. For example, Landsat TM has a 30 m spatial resolution and can

produce a map with scale of l:80 000 or l. 50 000. This map scale can produce land

use/land cover classification at level III as in Anderson's system or level II as in the

Indonesian system. New satellite such as Ikonos has a I m spatial resolution and can

produce a map with scale up to 1: 2000. Based on this spatial resolution, land use/land

cover classification system can be derived at a parcel level. At this level land use/land

cover classification becomes very detailed capturing local environmental, socio-

cultural, demographic, economic and other details.

These spatial scales have an influence on the level of land use/land cover change

observed. For a given time interval, land use-land cover change may not be discernible

at higher spatial levels while at lower levels (e g. at the level of a settlement), very

4l
large changes may be measured. The role of the land use/land cover classification

system is critical in this context, as are also the source and acquisition techniques of

spatial data for the measurement of land use/land cover change.

Hence, regardless of the intended level of analysis, the results obtained and the ensuing

description of land use/land cover change refers to land use/land cover change at the

level of the spatial scale. The smaller the scale the broader the land use/land cover

category and the more generalised the analysis.

2.4.6 The importance of land use/land cover change detection in

spat¡al planning

In many countries there have been significant changes in land use/land cover as a

consequences of development process (Scholten et al., 1999). Historically, more land

has been used for agricultural and urban purposes. The lack of collection of land

use/land cover data or information will have implications for land use planning or

spatial planning processes. Scholten et al.(1999) state that the need for land use/land

cover change identification at national, regional or local level is increasingly important

to determine the likely impact of specific projects or policies in advance of their

implementation Recently, with land use/land cover data available on a digital basis

from remoteþ sensed data and GIS, the lack of information on land use/land cover

change can be eliminated. An illustration of a model of spatial planning related to land

use/land cover supported by a digital model of spatial data is shown in Figure 2.4. This

figure shows that when land use/land cover data are available in digital format, they

42
may be integrated with other data attributes to create suitability maps for each land use

type and finally to support the land use allocation process for each parcel. It is a

dynamic model because it deals with changes in land use taking into account the

present land use pattern.

Current land
use/land cover Selected Selected
attribute per grid spatial policy planning
cell optrons scenario

Exogenous
Grid cell Suitabilrty sector specific
attributes map regional model
specification results

Suitability map for each land use type

Model engine: Regional

Exogenous allocate land use projections for
allocated land to specific grid each land use
use cell type

Figure 2.4. Overall model spatial planning structure related to digital data of land
use/land cover. (Adopted and modifiedfrom Scholten et al., 1999)

2.4.7 Land use evolution and demographic pressure in lndonesia

Historically, land use inventories were initiated in 1930 when Sir Dudley Stamp for the

first time undertook a land use inventory of the British Isles (Stamp, 7931) In

Indonesia, systematic land use inventory started in 1960 (Sandy, 1995) as a follow-up

of the Master Agrarian Act (UUPA) 1960, and land use inventory has been continually

43
conducted since 1969 (Silalahi, 1999). Since that year, land use in all provrnces rn

Indonesia has been mapped. The result is a land use map consisting of information of

the pattern and spatial distribution of land use in all provinces in Indonesia.

Based on the spatial information of land use condition in all provinces, in 1977, Sandy

(1977) produced a simplified model of land use evolution in Indonesia called the

"schema of Land use Evolution in Indonesia" or "Skema Evolusi Penggunaan Tanah di

Indonesia". This schema shows the evolution of land use pattern from schema A to

schema I (see Figure 2.5). Schema A represents land use condition in area that humans

have not settled yet, and schema I represents the land use pattern on critical condition,

which is when some reserve area has been changed into marginal land.

1000

Figure 2.5
The concept of land use
evolution schema in
500
Indonesia

25

0 Í
1000 t ll Legend

ffi Forest

ffi Rice Field

500 W Dry land agriculture

ffiT Mixed Garden

Shifting Cultivation
c
'*u**"t

25
ffi..E
tÉi=;:t
Swamp

Marginal land
0

Source: Sandv 11977)

44
This land use pattern evolution is clearþ recognized in Indonesia. One example can be

seen in Figure 2.6 below that shows the land use evolution in the Indramayu area, West

Java. Figure 2.6 (A) shows the land use condition as the same as schema D and E in

1857 when people start to develop their agricultural practices. Some dry agriculture

areas have been converted into rice field and rain fed areas with I cycle and improved
to 2 cycles rice field when supported by imigation. Figure 2.6 (B) illustrates land use

condition that has been changed similar to schema G and H by 1940. As an effect of
population growth and improving agricultural practices, shifting cultivation has ceased,

and some swamp areas have been drained and used for other agriculture practices.

Land use condition the same as schema I is shown in figure 2.6 (C) for a depression
area and Figure 2.6 (D) at high elevation area (mountain) in 1969. At this stage some

marginal land present in the upper and lower areas begins to show effects of overuse.

í€IERåilÊè8. ¡s¿t
ffiñr
kiÈùÈ\96ãr
e*"-a, f"til{ f.åHtliffi
c i êru
*sÈr ts49r.
t. a 3&
å +å¡q* ì.ìe,Þe3
1sìù ltre d-lÞs¡q- .î4þìêe
ÞeãÊ{4 G-11
T¡u.,ç S;i$ilô F* F

PAt{tÁl

tr
lrp
1tðc
Pã€L!¡li-ËtGAN iS Êã

q*-*--L.:*Jx{
'lAi-¡\P ÈäËiii& 3 *--+.

'West
Figure 2.6Land use evolution in Indramayu and its surrounding, Java

(Kartono et a|.,1998)

45
Silalahi (1999) conducted another study of land use change in Indonesia, which

focuses on analytic comparison between conservation and cultivation areas. The study

concludes that many factors affect land use change but the most significant factor is

population, specifically the density, growth rate and the percentage of poor people^

Land use changes in conservation areas are affected by population density and

increases in land ownership (Silalahi, 1999).

The evolution of land use pattern that has been shown above and the study conducted

by Silalahi (1999) focuses on the general scale and does not contain actual figures of

land use conversion such as settlement and urban land use. The urban land use will

continue to change with population growth and technological achievement. Firman

(2000) conducted a study of rural to urban land conversion in Indonesia. This study

concludes that the land conversion on the fringe areas of large cities in Indonesia in the

1990s has been largely due to the issuing of excessive land development permits

granted by the National Land Development Agency. This conversion mainly involves

land use change from agricultural into urban land use such as settlement, industrial and

other commercial areas. Competition of land use in the urban areas will have an effect

on land use in the countryside. The growth of Jakarta metropolitan city in Indonesia,

for example, has a strong influence on land use change in other areas such as Bogor,

Tanggerang, Bekasi and now through to Karawang and Cikampek as has been shown

inFigare2.2.

46
2.4.8 Land use/land cover change detection and spatial planning in

lndonesia

Documentation of land use as well as other biophysical spatial data was conducted in

Indonesia in the earþ 1960s (Sandy, 1995), but documentation in digital format began

since the RePPPToT @egional Physiscal Planning Programme for Transmigration)

project which was launched in the 1980s (Rais, 1997). Since this year several studies

and a pilot project dealing with land use inventory and spatial planning on a digital

basis have been conducted. Land Resources Evaluation and Planning (LREP), for

example, is a project at the Nationallevel to support spatial planning in Indonesia. This

project began in 1987 and finished in 1993 and was coordinated by Bakosurtanal (The

National Coordinator of Survey and Mapping Agency). Marine Resources Evaluation

and Planning (MREP) is another National level project to support spatial planning, but

mainly focuses on marine environments. Part of the results of these projects is

providing spatial maps (base maps as well as bio-physical and land use/land cover

maps) in digital format. Another pilot project that deal with spatial planning is the

Land Use Planning and Mapping (LUPAM). This project, under technical aid from the

German government, was implemented in the 1990s in the East Kalimantan province

(GTL,2000). Some important information from these projects is that the land use/land

cover change detection should be pursued in tandem with a streamlining of the spatial

planning process. Spatial planning in Indonesian areas formerþ controlled by central

government (strictly a top-down approacþ (JU No.2411992) has been entangled in

manifold inconsistencies with regard to competencies of different levels of government

and the roles and required contributions from its stakeholders. The new paradigm in

the period of Reformasi year 1999 supported by law No.221I999 attempts to do the
47
opposite (bottom-up planning approach) Gru No.22l1999). Through this new

approach, the spatial planning process must be supported by adequate spatial data or

information at appropriate scale. The government regulation No. 10/2000 supports this

new spatial planning process with regards to requirement of an accurate and an

appropriate map that has to be used. This law guides the office or institution

responsible for spatial planning activities at any level (National, Province and District)

to use maps of appropriate scales (Table 2.9).

Table 2.9: Scale map at different level of spatial planning application.

Name Scale Purpose

National Reqion Mao I :1000.000 Soatial olannino at National level
Provincial reqion Mao 1: 250.000 Soatial olannino at Provincial level
1: 100.000 Spatial planning at the provincial level
that has small area of authority
Kabupaten/Kota madya 1: 50.000 Spatial planninq at Kabupaten/Kota
[DistricUM unicipalitv) madva level
1: 25.000 or Soatial olannino at the Kabuoaten or
I : 10.000 Kota madya that has small area of authority
Source.' Govemment regulation No. 1 0/2000.

As can be seen in Table 2.9, different scales are to be used for different level of spatial

planning. To support spatial planning at the district level, for example, the map used

required by the law, is a map with minimal scale of 1: 50.000 and for the small

district/municipality a map of scale 1:25.000 or l:10.000 must be used. This new law

has implications for providing spatial data, include land use/land cover data, for spatial

planning process. Figure 2.7 shows the integrated planning system for regional

development in Indonesia, which uses current land use/land cover condition and other

spatial and non-spatial data as a basis of this spatial planning process. Spatial planning

covers physical planning and land use planning, which is closely related to, and should

48
provide the basis for, regional or local development planning and sectoral planning as

well as land resources management.

Regional
Cunent Land use/land
cove¡ condition Development
and others spatial and Planning
non spatial data

Sectoral
Physical Land Use
Planning
Planning Planning

Land
Resources
management

Figure 2.T Integrated Planning Systems for Regional Development

(Adopted and modified from GTZ, 2000)

Spatial planning is the fundamental process to identiSr, analyse, and organise the

physical requirements of a society's need and interests on a spatial and functional

basis.

Physical planning is a more specific form of planning, which is the designing of the

physical infrastructure of an administrative unit, such as transport facilities - roads,

railway, airports and - facilities for towns and other human settlements in anticipation

of population increase and socio-economic development. Physical planning is usually

carried out by sectors in charge of planning and managing particular infrastructure

49
Land use planning is a process that facilitates the allocation of land to the uses that

provide the greatest sustainable benefits (UNEP, Agenda 21, paragraph 10.5).

This schema has been introduced by GTZ to avoid the confusion of terms that indicate

the differences in interrelations between the various plans and the inherent danger for

overlaps and hence contradictions, between the spatial plans, development plans and

land use plans in Indonesia.

2.5 Gonclusion

This chapter has provided a review of land use/land cover change dynamics as well as

land use/land cover definition and understanding, including some aspects related to

land use/land cover change. The evolution of land use pattern in Indonesia has been

presented to illustrate the relation of land use change and demographic pressure in the

study area.

Land use and land cover definition depends upon the purpose of the application and the

context of the study. Land use and land cover in this study has been defined as terms of
"land use/land cover", which is land use representing human-induced intervention or

manipulation of the land to provide food, building material, building site etc. Hence,

land cover referred to the biophysical state of the earth's surface. It may incorporate

the observed physical properties, including the vegetation (natural or planted) and

human construction, which cover the earth's surface.

50
The interaction of human and biophysical factors influences the dynamics of land

use/land cover change. This interaction is complex and needs to be simplified and

categorized to identiff the most influential human and biophysical causes of land
use/land cover change. In relation to demographic pressure, land use/land cover change

is associated with population growth. Population has a direct association with demand

of land for their life and their activities, including agriculture activities. The initial

response to increases in population pressure is agricultural extensification, when land

becomes more difücult to access or only very marginal areas remain, after which

intensiflcation occurs.

The magnitude of lands useJand cover change is different at the global, national and

regional level. It requires different scale maps or spatial resolutions. The more global

the view the more simple is the land uselland cover category to be analysed and a

smaller scale of map or lower spatial resolution required.

The spatial information of land use/land cover is the most important part of the spatial

planning process. The importance of spatial data on spatial planning process in

Indonesia is supported by the government regulation No.20l2000, which states that

spatial planning must be supported by appropriate scale at each spatial planning level;

National, Province and District.

5l
 Chapter Three

THE STUDY AREA AND SPATIAL DATA

3.1 lntroduction
This study conducts an analysis of land use/land cover change in the downstream and

strrounding areas of the Ci Tarum watershed. The Ci Tarum watershed is located in

'West o
Java province. Geographically, the area lies between 106 57' - 107" 60'E and

5" 54,- 7o18' S. Administratively, the Ci Tarum watershed contains all or part of the

districts (Kabupaten) Bandung, Sumedang, PurwakartA Karawang, Bekasi, Bogor and

Cianjur (see Figure 3.1). The upper watershed has a hitly and mountainous topography

whilst the downstream area is coastal alluvial plain.

Sumedang

Sukabumi

Cianjur

Garut

Å
20 0 ã] 40
æ Kllonôlars

Figure 3.1. Study area as part of the Ci Tarum Watershed

52
This area was covered by two Landsat TM scenes 122164 and 122165 þath/row)'

Scene 122164 covered the downstream and surrounding areas and scene 122165

available
covered the upper watershed. Scene 122164 is the only satellite data that are

for 5 consecutive yeafs. Therefore, the area of study was focused especially in the

downstream and surrounding areas of the Ci Tarum watershed consisting

of all or part

of the five districts (Kabupaten) Bekasi, Bogor, Karawang, Purwakarta, and Subang

(Figure 3.2). This is an area of 503687 ha'

This chapter describes the social and biophysical environmental conditions in the

selected study area aflecting land use/land cover change and explains
the availability

ofspatial data for the area.

3.2 Selection of StudY Area

the selected
Since satellite data were not available for the whole Ci Tarum watershed,

area of this study is only the downstream watershed and its surroundings'
The region of

being rice fields and

interest is mainly lowland, with the dominant land use/land cover

settlement. This area has been a prime rice-producing region of Indonesia (World

Bank, 1990) and also includes fast-growing industrialization and urbanization

associated with the expansion of Jakarta, a city of 9 million people (Figure 3'3)' Three

large reservoirs have been constructed in the region; one is located in study area

(namely the Jatiluhur reservoir). The other two are located in the upper Ci
Tarum

watershed - namely the Ci Rata and Saguling reservoirs. Jatiluhur is the largest

paddy/rice field
reservoir and is a multipurpose reservoir, which supports irrigation for

53
downstream and as part of a hydroelectric power generation and supply system for the

whole of Java and Bali, especially the growing mega city of Jakarta.

Bogor

Kabupatên boundary
Sludy area boundary
A
Downstream of Citarum \^Æ 20 0 20 Kilometers
Jatiluhur reservoir

Figure 3.2The five Kabupatens and Jakarta within the study area

,/ \./ Adm¡nrslration bqundary

'.' . Mâih roads A
.: 500 10010 Kìlom dêF
o -- 500

-
Figure 3.3. The New Town and Industrial estates development in Jakarta and
surrounding areas. (Source: adaptedfromJayadinata, 1999: p 18a)

54
3.3 Socio-Economic Gharacteristics

The socio-economic and demographic conditions are generally similar in each

kabupaten of the study area except kabupaten Bekasi that adjoins the special capital

city district of Jakarta. The settlements are characterized by ribbon development along

the roads. Currently, since the government of Indonesia constructed the Cikampek

highway in mid 1985, some real estate and industrial plants, including a golf course'

are located along this highway (Figure 3.4).

Java 59"

Sï41Ê roads

ffi Toll way

use

tæ
'l|ì1,Þj Fo re sl
SÉttìemÊnl
Plantation estete
Flicc fiùl ? cycloo
A

illilll Rice field 1 cycla

I ReEeru oìr
# Fish ¡nnrl
Opan/diy land
ii:"ì:.:.'i 2fl 0 20 Kibm alers
ìiÌ{j.,$ Open/dry lEnd

Figgre 3.4. Land use of 1993 in five kabupaten within the s]udy area and Jakarta city
Source: SARI Project BPPT, 1998.

55
3.3.1 Population and DemograPhY

The total population in kabupaten Bekasi, Bogor, Karawang, Purwakarta and Subang

in 1990 was 9,1 million and 11.3 million in 2000. The population growth rate in

kabupaten Subang is low, 0.93%lyr during the period 1990-2000 (Table 3.1). On the

other hand, population growth rates during the same period in kabupaten Bogor,

Karawang, Purwakarta were high; about 1.33 - 2.40 %lyt and extremely high in

kabupaten Bekasi (about 5.60 %lyr} For the whole kabupaten within study area,

population growth rates during 1990- 2000 was 2.41 %lyr compared with 2.03 % n
west Java and l.4o/o in Indonesia.

Table: 3.1 Populations and Demographic Indicator

lndicators Kab. Kab. Kab. Kab. Total

Bekasi Karawang Purwakarta Subanq
Area {sgkm) 1990 1284.24 3357.92 1578.47 830.01 1864.02 8914.66
2000 1284.24 3204.44 1578.47 830.01 1864.02 8761.18
Ghanqe (%/Y) c -4.57 0 0 0 -0,13
No.of Population 1990 2104392 3736870 1491914 563039 1206664 9102879
1129834
2000 3282238 4233224 1765263 698353 1319264 2
Chanse (%lY) 5.60 1.33 1.83 2.40 0.93 2.41
1990 1639 1113 945 678 647 1021
Population DensitY
(per sqKm) 2000 255ê 1321 1118 841 708 1290
Chanqe (%lY) 5.60 1.87 1.83 2.40 0.93 2.63
of 1990 246934 708879 337229 120466 294171 1707679
2000 864412 1 029895 494158 1 86290 393337 2968092
Chanqe (o/o/Y) 22.73 4.11 4.23 4.96 3.06 6.71

Sex Ratio rn/f t990 1.00 1.03 1.00 1.O2 0.98 0.98
2000 1.01 1.03 1.01 1.O2 1.00 1.01

Change 0.01 0 0.01 0 0.02 0.03

Source : Central Burequ of Statistics ; Compiledfrom Java in Figures, 1990, and 2000,

Table 3.1 also reveals that the number of households (one household consists of 4 to 5

persons) increases more rapidly than the population. In kabupaten Bekasi, and also in

other kabupateq the number of households increased compared to the population

56
 %olyr whereas the
growth rate. ln total, the number of households increased at 6'77
increases' the
population growth rate was 2.41%lyr. As the number of households
increases. Thus, the trend (i'e'
demand for housing units and settlement areas also
signals the pressures from
relative high growth rates of the number of households)

another forms of population dimension. This signal is the need for more spacious

housing units.

3.3.2 EmPloYment

The population aged 10 years or above in all kabupaten

in year 1997 was 8'9 million'

of Of this part of the

while in year 2000 it was 10 million or an increase
72.4Vo.

populatior¡ a total of 4.4 mitlion n lggT and 4.8 million in 2000 were in the labour
high level of employment or
force (Table 3.2). This firther comes with a relatively

employment ratio; 3.9 million working and 0.5

million seeking job in 1997 which

job'
increased in 2000 by 4.5 million working and 0.37 million seeking

At the kabupaten level, the role of the agriculture sector in offering employment

opportunities for the ever-increasing labour force

is still very significant; despite the

fact that in the heart of the area near to the capital

city, industrialization is progressing

near to Jakarta there are services'

intensively. In kabupaten Bekasi and Bogor, located

trade, hotel and restaurant (tourism), which offer considerable

employment

(manufacturing) as the major

opportunities for the labour force. The role of industry

sector offering employment opportunities can be

clearly seen in the Municipality

force between kabupaten

(Kotamadya) Bogor (Table 3.2). When comparing the labour

municipalit¡ it clearly recognised that labour force for industry

is concentrated in
and

57
the municipality. This indicates that a lot of areas have been changed to industrial or
land use
other commercial use such as for hotels, restaurants and golf. Therefore,

condition in this area changed from agricultural to industrial area'

Table 3.2 Labow Force and occupations in five Kabupaten and

Kodya Bogor Yeat 1997-2000

Total in five kabuPaten Chanqe in Kodya Bogor Chanqe

lndicators
1997 2000 1 997-2000 1997 2000 1997-2000
8878373 10052676 1174303 235638 624264 388626
No.Of Population;=> 10 Years

3903838 4509838 606000 87344 258725 171381

Labour Force: Working
373959 -166931 16218 26786 10568
Seeking job 540890
4883797 439069 10 285511 1 81 949
Total 4444728

Occupation bY Sector;% of
Workino labour force: ,l
152,67 6 3 -2
a. 147,41
3,39 -2 0,97 0,59 -0,38
M and 6,21
76,6 -11,32 25,24 22,6 -2,64
I
87
1,71 -1 ,4 3 -0,36
EI Gas & Water Su 3,11
Construction 31,95 21,28 -10,67 I 6 1 -2,82
21
oÃ
111,41 129,27 17,86
Trade Hotel & Restaurants
39,25 52,67 13,42 61 0,15
Tran
5,47 2,52 -2,95 1 I -1,02
h Financial lntermediaries
59,89 -7,08 27 27 0,2
Services 66,97
0 -0,3 1 -0,16
Others 0,3
100 1 1
Total 100
Source: West Java in Figures, 1997 and 2000.

3.3.3 The Land Tenure.

Between the two most recent agricultural censuses

(i.e. 1983 and 1993) there was a

especially in Java'
substantial decrease in the area of agricultural land in Indonesia,
number of agricultural
However, Table 3.3 shows that there was an increase in the
land in Indonesia was
households over the same period. The reduction of agricultural

hectares per yeaf' On the

l.lO-milion hectares in ten yeafs or approximately 110,000

58
other hand, the number of agricultural households increased by about 2.49
million or

of this
approximately 249,000 households per yeaf. The most important consequence

change is in the average size of landholdings. In Indonesia, the average size of

landholding per household has declined from 0.98 hectare in 1983 to

0'83 hectare in

1993; approximately a decrease of 1531%'

Table 3.3. Changes in Agricultural Land and Households in Indonesia

1983 - 1993

1993-Agrieultural Chanp

1. Area (Million Hectare)

5.42 4.41 -1.01
Java
t.22 t.07 -0.15
& Nusa T
s.66 s.42 -0.24
t.64 l.78 0.14
Sulawesi
2.23 2.r9 -0.04

0.3 8 0.4 0.02

0.r7 0.18 0.01
J
t6.72 15.45 -1.01
otal
Number of Households

11,L08 1 488
Java
18,693 21,183 2,490
Indonesia
Source: Central Bureau Statistics

or 101 000 hectares

In Java, the decrease of agricultural land was 1.01 million hectares,

per year, whereas the number of agricultural households increased

by 448 000 over the

from 0'58 hectare

decade. This has caused a reduction of average size of landholdings,

in 1983 to 0.47 hectare in 1993; or 18'97%o decrease'

In
Apart from farm size, the distribution of land holdings needs to be considered'
less than 0.50
Indonesia, in 1993, the number of households with size of land-holdings

hectares was 10.94 million or 51.63o/oof the total agricultural households. In Java, the

59
distribution was even worse. In the same year, numbers of households
with farm size

less than 0.50 hectare was 8.10 million or 69.85%o of total households
in Java. The

by 795'000 or
number of households of this category (farm size<0.50 ha) has increase

approximately llo/o compared to that in 1983'

productivity of the
The type of agriculture land use is also significant since it impties
most productive
land and, thus, farm incomes to the owÏrers and operators' One of the
facilities'
agricultural lands is wetland or irrigated rice farms supported by inigation

Wetland rice in Java has decreased from 2.95 million hectares

in 1983 to 2.5l million

hectares in 1993; a reduction by 14.87% or 43,800 hectares per ye¿r' The reduction in

agricultural land, particularly in Java, gives more space for the development of

purposes (Table 3'4)'

industrial sites, settlement, infrastructure and facitities, and other

Table 3.4: The change of Agricultural land to other land use:

Year 1981-1999 (Ha)

Change from
West Java Central Java* East Java Indonesia
land to:
15847.00 22405.45 20872.87 72994.12

10771.00 t9s9.45 2870.42 t8262.38

Industrv
8080.99 26200.28 14177.67 50640.16
Land
6952.00 1338.00 5 1577.00 61599.85

55.00 8.33 30.23 106.66

Land
1248.82 1349.93 6316.25 9004.23
rter
6809.00 10000.02 10626.90 33885.23

49763.8r 63261.46 t06471.34 246492.63

Total
Source: Dep. Of Agriculture and National LandAgency
Note: * including Yogtakartø

where rice
In the case of wetland rice freld, the developments have led to a situation
such as horticulture'
farming no longer is competitive, even compared to other crops

Near the major growth centres, wetland rices farms have

given away to horticulture

60
crops (vegetable, flowers, etc.) and for animal husbandry (especially
broiler)' on the

other hand, industrializationdoes not guarantee more employment

opportunities for the

people.

3.3.4 Existing Land Use

in the discussion
Land use and land tenure are two aspects that should not be neglected

since the two are strongly related. Information on present

land use in the study area can

reveal the decrease of the

be seen in Table 3.5. The figures in Table 3.5 and Figure 3.4

area of inigated wetland rice field (i.e. technical, semi technical and non technical
hectare
irrigation); from349437 hectare in 1989 to 288108 hectare :lr'1997 and272604
The areas of dry land
in 2000; which confnms what has been discussed in land tenure.
compound) have
farm (i.e. dry field, grassland, temporary fallow etc' except housing
in land used for house
also decreased. The reduction of this area is due to the increase

increased from 127729

compounds and yards. The area of housing compounds has

hectare in 1937 to 166761hectare :[r-1997 and to 176679 in 2000'

6t
Table 3.5 Existing Land use in the five kabupaten within study area (Ha)

Iechnical Inigation 198185 196409 l 88580

WETLAND
technical 37682 27758 259t5
on technical 62339 29946 20t91
Rain fed s0003 33965 37918
1228 30 0

Total 349437 288 108 272604

Com and Yæds 127729 166761 176679
DRYLANI)
field 126015 146747 t29961
Grass land 9t542 26r3 2513

Iemporary fallow 4753 3218 2696

vate wood Forest land 5 1696 26988 36456
State Forest t40236 9slt4 9655',1

Estates/Tree CroPs 82453 60324 s 8003

Swamps 11550 8855 t7029

16040 26t72 t6564
0 66509 6015I
Total stt778 60330 1 5966t6
Source: Central Bureau of Statistics; West Java in Figures, 1989,1997 and2000.

I 400

1200

1 000
(t
E +RlcêFieH
o 800 +Houæ comPound
8 +St.b Forott
È- +Pond & W€ü¡nd
(! 600
o

400

200

0
Year'89 Year'97 Year'98 Year'99 Year'20
Yoar

Figure 3.5 Main Land use in West Java: 1989-2000

62
3.3.5 lnfrastructure

The provision of infrastructures (i.e. roads, electricity,

water supplies) is a prerequisite

3'6 shows the road

for development, in particular to promote industrialization. Figtre
several big cities surrounding
network that passes through the study area and connects
Jakarta' This road
of the study area, i.e. cirebon, Bandung cianjur, Sukabumi and
development process and land
accessibility will have an influence on accelerating the

use allocation.

Bogor

¡o

Provlnclol Road
RoEd À
TollWay
way n 0 20 Kilometers
Adm lnlstratlve boundarY

the study area'

Figure 3.6. Roads networks in five kabupaten within

63
 in the
Table 3.6.below provides also some information on the infrastructures situation

study area. Road length in whole area has increased by 2334.94 kilometres
or 54'82 %ó

in an eleven-year interval (V/est Java Statistic ofüce). During the 1989 to 2000 time
Karawang
interval, there was a significant increase in road length in kabupaten Bekasi,

and Subang. This increase related to construction of the Jakana - Cikampek highway
(Toll way); this had an effect on construction of connecting roads in this surrounding

kabupaten.

Table: 3.6. Infrastructure in five kabupaten within the study Area'

Kabupaten Kabupaten Kabupaten Kabupaten Total

Infrastructures Period Kabupaten
Bekasi Bogor Karawang Purwakarta Subang

ROADS (km)
Type of surface:
1219.00 3t3.45 307.32 238.41 2344.88
Asphalted 1989 266.70
1288.84 768.23 330.80 722.61 39s3.30
2000 842.82
298.20 169.00 79.97 228.37 949.54
Stoneigravel's 1989 174.00
121.70 315.32 552.09 72.20 139.90 t20l.2l
2000
113.00 s5.15 1s0.70 65.81 s80.24 964.90
Soils 1989
55.67 t204.48 0.00 63.30 1439.75
2000 116.30
Condition:
489.80 319.81 252.63 238.04 13 5 s.08
Good 1989 s4.80
r33.96 479.45 663.4r r14.70 1s0.ó6 t542.r8
2000
770.20 150.79 111.91 228.38 1547.98
Moderate r.989 286.70
404.11 77.83 204.06 342.2s 1248.3r
2000 220.06
2r3.00 20r.75 68.60 76.6s 5 80.25 It40.2s
Damaged 1989
719.6s r 139.1 26.99 32.40 276.20 2t94.34
2000
t6.20 110.60 93.9s 1191.00 0.00 I4tI.75
Heavily damaged 1989
t02.34 t756.57 51.84 2s7.50 2295.40
2000 t27.r5

CTRICITY
t53791 66001 90828 t0975 data 101
Installed
1617586 940905 5 38961 I o data 353 1069

218795 389661 11 8004 184148 No data 910608

No.Of Customers 1997
444867 5 13348 283 183 335770 No data t577t68
2000
954057 115 No data 6217420
Electricity sold 1991 2176458
2r9r483 1607769 12768 No data 77s5984
(KwÐ 2000 2679847
1997 and 2000
Source: Compiledfrom llest Javq in Figures, 1989,

64
Rapid increases of the road length suggest rapid development of industrial and
settlement areas located in the back-up area surrounding the highway (Karawang and

Bekasi) where the price of land is still relatively cheap compared to areas close to

Jakarta. Only with new road construction can the accessibility of those back-up areas

be enhanced meaningfullY.

The development of industry and services in the metropolitan area of Jakarta and other

cities within the study area such as Boor, Bekasi and Karawang, also places more

pressure on the demand for electricity. Factories, hotels, restaurants, resort areas'

recreational facilities and places, shopping centres, education centres, etc. demand

more electricity as compared to when such activities have not been intensiffing. Table

3.7 shows that the number of electricity customers signifrcantly increased in all

kabupaten except kabupaten Subang (no data available) from 910608 in 1997 to

1577168 in 2000, or an increase of 73.20 %:[rIa3-yeat time interval.

3.4 Biophysical Gharacteristics

The physical environment conditions such as topography, physiography, soil and

climate, influence the variations in land use/land cover type both temporally and

spatially. Robinson (1994 pp 4) argued that physical conditions such as soil,

topography and physiography are the primary constraints of agricultural expansion and

other land use practices. Therefore these parameters have to be known and recognised

in order to analyse land use/land cover change in the study area.

65
3.4.1 The Physiography and Soil type

Java consists largely of a series of east-west trending physiographic regions: a northern

based coastal belt of alluvial plain, a central belt of volcanic mountains with a series of

inter-mountain basins, and a southern belt of limestone hills. The study area is a part of

north coastal belt of West Java. In more detail, this area can be divided into four types
*Rolling hill and Volcano".
of physiography: "FIat", "Rolling hi11", 'Volcarto", and

The flat area is mostly located within elevation between 0- 499 m above sea level; for

the rest, elevation varies from 500 to 1499 and > 1500 m. The volcanic area is mostly

located within elevation between 500 - 1499 mabove sea level (Figure 3.7 and 3.8).

,¿ a

s<dl.llñ I

| çl@ 50

Figure 3.7. The Relief of v/est Java (source: adapted fromHugo, 1981: p 20)

66
 Java 5E"

f---'l Ksbupåtên boundBrY

Physiography
j:i:l:i:i::: Flqt
À
Rûlling Hill
Volcsnic and Rolllnü Hlll æ 0 20 Kibm dêts
¡ûlltli Volranic

Figure 3.8. The Physiography in five kabupatens withinthe Study Area

- (Source: BAKOSURTANAL, 1992)

Figure 3.9 shows the soil type in the five kabupaten within study area. Distribution of

soil type in this area was influenced by active volcanos in the south part of study area;

the type of soils are varied and underlined sedimentation of tuffvolcanic. The alluvial

soils mainly lay on the flat area in the north part, which was influenced by tuffvolcanic

sedimentation. Some andosol lay on the mountainous ateainthe southern part of study

area, while latosol and podsolic are dominant in the central of the study area.

67
 Java 56"

-l
Soilfype
Kabupeten bündary

iiii:i:i Alluvial associaton A

i Andosol association
GreY Regosol
Grumusol 20 0 æ KilomdËrs
Latosd assæiation
äiìllìi Poosotic -:-

Figure 3.9. Soil type in five kabupaten within the study area
- lsources: RePPProt, 1986)

3.4.3 Climate, lrrigation and Growing season

In general, .urest Java receives rainfall under the influence of the West Monsoon

(November to April) and East Monsoon (May to october). Due mainly to topographic

effects, the rainfall is lowest along the northern coast and tends to increase as one

moves inland to the more elevated areas. Figure 3.10 shows rainfall zones
of the region

differentiated according to the average number of dry (less than 60 mm of rainfall)

and

wet (more than 100 mm) months recorded (Hugo, l98l)'

68
 ü

llol¡tto !

Figure 3.10. The Rainfall Regimes in'West Java.

(Source: Hugo, 1981: P 2I)

The rainfall distribution throughout the year is an important limiting factor to

agriculture and hence density of settlement. The monthly rainfall regime throughout

the year shows the dry and wet season cycle in the study area (Figure 3.11). The wet

and dry season affect the growing season of rice. In the areas that are not irrigated, rice

has only one growing season. The growing season cycle is improved to two cycles with

the support of an irrigation system. The Jatiluhur reservoir has as its main function to

irrigate the agricultur aI area on the northern part, which is mainly wetland rice fields,

and maintains two growing cycles in this area (Table 3.7).

69
 Wet season Wet
@
?m
È
-uÆ Dry season
I
ï0
¡m
'Eo
I

0
JFN FB [/lAR AFR iVE JN JI- ÊCÐ SF O(r IIP EB
tvtrtì
Figure 3.11. The monthly rainfall regimes in the study area
(Source: Meteorology and Geophysics Agency)

Table 3.7.The Growing season schedule

Cycle I
Bloc Inigating Planting
I October lst November lst January I st
II October 16st November 16st Ianuary 16st
tII November lst December 1st February lst
N November 16st December 16st February 16st

Cycle II
Bloc Inigating Planting
t March lst April lst Iunelst
tI Mæch 16st April l6st Iune l6st
il April lst May lst July lst
IV April l6st Mav 16st Iuly 16st
Source: fafifuhu¡ ComPønY, 1998

the rice
There are three main irrigation systems from Jatiluhur reservoir that inigate
part of
field area in the study area. One is the east Tarum, which irrigates the eastern

rice field afea $rch as in Subang and Indramayu. Another is west Tarum which
fresh
irrigates the western part of rice field area such as in Bekasi, as well as supplying

70
 the rice
water to the Jakarta metropolitan city. The third is central Tarum that irrigates

freld area mainty in the kabupaten Karawang (Figure 3'12)'

Jâva c" rl

Bo0or

Irr ig'atiot t

Citarum river A
Jatiluhur regervoir l<ilem qlere

Adm i nistrative b oun darY

Figure 3.l2.TheCi TarumRiver and irrigation inthe Study area

3.5 The Availability of Spatial Data

3.5.1 Existing spat¡al data

analysis and
The main aims of this study are land use/land cover change detection,

prediction by using remote sensing and GIS techniques' Spatial as

well as temporal

scales have an implication on the magnitude of land use/land cover change analysis'
at the spatial scale
This study is concerned with land use/land cover change analysis

based on the resolution of Landsat TM satellite imagery. The spatial resolution of

data were collected'
Landsat TM is 30 m. Several types of spatial data and non-spatial

7l
 a spatial
constructed and converted into digital format and used as sources to construct

database for further qualitative and quantitative analysis in relation to socio-economic

and bio-physical driver factors of land use/land cover change.

The spatial data were available from several government agencies such as

BAKOSURTAI.{AL (The National coordinator for survey and Mapping Agency),

Department of Forestry, BPN (National Land Agency) and Department

of Mining and

Energy. Each agency has its own form and content of maps. BAKOSURTÆ{AL,
maps for
which is the main agency to provide base Ílaps, for example, Topographical

the whole Indonesian authority at scales l:250 000, l: 100 000, and 1: 25 000' Other

institution such as the Department of Forestry, the Centre of Soil and

Agro-climate

Research, and the Department of Mining and Energy are the institutions
which provide

addition spatial data such as forest cover' soil types, and geology as
well as

physiography. The Department of Forestry concentrates on forest mapping

at scales of

1: 25 000 and l: 50 000, while BPN creates land use maps at scales of l:25 000 and 1:

at least every
50 000. These maps consists of land use type which have to be updated

5-year (silalahi, lggg). Table 3.8 shows the agencies that are
involved in spatial data

collection.

Several spatial datasets that were used in this study consists of administrative

(Kabupaten and Kecamatan) boundary, roads, rivers and irrigation of west Java

sourced from BAKOSURTANAL topographical map for the

yeat 1993 with a scale of

l: 250 000; soil type and physiographic condition maps from RePProt
for the year

72
1986 with a scale of l: 250 000; land use type for the yeat 1993 source from sARI

project BPPT with a scale of 1: 250 000'

Agencies in Indonesia
Table 3.g. partial list of Land Resources Inventory and Mapping

Armed Forces Survey and Mapping Service

Army Topographic Service (PUSURTAABRI)
Naval Hydro-Oceanographic Service
Air Force Aerial Photographic Service

BAKOSURTANAL (National Coordinating Agency for Surveying and

Mapping)

Directorate of GeologY
Geological MaPPing Division

Meteorology and Geophysics Agency

LAPAI{ (National Space and Aeronautics Institute)

Ministry of Agriculture
Centre of Soil and Agro-climate Research
Directorate General of Estates

Ministry of ForestrY (MOF)

Agency for Forest Land use Inventory

Ministry of Home affairs (I\ftIA)

Directorate General, Agraria
BAPPEDAS

Ministry of Public Work (Nß\Ð

Directorate Generàl for Water Resources Development (DGWRD)
Directorate General of Human Settlement (Crpta Karya, DGCK)
MPW MaPPing Centre (PUSDATA)

Ministry of Transmigration
Directorate õeneral for Site Selection Planning and Programming

Pertamina (the national oil production corporation)

Source: World Bank (l eeO).

73
3.5.2 Satellite imagerY

Seven Landsat TM images path/row (122164) taken at the Dry season and the late Wet

season were available for the years 1989, 1990,199I,7992,7993,1995

and 1998 (Table

in this area
3.9). It is difficult to get imagery for the same times due to the cloud cover

by
- the acceptable cloud cover is < 10% (NRCT, 1997). These images were provided

BPPT (The Agency of the Assessment and Application of Technology) of the

Government of Indonesia under Land use/land cover (LULC) Indonesia

case study

program (SARCS, 1997)

Table 3.9 Landsat TM data

No Path/Row Level Date of acquisition Climate condition

1 122164 5 6-Jun-89 Dry Season
2 122164 5 9-Jun-90 Dry Season
rJVet Season
J 122164 5 25-May-91
4 122164 5 30-Jul-92 Dry Season

5 122164 5 12-Apr-93 Wet Season

6 122164 5 24- Dry Season

7 122164 5 12-lan-97 Dry Season

This satellite data were acquired at NRCT's (National Remote Sensing Centre
of

Thailand) receiving station in Bangkok at level 5. Level 5 means radiometrically

corrected with two-dimensional re-sampling to a map projection' All images were

data at
rectified to Universal Transverse Mercator (UTM-48) coordinates using GPS

selected sites within the study area. Landsat TM has 30x30 m spatial
resolution and 16

day temporal resolution.

The satellite image for the Indonesian afea can also be acquired from the LAPAII
receiving
ground station in Pare-pare South Sulawesi. This ground station is capable of

and processing SpOT and Landsat TM data. These satellite data were useful to provide

74
information of natural fesources on spatial basis, which overpass the same area every
data is
2g days for SpOT and 16 for Landsat satellite. The limitation of these satellite

that the spatial resolution is less that of terrestrial mapping.

3.6 Gonclusion
The aim of this chapter was to describe the social and biophysical environmental

background of the study atea, as well as the availability of spatial data

for the purpose

of the Ci
of the study. The selected study area is the downstream and its surroundings
an area of
Tarum watershed. This area was a prime rice-producing region and includes

the mega city

rapid industr talp¿atios.and urbanization associated with the expansion of
imagery on
of Jakarta. The other reason to select this a¡ea is the availability of satellite
is available in 5-
an annual basis. Only the scene 122164 covers the study area, which

was needed to
yeafs consecutive and free of cloud cover. The annual satellite image

detect annual land use/land cover change in the study area.

land use/land
The region of interest of this study is mainly lowland, with the dominant

cover being rice fields and settlement. Physiographycally, this area

mostly consists of

a"flat" physiographic with elevations between 0 - 500 m above sea level'

population growth is relatively high compared with population growth at the national

other kabupaten'
level. Population growth in Kabupaten Bekasi was higher than in the

people'
Kabupaten Bekasi was located near to Jakarta the city of 9 million

75
Several types of spatial data and non-spatial data that related to this study were

collected from different agencies. The satellite images were acquired for the
dry and

1997 '
the late wet season on annual base years: 1989,1990,1991,1992,1993,l995,and

76
 ChaPter Four

LAND USE/LAND COVER GHANGE DETECTION;

using Remote sensing and Gls Techniques

4.1 lntroduction
factors'
Land use/land cover dynamically changes as a result of social and biophysical

use and is often

The term land cover has commonly been used in association with land

the focus of remote sensing analysis. Remote sensing provides a

variable source of

and enables the

data from which updated land cover information can be extracted

detection of land use/land cover change.

Land use/land cover change when one land use is converted to another
is relatively

easy to detect, in contrast to detecting land use transformations. The difüculty of

detecting land use/land cover transformations is due to the mixes

within land cover

types. For example, it is diffrcult to differentiate residential and commercial areas

land use'
within urban land use or between food crop plantations within agricultural

There are many image processing techniques (called digital change detection
The selection ofan
techniques) that are used to identiff land use/land cover changes'

appropriate technique needs to take into account landscape structure, spectral

signatures and radiometric corrections'

can
In Indonesia, which is a tropical area,land use/land cover is often dynamic and

change due to weather or climate conditions and seasonal growing cycles' Sharkov
such
(1998, p. 175) stated that the various components of the physical environmental

77
as climate, soil, vegetation type, weather and relief contribute significantly to land

use/land cover change. V/ithin these conditions, land use/land cover can be difficult to

detected using conventional techniques such as terrestrial mapping or swveying on

short time scales. Although these techniques may achieve more detail, in terms of

spatial resolutior¡ they need more time to collect information over large areas. Land

use/land cover may change significantly while the mapping is still being undertaken.

Therefore, techniques that can quickly identiff and detect land use/land covsr, such as

remote sensing analysis, can be very useful.

The spatial accuracy of land use/land cover detection is also related to the spatial

resolution of the imagery used. The spatial resolution of Landsat TM imagery is 30 m,

which can differentiate the detail of urban land cover from vegetation and agricultural

land covers (Johnson et ql., 1989).

This chapter presents a review of digital change detection methods as well as the

results of land use/land cover change detection in the study area. The results of the land

use/land cover change are then used as an input into the land use/land cover change

model described.

4.2 Land use/land cover change detection techniques

There are a number of methods for obtaining land use/land cover data, such as

terrestrial surveys or mapping and remote sensing techniques. Terrestrial surveys or

mapping are direct mapping technþes that have the advantage of obtaining more

78
detailed information and can produce large-scale maps. The main disadvantage of this

technique is the time required to cover large areas. Other techniques include mapping

from aerial photography from aircraft or by remote sensing from airborne or satellite

platforms. Aerial photography has proven to be a popular approach to the description

and analysis of land use/land cover and land use/land cover change, both at alocalized

scale and for broader areas (Foster, 1985; Bryant et al., 1986). Some disadvantages of

aerial photography include timeliness, cost and comparability of photo interpretation

over time when different interpreters are involved. Using remote sensing techniques,

land use/land cover can be recorded over relatively large area and with reduced time

between the completion of data collection and actual field conditions. However, the

disadvantage of remote sensing techniques is that the spatial resolution is generally less

than the result from terrestrial mapping.

In spite of its timitations, remote sensing ptays an important role in the acquisition of

land use/land cover data. Remote sensing techniques enable the collection of land

use/land cover data in a digitalformat and the easy integration and analysis with other

digital spatial databases. Table 4.1 shows the comparison of the terrestrial and remote

sensing techniques to map land use/land cover conditions'

79
 Table 4.1 The Comparison of the Terrestrial, Air photos and remote sensing
Techniques.

Approach Product Advantaqes Disadvantaqes

Terrestrial Direct ground surveY and Map analog or Detailmap with High cost and
techniques mapping Digital high spatial time consuming
resolution, good
for small area
Aerialmapping Air photos interpretation Map analog or Semi-detailmap High cost and
Digital medium spatial time consuming
resolution
Remote Sensing lmage Processing lmage and map large Less spatial
techniques Digital uick, multitem resolution, need
acquisition to be improved
to be pro- by ground truth
for further

Source: From various literature: Jensen, 1996, Lillesand, 1987 and Lindgren, 1985.

Over the last two decades, remote sensing techniques have been increasingly

developed and intensively used for natural resource inventories including land use -

Iand cover monitoring (Weismiller et al., 1977; Angelici et al., 1977; Brera and

Shahrokhi, 1978; Gordon, 1980; Jensen 1986; Quarmby and Cishnie, 1989; Green ¿/

al., 1994;Liet a1.,1993). Since the first Earth Resources Technology Satellite (ERTS-

l), later called Landsat 1, was launched tn 1972, activities dealing with the natural

resources inventory and land use/land cover change detection using remote sensing

techniques have been conducted. Landsat 1 (MSS) had a spatial resolution of 80 m.

with this spatial resolution, land use/land cover types smaller than 80 m x 80 m could

not be recorded. Later, other natural resources satellites with finer spatial resolutions

were launched. Spot 1 was launched in 1985, Spot 2 in 1990, and Spot 4 in 1998, all of

which have four bands, two in the visible infrared, and one in infrared with a spatial

resolution of 20 m and abroad visible infrared (panchromatic) with spatial resolution

of l0 m. Landsat 5 TM was launched in 1982 and Landsat 7 TM was launched in 1999

80
with a 30 m spatial resolution and has eight spectral bands including thermal infrared

band and panchromatic (Foster, 1985: p. A\. The resolution element (spatial and

spectral) of these later systems is more comparable with the size of features requiring

detection. V/ith this improvement, land use/land cover change detection and
monitoring using remote sensing can be more accurately conducted.

Many studies have been conducted regarding the application of satellite imagery data

to monitoring land use/land cover change. Conventional classification methods such as

Maximum Likelihood using the original Landsat TM5 band combination (4,3 and 2)

has been extensively used to identiff land use/land cover in the Southeast Asia region -

e.g. in Indonesia, Malaysi4 Philippines, and Thailand (SARCS, 1997). All of these

studies result in land use - land cover change maps at the scale of 1: 100 000. Another

conventional approach is Band Ratios, which have been used to assess the capabilities

of Landsat TM in regional mapping in West Africa (Thenkabail, 1996). This study

demonstrated that finer scale maps can be produced from Landsat TM at scales

between 1: 50 000 and 1:200 000 (Thenkabai\1996: p. 105).

Image enhancement techniques such as Normalised Difference Vegetation Index

(NDVÐ and Principal Components have also been used to identiff land use/land cover

types (Byrne et al., 1980; Fung and LeDrew, 1985; Jadhav et al., 1993; Qi, 1993;

Elvidge and Chen, 1995; Salem et al., 1995;Lyon et al., 1998; Li and Yeh, 1998; Fung

and W. Siu, 2000; Zttanet a1.,2002;Yte et aL.,2002; Helmschort et a1.,2002; Filho e/

al.,2002).

81
Table 4.2list some of the various techniques used to identifu and monitor land use/land
cover using remote sensing data.

Table: 4.2.The various techniques to identiff and monitor land use/land cover
condition

Techniques Purpose Data Reference

1 Supervised Classification Fruit Tree inventory Landsat TM Gordon ef a/.,1986

band 3,4 and 5.
Texture extraction
procedureB3and4
2 Maximum Likelihood Grass Mapping Landsat MSS &TM Jadhav et al. ,1993
Classification (MLC) IRS-IA
I Visual interpretation of Monitoring Land use Landsat MSS,TM Ram, 1993
FCC band 1,2,3 and 4 Change IRS LISS-1
combination
4 Vegetation lndex, EVI Detect the current Landsat TM Salem, 1995
NDVI and MLC land cover classes
5 Unsupervised of band Regional Agroecosystem s Landsat TM Ihenkabil ef a/.,1996
ratio 514,413,411 Mapping
and combined with
oriqinal band 4,3,5
6 Photo interpretation and fo assess of Air photos and Salami, 1999
MCL of lcc 4,5,2 original environmental change Landsat TM
data
7 Post +lassification and Land cover change and Landsat MSS and Merkus et a1.,2000
Band differencino Deforestation SPOT
I MCL Examine the differences Landsat TM and Hashiba et a1.,2000
of the TM and MSS on MSS
Wetland classification
o Unsupervised Fuzzy Land cover detection Landsat TM Deer, 1999
Classification
10 H ierarchical unsupervised To discriminate urban land Landsat TM Ward ef a/., 2000
cover tvpes
11 I ntegrating spectral and Land cover Mapping Landsat TM Berberoglu et a1.,2000
texture using Neural
Network
12 The Vegetative cover Deforestation and [/toDls 250 Zhan,2002
Conversion UCC) llood detection
13 Curve theorem by using Detecting land cover Landsat TM Yue,2002
NDVI Chanqe
14 NDVI Veqetation cover Landsat TM Helmschort 2002
15 PCA Mapping and monitoring Landsat TM Filho, 2002
deqradation area

82
4.2.1 Digital change detection of land use/land cover

Land use/land cover change detection aims to measure different features of land

use/land cover categories at two diflerent dates. Using large-scale aerial photography

at two dates of the same geographic area enables land use/land cover change to be

detected, but visual interpretation has some limitations and it is diffrcult to undertake

subsequent analysis due to the non-digital nature of the data. The digital change

detection method uses digital imagery to correlate and compare two sets of images to

identiff changes. Four main conditions are recoÍrmended for successful change

detection analysis (Jensen, 1986: p.236). The researcher should:

o Have a prior knowledge of the study area characteristics (cultural and

biophysical)
o Know the precision with which the multiple-date imagery is registered. The
spectral response that is not associated with land use/land cover should be
minimized
o Have an understanding of the limitations of change detection techniques.

Singh (1989: p. 990) categorizes the digital change detection techniques into two

categories. The first is the independent analysis at different dates, and the second is the

simultaneous analysis of multi-temporal datasets. These two categories of digital

change detection are still considered the basis of digital change detection and are called

pre- and post-classification approaches (Metternicht, 1999; Lunetta, 1999). All

methods can be classified as one of these two approaches. In the post-classification

change detection approach, two different date images are independently classified and

labelled. From the result, the areas that have changed are extracted. The pre-

classification spectral change detection method is based on the assumption that

83
different spectral signatures over time represent a change in land surface conditions.

This technique involves the transformation of two original images to a new single-band

or multi-band image in which the areas of spectral change are highlighted (Yuan et al.,

1999: p.22). The advantage of the pre-classification approach is the ability to eliminate

the effect of mix-pixels, while the advantage of post-classification is overcoming

difficulties in change detection associated with the analysis of images acquired at

different times or by diflerent sensors.

The disadvantage of pre-classification is the need for image co registration and the

difficult in discriminating "change" and "no change" areas. Hence, the disadvantage of

post-classification is highly dependent on the individual classification of each different

image.

Recentl¡ many studies have used the combination of pre-classification and post-

classification approaches in a hybrid approach that complement each other in

overcoming the complexity of land use/land cover change detection (e.g. V/ard et al.,

2000; l./loy et a\.,2001; Stefanov et a1.,2001; V/ilson and Steven, 2002).

Table 4.3 shows the summary of digital change detection categorised by pre-

classifrcation as a simultaneous, post-classification as an independent, ild Hybrid

approaches. All of these techniques can include three basic approaches: pixel level

approach, feature level approach and object level approach (Deep, 1999). Pixel level if
detection refers to a numerical value attached to each band of each pixel in an image

such as digital number (DN). Feature level refers to some transformation of pixel level

84
values, and object level refers to symbolic or thematic labelling or classification of the

data. From this point of view pre-classification can be classified as feature level or

pixel level differencing, whilst most of the post-classification belongs to an object level

differencing , and ahybrid approach is a combination of feature level or pixel level and

object level differencing.

The common digital change detection techniques that were categorised as Pre-

classification, Post-classification and Hybrid are reviewed in the following sections.

85
 Table: 4.3. Comparison of digital detection techniques Categorized by Independently, Si

Categories Method Procedure/ Approach Required Advantage

Independent Post-Classification co-registration nature of the
lassification correction cover is identified
Obj ect level differencing correctly the
areas

Simultaneous Band Differencing Subtraction the image method

one date from that cost
another( Direct band to level for massive
subtraction) rectification processmg
level differencing Radiometric

Band Rationing before rectification method

subtraction: Radiometric cost
RVI,TNDVI,SAVI for massive
processmg
level differencing
level

Principal Component the original rectification method

into a component Radiometric cost
new lmage for massive
level differencing processmg
classifi-
level accuracy
data pre-

86
Table 4.3 continues
Categories Method Procedure/ Required Advantage
Multi date ingle Classification rectification and Efficient method
lassification two or more dataset correction Direct change detectio
class change accurate identification
analysis change area
ect level differencing normalization or
sformation image

vector ector deference rectification and Eflicient method

& direction) Radiometric correction Able to monitor
of the pixel eed reference for subtle changes
two dates different calculating change vector within individual land
or feature level differencing. cover class.

Hybrid Change Simultaneous Spectral rectification

and identification Radiometric
and independently correctly the
lassification areas
and Object level

Source: Jansen. Rf, ß96 andvarious Literalure

87
4.2.1.1 Pre-classification

4.2 .1.1.1 lmage Differencing

In image differencing techniques, two images at different dates are spatially registered

and subtracted one from the other. The corresponding pixel values are subtracted to

produce a new image that represents the change between the two dates.
Mathematicall¡ this method produces a new change image with the same number of

bands as the input images. Pixels presenting a signiflrcant radiance change can be

expected to lie in the tails of the distributions of the different image, whereas the

remaining pixels should be grouped about the mean. Images used in this technique

needs to be corrected to reduce radiometric differences between the input images.

Image differencing is simple to implement and has been widely used in a variety of

applications. Brera and Shahrokhi (197S) applied this technique for the analysis of

desertification in the Sahara region, and Vogelmann (19SS) investigated the detection

of change in temperate forests. Criticism of this method comss from Riordan (1980 in

Sing, 1989: p. 992). The method is sensitive to image miss-registration, to the

existence of mixed pixels, and to radiometric differences between the input images.

This method fails to speciff the mechanism of types of information loss (Singh, 1989:

p. 993). For example a forest with a radiance value of 170 in band 4 on one band and

140 on the second date shows a change of 30. This nurnber could come from the

difference between radiance values of 70 and 40 which does not represent the radiance

value of the forest pixel. Thus, there is a potential loss of information with the use of

simple image differencing technique. The other disadvantage is deciding the threshold

boundaries between change and no-change pixels displayed in the histogram (Jansen,

88
1986: p.2a\. However, the normalized or standardized image differencing of two or

more images from different dates that have been radiometrically corrected have the

potential to be used for digital change detection. In a resent stud¡ Fung (2000) was

successful in monitoring the change of environment quality in Hong Kong city by

using image differencing. Image difÊerencing was also determined to provide the best
'Wiregrass
results for detecting change in Long life Pine ecosystems (Houhouis and

Micher, 1996).

4.2.1.1.2 lmageRatioing

This method calculates the ratio of the values of corresponding pixels of registered

images at different dates. The basic concept of this approach is that if the intensity of

reflected energy is nearly the same in each image then this indicates no change. If
change occurred in a particular pixel, the ratio is expected to be either higher or lower

than the ratio in the no-change areas. Similar to other techniques, the standardization of

some data or radiometric correction between dates may be necessary.

This technique has not been as intensively investigated as image diflerencing (Nelson,

1983: p. 130a). Todd (1977) used this method to determine urban change in Atlanta,

Georgia. As a result 91.4% of land cover change was correctly identified (Todd,1977

in Nelsoru 1983: p. 1304). Fung (1988) tested the accuracy of this method by
comparing it with image differencing and principal components methods and found it

effective in detecting change from cropland to construction with a producer's accuracy

of over 9}Yo,but it was difhcult to detect change from cropland to residential (Fung,

1988: p. 14s3).

89
Howarth (1931 : p.287) found that the use ofband ratios using data from different dates

produced an enhanced image of the study area. He used this technique successfully to

detect flooding area in Mamawi Lake, Canada. Rachman (1997) concluded that the

band ratioing technique is suitable and efficient for the study of detecting altered rocks,

soil-moisture, drought, vegetation, and canopy change.

Ratios of band 3 and band 7 of Landsat TM for example, is best suited for recognizing

roads or other man-made features that appear in a lighter tone due to their relatively

high reflectance in the red band (band 3) and low reflectance in the mid-infrared

(band7) (Lillesand, 2000: p.52Ð. With a strong local knowledge and correct selection

of the band ratio, the change detection using this method can be successfully

conducted. This technique is well established and is stilt being used for purposes such

as the pre-classification process as an aid to improving thematic and spatial accuracy

(Watson et a1.,2001).

4.2.1.1.3 VegetationlndexDifferencing

This technique also has been established for a long time and is the most common

method used to identify vegetation. In the vegetation index differencing technique, the

digital spectral radiance values can be analysed independently on a band-by-band

basis, or in combinations of two or more bands. The most commonly used band

combination techniques involve the calculation of vegetation indices. These have been

developed on the basis of the strong absorption of red and strong reflectance of near-IR

by vegetation. The formula of vegetation indices could be a simple combination, or

band ratios, or a combination of both. It has been found that the ratio of near-IR to red

90
reflectance is significantly correlated with green leaf biomass (Tucker, 1979). There

are many vegetation indices in use (Qi, 1993; Elvidge and Chen, 1995, Salerr¡ 1995)

but the most common is the Normalized Difference Vegetation Index (NDVI)

(Eastmar¡ 2001), given by the ratio of near-IR minus red reflectance and near-IR plus

red reflectance.

It is important to note that ratioing two spectrat bands negates the effect of any
extraneous multiplicative factors in sensor datatl:øit act equally in all bands (Lillesand

and Kiefer, tggT). Singh (1939) underlined that the band ratio technique might

enhance noise that is not correlated in diflerent bands. Qi et al.(1993) stated that it is

useful especially when satellite data are available at high temporal frequencies, wffi

fine spatial resolutior¡ and with atmosphere correction of data, they advocate a

correction to account for first-order soil background effects when using vegetation

indices. In a study on the use of vegetation indices for change detectior¡ Lyon et al.

(tggg) found that the Normalised Difflerence Vegetation Index (NDVI) produced

results superior to those of other vegetation indices. Nelson (1983) examined this

method qualitatively in a study of gypsy moth defoliation in Pennsylvania. The result

indicated that of three methods tested (i.e. image diflerencing, ratioing, and vegetation

index difterence) the vegetation index difference most accurately delineated forest

canopy change. Angelici et at. (1977) employed the difFerence in the ratio between

NIR and red band and the thresholding technique to delineate the change area. The

disadvantage of this technique is that it's sensitive to miss-registration and the

existence of mixed pixels. However, this technique could be used on land use/land

cover change detection combined with other techniques (i.e. principal component

91
techniques) to delineate other vegetation signatures. Recently, these techniques have

been intensively used to identiff vegetation and change detection (V/ard et al., 2000;

Fung and Siu, 2000; Lanjeri et a1.,2001). Fung and Siu (2000) conclude that NDVI

values were related to woodland, tall scrubland and high-density urban areas. They

also found the decreasing NDVI values covered a large part of rural area revealing

urban expansion. Ward (2000) found the NDVI band 5 and band 3 can successfully be

used to classifr vegetation cover (woody and non-woody).

2.5.1.1.4 Principal Components Analysis

Byren (1980) stated that Principal Components Analysis (PCA) is a powerful method

lor analyzing correlated multidimensional data. It can be used in such systems to

facilitate the visual interpretation of a mass of data having uniform a priori

significance, by reducing redundancy. Lillesand and Kiefer (1987: p. 655) also stated

that PCA is a powerful data transformation method for information extraction in

remote sensing. This transformation can be used either as a pre-processing operation or

as an enhancement operation prior to visual interpretation of the data.

Principal components transformation is a linear transformation that defrnes a ne%

orthogonal coordinate system such that the data can be represented without correlation.

Therefore the degree of correlation between bands in the original multi-spectral image

can be eliminated and the redundancy of information can be reduced.

Principal Components Analysis (PCA) can also be applied to compress image datasets

from two or more dates i.e. to multi-temporal image data (e.g. Byrne et al., 1980: p.

175; Richards, 1984: p. 35; Ingebritsen and Lyon, 1985: p. 687; Fung and LeDrew,

92
1987; Ceballos and Bottino, 1997; Li and Yeh, 1998; Edge, 2001). There is a high
correlation between image data for regions that have not changed significantly and a

relatively low correlation between regions that have changed substantially. Richards

(1934) used a principal component transformation technique to highlight regions of

localized change, associated with bushfire damage and vegetation re-growth following

fire burns. Singh and Harrison (1985: p. 895) compared the use of a covariance matrix

and a correlation matrix. They found that there is a significant improvement in signal-

to-noise ratio and image enhancement by employing the correlation matrix rather than

the variance matrix. A non-standardued principal components analysis could be

justified due possibly to the differences in radiometric resolution between the spectral

bands of sensor. Fung and LeDrew (1987: p.1657) also examined image enhancement

for change detection based on standardized versus non-standardued approaches. They

stated that principal components analysis is a "scene dependent" technique and the

exact nature of the principal components derived is not known without an examination

of the eigenstructure and visual inspection of the image. However, it is a powerful data

reducing technique, and should be used with a thorough understanding of the

characteristics of the study area to avoid drawing any faulty conclusions. Ceballos and

Bottino (lgg7) conducted a further study on the discrimination of scenes of multi-

spectral imagery. They use principal component analysis to discriminate vegetated

landscapes in Northeast Brazil. This approach also has been used to discriminate

healthy and non-healthy corals in Fiji (Holden and LeDrew, 1998). Li and Yeh (1998)

used principal component analysis to reduce over-estimation of land use/land cover

change in monitoring rapid land use/land cover change and urban expansion in the

Pearl River Delta. A combination of principal component analysis and linear

93
programming methods was conducted to calculate fraction botmds of main seaweeds

and inter-tidal components for mixed pixels (Bajjouk et al., 1998). A recent study that

has been successfully conducted is discrimination of urban and non-urban area using

principal component analysis (Edge, 2001). This study has successfully discriminated
'West
urban and non-trban areas as a basis to predict urban growth in part of Java.

Principal component techniques can therefore be usefully applied to enhance images

prior to image classification in order to delineate change categories.

4.2.1.1.5 Change Vector AnalYs,b

Change vector analysis is the analysis of change based on the vector difference

between the multi-band digital vectors of the pixel on two dates. A particular pixel in

an image can be plotted as a point in this vector space with co-ordinates that

correspond to its brightness values in the appropriate spectral components. The data

values associated with each pixel thus define a vector in multi-dimensional space. If a

pixel undergoes a change from time tl to time t2, avector describing the change can be
defined by the subtraction of the vector at t1 from the vector at t2. This is called the

spectral change vector. It may be calculated from either the original or transformed

(e.g. PCA or NDVI) data, and using either individual pixel or clusters formed by a

spectral clustering or spatial segmentation algorithm. If the magnitude of the computed

spectral change vector exceeds some specified threshold criterior¡ it may be concluded

that change has occurred. The direction of the vector contains information about the

type of change. Malila (1980) introduced this method for the fnst time for detecting

forest change in northern Idaho (Yuan, 1999: p.25).Colwell (1981) used this technique

to detect forest change in the Palouse District of the Clearwater National Forest in

94
Idaho. He used two different strategies of digital change detection: delta-classification

(post-classification) and change vector analysis (Colwell" 1981: p. 3a0). He concluded

that using this two-change detection procedure could significantly affect the estimates

ofamount ofchange.

Lambin and Strahler (I99$ used this method to detect and categorue land cover

change processes. They concluded that change vector analysis combined with a PCA

on the change vectors proved effective in detecting and categotzng inter-annual

change between time trajectories of NDVI data. Cohen et al. (1999) used this

technique as a multi-temporal linear transformation compared with different images

techniques and composite analysis on conifer forest change detection in Pacific

Northwest region. They conclude that a reference image was important for accurate

characteization of forest change. If change vector analysis without a reference image

was compated with image differencing techniques, the image differencing technique

performed better than change vector analysis (Cohen, 1999: p. 99). The disadvantages

of this method are similar to those of other methods such as the raw image differencing

method, in that is that it is sensitive to miss-registràtion, to the existence of mixed

pixels, and to radiometric differences.

4.2.1.1.6 DirectMulti-DateClassification

Direct multi-date classification uses a single analysis of a combined dataset of two or

more datasets to identifr areas of change. The combined data set whether, an original

image or a transformed image, can be classified by a supervised or unsupervised

approach. In the supervised approacl¡ training sets pertaining to change and no-change

95
areas are used to derive statistics to defure sub-spaces of the feature (normally spectral)

space. In the unsupervised approacl¡ spectral classes are determined by cluster

analysis, and subsequent inspection can reveal where changes have occurred. Hoffer

and Lee (1939) used this method for forest change detection and combined it with PCA

to reduce the number of bands. They were able to detect at an accuracy of 90.6Yo for

the fuIl 12 band, and 90.4Yo using 6-band after reduction by principal components

analysis. Muchoney and Haack (1994) conduct a study of change detection for

monitoring Forest defoliation. They conclude that by using multi-date classification

gnder the similar condition, change areas were significantly different. A disadvantage

of this method is that it needs accurate identification of the training areas for each

change class. Other disadvantages are: the data need to be re-processed whenever a

new image is added, it is difficult to label the change class, and there is limited
information on "from-to" changes class. However, it is a simple technique because it

requires only a single classification.

4.2.1.2 Post - Classification Comparison

post - classification comparison is the most common method of detecting change' This

method involves the classification of each image independently, and is followed by a

comparison of the corresponding pixel signature to identifu areas where change has

occurred. The process of classification can be carried out separately by a supervised or

unsupervised approach. It reveals that change has occurred, and also reveals the precise

nature ofthe change, but requires an accurate rectification and classification processes.

The comparison of classified images can be conducted visually or by computers. The

comparison can also be undertaken using a Geographic Information System (GIS).

96
Joyce et al. (1980) concluded that post - classification comparison "appears suitable for

detecting land cover change with Landsat MSS data in sites where large areas of

forestland are being converted to cropland". Fisher and Pathiram (1993) found that

(conventional) post - classification comparison indicated that over 40 Yo of the land

area studied had undergone change, whereas it was known that little change had
actually occurred.

Jensen ¿/ al. (1937) used post - classification comparison with unsupervised

classification of aircraft MSS data for wetland change detection, but achieved limited

success. Martin (1939) concluded post - classification comparison gave better results

than either multi-date classification or principal component analysis for change

detection in the rural-urban fringe. Jensen et al. (1995) reported on a later study also

using post - classification comparison for wetland change detection. The result was that

cattail and cattail mixture (sawgrass) classes could be detected. This technique is

commonly used and combined with image enhancement processes to get better feature

signatures in order to classiff land use land cover change in what are called hybrid

techniques (e.g. Deep, 1999; Ward, 2000 and Madhavar¡ 2000). The disadvantages of

the method are: it needs ground truth information to get precise training areas, it is

dependent on accuracy of individual classifications, and it requires two separate

classif,rcations. The advantages are: it reveals the area that has been changed, it
provides "from-to" change class information and bypasses the difficulties associated

with the analysis of images acquired at different times, in that radiometric

normalization is not required.

97
4.2.1.3 Hybrid

The Hybrid approach combines spectral change identification methods and

classification based methods in various ways to minimize effors in land cover change

analysis. This method can substantially reduce the errors of commission

(misidenti$ing no-change and change pixels). The combination between

radiometrically normalized image differencing and post-classification comparison, for

example, can be used in land use/land cover change detection. Radiometrical image

differencing may be used to identifr the area of significant spectral change, and then a

post-classification comparison applied within areas where spectral change was detected

to obtain class-to-class change information.

Pilon et at. (1988) developed this method further and applied it to the detection of

changes in a northwestern Nigeria semiarid environment. They conclude that this

hybrid approach can reduce commission effors. Recent studies have been conducted

using new approaches to classification such as fiizzy logic and neural networks. These

techniques can be used to classiff land use/land cover categories on digital change

detection techniques . Fvz-zy logic was successful in recognizing thresholds of land use

by using the fivzy membership principle. ANN, as well as the frrzzy membership

approacl¡ was successfully used to classiff land use categories when the spectral

differences between classes are subtle (Berberoglu et aL.,2000: p.386).

Berberoglu et al. (2000) used hybrid methods of land cover mapping in the

Mediterranean basin with the integration of spectral and textural information using

neural networks. The ANN improved the accuracy of classification results compared

98
with the maximum likelihood approach. Berberoglu concluded that when using

spectral data alone and using spectral and textural data, the accuracy of ANN was

greater than that of maximum likelihood. Ward et al. (2000) used hybrid methods as a

hierarchical approach to detect urban land-use-land cover change. They combined

spectral change detection to separate impervious surfaces and exposed soil by using

band ratio (NDVI and Mineralogical Ratio) and post-classification by unsupervised

classification. The result improved the accuracy from 75 % to 83 Yo. Madhavan et al.

(2001) detected spatial growth of the Bangkok Metropolitan area using post-

classification comparison techniques combined with a vegetation-impervious-soil

model. The vegetation-impervious - soil model has been used to distinguish urban from

non-urban morphology.

An advantage of hybrid methods is overcoming the mixed pixel problem and providing

information on "from-to" change classes information. The disadvantage is that two

processing steps are needed (spectral detection and post--classification comparison).

4.2.2 Selection of the appropr¡ate detection and identification techniques

The focus on change detection in this study was to provide an approach that could be

used to detect and identifr land use/land cover change in the study area. Considerations

taken into account in selecting an appropriate detection and identification techniques

for this study were:

99
 l. The study area is in the Tropics area that has varied physical environmental

conditions such as climate, soil vegetation type and weather as well as growing

cycle.

2. Conventional change detection based on the post-classification comparison of

an original band combination constrained to get correct land use/land cover due

to some noise or disturbances.

3. Transformation techniques such as Principal Component Analysis (PCA) and

Normalized Difference Vegetation Index (NDVÐ are techniques that produce

new transformed images and can be used to replace the original band image to

detect land use/land cover change.

A Post-classification comparison approach was selected to detect land use/land cover

change in the study area. This change detection was based on Maximum Likelihood

supervised classification of transformed images principal component analysis (PCA)

and NDVI.

Principal Component Analysis (PCA) transformation was selected based on the

following corsiderations: 1) it may reduce effects of atmospheric differences (Bryne,

1980); 2) it has been successfully employed in land cover change analysis (Fung and

LeDrew, 1987; Holden and LeDrew, 1998; Ceballos and Bottino, 1997; Lie and Yeh

1998); 3) it has been successfully demonstrated to separate crop types cover without

performing data pre-processing (Jiaju, 1988) and 4) it has been successfully used to

recognise spectral features in order to distinguish human settlement (Edge, 2000).

100
NDVI transformation was the most coÍrmon vegetation index that was successfully

used to identifr vegetation cover (Tucker, l979,Eastmar¡ 2001).

4.3 Methods

The main method of digital change detection using remote sensing in this study is

shown in Figure 4.1. A combination of image processing, image enhancement and

image classification were used to identifu land use/land cover categories. Change

detection was conducted through post classification comparisons. The following

sections describe the integration of remote sensing and GIS operations to detect land

use/land coveÍ change in the study area.

101
 Multitemporal
geocorrected
TM

TM Composite
432

NDVI

Training
Classification Area

I 2 3 4 5 6 7 8 9

Check
No
Yes

LU-LC
Classification
Map

Change dstection,
LU-LC identifi cation and analysis
GIS Database

Note:
1 = Forest 4 = Open land 7 = Dry land
2 = Plantation 5 = Settlement 8 = Fishpond
3 = Rice fîeld 6 = Rice field unplanted 9 = Water/reservoir

Figure 4.1. The integrated model of detectior¡ identification and analysis of land
use/land cover change.

r02
4.3.1 Digital change detection techniques

Three main processes Pre-processing, Image Enhancement and Classification are

cofltmon image processing procedures to detect and identifr land use/land cover from

the satellite imagery. These procedt[es are described as follows.

4.3.1.1 lmage Pre-Processing

The satellite image could not be used directly to interpret or delineate physical

features. It needs to be processed, including corrections to eliminate the effors

introduced during image acquisition. The ideal or perfect remote sensing system has

yet to be developed (Jenser¡ 1936). Remote sensing devices have constraints such as

spatial, spectral, temporal, and radiometric resolution. Consequently, this has an

influence on data quality during data acquisition. Therefore, it is usually necessary to

pre-process the remotely sensed data before analysing it in order to remove some of

these errors.

Pre-processing aimed at the correction of distortions, degradations, and noise introduce

during the imaging process (Moik, 1930) will produce a corrected image that is as

close as possible, both geometrically and radiometrically, to the radiant energy

characteristics of the original scene. Radiometric and geometric errors are the most

coÍrmon types of errors encountered in remotely sensed imagery.

4.3.1.1.1 Geometric Correction

Most digital satellite images have already had systematic error removed. Landsat

satellite data for example, which are acquired through the EOSAT Company, have

103
consists of spatial enhancement, radiometric enhancement and spectral enhancement.

These enhancements cover image reduction and magnification, contrast enhancement

and edge enhancement. Table 4.4 below shows the Image enhancement types and their

function:

Table 4.4. The type of Image enhancement and function

Type ofEnhancement Algorithm Function

Image Reduction &, Image Reduction, To reduce the original image data set
Magnification down to one or a few manageable
images.

Image To improve the scale of display for

Magnification visual interpretation purposes or
occasionally to match the scale of
another image.
Contrast Enhancement Linear CE To improve the contrast of
(CE) Non-linear CE reflectance property among different
Rationing material. Therefore easy to
Filtering diflerentiate the featwes in the
surface.
Edge Enhancement Linear EE To enhance the edges surrounding
(Ee¡ Non-linear EE various object or features interest.
Source: from variotn image processing modules (Erdas Imagine, EkMapper and ldrisi)

4.3.1.2.1 PCA and NDVI Transformation

These two transformation techniques enable one to enhance and provide new set

images, which have strong spectral signatures of built-up areas or settlement and

vegetation cover. Using the image-processing module in ldrisi32 that provides PCA

andNDVI transformation algorithms, a new set of PCA and NDVI transformed images

were created

105
4.3.1.2.2 Filtering

The result of image classification produces some noise due to a scattering of individual

or small groups of pixel. Filtering processes can be used to enhance the spatial features

of the image. A filter creates a new image by calculating new values using a

mathematical operation on the original cell value and its neighbours. The mode or

focal majority filter, which assigns the most coflrmon value to the centre pixel, is

commonly used to remove very small areas from a qualitative image. All of the images

classified in this study were enhancing by mode filtering to remove isolated pixels.

4.3.2 Land use/land cover Classification

Multi-spectral classification is the method most often used to extract the information

from remotely sensed data. This procedure assumes that imagery of a specific

geographic area is collected in multiple regions of the electromagnetic spectrum and

that the images have good registration (Jenser¡ 1996). Therefore the physical features

of the earth, in particularly land use/land cover types, could be extracted through the

spectral reflectance properties ofthose objects or features analysis.

The supervised maximum likelihood classifier was selected in this study, which has

been commonly used in many studies. Three to five training areas for each land

use/land cover (LULC) category in this study were selected based on the object feature

of each category in the field. These categories are:

1: Forest 4: Open land 7 : DrY land

2:Plantation 5:Settlement 8:Fishpond
: Water/reservoir
:
3 Rice field 6: Rice field unplanted 9

106
4.3.3 Land use/land cover change detection

This study used a post-classifrcation comparison approachto detect land use/land cover

changes in the study area. This approached is based on Maximum Likelihood classifier

of the PCA and NDVI transformed images. Figure 4.2 below shows the algorithm of

this approach.

1989

t *
Post
Max likelihood
classifier
+ Classification
Comparison

1997

AnalYsis
Original band Transformation classification

Figure 4.2. Post-classification comparison approach of land use/land cover change

detection and identification

The first step is to transform the original image by using principal component (PCA)

and normalized different vegetation index (NDVI). The second step is to combine

these two sets of images and then classiff using supervised Maximum Likelihood

classification techniques. Finally, these two image representing difFerent dates were

compared with each other to detect and identiff any land use/land cover change. GIS

107
techniques, including overlaying and cross tabulatiorU were used to calculate the area

ofeach land use/land cover category that has been changed.

4.3.4 Accuracy Assessment

The Kappa index of agreement (KIA) has been selected to assess the accuracy of the

classification result. This method compares two images; one image contains the

interpreted land cover map while the second image contains the result of ground truth

investigation. It is mostly used in the post-classification assessment of land cover

classifications derived from remotely sensed data (Eastmar¡ 2001). This operation

creates an e¡¡or matrix that tabulates the different land cover classes to which ground

truth cells have been assigned. Output also includes column and row marginal totals,

effors of omission and commissior¡ an overall effor measure, confidence intervals for

that figure, and a Kappa Index of Agreement (KIA), both for all classes and on a per

category basis.

4.4 Result digital change detection and identification

in the study area.

4.4.1 lnitial recogn¡t¡on

Some physical characteristic of the study area have been described and dispþed in the

previous chapter. Physiographically, the study area is flat and consists of alluvial plain

deposits. Most of the land use/land cover conditions in the study area have been

modifred through human settlement and agriculture practices. Population pressure has

108
resulted in the expansion of the agriculture area, especially for wetland rice fields as

well as settlements.

As a result of initial image analysis using simple original band combination (FCC TM

band 4 red, TM band 3 green and TM band 2 blue), the initial identification of land

cover was recognued. There were significant land cover changes as shown in

Figure.4.3 a,b,c, and d).

Figure 4.3a
TMComposite 432
1989
l.Kabupaten Bogor
2.Kabupaten Bekasi
3.Kabupaten Karawang
4.Kabupaten Subang
S.Kabupaten Purw akarata

r09
 Figure 4.3b
TMComposite 432
1990
LKabupaten Bogor
2.Kabupaten Bekasi
3.Kabupaten Karawang
4.Kabupaten Subang
5. Kabupaten Purw akanla

Figure 4.3c
TMComposite 432
1995
l.Kabupaten Bogor
2.Kabupaten Bekasi
3.Kabupaten Karawang
4.Kabupaten Subang
5 .Kabupaten Purwakarata

110
 Figure 4.3d
TMComposite 432
1997
l.Kabupaten Bogor
2.Kabupaten Bekasi
3.Kabupaten Karawang
Subang
5 .Kabupaten Purwakarata

Four images that have been taken for the same season (dry season): June 1989 Fig.

4.3a, June 1990 Fig. 4.3b, August 1995 Fig. 4.3c and June 1997 Fig.4.3d show the

different land use/land cover conditions in the study area. Red represents vegetative

cover (mostly rice field arca); light blue represents open or arable land and dark blue

represents wetland and water. As seen on those images, the light blue area in 1990 is

slightly larger than in 1989, indicating that open and arable land has increased.

Vegetation is the dominant cover in this area in both 1989 and 1990 (see in kabupaten

Karawang no. 3 and kabupaten Bekasi no. 2). This is very different when compared

with conditions in years 1995 and1997.In 1995 arable and open land (light blue area)

was larger than in L997. The other interesting feature is the vegetation (red area). This

area shifts between the north and south parts of Karawang no 3. In 1995 (Figure 4.3c)

vegetative area w¿Ìs located in the north part, while n 1997 (Figure 4.3d) vegetative

111
atea was located in the south part of Karawang. If this area is rice field, it easy to
understand the shifting of this area as it could be related to the local cycle of growing

season in this area.

From this initial analysis it can be seen that land cover in the study area is dynamic

and changes over time. However the investigation of this change is still at a broad

category level, for example, settlement areas are still not clearly recognised.

4.4.2 lmage Enhancement

Table 4.5 shows the eigenvector matrix as well as the loading matrix from principal

component transformation for 1989 (see Appendix Chapter IV for other years). The

total eigenvalue indicates that the first principal component generally contains 85 - 95
per cent variance. This means the first component contains the majority of the

information. On the other hand, the loading matrix expresses the degree of correlation

between each component and the original bands. The highest correlation is between

component I and TM batd4 (Table 4.5).

tt2
Table 4.5: Principal components analysis transformation TM 1989

Comp I Comp 2 Comp 3 Comp 4 Comp 5 Comp 6

o/o var. 90.79 s.50 3.24 0.29 0.13 0.04
Eigen value. 2680.96 162.51 95.71 8.68 3.72 1.19

eisvec.l 0.45 0.60 -0.54 -0.38 -0.01 -0

eigvec.2 0.10 0.23 -0.02 0.31 0.00 0.91

eigvec.3 0.13 0.42 0.07 -0.01 -0.40

eisvec.4 0.64 -0.62 -0.33 0.25 0.20 0.00

eiqvec.5 0.57 0.03 0.62 -0.17 -0.51 0.00

eigvec.6 0.20 0.18 0.45 -0.1s 0.00

Comp I Comp 2 Comp 3 Comp 4 Comp 5 6

TM Band I 0.93 0.31 -0.2t -0.04 0.00 0.00

TM Band 2 0.83 0.49 -0.03 0.15 0.00 0.16

TM Band 3 0.76 0.5 8 0.08 0.26 0.00 -0.05

TM Band 4 -0.23 -0.09 0.02 0.01 0.00

TM Band 5 0.97 0.01 0.20 -0.02 -0.03 0.00

TlVt Band 6 0.90 0.20 0.39 -0.04 0.14 0.00

The strong correlation in this table shows that component 1 can represent the entire

original band and that the most representative band is band 4. The eigenvector value

within the matrix show that principal components 4 and 5 have a consistently high

value and correlate with eigenvector 3 and 6 for whole year, which is related to

settlement and other features that are sensitive to soil moisture content. Others

principal components did not demonstrate a consistently high value except principal

component 6. Principal component 6 has a high conelation with eigenvector 2, which

is related to coastal sedimentation and this feature is not included in land use/land

cover category inthe studYarea.

Although other components such as PC 1, PC 2 and PC 3 didn't have high values

consistently correlated to a certain band, these principal components (especially PC I

and PC 2) contain of the majority of information in the image. Therefore, all principal

113
components, except principal component 6, were to be combined with the NDVI image

Io anaLyze land use/land cover in the study area. Figure 4.5 shows spectral features of

principal component 5 (4.4 a) and spectral features ofNDVI (4.4b) from TM 1989'

4.4a

Figure 4.4a.The spectral features of Principal component 5 (PC 5)

The dark black colour in Figure. 4.4a represents settlement features and the light
white

colours on Figure. 4.4b rcpresent forest and other vegetation feattrres. Through the

combination of pCA and NDVI transformed images, the settlement and vegetation

types could be separately recognized.

tt4
 4.4b

Figure 4.4b. The spectral features of Band Ratio NDVI (4.5b)

4.4.3 Land use/land cover Classification

The spectral signature information from the new two transformed images indicates that

land use/land cover types have a strong feature signature. Nine categories of land

use/land cover in the study area have been classified by using a supervised maximum

likelihood classification approach. Table 4.6 shows the description of the nine

categories of land use/land cover (LULC) in the study area'

115
 Table 4.6The Land use/land cover Classification Scheme

LULC category Definition

1. Forest Secondary forest mainly broadly leaf wood
2. Plantation Plantation estate; Orange, Tea, Clove and Rubber
3. Rice field WeUdrainage rice field (sawah)
4. Open land The arable or bare land
5. Seftlement Built-up area; housing, industry and manufacturing area
ô. Rice field unplanted WeUdrainaqe rice field (sawah) unplanted
7. Drv land Grass land and dry land aqriculture
8. Fishpond Pond mostly near to the coastal area
9. Water Water body and Reservoir

The results of final classification fromthe image 1989 to 1997 after filtering are shown

in Figures 4.5ato 4.5 g.

use.land c,over 1989

Forest
Plantation
Rice field
Open/bare land
Settlement
Rice unplanted
Ä
Dry land 10 0 10 20 Kilometers
Fish pond
VvstenReseNoir

Figure 4.5a. The final classifred image of land use/land cover 1989

116
 Land use-.land cover 1990
Foresl
Plantatìon
Rice Reld
Oponrberc lend
Settlement
Rice unFlanted
Ä
Dry land 10 0 10 20 Kilometêrs
Fish Þond
Water/RêseNoir

Figure 4.5b. The final classif,red image of land use/land cover 1990

Land useland c,over 1991

Forsst
Pl€ntat¡on
Rico field
Opeil/þurt larrl
Ssttiêmsnt
Rrce unplant6d
Â
DJy land 10 0 10 20 Kilomoters
Fìsh pond
Water/RssÊrvo¡r

Figure 4.5c. The final classiflred image of land use/land cover 1991

tt7
 Land u6eland cover 1992
Fùrosl
Plantetion
Rìce field
Openibare land
Settlemont
Ricâ unplanted
À
Dry lsnd 10 0 10 20 Ktlomqters
Flsh pqnd
Water/RosBNoir

Figure 4.5d. The final classified image of land use/land coYer 1992

Land uss-land cover 1gg3

Forêst
Plsntatìon
R¡ca n€ld
Op6nf:,üfe lsnd
Settlement
Rìcê unplsnted
A
Dry lEnd 10 0 1n ?0 Kilomsters
Fish pond
W6ter/Ressruo¡r

Figure 4.5e. The final classified image of land use/land cover 1993

118
 Land use-land çover 1995
Forest
PlåntÉtion
Rìce field
Openibare land
Settlsment
Rice unplántêd
A
Dry land 10 0 10 20 Kilometers
Fish pond
Watar/RåseNo¡r

Figure 4.5f. The final classified image of land use/land cover 1995

Land uso-+and cover 1gg7

Foresì
Plar¡tction
Ric6 nBld N
Openñare lend
Settlemènt
Rirê unplarìtêd
A
Dry land 10 0 10 20 KitometÊrs
FiÉh pond
v1/etêtRêsefvoìr

Figure 4.5g. The final classified image of land use/land covet 1997

119
Table 4.7 shows the accuracy assessment of years 1989 and 1997. Overall accuracy

was 90 %o for 1989 and 89Yo for 1997. From this table it can be seen that commission

enor (misidentiffing no change and changed pixels) in 1989 was an average of 0.15

and omission (misidentifying change and no change) error was an average of 0.06. In

1997 commission error was 0.20 and omission error was 0.08. The result was an

overall accuracy of 90Vo and 89%o, which is good when compared with 85% as a

standard minimum accuracy of land use/land cover classification from satellite data

(Janser¡ 1996).

Table 4.7: Summary of Kappa index of agreement (KIA) assessment

Clãss vear 1989

Forest 0.23 0.03 0.96
Plantation 0.05 0.02 0.97
Rice field 0.03 0.07 0.87
Open/bare land 0.03 0.03 0.95
Settlement 0.01 0.06 0.93
Rice unplanted 0.r8 0.23 0.73
Dry land 0.73 0.0s 0.94
Fishpond 0.07 0.05 0.93
Water/reservoir 0.00 0 1

Average comm. error 0.15

Average om. error 0.06
Overall KIA 0.90
Commission error
Forest 0.31 0.03 0.96
Plantation 0.08 0.02 0.98
Rice field 0.01 0.06 0.89
Open/bare land 0.06 0.15 0.84
Settlement 0.51 0.10 0.90
Rice unplanted 0.07 0.28 0.70
Dry land 0.75 0.08 0.92
Fishpond 0.06 0.03 0.97
Water/reservoir 0.00 0.01 0.99
Average comm. effor 0.20
Average om. error 0.08
Overall KIA 0.89

t20
4.4.4.The land use/land cover change detection

Once the classifications have been generated, it is possible to carry out a theme
comparison between two classifications. This operation is performed by cross

tabulation on a pixel-by-pixel basis resulting in a complete matrix of annual changes in

land use/land cover within the study area'

Figure 4.6 shows the annual fluctuation of land use/land cover in the study area- From

this figure it can be seen that, in general all land use/land cover categories fluctuate

annually. Some categories increased and others are decreased. The land use/land cover

category that fluctuated the most was dry land. Settlement constantly increased in the

five consecutive years (1989-1993) and increased significantly between 1995 and

1997. On the other hand, rice fietd was tending to decrease and rice field unplanted

increased but decreased significantly from 1995 to 1997; fishpond and water/reservoir

remain constant.

300000

250000 *Forest
*Plantataion
200000
G -#Ricefield
QPg¡ l¿¡d
IU 150000
o -..#Settlement
100000 -Ë{- Rice unplanted
#Dryland/Grass
50000 *FFish pond
r
Lî: -l[-Water/Reservoir
0
1989 1990 1991 1992 1993 1995 1æ7

Year

Figure 4.6. The arurual land use/land cover change in the study area

r2t
More detailed analysis of land use/land cover change detection in annual 5-year

consecutive (1939 to1993), 2-year time interval between 1993-1995 and 1995-1997,

and 8-year time interval between 1989 and 1997 n the study area ate explained and

discussed as follow.

4.4.4.1 Annual Land use/land cover change from 1989 to 1993

The cross tabulation of annual land use/land cover classiflrcation generates a land

use/land cover change matrix (in percentage and hectares) within the study area

(Appendix Chapter IV). Tables 4.8a to 4.8d show the summary of major land use/land

cover changes in the study area.

Table 4.8 a: The summary major changes in land use/land cover category
1989 I 1990 (dryldry season)

Categorv Change Conversion shift

Open land 850/o increase (27943 ha) from rice field
Dry land J7o/o de,crease (18954 ha) lo rice field (unplanted) Within rice field (planted)
and rice field (planted) and rice field (unplanted)
Plantation 30% decrease (18999 ha) to rice field (unplanted)

From Table 4.8a it can be seen which major land use/land cover category changed

between 1989 and 1990. Open land increased while dry land decreased. The increase

of open land was the result of conversion from rice field (planted), and decrease of dry

land and plantation was due to conversion to rice field (planted) and rice field
(unplanted). Both 1989 and 1990 images were acquired in the same season (dry season

at the harvesting time-Table 3.7 n Chapter 3). In the time interval 1990 to 1991 the

t22
land use/land covers that changed significantly were rice field and dry land. Rice field

decreased and dry land increased (Table 4.8b).

Table 4.8b: The swnmafy major changes in land use/land covef category
r990lt99t
Category Change Conversion shifr
Rice field 630/o deqease (116924 ha) to dry land and rice
held unplanted Within rice field
and rice field (unplanted)
Dry land 247Vo ncrease (78565 ha) from rice field (unplanted)
and rice field (planted)

The decreasing of rice field (planted) and increasing of dry land in this time interval

was due to the different seasons. The 1990 image was acquired in the dry season at

harvesting time and the 1991 image was acquired in the wet season at planting time. In

the 1990 image (harvesting time), rice fields were being planted while in 1991, rice

fields have been harvested and are beginning to grow new plants (planting time).

Therefore the rice field þlanted) area was greater in 1990 than in 1991. Rice field

(planted) in 1990 was converted to dry land and rice field (unplanted) in 1991. This

phenomenon also was found when one analyzes change between 1992 arnd 1993;

which is 1992 was dry season or harvesting time and 1993 was wet season or planting

time. The change pattern in this time interval was the same as in time interval 1990 -
l99l rice field (ptanted) decreased and dry land increased (Table 4.8c). The decrease

of rice field was due to conversion to dry land or rice field (unplanted)'

t23
 Table 4.8 c: The summary major changes in land use/land cover category
19921 1993 (dry/wet season)

CatesorY Change Conversion shift

Forest 48olo increase (6931 ha) from plantation
Rice field 30olo decrease (41975 ha) to rice field (unplanted) Within rice field
and dry land (planted)
Open land 6IYo detease (32093 ha) lo rice field (planted)
Dry land 125%o incr ease (434 1 5ha) from rice field (unplanted)
and rice field (planted)

In the opposite direction of the change between dry (harvesting time) to wet (planting

time), is the change between 1991 and 1992 as show in Table 4.8 d.

Table 4.8 d: The summary major changes in land use/land cover category
199I I 1992 (weldry season)

Categorv Change Conversion shift

Forest 37%o decrøase (8320 ha) to plantation
Rice field 101% increase (69991 ha) from rice field (unplanted) Within rice field
and dry land (unplanted)
Dry land 69olo decrease (75608 ha) lo rice field (unplanted)
and rice field (planted)

Rice field (planted) in wet season was less than rice field (planted) in dry season. In the

\Met season of 1991, most of the rice fields were under preparation to grow new plants

and many rice fields were appearing as dry land or rice field (unplanted). Therefore in

this time period (1992-1992), rice fields (planted) were seen larger than dry land or rice

freld (unplanted). The increasing and decreasing of rice field (planted) is related to

decreasing and increasing of dry land and rice field (unplanted). This is the cycle of

change among these categories between different growing seasons (harvesting and

planting times).

124
The other interesting feature was the change in forest category. In the dry season,

forest area was smaller than in the wet season (see Table 4.8c and 4.8d). This

phenomena is related to the phenomenon of the "leaf on and leaf off' of broad leaf

reflectance (Yuan et al., 1999). In the dry season lots of leaves fall from broad leaf

plants, so at this time the reflectance of leaf decreases, while in the wet season the

reflectance increased from the leaves that have re-grown. Therefore decreasing and

increasing forest does not representing deforestation in this area. It was an

effect of the dry and wet seasons and could be a classification error due to a mix

between forest and broad leave plantation during these seasons.

4.4.4.2 Land use/land cover change in 1993 to 1995 and 1995 to 1997

Wet and dry season phenomena of land use/land cover changes can be continually seen

in the 2-year time interval (Table 4.9a). From Table 4.10U it can be seen that forest

and dry land were decreasing and rice field (planted) was increasing; this is similar to

the patterns of change from wet to dry season between l99l and 1992.

Table 4.9a: The summaly major changes in land use/land cover category
19931t995
Category Change Conversion shifr
Forest 4IVo de,crease (8933 ha) to plantation
Rice field 22%o increase (21209 ha) from rice field (unplanted) Within rice field
and dry land (unplanted)
Dry land 75%o deqease (58591 ha) to rice field (unplanted)
and rice field (planted)

Another 2-year interval detection is in the dry-to-dry season between 1995 and 1997.

In this time interval, the change pattern was different from the annual detection. In this

time interval settlement increased, while rice field (unplanted) decreased and dry land

t25
increased. The increasing of settlement areas related to decreasing of rice field

(unplanted) (Table 4.9b).

Table 4.9b: The summary major changes in land use/land cover category
1995 I 1997 (dryldry season)

Catesory Change Conversion shift

Settlement 245o/o increase (28979 ha) from rice field( unplanted) Within rice field
Dry land 23lo/o increase (45275 ha) from rice field (unplanted) (unplanted)
and rice field (planted)
o/o deuease (128877 ha) to rice field þlanted),
Rice field 50
unplanted settlement and dry land

In dry-to-dry season between 1989 and 1990 settlement did not increase, while in dry-

to-dry between 1995 and 1997 settlement was significantly increased. It means since

1995 in the study area a lot of land was converted to settlement.

4.4.4.3 Land use/land cover change in the 8-year interval 1989-1997

Table 4.11 shows a summary of land use/land cover change between 1989'1997 (the 8-

year interval). The change of land use/land cover in this time interval should represent

the real conversion or change ofthe land use/land cover in the study area after 8 years.

t26
 Table 4.10: The suÍìmary major changes in land use/land cover category
1989 I t997 (dryldry season)

Category Change Conversion shift

Open land 40% decrease (12930 ha) lo settlement and
dry land
Rice field t3%o decrease (21919 ha) to rice field (unplanted) Within rice field
(planted) and
And open land rice field (unplanted)
Settlement 81% increase(I8262ha) From rice field (unplanted)
ooen land and dry land
Plantation lTVo deqease (11109 ha) lo forest
Forest 57%o increase (6494 ha) from plantation

Within the 8-year interval there are many land use/land cover categories that changed

(Table 4.10). Open land and rice field þlanted) decreased and settlement increased,

while plantation decreased and forest increased. In this time interval it was also

recognized that forest increase related to decreasing of plantations. The increasing

settlement in this time interval related to the conversion from rice field (unplanted),

open land and dry land, while decreasing of rice freld þlanted) related to conversion to

rice field (unplanted). From this change trajectory, the sequence of land use/land cover

change in this aÍea caîbe recognized. Rice freld (planted) was converted to rice field

(unplanted) and open land, then rice field (unplanted) and open land was converted to

settlement.

There is no change or conversion from forest to rice field or settlement. Forest change

was related to plantations: if plantation decreases then forest will increase (confuse

classification). The mix between forest and plantation was recognized in the field when

ground truthing was conducted. In the study area, there were some broad leaf
plantation areas such as rubber and teak wood plantations that 3 or 5 year later will

t27
appear as forest on the image. These plantations were located in kabupaten Karawang

in the northern part of Jatiluhur reservoir.

4.4.5 Land use/land cover change identification and re-

classification

The change detection of nine categories of land use/land cover in the study area has

resulted in the various shifts and conversions among categories, especially in annual

time intervals. Change detection on the 8-year time interval shows no annual

variability except for rice field (planted) and rice field (unplanted). Rice field (planted)

and rice freld (unptanted) fluctuate due to different growing cycle. V/ith this condition

it was difficult to recognize the permanent change of rice field as well as open land

aÍeas. It was recognized in the freld that rice field þlanted) and rice held (unplanted)

have the same function as rice field. Open land and dry land categories are the same
as

land which is unused, or under preparation for construction or for dry land agriculture.

It is necessary, therefore, to regroup these two-land use/land cover categories in order

to recognize the change of rice field and open/dry land within the 8-year time interval.

Figure 4.7 shows land use/land cover after reclassificationinto seven categories. Rice

field (planted) and rice field (unplanted) were regrouped as rice fields. open land and

dry land were regrouped into open/dry land. From this figure it can be seen that rice

field area is the largest and was the dominant land cover in whole of the study area.

Settlement was seen to extend along the road corridor from the west to the east in the

middle of the study area, and open/dry land was located within rice field and settlement

area.

r28
 Lond u¡r 1989
Foro¡l
Phnlrllon
Rlcr fl.ld
Òprn/dry land
,S¡ül¡nirnl
flrh pond 10 0 10 20 Kilomêters
w6ler/feso]volr

L¡nd u¡c 1997

Forost
Plmtrtloñ
Rco flcld
Oprrwry l.nü
S¡llhm¡f{
'Flrh pond 10
'I_z+ 0 10 20 Kilomêtors
Wrtlr/rrsclvolr

Figrrre 4.7.Landuse/land cover in the study afea 1989 and 1997

Source: LANDSAT TM Classification

t29
 Table 4.ll:Landuse/land cover in the study arcayear 1989-1997 (ha)

Types ,ehauee. IolùChanee*)

Forest 113381 2.2s 178321 3.s4 64e41 5',l

Plantation 64063l 12.72 s29s9l 10.s 1 -t 1 1Oel -11

Rice field 29sß21 s8.71 280S661 ss.76 -143661
Open/dry land $4071 r6.s6 84s801 t6.79 t173l 0
Settlement 22ßsl 4.4'1 407971 8.10 r82621 81
Fishpond zlrosl 4.19 2118s1 4.2r 801 0

Water/reservoir sso2l
l.oe s4681 1.09 441 0

Total 5036S71 loc s036871 100 0 0

*) Percentage of change to total area of each category in year 1989

350000
A
300000
250000
tú
.C 200000
(ú
E 1 50000
1 00000
50000 tr 1989
E 1997
0

20000
B
1 5000 E Forest

1 0000
B Plantation
o E Rice field
! 5000
o g Open/dry land
c) 0

-5000
I Settlement
¿ Fishpond
-1 0000
E Water/resenoir
-1 5000
Land use change

Figure 4.8. The land use/land covers change 1989-1997 in study area (a) and Total
change of each category @).

130
From Table 4.11 and Figtre 4.8a and 4.8b. above, the change of land use/land cover

among seven categories between 1989 and 1997 can be seen. Rice field decreased as

well as plantatior¡ while settlement, open/dry land, and forest increased. The decrease

in rice field was related to increasing settlement and open/dry land. It is clearly shown

that there was conversion from rice field into settlement. Settlement increased by 8l %

or 18262 ha, while rice field decreased by 14866 ha or 5Yo. Plantation decreased, forest

increased, and water /reservoir and fishpond was relatively constant. As has been

mentioned, the increase in forest in the study area was due to the effect of "leaf on-leaf

off in the different seasons and confirsion between forest and broad leaf plantation.

4.5 Discussion

The focus on change detection in this study is to recognise land use/land cover change

dynamics inthe study area by using image processing.

There are three common digital change detection strategies: pre-classification, post -
classification and the hybrid approach. Pre-classification is the method that directly

differentiates between two dates' spectral signatures. This approach was based on the

assumption that different spectral signature over time represent a change in land

surface conditions. post-classification is based on images from two different dates that

are independently classified or labetled. The classification process can be based

on the

classifrcation of the standard/original image or a transformed image. The hybrid

approach is based on the combination of pre-classification and post-classification.

131
From the initial recognition based on the false colour composite analysis of the study

area, it was recognised that land cover in the study area is dynamic and changes over

time. However, this initial recognition did not give detailed information of land cover

category due to limitations of spectral signatures on the original images of the false

colour composite. The PCA and NDVI transformation was used to enhance the spectral

signature of the image. PCA transformation enhanced the spectral signatures of

settlement while NDVI enhanced vegetative cover. These two-transformed images

were combined and used for a post-classification comparison approach to detect more

detail in the land use/land cover change in the study area.

This study used post-classification comparison of two transformed (PCA and NDVI)

images for the following reasons: 1) the study area has varied physical environmental

conditions due to the weather/climate conditions; 2) the study area has varied land

cover types; 3) the land use in study area has extensively changed due to the population

and economic growth close to the Jakarta metropolitan atea; 4) PCA and NDVI

combination is an effrcient method and has potential for data processing and

maintaining classification accuracy without data pre-processing.

This approach was successful in detecting land use/land cover change in the study area.

It was found that weather or season and growing cycle strongly influenced the land

use/land cover conditions in the study area. In the classified image 1990 (acquired on

June-the early dry season) forest cover was less than in 1991 (acquired on May - the

late wet season) or within 1992 (acquired on July - dry season) and 1993 (acquired on

Aprit - wet season). Dry and wet seasons influenced the "leaf on-leaf off'

t32
phenomenon on the classified image. This phenomenon was clearly recognised in the

change of plantation and forest categories. In the dry seasor¡ forest cover was less than

in the wet season due to lots of leaves fall from plants ("leaf off'), while in wet season

lots of leaves have re-grown ("leaf on"). Therefore in wet season, forest cover

appeared to be increasing because of a lot of "leaf onl' and also mixing with increasing

"leaf on" of some broad leaf plantation such as rubber and teak wood The decreasing

and increasing forest and plantation in this study could also be an error in classifrcation

due to mix between forest and broad leaf plantation such as ruber and teak wood.

Rice field (planted) and rice freld (unplanted) were clearly recognised and changed

according to growing cycles. In the harvesting time (the month of June - JulÐ, a lot of

rice fields were covered by rice plantation that was ready to be harvested; therefore, in

this harvesting time the identified rice field area was large. In the planting time (the

month of April - MÐ a lot of rice fields were empty, because the land was being

prepared to grow new rice. In this planting time the rice field area \ilas less than in the

month of June-July (harvest time). This phenomenon was clearly recognised in the

field as well as from annual change detection.

From the summaries of annual change detection from 1989 to t993, it can be

recognised that for 1990 to 1991 the area rice fields (planted) was larger than rice field

(unplanted). The 1990 image was acquired in June in the early dry se¿rson or harvesting

time, while the 1991 image was acquired in May in the wet season at the planting time.

Therefore, the changes of land use/land cover within this time interval dry-to-wet

season were seen as conversion from rice field þlanted) to rice field (unplanted) or dry

133
land. On the other hand, change within the time interval 1991 - 1992 wet-to-dry season

was seen as conversion from rice field (unplanted) to rice field (planted). The 1992

image was acquired in July atthe harvesting time, when rice field areas were covered

with rice plantations that were ready to be harvested.

For the long-term time interval detection (the 8-year 1989-1997) as well as the 2-year

time intervals 1993 - 1995 and 1995 - 1997, seasonal variability was not an influence

on land use/land cover change. In this time interval, especially in the 8- year time

interval, it was recognized that some land use/land cover categories have permanently

changed. Settlement was significantly increased, but rice field (planted) and rice field

(unplanted) were still seen to fluctuate due to different growing cycles. Therefore,

regrouping of rice field (unplanted) and rice field (planted) into rice field as well as

open land and dry land into open/dry land, was necessary and resulted in more clearly

recognized long-term conversion of land use/land cover change. Settlement

significantly increased as a result of the conversion from rice field. Forest increased

related to decreasing of plantation. The other categories such as water/reservoir

(Jatiluhur reservoir) as well as fishpond remained constant'

4.6 Conclusion
Land use/land cover was clearly differentiated and could be classified into land

use/land cover categories as follows: 1) Forest; 2) Plantation; 3) Rice field (planted); 4)

Open land; 5) Settlement; 6) Rice field unplanted;7) Dry land; 8) Fishpond and 9)

Water/reservoir.

t34
The trajectory of land use/land cover change in this area \¡ras recognized as follows:

Rice field was converted to rice field (unplanted) and to open land, then rice field

(unplanted) and open land was convert to settlement. There was no conversion from

forest to rice field or settlement.

The change of forest in this study area did not represent deforestation or reforestation,

it was error in classification due to mixes between broad leaf plantation and forest that

related to dry and wet season ("leaf off- leaf on" phenomena). As found in the field, a

lot of some broad leaf plantation (such as rubber plantation and teak wood found

especially in the south part of Karawang near to the northern pafi of Jatiluhur

reservoir), was confi.rsed with forest area.

It was necessary, for long-term detection change analysis, to reclassifu or regtoup the

land use/land cover categories in this study area. This regrouping was important in

order to recognise which land use/land cover had permanently changed. Rice field

(planted) and rice field (unplanted) have the same function in the field as rice field

area, and open land and dry land categories have the same function as a land which is

unused or under preparation for construction or for dry land agricultwe- These

categories have been grouped as rice field and open/dry land.

135
After regrouping, the categories that really change are settlement and rice field.

Settlement increased while rice field decreased. There was no change or conversion

from forest to rice field or settlement.

136
 Chapter Five

LAND USE/LAND COVER GHANGE DRIVER

ANALYSIS

5.1 lntroduction
Land cover in the study area changed on an annual basis related
to seasonal variability

or growing cycle. For example, an area may be rice field (unplanted) in

one year and

detection, it
rice field (planted) the next year. From 8-year or long-term time interval
changed
was recog nzed that the land use/land covers that have permanently
are

settlement and rice field. Settlement increased while rice field decreased' The

year by year' while other

conversion of rice field to settlement increased permanently

temporarily, related
land use/land cover categories such as, open/dry land was changed

related to the
to the weather or seasonal conditions. Forest and plantation was mixed
..leaf on and leaf off' phenomenon aS an effect of weather or Season condition (see
study alea are interesting
chapter 4). Moreover, the land use/land cover changes in this

when analysed in relation to other socio-economic and biophysical factors' Factors

such as population growth, economic growth (income per capita,

GNP or

industrialization) as well as physical factors such as topography

or slope, soil and many

change of land use/land

others, are factors that could have a strong influence in the

as well as proximity to
cover in the study area. Population density, proximity to roads
factors that influence
the crty centre can represent social-economic growth as dynamic
and land system could
land use/land cover change; factors such as slope, physiography

represent static influence factors. The question is to what

extent and which of these

137
 strong influence on land
possible (dynamic and static) factors are the dominant or have

use/land cover change in the study arca?

factors of land use/land

This chapter presents an analysis of the possible influence
The analysis is both qualitative
cover change between 1989 and l9g7 ]rr_the study area'

and quantitative. The objectives of this chapter are:

o To analyzethe relationship between land use/land cover and the dynamic

proximity to urban
and static influence factors such as population density,

as well as physiography
and semi-urban center, proximity to roads, slope,

and land system in spatial dimension'

o To determine the factors that have the strongest relationship on land

use/land cover change in the study area'

5.2 Driver of Land use/land cover change.

can be divided into direct or

The factors that influence land use/land cover change
Direct drivers are associated with
proximate causes and indirect or underlying causes.

the physical environment' such as

activities which directly interact with and modify
and Meyer,lg94)' The indirect drivers
urbanization and agricultural expansion (Turner
and change the land use/land cover
influence how individuals or groups interact with,

lgg4). These are generally more complex as

they are built into the
(Blaikie et al.,
(Adger and Browr¡ 1994; Krummer and
human systen¡ underlying a land use activity

Tunner, 1994).

The complexity of driving forces is highlighted

by complex, interrelated influencing

and climate change' Land use

factors such as policy, institution, land management

138
change does not stem solely from population growth. Several studies argue
that human

(The
population aflects land use change more by human actions than by sheer numbers

National Academy of Science, 2001).

The focus of the driver or influence factors analysis in this study was divided
into static

and dynamic factors. Static factors were defined as factors that might never change,

such as slope, elevation, or physiography, while dynamic factors were defined

as

urban, and
factors that can be changed, such as population growtt¡ population density,

affluence
semi urban areas and transport routes, which represent impact of population,

(PAT). Social and economic factors such as income per capita'

and technology factors

GDp, industrialization and cultural structure, are also factors that can be included as

informatioru
dynamic drivers, but due to the diffrculties and limitation of collecting this

this analysis only includes factors such as population growttU population density'

proximity to the center of urban and semi urban areas and proximity to roads'

5.2.1 Discrim inant AnalYsis

between
Discriminant analysis is a technique for simultaneously examining differences

two or more groups of objects with respect to several variables (Klecka,

1980)' Fisher

introduced this technique for the first time in 1936 (Fisher, 1936). The concept

Linear combinations
underlying discriminant analysis is fairly simple (Marija, 1990)'

ofthe independent variables are formed and serve as the basis for classiffing cases into

one of the groups. For example, suppose \¡re measure height

in a random sample of 50

males and 50 females. Females are, on the average, not as tall

as males, and this

difference will be reflected in the difference in means (for the variable height).

Therefore, the variable height allows us to discriminate between males and females
139
with a better than chance probability: if a person is tall, then they are likely to be male;

if a person is short, then they are likely to be female (Marija, 1990).

Discriminant Analysis has been most commonly apptied in medical or biological

research (Rao, 1966; Reyment, 1,973; Lachenbruch, 1973 Chou at a\,1998;

Feighner

at a|,2000;Tu et al., 2003). Recently, this technique has been used for many
purposes

in areas such as physiology, educational researcl¡ banking, political science as well as

in geography and remote sensing (Sullivan, 1980; M*Üq 1990)' In pattern recognitior¡

(Du ef
linear discriminant analysis was successful in ctassifring a hyperspectral image

a1.,2001;Dt et a1.,2003).Dt et at. (2001) used linear - constrained distance-based

discriminant analysis (LCDA), which replaces the Fisher ratio used in linear

discriminant analysis (LDA) with the ratio of inter-distance to intra-distance as a

forced to separate.
criterion for optimality. The result was that all classes of interest are
et al'
By means of this direction" LCDA can detect and classifu similar targets. Dt¿tn

(1999) used rational discriminant analysis (RDA) based on a proximity description

of

instead of
the data, such as similarities or the distances to a subset of the object,

features. They argued that formal objects might also be represented

by a proximity

measgre (distances or similarities) to a set of prototypes or support objects' They

pattern recognition tool for
conclude that rational discriminant analysis is a valuable

applications in which the choice of the features is uncertain.

Tappeiner, et al. (1993) used discriminant analysis to identiff the influence of

environmental parameters and human impact on the Alpine

cultural landscape' This

study successfully integrated discriminant analysis and GIS

in modelling vegetation

patterns using natural and anthropogenic influence factors' The

model presented a

140
 and allowed the qualification and
useful tool for explaining the vegetation distribution

prediction of environmental ef[ects on it (Tappeiner et al', 1993)' Sinowski e/ ø/'

They used relief
(1999) applied discriminant analysis on geological stratification'

parameters such as altitude, slope and depth in discriminant analysis to stratify

properties' They concluded that
geological areas with different spatial variability of soil

determine the soil depth at which

discriminant analysis could be used objectively to

geology changes.

of artificial neural networks and

Blackard et al. (1999) studied comparative accuracies

types from cartographic variables' They

discriminant analysis in predicting forest cover

used the cartographic variables of elevatior¡ aspect and slope to predict forest cover
that a feed-forward artificial neural
type. The result was that the comparison indicated
type than did a traditional
network model more accurately predicted forest cover

statistical model based on Gaussian discriminant

analysis' h et al' (2003) improved the

Gaussian mixture discriminant analysis to

Gaussian discriminant analysis model with

capture sub-pixel heterogeneity. They used

a Gaussian mixture discriminant analysis

forest stands from Landsat TM images'

model for infening land cover fractions within
analysis model offers an
They conclude that the Gaussian mixture discriminant
problem in remote sensing'
attractive means for addressing the mixture-modelling

study; they used discriminant

Bertels, et al. (1999) conducted another comparative
performance' They concluded that
analysis and neural networks to evaluate company
analysis, except when they are
neural networks are not superior to linear discriminant

given highly uncertain information. Manel et al. (1999) compared discriminant

for predicting species distributions of

analysis, neural networks and logistic regression

t41
Himalayan river birds. They compared the perforrnance of multiple discriminant

analysis (MDA), logistic regression (LR) and artificial neural networks (Allf$. The

result was that MDA was preferable over LR or ANIN, but where there are complex or

non-linear influences on species distributioq ANN may well turn out to be

advantageous.

5.3 Method

5.3.1 Overlay analys¡s of possible driver factors

Overlay analysis in GIS was used to analyse the relationships between static as well as

dynamic factors and land use/land cover change in the study area. Static factors such as

slope, physiography and land system have been created as well as dynamic factor

(Table 5.1). The classes of each static factor followed the standard classification system

of slope class, physiography and land system used in Indonesia @irectorate Land use

and RePPProt, 1986), while dynamic factors were created by buffering a 10 km radius

from the centre of urban area and semi-urban area. This radius assumes that l0 km is

the longest distance people walk to work in agricultural areas. Urban and semi-urban

areas were defined based on the status of the city as a capital city.The capital city of a

kabupaten (district) was defined as urban and the capital city of kecamatan (sub-

district) was defined as semi-urban. The total population at kecamatan level in which

these cities were located was more then 50 000 people.

The buffer lkm from main roads (the state roads), others roads (the provincial and

kabupaten roads), and toll way the proximity factor. This assumes that I km is the

maximum distance of built up area located along the roads. This bufler area was

r42
overlayed with land use/land cover to identiff the relationship between roads as an

accessibility factor with land use/land cover change in the study area'

Another dynamic factor, population density, has been created based on the total

population per square km area of each kecamatans (a lower level of Kabupaten)'

The

The area
total population data were collected from the statistical offrce of Kabupaten.

of each kecamatan was calculated based on administrative boundary map that was
to
derived from topographic maps (chapter 3). Table 5.1 shows the method employed

create possible influence or driver factors in the study area.

t43
 factors
Table 5.1. The Method to create area of possible influence or driver

Features GIS Possible influence or driver

Method F'actors
Main City Buffer Proximity to Urban area
Other City Buffer Proximity to Semi Urban area

Main roads Buffer Proximity to Province roads area

Other roads Buffer Proximity to KabuPaten roads
Toll way Buffer Proximity to Toll way area

Population densitY Reclass Population densitY area

Slope Reclass 0 -3 %o area
3 -8o/o area
8 - 15 Yo atea
15 - 30 Yo
o/o
area
30 - 45 area
> 45 %o atea
Physiography Reclass Flat
Rolling hill
Volcano
Fold hill and volcano
Land system Reclass Coalescent estuarY area
Coastal beach ridges area
Hillocky plains area
Intertidal mudflats area
Low rounded hills area
Minor river floodPlain area
Permanently water lodged area
Steep hill on marls area
Undulating to rolling
Sedimentary Plains area.
v karts area

has been conducted to create

Buffering of 'þoint" features such as main and other cities

aÍr aÍea of proximity to the urban and semi-urban

as a possible dynamic influence

factor. while buffering of "line" features such as main roads, toll way and other roads
to create ar: areaof proximity to roads also as a possible dynamic
has been conducted

..polygon" or "afea" features such as population density areas, slope

influence factor,
been reclassified to create an
class as well as physisography and land system class have

then were used in overlay

influence area from this factor. These spatial data sets
cover change in the study area'
analysis of possible influence factors and land use/land
144
5.3.2 Simple linear and Discriminant Analysis

linear regression to
The quantitative analysis undertaken here consists of simple
use/land cover' and the
analyse relationships between populations and land
factors had the most influential on
discriminant analysis approach to determine which
analysis is used to build a predictive
land use/land cover in the study area. Discriminant

of group membership based on observed characteristics of each case' The

model
set of discriminant functions based on
procedure generates a discriminant function or a

provide the best discrimination

linear combinations of the predictor variables that

between the grouPs

or independent
In this stud¡ discriminant analysis was used to analyze the predictor

the best discrimation between the groups'

variables of land use/rand cover that provide

of six classes of land use/land cover (LULC)

The dependent variable consists
rice field, open/dry land' settlement and
categories in the study area (forest, plantation,

frshpond) and the independent variable consists

of five variables: slope' proximity to

(province road or main road)'

city center (urban and semi urban), proximity to roads
system is a nominal data type'
proximity to toll way and population density. Land
an independent variable in the discriminant
therefore this static factor is not included as

consist of an interval data type except for

analysis. Data for discriminant analysis must

group membership in any given cell or

pixel
the dependent variable (Marija, 1990). The

have strong influencs in discriminating

can be predicted based on the variables that

land use tyPe.

r45
 proximity to the city' to main roads'
Independent variables, such as population density,

as the dependent variable'

to toll way, and slope as well as land use/land cover type
ofrandom sample grid' In this
were created on a gTid system and masked with 20Yo

distance to the centre of urban area

analysis, proximity to the city was created as
was created as distance to
(capital city of Kabupaten), and proximity to main roads
that only urban areas that
provincial roads, this category being based on the hypothesis

are located along province roads have strong

influence on land use/land cover change'

distance intervals of 1, 5 km for roads

This database was classified into 30 classes with

ful for proximity to the city (urban area)' Another parameter is

and toll way and 2
(Class 0 no data Class I population
population density. This was divided into 4 classes

density 15-150 prs/sqkrn, class II population densþ

151 - 650 prs/sqkrn' and class III
population density 651-1200 prs/sqkm) (Table 5'2)'

Table5.2:Yaiableusedinthediscriminantanalysis

Variable Data type Range

Nominal 6 class
Land use/land cover
factots
to the lnterval 30 class
to main lnterval 30
P
to Toll lnterval 30 class
lation lnterval 4 class
SúaúÍc factor:
lnterval 6 class
Slooe

format using Arc/info to obtain a

This entire database was then converted into tabular

for discriminant analysis in SPSS software (Statistical

database that can be used

Package for Social Science)'

on the database of population at

Simple linear regression analysis was conducted
from the result of image classification
kecamatan level and land use/land cover derived

area in every kecamatan was derived from

of years 1989 and 1997.Landuse/land cover
146
intersection of the kecamatan boundary and land use/land cover for 1989 and 1997 '

Based on this database, simple linear regression was conducted to identiff relationships

between population and land use/land cover for the years t989 and 1997 '

5.4 Results
5.4.1 Overlay analys¡s

5.4.1.1 Static drivers

An intersection between land use/land cover and static driver factors resulted in
cover
identification of various relationships among driver factors and land use/land

change in the study area. Tables V.1, Y.2 arrd V.3 (Appendix V) show the matrix of

land systems'
intersection between land use/land cover versus slope, physiography and

use/land cover
From these tables it can be recognized that only fow categories of land

changed significantly. These \¡rere: forest, plantation, rice field and

settlement' Other

categories such as operVdry land, fisþond and water/reservoir did not change

significantly.

cover change
Table 5.3 shows the summary of the relationship between land use/land

categories of forest,
and slope. From this table it is evident that the land use/land cover

plantation, rice field and settlement significantly changed within all slope classes'

Forest decreased mostly within slope 30-4 5% and > 45yo and settlement increased

mostly within slope 0-3%, which is where settlement predominantly is located'

plantation decreased within slope g-15%o and ls -30yo, while rice fields decreased

shows that
within slope 0-3 Yobulincreased within slope more ttøî3%. This change

147
 (settlement
land use/land cover that represents human intervention or intensive land use

and rice field) is mostly located in the area within slope 0-3%.

Table 5.3: summary of land use/land cover change versus slope (1989-1997)

15-3Ûo/o >458/a
Tvpes\SloPe
-201
Forest 1) 6116 609 110 108 -191
-8.02 -9.16
2) 1.31 3.48 1.31 1.69
-2898 -1325 -1 049 -221 -151
Plantation 1) -5850
-1.25 -16.59 -15.85 -16.45 -9.28 -6.89
2)
345
Rice field 1) -1 9486 1677 1 166 909 434
18.23 15.71
2) -4.16 9.60 13.94 14.25
1 8043 330 -18 16 -4 -3
Settlement 1)
3.85 1.89 -0.22 o.24 -0.15 -0.13
2)
1) = unit in Ha
2)= Percentage of change area to total area of slope c/ass

the intersection
Another relationship that supports this phenomenon is also revealed by

land use/land cover and physiography (Table 5.4). Settlement and rice
fields
of
in rolling
significantly changed within flat areas, while forest and plantation changed

hill and volcano areas.

From both the intersection of land use/land cover with slope, as well as with

use/land cover
physiography, it was clearly recognized that, in general, the major land

located in the
in the study area such as settlement and rice field, were predominantly

flat areas within srope 0-3%, fo[owed by prantation and forest in rolling
hill and

was fitted with

volcano areas within slope 0-3% and more t}fin3o/o. This phenomenon

the schema of land use pattern in Indonesia (Chapter 2)'

148
 Table 5.4: Summary of land use/land cover change versus Physiography 1989-1997

Types Flat Rollinq hill(RH) RH and V Volcano(V)

Forest 1) 1090 3785 -1085 2761
7.37
2) 0.36 6.37 -1.02
-5377 -4647
Plantation 1) 2145 -5287
2\ 0.71 -8.90 -5.05 -12.40
Rice field 1) -19478 -2231 -2034 1610
4.30
2) -6.46 -3.75 -1.91
16877 499 9967 -46
Settlement 1)
2 5.59 0.84 9.35 -0.12

= unit in Ha
2)= Percentage of change area to total area of physiograPhY class

use/land
Further information was obtained from the result of intersection between land
cover
cover and land systems. From Table 5.5 it can be recognized that land use/land

such as rice field and settlement significantly changed in areas with land
systems of

forest changed in
coalescent estuary and undulating to rolling sedimentary plains, while

hillocky plains (Table 5.5). Land systems of coalescent estuary and undulating to

rolling sedimentary plains were located in the flat area, and consist of a fertile soil and
plains are located
are rich with minerals from volcanic sedimentary material. Hillocky

in the hill area consisting of volcanic tuff(RePPProt, 1986).

149
 land system (1989-
Table 5.5: Summary of land use/land cover change ver$$

Land system
Forest 11
t) 4
I 0 0
)" 0 0 0.

Plantation
353 1
t) -0 0
2.6 -7 1.3
2) 1

Rice field -10 -1

1
I
t) -37
)
371
518
-0.3 1
1l 0.1 I
l): UnitinHa
i): i"iurrt"s, of chønge area to total area of land system class'

tso
5.4.1.2 Dynamic drivers

Overlay analyses of land use/land cover change with dynamic driver factors were

conducted through examining the intersection between land use/land cover with

population density, proximity to the center of urban and semi-utban areas, and

proximity to roads. The results are presented as follows'

Population densitY

Table 5.6 shows the relationship between land use/land cover change and population

density. It is evident that settlement has consistently increased in area with higher

population density. Settlements increased graduaþ within areas with population

densþ class I (increased i1l+t-by23%),whi1e class II increased (by 3043, ha(64%),

and class III increase d (by aaal. ha (134%) (Figures 5.la and 5.1b). Areas of rice field

decreased with increasing population density. Rice field area decreased by 25277 ha (-

l l%) in areas with lowest population density (class I), while in class II it decreased by

9078 ha (-31%),and in class III it decreased by 3789. ha(-63%). There is no indication

of an effect of population density on decreasing or increasing forest and plantation'

Forest increased in area with population density class I and class II, while plantation

area decreased in low population density (class I) and increased in area with high

population (class II and III) (Table 5.6). The increasing forest and plantation in areas

with high population density could be due to areas of secondary forest, as well
as

plantations which consists of broad leaf plantation such as rubber and teak
wood which

are located close to the roads and settlement'

151
 Table 5.6: Land use/land cover and Population densþ Classes

LULC Ohanqe (ha) %Chanqe 3hanoe(ha) ToChanse lChange(ha) /oChanoe

Forest 2261 29 642 6el c c

Plantation -'10257 -21 2462 611 100 I

Rice field -25277 -11 -9078 -378s -63 -3rl
Ooen/drv land 25373 3e 3971 781 -66c EN

Settlement 3374 2X 3043 641 4441 134

Note: percentage change to total area of each category
C/ass / = 1 5 - 150 person/sqkm
-
C/ass // = 151 650 person/sqkm
C/ass /l/ = 651 - 1200 person/sqkm

Ä A

Flpulst¡il d€n*V 1989

Nlo dab Populdìm &nSY 1gg7
OÉs I {15 - 15Õ FrCaqhr} Ffu GK
Oßs ll (151 - 65Ù Þrdsdn) 10 û l0 KilÒmelss O6s l {15 - l5O trlsqkn)
oæs ll I 151- 650 prslsdsnl 10 0 10 K¡lomelffi
oss lll (651 - 1?00 pfsfs*m) 0æe iìl (661 - 1200 Prsjs$m)

Figure 5.la. The Classification of population density the study area

152
 À
^

/¡r,/ SudY ero bødrrY

/A\// Sìud!' roa boudrry .
1.nc* Roads
.1997 rO 0 l0 Kihdæ
10 O l0 KlbhdÌs I S¡tü€mmt
I S€tüeírertlgEg

Figure 5.1b. The settlement distribution inthe study area

-
Source: Analysis ofLandsat TM 1989 and 1997

lJrban and semLurban

greatest change
Rice field and settlement land use/land cover categories underwent
also occurred
within the 10 km radius from the urban centers (Table 5.7). This change
areas (Table 5.8). In this
in the area within l0 km radius from the center of semi-urban

area increasing settlement was accompanied by a decrease

in rice frelds' The largest

Jakarta metropolitan area'

increase of settlement is within the area which is close to the

Settlement substantially increased within 10 km of Bekasi,

Babakan' Cileungsi as well

as Cikarang. These areas afe about 30 km fiom Jakarta (Table 5'9 and Figure 5'2)'

153
 of urban area (Ha)
Table 5.7:Landuse - land cover change withinl0 km

Urban
Bekasi *)
;*J lr,,n,",,qJ
-11225 1756
,.nnr*"JF¡s¡pq!

0. -35.91 5.62 0
2) 0.
*) -1781 -795 1
Karawang 1

2) 0 2.77 -5.70 -2.il 4.5e1 o.2c

*) -717 -1 93 47 31
-277
Punarakarta
2)
---Z.nf -0.8e -2.30 -0.62 1 0.1

*) change -- unit in ha

á) percentage of change to totat area of each 10 km radius

km of semi-urban area (Ha)

Table 5.8: Land use -land cover change withinl0

Forest Rice field land

Semi-urban
*) -9030 8129 81 -17
Babakan 0 261
-28.89 26.O1 -0
2) 0.00 0.84
-5816 823
Cikarang *) 0 710
0.00 2.27 -18.61 2.63 I 0.
2\
77 -2791 -1201 -392
Cileungsi ")
-13.38 -5.76 -1.88 9.
2) -0.37
*) 68 852 708 -1111
Cikampek
-2.27 -3.55 2 0
2) 0.22 2.72
*) change = unit in ha
to totat area of each ctty in segment 10 km radius
2) Percentage of change

rice field and settlement within

Table 5.9: Percentage of decreasing and increasing
urban and semi-urban in the study area'

Distance
% Change of % Change of from
Urban and
Rice field Settlement Jakarta (km)
Sub urban
* -36 29 1f
Bekasi
-19 14 30
Cikarang**
-6 10 3C
Cileunqsi**
-6 5 5C
Karawang*
** -¿ 75
GikamPek
-2 2 7E
Purwakarta*
5.8
Source Summarized fromTable 5.7 and
**
* (Jrban area; semi-urban area'

154
 . Clty
[T:ïi-j Jakarta
Larid use 1997
Rice field 10 2tì Klþmders
Sdtlement

30
Clrange from 1989 to 1997
20

0
o
cD
0
(ú

o -10
s I Rice lield
-20
¡Settbnìent

-30

40
Belesi Cikarang CiÞungsi lGrawang Cikanpek R¡rwakarta

Figure 5.2.Theincreasing of settlement accompanied by decreasing rice field within

urban and semi-urbariin the study area. Sorting by distance from Jakarta city.
Source: analysis ofLandsat TM i,989 and 1997

155
From Figure 5.2 it can be seen that the percentage increase of settlement and

percentage decrease ofrice freld is greater for areas that are close to Jakarta than for

areas that are further from Jakarta. It can also be recognized that status as a capital city

of kabupaten (urban) and status as a capital city of kecamatan (semi-urban) does not

necessarily relate to increasing settlement. For example, Bekasi, Karawang and

purwakarta are capital cities of kabupaten (urban areas) but the large and significant

increasing area of settlement is only in Bekasi and Karawang, not in Purwakarta.

However, Cikarang and Cileungsi, which are the capital cities of kecamatan (semi-

urban areas), increased in their area of settlement (Table 5.9). This phenomenon

supports the view that the distance from the Jakarta metropolitan area strongly

influenced increasing settlement in the surrounding areas'

Another interesting feature was the increasing area of settlement in Cikampek, an

atea

more than 75 km away from Jakarta. This increase of settlement was related to

trade in
increasing industrial development in this area. Industry, manufacturing and

this area and surroundings increased from 4777 :urruts of industry and manufacturing in

19g9, to 6923 units in 1997 and 8354 units of industry and manufacturing in 2000 (he

local Statistic office of Karawang, 2000).

Accessibility

Roads have a strategic function as a transportation connector, facilitating

the flow of

cover
goods and the movement of people from one place to another. Land use/land

along roads tends to undergo rapid change due to the accessibility of the
surtounding

afeas

156
From Tables 5.10a to 5.10c it can be seen that rice fields and settlements have

undergone change along the 1 km buffer from the roads. The largest change
is along

the toll road. In this area, rice fields decreased by 3913 ha or '23 % and settlement

along
increased 3810 ha or 22o/o. The second largest change is rice field and settlement

the state road, where settlement increased by 3552 ha or 13 % and rice field was

decreased by 2t35 ha or -8.02 Yo. Alongthe state roads it was rècognised that open land

decreasedby -2346 ha or -8.81%. This decrease could be some open/dry land that has

been converted to settlement (housing or industrial estate) as has been

found in the

field. During the fieldwork in1999 it was found that inthis area a lot of new
housing

and industrial estates have been constructed'

Table 5.10a: Land use/land cover change in 1 km bufler of Toll Road

70. Lu¿Ul

301 r378 77 0.46

Plantation 1

9517 5604 -3913 -23.26

Rice field
3759 -14 -0.08
Ooen/dry land 3774
2231 6042 38 10 22.65
Settlement
oÁ Change
-- percentqge of change area to total area of
lkm buffer

Table 5.10b: Land use/land cover change in 1 kmbufter of state

Road

Land use
60 293 233 0.88
Forest
2124 2783 659 2.48
Plantation
ts28l 13146 -2t35 -8.02
Rice field
4234 887 -2346 -8.81
Ooen/drv land 1

4928 8479 3552 13.34

Settlement
oÁ Change : percentage ofchange areato total qrea of lkm buffer

t57
 Road
Tabte 5.10c: Land use/land cover change in 1 km buffer of other

1601 199 -401 -0.40

Forest 1

12767 r0064 -2702.1 -2.68

Plantation
58969 ss413 -35s6 -3.53
Rice field
r8754 19007 253.0 0.25
Ooen/drv land
Settlement 7689 r4t3l 6442.
ot6
Change : percentsge ofchange areato total area of lkm buffer

5.4.2 Quantitative analysis (Simple linear and Discrimant analysis)

such as settlement'
Four land use/land cover categories that had changed significantly
linear regression
rice field, forest and plantation have been analysed through simple
5'3d)'
with the total population within the study area (Figures 5.3a to

200000

y=76.217x+50250
R2 = 0.3927
o 50 000
o
ê o o
tr taa' O' a
o o
(E o a
E 00 000 o
CL
o
È
o
(É
o
F
50000
o
a
o a
a
l¡
0
0 500 1000 1 500

Settlement f 989 (ha)

settlement 1989
Figure 5.3a Linear regression of Total Population and

158
 8.00
a y=0.0017x+0.1129
6.00
R2 = 0.1704
s a a a o
4.00 a
5 a
ìo o
Ð
tr
!¿' o
a aa a
o
.g -500 00
a
500.û l ooo.oo 1500.00 2000.00 2500.00
ê
=
o
o-
-+ où
Settlement change I 989'l 997(Ha)
-6.00

Figure 5.3d Linear regression of the Population growth and Settlement change.

From Figgre 5.3 a to 5.3d it can be seen that there is positive relationship between total

..population and total area of settlement'., "population change and settlement change"

as well as "population growth and settlement change". This means that when

population increases settlement will increase, but the correlation of these relationships

was low with R2 below 0.65, and the lowest was correlation between population
growth

and settlement change (R2:0.17,Table 5.11).

Table 5. I I : Rf of finear regression between population and

land use/land cover
Totalof Totalof Population growth
Population 1989 Population 1997 1989-1997
Settlement 1989 R2 = 0.39
Settlement 1997 R2 = 0.66
R2 = 0.17
Settlement change
1989-1997
Rice field 1989 R2 = 0.0024
Rice field 1997 R2 = 0.0063
Rice field change R2 = 0.3477
Forest chanqe R2 = 0.0013

Plantation change R2 = 0|1402

160
 in Figure
Relationships between population and rice field, forest and plantation shown

5.3e-5.3g, were negative. This means that when population increases

rice field, forest

and plantation will decrease. This relationship was poor; with most of IÚ below 0'40,

and especially for forest category was very poor 0.0013 it could be due to

representatives' data of forest, which is only a small palt in the study

area'

4000
o¡' . 3000
y=-338.91x-57.939
às
o 2000 R2 = O.3477
lE oa
o 1000
E o oaa
o
ED o I
tr -b 4 6
o
-4
-reûo a a
a O
.E
=ÉL
-- C'c
-2000
o a
o-
-3000
a
-4000
Rice field change (ha)

Figure 5.3e Linear regression of Population growth and Rice field

change'

400
y=-3.1765x+3.4528
300
R2 = 0.0013
s 200 o
o
G 100 O
s
ìo
o -4 ^ô
-JU -1
-1 00
1 2 3
Ê
o -200 a
G

=e
o
-300
o-
-400
o

-500
Forest change (ha)

change'
Figure 5.3f Linear regression of Population growth and Forest

161
 700

dto a
èS O
o 'soo
6 y = -34.824x+ 176
400
.C R2 = 0.1402
ìo 300
aa
o oo a
C
o a
a 100
-!! o
=
o.
o o
È -4
., 2 4 o I
-1 00
o a a
-200
Plantation change (ha)

Figure 5.3g Linear regression of Population growth and Plantation change.

Discriminant function analysis results in variables that discriminate between six land

use/land cover categories in the study area. Table 5.12 shows the covariance matrices

for separate group. The values in this table gave some idea of how the relationship

between variables changes from group to group. For example, in group 5 (Settlement),

all variables have negative values except variable population density. It means these

variables have negative and positive relation with settlements. Settlement areas are

large in areas where population density increases (positive), and also will be small in

areas when slope decreases, and proximity to urban, proximity to main road, and

proximity to toll way all decrease (negative si$). This relationship makes sense, in that

settlement areas are close to roads, close to urban, have little slope, and are in high

population density areas. Other groups such as forest, plantation and rice field have

different relationships. For example, group I @orest) has a positive relation with

population density and a negative relationship with slope, proximity to road and

proximity to urban. It means that forest areas are large in areas with high population

t62
density and close to road and to urban. Rice fields also have positive relationship with

population density, slope and proximity to urban areas, and a negative relationship with

proximity to main roads and proximity to toll way. Oper/dry land has the same

relationship as the relationship of settlement to population density, which is a positive

relationship. This is a single land use/land cover (one year 1997) analysis. Forest,

plantation and rice field as well as settlement in this study area are located in areas with

high population density and near to the roads. Therefore, rice field as well as forest has

a positive relationship with population density and proximity to the roads. Especially

for forest it could be a misinterpretation of this category due to mixes with broad leaf

plantatior¡ which is located close to settlement and main roads.

Table: 5.12. Covariance Matrices

Population Prox. to main Prox. to toll
Group density Slope Prox. to urban road way

I : Forest 0.085 -0.033 -.2.76 108 0.098

Plantation 0.13'1 0.025 -0.106 -0.068 0.122

2=
3 : Rice field 0.169 0.007 0.098 -0.202 -0.394

4 : Open/ dry land 0.147 -0.003 -0.328 -0.348 -0.13 8

5 : Settlement 0.641 -0.001 -0.552 -t.252 -t.344

6 = Fish pond 0.252 0.000 t.7ll 0.625 -0.623

From this
The initial statistics from the discriminant analysis are shown in Table 5.13.

table it can be seen that the high values, occur only in four functions
with an eigenvalue

>0.1 or percentage of variance >15 vo. This means that only of this four function that

has strong relationship with land use/land cover'

Table 5. 13. Eigenvalues

Function Eigenvalue Vo of Cumulative Canonical
Variance % Correlation
1 .236' 35.5 35.s .43',¡

2 .t77', 26.(. 62.1 .388

148' 22.2 84.2 .359
4 .105' 15.7 100.c .308
5 000 .0 100.0 004

a First 4 canonical discriminant functions were used in the analysis

t63
úr Table 5.14a it can be seen that all variables have a different contribution in each

function. For example, the population density has a greater contribution to function 4

than the other, slope has a greater contribution to function 2, as well as proximity to

main road. It means factors such as proximity to urban areas have a strong contribution

in function 4, proximity to toll way has strong contribution in functioî 2, slope has

strong contribution in functions 1 and 3, population in function 3,and proximity to main

road in function 3. This relation was also supported by looking at the relationship

between the dependent variable and discriminant variables in the structure matrix

(Table 5.14b), which gives the canonical correlation coefficients. These values are
'When
comparable to factor loading and indicate the substantive nature of the variables.

some dependent variable has high canonical correlations while other have low, then
the

ones with high conelations contribute most to group separation (Bargman, 1970).

Table 5.I 4a. Standardized Canonical Discriminant Function Coeffrcients

Function
,)
I 4
Prox to urban -.2t4 .567 .27s 729
Prox. to toll way .245 -.769 -.162 .s96
Slope .660 130 .637 -.087
Population den -.t54 -.t49 .418 -.072
Prox. to main road .576 .689 .82t .282

Table 5.14b. Structure Matrix

Function
1 2 J 4
Prox to urban 767r .072 .62',7 -.066
Prox. to toll way 734* .38',1 -.485 100

Slope .368 .678* Il''l .625

Population den -.393 392 .34',1 7 53*

Prox. tomain road .672 130 2t0 .044

Pooled within-groups correlations between discriminating variables and standardized canonical
discriminant functions Variables ordered by absolute size of correlation within function.
* Largest absolute correlation between each variable and any discriminant function

164
 and toll way
From Table 5.t4b it can be seen that proximity to urban areas' main road

density in function
have a high correlation in function 1, slope in firnction 2, population

4. If it is compared with eigenvalue in Table 5.13, it can be clearly recognized that

factors such as population density, proximity to urban areas,
proximity to main roads

and proximity to toll way have a strong contribution' It means these factors have a

strong influence on land use/land cover in the study area, similar

to that found from

overlay analYsis.

The linear equation of the discriminant analysis is analogous

with multiple regressions'

The general linear equation is: Group = a * br*xr f b2 *x2*"""tb-*x'

group'
Table 5.15 shows the coefficients of linear equation of each

Table 5. 1 5 : Classification function Coeffrcients

<r vuy ,
1 2 3 4 5 o
6.868 6.975 7.447 6.911 9.808 9.941
6.543 3.249 2.888 2.762 2.783 1.852
Slope
0.689 0.689 o.773 0.511 1.41
Proximity to urban 0.853
-0.267 -0.237 -0.306 -0.190 -0.225 -o.240
to main road
0.759 0.656 0.928 0.553 0.675 1.041
to toll
-r7.609 -10.708 -12.093 -10.410 -13.560 -22.673
(Constant)
Fisher's linear discriminant function

function can be
From Table 5.15 the linear equation of discriminant classifrcation

divided into six groups as follows:

Groupl(Forest):.17.609+6.868*Popden+6.543*Slope+0.853*Prox.towban-0.267
*Prox to main road + 0.759*Prox to toll way

Group 2 (Plantation) : -10.708+6'975*PopderÈ3'?!?:Slope+0'689*Prox'to urban-

0.237*Prox to main road + 0.656*Prox to toll way

Group 3 (Rice freld) :-12.093+7.447*Popden+2.888*Slope+0'689*Prox'to urban-

0.306*Prox to main road + 0.928*Prox to toll way

urban-
Group 4 (Oper/dry land): -10.410+6.911*PopderÛ2362*Slope+0'773*Prox'to
O.tgO* Prox to main road + 0.553*Prox to toll way
165
Group 5 (Settlement) : -13.560+9.808*Popden+2J-83lS1ope+0'5ll*Prox'to urban-
0.225*Prox to main r< ad + 0.675*Prox to toll way

Group 6 (Fishpond) : -22.673+9.941*Popden+l.852*Slope*1'41*Prox'to urban-

0.240*Prox to main road * 1.041*Prox to toll way

grouped cases as well as cross-validated

This classification result was 43.7Yoof original

grouped cases coÍectly classifred. It means the accruacy of this linear equation of

discriminant classification function was 43'7 %;o'

5.5 Discussion

between land use/land cover

Many studies have been conducted to reveal the interplay
growth and changing patterns of population
changes and the influencing factors such as

and socio-economic factors. These studies have

mainly addressed the demographic and

between land use/land cover

socio-economic aspects of the non-spatial relationships

changes and the influencing factors, more

qualitatively than quantitatively'

from a spatial dimension using

This study focuses on analysis of their relationships
The methods used in the study might
GlS-based advanced spatial analysis techniques.

shed more light on the research freld of land use/land cover changes' both

and techniques can be

quantitatively and qualitatively. Such research methods
scale'
extrapolated to a broad dimension at a regional

use/land cover changes and possible

The results from overlay analysis between land
there is a close relationship between
influencing factors in this study demonstrate that
and static influencing factors' Slope and
land use/land cover changes and both dynamic
166
physiography were the static influence factors that had a strong relationship with land

use/land cover. The land use/land cover types that represent human intervention
<
(settlement, rice field, plantation) were mostly located in the flat areas with slope 3yo'

The dynamic factors such as population density, proximity to the urban areas and

proximity to the road also had a strong relationship with land use/land cover'

population
Settlement increased and rice field decreased significantly with increasing

density as well as within 10 km proximity to the city and within 1 km proximity

to the

roads. This land use/land cover distribution conformed to the schema of land use

pattem in Indonesi athathas been studied by I Made Sandy (Sandy, 1997).

In urban areas such as Bekasi, Karawang and Purwakart4 decreasing rice field and
such as in
increasing settlement are clearly recognized, as well as in semi-tlrban areas

cikarang, cileungsi and cikampek. The increasing of settlement accompanied by

decreasing rice freld in the study area was related to increasing population density'

Settlements increased within areas according to overall population density; class I

increased 23Yo, class II increased 64yo, and class III increased t34%. while rice field

areas decreased -llyo in area with lowest population density (class I), in class II it

decreased -3lyo, and in class III it decreased -63%. This condition indicates that the

decreasing rice
increasing population density has an effect on increasing settlement and

population
field in the study area. There is no indication of strong relation between
in areas with
density on decreasing or increasing forest and plantation. Forest increased

population density classes I and class II, while plantation area decreased in low

population density (class I) and increased in areas with high population (class II and

III). The increasing forest and plantation in areas that had high population density could

Purwakarta
be due to the type of forest in the study area, especially in the kabupaten

167
 progfams (Local
and Karawang, where there is secondary forest as part of reforestation

and settlements' Also'

Statistical Offrce, 7gg7), which are located close to the roads

as rubber and teak wood'

there are plantations which consist of broad leaf plants such

mostly located along the roads.

and decreasing rice

Another factor that has a strong influence on increasing settlement

field in the study atea, is the distance from the Jakarta metropolitan area
The

increasing of settlement accompanied by decreasing rice

field within urban and semi -

In the Bekasi area' 75

urban gradually decreased with increasing distance from Jakarta'

km away from Jakarta, settlement increased the greatest and rice field
decreased'

Jakarta' This supports the

followed by cileungsi and cikarang, which are 30 km from
developing the surrounding
view that the Jakarta metropolitan area has an influence on

afeas.

has a strong influence on

The factor of proximity to the main roads and tolt way also

decreasing rice field and increasing settlement' This proximity factor is easy to

understand in that roads have a strategic function as a transportation connector'

from one place to another' In

facilitating the flow of goods and the movement of people

effect, areas close to the road will have the greatest chance to be changed into

settlement or other commercial areas'

influence factors and land

From the discriminant analysis, the relationship between
independent (influence
use/land cover was clearly recognized- Relationships between
indicate that group
factors) and dependent variables (land use/land cover types/groups)

4 (open/dry land) and group 5 (settlement) had positive

relation with population density

168
and negative relation with, slope, proximity to urban, proximity to main road, and

proximþ to toll way. It means that open/dry land and settlement will increase in areas

with high population densþ, little slope, close to urban, close to main road, and close

to toll way. This relationship makes sense because most settlements are located and

extended in the flatateaalong the road and near to the urban afeas'

The other interesting feature is the positive relationship between group l(forest) and2

its negative relationship

þlantation) as well as rice field with population density and

with slope, proximþ to road and proximity to urban. It means large forests in areas

with high population density and close to roads and urban areas. A similar result was

found from the overlay analysis in that forest and plantation areas were mostþ located

in the areas with high population density and close to the main roads, especially in

kabupaten purwakarta and Karawang. Rice fields had a positive relationship with

population density, slope and proximity to urban, and negative relationship with

proximþ to main road and toll way. It represents the real condition that rice fields are

mostly located in areas vvith high population density, near to main road and toll way

and far from urban areas.

The canonical correlation on standardized canonical discriminant function coefficients

and structure matrix tables indicate the contribution of relationship between dependent

variable and discriminant variables. The canonical conelation is the correlation of two

other a
canonical (atent) variables, one representing a set of independent variables, the

set of dependent variables. These values are comparable to factor loading and indicate

the substantive nature of the variables. The high value of canonical correlation

represents the strong contribution to group separation. Therefore, the factors

population

169
density, proximity to main road and proximity to urban are the variables that strongly

contribute to separation of each group of land use/land cover in the study area.

5.6 Gonclusion

This chapter has analysed the influence of land use/land cover change drivers in the

study area. As the study area is mostly flat in the downstream area of the Citarum

watershed, land use/land cover predominantly consists of irrigated rice fields. Static

factors such as slope and physiography have a strong influence on land use/land cover

in the study area. The land use/land cover types that represent human intervention

(settlement, rice field, plantation) were mostly located in the region of little slope or flat

areas. Dynamic factors such as population densit¡ proximity to the urban areas and

proximity to the roads also had a strong relationship with land use/land cover. This

strong relationship is supported by discriminant analysis, where the factors population

density, proximity to the main roads and proximity to urban were the variables that had

the strongest contribution to discrimination of each group of land use/land cover in the

study area. Distance to the Jakarta metropolitan area strongly influenced the increasing

amount of settlement and the decreasing of rice field in the surrounding areas such as

Bekasi, Cikarang and Cileungsi.

170
 Chapter Six

LAND USE/LAND COVER CHANGE PREDICTION

IN THE STUDY AREA

6.1 lntroduction

Potential future spatial distribution of land use/land cover conditions is important

information that can be used by planners to identiff areas that require priority attention

or to anticipate poor arrangement of land resources. Many researchers have developed

models to simulate and e4plore land use/land cover changes. These models are various

and relate the purpose and the scale of studies (Berry, 7996; Clarke, 1997; Verburg e/

al., 1999;Irwin, 2001;Pontius et ø1.,2001; Fischer, 2001 and Gobin" 2002). Pontius e/

al. (2001), for example, developed a model to simulate the location of land use/land

cover change specifrcally for forest distr¡rbance. They present a GlS-based model,

GEOMOD2,whichqualifies factors associated with land use, and simulates the spatial

pattern of land use forward and backward in time. GEOMOD2 was able to classify

forest cover conectly to between 84 and 86 % with Kappa about 0.34 and 0.45. Gobin

et at. (2002) developed a model to predict the probability of private local agricultural

land use. The result was a binary logistic model for estimating probabilities of private

agricultural land use that correctly predicted 95.7% of the 300 sample plots. Significant

progress in spatial modelling of land use/land cover change has been made when

spatial data sets were available from remotely sensed data (Irwin et al., 2001).

Availability of digital spatial data from satellite imagery has assisted in

t7t
conceptualizngthe basic geographic and environmental processes associated with land

use/land cover change, and developing spatial models that fit the spatial process of

land use/land cover change.

Cellular automata (CA) have been used to simulate land use/land cover change (White,

IggT). CA is a mathematical technique in which the behaviour of a system is generated

by a set of deterministic or probabilistic rules based on the states of a cell and

neighbouring cells. Cellular automata, combined with Markov chain probability

principles, is a way to predict land use/land cover on a spatial basis, based on transition

probability and contiguous neighbouring cells.

This chapter presents the land use/land cover change prediction as well as simulation in

the study area based on Markov cellular automata models. The objectives are:

1. To use the Markov cellular automata model available in Idrisi32 to predict

land use/land cover change in the study area.

2. To simulate the future land use/land cover in study area based on scenarios

by using Markov cellular automata model'

3. To analyse the past and the future land use/land cover in the study area.

6.2 Land use/land cover change prediction model

Many studies have investigated land use/land cover change analysis, prediction and

modelling. These vary from those that focus on the concepts of land use/land cover

change to those on theorizing and modelling (Briassoulis, 2000). Models of land

172
use/land cover change have progressed since spatial data sets from remotely sensed

data have been available. Time series remotely-sensed data with multi-temporal data

acquisition allow prediction of the timing of changes, opening new avenues better to

link socio-economic transformation to land use changes (Kauûnann et aL.,2001) and to

understand time lags in the response of land use to socio-economic or natural

perturbations.

Serneels et al. (2007) assessed the driving forces of land use changes through a spatial

statistical analysis. They developed spatial statistical models of the proximate causes of

different processes of land use change. The descriptive spatial models developed in

their study suggest some important factors driving the land use that can be related to

some well-established theoretical frameworks. This study focused on understanding

proximate causes of the changes. It did not address the location of land use change

taking place. Pontius et al. (2001) studied land use/land cover changes related to

location and quantity issues. This study aimed at predicting the spatial patterr¡ location

and rate of tand use/land cover change. Other models that address the location and rate

of land use/land cover change include the Markov Cellular automata approach

(Veldkamp et a1.,2001). This is a spatial model that combines the Markov transitional

probability approach with cellular automata to predict land use/land cover change.

Through cellular automata techniques, behavioural models of land use can be made

spatially explicit (Veldkamp et aL.,2001).

The Markov Cellular automata model was selected in this study for three reasons.

Firstly, the Markov chain principle has the simple rule that transition probability can be

t73
used to predict future land use/land cover. Secondly, cellular automata can represent

the spatial dimension of the process, in this case land use/land cover states. And lastly,

Markov and cellular automata approaches complement each other to predict land

use/land cover change-addressing where and at what rate land use/land cover changes

are likely to progress.

6.2.1 Markov chain model

Landscape change and spatial diffusion processes can be simulated using linear,

stochastic techniques. A stochastic process is determined by random variables and is

describable only in probabilistic terms (Lambin, 1994). Probabilistic models in general

are appropriate for land use/land cover change processes given the complexity of

relationships between interacting variables, and the poor understanding of the driving

forces behind land use/land cover change. This approach has become common in

modelling land use/land cover change.

A Markov chain is a mathematical model for describing a certain type of process that

moves in a sequence of steps through a set of states. The Markov hypothesis, as

applied to land use/land cover change, is that past land use/land cover is helpful in

predicting future land use/land cover, given the present land use/land cover (Lambin,

1994). Bel et al. (1977) stated 'the conditional probability of land use/land cover at

any time, given all previous uses at earlier times, depends at most upon the most recent

use and not upon any earlier ones". The central mechanism of a Markov chain is a

probability pi¡ whichrefers to the likelihood of transition or movement from state i to7

in a given time interval, where i and j are either location-to model spatial diffusion

174
processes - or locationally relevant classes - to model landscape change processes

(Brown, 1970). For land use/land cover change studies, the states of the system are

defined as the amount of land covered by various land use/land cover, measured as

percentages ofarea ofeach land use/land cover type.

The Markov chain model will describe land use/land cover change from one period to

another and than use this as the basis to project future changes. This is accomplished

by developing a transition probability matrix of land use/land cover change from time

(t) to time (t+1), which will be the basis for projecting to a later time period. In

probabilistic terms for the sequence X (t), X (t+1) the Markov probability is expressed

as follows:

Plx(t+1)=ilx(t)=il
'Where j ofthe Markov series ofthe system (Bell, t974).
i and are indices

The Markov chain has been used in many studies. In the 1970s, for example, it was

used for analysing the location pattem of firms, migrations, and land use change (e.g.

Collin, 1973;Bell,1974; and Bell, 1977). Aaviksoon (1995) and Howard et al' (1995)

conducted further studies. Aaviksoon (1995) used a Markov chain model to simulate

vegetation dynamics; he concluded that the development of plant cover and land use

t)¡pes can be modelled well both on first and second order levels. The general trends

are well reproduced and the predictions are reliable to a few steps. Over longer time

periods (more than three or four steps), both approximations may fail to give conect

results. Howard et al. (1995) argued that a stationary Markov chain model allows the

future consequences of a given pattem of land use changes to be studied, and can be

175
used as an analytical tool. Logsdon et al. (1997) conducted another study on

Markov
probability mapping of land use change. They successfully showed that the

chain model affords a powerful descrþive and predictive model for

land use change

and for future land use distribution.

provide timely
In terms of predictioru the Markov chain models have the potential to

projection of land use/land cover changes with minimal data requirements,

but the

ignorance of interactions between contþous land areas by the Markov chain is a

major limitation for broad scale application (Lambin, 1994). Another

limitation is that

the Markov chain model does not help to address the "'Why?"
question. It can predict

future changes in land use/land cover-"When?' and "What?"- only as long

as the

stationary condition is met (Lambiru 1994)'

combined with GIs

Recently, the application of the Markov chain model has been
sensing
(Brown et al., 2000;Lopez et al., 2001;Hathout, 2002; weng, 2002). Remote

and GIS integration was used to provide spatial data of land

use/land cover type' weng

pattern of land use

(2002) was successful in analysing the direction, rate and spatial

that integration
change from the integration of remote sensing and GIS. He concludes
in
of remote sensing and GIS with Markov modelling \¡/as found to be beneficial

describing and analysing land use change. The integration

of GlS-based cellular

dependency and can be

automata with the Markov chain model can represent spatial

used to predict land use change the spatial dimension'

t76
6.2.2 Gellular automata

Spatial proximity underlies many dynamic processes in the landscape. The Game of

Life, for example, is a game based on the contiguous neighbour cell. Cells will live if 3

or 4 neighbour cells are alive and will die if 3 or 4 neighbouring cells are dead. This

can be very effectively modelled using a cellular automaton'

Cellular automata (CA) are dynamic models that are discrete in time, space and state

(Batty, l99S). Irwin (2001) stated that a cellular automaton is a cellular entity that

independently varies its state based on its previous st¿te and that of its immediate

neighbours according to specific rules. Clearly there is a similarity here to a Markovian

process. The only difference is application of a transition rule that depends not only

upon the previous state, but also upon the state of the local neighbourhood'

Cellular automata have been extensively used in many applications (White et al', 1993;

Butty, 1994; Cecchini, 1996;Xie, 1996; B"tty, 1997; V/u, 1998; Besussi et al', 1998;

Balzter et al., 1998; Candau et a|.,2000; Bryan, 2000; Sirakoulis, 2000)' White et al'

(1993), for example, simulate the spatial structure of urban land use using a cellular

automata model. Cecchini (1996) used cellular automata to build and test urban

development models, and Xie (1996) developed a generic model to simulate growth

dynamics for multi-sector land use in a sub-urban area. White et al. (1997) conducted

another study simulating the urban land use pattem using constrained cellular automata

for high-resolution modelling of urban land use dynamics. Xie (1996) integrated

cellular automata with GIS. He found that the integration a of cellular automata model

with a GIS would be advantageous because in many cases land use data that could be

177
used are already available in a GIS. Since that time, various studies have been

conducted which are related to integration of cellular automata with GIS (Takeyama et

al., 1997;Wagner, 1997;Couclelis, 1997;White et a\.,1997;White, 1997). Takeyama

and Couclelis (1997), develop a generalisation of map algebra (called "geo-algebra"),

of which cellular automata is a special case, and supporting spatial database

maniputations within GIS. V/u (199S) developed Simland, to simulate land conversion

through the integrated GIS and cellular automata. He uses multi-criteria evaluation

(MCE) to derive behaviour-oriented rules of transition with an analytical hierarchy

process method. A preliminary result of this model shows the interesting properties of

cellular automata models, in particular their spatial expression of complex micro-

macro dynamics. Batty et al. (1999) developed a model of urban dynamics through

GIS based cellular automata. He presents ways in which land uses are structured

through their life cycles, and ways in which urban activities spawn locations for new

activities.

On the urban growth application, Ward et al. (2000) developed a cellular automata

model integrated with a stochastic constraint to eliminate the broad scale factors of

constrained urban growth within GIS. The model can potentially simulate a wide

variety of urban forms depending on parameter values associated with the constraint

variables. All these studies show that cellular automata provide explicit handling of

dynamic spatial processes, potentially capturing in some detail the process of

interaction between the human systems and the natural environment within a GIS

178
Recently, studies have been conducted on developing integrated cellular automata and

GIS models that focus on spatial dynamics. Shi ¿r al. (2000) developed extended

cellular automata using Voronoi polygons to model dynamic interactions among

spatial objects. The Voronoi based cellular automata can model local interactions

among spatial objects to generate complex global patterns. The Voronoi spatial model

offers a ready solution to handling neighbourhood relations among spatial objects

dynamically. White et at. (2000) states that urban and regional models based on

cellular automata give good representations of the spatial dynamics of land use. He

developed a high-resolution integration of the spatial dynamics of urban and regional

systems.

Several researchers have conducted applications of cellular automata models to spatial

propagation. Bryan (2000) used cellular automata to model the propagation of land

clearance. He concluded that cellular automata reveal significant ecological

implications of the propagation process that are not considered by standard landscape

structural metrics. Sirakoulis et al. (2000) used cellular automata to study the effects of

population movement and vaccination on epidemic propagation. The model establishes

the acceleration of the epidemic propagation because of the increment of the

percentage of the moving population, or the maximum distance of population

movement. Cellular automata were found to be suitable for modelling this complex

non-linear problem.

The application of cellular automata in the simulation of spatial dynamic processes has

continued (Li Xia et al., 2001; Barredo et al', 2002; de Almeida et al', 2003; Arai et

t79
6.2.3 Prediction based on Markov Cellular automata Model within

GIS

Contemporary Geographic Information Systems (GIS) that can integrate and manage

spatial data have limited abilities to model dynamic spatial processes, and are poor at

modelling temporal change (Wagner, 1997). Using programming, GIS - especially

those based on tessellation models (using raster data structures) - can be used to build

CA models. The basic form of the CA model consists of a two-dimensional array, a set

of local states, neighbourhoods and transition rules. The states of cells in the anay

undergo change according to transition rules. Transition rules are functions of the cell's

state and the state of its neighboruing cells. This basic principle is able to extend GIS

capability into dynamic process modelling.

The next state of X cell is a function of the current state of X cell and the neighbour's

cells. If settlement is the state of the majority of neighbouring cells, then the X cell

will change and become settlement. The algorithm used to compute the next cell state

is referred to as the CA local rule. Usually, the same local rule applies to all cells of the

CA. A CA is characterized by five properties (Sirakoulis e/ a1.,2000):

1. the number of spatial dimensions (n);

2. the width of each side of the array (w). w; is the width of theTth side of the

artay, whereT : 1,2,3,....n (number of cells)7

3. the width of the neighbornhood of the cell (d).d¡ is the width of the
neighbourhood along theTth side of the arcayi.

4. the states of the CA cells;

5. the CA rule, which is an arbitrary function F.

181
The state of the X cell, at time step (t:1), is computed according to F. F is a function of

the state ofX cell at time step (t) and the states of the cells in its neighbourhood at time

step (t) (called transition rules).

The simplified figure of a two-dimensional CA (n:2), with neighbourhood width dr

:
3 and dz:3, is shown in frgure 6.2 below.

i-lj-l Lt.:i i-l j+r

i. j-l {ii) i,.¡+l

i+l.i-l i+l.i i+l.i+1

Figure 6.1. The neighbourhood ofthe (ij) cell is formed by the (ij) cell itself and the
eight adjacent cells (Sirakoulis et al., 2000).

Land use/land cover change is a dynamic spatial phenomenon. It is the result of spatial

interaction among socio-economic and biophysical environmental parameters. The

land use/land cover type in one cell is associated with those of its neighbours. The

simple rules of CA say that the state of a cell (or cells) changes according to transition

rules, and the transition rules (a function of the cell's state and those of the
neighbouring cells) can be applied on land use/land cover change at every cell. The

future land use/land cover type (according to Markov principles) is guided by the

present land use. In other words the conditional probability of land use/land cover type

at arry given time, given all previous uses at earlier times, depends at most upon the

most recent use and not upon any earlier ones (Bell and Hinojosa, 1977). This allows

the consideration of the land use/land cover category in which a parcel or cell will be

t82
classified at time (t+1) as dependent most upon its current classification at time (t)'

Logsdon et at. (1997) conclude that Markov chains afford a powerful descriptive and

predictive model for land use change and for future land use distribution. Therefore, in

Markov-based cellular automata models, the Markov transition probability approach is

used as a transition rule of CA.

The integration of Markov based cellular automata models within GIS in order to

predict land use/land cover change is relatively new. Idrisi32 vet.2 includes a Markov-

based cellular automata model to predict land use/land cover change (Eastmar¡ 2001).

This model is based on Markov transition probability as a transition rule of CA and

land use suitability as a constraint. In this model, the neighbouring cells that may

change are constrained by the land suitability of the neighbour cells. If neighbour cells

are suitable for settlement, for example, the cells will mostly likely change into

settlement and if neighbour cells suitable for forest it will change into forest.

6.3 Method

The Markov based cellular automata model in Idrisi32, called CA-Markov, has been

selected to model land use/land cover change in the study area. This model is based
on

the combination of cellular automata and Markov chain techniques. It adds an element

of spatial contiguity as well as knowledge of the likely spatial distribution of

transitions to Markov chain analysis.

183
Three data sets are required for operation of this model. Firstly, the latest land cover

classification is used as a basis for land use/land cover prediction (see Chaptet 4).

Secondly, the Markov transition that contains the Markov transition probability matrix

of land use/land cover change. Finall¡ the distribution of land use suitability for each

land use/land cover category. The transition matrix and suitability maps will be

discussed below.

6.3. 1 Markov transition probability calculation

A Markovian process is one in which the state of land use/land cover at time (t+1) can

be predicted by the state of the land use/land cover at time (t) given a matrix of

transition probabilities from each cover class to every other cover class. The Markov

module in the ldrisi3.2 can be used to create such a transition probability matrix. This

matrix is a result of cross tabulation of two images adjusted by the proportional error.

As input, it takes two land use/land cover maps. It then produces the following outputs:

(Eastmar¡ 2001)

o A transition probability matrix. These transition probabilities express the

likelihood that a pixel of a given class will change to any other class (or stay

the same) in the next time Period.

o A transition areas matrix. This expresses the total (in cells) expected to change

in the next time period.

o A set of conditional probability images - one for each land use/land cover

class. These maps express the probability that each pixel will belong to the

designated class in the next time period'

184
6.3.2 Multi-criteria evaluation approach to create land suitability

Change in state of cells not only depends on the Markov transition probability, but also

on the inherent land use suitability of each pixel. The suitability map was derived

according to criteria such as proximity to roads, river, city and existing land use/land

cover. The production of these maps follows the procedure outlined in the multi-

criteria evaluation (MCE) module of Idrisi32. The criteria were developed empirically

in relation to tmderlying land use/land cover change dynamics between two different

times (in this case between 1989 and 1997). The change of settlement category for

example, was related to factors such as proximity to main roads, proximity to urban

and semi-urban centres and mostly located in area within slope less than 45Yo.

Therefore in creating the suitability for settlement, it was based on the factors of

proximity to roads and city centres as well as slope less than 45Yo and existing

settlement. Suitability for rice field was based on proximity to inigation or river, slope

less than less than 45Yo and existing rice fields. The factors that contribute to each land

suitability analysis are listed in the Table 6.1.

Table 6.1. Land suitability factors of each category

Suitable for Factor

Forest Slope > 45 yo and proximity to roads, existing forest
Plantation Slope < 45Vo and proximity to roads, existing plantation
Rice field Slope < 45 o/o, proximity to roads, proximity to inigationhivet,
existing rice field
Open land/dry Iand All class of slope and proximity to roads, existing open land and
settlement
Settlement Slope < 45 Vo, Proximity to roads and proximity to city, existing
settlement
Fish pond Slope < proximity to
beach and proximity to
irrigation/river

185
Weighting factors have also been included in order to create the suitability map. In the

procedure for Multi-Criteria Evaluation using a weighted linear combination in

Idrisi32 based on pairwise comparisons (Eastman, 2001), it is necessary that the
weights sum to one. The consistency ratio of overall weighting value must have an

acceptable value. Using the WEIGHT module in ldrisi32, the weight value of each

factor for every land suitability category can be calculated.

6.3.3 Prediction process base on Markov cellular automata model

Markov cellular automata combine both the concept of a cellular automata filter and

Markov model procedure. The Markov rule is that the future state can be predicted

based upon the recent state, but with the Markov model alone it provides no sense of

geography. The Markov transition probabilities may be accurate on a per category

basis, but since there is no knowledge of the spatial distribution of occurrences within

each land use/land cover category, there is no spatial component on the modelling

outcome. A cellular automata (CA) model is used to add spatial characteristic to the

model. The CA model has the ability to change its state based upon the application of a

rule that relates the new state to its previous state and those of its neighbours. The

contiguþ filter was used to develop a spatially explicit contiguity-weighting factor to

change the state of cells based on those of their neighbours.

The contiguity filter was applied to a series of suitability maps already identifred for

each land use/land cover class. The process of the Markov Cellular automata model

was ¿Ìs follows: The first step is to create the transition probability and areas file from a

Markov Chain analysis (using the MARKOV module) of two prior land use maps,

186
establishing the quantity of expected land cover change from each existing category to

each other category in the next time period. The second step was selecting the base

image to use as the starting point for change simulation. The third step, to choose the

number of iterations according to how many years the simulation will conducted, i.e.,

if l0-year simulation, then chose 10 time iterations. Within each time step, each land

cover is considered in turn as a host category. All other land cover classes act as

claimant classes and compete for land (only from within the host class) using the

MOLA (multi-objective land allocation) procedure. The area requirements for each

claimant class within each host are equal to the total established by the transition areas

file divided by the number of iterations. The results of each MOLA operation are

overlaid to produce a new land cover map at the end of each iteration. The cellular

automaton component arises in part from the iterative process of land allocatior¡ and in

part from a filtering stage with each iteration that reduces the suitabilþ of land away

from existing areas. The filter is integral to the action of the Cellular Automata

component. Its purpose is to down-weight the suitability of pixels that are distant from

existing instances of the land cover type under consideration. The net effect is that to

be a likely choice for land cover conversior¡ the pixel must be both inherently suitable

and near to existing areas ofthat class.

This process is run for the number of iterations. Therefore, the result of prediction from

this model will depend on the number of iterations and filter type as well as the basic

land use/land covers data and transition probability. The filter and iteration can be

adjusted to get a more suitable result.

187
6.4 Selecting Parameters and Testing the Model

The Markov Cellular automata prediction model used in this study in Idrisi32 is an

experimental module. The operation of the Markov Cellular automata module to predict

land use/land cover change depends heavily on the quality of data input, such as

Markov transition probability, suitability maps and the base land use/land cover image.

The number of iterations and the contiguity filter type are other factors that have to be

considered. Therefore in order to perform suitable land use/land cover prediction these

parameters need to be tested.

In assessing the model in this study, the latest image that was used as a base image was

the 1989 classified image used to predict land use/land cover conditions for year 1997.

Different time intervals of the Markov probability matrix calculation were tried, as well

as dif[erent types of contiguity filters and number of iterations in order to get a suitable

result of prediction or simulation. The land suitabilþ map was also improved
especially for land suitability for settlement that very clearly associated witlì factors of

proximity to urban areas and proúmity to the main roads. The best result of prediction

is then compared with the classified 1997 image to examine the precision of the result.

The result oftest some par¿Ìmeters is described below.

6.4.1 Transition probabilitY

The transition probability was calculated based on Markov Chain analysis (using the

MARKOV module). Many parameters are required to create the Markov transition in

this module. These options include how many years or time intervals will be used and

188
for how long the prediction or simulation will be conducted. Table 6.2 shows the

Markov transition of land use/land cover between 1989 and 1997 n the study area. It

was an 8-year difference between 1989 and 1997; therefore, the time interval selected

to produce this transition matrix w¿rs an 8-year time interval and an 8-year prediction

time. From Table 6.2 lt can be seen that the diagonal direction has the higher
probability value. For example, class I to classl (Forest), class 2 to class 2 (Plantation)

and so on. It means that the high value of transition probability occurs between the

same categories, while for a category that has no chance to be changed or low

probability to change, the value was low, such as from settlement to forest or to

plantation as well as to water/reservoir (Table 6.2). These low values make sense

because the change of settlement to forest or to fishpond or to water/reservoir is

unlikely.

Table: 6.2.The Markov transition of land use/land cover 1989-1997,

8-year interval and 8-year time to predict.

Cells in Ex transition to
LULC ct.1 ct.2 ct.3 cr.4 cr.5 ct.6 ct.7
C1.1. Forest 57150 22300 47773 1092 116 0 26
Cl.2.Plantation 104503 234134 297490 56736 I 8865 142 0
Cl.3.Rice field 44330 287651 1985648 6101 10 271561 82092 2299
C|.4. Open/dry land 4633 80063 405784 357597 75893 2502 185
C1.5. Settlement 250 16424 68175 22583 141809 1052 75
C1.6. Fishpond 47 117 79397 4406 4476 145882 23
C1.7. Water/reservoi 1 361 581 73
Given Probabil of to
LULC c t. 1 ct.2 cr.3 ct.4 cr.5 ct.6 ct.7
C|.1. Forest o.4449 0.1 736 0.3719 0.0085 0.0009 0 0.0002
Cl.2.Plantation 0.1468 0.3289 0.4179 0.0797 0.0265 0.0002 0

Cl.3.Rice field 0.0135 0.0876 0.6047 0.1858 0.0827 0.025 0.0007

C|.4. Open/dry land 0.005 0.0864 0.4379 0.3859 0.0819 0.0027 0.0002
C1.5. Settlement 0.001 0.0656 0.2723 0.0902 0.5664 0.0042 0.0003
C|.6. Fishpond 0.0002 0.0005 0.3388 0.0188 0.0191 0.6225 0.0001
C|.7. Water/reservoir 0.002 0.0059 0.1241 0.0095 0.0012 0.0381 0.8192

189
6.4.2 Land suitability of each land use/land cover category

Suitability maps are another input parameter in Markov Cellular automata. Suitability

maps were developed in relation to the underlying land use/land cover change

dynamics in the study area between 1989-1997. Using the Multi-Criteria Evaluation

module with factors as shown in Table 6.3 and using water bodies as a constraint,

produced the suitability maps. These maps were:

1. map
Forest suitability 4. Oper/dry land suitability map
2. map
Plantation suitability 5. Settlement suitability map
3. map
Rice field suitability 6. Fishpond suitability map
Table 6.3. The weight value of suitability factors

For land For Rice field

Existinq Ooen land 0.5544 Existing Rice field o.2024

Proximity to road 0.1556 Proximity to irriqated 0.5434
Existino Settlement 0.2332 Slope 0.0865
Slooe 0.0569 Proximity to road o.0922
Consistency ratio = 0.04 Existinq Settlement 0.0756
Consistency is acceptable Consistency ratio = 0.07
Consistency is acceptable

For Plantation For Settlement

t r5¡ltr

Existinq Plantation o.5278 Existinq Settlement 0.0848

Proximity to road 0.1396 Proximity to DKI 0.2160
Slope 0.3325 Proximitv to citv 0.2402
Consistency ratio = 0.05 Proximity to road 0.3836
Consistency is acceptable Slope 0.0755
Consistency ratio = 0.08
Consistency is acceptable
For Forest

Existinq Forest 0.4869

Proximity to road 0.0778
Slope 0.4353
Consistency ratio = 0.01
Consistency is acceptable

190
Table 6.3 above shows the weight value of factors that have been selected for every

land use/land cover category. For the category open land, for example, the existing

open land has the highest weight value, follow by existing settlement, proximity to

roads and slope. While in the settlement category, the factor proximity to road has the

highest weight value, following by proximity to city and to Jakarta metropolitan area.

These weighting processes were created based on the change characteristic associated

with land use/land cover change dynamic between years 1989-1997. For example,

settlement was mostly changed within the area close to Jakarta, urban and semi urban

(city) center and the main road.

Figure 6.2a and b show the suitability map for settlement category produce by multi

criteria evaluation procedure.

Lorv

+
High

Figure 6.2a. Suitable for Settlement without factor proximity to Jakarta city

t9t
 Lo*'
+
High

I
::

Figure 6.2b. Suitable for Settlement with factor proximity to Jakarta city

suitability
The suitability map in Figure 6.2a and 6.2b shows the gradient of land use

for settlement. The different colours show the gradient of suitability from unsuitable

(black, red, yellow), to suitable (green) and very suitable (dark green)' It can be seen

that high suitability for settlement is located along the road and close
to cities (Figure

6.2a) and concentrated in the area close to Jakarta (on the left side image) (Figure
gradient from unsuitable
6.2b). Other category suitabitity maps have the same colour
(see APENDIX
(black, red, yellow), suitable (green) and very suitable (dark green)

VI). These suitability maps \Mere created based on examining some factors such as

each category in the

consistency ratio as well as underlying change characteristic of

study area.

192
6.4.3 The result of simulation or prediction process

Once the Markov transition area and suitability maps were constructed, the Markov

Cellular automata prediction model can be run. Land use/land cover for 1997 in the

study area was predicted by selecting the 1989 classified image as a base image,

combining it with the Markov transition and the suitability maps. In this prediction

process, different parameters such as contiguity filter type as well as suitability map

and Markov transition value, were tested.

Three kernel filter types (3, 5 and 7-kernel filters) were tried in order to obtain a

suitable result. The 3x3 kernel filter was the best contiguity filter when compared with

the 5-and 7-kernel filters. The number of iteration was 8, the same as the number of

years to predict (8 years from 1989 to 1997). Different types of suitability maps have

also been used, such as before and after being adjusted for proximity to Jakarta city'

Figure 6.3a and 6.3c show the prediction result based on this parameter. Figrne 6.3a

shows the results of prediction process based on the suitability map before being

adjusted and 6.3c the prediction process based on a suitability map that has been

adjusted.

The result obtained when the suitability map was adjusted for proximity to Jakarta

(Figure 6.3c) appears better than that for the unadjusted suitability (Figure 6.a).

Settlement in areas that close to main roads and urban and semi-wban areas wers

improved, such as in Bekasi, Karawang, Cikampek and Purwakana. Settlement and

t93
rice field as well as plantatior¡ fishpond and reservoir, look appropriate as predicted

except open/dry land and forest (Figure 6.3c).

Overall KIA of the predicted image before being adjusted was 56 %o and after being

adjusted was 67 o/o, an improvement of ll % (Table 6.4a). In the image after

adjustment, settlement, fishpond and water/reservoir all have a value above 56 %.

Other land use/land cover category such as open/dry land and plantation as well as

forest have KIA value around 4l to 46 Vo, while in the image before adjustment all

categories were below 55 yo, except fishpond and water/reservoir with 7l Yo and96Vo

respectively (Table 6.4a).

Plantdlwt
Rice field
0psnffly land

Fish pond
Wde/nsewoh

Fignre 6.3. a)prediction l; of 8-iteration, 3-kernel filter, suitability map of settlement

without proximity to Jakarta

194
 Figure 6.3. b). Original classified 1997 tmage

B ekasi

Karawang

Cikampek

Pr-rnvakarta

Figure 6.3c) predicted2;of 8-iteration, 3-kernel filter, suitability map of settlement

with proximity to JakartacitY.

195
Table 6.4a .The Kappa Index of Agreement (KIA) value between
original 1997 -1997
predictedl and original 1997- 1997 predicted2'

Forest 0.34 0.41

Plantation 0.43 0.43
Rice field 0.55 0.56
Open/dry land 0.43 0.46
ent 0.48 0.67
Fishpond 0.71 0.70
Water/reservoir 0.96 0.97
OverallKlA 0.56 0.67
") = 1997 original lmage with 1 997 image Predictedl
**) 1997 original lmage with 1997 image Predicted2
=

Another way to validate the prediction result is by comparing KIA

of the two cross-

images and between the

tabulated images between the original 1989 and original 1997

original 1989 and predicted 1997 images (Table 6'4b)'

between 1989 original

Table 6.4b.The Kappa Index of Agreement (KIA) value
tol997 oiigiout and original 1989 to 1997 predicted image

Land use/land KIA 1989 to 1997 KtA 1989 to KIA 1989 to

1997 predictl) 1997 predict2)
cover original
0.32 0.79 0.63
Forest
0.42 0.85 0.74
Plantation
0.59 0.80 0.70
Rice field
0.38 0.84 o.77
Open/dry land
Settlement 0.35 0.95 0.4
0.72 096 0.87
Fishpond
0.97 0.99 0.97
Water/reservoir
0.54 0.88 o.74
OverallKlA
based on suitabilitY maps before adjusted
Note: 1:the
after adjusted
Predicted2 = the result based on suitability maps

have a low value (Table

The KIA value between the lggg and 1997 original images

6.4b) because land useiland covers between these two images

were different'

Therefore, the KIA value between the original l9s9 and 1997 predicted images also

should have a low value - the same or close to the KAI between the 1989-197 original

t96
was better than predictedl which used the suitability map before it was adjusted. The

KIA value of predicted2 is lower than KIA value of predictedl, and has smallest

difference values with KIA value between the 1989 - 1997 original. For example, rice

field, settlement, fishpond as well as water/reservoir in predicted2 have a difference

value less than 0.15, and other categories such as forest, open/dry land and plantation

have difference value more than 0.31. While in predictedl all categories have

difference value more than 0.35. It means in predicted2 that the prediction result is

better than that in predictedl, especially for settlement and rice field categories. This

condition indicates that settlement and rice field was well predicted.

Modification of the Markov transition matrix improves the amount of cells that are

expected to change but did not improve the land use/land cover change on spatial

distribution. An inappropriate suitabilþ map of open/dry land category could cause the

poor result on open/dry land, forest and plantation change distribution' It was no

significant information to adjust the suitability map of this category. For example, there

are no systematic patterns of open/dry lands. Slope and proximity to tlìs roads as well

as existing open/dry land have been selected to create suiøbility maps of these

categories. A suitability map with or without existing land use category does not

improve the prediction result of open/dry land as well as forest and plantation.

r97
6.5 The potential outcome of future land use/land

cover in the studY area

cellular automata model

The assessment of the results of the application of the Markov
or predicting
in the study area indicates that this model was successful in simulating
rice field categories' Based on
land use/land cover change, especially for settlement and

was modelled' The

this outcome, the future land use/land cover in the study area
on the classified image year
following section presents l0 and 2}-year simulation based

1997 as a base image. This simulation was based

on two types of Markov transition

from 1989 and 1997 images

matrix. one is the Markov transition based on a calculation
(as Markov transition rule)'
with the assumption that transition probability stays same

The second is the Markov transition based on a scenario.

Prediction based on the transition matrix derived from

the 1989-1997 images assumes

future. This is not always the

that similar conditions from this period extend into the

in 1998 Indonesia suffered an economic crisis, which

will likely
case. For example

influence the future rate of land use/land cover change. Therefore, to try and to

incorporate the changed circumstances, two scenarios have been developed which

the IgSg-Igg7 situation. These scenarios are:

change the transition probabilities

1. Scenariol assumed urbanization will continue related to population

and

1989-1997; OperVdry land

economic growth but not as high as rate between

will remain as idle land ("lahan tidur"), due to slow down on development

housing or industrial estates; Forests have to be

well protected (at least 60%

could be kept to conserve Jatiluhur reservoir);

198
 category as
There is no change from settlement to other land use-land

well as from rice to forest or to water/reservoir'

Scenario2 assumed same as scenario 1 but very low

in urbanisation and
2

high change in agricultural area.

below.
The process and result ofthese simulations presented

6.5.1 Land use/land cover simulation200T (10 Year)

The time frame for simulation was ten years, and the
Markov transitions for the next

module' This operation was

lO-year period from 1997 were created from the Markov

based on the assumption that the characteristic

of land use/land cover change from

1997 to 2007 is same as that between 1989 and 1997 (Markov chain
ruþ' other

assumptions were that the rates of land use/land

cover change would be slightly lower

between 1989'L997 due to

(secnariol) and much lower (scenario2) than that change

economic crisis in 1998 as mentioned previously'

Based on this assumption a Markov transition

matrix was created' Table 6'5 shows the

land use/land cover categories'

sunmary of probability "expected urachanged" of each

t99
 Table 6.5 summary of probability expected to stay as same category
lO-year scenario

From 1989-1997
lmages Scenariol Scenario2
Expected Expected Expected
unchanged unchanged unchanged
LULC
36% 60% 500Á
Forest
23o/o 50% 45o/o
Plantation
Rice field 56% 65% 60o/c

29o/o 30o/o 4Oo/(

land
50% 85% 90o/o
Settlement
57o/o 55o/o 550¿

81o/o 100o/o 10001

Water/reservoir

higher in
From Table 6.5 it can be seen that probability of "expected unchanged" are

This means that for the next

both of the scerurios than that for the lggg-1997 images.

10 years, land use/land cover will not change as much as between 1989-1997' For

100 o/o, means there is no change

example a transition probability of water/reservoir of

a high probability to
to any other category from water/reservoir. Settlement also has

stay the same. Fishpond has lower value than in 1989'1997, which means it has more

chance to be changed to an other category'

use/land cover
A more detailed composition of transition probability for each land

category can be seen in Table 6.6a and 6.6b. The

composition of transition probability

the expected change

for each land use-land category in Table 6.6b and 6.6c represents

of land use/land cover for the next 10 year from 1997 '

200
 Table 6.6a. The Markov transition of land use/land cover for 10-year prediction
(1997-2007), based on 1989-1997 image'

Cells in transition to
LULC cr. 1 ct.2 ct.3 cr.4 ct.5 cr.6 cL7
70617 37349 85854 3943 277 40 59
C|.1. Forest
93561 136046 282742 19124
56372 588 0
C1.2. Plantation
53052 306456 1729823 641311 298654 88941 2497
G|.3. Rice field
8364 90688 471109 275542 90312 3571 188
C1.4. Open/dry land
C|.5. Settlement 1043 33453 142653 47324 226468 2176 181

0 612 89566 6097 5155 133937 24

C|.6. Fishpond
Water/reservoir
1.7. 1
2394 4931
Given P of to
LULC c I ct.2 cr.3 ct.4 ct.5 cr.6 cl.7
0.36 0.19 0.43 0.02 0.00 0.00 0.00
C|.1. Forest
0.16 0.23 0.48 0.10 0.03 0.00 0.00
C|.2. Plantation
0.02 0.10 0.55 0.21 0.10 0.03 0.00
C1.3. Rice field
0.01 0.10 0.50 0.29 0.10 0.00 0.00
C|.4. Open/dry land
0.00 0.07 0.31 0.10 0.50 0.00 0.00
Cl.5. Settlement
0.00 0.00 0.38 0.03 0 0.57 0.00
C1.6. Fishpond
0.00 0.01 0.13 0.01 0.00 0.04 0.81
CI.7. Water/reservoir

(scenario 2007)
Table 6.6b. The Markov transition of land use/land cover for l0-year
assumed that changes willlow
Cells in transition to
LULC ct. 1 ct.2 cr.3 cl.4 ct.5 ct.6 cL7
8884 39628 29721 9907 0 0 0
1. Forest 11

147108 2942'17 88265 58843 0 0 0

C1.2. Plantation
202847
C|.3. Rice field 0 I 56037 I 748976 156037 31207 0
281932 93977 0 0
C|.4. Open/dry land 0 93977 469887
0 0 0 67995 385303 0 0
C|.5. Settlement
0 941 56 11770 0 129465 0
C|.6. FishPond 0
607
7. Water/reservoir
Given of to
LULC cl. 1 ct.2 cr.3 cr.4 cl.5 ct.6 cL7
0.2 0.15 0.05 0 0 0
1.1. Forest 0.6
0.25 0.5 0.15 0.1 0 0 0
.2. Plantation
0 0.05 0.65 0.24 0.05 0.01 0
C|.3. Rice field
0.1 0.5 0.3 0.1 0 0
C|.4. Open/dry land 0
0 0 0.15 0.85 0 0
C|.5. Settlement 0
0 0 o.4 0.05 0 0.55 0
C|.6. Fishpond
0 0 0 0 0 0 1
C|.7. Water/reservoir

201
Table 6.6c. The Markov transition of land use/land cover for 10-year (scenario 2007)
assumed that changes will very low
Cells in transition to
LULC cr. 1 ct.2 ct.3 cl.4 cr. 5 cr.6 ct.7
C|.1. Forest 99070 59442 29721 9907 0 0 0

C1.2. Plantation 147108 264795 88265 88265 0 0 0

Cl.3. Rice field 0 156037 1872441 873806 1 56037 62415 0

C|.4. Open/dry land 0 93977 469887 37591 0 0 0 0

C1.5. Settlement 0 0 0 45330 407968 0 0

C1.6. Fishpond 0 0 94156 11770 0 '129465 0

C|.7. Water/reservoir 0 0 0 0 0 0 60759

Given Probabil of to
LULC cl. 1 ct.2 ct.3 cl.4 ct.5 ct.6 cL7
C|.1. Forest 0.5 0.3 0.15 0.05 0 0 0

C1.2. Plantation 0.25 0.45 0.15 0.15 0 0 0

C1.3. Rice field 0 0.05 0.6 0.28 0.05 o.02 0

C|.4. Open/dry land 0 0.1 0.5 0.4 0 0 0

C1.5. Settlement 0 0 0 0.1 0.9 0 0

C1.6. Fishpond 0 0 0.4 0.05 0 0.55 0

C|.7. Water/reservoir 0 0 0 0 0 0 1

Figure 6.4 shows the predicted land use/land cover for 2007 as a result of lO-year

simulation based on the Markov transition from transition between the 1989 and 1997

images (Table 6.6a), Figure 6.5 shows land use/land cover prediction lot 2007 based

on scenariol (Table 6.6b) and Figure 6.6 shows the land use/land cover prediction

based on scenario2 (Table 6.6c). Figure 6.7 shows the original 1997 land use/land

cover fiomthe classified image.

202
 Fish pond
Waìer/reseruoir

simulation based on
Figure 6.4. The result of land use/land cover change of 1O-year
Markov calculated from 1989- 1997 images (simulation2}}7)

Figure 6.5. The result of land use/land cover change of

l0-year simulation
- based on Markov scenario (Scenariol 2007)'

203
 Fish pond
Water/reseruolr

simulation
Figure 6.6. The result of land use/land cover change of lO-year
based on Markov scenario (Scenario2 2001)'

Figure 6.7. The classified 1997 image (Original)

204
From these figures it can be seen that there are diflerences in land use/land cover

change patterns, especially for the settlement category. As it has been assumed that

land use/land cover changes for the next ten years will be low in scenariol and very

low in scenario2 compared with change between 1989-1997, the land use/land cover in

scenariol as well as in scenario2 visually was slightly different from land use/land

cover simulation for 2007 (Figure 6.4). Settlement along main roads in the simulation

does not show much difference from settlement in scenariol (Figure 6.5), but higher

than in scenario2 (Figure 6.6).

Different land use/land cover changes between the original simulation and the

scenarios can also be seen by comparing the change characteristic between these

based
images (Table 6.7 and Figure 6. 8). The change characteristics were calculated

on percentage change oftotal cells for each land use/land cover category from
a cross-

tabulation of 1997 original and simulation2}}T images, and cross-tabulation of t997

original and scenario 2007 images.

From Table 6.7 and Figure 6.8 it can be seen that there are differences in the change

2007.
characteristic between simulation 2007 and scenariol 2007 as well as scenario2

yo,while in scenario I 2007 forest increased 34

In simulatio n2007,forests increased 14

vo and in scenario2 increased 24Vo; other categories such as plantation in simulation

2007 decreased 3 0/o, while in scenario | 2007 decreased I %o and in scenario2

in simulation increased 4I yo, while in scenariol increased

decreased 2%. Settlement

40%o and increased 24% n scenario2. These change characteristics result from the

assumption that activities for construction or development of new housing

or industrial

205
estates would slow down within the ten years from 1997. For example, the increased

percentage of open/dry land as well as decreasing fishpond in scenariol and 2

represents a lot of idle land ('tanah tidur'). Slightly low settlement (lo/o difference) in

scenariol and very Iow (20%o difference) in scenario2 represent low activity for the

development of new housing or industry or low urbanisation.

Table 6.7. The lO-year change characteristic between Simulation and Scenario.
Change characteristic Change characteristic Change characteristic
Land use- land 1997 originalto 1997 originalto 1997 originalto
cover simulation 2007 scenariol 2007 scenario2 2007
Forest 14 % increased 34olo increased 24% increased
Plantation 3% decreased 1% decreased 2% decreased

Rice field 10% decreased 13% decreased 18% decreased

Ooen/drv land 10 % increased 26 % increased 49 % increased
Settlement 41% increased 40% increased 24% increased
Fishpond 2% decrease 32% decreased 18% decreased

Water/reservoir no chanqe No change No chanqe

60
50
40
30
t,zo
tr
.E 10

àñ
2 4 5 7
-10 1

-20
-30 land useland covertYPe
40
E lùyear sinulaüon I lGyear sccnariol I10-¡æar scenario2

1. Forest; 2. Plantation; 3. Rice field; 4. Open/dry land; 5. Settlement

6. Fishpond i 7 . W atetheservoir.

Figrne 6.8. lQ-year change characteristic between simulation and scenario

206
6.5.2 Land use/land cover Simulatio n 2017 (2O-Year)

In section 6.5.1 two scenarios were used which adjusted the transition probabilities

from the lgSg-lgg7 matrix. This was based on changing economic circumstances

following the 1998 economic crisis. The use of these adjusted probabilities probably

produced a more realistic result; however, it is not expected that these economic
conditions will remain constant.

Therefore, the simulation for both scenarios was extended to 20 years. In this case the

assumption were:

1. Scenariol assumed the economic conditions was recovered and the

activities of development will increase; therefore, in this scenario, land

use/land cover will be increasingly changed, such as urbanisation will

increase and open land will decrease'

Z. Scenario2 assumed same as scenario 1 but the rate of land use/land

cover change is higher than in scenariol, especially high in urbanisation.

The adjusted probabilities can be seen in Table 6.8a and 6.8b as follow'

207
 rabre 6 sa M*k*;åî::?üxåîr3if1:ye*iï. use/land cover for

Cells in transition to
cr. ct.2 ct.3 ct.4 ct.5 ct.6 ct.7
LULC 1

19814 9907 0 0 0
1. Forest 1 18884 49535
132397 88265 0 0 0
.2. Plantation 73554 294217
2028478 702165 234055 0 0
0 1 56037
CI.3. Rice field
394705 87955 140966 93977 0
1.4. land 28193 93977 1

0 90660 362638 0 0
5. Settlement 0 0
0 0 58848 23539 11770 141234 0
C1.6. Fishpond
1.7. Water/reservoir
Given Probab of to
cl. cl.2 cr.3 ct.4 cr.5 ct.6 ct.7
LULC 1

0.25 0.10 0.05 0.00 0.00 0.00

C|.1. Forest 0.60
0.50 0.22 0.15 0.00 0.00 0.00
.2. Plantation 0.12
0.05 0.65 0.23 0.08 0.00 0.00
3. Rice field 0.00
0.10 0.42 0.20 0.15 0.10 0.00
C|.4. Open/drY land 0.03
0.00 0.00 0.20 0.80 0.00 0.00
C1.5. Settlement 0.00
0.25 0.10 0.05 0.60 0.00
Cl.6. FishPond 0.00 0.00
0.00 0.00 0.00 0.00 1.00
C|.7. Water/reservoir 0.00 0.00

rable.6.sb.Markovtransitioi"ï"J31'f'::*tåîtuse/landcoverfor

Cells in transition to
cr. cL2 ct.3 cL4 ct.5 cl.6 ct.7
LULC 1
9907 0 0 0
1. Forest 148605 29721 9907
88265 0 0 0
.2. Plantation 73554 294217 132397
624147 4681 10 0 0
C|.3. Rice field 0 0 2028478
187955 281932 46989 0
4 land 0 140966 281932
0 90660 362638 0 0
5. Settlement 0 0
0 0 58848 23539 11770 141234 0
cr.6.
0 0
Water/reservoir
1.7.
Given of to
cr. ct.2 ct.3 cL4 cl.5 ct.6 cl.7
LULC 1
0.05 0.05 0.00 0.00 0.00
Cl.1. Forest 0.75 0.15
0.22 0.15 0.00 0.00 0.00
.2. Plantation 0.12 0.50
0.65 o.20 0.15 0.00 0.00
3. Rice field 0.00 0.00
0.30 0.20 0.30 0.05 0.00
C|.4. Open/dry land 0.00 0.15
0.00 0.20 0.80 0.00 0.00
.5. Settlement 0.00 0.00
0.25 0.10 0.05 0.60 0.00
.6. F 0.00 0.00
0.00 0.00 0.00 0.00 1.00
cl.7 Water/reservoir 0.00 0.00

208
When the simulation was run using these assumptions, the result showed
that land

Figure 6'9 shows

use/land covef was changed significantly (Figure 6.9, 6.10 and 6.11).

the result of 2o-year simulation based on Markov transition from 1989'1997

(Simulation 2017), Figwe 6.10 shows the result of 2o-year simulation based on

scenariol and Figure 6.11 scenario2. From these figures it can be seen that land

use/land cover change was quite different, especially in scenario2 where settlement

increased, while rice field decreased as result of this urbanisation'

A summary of the changes can be seen in Table 6.9 and Figure 6'L2' It can be clearly

seen that the two scenarios produce results difflerent from that of the original

simulation.

Planl¡llon
Rþê fiEld
þñd

(simulation 2017)'
Figure 6.9. The result of land use/land covef 2}-yeat simulation

209
 2017)'
Figure 6.10 The result of land use/land cover 2}-year simulation (Scenariol

(Scenario2 2017)
Figure 6.1 1. The result of land use/land cover 2}-yeat simulation

2r0
 Table 6.9. The 2}-year change characteristic between Simulation and Scenario

Land use- land Ghange characteristic Ghange characteristic Ghange characteristic

cover 1997 originalto 1997 originalto 1997 originalto
Category simulation 2017 scenariol 2017 scenario2 2017
12 o/o increased
Forest 20% increased 11 % increased
Plantation No chanqe 1% decreased 21% decreased
Rice field 12% decreased 16%decreased 2O%decreased
Open/dry land 6 % increased 17% increased 9 % increased
Settlement 60% increased 65% increased 148% increased
Fishpond 1% decreased no change ro chanqe
Water/reservoir 14% decreased no change no change

Figure 6.12 shows the comparison of land use/land cover change characteristics

between simulatior¡ scenariol and 2 of 2}-year prediction. It can be seen that in

scenario2, land use/land cover change was higher than in scena¡iol and the original

simulation. For example, settlement in scenario2 increased 148% and rice field

decreased 20Vo, wlfike in the original simulation and scenariol settlement increased

60Yo and 65%. This indicates that the assumption of high urbanisation in scenario2

resulted in high increased settlement and supports the assumption that within twenty

years from 1997 the economic activity has recovered.

2tl
 160
140
120
100
o
ctt 80
c(É
60
o 40
s
20
0
-20 1 2 4 5 6 7

40
Land useland covertYPe

@ 2O-year simulation l2Gyear scenariol 120-year scenario2

1. Forest; 2. Plantation; 3. Rice field; 4. Open/dry land; 5. Settlement

6. Fishpond ; 7 . W ater heservoir.

Figure 6.12. 2}-year change characteristic between simulatioru

Scenario I and scnario2.

The expansion of settlement along the roads was clearly recognised from the result of

simulation both in scenario2 2007 and scenario2 2017. Figure 6.13 shows an overlaid

roads network with scenario2 2007 and scenario2 2017 images. From Figure 6.13 it

can be seen that settlement increased mostly along main roads and areas that are close

to city centres, especially to Jakarta. In some places it is seen that open/dry land

changed to settlement as well as from rice field. Forest remained unchanged.

212
 2001

A
10 0 10 20 Kilometers

I
x Foresl
Pbnlrtlon
Rlc¡ fleld
Jrkad¡

:]I Opin¡drylmd
Scül.mênt
FlEhpond
Waler/resewok

Figure 6.13a Overþ roads network with scenario2 image 2007

2017

lU 0 20 Ktlomêlers

Folêsl Jrk.rlt
Plrnlallon
Rlce f¡êld
Opon dry lrnd
Såttlemêrd
Flshpond
Wdgilrasgfvolr

Figwe 6.13b Overlay roads network with scenario22017 image.

213
6.5.3 Analysis of Land use/land cover, 1989, 1997,2007 and 2017

in the study area.

It has previously been mentioned (Chapter V), that proximity to roads and proximity to

urban and semi-urban areas are the main factors influencing land use/land cover

change in the study atea, especiaþ the change of settlement and rice field categories.

Therefore, spatial analysis of the future land use/land cover in this section focuses on

land use/land cover within 10 km of main roads and 10 km of urban and semi-urban

centres. This analysis includes the 1989, 1997 classified images, 2007 and 2017

simulation images (Scenario2).

As has been mentioned previously, scenario2 2007 was based on the assumption that

land use/land cover change will slow down with very low in urbanisation due to the

economic crisis in 1998. Scenario2 2017 assumed that economic activity recovered,

and land use/land cover change increased with high rubanisation. Therefore, land

use/land cover scenario2 year 2007 and20l7 was selected, as it is assumed that these

two scenarios will represent the next 10 and 20 years of land use/land cover condition

in the study area.

Table 6.10 and Figure 6.14 show land use/land cover within 10 km of main roads in

1989, 1997, 2007 and 2017 .

214
 Table 6.10. Land use/land cover within 10 km of main roads lr;r1989,1997
2007 and 2017.

LULC Area (Ha) Area (Ha) Area (Ha) Area (Ha)

1989 1997 2007 2017
589ç 6443 7374 6643
Forest
29351 33659 21939 26312
Plantation
154045 3698C 11 509C 9799€
Rice field 1

57162 5081 I 76019 51673

Cpen/dry land
1430f 32693 40760 77579
Settlement
Fishpond
ooà 1173 57Í 549
2528 252e 252e 2528
Water/reservoir
Note: Year 1989 and 1997 from classified lmage
Year 2007 scenario2 and 2017 scenario2 image

some land use/land

From Table 6.10 it can be seen that for 1989, 1997,2007 and20l7
forest' plantation'
cover categories in areas along the main roads decreased, such as

rice field, and fishpond. Settlement increased. In 1989 settlement areas were less than

was more
plantations, open/dry land and rice field, but eighteen yeafs later settlement

than plantation, and tn 2017 settlement vvas more than open/dry

land (Figure 6'14)'

This indicates that within l0 and 20 years from 1gg7, the agricultural areas such as

plantation along the main roads witl be less than settlement afeas'

Other information that can be seen from Figure 6.14 and

Figure 6.15 is that the

increase in open/dry land in 2007 was higher than that

for settlement' It indicates a lot

and in 2017
of idle land ("lahan tidur"), due to slow down in development activity,

settlement increased more than open/dry land. This indicated

that development

recovered'
activities increased as it was assumed that economic activity

215
 1 80000
160000
140000 ! Forest
@ Plantation
120000
ag A Rice field
- 1 00000
g Open/dry land
(ú
o 80000
I Settlement
60000
n Fishpond
40000 6 Water/reservoir
20000
0
1989 ',1997 2007 2017
Year

Figure 6.14. The tend of land use/land cover change within 10 km of the main roads,
1989 to 2017 (2007 and20l7 the result of scenario2)

2t6
 1989

I 9S7

20 0 20 Kllometers

Foroct
'
Fi_Ijil Jd<arta
r Plairtaüon
Rice field
openidrv land
Pror¿incid Road
StEte Road
Sêttlernent Toll W+'
Fishpond Rail Waf
Wder/reservoir

Figure 6.15a. Land use/land cover within l0 km of the main roads

2t7
 2007

2011

2û Kilomêters
20
^
0

Pldntdtion
trf,i-ì Jd(dttr
Rice tield Prw¡nc¡d Roed
Opeildry lÊnd $ete Road
Seül€mPnt TqllWåy
Fishpøtd Rail Wey

roads
Figure 6.15b. Land use/land cover within 10 km of the main

218
Settlements continued to increase and agricultr.ral areas to slightly decrease in the area

within 10 km of the main roads. It was more drastically changed in areas within 10 km

of the urban and semi-urban center (Figure 6.16 u b and c). For example, in area

within 10 km of Bekasi urban center in 1989, settlement was still less than rice field,

but after 1997 settlement was greater than agricultural areas. By 2017 rice field will be

nearly zero inthis area as it was urbanized to adjoin the Jakarta metropolitan city. The

semi-urban centers next to Bekasi are Cileungsi and Cikarang. In these two semi-urban

ateas, rice fields in years 1989 to 2007 was greater than settlement, but then decreased,

and finally between years 2007 and 2017 settlement increased and exceeds the rice

field area. In2007 Bekasi, Cikarang and Cileungsi would become one big settlement

area adjoining Jakarta cþ (Figure 6.17).

In other urban and semi-urban centers such as Karawang, Cikampek and Purwakarta,

rice field was slightly decreased and :urrrtil 2017. Rice field was still greater than

settlement afeas, except in Karawang, where settlement areas in 2017 werc same as

rice field (Figrne 6.15b).

219
Settlements continued to increase and agricultural areas to slightly decrease in the area

within 10 km of the main roads. It was more drasticalþ changed in areas within 10 km

of the t¡ban and semi-urban center (Figure 6.16 a, b and c). For example, in area

within 10 km of Bekasi urban center in 1989, settlement was still less than rice field,

but after 1997 settlement was greater than agricultwal areas . By 2017 rice field will be

nearly zero'tnthis area as it was urbanized to adjoin the Jakarta metropolitan city. The

semi-urban centers next to Bekasi are Cileungsi and Cikarang. In these two semi-urban

areas, rice fields in years 1989 to 2007 was greater than settlement, but then decreased,

and finally between years 2007 and 2017 settlement increased and exceeds the rice

field area. In2007 Bekasi, Cikarang and Cileungsi would become one big settlement

area adjoining Jakarta cþ (Figure 6-17)-

In other urban and semi-urban centers such as Karawang, Cikampek and Purwakarta,

rice field was slightly decreased and vrfti| 2077. Rice field was still greater than

settlement areas, except in Karawang, where settlement areas in 2017 were same as

rice field (Figure 6.15b).

219
 25000 Bekasi
20000
rú
- 1 5000
(!
o 10000
5000
0
1989 1997 2007 2017
Year
+Forest +Plantation +Rice field +eopen / dry land
+Settlement +Fishpond +Water/ reservoir

25000
Gileungsi
20000
aú
1 s000
aú
0, 1 0000

5000
0
1 989 1997 2007 2017
Year

-+Forest +Plantation +Rice field +eOpen/dry land

+Settlement +Fishpond +Water/resenoir

Figure 6.16a. Graphic of decreasing and increasing land use/land cover within 10 km
of urban and semi-urban areas.

220
 25000 Gikarang
20000

$ ''uooo

E roooo
5000

0
1989 1997 2007 2017
Year

+Forest +Plantation +Rice field -leOpen/dry land

+Settlement +Fishpond +Water/reservoir

25000 Karawang
20000
-g 1 5000
G
o 1 0000

5000

0
1 989 1 997 2007 2017

Year

+Forest +Plantation +Rice field +eOPen/dry land

+Settlement -o- Fishpond +Water/resenoir

Figrue 6.16b. Graphic of decreasing and increasing land use/land cover within l0 km
of urban and semi-trban

221
 25000 Gikampek
20000
.ú
1 5000
aú
o 1 0000
5000
0
1 989 1 997 2007 2017
Year

-+ Forest + Plantation -+-Rice field -r+Open/dry land

+-Seftlement + Fishpond +Water/reservoir

25000
Pururakarta
20000

f(! rsooo
o 1 0000

5000

0
1989 1997 Year 2007 2017

+Forest +Plantation +Rice feld -reOpen/dry land

+Settlement +Fishpond +Water/resenoir

Figure 6.16c. Graphic of decreasing and increasing land use/land cover within 10 km
of urban and semi-urban

222
 1989

€
^*o r
f

2Ð07

¡l

A
n A 20 Kíþmeters

Figwe 6.17.Land use/land cover within

10 km of urban and semi-urban centre fr Foresf.
Plarfiaition
Ricefield
: Apenldry land
r Settlêment
,Iì F¡shpffd
I Wteîlreservotr

223

-
6.6 Discussion
The Markov chain model is a powerful descrþive and predictive tool for land use

change and for future land use distribution. The Markov-Cellular automata model was

successful in predicting land use/land cover change in the study areq especially for

settlement and other categories that located close to main roads or trban centres. This

model depends heavily on the quality of data input, such as Markov transition

probability and suitability maps. The number of iterations and the contiguity filter type

are other factors that also have to be considered.

Different combinations of filter types and number of iteration did not improve the

result. Selecting different filter types shows that the groups of cells/pixels become

more compact but did not improve the KIA value. The only efÊective way to improve

the prediction results was by adjusting the suitability maps. The adjusted suitability

map of settlement with added proximity factor to Jakarta and other urban areas,

improved the predicted result for the settlement category. It was shown that a modified

suitability map improves the spatial distribution of land use/land cover change.

Diffrculty in adjusting the suitability maps of open/ dry land and other categories such

as plantation and forest, resulted in poor prediction of these categories. In this case,

knowledge of the background of change characteristics of each land use/land cover

category in the study area needs to be considered. For example, the change

characteristics of settlements was known to relate to the proximity to main roads and

cities; therefore, factors such as proximity to roads and proximity to the cþ were

224
appropriate factors in creating a suitability map of the settlement category. Change

characteristics of open/dry land, plantation as well as forest were not clearly known.

Factors such as slope and proximity to the roads that have been used in this study are

not really dominant factors in the suitability maps for these categories. Further study is

need with regard to obtaining more suitable prediction results for categories that still

had poor predicted.

Although there were some erroneous results for some land use/land cover categories

such as forest, plantation and open/dry land, good results were obtained for settlement,

rice field and fishpond categories especially in areas close to main roads and urban

centres. Based on this outcome, further work was undertaken to predict or simulate the

future (10- and 2}-year) land use/land cover cbange in the study area. Two Markov

transition matrices for l0-year as well as 2}-year simulation were created. One was

from Markov transition based on transition land use/land cover change from 1989 and

1997 images, and the other one based on scenarios. A transition scenario was created

that was assumed to be representative of future land use/land cover change in the study

area. The economic crisis of 1998 in this country was also considered in creating the

Markov transitio n probabilities.

A Markov Cellular automaton depends heavily on Markov transition, which assumes

that the future land use/land cover change was guided by the transition from the past to

the recent/present land use/land cover condition (Markov principles). The future land

use/land cover change simulation based on the Markov transition from 1989-1997 was

successful in simulating land use/land cover 2007 and20l7. The economic crisis in

225
1998 was considered as an efflect in declining some economic activities such as

development housing or industry. Therefore the Markov transition was adjusted in

order to obtain the representative transition for the next lO-and 2O-years of land

use/land cover change from 1997. Two Markov transition scenarios for 10-and 2D-year

simulations were created based on different assumptions. Secnariol for simulation to

2007 assumed that land use/land cover will be slightly lower and scenario2 assumes

very low urbanisation and high change within agricultrual area. Scenariol of the

simulation to 2017 assumed that tand use/land cover change will slightly increase and

there will be high level of urbanisation.

The results show that the intensive land use category in scenariol 2007, such as

settlement, stightly increased while rice fields decreased. In scenario2 2007 there was

very low change especially for the settlement category; it is assumed in this scenario

thatthere will be lowurbanisation. In scenariol and scenario2 2017, it was assumed

that economic activity recovered and the change of land use/land cover was higher than

lr¡'2007.

These results indicate that Markov Cellular automata model can be used to simulate the

future land use/land cover change based on transition scenarios as far as the transition

can represent the future land use/land cover change.

Further analysis from overlay of land use/land cover within l0km of main roads, wban

and semi-wban centre has shown that land use/land cover, especially settlement and

rice field, changed for the years 1989 , 1997,2007 and20l7. In an area within 10 km

226
of main roads between Jakarta to Purwakarta" settlement increased and rice field

decreased, while plantation, forest and fishpond slightly decreased. In the area within

10 km of urban and semi-urban centre that are close to Jakarta such as Bekasi,

Cikarang and Cileungsi, settlement and rice field were drastically changed. In these

three areas, settlement was larger than agricultural area and n 2017 these areas will

become one a big settlement area adjoining Jakarta city. This phenomenon can easily

be understood, since proximity to main roads as well as urban and semi-urban centres

are the main factors that influence land use/land cover change, especially in settlement

and rice field categories.

The present as well as the past and the future of land use/land cover information

resulting from this study have implications for the spatial planning process in this area.

Information such as that the future settlement area will increase at the expense of

agricultgral areas such as rice field, indicates that these areas will be urbanised. This

information has the implication that urbanisation will have to be anticipated to

eliminate future conflict in land use arrangement or land allocation as part of the

spatial planning process. This would also have implications on monitoring

environmental degradation such as sedimentation, flooding or decreased carrying

capacity with regard to the expansion of settled areas over agriculttral areas.

the main
Other important information from this result is that settlement increased along

roads and in areas that are close to the Jakarta metropolitan city, as the model
rules

stated. This indicates that the development of housing and new settlement
in the fringe

Jakarta has interaction between Jakarta, as the core metropolis, and

the
area of

227
suroundings areas. This also indicates that new settlements surrounding Jakarta area

socio-economically heavily dependent on the core, and has implication on the change

of labour force, for example, from agricultural labour to non-agricultural labour.

Another implication is that exceed settlement areas the agricultural areas, for example ;

or a lot of open land is idle from this simulation, shows that environmental conditions

will decrease-such as increasing runoff or decreasing carrying capacity in this area.

Simulation of the future land use/land cover change based on the transition Markov

principle, as well as scenarios based on justification to obtain the representative

transition for the future land use/land cover change, showed that the future land

use/land cover distribution could be successfully simulated. This simulation indicates

that the future land use/land cover can be simulated based on some policy scenarios in

order to obtain the optimal land use planning in this area, which could have a
positive

outcome on economic and environmental development'

The success of the Markov Cellular automata model to simulate the futtlre land

use/land cover in this study area also has a contribution on development of land

use/land cover change modelling on the spatial basis.

6.7 Gonclusion
This chapter has provided the results of an application of the Markov Cellular automata

model to predict land use/land cover n 1997 from image 1989, as well as simulation

228
and analysis of the future land use/land cover change in the study area. The result

indicates that more appropriate suitability maps especially for forest, plantation and

open/dry land category were needed.

The prediction result shows that overall KIA value is 56Vo before the suitability map

was adjusted and 67Yo after it was adjusted, with KIA value of forest, plantation and

open/dry land below 50%, KIA values for settlement and rice field 67Yo and 56o/o, arrd

KIA values of fishpond and water being 70o/o and 90o/o. The KIA of land use/land

cover category with a value below 50% shows the prediction result of this category

was not good enough in the categories of open/dry land, forest and plantation'

Settlement and rice field categories were adequately predicted, with a KIA value over

50%. The poor prediction result on land use-land categories such as open/dry land,

plantation and forest was due to limitations in the information of the background

change characteristics of these categories.

The information of the distribution of land use/land cover change characteristic is

important in order to select factors and criteria on creating a suitability map. The

information of settlement growth along the roads and in utban and semi-urban areas,

for example, was the appropriate information in order to select factors such as
proximity to roads and city in creating a suitability map of settlement category'

This model was successful in simulating or predicting land use/land cover change
cover for
especially for settlement and rice field categories. The future land use/land
growing
years 2007 and 2077 as a result o the simulation shows that settlement was

has an
along the main roads and close to Jakarta metropolitan city. This simulation

229
implication that there \¡ras conversion from rice field into settlement especially in area

that are close to Jakarta and along the main roads.

As a Markov transition principle, it was assumed that transition change from the past to

the recent time stayed the same for the future land use/land cover change, it has

weakness since the future may have significant change in economic conditions that

could affect the land use-land change behaviour. Modifred Markov transition with

some scenarios successfully created the new simulations that could represent the future

land use/land cover condition.

From some simulations it was found that Markov Cellular automata allow one to

simulate the future land use/land cover change, based on the Markov transition created

from information from the past to the recent time of land use/land cover change, and

also based on scenario. The appropriate background information that could have an

effect on land use/land cover change is needed in order to create a suitable transition

scenario that can represent the future land use/land cover change.

230
 Chapter Seven

CONCLUSION

7.1 Summary and Findings

The main aims of this thesis were to develop a methodology effectively to detect land

use/land cover change, to understand the inter-annual dynamics of land use/land cover

changes and to analyse, as well as to predict, land use/land cover change dynamics

using remote sensing and GIS. The Post-Classification Comparison approach was

chosen because it addresses the relevant and important aspects of change: detectior¡

identification and location. A combination of PCA and NDVI transformed images was

selected instead of the original band to classifr/identifu land use/land cover in the

study area. PCA gives a strong spectral signature of settlement or built-up area, while

NDVI gives a strong spectral signature of vegetation cover. It was hypothesised that

through the two transformed images, the detail of land use/land cover category in the

study area could be detected and identifred.

Chapter 2 reviewed and defined land use/land cover change, and described some

aspects that related to land use/land cover change. It was argued that population or

demographic pressure as well as economic growth and physical environmental

conditions are the factors which simultaneously, or independently, affect changes of

land use/land cover in the study area. The demand and supply relationship of the land

use/land cover change model shows that the demand structure of land use is very

dynamic; the demand for using land for human activities increases fromtime to time

231
under the influence of population growth, community structure and economy. On the

supply side, structure is relatively static; the surface area remains constant. The

constant size of the surface area with increasing demands being placed upon it may

result in land use conflict. Land use is mostly converted from less intensive uses to

more intensive uses, such as from forest or open/dry land agriculture to rice field or to

settlement, but it is also converted from one intensive land use to other, such as from

rice field to settlement or from intensive agriculture land into non-agriculture land use,

or from housing to other commercial land use. As an effect of population and

economic growth, demands from the non-agricultr¡re sector are usually much stronger

than those fiom agriculture, due to the much higher incomes which this sector can

obtain from the same area of land. In the long tenr¡ this could result in scarcity of food

supply since insufficient land is available for agriculture activities. Land use allocation

and management supported by the appropriate techniques such as remote sensing and

GIS are required to avoid such land use conflict.

Chapter 3 describes the socio-economic and biophysical conditions in the study area.

Physiographically the study area is a flat alluvial plain. Land use/land cover in the

study area has changed over time as an effect of demographic pressure and economic

growth, especially because the area is close to the Jakarta metropolitan arca. Land

use/land cover along the main roads from Jakarta to Purwakarta is very dynamic and

has become more intensive over time. S/ithin this area many rice fields have been

converted into settlement or builfup areas for housing and industrial estate

development.

232
Chapter 4 presented the analysis and results of land use/land cover change detection

using image processing and GIS. The results of the Maximum Likelihood supervised

classification of PCA and NDVI transformed image were successful, with a Kappa

index agreement (KIA) value >85yo. Land use/land cover was clearly differentiated

and able to be classiflred into land use/land cover categories namely: 1) Forest; 2)

Plantation; 3) Rice field (planted); 4) Open land; 5) Settlement; 6) Rice field

(unplanted);7) Dry land; 8) Fishpond and 9) Water/reservoir.

Post-classification comparison of this classified image was used successfully to detect

land use/land cover change in the study area. At the annual interval, detection found

that weather or season was related to a "leaf on and leaf off' phenomena. In the dry

seasor! forest was less than plantation because of a lot of "leaf off' (leaf had fallen),

while in the wet season forest was larger than plantation. In this time leaf leaves had

regrown ("leaf oni'). Therefore the change of forest area in the study area did not

represent deforestation or reforestation; it was error in classification as an efFect of the

"leaf off - leaf on" phenomena. It was clearly recognized in the field that there are

present some broad leaf plantations such as rubber plantations and teak wood,

especially in the southem part of Karawang to the north of Jatiluhur reservoir, which

can be confused with forest.

On the other hand, rice field (planted) and rice field (unplanted) were alternates

associated with the growing cycle. It seen at harvesting time that the land was covered

with rice plantation ready to be harvested, while at planting time there were no rice

plantation due to the fact the land is under preparation. Therefore at harvesting time

233
rice field (planted) was recognised to be larger than rice field (unplanted) but at

planting time rice fietd (unplanted) was larger than rice field (planted). This

phenomenon was clearly recognised in the field. Clear relationships of season or

growing cycle with changes of other land use/land cover was not recognised, except

the change in open land and dry land agriculture associated with dry and wet season.

Settlement increased over time, fishpond and water/reservoir remain constant.

From annual detection it was recognised that rice field (planted) and rice field

(unplanted) changed over time in association with the rice growing cycle. For the long-

term change analysis, it was necessary to reclassifr or regroup these two categories

including both the open land and dry land categories. Rice field (planted) and rice field

(unplanted) have the same function in the field as rice field area, and so were included

as rice field. Open land and dry land categories also have the same function as land

which is unused or under preparation for construction or for dry land agriculture, and

so were regrouped as open/dry land. After regrouping of these land use/land cover

categories it was found that settlement and rice field areas had permanently changed

over time. Settlement had increased while rice field had decreased. The trajectory of

land use/land cover change in this study area was recognized as follows: Rice field was

converted to open/dry land or directly to settlement, and open/dry land was converted

to settlement. There was no change or conversion from forest to rice field or to

settlement.

Chapter 5 commences the analysis of factors influencing land use/land cover change.

The large number of possible elements were simplified into static and dynamic factors.

234
Static factors were defined as factors that might never change, such as slope, elevation,

or physiography, while dynamic factors were defined as factors that can cltange, such
transport
as population growth, population density, urban, and semi-urban areas and

routes. Overþ analysis of these two factors with land use/land cover found that

population density, proximity to urban and semi-urban centers as well as proximity to

cover
roads are the dynamic factors that have a strong relationship with land use/land

in the study area. Settlement increased and rice field decreased significantly with

increasing population density as well as within the 10 km buffler of proximity

to urban

and semi-urban centers and proximity to roads. As the study area is mostly a
flat

(settlement
alluvial plain, land use/land cover types that represent human intervention

and rice field) were generally located in flat areas with a slope of less
than3o/o.

In urban areas such as Bekasi, Karawang and Purwakart4 decreasing rice field and
such as in
increasing settlement are clearly recognized as well as in semi-urban areas

cikarang, cileungsi and cikampek. The increase of settlement accompanied by

population density'
decreasing rice field area in the study alea was related to increasing

V/ithin an 8-year time interval (1989-1997), there was a gradual change from low
population
population density to high population density. Settlement in the areas with a

density of 15-150 person/sqkm (class I) increasedby 23o/o, in areas with population

density of 151-650 person/sq.km (class II) increase d 64Vo, and in areas with population

density of 651-1200 person/sqkm (class III) increased of 134%. The rice field area

decreased by -1 1% nareas with the lowest population density (class I), in class II it

decreased by -3lo/o, and in class III it decrease d by -63%. This indicates that the
rice
increasing population density is an eflect of increasing settlement and decreasing

235
field in the study area. There is no indication of a strong relationship between

increasing population density and decrease or increase in forest and plantation area.

Forest increased in areas with low population density (class I and class II), while

plantation decreased in areas with low population density (class I) and increased in

areas with high population density (class II and III). The increasing plantation areas in

areas of high population density could be due to a lot of plantations being located close

to settlement areas. This is the case with rubber and teak wood plantations that are

found in the study area, especially in kabupaten Purwakarta and Karawang.

Another finding is that distance from the Jakarta metropolitan area has a strong

influence on increasing settlement and decreasing rice field in the study area" The

increasing density of settlement accompanied by decreasing area of rice field within

urban and semi - wban areas gradually decreases with increased distance from Jakarta.

In the Bekasi area 15 km away from Jakarta, occutred the greatest increase in

settlement density and the largest decreases in rice field, following by Cileungsi and

Cikarang, which are 30 km from Jakarta. This supports the contention that the Jakarta

metropolitan area exerts an influence on land use/land cover change on the surrounding

areas.

A positive relationship was found between total population and settlement, while the

relationship between total population and rice field, forest and plantation was negative.

It was found that for an increase in total population, the area under settlement will

increase and the area under rice field, forest as well as plantation will decease.

236
Discriminant analysis supported the strong relationship between the static and dynamic

factors that influence land use/land cover. Factors such as population density,

proximity to main roads, proximity to urban centre and slope are the factors that

strongly influence land use/land cover change in the study area.

Chapter 6 examines, as well as applies, a Markov Cellular automata model to predict

and simulate future land use/land cover change in the study area. It is shown that by

selecting different filter types, the groups of cells/pixels become more compact but it

did not improve the KIA value. The only eflective way to improve the prediction

results was by adjusting suitability maps. The adjusted suitability map of settlement

which adds a proximity factor to Jakarta and other urban areas improved the predicted

result for the settlement category.

The diffrculty experienced in deriving an adjusted suitability map of openl dry land and

for other categories such as plantation and forest, resulted in poor prediction of these

categories. In these cases, knowledge of the background of change characteristics of

each land use/land cover category in the study area needs to be considered.

Markov transition probability was effective in determining the amount of area that can

be expected to change, while suitability maps are eflective in determining the pattem

of land use/land cover change distribution. Two different Markov transitions have been

used to simulate the future land use/land cover in the study area. The two Markov

transitions are a Markov transition based on information that was obtained from 1989-

1997 images (assuming that the future land use/land cover change transition stay the

237
same as that transition between 1989-1997), and a Markov transition based on a

scenario that was assumed to represents the future land use/land cover change

transition, considering the economic crisis since 1998. It was found that the future land

use-land cover change could be simulated based on the Markov transition from 1989-

1997 'mrages and the Markov transition scenario. The future land use/land cover from a

simulation based on the Markov transition scenario can be more or less changed

depending on the scenario assumed if comparing with the simulation based on Markov

rules (transition from the past land use/land cover condition to the recent condition). It

found that the trends of change which are characteristic for each land use/land cover

category in the two simulation images are diflerent. In the scenariol 2007 image as a

result of simulation based on a Markov scenario that assumed the future land use/land

cover will slightly change related to the economic crisis 1998, it was shown that the

trend of land use/land cover change was lower than that simulation based on Markov

transition 1989-1997.

Percentage of change, especially for settlement category, in a simulatíon2}D7 based on

a Markov scenario I and 2 was low compared with percentage change in the

simulation image based on transition from 1989-1997 and high percentage in

increasing open/dry land as well as decreasing fishpond. This represents a considerable

amount of idle land ('tanah tidur" literally "sleeping land") within the 10 years from

lggT-2007 and is assumed to be an effect of the economic crisis. While in a simulation

2017 basedon scenariol and scenario2, the percentage change increased especially for

intensive land uses such as settlement, and rice field decreased, as it was assumed that

238
within the 20 years from 1997 Iand use/land cover change will increase due to

economic recovery fiom the crisis, especially aftet 2007 .

It was found that the future land use/land cover within 10 km of main roads and urban

and sub-urban centres will continue to intensifu, with settlements continuing to

increase and agricultural area slightly decreasing. Change was more drastic in areas

within l0 km of the urban and semi-urban areas. For areas within 10 km of Bekasi

urban center for example, settlement area in 1989 was less than rice field, but since

1997 settlement was more than agricultural area, and no more than rice field area in the

year 2017.In two semi-urban areas next to Bekasi, Cikarang and Cileungsi, it was seen

that rice field in 1989 was greater than settlement, but it continued to decrease and

finally 1¡-2007 settlement exceeded the rice field area. 1n2007 Bekasi, Cikarang and

Cileungsi will become one big settlement area adjoining the Jakarta metropolitan city.

This area is becoming urbanized as an influence of the Jakarta metropolitan complex

which has the largest urban population in Indonesia. Construction of new housing or

industry in this area still continues, as well as the construction of the toll way between

Cikampek-Purwakarta- Padalarang.

Finally the results of future land use/land cover change simulations found that the

Markov Cellular automata model enables us to simulate the future land use/land cover

change. This simulation can run based on the Markov transition that is created from

information relating to past and recent land use/land cover change (Markov rules) and

based on a Markov transition scenario as far as the scenario could represent future land

use/land cover change. In this case, the appropriate background information and

239
significant factors that could have an eflect on the future land use/land cover change

are needed in order to create suitable a Markov transition scenario.

7.2 lmplications

7 .2.1 Policy

Housing and industrial estate expansion has had an influence on the booming

development around the big cities of Indonesia, especially in Jakarta and its adjoining

areas- JABOTABEK. This pattern is being repeated around other mega city centres in

Southeast Asia such as Manila, Bangkok and Ho Chi Min city. Up until July 1997 the

Indonesian National Land Agency (BPN) had issued permits for development of about

l2l 000 hectares of land on the periphery of Jakarta (Bogor-Tanggerang-Bekasi) to

private developers, of whict¡ developers have acquired 46000 hectares (FirmarU 1998).

This policy has had the implication that land use/land cover in this area has changed

rapidly. The recent physical development of Jakarta and its periphery is characterised

by conversion of agriculture land to urban land uses. This conversion of agriculture

land to urban land uses such as settlement was clearly recognised in the land use/land

cover change detection in this study using satellite image for the years 1989 to 1997.

On the other hand, the future land use/land cover simulation obtained by using a

Markov Cellular automata model indicates that urbanisation in this area will continue,

from both a simulation based on Markov transition of land use/land cover from 1989-

1997 añ a Markov transition scenario which assumes that the economic crisis since

1998 will affect the rate of land use/land cover change in the study area. This

240
urbanisation will also happen around other big cities in Indonesia such as Surabaya'

Medan and Makasar.

The implications on the government policy of the results are:

l. Land use/land cover change detection using remote sensing and GIS technique

is useful to obtain the recent land use/land cover condition on the spatial base.

2. Detection of land use/land cover change by using remote sensing and GIS as

well as simulation of the future land use/land cover change are useful to support

government policy in order to obtain optimal land use or land allocation

arrangement that can eliminate land use conflict. For example, not taking

productive land out of agricultural until it is necessary to do so.

3. The government has to consider that rapid land use change conversion from

agricultural land to urban uses that were found in this study can be used as an

input in order to arrange the land permits issued and to avoid a lot of idle land

(,lahan tidur") as well as other environmental problems such as sedimentation,

flood and dectining carrying capacity ofthis area.

7.2.2 New Research

As satellite images have different levels of spatial, as well as spectral, and temporal

resolution, further study is needed in order to compare the results from this study with

results from studies with more detailed classes of land use/land cover categories. Soft

classification using fuzzy classifiers, for example, could be used to try and obtain

better land use-land cover classification within agriculture land uses or within urban

land use whicl¡ in this study, was generalised to avoid mixed classification.

24r
Markov Cellular automata model based on Markov transition from the past to the

recent land use/land cover condition has a limitation on representing the future land

use/land cover change if the past has significantly different condition from the future,

such as declining or increasing economic conditions that could have an effect on land

use/land cover change. This limitation can be eliminated by using a modified Markov

transition based on consideration of some significant factors to obtain the transition

that can represent the future land use/land cover change.

The further study on this simulation needs to obtain the future land use/land cover

change which is simulated correctly, simulated based on the Markov transition that can

represent the future land use/land cover. For example, a Markov transition scenario

that was created based on the accurate and appropriate background information that

could affect the future land use/land cover change, such as information of relationship

between population growth or economic growth and land use/land cover change rate.

There is a pressing need to do similar studies in other mega SE Asia cities and other

smaller Indonesian cities such as Surabaya, Medan and Makasar. Further research also

needs to develop new factors of significance to be included in simulations ofthe future

land use/land cover change based on the Markov Cellular automata model.

242
 REFERENCES

Aaviksoo K. (1995). Simulating vegetation dynamics and land use in a mire

landscape using a Markov model, Landscape and Urban Planning., 3I:129'142.

Adger W. Neil and Katrina Brown (1994). Land use and the causes of Global
Warming. John S/iley, England.

Alba Ma. J.B, Flaviana.H., Anita. J., and Estella. B.J. (1997). The Philippine Land
use and Landover change case study. SARCS-START-IGBP, Bangkok, Thailand.

Alomar D, C. Gallo, M. Castaneda, R. Fuchslocher (2003). Chemical and

discriminant analysis of bovine meat by near infrared reflectance spectroscopy
(NRS), Meat Science. 63 :441-450.

Anderson J.R., E.E. Hardy, J.T. Roach, and R.E.Witmer (1976). A Land Use and
Land Cover Classification System for Use with Remote Sensor Data, Profesional
Paper 964,U5. Geological Survey, Reston, VA.

Angelici G.L., Brynt, N.A, and Friedman, S.Z. (1977). Techniques for Land Use
Change detection using Landsat imagery. Proceeding of the American Society of
Photo grametry, 217 -228 .

Arai T and Tetsuya A. (2003). Empirical analysis for estimating land use transition
potential functions-case in the Toþo metropolitan region. Computer, Environment

and Urban Systems, (Article in Press)

Balz¡er H., Paul'W. Braun and V/olfgang Kohler (1998). Cellular automata models
for vegetation dynamics. Ecological Modelling, 107: ll3-125.

Bargman,V. (1970). Group representations in mathematics and physic. Springer-

Verlag, Berlin.

243
Baredo J.I, Marjo.K, Nialt.Mccormick, carlo.L.(2002). Modelling dynamic
spatial processes: simulation of urban future scenarios through cellular automata.
Landscape and Urban Planning, 981 :1 -16'

Batty M and Kim K.S. (1992). Form follows function: reformulating urban
population density function. Urb an Studie s, 29 :l 043'107 0 .

Batty .M and Y.Xie (1994). From cell to Cities. Environment and Planning B:
Planning Design, 2l : 83 I -848.

Batty M, H.Coucelis, M. Eichen (1997). Editorial: Urban system a cellular

automata. Environment and Planning B: Planning Design, 24:159-164.

Batty M, and Y. Xie (1997). Cellular Automata as the basis of integrated dynamic
regional modelting. Environment and Planning B: Planning Design,24:235'246'

Batty M. (1993). Urban evolution on the desktop: Simulation with the use of
extended cellular automata. Environment and Planning, A.30:1943-1967 .

Batty M, Yichun Xie and Zhanh Sun (1999). Modelling Urban dynamics through
GlS-based cellular aatomata.Computer, Environment and Urban System,

23(3):205-233.

Baulies X. and Szejwach, G. eds. (1997), Survey of need, gaps and priorities on
data for land use -land cover change research. LUCC Data Requirements
Workshop, LUCC Report Series No.3, Barcelona, Spain.

Bajjouk T, J.Populus and B. Guillaumont (1998). Quantification of Subpixel

Cover Fractions Using Principal Component Analysis and a Linier Programming

Method: Application to the Coastal Zone of Roscoff (France). Remote Sensing of
Environment, 64:1 53 -165 .

244
Beberoglu, C. D.Lioyd, P.M Atkinson and P.J. Curran (2000). The Integration of
Spectral and Textural Information using Neural networks for Land Cover Mapping
in the Mediterraneat. Computer and Geoscienc, 26:385'396

Bell E.J. (1974). Markov Analysis of Land Use Change an Application of

Stochastic Processes to Remotely Sensed Data. Geographical Analysrs, 28(1): 56-
66.

Bell E.J. (1977). Markov analysis of Land use change: continuous time and

stationary process. Socio-Econ Planning Science, ll:13'17 .

Bell E.J and Hinojosa, R.C. (1977). Markov anaþsis of Land use change:
continuous time and stationary processes. Socio-Econ Planning Science. Il:.13'17.

Benstein J.D. (1994). Land use consideration in urban environmental management,

Washingfon DC.

Bertels K, J..M. Jacques, L. Neuberg and L Gatot ( 1999). Qualitative company

performance evaluation: Linear discriminant analysis and neural network models.

European Journal of Operational researcft. 115:608-615.

Besussi E, Arnaldo.C and Enrico.R.(1998).The Diffirsion City of the Italian North-

East: Identification of Urban Dynamics Using Cellular Automata Urban Models.
Computer, Environment and Urban Systems, 22(5):497 -523.

Bilsborrow R.E, Okoth-Ogendo, H.W.O. (1992). Population driven changes in land

use in developing countries. Amabio,2l:3745.

Bicik lvan, Leos.J., Vit.Stepanek.200l Land use change and their Social
driving forces in chechia in the 19th - 20ú Centuries. Land use Policy,

18:65-73

245
Blaikie P, Cannon.T, Davis.I and ìVisner.B. (1994) At Risk: Natural Hazards,
People s vulnerability and Disaster, London.UK.

Blackard J.A and Denis J.DDD. (1999). Comparative accuracies of artificial neural
networks and discriminant analysis in predicting forest cover types from
cartographic variables. Computers and electronics in Agriculture.24:131-151.

Briassoulis H. (2000) Analysis of Land Use Change: Theoretical and Modellling

Approaches, Regional Science.

Borak J.S., E.F.Lambin and A.H.Strahter (2000). The uses of temporal metrics for
land cover change detection at coarse spatial scales. International Journal of
Remote Sensing, 2l : l4l5 -1432.

Brera A.M. and F. Shahrokûi (1973). Application of Landsat Data to Monitor

Desert Spreadin in the Sahara Region. Proceedings of the l2th International
Symposium on Remote Sensing of Environment, AnnArbor, Michigan.

Bressan M, J. Vitria (2003). Nonparametric discriminant analysis and nearest

neighbor classifi cati on, P att er n Re c o gniti on Le tt er s . In Pre ss.

Bruzzone L. and D. Fernandez Prieto (2000). An adaptive parcel-based technique

for unsupervised change detection. International Journal of Remote Sensing,

2l:817-822.

Bryan B. A. (2000). Modelling the propagation of land clearance using cellular

for nature conservation. The 4tr Intemational
automata and the implications
Conference on Integrating GIS and Environmental Modelling (GIS/ÆM4):
problem, Prospects and Research Needs, Banü Alberta Canada'

Brown L.A.(1970) On the Use of Markov Chains in Movement Research.

Economic Geography, 46:393'403 .

246
Brown .D.G, B.C.Pijanowski and J.D.Duh.(2000).Modeling the relationships

between land use and land cover on private lands in the Upper Midwest, USA.
Journal of Environmental management, 59 ; 247 -263 .

Burrough P.A. (1986). Principal of Geographical Information System for Land

Resources Assessment. Oxford University Press, England. UK.

Byme G.F., P.F. Crapper, and K.K. Mayo (1980). Monitoring Land Cover Change
by Principal Component Analysis of Multitemporal Landsat Data. Remote Sensing
of Environment, l0 : 17 5 -184.

Bryant C.R. (1936). "Agriculture and Urban developmenf in Pacione,M (eds.),

Progress in Agricultural Geography, Bechenham,UK.

Candau .J, Steen. R, Keith C.C. (2000). A couple cellular automaton model for
land use/land cover dynamics. The 4ú Intemational Conference on Integrating GIS
and Environmental Modelling (GIS/ÆM4): problem, Prospects and Research

Needs, Banff, Alberta, Canada.

Ceballos and M.J.Bottino (1997). The Discrimination of Scenes by Principal

Component Analysis of Multi-Spectral Imagery. International Journal of Remote
Sensing, 18:2437 -2449.

Chapin F.S. (1972). Urban Land use planning. University of Illinois, Urbana

Charlie N, Suwit. O, Kamerun .M. (1997). The Thailand Land use and Land cover
change case study. SARCS-START-IGBP, Bangkok, Thailand.

Chou Kue-Chen and David S/. Elrod (1998). Using Discriminant Function for
Prediction of Subcellular Location of Prokaryotic Proteins, Biochemical and
B i ophy s i c al Re se ar c h C ommuni c at ion. 252 :63 -68 .

247
Cecchini A. (1996). Urban modelling by means of Cellular Automata: generalised
urban automata with the help onJine (AUGH) model. Environment and Planning
B : Planning Design, 23 :721'7 32.

Cihlar J., Qingþan Xiao, J. Chen, J. Beaubien, K. Fung and R. Latifovic (1998).
Classification by Progressive Generalization: a New automated Methodology for
Remote Sensing Mutti Channel data. International Journal of Remote Sensing, 19:
2685-2704.

Claramunt C, Christine Parent, Stefano Spaccapietra and Marius Theriault. (1999).

Database Modelling for Environmental and Land Use Changes, in Srillwell.J, Stan
Geertman and Stan Openshaw (eds.) Geographical Information and Planning.
Springer-V erlag, Berlin.

clarke K.c., søcy Hoppen and Leonard J. Gaydos (1997). Methods and
Techniques for Rigorous Calibration of a cellular automaton model of urban
growth.

Cloke Paul.J. (19S9). Rural Land use planning in developed nation. Unwin
Hyman.ltd, UK

Coucelis H. (1997). From Cellular Automata to Urban models: new principles for
model development and implementation. Environment and Planning B: Planning
Design, 24:165-174.

Cohen W.B. and Maria Fiorella (1999)." Comparison of Methods for Detecting
Conifer Forest Change with Thematic Mapper Imagery". in Lunetta and
Christopher D.E.(eds). Remote Sensing Change Detection: Environmental
Monitoring Methods and Applications. Sleeping Bear, UK.

248
Coiner J.C. (1980). Using Landsat to Monitor Change in Vegetation Cover Induced
by Desertification processes. Proceedings of the l4th International Symposium on
Remote Sensing of Environment, AnrrArbor, Michigan.

Collin Clark (1977). Population growth and Land use. Macmillan. London.

Collin L. (1973).Industrial Size Distributions and Stochastic Process. Progress in

Geography, 5:12l-65.

Colwell J.E. and F.P. Weber (1931). Forest Change Detection. Proceedings of the
l5th International Symposium on Remote Sensing of Environment, Attrt Arbor,
Michigan.

Dawson A.H. (1934). The Land problem in the developed Economy. Croom Helm,

NSV/, Australia.

deBie CA, van Leeuwen JA and Zuidema PA.(1996). The Land Use Database. A
knowledge-based software prograrnme for structured storage and retrieval of user-

defined land use data set. Users' Reference Manual (v104, DOS). ITC, Enschede;
FAO, Rome: Agricultural University,'Wageningen.

Deep Peter. (1999). Change Detection in Remote Sensing Using Unsupervised

Fuzzy Classification. PhD dissertation University of Adelaide, Australia.

de Almeida C.M, Michael.B, Antinio.M.V.M, Gilberto.C, Britaldo.S.S.F,

Gustavo.C.C, Cassio.L.P.(2003). Stochastic cellular automata modeling of urban
land use dynamics: empirical development and estimation. Computer, Environment
and Urban Systems, (Anicle in Press)

de Koning G.H.J, P.H. Verburg, A.Veldkamp, L.O.Fresco (1999). Multi-Scale

Modelling of Land use change dynamics in Ecuador. Agricultural Systems,6l :77-
93.

249
Djoko S. (1996). Problems and prospects of Indonesia new towns development.
Paper presented at the International Seminar Planning and Development of
Satellite Towns in Southeast Asia. Organized by LIPI-The Japan Foundation,
Jakarta 25-27 June 1996.

Dobson, M. C., Pierce, L. E. and Ulaby, F.T. 1996, Knowledge-based land cover
classification using Ers-l/Jrs-l SAR composites, IEEE Transactions on Geoscience
and Remote Sensing, 34(l), 83 -99.

Du Qian, Chein-I Chang (2001). A linear constrained distance-based discriminant

analysis for hyperspectral image classification. Pattern Recognition.34:361-373.

Du Qian, Hsuan R (2003). Real-time constrained linear discriminant analysis to

target detection and classification in hyperspectral imagery. Pattern Recognition
.36:l-72.

Duin Robert P.W, Elzbieta P, Dick de Ridder (1999). Relational discriminant

analysis. P attern Re co gnition Letters.2O : 1 1 75 - 1 I8I .

Eastman. J.Ronald (2001). Tutorial ldrisi32 Release 2. Clark Labs. Clark

University, USA.

Edge Dustin (2001). Urban Growth Simulation using GlS-based Cellular

Automata. MSc Thesis Universþ of Adelaide.

Ehrlich Paul.R, Anne H, Holden John P. (1977). Ecoscience. Population, resources,

environment. W.H.Freeman and Company, USA.

Elvidge C.D. and Z. Chen (1995). Comparison of Broadband and Narow-Band

Red Versus Near Infra Red Vegetation Indices. Remote Sensing of Environment,
54:38-48.

2s0
ERDAS. (1997). Image Manual V.8.3, 4ú edn. ERDAS, Inc. 2081 Buford
Highwas, NE, Atlanta, Georgia, USA.

Feighner J.P, LLLL. Sverdlov, H. Abajian, J. Hlavj4 G. Nicolau (2000).

Discriminant analysis to separate a study population by treatment subgroups in a

pilot clinical trial with a new pontapeptide antidepressant (INN00835). European
Neuropsycho pharmacologt. 1 I (3).5184.

Firman Tommy. 1997. Land Conversiton and Urban Development in the Northem
Region of West Java, Indone sia. Urb an Studie s, 3 4(7), I 027 -1046.

Firman Tommy.2000. Rural to urban land conversion in Indonesia during boom

and bust periods. Land Use Policy,lT: 13-20.

Fisher, R.A.(1936). The use of multþle measurements in taxonomic problems,

Ann. Eugenics, 7, I 79-1 88.

Fischer Gunther and Laixiang Sun. (2001). Model based analysis of future land-use
development in China. Agriculture, Ecosystems and Environment, 85: 183-176.

Fisher P.F. and S.Pathirana (1990). The Evaluation of Fuzzy Membership of Land
Cover Classes in the Suburban Zone. Remote Sensing of Environment, 34: l2l-132.

Fisher P.F. and S.Pathirana (1993). The Ordering of Fttzq Land - Cover

Information Derived from Landsat MSS Data. Geocarto Internatioanl, 3:5'14.

Filho R. 4., Yosio E. Shimabukuro Q002). Digital processing of a Landsat-TM

time series for mapping and monitoring degraded areas caused by independent gold
miners, Roraima State, Brazilian Amazon. Remote Sensing of Environment,T9:42-
50.

251
Foody G.M. (2001). Monitoring the Magnitude of Land - cover change around the
Southern Limits of the Satrara. Photogrametric Engineering and Remote Sensing,
67:841-847.

Foody G.M. (2000). Estimation of sub-pixel land cover composition in the

presence of untrained classes. Computers & Geosciences, 26:469-478.

Foody G.M and D.S. Boyd (1999). Detection of partial land cover change

associated with the migration of inter-class tansitional zones. International

Journal of Remote Sensing, 20:2723'27 40.

Foran B.D. (1937). Detection of Yearly Cover Change with Landsat MSS on
Pastoral Landscapes in Central Australia. Remote Sensing of Environment, 23:333'
350.

Foster B.C. (1935). An examination of some problems and solution in monitoring

urban areas from satellite platforms. International Journal of Remote Sensing,

6:139-151.

Frederick M., Iwan G., Handoko., Asep Karsidi, M. Muchlis., Andi Rahmadi.,
Hartanto Sanjaya., Laju Gandarum. (2001). Case Studies and Diagnostic Models of
the Inter-a¡nual Dynamics of Deforestation in Southeast Asia: Is the missing sink
for Carbon in Land cover Change?. Indonesian Case Study. Final report submitted
to: The United Nations OfEce of Project Services (LINOPS).

Fung T and W. Siu (2000). Environmental qualþ and its changes, an analysis
using NDVI. International Journal of Remote Sensing, 2l:I0ll'I024.

Fung T. and E. LeDrew (1937). The Application of Principal Component Analysis

to Change Detection Using Various Accuracy Indices. Photogrametric Engineering
and Remote Sensing,53: 1649-1658.

252
Fung T. and E. LeDrew (l9SS). The Determination of Optimum Threshold Levels
for Change Detection. Photogrametric Engineering and Remote Sensing, 54:.1449-
1454.

Gemmell F., Jari Varjo and Mikael S. (2001). Estimating Forest cover in a boreal
Forest test site using Thematic Mapper data from two dates. Remote Sensing of
Environment, 77 : 197 -2ll .

Gobin 4., P. Campling, J. Feyen (2002). Spatial analysis of rural land ownership.
Landscape and Urban Planning,55: 185-194.

Goldblum C., Tai-Chee Wong (2000). Growth, crisis and spatial change: a study of
haphazard urbanization in Jakarta, úrdonesia. Land Use Policy, 17:29-37 .

'Warren
Gordon D.K, R.P. and \I/illiam D.P. (1986). Fruit Tree Inventory with
Landsat Thematic Mapper Data. Photogrametric Engineering and Remote Sensing,
52: l87l-1876.

Gordon S. (1930). TJtilizng Landsat Imagery to Monitor Land Use Change'

Remote Sensing of Environment, 9: 1 89-1 96.

Government Regulation of the republic of Indonesia Number 10 year 2000

Conceming the Map accuracy forthe Spatial Planning in Indonesia.

Green. K., Dick Kempka, and Lisa Lackey (1994). Using Remote Sensing to
Detect and Monitor Land Cover and Land Use Change. Photogrametric
Engineering and Remote Sensing, 60: 33I -337 .

GTZ. (2000). Final Report Private Public Parbrership (PPP) Project to Promote
Spatial Planning and Forest Mapping, Jakarta, Indonesia

253
Hashiba H., K. Kameda, S. Uesugi and S. Tanaka (2000). Land use change analysis
of Tama river basin with different spatial resolution sensor databy Landsat MSS
and TM. Advance Space Research, 26 (7):1069-1073.

Hathout S. (2002). The use of GIS for monitoring and predicting urban growth in
East and West St Paul, V/innipeg, Manitoba, Canada. Journal of Environmental
Mana ge me nt, 66 :229 -23 8.

Haub Carl (1996). Future Global Population Growth, in Ramphal, Shridath and
Steven W.S (eds). Population Growth and Environmental issues.Praeger

Publishers. London.

Helmschrot J., Wolfgang Albert Flugel Q002). Land use characterisation and

change detection analysis for hydrological model parameterisation of large scale

afforested areas using remote sensing. Physics and Chemßtry of the Earth, 27;
7tl-718.

Henderson, F.M. and Xia, Z.G. 1997, SAR applications in human settlement

detection, population estimation and urban land use pattern analysis - a status

report, IEEE Transactions on Geoscience and Remote Sensing,35(l), 79'85.

Hidding M.C and Andre, T.J.T. Q002). Beyond fragmentation: new concepts for
urban-rural development. Landscape and Urban Planning, 58:297'308.

Heilig, G. K. 1997, Anthropogenic factors in land use change in China, Population

and Developemnt Review, 23(l), 1 39 -l 68.

Holden H and E.LeDrew (1993). Spectral Discrimination of Healthy and Non-

Healthy Corals Based on Cluster Analysis, Principal Components Analysis and
Derivative Spectro sco py. Remote Sensing of Environment, 65 :217'224 .

254
Hoover E.M and Frank Gianatani (1999). An lntroduction to Regional Economics.
htþ ://www.rri.wvu.edu/WebBook/Gia¡ratani.

Howard D.M, P.J.A. Howard and D.c. Howard. (1995). A Markov Model
Projection of Soil Organic Carbon Stores Following Land Use Changes. Journal of
Environmental management, 45 : 287'302.

Howarth P.J. and G.M. Wickware (1931). Procedures for Change Detection Using
Landsat Digital Data. Inte rnational Journal of Remote Sensing, 2: 27 7 -29 I .

Hubefy C.J. (1994). Applied Discriminant Analysis. John Wiley and Sons, Inc.
New York.

Hugo, Graeme J. (1931). Population Mobilþ in 'west Java, Gajah Mada

University Press. Yo gyakarta.

Hugo, Graeme J. (1932). Evaluation of the Impact of Migration on Individuals,

Households and Communities, in ESCAP, National Migration Surveys. X.
Guidelines for Analyses, Comparative study on Migration, Utbanization and

Development in the ESCAP Region, Survey Manuals, New York: United Nation,
pp.l89-215.

Ingebristsen S.E. and R.P.J. Lyon (1985). Principal Component Anaþsis of

Multitemporal Image Pairs. Internqtional Journal of Remote Sensing, 6:687-696.

Irwin E. G., Jacqueline Geoghegan (2001). Theory, datA and methods: developing
spatially explicit economic models of land use change. Agriculture Ecosystems and
Environment, 85:7-23.

25s
Jadhav R.N., M.M. Kimhoti and A.K.Kandya (1993). Grassland

Mapping/lvlonitoring of Banni, Kachchh (Gujarat) using Remotely Sensed Data.

International Journal of Remote Sensing, 14:3093-3103.

JayadrnataJ.T. (1999) .TalaGuna Tanah dalam Pembangunan Pedesaan, Perkotaan

dan Wilayah. ITB, Bandung.

Jensen J.R (1936) Introductory Digital Image Processing. Prentice Hall,

Englewood Cliffs, New Jersey.

Jensen J.R., E.W. Ramsay, H.E.Mackey, E.J. Christensen and R.R. Sharitz (1987).

Inland Wetland Change Detection Using aircraft MSS Data. Photogrametric

Engineering and Remote Sensing, 53 : 251 -529.

Jensen J.R., K. Rutchey, M.S. Koch and S. Narumalani (1995). Inland Wetland
Change Detection in the Everglades Water Conservation Area 2A Using a Time
Series of Normalized Remotely Sensed Data. Photogrametric Engineering and
Remote Sensing, 6l : 199 -209.

Jensen J.R (1996) Introductory digital image Processing: a remote sensing

perspective. Prentice-Hall, USA.

Jha C.S. and N.V.M.Uni (1994). Digital Change Detection of Forest Conversion of
Dry Tropical Forest Region. Internationol Journql of Remote Sensing, 15:2543-
2552.

Joyce A.T., kvey,J.H,and Bums,G.S.(1980). The use of Landsat MSS data for
detecting Land use change in forestland. Proceedings of the l54th International
Symposium on Remote Sensing of Environment. Held in Costa Rica in 1982, pp.
979-988.

2s6
Ju Junchang, eric D.Kolackzyk, Sucharita Gopal (2003). Gaussian mixture
discriminant analysis and sub-pixel land cover charactenzation in remote sensing,
Remote Sensing of Environment. 84 :5 50 :5 60.

Juppenlatz. M, Xiaofeng Tian (1996). Geographic Information Systems and remote

sensing. McGraw-hill Book Company. Asutralia.

Karsidi 4., SHM. Tampubolon., Moch. Muchlis, Hartanto. S. (1997). The

Indonesia Land use and Land cover change case study. SARCS-START-IGBP,
Bangkok, Thailand.

Karathanassi V., CH. Iossifidis and D. Rokos (2000). A Texture-based

Classification method for classiffing built areas according to their density.

International Journal of Remote Sensing, 2l: 1807-1823.

Kaufrnann R.K., Karen C.S. (2001). Change Detection, accufacy' and bias in a

sequential analysis of Landsat Imagery in the Pearl Delta, China: Econometric

techniques . Agriculture, Ec osystem and Environment, 85 :95 - I 05.

Klecka W.R. (1930). Discriminant Analysis. In Sullivan John.L (eds). (1980).

Series: Quantitative Applications in the Social Sciences. Sara Miller McCune, Sage
Publication, Inc.

Krummer D and Tumer II B.L.(1994). The human causes of deforestation in

Southeast Asia. Bio scienc e, 44(5):323 -330.

Kwarteng A.Y and P.S. Chavez, Jr. (1998). Change Detection Study of Kuwait
City and environs Using Multi-temporal Landsat Thematic Mapper dala.
International Journal of Remote Sensing, 19:1651'1662.

257
Lachenbruch, P.A. (1973). Some Results on the multþle group discriminant
problem, in T.Cacaullus (eds), 1973, Discriminnat Analysis and Application,
Academic Press. New York.

Lambin E.F. and A.H. Strahler (1994). Change Vector Analysis in Multitemporal
Space: a Tool to Detec and Categorise Land Cover Change Processing Using High

Temporal Resolution Satellite Data. Remote Sensing of Environment,4S:231-144.

Lambin E.F. (1994). Modelling Deforestation Processes. A Riview. TREES

SERIES B. Research ReportNo.l, Luxembuourg.

Lambin E.F, Baulies X, Bockstael N, Ficcher G, Krug T, Leemans R, Moran E.F,

Rinfrrss R.R, Sato Y, Skole D, Tumer II B.L and Vogel C. (1999). Land Use and

Cover Change Implementation Strategy. IGBP Report No.48:IHDP Report No.l0,

Stockholm: Intemational Geosphere-Biosphere Programme Secretariat.

Lindgren, David.T (1985). Land use planning and remote sensing. Martinus
Nij hoff Publishers, Dordrecht, The Netherlands.

Liu X.H, Claes A. (2003). Assessing the impact of temporal dynamics on land -use
change modeling. Computers, Environment and Urban Systems, A¡ticle in Press.

Li Xiubin and Hong Yang (2000). Cultivated land and food supply in China. Land
Use Policy,lT:73-88.

Li X., A.G.O. Yeh (1998). Principal Component Analysis of Stacked Multi-

Temporal Images for the Monitoring of Rapid Urban Expansion in the Pearl River
Delta. International Journal of Remote Sensing, 19:1501-1518.

Li Xia, and Anthon.G.O.Yeh. (2001). Modelling sustainable urban developmentby

the integration of constrained cellular automata and GIS. Int.J.Geographical

Information Science, 14 Q):13 l'152.

258
Li Xia, and Anthony.G.O.Yeh (2001). Calibration of cellular automata by using
neural networks for the simulation of complex urban systems. Environment and
P I annin g. A, 23 :l 445 -l 462.

Li Xia, and Anthony.G.O.Yeh (2001). Zontng Land for Agricultural Protection by

the Integration of Remote Sensing, GIS and Cellular Automata. Photogrametric

Engineering and Remote Sensing, 67 (4):47 l-477 .

Li Xia, and Anthony G.O.Yeh Q002). Urban simulation using Principal

Components Analysis and Cellular Automata fro Land-Use
Plannng.P hoto gr ametric Engine ering and Remote Sensing, 6 I (4): 3 4l -3 5l .

Lillesand T.M. and R.W. Kiefer (19S7). Remote Sensing and Image Interpretation.
Wiley, USA.

Lillesand T.M. and R.W. Kiefer (2000). Remote Sensing and Image Interpretation.
Wiley, USA.

Li Lin, Yohei.S, Harhong.Z. Q002\ Simulating spatial urban expansion based on a

physical process. Landscape and Urban Planning, 975:.l'10.

Logsdon.M.G, Earl.J.B and Frank. V.W. (1997). Probability Mapping of Land Use
Change: A GIS lnterface for Visualizing Transition Probabilities. Computers,

Environment and Urban Systems, 20(6): 3 89-3 98.

Lopez E., Gerardo Bocco, Manuel Mendoza, Emilio Duhau (2001). Predicting
land-cover and land-use change in the urban fringe, a case in Morelia city, Mexico.
Landscape and Urban Planning, 55:271-285

Ltrnetta R.S., Cristopher D.Elvidge (1999). Remote Sensing Change Detection:

Environmental Monitoring Methods and Applications. Sleeping Bear, UK.

259
Lyon J.G., D. Yuan, R.S. Lunetta and c.D. Evidge (1998). Change Detection
Experiment Using Vegetation Indices. Photogrametric Engineering and Remote
Sensing,64:143-150.

Madhavan B.B., Sachio Kubo, N.Kurisaki and T.V.L.N. Sivakumar (2001).

Appraising the anatomy and spatial growth of the Bangkok Metropolitan area using
a Vegetation-impervious-soil model through remote sensing. International Journal

of Remote Sensing, 22:7 89'806.

Marija J.Norusis (1990). SPSS.Advanced Statistics User's Guide.SPSS Inc,

Chicago.

Martin L.R.G and Howarth,P.J. (1989). Change detection accuracy assessment

using Spot multispectral imagery of the rural-urban fringe. Remote Sensing of

Env ir onment, 3 0 :5 5 -66.

Malila V/.4. (1930). Change vector analysis: and approach for detecting forest
changes with Landsat.Proceeding of the 6th Annual Symposium on Machine
Processing of Remotely Sensed Dafa. Purdue University,pp.326-335.

Manel S, Jean-Marie D, Steve J.O. (1999). Comparing discriminant analysis,

neural networks and loistic regression for predicting species distributions: a case
study with a Himalayan river bird, Ecological Modelling.l2}:331-347 .

Mather A. S. (1986). Land (Jse, Longman, New York

Matsura S., Abduratrim.A., Mochtar. J. (1997). The Malaysia Land use and Land
cover change case study. SARCS-START-IGBP, Bangkok, Thailand.

Meyer V/. B. and B.L. Tumer II. (1992). Human Population Growth and Global
Land-Use / Cover Change. Annual. Rev. Ecol. System,23 :39-61.

260
Meyer W. B., and Tumer II, B. L. (eds). (1994). Change in Land Us and Land
Cover: A Globlal Perspectiv¿, Cambridge University Press, UK

McConnell W.J and Emilio F.M (eds.). (2000). Meeting in the Middle: The
challenge of meso-level integration. LUCC Report Series No.5. LUCC Focus 1

Office.

McNeil (1994. Toward a Typology and Regionalization of Land - Cover and Land
'W.
use Change: Report of Working Group B. in Meyer 8., and Tumer II, B. L.
(eds). (1994). Change in Land Us and Land Cover: A Globlal Perspective,
Cambridge University Press, UK

Mettemicht G. (1999). Change Detection Assessment Using Fuz'zy sets and

Remotely Sensed data an Application of Topographic map revision. /,SPRS
Journal of Photogrammetry qnd Remote Sensing,54:221-233.

Morisette, J.T., Khonam, S., Maco, T. 1999, Land cover change detection

enhanced with generalized linier models, International Journal of Remote Sensing,

20(t4),2703 -2721.

Moser S.C. (1996). A Partial Instructional Module on Global and Regional Land
use/cover change: Assessing the Data and searching for General Relationship.
Geoj ournal, 39 (3). 241 -250.

Moy H.L., A. Cotonnec, L. Le Du, A. Chardin and P. Petez (2001). A

Comparison of Parametric Classification Procedure of Remotely Sensed Data

Applied on Different Landscape Units. Remote Sensing of Environment,T5Q): na

-187.

Muchoney D.M. and B.N. Haack (1994). Change detection for Monitoring Forest
Deforestation. Photogrametric Engineering and Remote Sensing,60:1243-1251-

261
Muniyati (2000). Wetland change detection on the Kafue, Zambia, by
Classification of a multitemporal Remote Sensing Image dataset. International

Journal of Remote Sensing, 2l :1787 -1806

National Academic of Science (2000). Growing Population, Changing Landscapes.

1.

Nasution L.I. (1999). Kebijakan Penatagunaan Tanah di Indonesia, in Proceeding

Perubahan tutupan lahan dan penggunaan tanatr di Indonesia dalam kaitannya

dengan study LUCC-IGBP, Karsidi,A., Hartanto,S, Laju.G (eds). Dit. TISDA-BPP

Teknologi Jakarta.

Nelson R.F. (1933). Dekcting Forest Canopy Change Due to Insect Activity Using
Landsat MS S. P hoto gr ametric Engine ering and Remote Sensing, 49 :7303 -13 I 4'

Reppprot (19S9). Review of Phase I Results Java and Bali. Direktorat Bina

Program, Departemen Transmigrasi, Jakarta.

OECD. (1976). Land use policies and agriculture. Organization for Economic Co-
operation and Development, Paris.

Olson H. (1995). Reflectance Calibration of Thematic Mapper data for Forest

Change Detection. International Journal of Remote Sensing,l6: 81-91.

Oetter D.R., Wanen B.C, Mercedes B, Thomas K.M and Robert E.K. (2001). Land
cover Mapping in an Agricultural setting using Multi-seasonal Thematic Mapper
data. Remote Sensing of Environment,76Q): I39-I55.

Pastor .J and Carol.A.J. (1992). Using Simulation Models and Geographic

Information System to Integrate Ecosystem and Landscape Ecology, in Naiman.

262
J.R (eds.). W'atershed management: balancing sustainability and environmental
change. Springer-V erlag, New York.

Pathan, S. K., 7 others, et al. 1993, Urban growth trend analysis using GIS
techniques a case study of the Bombay Metropolitan, International Journal of
Remote Sensing, 14(17), 3169 - 3179.

Pilon P.G., P.J. Howarth, R.A. Bullock and P.O. Adeniyi (1988). An Enhanced
Classification Approach to Change Detection in Semi - Arid Environment.
Photo grametri c Engineering and Remote Sensing, 54 : 77 09 - 17 16.

Pontius R.G. Jr., Joseph D. Cornell, Charles A.S. Hall (2001). Modeling the spatial
pattem of land-use change \¡rith GEOMOD2: application and validation for Costa
Rjca. Agriculture, Ecosystems and Environment, 85: 191-303.

Qi J., A.R. Huete, M.S. Moran, A. Chehbouni and R.D. Jackson (1993).
Interpretation of Vegetation Indices Derrived from Multi-Temporal SPOT Images.
Remote Sensing of Environment,44: 89-101.

Quarmby N.A. and J.L. Cushnie (1939). Monitoring Urban Land Cover Changes at
the Urban Fringe from Spot HRV Imagery in South - East England. International
Journal of Remote Sensing, l0: 953 -963.

Rachman.Mahbub Muhammed.(1997). Identification of Land Use and Land Cover

Using Band Rationg Technique.
anl.

Rais Jacub. (1997). Planning sustainable land management at regional level: the
Indonesian case, ITC Journal. 3:4.

Ram Balak and A. S. Kolakar. (1993). Remote sensing application in monitoring

land use changes in Arid Rajasthan, International Journal of Remote Sensing,

t4(17),319t- 3200.

263
Rao, C.R.(1966). Discriminant function between composite hypotheses and related
probleq Biometrikø, 53, 339-345.

Rao K.S., Rekfia Pant (2001). Land use dynamic a Landscape change pattem in a
typical micro watershed in the mid elevation zone of central Himalay4 India.
Agricultur e, Eco sy stems and Environment, 86 :l 13 -123 .

Reyment,R.A. (1973). The discriminant function in systematic biology, in

T.Cacaullus (eds), 1973, Discriminnat Analysis and Application, Academic Press.
New York.

Rhind David, Ray Hudson (1980). Land Use. Mathuen, New York.

Richard J.A. (1984). Thematic Mapping from Multitemporal Image Data Using the
Principal Components Transformation. Remote Sensing of Environment,16 3546.

Rindfuss, R.R. and Stern, P.C. 1998, Lingking remote sensing and social science:
the need and the challenges, ln People and Pixels, Eds. Diana Liverman,
Emilio,F.Moran., Ronald,R.Rendfuss. And Paul,C.Stem., National Academy Press

,Washington.

Reid R.S., Kruska R.L., Muthui N., Taye 4., Wotton S., Wilson C.J., Mulatu W.
(2000) Land use and land cover dynamics in response to changes in climatic,
biological and socio-political forces: The case of south western Ethiopia.
Landscape Ecologt, I 5 : 339-355.

Ridd M.K and Jiajun Liu (1998). A Comparison of Four Algorithms for Change
Detection in an Urban Environment. Remote Sensing of Environment,63:95-100.

264
Robinson C.J. (1979). A critical review of the change detection and urban
classification literature. Technical Memorandum CSCÆM-7916235, Computer
Sciences Corporation, Silver Spring, MD.

Robinson J, Stephen Brush, Ian Douglas, T.E. Gradael, Dean Cttactz, Winifred
Hodge, Diana Liverman, Jerry Melillo, Richard Moss, Alexei Naumov, George
Njirv, Joyce Penner, Peter Rogers, Vemon Ruffan, James Sturdevant.(l99a). Land

Use and Land Cover Projections: Report working groups C.in Changes in Land
Use and Land Cover: A Global Perspective.

Saefulhakim S, Eman.R, Dyah R.P, Dyah T.S. (1999). Pengembangan model

system interaksi antar aktiffitas sosial ekonomi dengan perubahan penggunaan
lahan, Lokakarya HDP-LUCC Indonesia, 16 Nopember 1999,Jakarta.

Sage Colin.1994. Population and hcome in Mayer and Turner II (eds). Change in
Land use ønd land cover: A Global Perspective, Cambridge, UK: Cambridge
University Press: 263 -285 .

Sandy I Made (1995). Tanah: Muka Bumi. PT.Indographia-FMIPA. Universitas

Indonesi4 Jakarta.

Sandy I Made (1977). Penggunaan Tanah (Land Use) di Indonesia. Publikasi

No.75. Direktorat Tata Guna Tanah, Direktorat Jenderal Agraria, Departemen

DalamNegeri, Jakarta.

Santoso Harry (1999). Pemantauan Perubahan Penutupan dan Penggunaan Lahan

Hutan di Indonesia. in Prosiding Perubahan tutupan lahan dan penggunaan tanah di
Indonesia dalam kaitannya dengan study LUCC_IGBP, Karsidi,A., Hartanto,S,
Laju.G (eds). Dit. TISDA-BPP Teknologi Jakarta.

Salami A.T (1999). Vegetation Dynamic on the Fringes of Lowland Tropical

Rainforest of Southwestern Nigeria-an Assessment of Environmental Change with

265
Air Photos and Landsat TM. International Journql of Remote Sensing, 20:1169-
1181.

Salem.B.B., A. El-Cibahy, and M.El-Raey (1995). Detection of Land Cover

Classes in Agro-ecosystems of Northern Egypt by Remote Sensing. International

Journal of Remote Sensing, 16:2581'2594.

SARCS (1997). The Finat Report of LULC Case Study in the four countries:
lndonesia, Malaysia, Philippine and Thailand. SARCS Secretary Bangkok.
Thailand.

Scholten H.J, Rob van de velde, Piet Rietveld and Maarten Hilferink. (1999).
Spatial Information Infrastructure for Scenario Planning: The Development of a
Land Use Planer for Holland, in Srillwell.J, Stan Geertman and Stan Openshaw
(eds.). Geo graphical Information and Planning. Springer-Verlag, B erlin.

Sembaloni .F. (1997). An Urban and regional model based on cellular automata.
Environment and Planning B: Planning Design,24:589-612.

Serneels S, Eric F.Larrbin (2001). Proximate cause of land use change in Narok
District, Kenya: as spatial statistical model. Agriculture, Ecosystems and
Environmenf. 85:65-81

Shi W and Matthew, Y. C.P. (2000). Development of Voronoi-base cellular

automata - an integrated dynamic model for Geographic lnformation System,
Intemational Journal Geographic Information Science, 1 4(5) 45 5 -47 4.

Silatahi, S.B. and Putu Suweken (1999). Tingkat Perkembangan Penggunaan

Tanah. In Karsidi. A et al (eds): Prosiding Lokakarya Prubahan Tutupan Lahan dan
Penggunaan Tanah di Indonesia dalam kaitannya dengan Study LULCC-IGBP.
BPPT, Jakarta.

266
Singh A. and A.Ha:rison (19S5). Standardized Principal Component.International
Journal of Remote Sensing, 6: 883-896.

Singh A. (1939). Digital Change Detection Techniques Using Remotely Sensed

Data.International Journal of Remote Sensing, 10: 989-1003.

Sinowski W, KKK. Auerswald (1999). Using relief parameters in a discriminant

analysis to stratifr geological areas with different spatial variability of soil
properties, Geoderma. 89:l 13-128.

Sirakoulis G. Ch.,L Karaffllidis, A. Thanailakis (2000). A cellular automaton

model for the effects of population movement and vaccination on epidemic
propagatio n. Ec olo gic al Modelling, 133 : 209'223 .

Skole. D and Tucker C.(1993). Tropical deforestation and habitat fragmentation in

the Amazon: satellite data from 1978 to 1998. Science,260:1905-1910.

Solecki W.D. (2001). The role of Global-to local linkages in land use/land cover
change in south Florida.Ec olo gic al Economics.3T :339 -3 5 6.

Sommers, L.M. 1981. The Spatial Character of Land Use. In Lounsbury,

J.F.,L.M.Sommers and E.A. Fernald (eds.). Land Use: A Spatìal Approach.

KendallÆItmt. USA.

Song C., Cutis E.S/, Karen C.S, Mary P.L and Scott A.M. (2001). Classification
and Change Detection Using Landsat TM Data: When and How to Correct
Atmo spheric Effects? Remote Sensing of Environment, 75 :230'244.

Srillwell. J, Stan Geertman and Stan Geertman and Stan Openshaw (eds.) (1999).
Geographical Information and Planning. Springer-V erlag, Berlin.

267
Schmidt, M. G., Schreier, H. E., Shalì, P. B. 1995, A GIS evaluation of land use
dynamics and forest soit fertility in a watershed in Nepal, International Journal of
Geographical Information System,9 (3), 317 -327.

Srillwell. J, Stan Geerfinan and Stan Openshaw (1999). Developments in

Geographical Information and Planning, in Srillwell.J, Stan Geertrnan and Stân

Geertrnan and Stan Openshaw (eds.) Geographical Information and Planning.

Springer-V erlag, B erlin.

Star. J.L, John. E.Estes, Kenneth. C. McGwire (eds) (1997). Lrtegration of

Geographic Information System and Remote Sensing. Carnbridge Universþ Press.
New York.

Statsoft (2 003 ). Discriminant fu nction Analysis.

http ://www. statsoftinc.corn/textbook/stdiscan.html

Stefanov W.L., Michael S.R and Philip R.C. (2001). Monitoring Urban Land cover
Change. An expert system approach to land cover classification of semiarid to arid
urban centers. Remote Sensing of Environment,TT:173'185

Stamp, L.D. (1931). The Land utilization survey in Britain. Methuen & Co Ltd.
London.

Stow D.4., L.R Tinney and J.E.Estes (1980). Deriving Land use/Land Cover
Change Statistics From Landsat: a Study of Prime Agricultural Land. Proceedings
of the l4h International Symposium on remote Sensing of Environment, Ann Arbor
Michigan, ppl227 -1237 .

Sudradjat Adjat, Apik Karya, Atmen. Z, Dedth. F.R, Diana.H, Ekowati Ch.,
Ening.V/., Gatot.P., Giyati, Gunawan.S., Mulyono, Prijanto.P., Slamet.S','Wastra,

268
S.Y. Randha and Triyono (2001). Dasar Penerapan Kebijakan Proyek Lahan
Gambut Sejuta Hektar. Program Pasca sarjana IPB, Bogor.

Sullivan John.L (eds). (1980). Series: Quantitative Applications in the Social

Sciences. Sara Miller McCune, Sage Publication, Inc.

Tampubolon S.M.H. (1999). Pengaruh beberapa factor sosial-ekonomi dan

kependudukan terhadap perubatran tutupan and penggunaan lahan di lndonesia, in
Prosiding Perubahan tutupan lahan dan penggunaan tanah di Indonesia dalam

kaitannya dengan study LUCC_IGBP, Karsidi,A., Hartanto,S, Laju.G (eds). Dit.

TISDA-BPP Teknologi Jakafia.

Takeyama M, and Helen.C. (1997). Map dynamics: integrating Cellular Automata

and GIS through Geo-Algebru.Int.J.Geographical Information Science, 11 (l) 73-

91.

Tandy Clifford R.V. (1973). Industrial Land use and Dereliction, in Land use and
landscape Planning, Lovejoy Derek (eds.). Leonard Hill Books. Great Britain.

Tappeiner U, Erich Tasser, Dottfried T. (1998). Modeling vegetation patterns using

natural and anthropogenic influence factors: preliminary experience with a GIS
based model applied to an Alpine area' Ecological Modelling.ll3:225'237

Thenkabail P.S. and C. Note (1996). Capabilities of Landsat-5 Thematic Mapper

rM Data in Regional Mapping and Characterization of Inland Valley Agro
ecosystems in West Africa. International Journal of Remote Sensing, 17:1505-
1538.

Thomas, D.S.G., Deborah, S. (1997), Understanding the dynamic of social and

environmental variability: the impacts of structural land use change on the

269
environment and peoples of the Kalahari, Botswana, Applied Geography, l7(l),
tt-27.

Thomascik Tomas, Amarie Janice Mah., Anugerah Nontjik., Mohamad Kasim

(1997). The Ecological of Indonesian Seas. Part One. Periplus edition (Hk) Ltd.
CV. Java Books

Todd W.J. (1977). Urban and Regional Land use change detected by using Landsat
datz. Journal of Research by the US. Geological Survey, 5:527'534.
Tumer B.L., Hanham, R.O. and Portaharo, A.V. (1977). Population pressure and
agricultural intensity, Annals of the Association of American Geographers, 6(3),
384-396.

Tu Qiang, Dafu Ding (2003). Detecting pathogenicity islands and anomalous gene
clusters by interactive discriminant analysi s, FE MS Micr obiolo gt Le tter s, 221 :269 -
275.

Tumer B.L, Robert Q.Hanham and Anthony V.Prtararo ( 1977). Population

pressure and Agricultural Intensity. Annals of the Association of American
Ge o gr apher, 6 : 3 84 -396.

Tumer B.L, and Meyer. W.B. (1994)."Global land use and land-cover change: an
overview," in W. Meyer and B.L. Turner (eds), Changes in Land Use and Land
Cover: A Global Perspective, Cambridge, UK: Cambridge University Press, pp 3-
10.

Tumer 8.L., D. Skole, S. Sanderson, G.Fischer, L. Fresco and R.Leemas (1995).

Land use and land cover change: Science/Research Plan. IGBP Report No.35.HDP
ReportNoT.IGBP and HDP Stockfiolm and Geneva.

Tumer, B.L and W.B. Meyer.199I. Land use and land cover in global environment
change: Considerations for Study, International Social Science Journal, 130:669-
679.

270
Tucker C.J. (1979). Red and Photographic Infrared Linier Combinations for
Monitoring Vegetation . Remote Sensing of Environment, I :127 -150 .

Túcker C.J, and Maxwell, E.L (1976). Sensor design for Monitoring Vegetation
Canopies. P hoto grametric Engine ering and Remote Sensing, 42: 1399 .

van Deusen P.C. (1994). Conecting Bias in Change Estimates from Thematic
Maps. Remote Sens ing of Envir onment, 50 :67'7 3 .

Tucker C.J., Dregden,H.E and Newcomb,W.'W (1991). Expansion and Constaction

of the Salrara Desert from 1980 to 1990. Science, 255:299'301'

UU. Number22year 1999 regardiing Regional Govemance. Republic of Indonesia.

UU. Numb er 24 year 1992 regarding Spatial planning. Republic of Indonesia.

Volgemann J.E. (1938). Detection of Forest Change in the Green Mountains of

Vermont Using Multispectral Scanner Data. Remote Sensing of Environment,
9:1187-1200.

Veldkamp 4., E.F. Lambin (2001). Predicting Land use Change. Agriculture,
Ecosystem and Environment, 85 :1 -6.

verburg P.H., G.H.J. de Koning, K.Kok, A. veldkamp, J. Bouma. (1999). A spatial

explicit allocation procedure for modelling the pattern of land use change based
upon actual land use , 116 : 45-61.

Vink A (1975). Land (Ise in Advancing Agriculture, Springer-Verlag, Berlin,

Germany.

Wagner D.F. (1997). Cellular automata and geographic information systems.

Environment and Planning B: Planning Design,24:219'234.

271
Ward D., Stuart R.P. and Alan T.M. (2000). Monitoring Growth in Rapidly
Urbanizing Areas Using Remotely Sensed Data. Professional Geographer,52(3):
371-386.

V/atson N. and David W. (2001). Pre-classification as an Aid to Improvement of

Thematic and Spatial Accuracy in Land Cover Maps Derived from Satellite
Imagery. Remote Sensing of Environment,Ts Q):267'278.

rùVeismiller R.4., S.J. Kristot D.K.Scholz, P.EEE. Anuta and S.A. Momin (1977).
Change Detection in Coastal Zone Environments. Photogrametric Engineering and
Remote Sensing, 43 : 7 533 -1539.

V/eng Q. (2002). Land use change analysis in the Zhujiarg Delta of China using
satellite remote sensing, GIS and stochastic modelling. Journal of Environmental
Management, 64 :27 3 -283 .

Whitaker Jean.S. (1997). Use of Stepwise Methodology in Discriminant

Analysis.htþ I I encae.neUff/tamu/STEPWls.htm

White R, Engelen G. (1992). Cellular dynamics and GIS: modelling spatial

complexity. WP-9264. Research Institute for Knowledge Systems, Maastricht, The
Netherlands.

White R., Engelen G. (1993). Cellular automata and fractal form: a cellular
modelling approach to the evolution of urban land use pattems. Environment and
Planning A, 25 : ll7 5 -1199.

White R. (1997). Cellular automata as the basis of integrated dynamic regional

modelling. Environment and Planning : P I anning and Des i gn, 24 : 23 5'246.

272
White R, and G.Engele.I.U. (1997). The use of constrained cellular automata for
igh-resolution modelling of urban land-use dynamics. Environment and Planning
B : Planning Design, 24:323 -347 .

White R., G. Engelen (2000). High-resolution integrated modelling of the spatial

dynamics of urban and regional systems. Computers, Environment and Urban

Systems,24:383400.

V/orld Bank (1990). A World Bank Country Study. Indonesia: Sustainable,

Development ofForest, Land, and water. The World Bank Washington D.C.

World Bank (1999). Indonesia-Jatiluhur Water Resources Management. The World

Bank Washington D.C.

World Bank (2001). Indonesian Environment and Natural Resource Management

in a Time Transition. htp ://eap.worldbank.org/indonesia/environment.

Wilsson E.H, and Steven,A.S.Q002). Detection of forest harvest type using

multþle dates of Landsat TM imagery. Remote Sensing of Environment, 80:385-
396.

Wolman M Gordon (1993). Population, Land Use, and Environment: A Long

History in Jolly Carole.L and Barbara Boyle Torrey (eds) Population and Land Use
in Developing Countries. National Academy Press. Washington, D.C.

Wu F. (1998). An experiment on the generic polycentricity of Urban Growth rn a

cellular automatic city. Environment and Planning B: Planning Design,25:.731-

753.

V/u F. (1993). Simulating urban encroachment on rural land with fuzzy-logic-

controlled cellular automata in a GIS. Journal of Environmental Management, 53;
293-308.

273
Wu F and C.J.Webster (1998). Simulation of Land development through the
integration of cellular automata and multi criteria evaluation. Environment and
Planning B : Planning Design, 25 : 103 -126.

V/u Fulong. (1998). Simland: a prototype to simulate land conversion through the
integrated GIS and cellular automata with AHP-derived transition rules.

Int.J.Geographical Information Science, 12 (1): 63-82.

Wu Fulong, and Christopher.J.V/. (2000). Simulating artificial cities in a GIS

environment: Urban Growttr under alternative regulation regimes.
Int. J. G e o gr aphi c al Info r mati on S c i enc e, 1 4 (7 ) : 625 -648 .

Xie Yichum. (1996). A Generalized Model for cellular Urban Dynamics.

Geographical Analysis, 28 (14): 3 50'37 3 .

Yeh A.G.O, ffid Xia Li (1993). Sustainable land development model for rapid
growth areas using GlS. Int.J.Geographical Information Science, 12 Q):169-189.

Yeh A.G.O, and Xia Li (2000). A cellular automata model to simulate development
density for urban planning Environment and Planning B: Planning Design, 29
431-450.

Yeh A.G.O, and Xia Li (2001). A constrained cellular automata model for the
simulation and planning of suitable urban form by using GIS. Environment and
Planníng B : Planning Design, 28: 733-7 53.

Yuan Ding, Christopher D.Elvidge and Ross S.Lunetta .(1999). Survey of

Multispcetral Methods for Land Cover Change analysis, in Ross S.Lunetta and
Christopher D.Elvidge (eds). Remote Sensinf Change detection Environmental
Monitoring Methods and Applications. Taylor & Francis.Ltd. Londodn.

274
Young S.S, Benyamin,P., Jokisch,B., Ogneva,Y., and Grren,A.(I991). Global Land
use/cover: Assessment of data and some general relationships. Apendixt I of The
Report of the Ad-hoc Working group on Global Land Use Change, Committee for
Research on Global Environmental Change, Social Science Research Counsil.

Young S.S and Anyamba,A. (1999) Comparison of NOAAÆrIASA PAL and

NOAA GVI data for vegetation change studies over China. Photogrametric
Engineering qnd Remote Sensing, 65 :679'688.

Young S.S. and C.Y.Wang (2001). Land cover change analysis of China using
global-scale Pathfinder AVHRR Land cover (PAL) data,1982-1992. International
Journal of Remote Sensing, 2l: 1457 -1477 .

Yuan D, ChristopherD. Elvidge and Ross S. Lunetta (1999)." Survey of

Multispectral Methods for Land Cover Change Analysis". in Lunetta and
Christopher D.E. (eds). Remote Sensing Change Detection: Environmental
Monitoring Methods and Applications. Sleeping Bear, UK.

Yuan D, and Chris Elvidge (1993). NACL Land Cover Change Detection Pilot
Study: V/ashington D.C. Area Experiments Remote Sensing of Environment, 66:.
166-18.

Yue T. X., Ji Yuan Liu, Sven Erik Jorgensen, Qin Hua Ye (2003). Landscape
change detection of the newly created wetland in Yellow River Delta. Ecological
Modelling,3225: l-ll.

ZhanX., R.A.Sohlberg, J.R.G.Townshend, C.DiMiceli, M.L.Caroll, J.C.Eastman,

M.C. Hansen, R.S. DeFries (2002). Detection of land cover changes using MODIS
250 mdata. Remote Sensing of Environment, 33: 336-350.

275
 APENDIX IV
Tablel com nents transformation TM 1990
Gomp 1 Comp 2 Comp 3 Gomp 4 Gomp 5 Gomp 6
Tova¡. 89.12 7.83 2.69 0.17 0.13 0.07
Eisen value. 1895.66 166.53 57.32 3.52 2.68 1.40

eiqvec.l 0.15 0.30 0.49 0.06 -0.55 -0.59

eiqvec.2 0.11 0.21 0.36 0.04 -0.39 0.81
eigvec.3 0.17 0.46 0.47 0.03 0.00
eiqvec.4 0.68 -0.65 0.29 -0.16 0.08 -0.01
eigvec.5 0.65 0.34 -0.51 0.45 -0.05 0.00
eigvec.6 0.22 0.34 -0.25 -0.07 0.00

Comp I Comp 2 Gomp 3 Gomp 4 Comp 5 Gomp 6

TM Band I 0.75 0.46 0.44 0.01 -0.10 -0.08
Tllll Band 2 0.79 0.44 0.44 0.01 -0.10 0.15
TM Band 3 o.72 0.59 0.35 0.00 0.12 0.00
TM Band 4 0.96 -0.27 0.07 -0.01 0.00 0.00
TM Band 5 0.15 -0.13 0.03 0.00 0.00
TM Band 6 0.89 0.40 -0..¡7 -0.15 -0.01 0.00

Table2: Princi transformation TM 1991

Gomp I Gomp 2 Comp 3 Comp 4 Comp 5 Gomp 6
o/o
Yãi. 85.90 8.83 4.76 0.20 0.18 0.12
Eigen value. 2505.61 257.68 138.77 5.82 5.33 3.58

eiqvec.l 0.18 0.37 -0.36 -0.38 -0.01 0.74

eigvec.2 0.24 0.28 -0.45 -0.47 -0.04 -0.66
eiqvec.3 0.19 0.51 -0.31 -0.06 -0.06
eiqvec.4 0.62 -0.65 -0.36 0.15 0.18 0.08
eigvec.5 0.65 0.16 0.57 -0.08 -0.46 0.00
eiqvec.6 0.25 0.27 0.33 -0.04 0.8t -0.05

Comp I Gomp 2 Comp 3 Comp 4 Comp 5 Com 6

TM Band I 0.76 0.50 -0.36 -0.08 0.00 0.12
TM Band 2 0.87 0.32 -0.38 -0.08 -0.01 -0.09
TM Band 3 o.73 0.61 -0.27 0.14 -0.01 -0.01
TM Band 4 0.94 -0.32 -0.13 0.01 0.01 0.00
TM Band 5 0.08 0.20 -0.01 -0.03 0.00
TM Band 6 0.89 0.31 0.28 -0.01 0.14 -0.01
Table3 com transformation TM 1992
ana
Comp I Comp 2 Comp 3 Comp 4 Gomp 5 Comp 6
o/o
va¡. 91.22 4.85 3.65 0.15 0.12 0.02
Eiqen value. 4471.12 237.69 178.69 7.20 5.85 1.10

eigvec.l 0.58 -0.27 -0.57 -o.44 0.01 -0.2

0.24 -0.03 -0.25 0.09 -0.05 0.93
eiqvec.3 0.28 0.19 -0.32 -0.19 -0.24
eigvec.4 0.46 -0.60 0.56 0.25 o.24 0.00
eigvec.S 0.52 0.53 0.44 -0.23 -0.45 0.00
eigvec.6 0.22 0.50 0.00 0.00 0.00

Gomp I Gomp 2 Comp 3 Gomp 4 Gomp 5 Gom 6

TM Band I -0.1 1 -0.19 -0.03 0.00 -0.01

TM Band 2 0.97 -0.03 -0.20 0.01 -0.01 0.06
TM Band 3 0.95 0.14 -0.22 0.11 -0.02 -0.01
TM Band 4 0.93 -o.28 0.23 0.02 0.02 0.00
TM Band 5 0.95 o.22 0.16 -0.02 -0.03 0.0c
TM Band 6 0.89 0.46 0.00 0.00 0.12 0.00

Table4: Princi com s transformation TM 1993

Comp I Gomp 2 Gomp 3 Gomp 4 Comp 5 Gomp 6
Yova¡ 92.55 4.57 1.71 0.98 0.16 0.03

Eiqen value. 7325.56 361.61 134.99 77.69 12.60 2.41

eigvec.l 0.70 -0.30 -0.03 -0.52 -0.02 -0.

eigvec.2 0.49 -0.29 -0.16 o.26 -0.05 0.77

o.20 -0.21 -0.22 -0.06 -0.51
eiqvec.3
eiqvec.4 0.36 0.34 0.81 0.24 0.20 0.00
eigvec.S 0.31 0.78 -0.35 0.00 -0.42 0.00

eiqvec.6 o.12 o.26 -0.38 0.00 0.0c

Comp I Comp 2 Gomp 3 Gomp 4 Comp 5 6

I -0.09 -0.01 -0.08 0.00 -0.01
TM Band
TM Band 2 0.99 -0.13 -0.04 0.05 0.00 0.03
0.91 -0.21 -0.13 0.37 -0.01 -0.04
TM Band 3
TM Band 4 0.94 0.19 0.29 0.06 0.02 0.0c

TM Band 5 0.85 0.48 -0.13 0.00 -0.05 0.00

Tllll Band 6 0.80 0.41 -0.36 0.00 0.25 0.00
Table 5 com transformation TM 1995
Comp I Gomp 2 Gomp 3 Comp 4 Gomp 5 Gomp 6
o/o
Yài. 90.09 5.43 4.10 o.22 0.14 0.03
E¡gen value. 4187.64 252.23 190.53 10.16 6.32 1.32

eiqvec.l 0.56 0.05 -0.66 -0.43 0.02 -0

eigvec.2 o.24 -0.07 -0.25 0.09 -0.04 0.93
eiqvec.3 o.27 -0.32 -0.23 0:82 -0.19 -0.25
eiqvec.4 0.46 0.77 0.28 0.27 0.21 0.00
eiqvec.S 0.54 -0.32 0.58 -0.25 -0.46 0.00
eigvec.6 0.24 -0.44 0.20 0.00 0.00

Comp I Gomp 2 Gomp 3 Comp 4 Gomp 5 Comp 6

TM Band I 0.02 -0.24 -0.04 0.00 -0.01
Tllll Band 2 0.97 -0.07 -0.22 0.02 -0.01 0.07
TM Band 3 0.94 -o.27 -o.17 0.14 -0.03 -0.02
TM Band 4 0.92 0.38 o.12 0.03 0.02 0.00
TM Band 5 0.96 -0.14 o.22 -0.02 -0.03 0.00
TM Band 6 0.90 -0.41 0.16 0.00 0.12 0.00

Table 6: Princi I com ents a transformation TM 1997

Comp I Gomp 2 Comp 3 Gomp 4 Comp 5 Comp 6
Ù/o
va¡. 86.41 8.76 4.34 0.25 0.17 0.07
Eiqen value. 2279.39 230.96 114.49 6.55 4.59 1.97

eigvec.l 0.22 0.26 0.55 0.03 -0

eiqvec.2 0.12 0.14 0.31 0.02 0.33 0.87
eiqvec.3 o.25 0.35 0.51 0.14 -0.73 0.00
eigvec.4 0.52 -0.79 0.25 -0.21 -0.07 0.00
eiqvec.S 0.70 0.18 -0.48 0.48 0.09 0.00
eiqvec.6 0.33 0.37 -0.23 -0.02 0.00

Gomp I Gomp 2 Comp 3 Gomp 4 Comp 5 Gomp 6

TM Band I 0.82 0.31 0.47 0.01 0.10 -0.05
TM Band 2 0.81 0.31 0.47 0.01 0.10 0.17
TM Band 3 0.84 0.37 0.38 0.02 -0.1 1 0.00
Tllll Band 4 0.90 -0.43 0.09 -o.02 -0.01 0.00
TM Band 5 0.08 -0.15 0.04 0.01 0.00
TM Band 6 0.92 0.33 -0.14 -0.13 0.00 0.00
TableT: matrix of Land use/Land cover 1989 vs. 1990
1 2 3 4 5 6 7

1. Forest 1) 5907 5417 8588 29 219 2808 298

2) 1.17 1.08 '1.7'l 0.01 0.04 0.56 0.06
2. Plantation 1) 677 25469 81 30 202 1293 4190 4202
2) 0.13 5.07 '|'.62 0.04 0.26 0.83 0.84
3. Rice field 1) 4466 25909 107979 2921 2176 22244 18939
2l 0.89 5.1s 21.48 0.58 0.43 4.42 3.77
4. Open land 1) I 472 8891 175'14 3500 21948 7058
2) 0.00 0.09 '|'.77 3.48 0.70 4.37 '|'.40

5. Settlement 1) 26 611 1775 409 9858 2562 1895

2l 0.01 0.12 0.35 0.08 1.96 0.51 0.38
6. Rice unplanted 1) 229 4502 27610 4687 4630 57444 10943
2) 0.05 0.90 5.49 0.93 0.92 11.43 2.18
7.Dry land/Grass 1) 11 752 826',| 6743 714 7896 7307
2l 0.00 0.15 1.U 1.U 0.'t4 1.57 1.45
B. Fish pond 1) 14 27 140 234 144 51 13 23
2) 0.00 0.01 0.03 0.05 0.03 1.O2 0.00
0 10 16 2 1 138 1
9. Water/Reservoir 1)
2) 0.00 0.00 0.00 0.00 0.00 0.03 0.00
Total 1989 1) 1 1338 63168 17'1389 32741, 22535 124343 50666
2) 2.26 12.56 34.09 6.51 4.48 24.73 10.08
1) unit = Hectare
2) percentage of that category in the study area
TableS: matrix of Land use/Land cover 1990 vs. 1991
'l 2 3 4 5 o 7
1. Forest 1) 9821 3287 8735 248 126 380 53
2) 1.95 0.65 1.73 0.05 0.03 0.08 0.01
2. Plantation 1) 2704 18053 14061 268 173 1466 274
2) 0.54 3.58 2.79 0.05 0.03 0.29 0.05
3. Rice field 1) 1353 1275 29970 16834 1304 10008 7853
2) o.27 0.25 5.95 3.34 o.26 1.99 1.56
4. Open land 1) 422 587 17',158 24975 '1777 17770 11294
2) 0.08 0.1'2 3.41 4.96 0.35 3.53 2.24
5. Settlemenl 1) 226 801 '1652 1259 7794 3937 274
2) 0.04 0.16 0.33 0.25 1.55 0.78 0.05
ô. Rice unplanted 1) 4695 2197 68852 9173 2247 47669 4020
2) 0.93 0.44 13.67 1.82 0.45 9.46 0.80
7.Dry land/Grass 1) 3837 17934 43619 6083 3664 27152 7858
2) 0.76 3.56 8.66 1.21 0.73 5.39 1.56
8. Fish pond 1) 218 28 1562 1793 98 3923 80
2) 0.04 0.01 0.31 0.36 0.02 0.78 o.02
E
9. Water/Reservoir 1) 37 I 54 51 I 128
2) 0.01 0.00 0.01 0.01 0.00 0.03 0.00
Total 1990 1) 23312 44169 185664 60684 '17191 112432 317'12
2) 4.63 8.77 36.86 12.05 3.41 22.32 6.30
1) unit = Hectare
2) percentage of that category in the study area
Table9: matrix of Land use/Land cover 1991 vs. 1992
1 2 3 4 5 6 7

1. Forest 1) 10819 1569 262 41 20 172 142

2) 2.15 0.31 0.05 0.01 0.00 0.03 0.29
2. Plantation 1) 5371 22017 668 218 137 1034 14304
2) 1.07 4.37 0.13 0.04 0.03 0.21 2.84
1) 5202 8713 6958 19971 1281 6881 1 26199
3. Rice field
2) 1.03 '1.73 1.38 3.97 0.25 13.66 5.20

4. Open land 1) 42 245 24129 13460 297 9940 4054

2) 0.01 0.05 4.79 2.67 0.06 1.97 0.80
5. Settlement 1) 168 357 1635 2882 7914 2882 4481
2) 0.03 0.07 o.32 0.57 1.57 0.57 0.89
6. Rice unplanted 1) 869 1970 33920 29018 6116 52436 391 15
2) o.'17 0.39 6.74 5.76 1.21 10.41 7.77

7.Dry land/Grass 1) 188 2141 366 8665 201 2494 20580

2) 0.04 0.43 0.07 1.72 0.04 0.50 4.09

8. Fish pond 1) 2 2 708 251 15 5249 96

2) 0.00 0.00 0.14 0.05 0.00 1.04 0.02
9. Water/Reservoir 1) I 2 4 0 0 2 5
2) 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Total 1991 1) 22668 37016 68650 74506 15980 143021 110277
2) 4.50 7.35 13.63 14.79 3.17 28.40 21.90
1) unit = Hectare
2) percentage of that category in the study area
Table 10: matrix of Land use/Land cover 1992 vs. 1993
'l 2 3 4 5 6 7
1. Forest 1) 8690 5080 6756 'l 0 4 'l 459 224
2) 1.73 1.01 1.34 0.00 0.01 0.09 0.04
2. Plantation 1) 1088 17108 5691 u 32 985 979
2) 0.22 3.40 1.13 0.01 0.01 0.20 0.19
3. Rice field 1) 3164 14190 26037 4612 4217 36479 7700
2) 0.63 2.82 5.17 0.92 0.84 7.24 1.53
4. Open land 1) 549 171 1 189 7713 446 923/. 973
2) 0.11 0.03 0.24 1.53
0.09 1.83 0.19
5. Settlement 1) 62 32 178 1420 8795 4282 25
2) 0.01 0.01 0.04 o.28 1.75 0.85 0.00
6. Rice unplanted 1) 260 1422 84355 33586 4414 81723 6048
2) 0.05 0.28 16.75 6.67 0.88 16.23 1.20
7. Dry land/Grass 1) 516 5759 14400 4987 2374 313'12 18705
2) 0.10 1.'|'4 2.86 0.99 o.47 o.¿¿ 3.7'l
8. Fish pond 1) 1 I 33 36 188 60 1 165 15
2) 0.00 0.01 0.01 0.04 0.01 0.23 0.00
1) 0 0 0 0 0 1 0
9. Water/Reservoir
2) 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Total 1992 1) 14348 43794 138641 52551 20379 165641 34669
2) 2.85 8.69 27.53 10.43 4.05 32.89 6.88
1) unit = Hectare
2) percentage of that category in the study area
Tablel l matrix of Land use/Land cover 1993vs.1995
I 2 3 4 5 6 7 I
l. Forest 1) 7282 1468 2926 239 37 150 240
2) 1.45 0.29 0.58 0.05 0.01 0.03 0.05
2. Plantation 1) 8657 14974 9401 386 43 396 2251
2) 1.72 2.97 1.87 0.08 0.01 0.08 0.45
3. Rice field 1) 2581 3037 15107 2494 126 78208 16263
2) 0.51 0.60 3.00 0.50 o.o2 15.53 3.23
4. Open land 1) 39 47 2805 2919 495 8108 3664
2) 0.01 0.01 0.56 0.58 0.10 1.61 0.73
5. Settlement 1) 35 40 1295 347 7339 1728 977
2) 0.01 0.01 0.26 0.07 '|'.46 0.34 0.19
ô. Rice unplanted 1) 2532 5993 61681 12740 6789 118/.51 45682
2) 0.50 1.19 12.25 2.53 1.35 23.52 9.07
7.Dry land/Grass 1) 130 356 2989 1235 10 5795 8970
2) 0.03 0.07 0.59 0.25 0.00 1.15 1.78
8. Fish pond 1) 4 0 46'l 97 49 1781 36 1
2) 0.00 0.00 0.09 o.02 0.01 0.35 0.01
,l I
9. Water/Reservoir 1) 20 2 2 0 I8
2) 0.00 0.00 0.00 0.00 0.00 0.00 0.00
fotal 1993 1) 21279 25917 96666 20458 14888 214635 78084 2
2) 4.22 5.15 19.19 4.06 2.96 42.61 15.50
1) unit = Hectare
2) percentage of that category in the study area
Table 12: C matrix of Land use/Land cover 1995 vs. 1997
1 2 3 4 5 6 7 I
1. Forest 1) 6581 8445 745 43 2 I
1.31 1.68 0.01 0.00 0.00
Plantation 1) 2230 14060 61 2050
o.44 2.79 1.55 0.01 5.29 0.41
Rice field 1) 3222 12797 32028 93053 1982
0.64 2.54 6.36 1.11 18.47
18.47 0.39
Open land 1) 19 31 1683 6507 9036 2087
0.00 0.01 0.33 1.29 1.79 0.41
Settlement 1) 12 1384 1119 ng52 142
0.00 0.01 0.27 o.22 1.95 5.55 0.03
Rice unplanted 1) 187 413 60752 1088 952 6211 937
0.04 0.08 12.06 0.22 0.19 12.33 0.19
7.Dry land/Grass 1) 93 338 10985 3664 435 36764 12287
2) 0.02 0.07 2.18 0.73 0.09 7.30 2.44
8. Fish pond 1) 1 0 2501 27 27 2609 1 1
2) 0.00 0.00 0.50 0.01 0.01 0.52 0.00
9. Water/Reservoir 1) 0 0 3 0 0 91 0
2) 0.00 0.00 0.00 0.00 0.00 0.02 0.00
Total 1995 1) '12u6 36'|'2'l 117875 18106 1 1818 260273 19493 2
2) 2.45 7.17 23.40 3.59 2.35 51.67 3.87
1) unit = Hectare
2) percentage of that category in the study area
Table'13: matrix of Land use/Land cover seven 1989 vs. 1997
1 2 3 4 5 6
1. Forest 1) 5934.87 8591.4 2913.66 370.62 17.37 2.43
2) 1.18 1.71 0.58 0.07 0.00 0.00
2. Plantation 1) 1689.93 24786.9 18920.7 6410.34 1137.06 7.47
2) 0.34 4.92 3.76 1.27 0.23 0.00
3. Rice field 1) 3619.98 24458.94 210381.6 32476.59 4721.49 5070.69
2) o.72 4.86 41.77 6.45 0.94 1.01
4. Open / dry land 1) 82.35 4668.57 40107.96 37864.08 '1564.92 281.43
2) 0.02 0.93 7.96 7.52 0.31 0.06
5. Settlement 1) 9.09 1549.44 17860.5 6074.55 15015.51 286.29
2) 0.00 0.31 3.55 1.21 2.98 0.06
6. Fish pond 1) 0.09 12.06 5406.39 197.55 72.il 15454.53
2) 0.00 0.00 1.07 0.04 0.01 3.07
7. Water/Reservoir 1) 2.O7 0.99 141.21 13.59 5.94 1.71
2) 0.00 0.00 0.03 0.00 0.00 0.00
Total 1989 1) 11338.38 64068.3 295732 83407.32 22534.83 21104.55 5
2) 2.25 12.72 58.71 16.56 4.47 4.19
1) Unit = hectare
2) Percentage of that category in the study area
 APENDIX
CHAPTER V

Table V.1: Land use/Land cover versus Slope

Types\Slope
Forest l) 330( 1962 1701 2378 1263 62C

2) 0.65 0.39 0.34 0.4'1 0.25 0.12

Plantation 1) 51380 6591 3404 2030 377 t92
a 0.40 0.07 0.04
z) 10.1? 1.30 0.6'l
Rice field 1) 286696 5566 2473 t7t4 710 1375
2) 56.75 1.10 0.49 0.34 0.14 0.2'1

Jpen/drv land 1) 8030t 1801 563 18? 24

2) 15.9{ 0.36 0.11 0.04 0.00 0.00
Settlement l) 21280 532 I l5 35 4 4

2) 4.21 0.11 0.02 0.01 0.00 0.0t

(
Fishpond l) 21021 0 0 0 0

) 4.16 0.00 0.00 0.00 0.00 0.00

Types\Slope
Forest 1) 9422 2570 1811 2486 r072 418
2) 1.8i 0.51 0.36 0.49 0.21 0.08
Plantation l) 45529 3693 2078 981 156 4C

2) 9.01 0.73 0.41 0.19 0.03 0.0r

Rice field 1) 26721t 7243 3639 2622 tt44 1721
1 0.72 0.52 0.23 0.34
s2.90 1.43
Open/dry land 1) 81473 2082 644 2t9 5 t7
) 16.13 0.41 0.13 0.04 0.00 0.00

Jettlement 1) 39323 862 97( 50 I 2

) 7.78 0.17 0.02 0.01 0.00 0.00

Fishpond 1) 2102f 4 0 C 0 0

2) 4.16 0.00 0.00 0.00 0.00 0.0c

1) = unit in Ha
2): Percentage ofeach category to total study area
Table V.2: Land useiI,and cover versus Phsyog¡aphy (1989-1997)

Types
Forest 42 3431 603',l t722
Plantation 985? r7135 t3369 230r6
Rice field 214006 19578 12762 s2t99
)oen/drv land 474s8 1r782 3r21 20523
Jettlement 11346 926 635 9064
20976 23 5 44

Types
Forest IT32 7216 8798 636
Plantation 9669 t6448 8722 17638
Rice field z0n09 17348 14372 50165
Open/drv land 47090 14988 3439 18923
Settlernent t9289 1425 589 1903 I
Fishpond 20791 68 32 159
 Table V.3: Land use/Land cover change versus Land system

Land use types 1989 (ha) [qqz (ha) Chanee (ha) P/o change lfi.and use types I
1989 (ha) 1997 (ha
Forest sl 47s 4791 o.o9lþorest 0

4830 3fi71 o.7olÞlantation 0

Plantation r2e4
-l _) I field 8614 r or
Rice field t7ßo2l 168s06
f,pen/dry land 31 l93l 30804 -3 0 land 158 2

Settlement 877sl 13963 51 I ettlement 122

Fish pond 3o7q so6s 1 t55ti l3

Land use types 989 997 Chanee (ha) þ/o change lþand use types 1989 (ha) 1997 (h
Forest ol o.oolþorest 5588 sl
Plantation rl tot l00l O.021þlantation 10191

field J 3ll 23 -0 field 741 113

land 1 0. land I I
38 0. 0 I 23 t
9 0. ish

use 997 use 989 997

Forest 15 5 7 sg44l l.l8lþorest 51 t

Plantation t82671 1470t 35661 -0.7l1þlantation 1825 I
Rice field l5ls2l 10863 -428e1 -o field 9001
Ooen/drv land t22jSl l4o5l lsl3l 0.36llOpen/dry land 5960 6

Settlement 3ssl 44',1 s9l o.o2llsettlement 635 lo

Fish pond 8l 15 7l o.oolþistr 115
 Table V.3 : Land use/Land cover change versus Land system

Land use types 1989 (ha) 1997 0¡,a\ 3hanee (ha) use rqsXha) ltggt $
Forest 0 0 0 0. orest 2os5l
Plantation 0 29 29 0.01 26ß41
Rice field 2681 t3t4 -1361 -0.27lh.ice field
Open/dry land 93( 2223 1288 0.25llopen/dry land
,1
Settlement 69 62 0.0lllSettlement
ll -11 0

use 989 997 use 989 997

orest 5 0. l3
-0.1 28

field 244 field 131

Cpen/dry land 566 961 395 0.0SllOpen/drv land tßl

t02 64 31 -o.o lllsettlement J
Settlement
Fish pond c 0 c o.00lFish pond
Tabte V.4: Land use/Land cover and Classes

LULC 1989 (ha) 1997 (ha) 1989 (ha) 1997 (ha) 1989 (ha) 1997 (ha)
Forest 770C 9961 933 1074 c c

Plantation 48331 38075 4032 6494 € 1€

Rice field 239792 214515 2945ç 20372 605t 226ç

Open/dry land 70117 9549C 5111 908€ 126C 591
Settlement 14605 1997€ 476! 7701 3302 7743
Fishoond 9595 1',1238 354 412 4 '|'2
Note:
Class I = 1 5 - 1 50 personlsqkm
C/ass t/ = 151 - 650 person/sqkm
Class //l = 651 - 1200 person/sqkm
 APENDIX VI

Lrlu'
+
High

Figure 1. Suitable for Forest

Lou'
+
High

Figure2. Suitable for Plantation

 Lorv

+
High

Figure 3. Suitable for Rice field

Lon
+
Higli

Figure 4. Suitable for Open/dry land

 Lorv

+
High

Figure 5. Suitable for Fishpond

 Table: Land use/Land cover within l0 km of urban and semi urban (Ha)
Urban and semi-urban Year Forest lPlantation Rice field Ooen / dry land Settlement Fis
Bekasi 1989 d 4oe 19921 331€ 532€
1997 d 64s 92ú 5't21 1493€
Scenario2.2007 d oe€ 2930 540C 20926
Scenario2.2017 ol 5e3 858 239i 26098
Cikarang Forest lPlantation Rice field Open/dry land Settlement Fis
198ç ol 308 23085 6333 1530
1997 d 1oo€ 1732t 7131 578C
Scenario2.2007 q ees 12903 7802 9556
Scenario2. 2017 d 722 949i 4768 1626€
0ileungsi Forest lPlantatÍo¡ Rice field Ooen/drv land Settlement Fis
'1989 77 4127 1245e 283€ 133€
1 997 ol 1301 12788 239C 334/,
Scenario2.2007 d 1æ1 9258 370i 6',172
Scenario2. 2017 d 1260 4457 238C 12885
l(arawang Forest lPlantation Rice field Ooen/drv land Settlement Fis
198ç 3l 't7U 1968f 7974 1861
1 997 187 25e5 't7953 7157 327ê
Scenario2.2007 32sl 2477 1428i 9800 4372
Scenario2.2017 204 1556 11597 9373 8527
3ikampek Forest lPlantation Rice field Ooen/drv land Settlement Fis
1989 21d 6374 1679[ 6719 1114
1997 274 7227 16131 558€ 1958
Scenario2. 2007 æd 7651 13500 709C 2422
Scenario2. 2017 æ81 6153 't3't24 673C 486
Purwakarta Forest lPlantation Rice field Open/dry land Settlement Fis
1989 109d 8008 8862 265t 1042
1 997 18æl 775/l 853C 2441 1500
Scenario2. 2007 35781 6228 77æ, 3252 '1879
Scenario2.2017 26æl so02 72U 2909 3907
 APENDIX VI.2
Markov transition matrix 8-year interval, 8 year to predict
1989 to 1997
Cells in transition to
LULC Class 1 Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Total 1989
Class 1 571 50 22300 47773 1092 116 0 26 12845
Class 2 104503 234134 297490 56736 I 8865 '|'42 0 71187
Class 3 44330 28765'l 1985648 6101 10 271561 82092 2299 328369
Class 4 4633 80063 40578/, 357597 75893 2502 185 92665
Class 5 250 16424 68175 22583 141809 1052 75 25036
Class 6 47 117 79397 4406 4476 145882 23 23434
Class 7 122 361 7587 s81 73 2329 50083 6113
fotal
exoected 211035 641051 2891855 1053105 512793 234000 52692 559653
Given Probabil of to
LULC Class 1 Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Total
Class 1 0.4449 0.1736 0.3719 0.0085 0.0009 0 0.0002
Class 2 0.1468 0.3289 0.4179 0.0797 0.0265 0.0002 0
Class 3 0.0135 0.0876 o.6047 0.1858 o.0827 0.025 0.0007
Class 4 0.005 0.084r 0.4379 0.3859 0.0819 0.0027 0.0002
Class 5 0.001 0.0656 0.2723 0.0902 0.5664 0.oo42 0.0003
Class 6 0.0002 0.0005 0.3388 0.0188 0.0191 0.6225 0.0001
Class 7 0.002 0.0059 o.1241 0.0095 0.0012 0.0381 0.8192
Total 0.6134 0.7485 2.5676 0.77U o.7787 0.6927 0.8207
 Markov transition matrix 8-year interval, 8 year to predict
1989 to 1997 (after rationalisation)

Cells in transition to
LULC Class 1 Class 2 Class 3 Class 4 Class 5 Class 6 Glass 7 Total 1989
Class I 57150 22326 47773 '1092 116 0 0 12845
Class 2 104503 234134 297490 56736 19007 0 0 71187
Class 3 30538 287651 20651 13 6101 10 208186 82092 0 328369
Class 4 4633 80063 405784 360099 76079 0 0 92665
Class 5 0 1377 431 39 37605 168248 0 0 25036
Class 6 0 0 79397 4406 4453 146093 0 23434
Glass 7 0 0 251 581 0 501 59804 6113
fotal
expected '196824 625551 2938947 1070629 476088 228687 59804 559653
¡ability of :hange to
Given Probabil
LULC Class 1 Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Total
Class I 0.4449 0.1738 0.3719 0.0085 0.0009 0 0
Class 2 0.1468 0.3289 o.4179 0.0797 o.0267 0 0
Class 3 0.0093 0.0876 0.6289 0.1858 0.0634 0.025 0
Class 4 0.005 0.0864 0.437S 0.3886 0.0821 0 0
Class 5 0 0.0055 0.1723 0.1502 0.672 0 0
Class 6 0 0 0.3388 0.0188 0.019 0.6234 0
Class 7 0 0 0.0041 0.0095 0 0.0082 0.9782
Total 0.606 o.6822 2.3718 0.8411 0.8641 0.6566 0.9782
 Markov transition matrix 8-year interval" 20 yeat to predict
From 1997

Cells in transition to
LULC Class 1
Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Total 1997
Glass 1 37548 30890 101903 17932 7827 1922 119 19814

Class 2 68023 78026 315341, 79085 40955 6767 235 58843

Class 3 97991 339536 1521046 645056 401639 111722 3745 312073
Class 4 24152 94917 513117 1776'17 115592 13815 564 93977
Class 5 7389 39120 18970s 61739 149452 5576 317 45329
Class 6 1224 7109 104584 '17772 1 1981 92626 94 23539
Class 7 213 723 8385 1349 547 2455 47088 6075
Total
expected 236539 590321 2754082 1000550 727993 234883 52163 559653
Given
Given Prc ;hange to
LULC Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Total
Class 1 0.1895 0.1559 0.5143 0.0905 0.0395 0.0097 0.0006
Class 2 0.1156 0.1326 0.5359 0.13/,4 0.0696 0.0115 0.0004
Class 3 0.0314 0.1088 o.4874 o.2067 o.1287 0.0358 0.0012
Class 4 0.0257 0.101 0.s6 0.189 0.123 0.0147 0.0006
Class 5 0.0163 0.0863 0.4185 o.1362 0.3297 0.0123 0.0007
Class 6 0.0052 0.0302 0.4443 0.0755 0.0509 0.3935 0.0004
ass 7 0.0035 0.0119 0.1 38 0.0222 0.009 0.0404 0.775
otal 0.3872 0.6267 3.0844 0.8545 0.7504 0.5179 0
 Markov transition matrix 8-year interval, 20 yeat to predict
From 1997 (Scenario)

LULC Class 1
Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Total 199
Class 1 148605 29721 9907 9907 0 0 0 19814
Class 2 73554 294217 132397 88265 0 0 0 58843
Class 3 0 0 2028478 624147 4681 10 0 0 3',12073

Class 4 0 140966 281932 1 87955 281932 46989 0 93977

Class 5 0 0 0 90660 362638 0 0 45329
Class 6 0 0 58848 23539 11770 14123/' 0 23539
Class 7 0 0 0 0 0 0 60759 6075
Total
expected 222159 464904 2511562 1024472 1124451 188223 60759 559653
Given Probability of
to
LULC Class 1
Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Total
Class 1 0.75 0.15 0.05 0.05 0.00 0.00 0.00
Class 2 0.12 0.50 0.22 0.15 0.00 0.00 0.00
Class 3 0.00 0.00 0.65 o.20 0.15 0.00 0.00
Class 4 0.00 0.15 0.30 0.20 0.30 0.05 0.00
Class 5 0.00 0.00 0.00 0.20 0.80 0.00 0.00
Class 6 0.0c 0.00 o.25 0.10 0.05 0.60 0.00
Class 7 0.00 0.00 0.00 0.00 0.00 0.00 1.00
Total 0.87 0.80 1.47 0.90 1.30 0.65 1.0000
 Cross-tabulation LIJLC 1997 and 2007 simulation based on Markov transition 1989-1997
LULC Forest Plantation Rice fieldl Open/dry land Settlement Fishpond Water/reserv
163154 47388 104361 4833 924 2
Forest
Plantation 2633 46089'.1 eos6d 36538 13843 190

Rice field 3219: 66309 26s464d 207s3 12303 16679

Open/dry land 73 965 I lSossl ß498',1 3281 2529
Settlement 49 4051 16760il 42s40 422618 3298
Fishpond 34 136 r6447 124 319 2t2692
I 0 re4l 0 5 I 6
Water/reservoir
lotal1997 198139 s88432 3t2}ßsl %977: 4s3299 235391 60

Cross-tabulation LULC 1997 utd20l7 simulation based on Markov transition 1989-1997

LT]LC Forest Plantation hjce field lOpen/dry land Settlement Fishpond Water/reser
Forest 159321 3848 3lsssl 7l8l 576 38

Plantation 8942 47s7881 772J71 22052 s730 483

Rice field 27791 soesel 26on42l 39380 13731 t2601

Open/dry land 384 lrss6l l59o49l 818s20 6669 381 I
Jettlement 1118 1 109 230 426180 6429

Fishpond 583 s27 t87s4l 155 413 211983

0 106l 5 c 46 6
lVaterlreservoir
lotal 1997 198 139 5384321 3t2}ßsl %977s 453299 235391 6
-d
Cross-tabulation Lu 1 997 and 2007 simulation based on Markov scenario
IULC Forest lPlantation lRice field Ooen/drv land lsettlement Fishpond uVater/reservoi fl
10601 1 0
Forest 16e86d 8668¿ 8380
Plantation 117d 4eß8el 85103 325381 15Í 13

Rice tield 21ßd 2316S1 259636€ 7nd I 62585 d

Open/dry land 587d ß'l4d 269651 86e97el 8106 1267.
Settlement e4l 547 160261 28eæl 445031 42?
:ishoond dde74 d c 159697
rAlater/reservoir o d c 0 607
dol 93977d 45329ç 235391 60759
Iotal 1997 1981391 588ß2 3120735

Cross-tabulation Lu 1 997 nal and 2017 simulation based on Markov scenario

LULC =orest lPlantation Rice field þpen/dry land þettlement Fishpond ¿llater/reservoir
:orest 17æ561 43472 121 33d c

60181 40949€ loee/ 3æ47 31 c

?lantation
Rice field 842d 106791 æ,32eill 10644 c 52743 0
Cpen/dry land ß151 15601 726o17 443 956
Settlement 10271 13osg 49s79d 151 10093
Fishpond d1€ -ztaszd
445e1 1215d 2 171599
Water/reservoir c d d c 607
d
fotal 1997 1981391 588432 s12oßd 9397751 453299 235391 60759