LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

 

DATE DUE DATE DUE - DATE DUE
AUGIIFN‘SJ , , . ..
a 30 DEB‘Oifan

 

\J

;Q"3¥1§ 90‘1“”

 

 

 

 

 

 

 

 

 

 

 

 

 

8/01 cJCIFiC/DateDuepes-pJS

DEVELOPMENT OF INTEGRATED PROGNOSTIC MODELS
OF LAND USE/LAND COVER CHANGE: CASE STUDIES IN BRAZIL AND CHINA
By

Yushuang Zhou

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Geography

2002

ABSTRACT

DEVELOPMENT OF INTEGRATED PROGNOSTIC MODELS
OF LAND USE/LAND COVER CHANGE: CASE STUDIES IN BRAZIL AND CHINA

By

Yushuang Zhou

Human beings are transforming significant portions of the earth’s land surface, which has
been of central concern to the international research community for most of the past
century. The conversion of natural forest biomass to agricultural activities plays a
significant role in global climate change, while the loss of agricultural land as results of
urban sprawl brings more concern about global food security. This dissertation aims to
improve the understandings of land use/land cover change (LUCC) dynamics, to
investigate activities and factors that control LUCC processes at various scales, and
ultimately, to build integrated prognostic models that can provide scientific projections
for land use/land cover change as reference to filture policy designs in developing

countries.

Case studies are carried out in the Brazilian Amazon and the East Region of China. As an
agricultural frontier in Brazil, the Amazonia is chosen to be a representative case to study
the consequences of agricultural activities on a primarily forested landscape. In contrast,
the East Region of China is a rather developed area with a long history of agricultural
development. It is selected to be a case for studying the change of agricultural landscape

in highly urbanized areas. The State of Rondonia in the Brazilian Amazon and the

Shanghai Municipality located in the East Region of China are further studied to account

for the spatial variations of land use/land cover change.

Remote sensing technique is applied to measure the magnitude and rate of land use/land
cover change, and various modeling method are developed for different modeling
purposes and based on data availability. A multinomial logit model of cultivated land use
change in the East Region of China shows a declining pace of cultivated land loss that is
highly subject to government policy intervention. A principal component regression
model is developed to examine how much agricultural land will be taken over by urban
sprawl in Shanghai, and how the frontier of urban sprawl is advancing deeper into the far
suburb areas. Results of the systematic model of land use/land cover change in the
Brazilian Amazon and the State of Rondonia demonstrate an accelerating speed of
deforestation in the coming decades. Nearly one third of the standing forest in the
Amazonia will be cleared within next twenty years, and more than 70% of primary forest

for Rondonia, a frontier state of agricultural development, will disappear from sight.

DEDICATION

To my family

For their infinite love and constant inspiration

And to my husband Gao

For being a wonderfill spouse

iv

ACKNOWLEDGMENTS

First and foremost I would like to thank Dr. David Skole for his constant support and
encouragement throughout my graduate studies. His academic guidance was essential to
the completion of this dissertation and has provided important insights to scientific
research. I would also like to express the gratitude for my major committee members, Dr.
Jiaguo Qi, Dr. Robert Walker and Dr. Jianguo Liu, for reading previous drafts of this
dissertation work and providing valuable comments that helped to improve the

presentation and contents of this dissertation.

I am also grateful to all the faculties and staff at the Department of Geography for their
great help to make me become a good geographer. I would like to thank Dr. John Hunter
for his constant support and inspiration. Many thanks also go to Walter Chomentowski
who provided valuable advice and technical assistance to this study, and to Mark
Cochrane, Eraldo Matricardi and Andrea Silveira for providing thoughtful comments and
useful information. I owe special thanks to Diane Cox and Deana Haner, two wonderfirl
ladies for their finest assistance and support. I also appreciate the friendship with other

colleagues and fellow graduate students at BSRSI and in the Department of Geography.

Last, but not least, I would like to thank my husband Gao for his support and love during
the dissertation writing. His encouragement made this dissertation possible. My parents,
my sister and my brother receive my deepest gratitude and love for their dedication and

the many years of support.

TABLE OF CONTENTS

List of Figures ............................................................................... xi
List of Tables ................................................................................ xiv
Introduction ................................................................................. 1
Reference .............................................................................. 4
CHAPTER 1. Land Use/Land Cover Change Monitoring and Modeling ........... 6
1.1 Land Use/Land Cover Change ..................................................... 6
1.2 Study Areas and Research Tasks .................................................. 7
1.2.1 LUCC Monitoring ...................................................... 9
1.2.2 From Pattern to Process ................................................ 10
1.2.3 LUCC Modeling ........................................................ 11
1.3 Spatial and Temporal Dimensions of LUCC .................................... 11
1 31 Spatial Dimension ..................................................... 12
1 .3 .2 Temporal Dimension ................................................. 15
1.4 LUCC Significance .................................................................. 19
1.4.1 Urban Sprawl and Agricultural Land Loss ......................... 19
1.4.2 Agricultural Expansion and Deforestation .......................... 21
1.5 Reference .............................................................................. 23
CHAPTER 2: Cultivated Land Use Change in the East Region of China .......... 29
2.1 Introduction ........................................................................... 29
2.2 Study Area: the East Region of China ............................................ 31

vi

2.3 Land Use Data ........................................................................

2.4 Distribution and Changes of Cultivated Land ....................................
2.5 Driving Forces of LUCC ............................................................
2.51 Agriculture Structure Change .........................................
2.5.2 Urbanization ............................................................
2.5.3 Rural Development ....................................................
2.5.4 Other Causes ............................................................
2.5.5 Summary ................................................................
2.6 Land Use Change Model ...........................................................
2.7 Modeling Results ....................................................................
2.8 Conclusions and Discussions .......................................................
2.9 References .............................................................................
CHAPTER 3: Urban Sprawl and Land Use Change in Shanghai, China ............
3.1 Introduction ...........................................................................
3.2 Shanghai: Urbanization and LUCC ...............................................
3.2.1 Urban Sprawl ...........................................................
3.2.2 Urbanization and Suburbanization ................. .. .................
3.2.3 Land Use Change Dynamics ..........................................
3.3 LUCC Monitoring ....................................................................
3.3.1 Image Pre-Processing ..................................................
3.3.2 Classification and Editing .............................................
3.3.3 Field Validation .........................................................
3.4 LUCC Modeling Method and Data ................................... . ............

vii

34

35

39

41

43

45

47

48

49

58

63

64

67

68

72

72

74

79

81

81

83

86

87

3.5 LUCC Modeling Results ............................................................ 91

3.5.1 PCA Results ............................................................. 91

3.5.2 MLR Results ............................................................ 94

3.5.3 PCR Results .............................................................. 96

3.5.4 Land Use Projections .................................................. 99

3.6 Conclusions ........................................................................... 105
3.7 References ............................................................................. 107
CHAPTER 4: Land Use/Land Cover Change in the Brazilian Amazon ............ 11 l
4.1 Introduction ............................................................................ 112
4.1.1 LUCC Monitoring ...................................................... 113

4.1.2 Process Understanding ................................................. 115

4.1.3 LUCC Modeling ........................................................ 116

4.1.4 Study Areas .............................................................. 119

4.2 Modeling Method and Data ........................................................ 120
4.2. 1 Conceptual Model ...................................................... 120

4.2.2 Modeling Method ...................................................... 121

4.2.3 Modeling Data .......................................................... 123

4.3 Land Cover Change Model ......................................................... 125
4.4 SociO-Economic Model ............................................................ 129
4.4.1 Importance of Agricultural Development ........................... 130

4.4.2 Agricultural Economic Processes .................................... 130

4.4.3 Modeling Agricultural Land Use Processes ........................ 133

viii

(1) Demand Expansion ................................................

(2) Expection Adjustment .............................................
(3) Inventory Correction ...............................................
(4) Cobb-Douglas Production Function ............................
(5) Agricultural Land Use Change ....................................
4.4.4 Other Proximate Sources ..............................................
(1) Logging ..............................................................
(2) Road Construction ..................................................
(3) Forest Fire ...........................................................
(4) Mining ................................................................
(5) Modeling ............................................................
4.5 Population Model .....................................................................
4.5.] Age Structure and Gender Composition ............................
4.5.2 Population Distribution: Urban vs. Rural ...........................
4.5.3 Population Occupation ................................................
4.5.4 Population Growth ......................................................
4.5.5 Modeling Population Growth and Distribution ...................
4.6 Sensitivity Analysis ...................................................................
4.6.1 Secondary Growth Regrowth Rate ..................................
4.6.2 Autonomous Abandonment Rate ....................................
4.6.3 Clearance Rate ..........................................................
4.6.4 Sensitivity Analysis Results ..........................................
4.7 LUCC Modeling Results and Validation .........................................

ix

136

137

137

138

139

140

141

142

143

143

144

146

147

148

150

151

152

157

158

158

159

160

160

4.7.1 Population Modeling Results and Validation ....................... 161

4.7.2 Socio-Economic Modeling Results and Validation ................ 165
4.7.3 Land Cover Modeling Results and Validation ...................... 168
4.8 Conclusions and Discussions ....................................................... 174
4.6 References ............................................................................. 178
CHAPTER 5: Summary .................................................................. 185
References ............................................................................. 190

Figure 1.1
Figure 1.2
Figure 1.3
Figure 1.4
Figure 1.5
Figure 2.1
Figure 2.2
Figure 2.3
Figure 2.4
Figure 2.5
Figure 2.6
Figure 2.7

Figure 2.8
Figure 2.9.1

Figure 2.9.2
Figure 2.9.3
Figure 2.9.4
Figure 2.10
Figure 2.11.1
Figure 2.11.2

Figure 2.11.3

LIST OF FIGURES

LUCC Triangle with Four Major Types of Changes ................
Maps Of Study Areas .....................................................
Bid-Rent Theory of Land Use/Land Cover Change ..................
Three Major Types of Diffusion Processes ............................
Land Use/Land Cover Change Diffusion Processes .................
Balance of Cultivated Land for All Economic Regions in China..
Study Area: the East Regions of China ................................
Distribution of Cultivated Land Per Capita by County ..............
Net Changes of Cultivated Land by County and by Province. . .
Conversion Processes of Cultivated Land .............................
Change of Multiple Cropping Indices over time .....................
Explanatory Powers of the Driving Forces in Each Province ......

A Conceptual Model for Cultivated Land Use Change .............
Urban Accessibility .......................................................

Rural Industry Output ....................................................
Elevation .......................................... . ........................
Slope .........................................................................
Flow Chart for Data and Modeling Methodology ....................
1980 Estimated ............................................................
1980 Observed .............................................................

1990 Estimated ............................................................

xi

13

16

17

32

33

37

38

40

47

49

50

54

54

54

54

57

61

61

61

Figure 2.11.4 1990 Observed ............................................................. 61

Figure 2.11.5 2025 Estimated ............................................................ 62
Figure 2.11.6 2050 Estimated ............................................................ 62
Figure 3.1 Study Area: the Shanghai Municipality ................................ 72
Figure 3.2 Urban Area Expansion for the Central City of Shanghai ............ 73

Figure 3.3 Urban and Rural Population Growth in Shanghai Municipality... 75

Figure 3.4 The Ring Structure of Urbanization in Shanghai Municipality. . .. 76
Figure 3.5 Distribution GNP and Population for 1985, 1992 and 1997 ........ 77
Figure 3.6 Urban Sprawl in Shanghai, 19705 - 19905 ........................... 78
Figure 3.7 A Systematic View of LUCC in Shanghai, China ................... 79

Figure 3.8 False Color Composition of Different Land Types in Shanghai... 83
Figure 4.1 Study Areas: the Brazilian Amazon and the State of Rondonia. .. 119
Figure 4.2 A system of LUCC ........................................................ 120
Figure 4.3 Land Cover Data in the Brazilian Amazon Based on Landsat

Satellite Image Classification for Year 1986 and 1992 .............. 124

Figure 4.4 Diagram of Land Cover Change Model with Four LUCC

Processes ................................................................... 126
Figure 4.5 Diagram for the Land Cover Change Model .......................... 128
Figure 4.6 Sources of Deforestation in the Brazilian Amazon .................. 129
Figure 4.7 Dynamic Agricultural Economic Growth Processes ................. 131
Figure 4.8 Diagram of Agricultural Development and Land Use Change. . 133
Figure 4.9 Interaction Loops between Deforestation Source Activities. 141
Figure 4.10 Modeling Other Proximate Sources of Deforestation ............... 144

xii

Figure 4.11
Figure 4.12
Figure 4.13
Figure 4.14
Figure 4.15

Figure 4.16

Figure 4.17.1
Figure 4.17.2
Figure 4.18.1
Figure 4.18.2
Figure 4.19.1
Figure 4.19.2

Figure 4.20.1

Figure 4.20.2

Figure 4.21.1

Figure 4.21.2

Population Pyramid in the Brazilian Amazon, 1990 .................
Growth of Urban Population Percentage (1940 — 1989) ............

Different Land Use Patterns for Different Occupational Groups.

Population Growth Model with Four Ecological Processes. . . . .

Population Distribution and Occupational Composition ............
Population Growth in the Brazilian Amazon and the State of

Rondonia (1986 — 2020) ..................................................
Urban/Rural Population in the Brazilian Amazon (1986 — 2020)..
Urban/Rural Population in the State of Rondonia (1986 - 2020)..
Agricultural Land for the Brazilian Amazon (1986 —- 2020) ........
Agricultural Land for the State of Rondonia (1986 - 2020) ........
LUCC for the Brazilian Amazon (1986 — 2020) .....................
LUCC for the State of Rondonia (1986 — 2020) .....................

Annual Deforestation Rate and Secondary Growth Rate in the
Brazilian Amazon (1986 — 2020) .......................................

Annual Deforestation Rate and Secondary Growth Rate in the
State of Rondonia (1986 — 2020) ......................................

Validation for Land Types in the Amazonia (1992, 1999) .........

Validation for Land Types in Rondonia (1992, 1999) ...............

xiii

148

149

151

153

156

162

163

163

166

166

169

169

171

171

173

173

Table 2.1
Table 2.2
Table 2.3
Table 2.4
Table 2.5
Table 3.1
Table 3.2
Table 3.3
Table 3.4
Table 3.5

Table 3.6

Table 3.7

Table 3.8.1
Table 3.8.2

Table 3.8.3

Table 4.1
Table 4.2
Table 4.3
Table 4.4
Table 4.5
Table 4.6
Table 4.7

Table 4.8

LIST OF TABLES
Comparison of Cultivated Land Data from SLA and MOA (1990).
Four Classes of Cultivated Land Per Capita .............................
Economic Assessment of Agricultural Structure Change .............
Urban Population Growth and Area Expansion ........................
Rural Economy Structure in the East Region ...........................
Land Use/Land Cover Types for Shanghai, 1978 — 2000 .............
Spatial Variation of Land Use Distribution ..............................
Accuracy Assessment for Land Use/Land Cover Classification. .
List of Variables Used in Principle Component Regression ..........
Component Matrix Extracted by Principal Component Analysis. . ..

Multivariate Linear Regression Results .................................
Parameter Estimates of LUCC Model Using PCR Regression .......

Model Results Validation and Projection for Build-Up Areas .......
Model Results Validation and Projection for Agricultural Land. . . ..
Model Results Validation and Projection for Forest Area ............
Equations for the Land Cover Change Loop ............................
Equations for Source Activities of Deforestation ......................
Equations of Agricultural Economic Processes ........................
Equations of Other Proximate Sources ..................................

Equations Of the Population Growth Model ............................

Equations of Population Distribution Model ............................
Validation Table for Population of Different Age/Gender Groups.

Validation Table for Agricultural Land Areas ..........................

xiv

34

36

42

45

85

86

87

89

93

94

98

101

103

104

128

129

134

145

154

157

164

167

Introduction

Land use/land cover change (LUCC) is one of the three most important global
environmental changes along with increasing carbon dioxide concentrations in the
atmosphere and alterations in global nitrogen cycle (V itousek, 1994). Human beings have
transformed significant portions of the Earth's land surface, with ten to fifteen percent of
land being dominated by agricultural or urban-industrial areas, and six to eight percent

being pasture (V itousek et a1. 1997).

The scientific community’s call for studies on land use/land cover changes could date
back to the Stockholm Conference on Human Environment in 1972, and later at the
United Nations Conference on Environment and Development (UN CED) in 1992. After
that Turner, Moss, and Skole (1993) surveyed many international programs and
organizations that expressed the need for addressing land use/land cover changes as a
research priority. Since then the subject of land use/land cover change has been of central
concern to the environmental research community and drawn considerable attentions
from various disciplines. It is now a science that brings together the human social
processes of land use change and the physical processes of environmental change under a
single conceptual umbrella in an effort to understand the fundamental processes of global

environmental change (Turner H et a1, 1994).

Land use/land cover changes are results from multidimensional interactions of

biophysical and socioeconomic processes (F rohn et a1, 1996; Fischer et a1, 1996). Land

cover is the physical and biotic character of the land surface while land use denotes the
purposes for which human beings employ or exploit land cover (Turner H et a1, 1994;
Lambin, 1994). Most of the early studies focused on the biophysical processes that drive
land cover change without accounting explicitly for land use changes driven by human
demands and economic activities (Fischer et al, 1996). However, past decades have
witnessed a growing recognition of human interventions as a major force shaping the
biosphere - as much as 40% of the global net primary production has been appropriated
by human activities (V itousek et a], 1986). Geographers are among the first scientists to
realize that it is the human actions rather than natural forces that are the sources Of most
contemporary change in the status and flows in the biosphere, and to sound the alarm that
human-induced land use changes are beginning to threaten the balance of life (Turner H
et a1, 1994). This is how the human dimension became a major theme in the agenda of

global change research (Gallopin, 1991;IGBP&HDP, 1993, 1995; Turner H et a1, 1995).

This dissertation work aims to improve the understanding of the dynamics of land
use/land cover change process, investigate factors that control land use/land cover change
processes at various scales, and explore the interlinkages of different factors and different
land types. The major theme is to monitor and model the spatial-temporal dynamics of
land use/land cover change as well as its relationship with the human socio-economic

driving forces using GIS, remote sensing and different modeling techniques.

This report is structured in five chapters as follows:

Chapter one provides an introduction to the background of LUCC research, discusses the
importance of LUCC as a link with global environmental change and food security, and

outlines the most important research questions to be addressed in LUCC studies.

Chapter two presents a study of agricultural land use change in the East Region of China,
where conflicts between urban and agricultural land use are the dominant subject of
controversy. This chapter is designed to analyze what drives the rapid loss of cultivated
land in highly urbanized areas. Based on this, a multinomial logit model is constructed to

predict future land use change in the East Region of China.

Chapter three zooms into Shanghai, the largest city in the East Region of China. It closely
examines the phenomena of urban sprawl and agricultural land loss in metropolitan
Shanghai. Remote sensing technique is applied to monitor the magnitude, rate and pattern
of land use/land cover change. A principal component regression model is constructed to
link remotely sensed observations of land use/land cover change to the understanding of

processes that resulted in the LUCC changes.

Chapter four is a study on deforestation and secondary grth in the Brazilian Amazon
Basin. It studies the consequences of agricultural development and various other tropical
activities on a primarily forested landscape. A systematic model is constructed by linking
remotely sensed land classification and census data to simulate future land use/land cover

changes in the Brazilian Amazon. The model is applied to the State of Rondonia as well.

Chapter five summarizes the empirical and theoretical findings fi'om the land use/land
cover change studies in China and Brazil. It also addresses the implication of the studies

and suggests avenues for future research.

Reference:

Fischer, G.; Ermoliev, Y.; Keyzer, MA. and Rosenzweig, C. 1996. Simulating the socio-
economic and biogeophysical driving forces of land-use and land-cover change: The
HASA land-use change model. Working Paper WP-96-010, Laxenburg: International
Institute for Applied Systems Analysis.

Frohn, R.C.; McGwire, K.C.; Dale, V.H. and Estes, J.E. 1996. Using satellite remote
sensing analysis to evaluate a socio-economic and ecological model of deforestation in
Rondonia, Brazil. International Journal of Remote Sensing 17: p3233-3255.

Gallopin, G.C. 1991. Human dimensions Of global change: Linking the global and local
processes. International Social Sciences Journal 130: p707-718.

IGBP&HDP. Land use and land cover change, science/research plan. IGBP Report No.
35 & HDP Report NO. 7, 1995, p52-60.

IGBP&HDP. Relating land use and global land and cover change. IGBP Report NO. 24 &
HDP Report No. 5, 1993, p21-25.

Lambin, ER 1994. Modelling deforestation processes: a review. TREES Publications
Series B: Research Report 1. European Commission Joint Research Centre, Ispra.

Turner H, B.L.; Meyer, W.B. and Skole, D.L. 1994. Global Land-Use/Land-Cover
Change: Towards an Integrated Program of Study. Ambio 23 (1): p91-95.

Turner H, B.L.; Moss, RH. and Skole, D.L. 1993. Relating Land Use and Global Land-
Cover Change: A Proposal for an IGBP-HDP Core Project. IGBP Report 24 and HDP
Report 5. International Geosphere-Biosphere Programme and the Human Dimensions of
Global Environmental Change Programme, Stockholm.

Turner H, B.L.; Skole, D.; Sanderson, S.; Fischer, G.; Fresco, L. and Leemans, R. 1995.
Land—Use and Land-Cover Change Science/Research Plan. IGBP Report No. 35 and
HDP Report No. 7. International Geosphere-Biosphere Programme and the Human
Dimensions of Global Environmental Change Programme, Stockholm and Geneva.

Vitousek, P. M. (1997). Human domination of Earth's ecosystems (vol 277, pg 494,
1997). Science 278(5335): p21-21.

Vitousek, RM. 1994. Beyond Global Warming: Ecology and Global Change. Ecology.
75(7):pl861-1876.

Vitousek, P.M., Erlich, P.R., Erlich, AH. and Matson, RA. 1986. Human appropriation
of the products of photosyntheses. BioScience 36:368-373.

CHAPTER 1

LUCC Monitoring and Modeling

1.1 Land Use/Land Cover Change

Broadly speaking, there are three main types of land: urban, agricultural and forest. The
term land use/land cover change can refer to any type of change that occurs within each
land type (such as agricultural intensification and forest thinning) or between two
different land types (for example, deforestation and urban sprawl). This project will focus
on the second category of change, that is, changes between different land types. Given
the irreversibility of urban land use, there are four possible types of changes that form an
LUCC triangle: (1) change from forest to agricultural land; (2) change from agricultural
land to urban use; (3) change from forest to urban land; and (4) change from agricultural

land to forest (Figure 1.1).

 

Urban

 

 

 

 

 

 

 

 

 

Agro 4 J Forest

 

 

 

Figure 1.1: LUCC Triangle with Four Major Types of Changes: 1- Agricultural land
encroaching on forest; 2 - Urban sprawling over agricultural land; 3 - Urban taking over
forest directly; and 4 - Agricultural land turning into forest.

 

Most studies focused on one type of land use/land cover change. These fragmented
studies do not integrate different land use/land cover change types, therefore are not able
to account for the dynamic interlinkages between different land types as shown on the
LUCC triangle. As a science whose research subjects across both social and natural
sciences, geography has a unique advantage and rich heritage of thinking systematically.
By its very nature, geography is a discipline that seeks to integrate and synthesize
knowledge, and to bring together and integrate the area association and spatial
organization of both natural and anthropogenic phenomena on the earth’s surface. This
has put geography in a position of advantage to study land use/land cover change in an

integrative way.

1.2 Study Areas and Research Tasks

This dissertation work aims to understand the major types of land use/land cover change
and their interlinkages across space and over time. The East Region of China and the
Brazilian Amazon are selected to be the main study areas. The Shanghai Municipality
located in the East Region of China and the State of Rondonia in the Brazilian Amazon
are further studied to account for spatial variations of land use/land cover change (Figure

1.2).

 

 

 

     

 

East Region

   

 

Shanghai

_. .. _._... ,......_I

    

./f
,4/ ~ ’  

Brazil Legal Amazon i Rondonia

Figure 1.2: Maps of Study Areas: (1) the East Region of China and the Shanghai
Municipality; (2) the Legal Amazon of Brazil and the State of Rondonia

Though both belonging to developing countries, the East Region of China and the
Amazonia in Brazil have very different levels of development and different types of
landscape as well. The Amazonia is an agricultural frontier in Brazil, and it is chosen to
be a representative case to study the consequences Of agricultural activities on a primarily
forested landscape. In contrast, the East Region of China is a rather developed area with a
long history of agricultural development. It is selected to be a case for studying the

change of agricultural landscape in highly urbanized areas.

There are three general research tasks in this dissertation work: to monitor land use/land
cover change using remote sensing and GIS, to understand the process of land use/land
cover change and to model land use/land cover change based on remotely sensed

observations and theoretical understandings of the processes.

1.2.1 LUCC Monitoring

Remote sensing and geographic information system (GIS) techniques are the powerful
tools to fulfill the task of land use/land cover change monitoring (Downton, 1995; Coppin
and Bauer, 1996). Since remote sensing data can be collected at multiple scales and at
multiple times, they provides a great opportunity to characterize and map land use/land
cover distribution and change patterns with great details that is not possible by using
census data or field surveys alone (Townshend et a1, 1991). Together with data available
from ground-based observations or surveys, remotely sensed data can be used to monitor
changes in space and time, to develop and validate dynamic models of regional
development and to forecast future land use patterns and changes (Moran et a1, 1994;
Frohn etal, 1996; Wood and Skole, 1998). This unique advantage of remote sensing has
stimulated a growing interest in making scientific progress through the use of remotely
sensed data in social science research community, so-called “Pixelize the Social”

(Pritchard et a1, 1996; Schweik and Green, 1999).

What makes remote sensing more appealing to land use/land cover change studies is that

remote sensing provides an opportunity to measure landscape attributes and record data

in accurate digital maps and files that can be stored and managed in Geographic
Information System (GIS). GIS has the capability to readily combine, integrate and
analyze remote sensing data within a spatial framework (Liu et a1, 1993; Martin, 1996). It
is the combination of remote sensing and GIS that makes data input and computational
limits for dealing with temporal sequences of land use/land cover information a less

significant issue (Lambin, 1994).

1.2.2 From Pattern to Process

However, monitoring the magnitudes and patterns offers only a partial picture of land
use/land cover change; it cannot provide much insight into the processes. We need to
move from pattern description to process understanding, to find out what processes are

involved that change land use/land cover, and what are the underlying driving forces.

Understanding the relationships between pattern and process at landscape scales was the
focus of the earlier works in ecology (Baker, 1989; Gardner, 1991). But most recent land
use studies tend to focus on spatial aspects of patterns rather than the processes that
resulted in the patterns (Grainger, 1995; Sklar and Costanza, 1991). Recent development
in landscape ecology has reemphasized the important relationships between spatial
patterns and processes (Turner 1990; Naveh and Lieberamn, 1998). The pattern to
process approach allows us to consider how spatial patterns of land use/land cover
change are resulted from social-economic and ecological processes, and how the

processes are influenced by the changing patterns.

10

1.2.3 LUCC Modeling

With the understanding of pattern and process, LUCC models can be built to link pattern
to process by quantifying the contribution from each driving force and representing the
sensitivity of land use/land cover change to economic, ecological or locational factors.
LUCC modeling has become a popular practice since policy makers and the public
became increasingly concerned about quantifying the environmental impacts of human

sociO-economic activities on land use/land cover change (Lambin, 1994).

Well-defined quantitative models can play an essential role in our efforts to untangle the
complexities of what drive land use/land cover change and to generate projections for
future land use/land cover changes. As a caricature and deliberate simplification of
reality, models simplify the multidimensional land use/land cover change processes by
highlighting the important variables as well as the causal/impacts relationships involved
in the processes. Good modeling skill may help us move closer to the reality and
represent the qualitative factors in a better manner; and good model projections may

contribute to the design Of policy responses to land use/land cover changes.

1.3 Spatial and Temporal Dimension of LUCC

Land use/land cover changes have two dimensions: space and time. Geographers have
special interests and unique advantages in dealing with the spatial component. The basic
spatial question is “what is where”: what is the magnitude of different types of land?
What is the rate of different types of land use/land cover change? And where are the

changes? The spatial pattern is also called into close attention because of its essential

ll

importance in understanding land use/land cover change process as well as its
environmental impacts on key ecological processes (Dunn et a1, 1991). Since landscapes
are in constant changes, it is important to go beyond the spatial dimension to account for
the temporal dynamics of land use/land cover change: how is land use/land cover
changing over time? And in the long term, do we know enough to generalize the trend of

land use/land cover change?

1.3.1 Spatial Dimension

Bid-rent function offers a powerful analytical approach for exploring the spatial
component of land use/land use change. As early as in 1826, Von Thunen started to
account for the spatial distribution of economic activities upon agricultural landscape
using a bid-rent function, which shows that the most productive activities will compete
for locations with the highest accessibility, and activities not productive enough will be
located further. In Von Thunen’s agricultural land use model, urban area is a
dimensionless central point. Later geographers such as Alonso (1964) and Fujita (1989)
succeeded in generalizing and reviving Von Thunen’s bid-rent function and applying it in
an urban context to account for the structure and dynamics of industrial and residential
land use changes. Muth (1961) was the first one to use bid-rent functions to account for
land use change in regional context where urban residential area growth pushes the city
limit into the agricultural hinterland. However, the agricultural land area in Muth’s land
use model is not bounded; therefore, there is no land conversion between urban (or
agricultural) land and natural land. Walker et a1. (1997), Walker and Homma (1996) and

Walker and Solecki (1999) described possible adaptations of the bid-rent models to land

12

cover change. Walker (2001) introduced natural land into the bid—rent function of land
cover change, a dynamic process in which forestry, agriculture and urban industry

compete for space.

Assuming a homogenous region like that of Von Thunen’s isolated state, we can imagine
a concentric zone model where the region is comprised of a set of concentric zones with
each zone devoted to a particular type of land use (Figure 1.3). An urban land zone is
located in the center as the core of the region, surrounded by an agricultural land zone in
a less intensively used manner, with forest area (natural wildness) as the outmost zone.
The bid-rent for urban land use is higher than that of agriculture near the urban core area,
but falls below agricultural rent at a sufficient distance from the city. The bid-rent for
agriculture exceeds that of forestry within certain distance to urban market, but becomes

profitless as it moves further away from the urban center.

 

 

 

 

 

 

 

 

 

Rent A
1 Urban
\\'\ AgrO
N Forest
; ii: Accessibility
: ‘Ezgrest -. L
a d2
d3 '

 

 

 

Figure 1.3: Bid—Rent Theory of Land Use/ Land Cover Change

The changing landscape patterns can be described in the following processes. As the city
grows over time, urban land area will exert pressure on the agricultural zone immediately
surrounding it. Outward expansion of the urban area will invade agricultural area and
cause agricultural area to expand outward into forestry areas. Therefore the agricultural
land is the advancing edge of urban sprawl and forestry land is at the threat of agricultural
expansion. The process will continue with each successive zone moving further fi'om the
center, but not in an unlimited way because of the environmental pressure as the result of

diminishing forest area.

The concentric zone model is a simple representation of a region’s land use patterns, i.e.,
an idealized model of land use and social spatial structure. There are many exceptions to
and variations on this theme, and there is no particular region that actually and precisely
fits this pattern. The introduction of realities such as local physical situation variables,
transport routes and multiple city centers will cause distortions of the concentric pattern
(Isard, 1956). The most plausible assumption of the concentric model is homogeneity of
the region. Ideally, land surrounding the urban core area is assumed to be entirely flat
with uniform fertility. But the fact is that interactions between physical, biological and
anthropogenic elements lead to a heterogeneous mosaics of landscape (Froman, 1995).
We have to take the site variables such as physical situation and transportation network
into account in location choices. Transportation accessibility is the decisive factor
affecting land use change. Hoyt (1939) developed the sector theory to account for the
impacts of major transportation routes on land use change. He theorized that cities tend to

grow in wedge-shaped patterns, emanating from the central business district (CBD) and

14

centered on major transportation routes. Therefore land use tends to conform to a pattern
of sectors rather than concentric rings. However, Harris and Ullman (1945) argued that
economic activities are built around several centers of nuclei rather than one single
center. Cities of greater size develop substantial suburban areas, some of which reach
significant size, develop into smaller business districts and act as satellite nodes, or

nuclei, of activity around which land use patterns formed.

1.3.2 Temporal Dimension

In terms of the temporal component of land use/land cover change processes, diffusion
theory provides a useful tool to capture the spread effects of land use/land cover change
process which can be viewed as moving across landscape in diffusion waves as the

passage of Hagerstrand’s (1953) innovation diffusion profile.

There are three general types of diffusion process: expansion diffusion, hierarchical
diffusion and relocation diffirsion (figure 1.4). Expansion diffusion is a contagious
spreading process operated by a distance decay fiinction. Urban sprawl and deforestation
are two good example of expansion diffusion. The further away [Tom the source of
expanding, the less likely a parcel of land will be changed. Hierarchical diffusion works
in a “jumping” manner where simple linear distance is not always the strongest influence
on a diffusion pattern. For example, logging activities will leap over intervening places
following a path that concentrates initially only on regions with high quality wood, and
from there spreading to regions in the lower hierarchy. Shifiing cultivation is an example

of relocation diffusion. Forestry land is cleared and crops are grown and harvested for a

15

couple of years till the fertility of the land is exhausted. Then the land will be abandoned
and farmers will move to a new location for new clearance. However, many diffusion
processes in reality represent a combination of expansion, hierarchical and relocation

diffusion, especially the combination of expansion with either hierarchical or relocation

diffusion (Robinson, 1998).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0
0R5 0 O O O
o O o . o o ’
t ° , 9 . .
O . . O
O
O O O . . .
O . .
QM -
O -
t + 1 N —'
O
O O O —
O ’
Expansion Hierarchical Relocation
Diffusion Diffusion Diffusion

Figure 1.4: Three Major Types of Diffusion Processes: Expansion Diffusion,
Hierarchical Diffusion and Relocation Diffirsion

Land use/land cover change is a built-up process. The contemporary urbanized
landscapes started a long time ago when agricultural and cattle ranching were the main
economy activities; agricultural expansion and deforestation were the major types

changes of land. It is through the built-up process that very advanced and complex levels

16

of landscape have been obtained with dominant presence of buildings and infrastructures
(F olch, R., 1997). Land use/land cover change processes start with a primary slow
increase stage, followed by a rapid increase stage, then a slow condensing stage again
before finally reaching a saturation stage. As time goes along, we can expect the

following general transitional trends for different land types (Figure 1.5).

 

 

 

 

 

 

Percentage

 

 

 

 

 

 

 

Figure 1.5: Land Use/Land Cover Change Diffusion Processes

Urban land grows slowly at the primary stage with a small urban population, but starts to
accelerate (the urban sprawl stage) when the region steps into an industrialization age,
and finally reaches a saturating stage marked by a slowing and eventual cession of the
urban growth process. Agricultural land has a rapid initial expansion due to the increasing

demand for agricultural products, reaches a peak size of agricultural area and then starts

17

to shrink as more and more agricultural land being encroached by urban land and other
uses. Forest area starts with an initial high stock, followed by shrinkage in size as it being
converted to agricultural or other uses. But forest cover will eventually experience a
slight increase due to environmental awareness and forest reconstruction projects. In the
land use transitional stages, only the shapes of the trends are universal, the slopes

(parameters) are dependent on specific region’s economy type and development stage.

The literature of forest transition theory was derived from the concept of the
environmental Kuznets curve, which says that environmental impacts of natural resource
use initially grow but later decline as income increases (Arrow et al., 1995; Mather et al.,
2000; Stem et al., 1996). Similarly, forest transition theory indicates that forest cover
would exhibit a U-shaped transition pattern; an initial decline in forest cover due to
deforestation is later reduced, offset, and eventually outweighed by secondary forest

expansion (Drake, 1993; Mather, 1992; Walker, 1993).

Regional level land use data from The Food and Agriculture Organization (FAO)
supported the general trends of these land use/cover changes. The Brazilian Amazon is
now experiencing a slight increase of urban land, a rapid expansion of agricultural land,
and a fast declining of forest area. This is very different from what is happening in China,
where agriculture system has been developed for thousands of years and industry has
been developing very fast more recently. Urban land in China is undergoing a rapid
expansion period at the cost of both agricultural land and forest area, and agricultural
land is taking over forest area due to the increasing tension between the supply and

demand of agricultural land. The land use transitional stage for highly developed regions

18

such as Shanghai bears much similarity to that of the USA, where urban area is still
growing, but with a tendency to stabilize. Forested area is also experiencing a slow
growth as its importance of being a vital environmental resources being increasingly

recognized, and the government is making massive efforts on aforestation projects.

1.4 LUCC Significance

1.4.1 Urban Sprawl and Agricultural Land Loss

From an economic perspective, change of agricultural land is a major concern due to its
explicit ties with agricultural production and food security (Simpson, 1993). Urban
sprawl has been blamed for eliminating productive agricultural land and imperiling food

production at an alarming pace (Berry, 1990).

Loss of prime farmland has been repeatedly reported worldwide (Nelson, 1992; Levia,
1998; Yeh and Li, 1997, 1999; Li, 1998). An estimated 5 million to 7 million hectares of
farmland are disappearing from the world each year. The United States lost nearly
400,000 hectares of farmland each year to urban sprawl (Feder, 1997), which also took
over about one million hectares of agricultural land in China each year between 1987 and
1992 (Tyler, 1994). Arable land per capita in many developing countries has declined
sharply. The average arable land of developing countries as a whole dropped from one
half of a hectare per person in 1961 to less than one fifth of a hectare in 1992. If current
trends in population growth and land use continue, the amount of arable land will be a
little over one-tenth of a hectare per person in 2050 (Robey, 1997). This situation has

brought the international attentions to the concern of food security, “a state of affairs

l9

where all people at all times have access to safe and nutritious food to maintain a healthy
and active life” (FAO, 1996). By this definition, about one-third of world population
lacks food security. F AO estimates that world food production would have to double to

provide food security for the 8 billion people projected for 2025 (Robey, 1997).

The amount of land that could be cultivated to improve food production was estimated to
be 2 billion hectares, or about 40% more than what is currently being cultivated (Robey,
1997). However, most of the uncultivated land has marginal productivity because of poor
soil quality or inappropriate weather condition. Without substantial new financial
investments or unforeseeable improvements in technology, increases in food production
will have to come fiom the existing agricultural land. Therefore, it is of great importance
to protect the existing agricultural land from being encroached by urbanization. This is
notably a challenge for heavily populated countries such as China, one of the nations with
the lowest per capita land resources in the world (CCICED, 1996; Yang and Li, 2000;
Bai, 2000). As populous agricultural areas become even more crowded, arable land is
likely to come under increasing pressure of being developed. While the overall
agricultural land area in China is slightly increasing, agricultural production ability is
decreasing due to the fact that most agricultural land being lost is likely to have been
fertile prime farmland located on the periphery of major cities and towns, yet most
agricultural land newly brought into cultivation constitutes inferior land whose
productivity is for the time being far less than what has been lost (Ash and Edmonds,

1998; Smil, 1999; NACOC, 1999). Thereby, the offset in quantitative changes of

20

agricultural land does not necessarily justify a similar conclusion regarding the

production capability of agricultural land and food security.

Besides its economic significance, urban grth related land use change also has
significant impacts on the ecological sustainability. According to the World Wildlife
Fund in China, at least 20 percent of all plant and animal species in China are threatened,
well over the world's average of 10 to 15 percent. At least 156 of the 640 species listed on
the International Convention on the Trade of Endangered Species of Flora and Fauna are
found in China. Nowhere is the threat more evident than in Shanghai, where urban
expansion projects have forced hundreds of species of wild birds and reptiles from their
natural environments (United Press International, 2000). Chapter three and chapter four
will proceed to explore urban sprawl and agricultural land use change in the East Region

as well as the Municipality of Shanghai in China.

1.4.2 Agricultural Expansion and Deforestation

From an ecologic perspective, however, loss of forest land is of greater importance due to
its significant role in global carbon cycle, ecological complexity and biodiversity
preserving (Skole and Tucker, 1993; Adger and Brown, 1994). Most deforestation is for

agricultural development purposes.

Deforestation has a profound effect on the global carbon cycle by increasing the amount

of C02 in the atmosphere, which has increased by more than 25% during the last two

centuries and has largely contributed to climate change and global warming. From 1850

21

to 1990, deforestation worldwide released 122 billion metric tons of carbon into the
atmosphere, while the current rate is approximately 1.6 billion metric tons per year. In
comparison, fossil fuel burning releases about 6 billion metric tons of carbon per year, so
deforestation makes a significant contribution to the increase of atmospheric C02,
Houghton, Skole, and Leflcowitz (1991) emphasized the importance of land use/land
cover data in calculating the global carbon balance. According to Houghton and Skole
(1990), approximately 30% of the current net flux of carbon dioxide is biogenic and
related to anthropogenic land cover changes. Long-term historical release of carbon
dioxide from human alterations of land use/land cover has been approximately equal to

the release of carbon from fossil fuel burning.

Most biotic carbon before the 20th century came from deforestation in northern
hemisphere temperate zone; but since mid-20th century, the biotic carbon emission center
has shifted to southern tropical areas (Ojima et al., 1994). Repetto (1990) reported an
accelerating pace of tropical forest resources depletion. The Food and Agriculture
Organization (F AO) estimates that 53,000 square miles of tropical forests were destroyed
each year timing the 19803, with an estimated 21,000 square miles of annual
deforestation in South America, mostly in the Amazon Basin. If the current rate of
deforestation continues, the world's rainforests will vanish within 100 years. Besides its
significant role in global environmental change, tropical deforestation also imposed great
threat to biodiversity preservation and ecological sustainability (Meyer and Turner 11,
1994). Covering only 7% of the total dry surface of the Earth, tropical forests have the

highest ecological complexity, holding over half of all the species in the world. Many of

22

the rainforest species require a special habitat to live in, therefore are vulnerable to
deforestation. Alarming discoveries of the relationship between tropical deforestation and
global warming and biodiversity preservation has stimulated a rising interest in land
use/land cover change researches in tropical areas, especially in the Brazilian Amazon,
the world’s largest piece of contiguous tropical forest reserve that contains about 40% of
the world's remaining tropical rainforest (Houghton et a1, 1987; Houghton et a1, 1991;
Henderson-Sellers, 1994; Adger and Brown, 1994). . Farmers slash an area of forest, burn
the tree trunks and grow agricultural product on the cleared land, so-called slash and burn
agriculture. Chapter four will study land use/land cover change in the Brazilian Amazon
and the State of Rondonia as the results of agricultural and other tropical development

activities.

1.5 Reference:

Adger, W.N. and Brown, K. 1994. Land use and the causes of global warming.
Chichester: John Wiley & Sons.

Agarwal, C., Green, G.L., Grove, M., Evans, T. and Schweik, C., 2000. A Review and
Assessment of Land-Use Change Models: dynamics of space, time and human choice.
Center of the Study of Institutions, Indiana University.

Arrow, K., Bolin, B., Costanza, R., Dasgupta, P., Folke, C., Holling, C.S., Jansson, B.-O.,

Levin, S., Miller, K-G., Penings, C., and Pimentel, D. (1995). “Economic growth,
carrying capacity and the environment.” Science 268, p520-521.

Ash, R. F. and Edmonds, R.L. 1998. “China’s Land Resources, Environment and
Agricultural Production.” The China Quarterly 156(December):p836-879.

Bai, W. 2000. Dynamic trend analysis for land use change in Shenzhen, China. Journal of
Natural Resource (Chinese), 15: (2). April 2000, pl 12-1 16.

23

Baker, W.L. 1989. A review of models of landscape change. Landscape Ecology 2, p111-
133.

Berry, B.J.L. 1990. Urbanisation. In W. B. Meyer and B. L. Turner H, eds. Changes in
Land Use and Land Cover: A Global Perspective. Cambridge University Press,
Cambridge, p103-120.

CCICED. 1996. The Seventh Report for the Fifth Conference of the China Council for
lntemational Cooperation on Environment and Development, “The Role of Sustainable
Agriculture in China: Environmentally Sound Development.” Shanghai, 23-25
September.

Coppin RR. and Bauer, M. E. 1996. Digital change detection in forest ecosystems with
remote sensing imagery. Remote Sensing Rev. 13, p207-234.

Downton, M.W. 1995. Measuring tropical deforestation: Development of the methods.
Environmental Conservation 22: p229-240.

Drake, W.D. (1993). Towards building a theory of population-environment dynamics: A
family of transitions, in Population-Environment Dynamics (G.D. Ness, W.D. Drake, and
SR. Brechin, Eds), p305-355, University of Michigan Press, Ann Arbor.

Dunn, C.P.; Sharpe, D. M., Guntenspergen, G.R.; Steams, F. and Yang, Z. 1991.
Methods of analysing temporal changes in landscape pattern. In: Turner, MG. and
Gardner, R. H. (eds). Quantitative methods in landscape ecology. The analysis and
interpretation of landscape heterogeneity. Springer-Verlag, NewYork.

FAO, 1996. Food for All. Rome, FAO 1996, p64.

Feder, B. Sowing preservation—Towns are slowing invasion of farms by bulldozers.
New York Times, March 20, 1997. p. D1, 19.

F olch, R. 1997. The Socioecological Landscape: Pattern and Process in the Western
Mediterranean. LUCC Newsletter 1, p 4-9, Feb 1997.

Frohn, R.C.; McGwire, K.C.; Dale, V.H. and Estes, J .E. 1996. Using satellite remote
sensing analysis to evaluate a socio-economic and ecological model of deforestation in
Rondonia, Brazil. International Journal of Remote Sensing l7: p3233-3255.

Froman, R. T. 1995. Land Mosaic. Cambridge: Cambridge University Press.

Fujita, M., 1989. Urban Economics. Cambridge University Press, Cambridge.

Gallopin, G.C. 1991. Human dimensions of global change: Linking the global and local
processes. International Social Sciences Journal 130: p707-718.

24

Gardner, R. H. (eds), 1991. Quantitative methods in landscape ecology. The analysis and
interpretation of landscape heterogeneity. Springer-Verlag, NewYork.

Grainger A., 1995. National land use morphology: patterns and possibilities. Geography
20: p235-45.

Hagerstrand, T., 1953. Innovations Diffusion as a Spatial Process. University of Chicago
Press, Chicago.

Harris C. and Ullman, EL. 1945. "The Nature of Cities". Annals, American Academy of
Political and Social Science 242.

Henderson-Sellers, A. 1994. Land-use change and climate. Land Degradation &
Rehabilitation 5: p107-126.

Houghton, RA. and Skole, D.L.l990. Carbon. In W. B. Meyer and B. L. Turner H, (eds).
Changes in Land Use and Land Cover: A Global Perspective. Cambridge University
Press, Cambridge, p393-408.

Houghton, R.A.; Boone R.D.; Fruci, J.R.; Hobbie, J .E. and Melillo, J.M. et a1. 1987. The
Flux of Carbon from Terrestrial Ecosystems to the Atmosphere in 1980 due to Changes
in Land Use: Geographic Distribution of the Global Flux. Tellus 3919: p122-39.

Houghton, R.A.; Leflcowitz, D.S. and Skole, D.L. 1991. Changes in the landscape of
Latin America between 1850 and 1985. Forest Ecology and Management, 38: pl43-172.

Hoyt, H. 1939. The Structure and Growth of Residential Neighborhoods in American
Cities. Washington, DC: Federal Housing Administration.

lsard, W., 1956. Location and Space Economy, Wiley, NY

Lambin, E.F. 1994. Modelling deforestation processes: a review. TREES Publications
Series B: Research Report 1. European Commission Joint Research Centre, Ispra.

Levia, DP. 1998. Farmland Conversion and Residential Development in North Central
Massachusetts. Land Degradation & Development, 9 (2): p123-130.

Li, X. 1998. Measurement of rapid agricultural land loss in the Pearl River Delta with the
integration of remote sensing and GIS. Environment and Planning B — Planning and
Design, 25(3): p447-46l, May 1998.

Liu, D.S.; Iverson, LR. and Brown, S. 1993. Rates and patterns of deforestation in the
Philippines: application of geographic information systems analysis. Forest Ecology and

Management 57: pl-l6.

Martin, D. 1996. Geographic Information Systems: Socio-economic Applications. 2nd
edition. New York. Routledge.

25

Mather, AS. (1992). “The forest transition,” Area 24(4), p367-3 79.

Mather, A.S., Needle, CL, and Fairbum, J. (2000). “Environmental Kuznets curves and
forest trends.” Geography 84(1), p55-65.

Meyer, W.B. and Turner H, B.L. (eds). Changes in Land Use and Land Cover: A Global
Perspective. Cambridge University Press, Cambridge.

Moran, E.; Brondizio, E.; Mausel, P. and Wu, Y. 1994. Integrating AmazOnian
vegetation, land use, and satellite data. BioScience 44(5): p329-33 8.

Muth, RF, 1961. Economic Change and Rural-Urban land Conversions. Econometrica,
29(1):1-23.

National Agricultural Census Office of China (NACOC), 1999. Abstract of the First
National Agricultural Census in China. Beijing: China Statistics Press.

Naveh, Z. and Lieberamn, AS, 1998. Landscape Ecology: Theory and Application.
Springer-Verlag, New York.

Nelson, AC. 1992. Preserving Prime Farmland in the Face of Urbanization: Lessons
from Oregon. Journal of the American Planning Association. 58(4): p 467-488.

Ojima, D.S.; Galvin, K.A. and Turner 11, BL. 1994. The Global Impact of Land-Use
Change. BioScience, 44(5):p300-304.

Pritchard, Jr. L.; Geoghegan, J.; Sanderson, S. and Turner 11. B. L. 1996. Socialising The
Pixel and Scalar Themes in Land-use/Cover Change (LUCC). Paper resented at the
Workshop on "People and Pixels: Using Remotely Sensed Data in Social Science
Research", Committee on the Human Dimensions of Global Environmental Change,
National Research Council (US), November 12th-13th, 1996.

Repetto, R. 1990. Deforestation in the tropics. Scientific American: 262 (4): p36—42.

Robey, B., 1997. Population and the Environment: the global challenge. John Hopkins
Population Information Program.

Robinson, GM. 1998. Methods & Techniques in Human Geography. Wiley.
Schweik, CM. and Green, G.M., 1999. The Use of Spectral Mixture Analysis to Study
Human Incentives, Actions and Environmental Outcomes. Social Science Computer

Review, 17(1 ): p40-63.

Simpson, J .R. 1993. Urbanization, Agro-Ecological Zones and Food Production
Sustainability. Outlook on Agriculture 22(4): p233-239.

26

Sklar, PH. and Costanza, R. 1991. The development of dynamic spatial models for
landscape ecology: a review and prognosis. In: Turner, MG. and R. H. Gardner, (eds).
Quantitative methods in landscape ecology. Ecological Studies Vol.82, New York:
Springer Verlag, p239-288.

Skole, D.L. and Tucker, C. 1993. Tropical Deforestatio and Habitat Fragmentation in the
Amazon: Satellite Data from 1978 to 1988. Science 260: p1905-1910.

Smil, Vaclav. 1999. “China’s Agricultural Land.” The China Quarterly 157, 414—429.

Stern, D.I., Common, MS, and Barbier, ES, 1996. “Economic growth and
environmental degradation: The environmental Kuznets curve and sustainable
development,” World Development 24(7), p1 151-1 160.

Townshend, J.R.G.; Justice, C.; Li, W.; Gurney, C.; and McManus, J. 1991. Global Land
Cover Classification by Remote Sensing: Present Capabilities and Future Possibilities.
Remote Sensing and the Environment 35: p243-355.

Turner H, B.L.; Meyer, W.B. and Skole, D.L. 1994. Global Land-Use/Land-Cover
Change: Towards an Integrated Program of Study. Ambio 23 (1): p91-95.

Turner, M. 1990. Spatial and temporal analysis of landscape patterns. Landscape Ecology
4: p21-30.

Turner, M. G. and Garden, R. H. 1991. Quantitative Methods in Landscape Ecology.
Springer-Verlag, New York.

Tyler, P. Nature and economic boom devouring China's farmland. New York Times,
March 27, 1994. p.1 and 8.

US Embassy in Beijing. China’s Farmland Loss Rings Alarm — Satellite Photographs
Reveal a Serious Problem. A report from US. Embassy Beijing June 1997.

Walker, RT. 2001. Urban Sprawl and Natural Areas Encroachment: linking land cover
change and economic development in the Florida Everglades. Ecological Economics 37
(21), p357-369.

Walker, RT. (1993). “Deforestation and economic development,” Canadian Journal of
Regional Science 16, 481-497.

Walker, RT. and Homma, A. 1996. Land Use and Land Cover Dynamics in the Brazilian
Amazon: An Overview. Ecological Economics, Vol. 18 (l), p67-80.

Walker, RT. and Solecki, W.D., 1999. Managing Land Use and Land Cover Change:
The New Jersey Pinelands Biosphere Reserve. Annals of the Association of American
Gergraphers, 82(2), p219-236.

27

Walker, R.T., Solecki,W.D. and Harwell, C. 1997. Land use dynamics and ecological
transition: the case of South Florida. Urban Ecosystem 1(1), p37-47.

Wood, CH. and Skole, D. 1998. Linking Satellite, Census and Survey Data to Study
Deforestation in the Brazilian Amazon. From: Pixel and Pixel: Linking Remote Sensing
and Social Science, Diana Livennan, Emilio Moran, Ronald Rinfuss, and Paul C. Stem
(eds); National Academy Press, Washing DC.

Yang, H. and Li, X., 2000. “Cultivated Land and Food Supply in China.” Land Use
Policy 17, 73-88.

Yeh, AG and Li, X., 1997. An Integrated Remote Sensing and GIS approach in the
Monitoring and Evaluation of Rapid Urban Growth for Sustainable Development in the
Pearl River Delta, China. International Planning Studies, 2(2): p193-210.

Yeh, AG. and Li, X. 1998. Sustainable land development model for rapid growth areas

using GIS. International Journal of Geographical Information Science, 12(2): p169-189,
Mar 1998.

28

CHAPTER 2
Cultivated Land Use Change Analysis and Modeling

in the East Region of China

2.1 Introduction

It will be a challenge for China to feed more than 20% of the world’s total population
with less than 10% of the world’s total arable land. As one of the most important grain
producers in the world, China has only 13-14% of its landmass under cultivation, in
contrast with 55% in India and 20% in the USA (Information Office of State Council in
China, 1995). Moreover, cultivated land per capita in China is only half that of India,

one-ninth that of USA, and about half that of the world average.

With its large population and low ratio of cultivated land per person, there is growing
pressure on China’s agricultural system, particularly for grain production. While
population is growing at a rapid rate of about 15 million per year, cultivated land is
increasingly being converted to orchards or fishponds in response to changes in diet
preferences, and to built-up areas for industrial and residential uses. The State Land
Administration (SLA) of China reported that 1.73 million ha of cultivated land were lost
between 1988 and 1995. To put this magnitude in perspective, it is more than a third of
Japan’s total cultivated land area. Assume the average grain yield to be 4.5 tons/ha, and
the grain consumption for each rural person to be 0.25 ton/year (China Statistic

Yearbook, 1997), the loss of land area was equivalent to a loss of grain production

29

capacity of 7.8 million tons, or enough to feed a rural population of more than 30 million

for one year.

The problem of cultivated land loss and food security in China has drawn considerable
international attention. Growing disparities between land supply and demand has brought
the specter of a Malthusian crisis, and aroused the debatable question of whether China
can feed itself in the future. For example, a recent report from the Worldwatch Institute
has predicted that, with economic modernization, China would lose a significant
proportion of its cultivated land in the coming two decades, which would result in a
major food security crisis and force China onto the international food market to import
food (Brown, 1995). The Chinese government has also realized the importance of food
security. “Hunger breeds discontent” is a frequently used idiom within official and public
discourse (UNDP, 1996). The Government regards food security as one of its top
priorities for attaining social order and political stability. National policies have been
promulgated to protect arable land to ensure grain self-sufficiency since 1990s In 1995
the central government initiated a new grain policy giving provincial governors primary
responsibility for maintaining the "grain bag" by stabilizing sown area and allaying urban

residents’ concerns about grain supply.

This study addresses some of the key issues related to land use change and food security.
It uses a case study of the East Region of China in the 19805 and 19905 to explore the
driving forces of China’s cultivated land loss problem within a socio-economic and

environmental fiamework. The use of political ecology as a conceptual and analytical

30

framework enables a reevaluation of the importance of shifting regimes -- from
centralized planning to free market, from collective to individual action -— for
understanding the roles of social relations within which land users interact and land uses
are changed. Quantitative models can help simplify the complex and multidimensional
processes of land use and land cover change by highlighting the important variables as
well as the causal relationships involved in the processes. The study also applies a well-
defined model to untangle the complexities of physical and social-economic driving
forces of land use/land cover change, and to quantitatively forecast future land use

changes and loss of cultivated land in the East Region of China.

2.2 Study Area: the East Region of China

We adopt a regionalization scheme used by the International Institute for Applied System
Analysis (IIASA). It divides China into eight economic regions: North, Northeast, East,
Central, South, Southwest, Plateau and Northwest regions. This study examines the East
Region, which had the highest percentage of cultivated land loss, though the Northwest
Region had the largest area of land lost from cultivation. As seen fiom Figure 2.1, every
economic region has experienced losses Of cultivated land from 1988 to 1995, except for

a minor increase in the Plateau Region.

31

 

-380 P J East

 

 

 

 

 

-287[ ] Central
-11 [1 South
-105 [:3 Southwest
15 l] Plateau
-389 L ] Northwest

_ __. . _ . __ . "1

-500 -400 -300 -200 -100 0 100

Figure 2.1: Balance of Cultivated Land for Eight Economic Regions of China (1988-
1995) Unit: 1000 ha. Data Source: State Land Administration. Edited from Fischer et al

The East Region is located in the lower reaches of the Yangtze River (Figure 2.2),
including the Shanghai municipality and Jiangsu, Zhejiang and Anhui provinces. It is an
important economic region for both agriculture and industrial development. The total
land area of the East Region is less than 4% of the national territory, but it accounts for
12% of China’s cultivated land area and 18% of the total population. Also, it produces
18% of the agricultural output and generates 63% of the national rural industry output.
The topography of north part of East Region is low plains, while the south part is fairly
mountainous. The plain areas have rich agriculture resources with fertile land and
abundant water resources. Roughly 80% of the cultivated land area is irrigated, the
highest of any region in China (Streets et al., 1995). It permits a multiple cropping system
of two or three crops per year, producing an average yearly grain output of 15 -18.75

tons/ha.

32

 

   

DEM

East Sea

East Region

 
    

Iiangsu A
Shanghai

 

 

 

Figure 2.2: Study Area: the East Regions of China (Shanghai,Jiangsu, Zhejiang and
Anhui Provinces). The graph on the right side is the Digital Elevation Model (DEM).

Historically, Jiangsu and Zhejiang provinces were known as China’s “home for fish and
rice”. Now these two provinces are among China’s most developed areas. Anhui province
by contrast is less developed and remains mostly rural. The city Of Shanghai is the most
important industrial, commercial and financial center of China. As a metropolis Of over
13 million people, Shanghai has little natural resources of its own. The peripheral
provinces provide most of its labor and raw materials supplies. Demographic statistics
show that approximately 75% of the immigration into Shanghai in the 19805 was from
the peripheral provinces of East Region itself. Shanghai’s booming economy has brought
the East Region rapidly into an industrial and technological age in advance of other
regions. It is playing a key role in China’s “economic development with Chinese
characteristics”. Thus, what is happening here is a prototype for what may happen in

other regions.

2.3 Land Use Data
Data sets from the Ministry of Agriculture (MOA) and the State Land Administrative
(SLA) of China, were used in this study. Comparison indicates a close agreement between

the land use data from these two sources (Table 2.1).

Table 2.1 Comparison of Cultivated Land Data from SLA and MOA (1990)

 

 

 

 

 

 

State Land Ministry
Province Administrative of Agriculture
(SLA) (MOA)
Shanghai 324 343
Jiangsu 5,069 5,482
Zhejiang 2,357 2,559
Anhui 6,023 5,472
EAST REGION 13,771 13,857

 

 

 

 

 

Unit: 1,000 ha

Each of these data sets has its own advantages and disadvantages. Data from the MOA are
available at the county level, thus have a higher spatial resolution than data from the SLA,
which is provided at the province level. However, the MOA data set differentiates eight
types of land use (cultivation, horticulture, forest, grassland, settlement, transportation,
water and unused land) for only two dates, 1980 and 1990. By comparison, the SLA data
set covers eight years (1988-1995) with detailed land conversion information, which
enables an analysis of the dynamics of cultivated land change. The combination of these

two data sets provides a good picture of the land use changes in the East Region of China.

34

2.4 Distribution and Changes of Cultivated Land

The average per capita cultivated land in the East Region was 0.087 ha/capita in 1980. It
declined to 0.080 ha/capita in 1990, far below the national average of0.117 ha/capita.
According to a study by the United Nations Food and Agriculture Organization (FAO), a
minimum of 0.053 ha of cultivated land per person is required to ensure self-sufficiency in
food supply, even with the existence of advanced agriculture production techniques (State
Land Administrative Research Project for Cultivated Land Protection, 1998). Yet 48 of 237
counties in the East Region were below this minimum level in 1980, increasing to 74 by
1990 (Table 2.2). Except for Anhui province, all provinces in this region experienced an
expansion in the number of counties with per capita cultivated land lower than the F A0
minimum requirement. Zhejiang province was the most critical case, with more than 60%

of the counties below the minimum level in 1990.

35

Table 2.2 Four Classes of Cultivated Land Per Capita

 

Total <0.053 0.053 — 0.117 0.117 — 0.250 > 0.250
Province County ha/capita ha/capita ha/capita ha/capita

 

Name MM" 1980 1990 1980 1990 1980 1990 1980 1990

 

 

 

 

Shanghai 9 l 3 8 6 0 0 0 0
Jiangsu 75 10 15 54 46 10 13 0 0
Zhejiang 73 18 45 48 29 l 0 7 0
Anhui 80 19 l 1 34 40 27 29 0 0

 

East Region 237 48 74 144 121 38 42 7 0

 

(‘70) (100) (20) (31) (61) (51) (16) (18) (3) (0)

 

 

 

 

 

 

 

 

 

 

 

 

Data Source: Ministry of Agriculture (MOA). Note 1: Since the county boundaries vary over
time, data of 1980 are re-coded to match the county administrative codes for 1990. Note 2:
Two counties (Jinhua and Daishan) in Zhejiang Province are excluded from the analysis due
to data inconsistency.

Figure 2.3 shows the spatial distribution of cultivated land per capita by county. Dark red
represents counties with cultivated land less than the F AO minimum requirement for grain
self-sufficiency (< 0.053 ha/capita). Light red is for counties with an average cultivated
land of more than self-sufficiency level but lower than the national average level (0.053 —
0.117 ha/capita). Light green indicates counties with per capita cultivated land larger than
the national average level, but lower than the world average level (0.117 - 0.250
ha/capita). And dark green shows counties where per capita cultivated land is higher than

the world average (>0.250 ha/capita).

36

Year 1980 Year 1990

     

Cultiland ha/capita

     

Cultiland h capita

- <0.053 - <0.053
l:| 0.053-0.117 I: 0.053-0.117
[:1 0.117—0.250 El 0.117—0.250
- >0.250 - >o.250

Figure 2.3: Distribution of Cultivated Land Per Capita by County. Data Source: Ministry
of Agriculture (MOA)

In 1980, the cultivated land per capita for most counties was between 0.053 — 0.250 ha.
Counties with cultivated land per capita less than the FAO minimum requirement were
located mainly in the mountainous areas of Zhejiang and Anhui provinces, but also in the
urban areas of J iangsu and Zhejiang provinces. There were only a few counties in
Zhejiang province whose cultivated land per capita was higher than the world average.
From 1980 to 1990, what emerges most strikingly is that counties with cultivated land per
capita of less than 0.053 ha became concentrated on the east and northwest part of
Zhejiang province. Another striking observation is that by 1990 no counties have a per
capita cultivated land of above the world average, while several did in 1980. Counties

within the group 0.053 — 0.250 ha/capita still led the predominant majority.

37

Figure 2.4 shows the spatial distribution of the net change in cultivated land. Dark red
indicates counties with a net loss of more than 3% and dark green represents counties
with a net increase of more than 3% during 1980 - 1990. The remaining counties, shown
as the white ones, had experienced a relatively small change of less than 3% in either
direction. Most of the counties with a loss of more than 3% were located in the southwest
mountainous area, along the boarder between Zhejiang and Anhui provinces as well as in
the coastal part Of Jiangsu province. Land newly brought under cultivation was
distributed unevenly to the north of the Yangtze River, mostly clustered in Anhui
province. This spatial distribution is very disproportionate to Anhui’s share of national

cultivated acreage and its importance Of agriculture development.

Net Change of Cultivated Land
at Province Level (1980-1990)

1,000 ha
7%

    

Land Use Change

|:1- :39" 7) Saga Jargw thuarg Mu East
—_ 04, ~ - % Rag'on
- > 3%

Figure 2.4: Net Changes of Cultivated Land by County and by Province. Data Source:
Ministry of Agriculture (MOA)

38

Analyses of cultivated land use change between 1980 and 1990 shows the change is the
net result of loss of cultivated land in Shanghai (30 thousand ha), Jiangsu (370 thousand
ha) and Zhejiang provinces (658 thousand ha), and new addition of cultivated land in
Anhui province (706 thousand ha). The total net loss of cultivated land for the East
Region is 352 thousand ha during the period. However, the decline in productivity
resulting from this land 1055 was not proportional to the decrease in area; most cultivated
land 1055 is from encroachment of urban and rural development on the periphery of cities
and results in the loss of above-average quality land (Ash and Edmonds, 1998). At the
same time, most cultivated land newly brought into cultivation does not have equal
quality. For example, reclamation is by far the most important means of obtaining new
cultivated land (Guenther et a1, 1998) and a large proportion of reclaimed land constitutes

inferior land, whose productivity is far less than what has been lost.

2.5 Driving Forces of LUCC

From our land use database, in particular the SLA data, we can ascertain the direct causes
of cultivated land use change. We estimated the magnitude of cultivated land loss and the
associated uses for converted land for each province during the period 1988 — 1995.

Figure 2.5 groups the cultivated land use change into four conversion processes.

39

Natural Disaster - 5.6%
Abandonment — 1 .27.

  

 

Uban spawl - 14.2%
Watercansemney— l2.2‘/o

TVEs — 42%
Rural housing — 4.77.
Water comrmncy — 6.7%
Otlers rural use - 4.07.

Horticultm - 33.8%
Fishpond -— 8 .67.
Forestry — 4.7%

Gmlmli — U 27.

Figure 2.5: Conversion Processes of Cultivated Land. Data Source: State Land
Administration (SLA)

Between 1988 and 1995, the total land area converted from cultivation to non-cultivation
uses was 526.7 thousand ha. The dominant source of change was conversion to other
forms of agriculture due to agricultural structure change, accounting for nearly half of the
total conversion (47%). The most important activity in this category was conversion to
horticulture, which alone accounted for 34% of all the conversions. The second most
important contributor was urbanization, including sple of urban areas into once
cultivated land and water conservancy for urban use. Urbanization accounted for 26% of
the loss of cultivated land. The third source of change was rural development, which
included rural housing construction, water conservation projects and Township & Village
Enterprise (TVE) development. This accounted for 20% of all conversions. The

remaining 7% were due to other causes such as natural disasters and cultivated land

40

abandonment. Among these land use changes, urban construction, rural housing and
Township & Village Enterprises (TVEs) are irreversible losses. Depending on the degree
to which natural disasters and abandonment are irreversible losses, as much as 50% of the

loss of cultivated land during the period was permanent.

2.5.1 Agriculture Structure Change

Agriculture structure change is the most important cause of cultivated land loss in the
East Region. Of the 526.7 thousand ha of cultivated land conversion occurred between
1988 and 1995, 246.7 thousand ha was due to agriculture structure change, which was the
result of changes in the overall agricultural production system as the government policy

and planning embraced a more market-oriented economy.

In the past, the Chinese production system had been organized around a farming team
system, under which the government decided what and how much to plant, and teams of
farmers were assigned points based on their participation in the system. Points were then
used as the basis for income distribution. But the adoption of the Household Production
Responsibility System and later the transition from a centrally planned economy to a
market-oriented economy have allowed farm households to decide what and how much
agricultural commodities to produce, i.e., the decision making at the farm level are now
being guided by household economic returns rather than directives from the government.
Thus, farmers who had been used to a fixed-price system were rapidly pushed into a

market system. Without appropriate guidelines and regulations, overreaction to market

41

Signals often occurred, which lead to a rush to produce hi gh-priced agricultural

commodities, such as horticultural products, over staples and grains crops

An increasing proportion of land has been taken out of crop cultivation and put into other
agricultural uses that provide higher economic returns. Table 2.3 presents land use
change statistic compared with net economic value for various land uses in the region. It
is clear that the highest gains occurred to land uses with the highest economic potential,
resulting in net replacement of cultivated lands by land uses with as much as a 3-fold

more economic value.

 

 

 

 

 

 

Table 2.3 Economic Assessment of Agricultural Structure Change
Agriculture Net Economic Return Land Use Change
Type (yuan/ha). (1,000 ha)

Fishpond 14,970 46
Horticulture 8,265 179
Grassland 585 1

Forestry 555 25
Cultivation 5,535 -250

 

 

 

 

 

Note‘: Edited from the National Office of Regional Planning for Agricultural Resources
(1997). l Yuan 5 8.3 US$

From these statistics, it is apparent that the process of land use change from cultivation to
horticulture and fishponds has by itself improved the region’s agricultural economic

output by approximately 900 million yuan during the period of 1988-1995. Economic

42

returns from grassland and forestry land are as low as 585 yuan/ha and 555 yuan/ha
respectively. However, the land converted from cultivation to grassland or forestry
generally had a low productivity either because of being infertile or environmental
degradation. But even if we assume these lands originally had the average cultivated land
productivity, we can still conclude that the total agricultural structure change has
produced a substantial net economic benefit of 800 million yuan for the East Region, an

economic benefit with the price of losing important land for grain and cereal production.

2.5.2 Urbanization

Urbanization associated with industrialization is well known as one of the most important
driving forces for land use/land cover change. Most metropolitan areas around world
have experienced massive land use change associated with rapid urban sprawl. This is
also the case for the four capital cities in the East Region -- Shanghai (Shanghai
municipal), Nanjing (Jiangsu province), Hangzhou (Zhejiang province) and Hefei (Anhui
province). Table 2.4 provides a summary of urban expansion and population growth for
the four capital cities during the period 1986 - 1996. The ratio of urban land growth rate
to urban population growth rate is defined as the elastic ratio of urban area expansion.
The urban population of Hangzhou increased dramatically, more than 160% over this ten-
year period. Shanghai’s population increased by 1.2 million within ten years. Nanjing as
well as Hefei also experienced rapid population growth. Correspondingly, the urban areas
of these cities grew rapidly. Shanghai’s urban area expanded from 20 thousand ha to 36
thousand ha, 8-fold faster than population growth. The urban area of Nanjing increased

from 12 thousand ha to 17 thousand ha, and Hangzhou expanded from 63 km2 to 102

43

kmz. Detailed urban area data about Hefei during 1986-1996 is not available for this

study. For the East Region, the overall urban population grew fi'om 43.7 million in 1980

to 50.4 million in 1990, an increase of 15.3% during within ten years. The urban area

increased from 34095 km2 to 35751 km2, an increase of 3.7%.

 

 

 

 

 

 

 

Table 2.4 Urban Population Growth and Area Expansion
Urban Population1 Urban Area2 Elastic
City Name (million) (1000 ha) Rafi” ”f
Urban
1985 1995 Change 1986 1996 Change Growth
1%) ("/o)
Shanghai 12 14 10 20 36 78 8.1
Nanjing 2 3 29 12 17 36 1.2
Hangzhou 2 4 164 6 10 62 0.4
Hefei l 1 44 n/a n/a n/a n/a
All Cities 44 50 15 3448 3575 4 0.2

 

 

 

 

 

 

 

 

 

Notel: from Population Projection of United Nation (1998); Note2: from State Land
Administration Research Project of Cultivated Land Protection (1998). Urban areas for
Shanghai, Nanjing and Hangzhou are based on remote sensing satellite images.

Besides population growth, urban development is another important contributing factor to

urban area expansion. Large areas of cultivated land were arbitrarily zoned for

development. However, oversupply, unbalanced and mismatched development, as well as

spiraling inflation are causing considerable problems in China’s development economy

(Chan, 1997). Consequently, large areas of cultivated land have been cleared and zoned

for urban development and yet lie idle without any subsequent development.

 

2.5.3 Rural Development

China is still a nation of farmers. Before the economic reforms of 1978, nearly the entire
rural labor force was engaged in agriculture, contributed more than 70% of the rural gross
products. As agricultural production becomes more intense and economic reforms
encourage the development of rural-based industry, the economic structure in rural areas
became more diversified. Urban industrialization was extended to rural areas with the
Township & Village Enterprises (TVEs). In mid-19805, the rural Township & Village
Enterprise (TVE) industry output contributed about 50% of the gross output of the rural
areas in the East Region, whereas Shanghai had the highest percentage (Table 2.5). By
1995, each province had experienced significant increases in TVE output. TVEs
contributed 85% of the rural total output, and became a dominant source of rural income

in the East Region. The pace of TVE development far exceeds agriculture growth rates.

Table 2.5 Rural Economy Structure in the East Region

 

 

 

 

 

 

 

1985 1995
"Win“ Agro TVE TVE % of Rural Agro TVE TVE % of Rural
Output Output Output Output Output Output
Shangh“ 31 94 75 182 1520 89
Jiangsu 289 352 55 1687 8890 84
Zhejiang 174 200 54 892 7478 89
Anhui 198 41 17 980 3012 75
East Region 692 687 50 3661 20900 85

 

 

 

 

 

 

 

 

 

Unit: 100 million yuan, in current price. Data Source: China Statistical Yearbook, 1986
and 1996.

45

The construction of TVEs explained about 13 % of the land withdrawn from cultivation
between 1988-1995. In addition to the land that was lost as the result of rural industry
construction, TVES are responsible for other long-term impacts on cultivated land use

change:

First of all, the development of TVEs stimulates rural home construction. Although
agriculture continues to provide work for the majority of rural population, grth of rural
industry has created more disposable income available for improving living standards. As
a result, the development of TVEs has stimulated increased demand for rural housing

construction.

Secondly, TVEs cause land degradation. Most rural industries are low technology
oriented and highly polluting activities. The intrusion of these industries into rural areas
has resulted in rural environmental pollution that causes major crop losses and land

degradation.

Thirdly, the development of TVEs increases land abandonment. The development of
TVEs is necessarily accompanied by a steep decline in the rural labor force involved in
agricultural activities. The labor force leaving agriculture production is mostly comprised
of younger and better-educated individuals. An increased industrial labor force has come
at the expense of the rural labor force, and this has intensified the shortage of human

capital in agricultural sector, leading to the abandonment of cultivated land.

46

Finally, TVEs lead to an “invisible” loss of cultivated land. As the economy develops, the
opportunity cost of crop farming increases. Farmers become less interested in labor-
intensive multiple-cropping systems. Double or triple cropping, a practice which once
played an important role in rice production, has declined in Zhejiang, Shanghai and

Jiangsu provinces (Figure 2.6).

 

250 '1

240 -

220 ‘ +Zhejiang

"a 200 _ + Shanghai

+Jlangsu

180 4 +Anhui
+ China

160 -
Year

140

 

 

Figure 2.6: Change of Multiple Cropping Indices over Time. Source: Statistical Yearbook
of China, 1980-1996. MCI is calculated by dividing sown area by cultivated area.

2.5.4 Other Causes

Two other striking contributors to cultivated land 1055 are natural disaster and soil
erosion. Between 1988 and 1995, 29.3 thousand ha of cultivated land in the East Region
were destroyed as the result of natural disaster. In 1991 alone, 11.3 thousand ha of
farmland were forced out of cultivation by such natural disasters. Another 6 thousand ha

were abandoned due to various other reasons such as soil erosion.

47

Located in the lower level of the Yangtze River, the East Region has suffered a lot from
natural disasters, particularly floods, which inundate and degrade vast tracts of cultivated
land every year. According to the 1995 China Environment Yearbook, deforestation in
the upper reach of the Yangtze River has resulted in a 40% increase in silt accumulation
in the Yangtze River since 19505, which has caused the riverbeds in the lower reach to
rise several meters higher than the surrounding farmland, causing flood ahnost every
year. Much of the soil erosion can be attributed to the region’s topography. More than
half of the land territory is made up Of mountainous areas, where top soil is typically thin
and unstable. Coupled with a semi-tropical climate, erosion has been accelerated by the

rains in these areas.

2.5.5 Summary

The forces driving changes in land use and cultivated land 1055 can be categorized into
four components: agricultural structural change, urbanization, rural development and
others factors related to land degradation and land abandonment. Figure 2.7 illustrates the

influence of these factors.

For Shanghai, the dominant sources Of change are urban construction and rural
housing/industry development. In 1988, agriculture structural change explained about 40%
of the total cultivated land 1055 in Shanghai. These sources of change fit well with
Shanghai’s changing economic and demographic profile of rapid urban sprawl and rising
output fi'om TVE initiatives. The dominant factor in Jiangsu province is agriculture

structural change. Freshwater fish production in fishponds and horticulture are very

48

important agricultural income sources in Jiangsu, and increasing rapidly. Zhejiang also has
felt the impact of agriculture structural change, as well as rapid urbanization. Here natural
disaster appears to be more prevalent than other provinces. In Anhui, agriculture structural

change, urban construction and rural development have approximately equal levels of

importance as sources of cultivated land 1055.

 

 

    

0% , . .
1%8 1%9 19901991 19921993 1%4 19% 1%8 1% 1S3) 1&1 1M 1%3 1%4 1%
m Ari“
lAgro Structure Change I Urbanization a Rural Industry I Others

Figure 2.7: Explanatory Power of the Driving Forces in Each Province. Data Source: SLA

2.6 Land Use Change Model
Now we present a model of future changes of cultivated land use for the East Region. The

East Region of China can be viewed as an ecological-economic system integrating urban,

49

rural, agricultural and natural subsystems (Figure 2.8). The driving forces of land use
change described above are factored into this model to project future cultivated land use
change. Since the region is already intensively used, an expansion in urban area, rural
housing, industry land or horticulture land is accompanied by a loss of cultivated land or, to

a lesser degree, natural land. Urban and rural built-up areas are assumed to be irreversible.

  

u... Otherhndim
charge

— Cultivated Land
Us chame

 

 

.9...’IQ'IOVPI'DIIllfi'IIQ'IOCIO.PIQOII

 

Figure 2.8: A Conceptual Model for Cultivated Land Use Change

The general modeling approach of this study integrated both Von Thunen and Ricardo land
use models. Traditional Von Thunen model regards proximity to commodity markets as the
prime determinant for economic returns of different land uses, while the Ricardo model
describes potential returns in terms of specific land quality characteristics. The land use
change model in this study has two major explanatory components: a Thunen component

representing the impacts from the regional urban centers, and a Ricardian component

50

representing the local physical conditions. Land rents are regarded as a function of the
spatial proximity to city as well as land characteristics such as elevation, slope or soil
quality, which measure the land suitability to different uses (Geoghegan et a1, 2001).
These two components are combined in an integrated model to predict land use changes

between urban, agriculture and nature (Konagaya et a1, 1999).

Spatial proximity to commodity markets is commonly used in empirical land use models
(Capozza and Helsley 1989, Ludeke et al, 1990, Chomitz and Gray, 1996, Bockstael, 1996,
Turner et a1, 1996, Nelson and Hellerstein 1997, Cropper et a1 1999). For example, Ludeke
et al. (1990) found strong relationships between the deforestation pattern and the road
accessibility as well as the proximity to a house or shelter. Many other research results also

illustrated the strong spread effect and the importance of accessibility in land cover change.

There are three groups of driving variables being included in the land use model in this
study: (1) Urban situation variables (distance-weighted central urban population and local
urbanization level); (2) Rural situation variable (nrral industry output per capita) and (3)
Site variables (mean elevation and mean slope). Four land use types are distinguished in the
model: (1) urban and rural built-up land area; (2) cultivated land area; (3) horticultural land
area and (4) natural land area (including forest, grassland, water and other unused land).
We differentiate horticulture from cultivation land use because of the dramatic conversion
between these two land use types as the result of agriculture structural change.

The observation units of the model are counties.

51

The urban situation variables play an important role in detemiining the influence of urban
centers. Variables that simply measure the population of selected cities do not describe
how the impact from each city changes as its proximity to land varies. Shi et a1 (1997)
include a gravity index as an explanatory variable in a county-level hedonic model of
agricultural land price to account for urban influences. In this study, a gravity urban
population index is also developed to integrate the impacts of both population and
proximity. Urban population of the four capital cities in the East Region is associated
with each county in proportion to the inverse of the distance between the county under
study and the capital cities. A county with a shorter travel distance to the provincial
capitals is said to have higher accessibility, and thus experience a greater influence from

that urban center. The calculation can be made using equation 1:

4 4 4
Yizzlpij =Zl{[(1/di,-)/_Zl(1/du)] * Pj} (1)
= J: J:
Where:
Y, ---- distance-weighted central urban population for county i

i ------ county number (i = l, 2, 3. 237)

j ------ urban centers (1 = Shanghai; 2 = Nanjing; 3 = Hangzhou; 4 = Hefei)
Pi,- ---- urban impact of county i from capital city j

P, --—- urban population for capital city j

dij ---- distance between county i and capital city j

52

concentric zones of accessibility extending from Shanghai to the west is apparent. This
spatial pattern resembles very much the distribution of Township & Village Enterprises
(TVEs) as depicted by Figure 2.9.2, which suggests that rural industry would spread to
the west in the future. In fact this appears to be the case today as shown in Figure 2.9.2.
Figures 2.9.3 and 2.9.4 show the average elevation and slope for each county. Most of the

high elevation and high slope land is in the south of the region.

53

     
  

   
  

Urban Access

[:1 Lowest
Low . 300-600
- High ‘ - 600- 1500
- Highest - >1500
Figure 2.9.1: Urban Accessibility Figure 2.9.2: Rural Industry Output

   

Elevation
:1 < 300
[:1 300-600
- 600-1500
- > 1500

 

Figure 2.9.3: Elevation Figure 2.9.4: Slope

54

These spatially articulated variables (accessibility, elevation and slope and rural industry
output) are incorporated into a land use change model in the following manner: for each
county i, the ratio of land use type c is defined as an exponential function of the urban
impact variables, local physical variables and rural development variable (Equation 2).
Because the observation units (counties) are heterogeneous in size, the absolute extents of
different land uses are not appropriate for modeling purpose. Instead, the ratio of each
land use type K,(c) is used as the dependent variables. It also describes the probability of
a given land use type occurring in a given county and at a certain point in time. Since the
dependent variable K,(c) is the ratio of land use type c in county i, all the four land use

types in county i should add up to 1 (Equation 3). The equations are specified as follows:

K103) : k0i(C) {CXP[31(C)*Yi+3z(C)*Xi]} * {expla3(c)*zii+34(C)*Zzili
* exp[a5(c)*R,] (2)

4
CEIKKC) E 1 (3)

i ------ county number (i = 1, 2, 3,. . ., 237)

j ------ urban centers (1 = Shanghai; 2 = Nanjing; 3 = Hangzhou; 4 = Hefei)

c ------ land use type (1 = Cultivate; 2 = Built-up area; 3 = Horticulture; 4 = Nature)
K,(c) ----- ratio of land use type c in county i

K0, --- constant term

Y, ----- distance-weighted central urban population as defined by equation (1)
X, ----- ratio of urban population to total population for county i
Z], ---- mean elevation of county i

55

22, ---- mean slope of county i
R, ----- per capita rural industry output in county i

An(c)-- model parameters for land use type c

The parameters, from al to as, are estimated from the transformations:

Lani(C)] = Ln[k0,-(C)] + 31(C)*Yi + 32(C)*Xi + 33(C)*Zli + 34(C)*ZZi + 35(C)*Ri
(4)

Equation 4 is a multinomial logit model, where the dependent variable is a four-
dimensional vector of land use ratios for cultivated land, built-up areas, horticultural land
and natural areas. Statistical significance of a logit model is tested using likelihood ratio
(P2) test statistics, which follow a Chi-square distribution with t degrees of freedom (t =
number of parameters been estimated in the model). The likelihood ratio index is similar
to the determinant coefficient (R2) of a regression model that is used to examine the
goodness of fit of a model. P2 = 1 when the model estimate fits the observed data
perfectly, and when P2 = 0, it means that the explanatory variables do not explain the

dependent variables at all.

Figure 2.10 shows sources from which data are obtained for the model. County-level land
use data come from the land survey conducted by the Ministry of Agriculture, China.
Urban population data are taken from United Nation urban population projection. Urban
accessibility is generated in a GIS, using the Euclidian distance between county centroids
and urban centers. Rural industrial production data came from regional statistic yearbook

published by the Statistic Bureau of China. Elevation and slope data are derived from a

56

Digital Elevation Model (DEM) of USGS. All the data are geo-coded to county-level

administrative boundary coverage.

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MOA Land UN Urban County County USGS
Use Population Boundary Statistical GTOPO3O
Survey Projection Map Data
. 1 .
f
Central City Distance to
* Population Central City
Pj Dii
Per Capita Mean Mean
Land Use Distance weighted Rural Elevation Slope
Ki(c) (c=l~4) urban Population Yij Industry Ri Zil 212
¥
Year 1980 Data ’ Maximum Likelihood Estimates
l alic). a2<c>, asic). a4(c). a5<c>

 

 

 

 

 

 

 

Trend Analysis of h ] . Model
[ Year 2025 and L Year 1990 Data Validation

Projection of year
+ 2025 and 2050

 

 

 

 

 

 

Figure 2.10: Flow Chart for Data and Modeling Methodology

Figure 2.10 also shows the major steps of model estimate. Land use data as well as the
corresponding explanatory variables for year 1980 are used to estimate the parameters of
the model using maximum likelihood method. Assuming the parameters remain constant
over time for each land use type, we apply the data of 1990 to the model and compare the
model-estimated results with the observed data to examine the explanatory power of the

model. Explanatory variables for 2025 and 2050 were projected by trend extrapolation,

57

i.e., with an underlying assumption that the current trend for either urban population
growth or rural industry development will continue through the projection year. Finally
we use the multinomial logit model to project future land use change for the years 2025

and 2050.

2.7 Modeling Results
We ran the model in the Guide to Available Mathematical Software (GAMS). The model
explains land use in 237 counties of the East Region with a P2 = 0.793. All parameters

were statistically significant at a =0.0S. The modeling results are as follows:

Urban Imgrcts Rural Industiy Local Physical

 

 

 

K,(1) = 2.41 [exp(-0.65*Yi —1.58*X,)] [cxp(-O.87*R,)] [exp(O.l 1*21i — 0.74*22,)] (4.1)

K,(2) = 1.89[exp(1.90*Yi+1.58*X,)] [exp(2.l7*R.-)] [exp(-0.95*Zli— 0.19*22,)] (4.2)

K,(3) = 1.26 [exp(0.97*Yi +0.73*x,)] [exp(5.76*R.)] [exp(-0.37*Zli + O.48*Z2,)] (4.3)

K,(4) = 2.16 [exp(-l .37*Y, 4.34*x,)] [exp(0.47*R.-)] [exp(0.04*Zl;— 0.42*22,)] (4.4)

Equations 4.1 — 4.4 are the land use model for cultivated land, built-up area, horticultural

land and natural land respectively. Taking a close look at the parameters for the

equations, we can note that impacts from the four provincial capital cities as well as the

local urban population will negatively affect a county’s cultivated land ratio and natural

land ratio, but positively related to built-up area as well as the horticulture land ratio. This

supports the hypothesis that urbanization will lead to a loss of cultivated and natural land.

58

A more urbanized regional landscape will have more built-up area, more high-profit

horticultural land, but at the cost of less cultivated land and natural area.

The model results also suggest the importance of rural industrial development as a
driving force for regional land use changes. The influence of rural industry is negative on
cultivated land, but positive on all other three types of land use. As analyzed previously
in the study, the development of TVEs in rural areas is essentially a part of the
urbanization process. Increased TVE development will bring more built-up area and
create a higher demand for horticultural land, as the result, impose a threat for the loss of
valuable cultivated land. However, equation 4.4 shows that rural industry development
has a positive coefficient for natural land use. This may be attributed to the
environmental reconstruction projects (for example, reforestation) sponsored by TVEs as

part of community constructions.

With respect to local physical variables, slope is negatively related to cultivated land
ratio, while elevation is positive. National guidelines for proper cultivation practices
stresses that cultivation is best on land with a slope less than 30% because of the high
correlation between slope and soil erosion. Therefore, any area with a slope greater than
30% is unlikely to be developed for cultivation. The positive sign of elevation may be
explained by the fact that land on high elevation will still be reclaimed in a densely
populated area like the East Region. It might also suggest that physical variables are a
less important constraint in regional development than other factors such as accessibility

and demographics.

59

Figure 2.1 1.1-2.1 1.4 compare observed cultivated land data with model estimates for
both 1980 and 1990. Comparison indicates that the model captures the pattern of
cultivated land distribution very well. For 1980, counties with more than 48% of
cultivated land were mainly located in the plains of the East Region. Counties with
cultivated land less than 32% or even 16% were mostly in the southern mountainous
areas. Counties with cultivated land between 32-48% can be found in a conversion belt in
between the former ones. When comparing observed cultivated land data in 1990 with
1980, we observe a decrease in the number of counties with a high percentage (>48%) of
cultivated land, and an increase in the number of counties with a medium percentage of
(36-48%). Evaluating the model with the driving force variables for 1990, there is good

correspondence between model results and the observed spatial distribution.

60

  

Cultivated Lan
[:1 <16%
[:1 16-32%
- 32-48%
- >48%

Figure 2.11.1: 1980 Estimated

Cultivated Land %
[:1 < 16%

[:1 16-32%
- 32-48%
- >48%

Figure 2.11.3: 1990 Estimated

  

  

61

  

Cultivated Land "/oI
1:] < 16%

:1 16-32%

- 32-48%

- >48%

  

Figure 2.11.2: 1980 Observed

  

Cultivated and%
[:1 <16%

[:1 16-32%
- 32—48%
- >48%

Figure 2.11.4: 1990 Observed

The model is used to project future land use changes for the East Region. Model
projection shows a reduction of cultivated land to the level of 12.3 million ha by the year
2025, a total loss of 1.24 million ha from year 1990, accounting for an average annual
loss of 35.3 thousand ha. Between 2025 and 2050, another 113.3 thousand ha of
cultivated land is estimated to be lost. In percentage terms, the model predicts a decrease
of cultivated land by 9.2% from 1990 to 2025. The rate of cultivated land loss will slow
during between 2025 and 2050, with a total loss during this period of only 0.9%. The per
capita cultivated land in the East Region will be 0.073 ha in 2025, and 0.072 ha in 2050.
The spatial distribution of cultivated land in the year 2025 and 2050 (Figure 2.11.5 and
2.11.6) shows that counties with a high cultivated percentage (>48% of the total land
area) will be clustered in the north part of Jiangsu and Anhui provinces. Most counties in

Zhejiang province, and the hilly part of Anhui province will have cultivated land of less

than 32%; half of those counties will have a percentage of cultivated land less than 16%.

      
   
   
     

   

Cultivated Land % Cultivated Land %>
< 16% Cl < 16%
CI 16-32% [:1 16-32%
- 32—48% - 32-48%
- >48% - >48%
Figure 2.11.5: 2025 Estimated Figure 2.11.6: 2050 Estimated

62

2.8 Conclusions and Discussions

This study shows that urbanization, agriculture structural change, rural development and
local physical environment have all played important roles in the cultivated land use
change process in the East Region of China and will continue to be important
determinants in the future. Results from prognostic modeling analysis suggest that
cultivated land will continue decreasing at a rapid rate for another two decades before it
reaches a stabilization level of 0.0727 ha/capita. This value exceeds the self-sufficiency
level defined by FAO (0.053 ha/capita), so we conclude that food security should not be a

large problem in the East Region.

Because all driving factors of cultivated land use change are susceptible to policy
intervention at least to some degree, policy-makers will have some flexibility in
addressing the cultivated land loss problem and food security issue. Urban development
has enjoyed a high priority in the national agenda for about half a century, and the
government has realized the need to slow urban expansion as a measure to reduce its
encroachment on arable land. Structural adjustment in the agriculture sector and
promotion for rural industry development are more recent policy priorities. Directly
associated with policies for agriculture structural change and rural industry development
are govemment’s efforts to stimulate increases in farmers’ income, which have the
potential to further increase the loss of cultivated land in the future. However, it is quite
possible that government policy will focus on the fundamental economic objective of

grain self-sufficiency and efforts aimed at guaranteeing ample domestic grain supply.

63

Such policy positions might lead to more restrictive regulations, hence limiting land uses

which result in conversion of cultivated land.

In this study we used county-level survey data to quantify land use change. Spatial
articulation was resolved at the county level, and hence we may have underestimated the
importance of considerable local variation. In addition to these issues of scale,
underreporting in land use data from government sources is another important concern
(Ash, 1998; Yeh and Li 1997). To resolve both issues we suggest the use of remote
sensing data for future analyses. Remote sensing data with high spatial resolution can be
a powerful tool for refining the measurements of the rate, extent and spatial pattern of and
use/land cover change. It is an efficient means of monitoring land use change, providing
fine resolution data of land use/land cover, and enabling spatially explicit quantitative
analysis and modeling of land use/land cover change. Since we can collect remote
sensing data at multiple scales and at multiple times, we are able to analyze various land
use changes synoptically from local to regional scales through time, and to consider the
real spatial arrangements of landscape from remote sensing scenes, characterizing and
mapping the land use pattern with a great detail that is not possible by using the census

data alone.

2.9 References

Ash, RF. and Edmonds, R.L., 1998. “China’s land resources, environment and
agricultural production”. The China Quarterly, issue 156, p836-879.

Brown, LR, 1995. “Who will feed China?” New York, W.W. Norton.

64

Capozza, D.R.; Helsley, R.W., 1989. The fundamentals of land prices and urban growth.
Journal of Urban Economics. 26: p295-306.

Chan, N., 1997. “Land-use rights in Mainland China: Problems and recommendations for
improvement”. Journal of Real Estate Literature, Vol.7, No.1, p53-63.

Chomitz, KM. and Gray, D. A. 1996. Roads, land use, and deforestation: a spatial model
applied to Belize. The World Bank Economic Review. 10(3): p487-512.

Cropper, M., Griffiths, C. and Mani, M., 1999. Roads, population pressures, and
deforestation in Thailand, 1976-89. Land Economics. 75(1): p58-73.

FAO/UNDP/Ministry of Agriculture in China, 1997. “ Promotion of sustainable
agriculture and rural development in China: elements for a policy framework and a
National Agenda 21 Action Program".

Fischer, G. et a1, 1998. “The balance of cultivated land in China during 1988-1995”. IR-
98-047/August. International Institute for Applied System Analysis, Laxenburg, Austria.

Geoghegan, J., Villar, S.C., Klepeis, P., Mendoza, P.M., Ogneva-Himmelberger, Y.,
Chowdhury, R.R., Turner H., B.L., Vance, C., 2001. Modeling tropical deforestation in
the Southern Yucatan Peninsular Region: Comparing survey and satellite Data.
Agriculture, Ecosystems, and Environment, V0. 84, Nos. 1-3, p25-46, 2001.

Hanink, D. M. and Cromley, R. G., 1998. “Land use allocation in the absence of a
complete market values”. Journal of Regional Science, Vol.38, No.3, p465-480.

Information Office of the State Council, P.R.China, 1995. “White Paper ---- Family
Planning in China”.

Konagaya, K., Morita, H., and Otsubo, K., 1999. “Chinese land use predicted by the
GTR-Model”. Discussion paper in the 1999 Open Meeting of the Human Dimensions of
Global Environmental Change Research Community, Tokyo.

Ludeke, AK, 1990: An analysis of anthropogenic deforestation using logistic
regression and GIS. Journal of Environmental Management, 31, p247—259.

National Office of the Regional Flaming for Agricultural Resources, 1997. “Analysis
and Assessment of the Ten-year Change of China’s Agricultural Resources, 1986 -
1995”.

Nelson, G.C. and Hellerstein, D., 1997. Do roads cause deforestation? Using

satellite images in econometric analysis of land use. American Journal of Agricultural
Economics. 79(1): p80-88.

65

Shi, Y.J., Phipps, T.T. and Colyer, D., 1997. Agricultural land values under urbanizing
influences. Land Economics. 73(1): p90-100.

State Land Administrative Research Project of Cultivated Land Protection, 1998. “The
recent change of cultivated land and medium-term trend analysis”. China Social Science,
2(1998), p75-98.

Streets, D., et al., 1995. “Remote sensing of global change: grth in China’s Jiangsu
province”. International Journal of Sustainable Development, 2(1995), p275-266.

Turner, M., Wear, D.N., Flamm, R.O., 1996. Land ownership and
land-cover change in the southern Appalachian Highlands and the Olympic
Peninsula. Ecological Applications. 6(4): p1150-1172.

UNDP, China Environment and Sustainable Development Resource Book II: a
Compendium of Donor Activities, 1996.

Wear, D.N., Bolstad, P., 1998. Land use changes in southern Appalachian landscapes:
spatial analysis and forecast evaluation. Ecosystems. 1: p575-594.

Yeh, A. and Li, X., 1997. An integrated remote sensing and GIS approach in the

monitoring and evaluation of rapid urban growth for sustainable development in the Pearl
River Delta, China. International Flaming Studies; Abingdon; Vol. 2, No. 2, Jun 1997.

66

CHAPTER 3

Urban Sprawl and Land Use Change in Shanghai, China

Urban sprawl has been reported to be one of the most significant development issues
globally and held responsible for a host of problems such as rapid loss of farmland,
natural vegetation and open spaces, profligate energy consumption, rising infrastructure
costs and decline of social fabric (Tjallingii, 2000; Yeh and Li, 1998). This study will
focus on its detrimental impacts on land use/land cover change. Shanghai, a fast-growing
metropolitan area in China that has little unused arable land and a fragile ecosystem, is

selected to be a case for studying urban sprawl and its impact on rural landscape change.

The study will proceed in five sections. Section one introduces the process of urban
sprawl and the consequent economic and ecological significances. Section two outlines
the urban sprawl problem in Shanghai and how urbanization and sub-urbanization drive
urban sprawl and land use/land cover change dynamics. Section three uses remote
sensing to monitor the magnitude, rate and pattern of land use/land cover change in
Shanghai. Section four constructs a land use/land cover change model using Principal
Component Regression modeling method that combines the remotely sensed satellite data
and statistical data. Section five draws conclusion from monitoring and modeling results,

discusses the limitations of the study and shows future research direction.

67

3.1 Introduction

Over the last half century, widespread emergence of urban sprawl has characterized city
growth patterns for major metropolitan areas in the world. An estimated 5 million to 7
million hectares of farmland are disappearing from the world each year due to urban
sprawl. Natural areas, such as forests, wetlands and other ecosystems, are also being

sacrificed to urban development such as residential estates, industry and tourism.

Although the amount of land converted to urban uses may be far smaller globally than the
amount of natural lands being lost to agricultural activities and grazing, urban sprawl has
more profound impacts on people’s lives. Cities are at the heart of a changing global
economy, playing an increasingly important role in the economic standing of nations.
Urban build-up area occupies only 1-2% of the Earth’s land surface, yet it houses nearly
half of the world’s population and uses three quarters of the resources we take from the
earth, marking the most massive change in the flows of energy, water and materials
(Douglas, 1994). For historical reasons, cities are ofien located on prime agricultural land
with abundant natural resources or valuable ecosystems near water bodies. Yet residential
and commercial development is replacing these valuable lands at an unprecedented rate.
In terms of urban sprawl, it is the type rather than the scale of land being changed that is

more important (WRI, 1997).

Urban grth generally experienced two major stages: urbanization and sub-

urbanization. The evolution of urbanization and suburbanization has been amply

documented. It is also well understood that cities exert centripetal and centrifugal forces

68

at different stages of their history that attract people to the cities or push them toward the
periphery (Zhou and Ma, 2000). Urbanization began roughly around the turn of the 19th
century when large-scale industries took root in urban centers in developed countries and
attracted massive population movement into cities. Urban residents represented only 3%
of the world’s total population in 1800, yet the percentage rose to 14% by the beginning
of the 20th century, 29% in 1950 and 49% in 1998. Researchers at World Resources
Institute estimated that worldwide movement towards cities is growing at three times the
rate of population expansion worldwide; in ten years from now, two thirds of human
beings will be living in the cities. 90% of the urban population growth will occur in the
developing countries where urbanization started at a much later time but at a
unprecedented fast pace (WRI, 1997). Urban growth in developed countries arrived at a
sub-urbanization stage in the middle of the 20th century when oversized urban population,
high land price, severe environmental pollution and crowd transportation situations in
urban centers drove the location of urbanization centers away from the central cities to
suburb areas. This new urban grth stage was also facilitated by the modern age of
transportation network and automobile that made it possible for suburbs to gradually take
the place of the central cities as the area with the fastest growing population and

economic development.

Both urbanization and sub-urbanization consume large amounts of land. Rural landscapes
surrounding most of our cities and towns are being converted to urban uses at an
accelerating rate. The land conversion pressures are especially high in developing

countries where cities are growing at least twice as fast as those of the developed world.

69

The general pattern of urban growth in developing countries today is marked by rapid
growth of urban population and expansion of urban area, not much different fi'om what
occurred a century ago in the developed world. What is unprecedented is that, in some
major developing world cities like Shanghai, China, urbanization is still at its most
intensive stage while sub-urbanization has started to play its role on stage, a case in
which urban growth would exert double pressures on the neighboring land, therefore,
have far-reaching impacts .on land use/land cover changes (UN ESCAPE, 2001; Zhou and

Ma, 2000.

In order to ensure food security and compensate the agricultural production due to the
loss of agricultural land, crop production on the remaining land has to be more intense, a
practice that is heavily dependent on substantial new financial investments or technology
improvements (Ash and Richard, 1998). An alternative is to reclaim more land into
cultivation. But reclamation can push agriculture to frontier areas with marginal
productivity and entail high ecological risks for fiagile ecosystems by forcing birds and
animals from their natural habitats and surroundings. Large-scale reclamation projects in
Shanghai have forced hundreds of species of wild birds and reptiles from their natural
environments (United Press International, 2000). Although the city government has
passed environmental laws aiming at protecting the wildlife, developers have flagrantly
violated the laws, and the number and variety of wildlife species in Shanghai had

declined by 10 percent in the past year.

70

This study aims to explore the impacts of urban grth on land use/land cover change,
with a case study in Shanghai, China where rapid development of industry and the
expansion of the city have systematically taken over large areas of agricultural land,
seized limited green land and natural reserve, threatening the urban economic and
ecological environment. There are two main research tasks in this project: to monitor
urban sprawl using remote sensing and GIS, and to model land use/land cover change
based on remotely sensed observations and theoretical understandings of land use/land

cover change processes.

LUCC monitoring intends to determine the pattern and magnitude of urban sprawl and
rural land use change. Remote sensing provides a strong tool to visualize the spatial
patterns and temporal sequence of urban growth patterns, and to convey how the progress
of modern urbanization and sub-urbanization can result in profound changes to the
landscape (Dai et a1, 1996; Li, 1997; Mei, 1992). The priority task of this project is
LUCC modeling based on the understanding of LUCC processes as results of urban
growth. The LUCC model can play an important role to generate projections for future
land use/land cover changes by simplifying the multidimensional processes and
highlighting the important variables as well as the causal/impacts relationships involved
in the processes. Good model projections may contribute to the design of policy

responses to urban growth management.

71

3.2 Shanghai: Urbanization and LUCC

The Shanghai municipality is located at the mouth of the Yangtze River Delta, bordered
by the J iangsu province on the north and west and Zhejiang province on the south (Figure
3.1). It includes fifteen districts (Huangpu, Nanshi, Luwan, thui, Changning, Jing’an,
Putuo, Zhabei, Hongkou, Yangpu, Pudong New Area, Minhang, J iading, Baoshan and

J inshan) and five counties (Nanhui, Fengxian, Songjiang, Qingpu and Chongming). With
an area of only 6,340 kmz, Shanghai houses a population of 16.74 million (data of March,

2001), therefore, is one of the world’s most populous urban areas.

Shanghai Municipality Shanghai Central City .

Zhabei Y
’ ' anpu
Chongnnng . Hongkou
Putou ‘ ‘_: .V-
_\“_g__'_”._,,/;J in I an H U31] gp U z

,’ ' Luwan r
Changnni , Nanshi
thui ’

Jiangsu

A
\g '

fiAVL\Shanghai

 

Zhefiang

Figure 3.1: Study Area: the Shanghai Municipality

3.2.1 Urban Sprawl
Like in most other metropolitan areas over the world, rampant urban sprawl has

characterized the urban growth patterns for Shanghai during the past half century (Chen

et a1, 2000; Mei, 1992). Despite the administrative criteria that make a clear distinction
between rural and urban areas (Zhang and Zhao, 1998), urban space in Shanghai has
expanded dramatically outwards to encroach on the countryside, and rural space has been
transformed into an extension of the urban landscape. Among all the large metropolitan
areas in China, Shanghai has the highest rate of agricultural land loss. Study of Chen et al
(2000) shows that the build-up area in Shanghai has expanded rapidly from 91 km2 in

1947 to 364km2 in 1996, a four-fold urban area expansion within half a century (see

 

    

 

Figure 3.2).
.7,— .2“? 53 ~ ,r";.r.é;::\;' .m-‘rfl—r/T»? ‘ 1 L‘s-z. \-.
Year 1947 "1’"? ,2 ‘ f i; YearI979 “ """ ‘ T ’
;;._:.’ a ’ J}?- l 4-31. ,.{.,.,x.:/“‘:"7

         

T

“5’

      

 

 

 

      

r, V 7‘" “- :‘flfi‘ :7 V ‘17.": VP; i l —‘ah‘

——"——,—t-:V ~. ‘ .' i 3‘
, Year 1 96 -,J ' i " ;
I“F:i-4}*W~ ‘ L

  

 

 

 

 

 

Figure 3.2: Urban Area Expansion for the Central City of Shanghai, 1947 - 1996.
Source: The Atlas of Shanghai.

73

3.2.2 Urbanization and Suburbanization

As the largest city of China and one of the world’s most important industrial, commercial
and financial centers, Shanghai has long been the largest urban economy and domination
over China's commerce (Murphey, 1953; Johnson, 1995). But urbanization in Shanghai
remained modest until the communist victory in 1949 (Atash and Wang, 1990; Chang,
1963), since when it has become an administrative entity directly subordinate to the
central government and enjoyed high policy priority on government investment. It is now
a leading model in the modernization of China and an industrial giant whose products
supply China's growing domestic and international markets (Y eung and Sung, 1996).
Except for policy incentives, many of the driving forces of urbanization in Shanghai
remain the same as what occurred a century ago in the developed world, with the most
important ones being the shifiing job opportunities from agriculture to industry and
services, and the concentration of economic opportunities in urban areas (Fung, 1982). It
is the large scale of immigration into the city that keeps the overall population of

Shanghai growing.

Rural population in Shanghai grew slightly from 3.2 million in 1959 to 3.5 million in
1998. In contrast, urban population in Shanghai increased from 7.2 million to 9.5 million
(Figure 3.3). Providing housing alone for such a large urban population would make a
large claim on land (Heilig, 1997; Verburg et a1, 1999). Besides removing agricultural
land from cultivation as the city expands, urbanization also affects rural land use change
by reducing the farmland area as more farmers move to cities or start to work for Town

and Village Enterprises (TVEs). Guildin (1997) argues that the indirect effects of

74

urbanization may cause a larger loss of agricultural land than physical conversion into
build-up areas. The booming of TVEs in Shanghai was resulted from the spill-out effects
of urban industry, and it has extended the urban industrialization to rural areas in

Shanghai, so-called rural sprawl (Zhou, 1997).

 

 

 

(n

O)

 

IUrban_pop

   

I mRuraLpop

 

t ' I T

1 968 1 973 1978 1983 1 988 1993 1 998

 

Figure 3.3: Urban and Rural Population Growth in Shanghai Municipality, 1959 — 1998

Decades of urbanization helped Shanghai to form a ring-structured spatial pattern that
corresponds remarkably well with the boundary of the city’s master plan (Zhou and Ma,
2000; UNESCAP, 2001). The central city, including the ten districts of Huangpu, Nanshi,
Luwan, thui, Changning, Jing’an, Putuo, Zhabei, Hongkou and Yangpu, is the core
area of urban growth (Figure 3.4). It covers an area of 280.8 kmz, accounting for 4.4% of
the total area. The first ring out of the urban core area is near suburbs, including the
districts of Pudong New Area, Minhang, Jiading and Baoshan. With an area of 1777 km2

(26% of the total area), the near suburb ring is the major supplier for fresh vegetable,

75

sideline food and crops. Outside for that is the far suburbs (including the district of
Jinshan and five counties - Nanhui, Fengxian, Songjiang, Qingpu and Chongming), a ring
of fresh water fishery, bases for fruits and crops, where intensively cultivated fishponds

and orchards hold a substantial percentage.

 

    

Far Subin'bs

 

 

 

Figure 3.4: The Ring Structure of Urbanization in Shanghai Municipality, China

Most of the early urban grth was concentrated in the central city (UNESCAP, 2001;
Ning and Yan, 1995). However, severe environmental pollution and high transportation
pressure started to force many industrial plants to relocate from the city center to the
outlying suburb areas in the 19803. The rapid growth of Township & Village Enterprises
(TVEs) in the later 1980s further dispersed the urban growth center away from the central
city; and the grand opening of the Pudong New Area under the strong support of the

central government in 1993 marked the beginning of a new suburbanization stage.

76

Suburbs started to take the leading role as the most active urban economic growth areas.
Figure 3.5 shows the distribution for the percentage of annual GNP value and population
in the three concentric rings of Shanghai. We can observe the decline of the central city
and the accompanied growth of the suburb areas, especially the near suburbs. The GNP in
the suburbs of Shanghai increased from 26 billion yuan in 1985 to 102 billion yuan
inl992 and 389 billion yuan 1997, indicating an annual GNP growth rate of 40% during
1985 — 1992, and 56% during 1992 - 1997. In the meanwhile, the GNP for the central city
experienced a growth from 59 billion in 1985 to 116 billion in 1992 and 153 billion in
1997, with an annual growth rate of 14% during 1985 — 1992, and 6% duringl992 —
1997. Shanghai is the only city in China that has experienced a negative natural
population growth in thel99OS. The population of the central city declined from 6.7
million in 1985 to 6.4 million in 1993 and 6.3 million in 1997, while the population for

the suburbs increased from 5.4 million in 1985 to 6.5 million in 1993 and 6.7 million in

1997.

   

1985 1992 1997 1985 1992 1997
.CentralCity mNearSubur’os .FarSuburbs 'CentralCity nNearSuburbs 'FarSuburbs

Figure 3.5: Distribution GNP Value and Population for 1985, 1992 and 1997

77

However, suburbanization in Shanghai is still at its primary stage despite the high speed
of development in the suburb areas (Zhou and Ma, 2000; UNESCAPE, 2001; Yusuf and
Wu, 1997). With nearly half of the municipality’s population and one third of the
municipality’s GNP, the central city still shows strong attractions for investment.
Therefore, the urban growth pattern in Shanghai since 1993 can be characterized as rapid
suburbanization in the near suburbs along with intense urbanization in the central city.
With the double pressures, Shanghai experienced the most extensive urban sprawl. Figure
3.6 displays the growth of urban build-up areas since the 19705, with a ring-structured
fashion of urban sprawl. The river running through the scene is the Huangpu River, and
the central city of Shanghai is located to the west of the river bank. Most urban sprawl
took place during the 19905 when the suburbs established most of its transportation

networks, housing complexes, parks as well as other recreational facilities.

 

Figure 3.6: Urban Sprawl in Shanghai, 19703 — 19905. Source: Jiangnan Remote Sensing
Institute, China

78

3.2.3 Land Use Change Dynamics

In order to examine the impacts of urban growth on land use change, we cannot isolate

the change of urban land from that of agricultural land or natural land; instead, we need

to look at the dynamic conversions between these land types as a coherent system. Figure

3.7 shows the dynamics of land use system in Shanghai.

 

 

 

 

 

 

Environmental Wetland &
Reconstruction Water Reclamation
Reclamation

 

I Forest I‘Environmental

Reconstruction

 

Urban Sprawl

 

 

! Agro land I

 

 

 

 

4 Urban Area ]

 

 

Urban Sprawl A
Rural Sprawl‘
, Y
I
I Rural Area I Urban Sprawl

 

 

 

Figure 3.7: A Systematic View of LUCC in Shanghai, China

There are four major activities contributing to the LUCC dynamics in Shanghai: urban

sprawl, niral sprawl, reclamation, and environmental reconstruction. As urban sprawl

forces large areas of prime agricultural land at the city edge out of cultivation, it also took

over neighboring rural communities and green open space.

79

But urban sprawl is not the sole contributor for the loss of agricultural land. Rural sprawl,
caused by the construction of rural Township & Village Enterprise (TVEs), takes more
agricultural land out of cultivation. In order to compensate the loss of agricultural land
and also to meet the growing demand for urban land use, Shanghai, “a city on the sea” in
Chinese, reclaims land from the sea on an annual basis. Since 1949, Shanghai has
reclaimed more than 660 km2 of land from the sea, an area equals to nearly two-thirds of
the total areas used in construction projects in the past decades (HASA, 1999). A large
proportion of the reclaimed areas are developed into farmland, and others are developed
into port or industrial construction use. For example, reclamation has played an important
role in the development of Chongming Green Food Zone, China's first state-level
pollution-free food production center on the Chongming island. The construction of
Baoshan Iron & Steel Co. on reclaimed land is another oflen-cited example of land

reclamation in Shanghai.

Environmental reconstruction programs also play an important role in Shanghai’s land
use change. A nice environment with green open space is very important in the city’s bid
for modernization. It can improve the urban environment by preserving water and soil,
adjusting the temperature and humidity of air, preventing pollution and dust, purifying air
and eliminating noises. The government of Shanghai municipality is making major
efforts to improve its environment by increasing the greenness percentage for the region.
The green coverage of the city has reported to increase from 8.2% in 1978 tol9. 1% in

1998.

80

3.3 LUCC Monitoring

The analysis of Chinese land use statistics is difficult due to large inconsistencies
between sources and the absence or unreliability of many critical variables. Official
statistics in China are generally assumed to have under-report problems (V erburg et all,
1999; Smil, 1995; Brown, 1995; Ash and Richard, 1998; Crook, 1991 and 1993). In this
project, we use remote sensing to estimate different land use types over time so as to
monitor urban sprawl and land use/land cover changes in Shanghai. Four scenes of
Landsat TM data with path of 118, row of 038 from 1987 to 2000 are acquired, with the
acquisition dates of 05/1 8/1987; 04/08/1990; 08/04/1998 and 06/15/2000. Each scene is
pre-processed and then classified. Filed validation is carried out for the scene of year
2000. Post-classification analysis is performed to analyze the magnitude, rate and pattern

of land use/land cover change.

3.3.1 Image Pre-Processing

Each TM imagery is co-registered (with RMS values less than 0.8 pixels) to a Universal
Transverso Mercator (UTM) map projection, and geo-referenced to north-up by using the
four comer points provided in the DDR as Ground Control Points (GCPs) to define the
linear transformation. Image rectification is performed using nearest neighbor resampling
technique to minimize convolution of the data. An image enhancement step (contrast
manipulation) is performed to increase the visual distinction between features in a scene
in order to increase the amount of information that can be visually interpreted from the

data (Ahem and Sirois, 1989).

81

False color composites are generated displaying bands 4, 3 and 2 as red, green, and blue
respectively (see figure 3.8). In this color combination, old urban build-up areas
(buildings, pavement etc.) appear with dark bluish tones, while new urban areas are much
brighter from all the reflective pavement and roofs. Pre-construction land appeared to be
medium gray-green, while construction land appears to be totally bright because of the
high reflectiveness of bare soil. Rural build-up areas have very similar reflectance like
that of urban build-up areas, but with very different spatial patterns that are easily
distinguishable fiom Landsat images. Rural villages are usually small clusters of single-
family buildings among agricultural fields, while urban residential areas rarely have
single-family buildings except in few recent exceptional westem-style suburban areas.
Photosynthesizing vegetation always adds a red tint on false colored TM images. The red
component corresponds to variation in vegetation, including agricultural land and
greenness area. There are two major types of agricultural area: crop cultivation land and
horticultural gardens. Cultivation land has more seasonal variation than horticultural
gardens. The April 1990 image was taken between harvest and the grth of the next
crop so there were less red tones on the crop fields, while the August 1998 image was
taken during the growing season, and it showed healthy crops with reddish reflectance.
Forestry and parks have the most intense vegetation, therefore appear bright red. For the
class of water, most water appeared to blue, while some others appears almost black

because the scatter less light back to the Landsat sensor.

82

   

Agricultural Land

 

Forest Area Water

Figure 3.8: False Color Composition of Different Land Types in Shanghai, Landsat TM

3.3.2 Classification and Editing

There are two basic methods for multi-temporal satellite data analysis: a pixel-to-pixel
combination of multi-temporal scenes without classifying individual scenes, or a
comparison of individually classified scenes of different dates (Singh, 1989; Yeh and Li,
1997). This project uses the post-classification comparison method because it can
facilitate the generation of land transition matrices and provide spatially explicit
information on “from—to” class changes. Each scene is individually classified, using

spectral image classification to aggregate the spectral response for different objects

83

within a pixel into three thematic features of land: build-up areas, agriculture land and

water/wetland.

A hybrid classification methodology is adopted by using an iterative, self-organizing,
unsupervised (ISO) clustering technique along with a knowledge-based classifier (Skole
et a1, 1998). For the ISO analysis, a new band is created by calculating the ratio of band 4
(NIR) to band 2 (RED), which expresses the increasing difference between red and near-
infrared reflectance with increasing green leaf area index. The unsupervised classification
uses only band 3 and this new band NIR/RED. An ISO algorithm is applied to assign the
45 output clusters. The statistical mean spectral values of clusters are shifted to locate the
actual spectral statistics for each cluster. This step is repeated until 95% of the pixels are
not reassigned to new clusters during the iterations that shifi the cluster means.
Assignment of individual pixels to a cluster is accomplished using a minimum spectral
distance decision rule such that all pixels are assigned to one of the 45 output clusters.
These output clusters are then recoded into a new classification containing three classes
by overlaying the 45 clusters on the original imagery to aggregate the 45 spectral clusters
into final land use/cover classes. The output classification is converted from ERDAS file
format to an ARC/INFO raster format (GRID file). The GRID file is then converted to
vector format in ARC/INFO. Polygons that are misclassified are identified using visual
interpretation and recoded into GIS. Vector coverages are plotted and checked for further

editing. These plotting and editing steps are repeated until the classification is completed.

84

Post-classification look-up tables are constructed for the magnitude of different land
types at different dates, and for the rate of different type of land use/land cover changes
over time. Urban growth pattern change over space and time are also analyzed. The
structure of land use, according to the land use/land cover classification based on Landsat

TM images from 1987 — 2000 is shown on table 3.1.

 

 

 

 

 

Table 3.1 Land Use/Land Cover Types for Shanghai Municipality, 1978 - 2000
Year Land Build-Up Area Agro Land Forest Water/Wetland
1987 Area (km’) 1625 3701 589 273

% (26.3) (59.8) (9.5) (4.4)
1990 Area (kmz) 1799 3572 600 195

% (29.1) (57.7) (9.7) (3.2)
1998 Area (kmz) 2393 3047 630 119

% (38.7) (49.2) (10.2) (1.9)
2000 Area (ka) 2570 2864 638 118

% (41.5) (46.3) (10.3) (1.9)

 

 

 

 

 

 

Comparison of Shanghai’s land use from 1987 to 2000 demonstrates that build-up area
grew from 1625 km2 to 2570 kmz, and build-up area percentage increased from 26.3% to
41.5%. In the meanwhile, there was 837 km2 of agricultural land being converted to urban
uses, accounting for 22.6% of the total agricultural land area in 1987. Agricultural land
percentage decreased from 59.8% to 46.3%. Forested area experienced a slight increase
from 9.5% to 10.3%, and water/wetland area decreased from 4.4% to 1.9%. By the year

2000, the region’s overall urban land held a percentage of 41 .5% while the percentage for

85

agricultural land, forested area and water/wetland was 46.3%, 10.2% and 1.9%
respectively. Table 3.2 illustrates the spatial variation of land use distribution for the
three rings of Shanghai municipality. The urban core area has a build-up percentage of as
high as 89%, while the near suburbs and far suburbs have 68.9% and 31.1% of land under
urban uses respectively. In contrast, the percentage of agricultural land is the highest in
far suburbs where 56.4% of land is under agricultural use, while in the urban core area

there is no land remains in agricultural use.

Table 3.2 Spatial Variation of Land Use Distribution

 

 

 

 

 

 

 

 

 

 

Urban Year Build-Up Area Agro Land Forest Water
Rings
Urban 1987 82.8 10.5 5.0 1.7
Core 2000 89.0 0.0 11.0 0.7
Near 1987 30.1 62.0 6.1 1.8
Suburbs 2000 68.9 19.7 9.5 1 .9
Far 1987 22.7 61.2 10.8 5.3
Suburbs 2000 31 .1 56.4 10.5 2.0
Unit: %

3.3.3 Field Validation

In order to ensure accurate classification, ground truthing is carried out on the study sites
using GPS. Because satellite system can typically acquire voluminous amounts of data
that can be spatially rectified and processed, characterizing the temporal dynamics of

large areas may be easier with remote sensing than through traditional field approaches.

86

Fieldwork is often necessary, however, especially when new or different relationships

must be established between the remote sensing data and field conditions. After such

relationships being established, much fieldwork can be eliminated and restricted to a

spot- checking of selected sites (Wood and Skole, 1998). A random stratified sampling

procedure is used to select points for classification accuracy assessment, with 50 points

for each class of build-up area and agricultural land, and 25 points for each class of forest

and water/wetland. The overall classification accuracy was 82%. Table 3.3 shows the l-
1

results for accuracy assessment.

 

 

 

 

 

 

 

 

Table 3.3 Accuracy Assessment for Land Use/Land Cover Classification
Reference Classification Row %
Build-Up Agro Forest Water Total Correct
Bufld‘UP 42 3 1 4 50 34
Ag") 2 41 5 2 50 82
Forest 2 4 19 0 25 76
Water 2 1 1 21 2 5 84

 

 

 

 

 

 

3.4 LUCC Modeling Method and Data

The modeling method in this project is a two-step Principal Component Regression
(PCR), which basically consists of Principal Component Analysis (PCA) followed by
Multivariate Linear Regression (MLR). Eigenvectors and component scores of the

explanatory variables are calculated in the Principal Component Analysis step. Then the

87

scores are regressed against the land use data using Multivariate Linear Regression

method.

A cross-sectional panel data at county/district level for the years of 1987, 1990 and 1998
are used for model parameter estimates. Table 3.4 lists the variables being used in the
model, including land use data derived from remote sensing satellite images and
demographic and socio-economic data from the Yearly Statistic Book for Shanghai,
County-level Statistic Yearbook for China as well as the National Population Survey

(July 1, 1990).

88

 

 

Table 3.4 List of Variables Used in Principle Component Regression
Variable Name Description Unit

Land Use Types -
Urban Ratio of build-up areas, including land for settlement,

industry and transportation -
AgroLand Ratio of agricultural land areas, including land for

cultivation, horticulture other agricultural uses -
Forest Ratio of forested land, including woodland, economic

forest, wind-protection forest and forestry parks -
Water Ratio of water bodies and wetland, including lakes,

rivers, beaches, reservoirs and marshlands

Explanatory Variables

PopDens
UrbanPop
PopGrowth
RuralActive
AgroLab
RurInduLab
InduOPT
RurInduOPT
GrainProd
RuralOPT
Machinery
Forestry

Fishery

Population density 10000 person/km2
Urban population percentage -
Population growth rate 1/ 10000

Active rural labor percentage

Rural labor in agricultural production

Rural labor in industry production

Urban industry output percentage

Rural Industry output as percentage of rural output
Grain production as percentage of agricultural output
Rural output as percentage of GNP

Percentage of land ploughed by machine

Percentage of forestry production in rural economy

Percentage of Fishery production in rural economy

 

89

Principal Component Analysis (PCA) involves a mathematical procedure that transforms
a number of possibly correlated variables into a smaller group of uncorrelated variables
called principal components. The objectives of performing PCA are to reduce the
multilinearity and dimensionality of the data set, and to identify new meaningful
underlying variables for land use changes. The factor scores of thee components being

extracted from PCA are calculated using the following equation:

Snxl = menTmel (1)

Where S is an n x 1 matrix of score values for all of the explanatory variables, F T is the
transpose of the m x 7: factor matrix F, while X is an m x 1 matrix of the explanatory
variables. The dimensions of the matrices are m representing the total number of
explanatory variable included in the PCA (m = 13), and n being the number of

components extracted from PCA (n = 3).

After obtaining the score matrix S from the PCA analysis, a Multivariate Linear
Regression (MLR) step is performed to regress the land use data on the PCA scores. The

model equation is therefore:
kaI =Akxl +kanSnxl +Ekxi (2)

where L is a k x I land use matrix where the dimension k is the number of land use types
under study( in this study k = 3, including build-up areas, agricultural land and forested

area). S is an n X] matrix of the scores from the PCA model, A is a k X 1 matrix of

90

regression intercepts, B is a k xn matrix of the regression coefficients, and E is a k x 1
matrix of estimate errors. The A intercept matrix and B coefficients matrix are solved by

Ordinary Least Square (OLS) regression.

Finally, by combining the MLR regression equation (2) with the PCA scores equation

(1), the final PCR model equation emerges:
1...... = A 1...] + [B a. (menfi Xma + E... (3)

where L is a k x 1 land use matrix, A is a k x 1 matrix of regression intercepts, B is an k
xn matrix of the regression coefficients, F T is the transpose of a m xn factor matrix, X is
a m x 1 matrix of explanatory variables, and E is a k x 1 matrix of estimate errors. If we

put the resulted matrix of [B kxn (men)T] as C kxm:
c a... = B 1... (may (4)

Then equation 3 becomes an equation like that of the ordinary least square regression:
kal =Akxl +Ckxm mei +Ekxl (5)

Thus the Principal Components Regression combines Principal Component Analysis and

Ordinary Least Squares Regression to estimate the matrices of A m and C km

3.5 LUCC Modeling Results
3.5.1 PCA Results

All the thirteen explanatory variables are included in the PCA analysis. Three

components with eigenvalues larger than one are extracted (Table 3.5). The first

91

component includes six variables: population density, urban population percentage,
population growth rate, urban industry output percentage, rural agricultural output
percentage as well as machinery plough level. This component accounts for as much as
45% of the variability in the data set. It can be referred to as the urbanization component
because all the variables are highly correlated with urbanization level. Population density,
urban population, urban industrial output and machinery plough level are positively
correlated with urbanization level, while population growth rate and rural agricultural

output have negative correlations.

The second component accounts for an additional 18% of the total variance of the
explanatory variables. This is a component indicating the level of rural agricultural
production. It consists rural active labor, rural agricultural labor, grain production, rural
industrial labor and rural industrial output percentage, where rural active labor, rural
agricultural labor, grain production are variables positively correlated with rural
agricultural production, while rural industrial labor and rural industrial output have
negative correlations. Spill-out effects of urban industry lead to the development of
industry in rural areas, driving rural labors to abandon their land to work for Town and
Village Enterprises (TVEs). Regions with higher output of rural industry tend to have
lower agricultural production and higher pressure for agricultural land to be converted to

other uses.

The third component, including forestry and fishery productions explains another 14% of

the total variance. Forestry and fishery are two important forms of rural economy in

92

Shanghai that are heavily dependent on natural resource endowment. Therefore this

component will be addressed as natural resource-based rural economy.

Table 3.5 Component Matrix Extracted by Principal Component Analysis

 

 

 

Component
Variables 1 2 3
PopDens .801 -. 150 -.394
UrbanPop .936 -.067 -.035
PopGrowth -.640 .486 -.647
RuralActive .053 .814 -.269
AgroLab -.115 .841 .258
RurInduLab .337 -.823 .054
InduOPT .841 -.389 -.001
RurInduOPT .468 -.777 -.l 1 1
GrainProd -.464 .479 -.487
RuralOPT -.865 .259 -.257
Machinery .578 -.01 1 .539
Forestry -.063 -.042 .494
Fishery -.173 .212 .836

 

 

 

 

Note: Extraction Method-Principal Component Analysis. Rotation Method-Varimax with
Kaiser Normalization. Rotation converged in 4 iterations

93

3.5.2 MLR Results

Remotely sensed land use data are used to regress against the factor scores of from the

three components extracted from the principal component analysis. The results are

presented in table 3.6. All the regression models have high overall significance level, and

the explanatory powers for the land use models are reasonably good. The R2 is 0.703 for

the urban land model, 0.646 for the agricultural land model and 0.447 for the forest land

 

 

 

 

 

model.
Table 3.6 Multivariate Linear Regression Results
Dependent Coefficient Independent Variables
J 2 & Constant (Component Scores)
Variable R
Significance 1 2 3
B .326 .1030 -.0678 -.0372
Urban '703 Std. 0 - .670 -.442 -.243
Sig. .000 .000 .001 .044
B .554 -.0963 .0449 .0203
“mun" “6 Std. B - -.716 .333 .151
Sig. .000 .000 .013 .236
B .095 -.0001 .0140 .0109
“"3“ “47 Std. B - -030 .527 .410
Sig. .000 .848 .002 .015

 

 

 

 

 

 

 

Note: 13 is the unstandardized regression coefficient from MLR while Std. B is the
standardized coefficient. Sig. indicates regression Significance.

94

For the urban land use model, all three components are significant at the level of 5%. The
urbanization level (component one) has a positive partial coefficient that is highly
significant, which shows that urbanization level contributes directly to the expansion of
urban areas as discussed throughout the first part of this study. The regression
coefficients for component two (development of agriculture) and component three
(development of other natural resource based rural economy) are both negative, which
means that areas with more traditional rural production (such as cultivation, fishery and

forestry) and less rural industry tend to have less urbanized area.

In the agricultural land use model, component one is negatively related with agricultural
land ratio and the coefficient is statistically Significant. This supports the hypothesis that
high urbanization level will be accompanied by lower ratio of agricultural land area. Both
component two and three have a positive relationship with agricultural land use, but the
coefficient for component two is significant at the level of 1.3% while the coefficient for
component three is not. Thus, agricultural cultivation and rural industry development
contribute to variation in agricultural land area, but other rural economy forms such as
forestry and fishery do not have a Significant statistical relationship with agricultural land

area.

For forest area, component one is negatively correlated while component two and three

have a positive relationship. But the probability of Significance for component one is as

high as 0.848, so urbanization level is not a statistically significant contributor to forest

95

area. But both components two and three are significant, which means rural areas tend to

have more land covered by forest areas.

3.5.3 PCR Results
The final step is to combine the MLR regression equation with the PCA component
matrix to estimate the coefiicients of each explanatory variable. The resulted estimates

are listed in table 3.7.

(l) Intercept

The constant terms for urban land, agricultural land and forest area are 0.326, 0.54 and
0.095 respectively. If all the explanatory variables are set to be zero, agriculture will take
up a majority of 54% of the total area, urban will account for 32.6% of the land and forest

area will be 9.5%.

(2) Component one

There are three variables (population density, urban population percentage and urban
industrial output) in component one that contribute positively to urban build-up areas, but
are negatively related with agricultural land and forest area. For example, one unit of
growth in population density would lead to an increase of 0.087 for the ratio of urban
build-up areas, in the meanwhile, a decrease of 0.078 units of agricultural land and 0.003
units of forest area. All other explanatory variables can be interpreted in the same way.
Rural output percentage and population growth rate are two variables that contribute

negatively to urban land but positively to agricultural and forest area. More rural

96

production implies more agricultural land area and more forest area but less build-up
area, and region with higher population growth rate tend to have lower urban area ratio.
Machinery plough level is the only variable in component one that has a positive
relationship with urban land and forest ratio, but a negative relationship with agricultural

area.

(3) Component Two

Rural active labor ratio, agricultural labor ratio and grain production are three variables in
component two that contribute negatively to urban area, but positively to agricultural and
forest area. On the contrary, rural industrial labor and output are two variables that
contribute positively to urban build-up land but negatively to agricultural and forest area.
Therefore, regions with more agricultural activities maintain a higher percentage of
agricultural and forest area, yet regions with more rural industry development have more

land being converted into build-up areas.

(4) Component Three

Both the variables (fishery and forestry) in component three contribute negatively to
urban land ratio, but positively to both agricultural and forest land. This is because
fishery and forestry are only important in the far suburbs where urbanization level is
much lower than the urban core area. The growth of these lines of agricultural activities

tends to prevent agricultural land and forest area from being encroached by urban uses.

97

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 3.7 Parameter Estimates of LUCC Model Using PCR Regression
Model LAND Intercept
1 Urban 0.326
2 Agro 0.554
3 Forest 0.095
Component One
Model LAND
PopDens UrbanPop PopGrowth InduOPT RuralOPT
1 Urban 0.087 0.102 -0.104 0.1 13 -0.097
2 Agro -0.078 -0.094 0.079 -0.099 0.090
3 Forest -0.003 -0.002 0.014 -0.006 0.002
Component Two
Model LAND
RuralActiv AgroLab GrainProd RurInduLab RurInduOPT
1 Urban -0.040 -0.079 -0.062 0.089 0.105
2 Agro 0.026 0.054 0.056 -0.068 -0.082
3 Forest 0.008 0.015 0.002 -0.011 -0.012
Component Three
Model LAND
Forestry Fishery
1 Urban -0.022 —0.063
2 Agro 0.014 0.043
3 Forest 0.005 0.012

 

 

98

The statistical relationships Show that urbanization (in both urban and rural areas) takes
place at the cost of agricultural land and forest area, but agricultural activities (including
traditional cultivation, fishery and forestry) tend to preserve agricultural and forest area.
A comparison of the coefficients for agricultural land model and forest area model shows
that the coefficients for agricultural land are generally larger than that of the forest area,

which means urban sprawl will be more of a threat to agricultural land than to forest area.

One limitation of the Principal Component Regression model, though, is the fact that it
cannot calculate the statistical significance of each individual variable like in Multivariate
Linear Regression. Therefore it is impossible to determine the statistical significance for

individual explanatory variables.

3.5.4. Land Use Projection

The models are used to estimate land use ratios of different types using statistical data of
year 2000. The results are compared with observed land use data of the same year from
satellite image classification for model validation purpose. Trend analysis of time series
demographic and socio-economic data enables the forecasting of the explanatory
variables for year 2025, which are then used to project land use types for year 2025.

Tables 3.8.1 — 3.8.3 present the model validation and projection results.

(1) Urban land
In 2000, the overall urban land ratio for Shanghai was 41.5% while the model estimate is

43.8%, with a marginal estimate error of 2.3%. The estimate error percentage is the

99

lowest in the near-suburb areas, where there is generally a high agreement of more than
99% between the model estimates and remote sensing observations. The model tends to
underestimate the urban area by 4.4% in the central city, and overestimate by 4.4% in the
far-suburb areas. The error percentage at the county/district level ranges from 0.0% in

the central city to 6.7% in Fengxian county.

According to the urban land use model projection, build-up areas will take over an "
additional 10% of the land area in Shanghai and lead a majority of 51.5% by year 2025. E”
While all the counties and districts will experience a positive grth of urban area, the

growth is not evenly distributed over space. The urban grth rate between 2000-2025 is

4.1% for the central city, 7.0% for the near suburbs and 11.2% for the far suburbs.

Therefore the urban are center will gradually shift away from the old city in the next two

decades. Rural industry development are penetrating into far suburb areas, bringing more

urban and industrial activities into the remote regions.

A closer examination at the counties/district level shows spatial variations of the urban
area growth rate, ranging from 2.8% in the Pudong New Area and 16.6% in Songjiang
county. Counties with the highest urban land grth rate (Songjiang, Jinshan and
Chongming) are all located in the far suburbs of Shanghai Municipality. Jiading is the
only near suburb county where the urban land area will grow by more than 10% during

2000-2025.

100

Table 3.8.1 Model Results Validation and Projection for Urban Build-Up Area

 

 

 

 

 

 

 

 

 

Zones County/ 2000 2000 Estimate 2025 Urban Am
Growth%
District Observed Estimated Error % Estimated (2000—2025)
Center Central City 89 84.6 -4.4 93.1 4.1
Minhang 83 83 0.0 90.4 7.4
Near Jiading 50.3 53.7 3.4 61.5 11.2
Suburbs Pudong 72.5 67.6 -49 75.3 2.8
Sub-total 68.9 68.3 -0.6 75.9 7.0
Jinshan 27.1 27.8 0.7 42.4 15.3
Nanhui 43.5 45.4 1.9 52.1 8.6
F“ Fengxian 37.6 44.3 6.7 46.1 8.5
“mm" Songjiang 33.8 37.9 4.1 50.4 16.6
Qingpu 35.6 41.5 5.9 40.7 5.1
Chongming 19.5 20.8 1.3 32 12.5
Sub-total 31.1 35.5 4.4 423 11.2
Shanghai TOTAL 41.5 43.8 2.3 51.5 10.0

 

 

 

(2) Agricultural land

The agricultural land model estimates the overall percentage of agricultural land area in
Shanghai to be 46.3% in 2000, approximately the same as what was detected from
satellite images (46.2%). A comparison of model estimates with remotely sensed land use

data at the county/district level Show a general overestimate of agricultural land in the

101

central city and near suburbs(except for the county of Jiading), and an underestimate in

the far suburbs.

Model results also Show that agricultural land percentage will fall to 39% by year 2025.
Therefore, an additional 16% of agricultural land will be lost within next 25 years. In
terms of land area, this means a loss of 456 km2 of agricultural land, equivalent to 2.3
times of the central city area. Spatial distribution of agricultural land use change Shows
that only the central city and Pudong New District will have a Slight increase in
agricultural area, all other counties and districts will lose part of their agricultural land
stock. Songjiang, Jiading and Chongming are the counties that will have the highest rate
of agricultural land loss. The far suburbs will lose 10% of the agricultural land. Again,
this agrees with the trend that urbanization is slowing down in the central city, and

suburbanization is spreading from near suburbs to far suburb areas.

102

tr:- mam-5|

Table 3.8.2 Model Results Validation and Projection for Agricultural Land

 

 

 

 

 

 

 

 

Zones County/ 2000 2000 Estimate 2025 Agro Land
Growth %
District -0 -M Error % -M (2000-2025)
Center Central City 0.4 8.6 8.2 1.8 1.4
Minhang 5 8.8 3.8 4.6 -0.4
Near Jiading 40.5 37.4 -3.1 29.8 -10.7
Suburbs Pudong 14.6 24.6 10.0 17.7 3.1
Sub-total 19.7 23.4 3.7 17.2 -2.5
Jinshan 54.9 51.5 -3.4 47.1 -7.8
Nanhui 46 45.3 -0.7 38.5 -7.5
Far Fengxian 51.3 46.2 -5.1 43.6 -7.7
Suburbs Songjiang 54.2 53 -l.2 40.1 -14.1
Qingpu 56.2 48.5 -7.7 47.8 -8.4
Chongming 66.5 66 -0.5 56.6 -9.9
Sub-total 56.4 53.7 -2.7 47.2 -9.2
Shanghai TOTAL 46.3 46.2 -0.1 38.9 -7.4

 

 

 

(3) Forest area

The observed forest coverage in Shanghai was 10.3% in 2000, while the model estimates
a similar percentage of 10.0%. Despite the Slight increase of forest coverage from 9.5%
in 1978 to 10.3% in 2000, the forest model shows that Shanghai’s total land under the
Class of forest will fall to 9.6% by 2025. The distribution of forest area growth’has more

spatial variations than urban or agricultural land change. The forest coverage in the

103

central city will decrease by 5.8% percent. In the near suburbs, Minghang and Pudong
Near Area will also have a decrease of forest coverage rate, while Jiading will experience
a slight increase. In the far suburbs, Jinshan and Chongming are the two counties that will
have less land under forest cover, while Nanhui, Fengxian, Qingshan and Qingpu will

have more forest area.

Table 3.8.3 Model Results Validation and Projection for Forest Area

 

 

 

 

 

 

 

 

Zones County/ 2000 2000 Estimate 2025 Forest Area
Growth %
District -0 -M Error % -M (2000-2025)
Center Central City 11 6.8 -4.2 5.2 -5.8
'Minhang 10 8.2 -1.8 5.1 -4.9
Near Jiading 7.5 8.9 1.4 8.7 1.2
Suburbs Pudong 10.9 7.8 -3.1 6.9 -4.0
Sub-total 9.5 8.3 -1.2 6.9 -2.6
Jinshan 14 10.7 -3.3 10.5 -3.5
Nanhui 8 9.3 1.3 9.4 1.4
Far Fengxian 9 9.6 0.6 10.3 1.3
Suburbs Songjiang 9 9.1 0.1 9.5 0.5
Qingpu 8 10.1 2.1 11.5 3.5
Chongming 13 13.1 0.1 11.4 -1 .6
Sub-total 10 10.7 0.7 10.6 0.6
Shanghai TOTAL 10.3 10 -0.3 9.6 -0.7

 

 

 

104

3.6 Conclusions

This project demonstrates an application of the remote sensing technology to the study of
land use dynamics related to human socio-economic activities. Temporal remote sensing
data are used to characterize urban land use Change. Since underreporting of agricultural
land area has been a long tradition in Chinese statistics as the consequence of linkage
between the statistical system, the tax system and the production quota (Crook, 1991;
Verburg et a1, 1999), remote sensing provides a more objective tool for measuring the
magnitude and rate of land use change. It also provides a strong visual portrayal of
recognized growth patterns, and dramatically conveys how the progress of modern

urbanization results in profound changes to the landscape (Ward et al, 2000).

Land use models are developed using Principal Component Regression modeling
method. This is a method that combines the advantages of principal component analysis
and multivariate linear regression to explore the relationship between land use changes
and the driving forces. Remotely sensed data on urban extent, agricultural land and forest
area along with statistical data on urban and rural population growth and economic
development are used to calibrate the land use models. Model validation shows high

agreements between the model-estimated and remotely sensed land use data.

The land use models are used to project future land use changes and the regional pattern
0f urbanization. Modeling results Show urban build-up areas will keep eating up a large
aInount of land, mostly agricultural land, in the next twenty-five years. The growth of the

central city has stepped into an old age. Market economies as well as a deeper recognition

105

of environmental protection have urged factories to leave urban areas through land
exchange, promoting the decentralization of urbanization. The booming of rural industry
in the suburb areas further disperses the region’s urban growth. As the result, the frontier
of urban sprawl is advancing deeper into the far suburb areas. As the process advances,
the agricultural land loss problem that has perplexed Shanghai for many years is expected
to deteriorate. Therefore, it is of great importance for the municipality government to
strengthen its land use regulation and planning in order to protect its limited agricultural

land resources.

Despite the government’s propaganda on afforestation efforts, the increase for forest
coverage in the region was very modest during 1978 — 2000. The forest model projects a
slight decrease of forest area ratio during the next 25 years. In order to alleviate the
environment problem and achieve sustainable development in the region, more intensive

measures for afforestation work have to be taken immediately.

Land use/land cover changes are results from multidimensional interactions of
biophysical and socioeconomic processes. The metropolitan area of Shanghai is very
plain with little Spatial variation in the terrain and soil quality. Therefore none of the
biophysical variables are included in the modeling framework of this project.
Transportation is has long been known as an important driving force of land use change.
It is a crucial variable that changes the concentric shape of land use zones and facilitates
the process of urban sprawl. However, little information about transportation

development at the county/district level is available for this study, so it is not included in

106

any of the modeling component. A more thorough analysis of model performance is

highly recommended as soon as more transportation information becomes available.

3.7 References

Ahem, F. J ., and Sirois J ., 1989. Reflectance Enhancements for the Thematic Mapper An
Efficient Way to Produce Images of Consistently High Quality. In: Photogrammetric
Engineering and Remote Sensing, 55(1):p61-67.

Ash, R. F. and Richard LE. 1998. “China’s Land Resources, Environment and
Agricultural Production.” The China Quarterly 156(December):p836-879.

Atash, F. and Wang, X.H., 1990. Satellite Town Development in Shanghai, China: An
Overview. The Journal of Architectural and Planning Research, Vol.7, p245-257.

Berry, B.J.L. 1990. Urbanisation. In W. B. Meyer and B. L. Turner 11, eds. Changes in
Land Use and Land Cover: A Global Perspective. Cambridge University Press,
Cambridge, p103-120.

Brown, GP, 1995. Arable Land Loss in Rural China: Policy and Implementation in
Jiangsu Provincel Asian Survey, 35(10), p922-944.

Chang, SD, 1963. The historical trend of Chinese Urbanization. Annals of the
Association of American Geographers, Vol.53, p109-143.

Chen, S., Zeng, S. and Xie, G, 2000. Remote Sensing and GIS for Urban Growth
Analysis in China. Photogrammetric Engineering & Remote Sensing, Vol 66, No. 5, May
2000, p293 — 598.

Douglas, 1. 1994. Human Settlements. In: Meyer, W.B. and Turner II, B.L. (eds).
Changes in Land Use and Land Cover: A Global Perspective. Cambridge University
Press, Cambridge, p149-170.

Crook, F,W, 1991. China: Agriculture and Trade Report. Washington, DC: US
Department of Agriculture Research Service RS-9l-3.

Crook, F.W., 1993. Under-reporting of China’s Cultivated Area: Implications for World

Agricultural Trade. In China: International agriculture and trade report. Washington, DC:
US Department of Agriculture Research Service RS-93-4.

107

Dai, C., Tang, L., Jiang, P. and Lin, J ., 1996. Satellite Remote Sensing for Monitoring
Urban Expansion and Environmental Remediation. Population Dynamics and
Management of Urban Environment, the 47th IAF Congress, Beijing, p9-16.

Fung, KL, 1982. Urban Sprawl in China: Some Causative Factors. In Ma, L.J.C. and
Hanten, B.W. (Editors), Urban Development in Modern China. Boulder, CO: Westview,
pl94-221.

Guildin, E. G., 1997. “Desakotas and Beyond: Urbanization in Southern China.” In G.E.
Guildin, ed., Farewell to Peasant China: Rural Urbanization and Social Change in the
late 20’” Century. New York: ME. Sharpe, p. 47-69.

Heilig, G., 1997. Anthropogenic Factors in Land-Use Change in China. Population and
Development Review, 23(1):p139-168.

HASA, 1999. Monitoring Urban Growth/Change in the Shanghai Region with Earth
Obsrevation. Project proposal.

Johnson, LC, 1995. Shanghai: from Market Town to Treaty Port, 1704-1858. Stanford,
CA: Stanford University Press.

Li, X., 1997. Integration of Remote Sensing and GIS for Monitoring and Evaluating
Agricultural Land Loss: a case study in Dongguan, Acta Geographica Sinica, 52(2):p279-
287.

Mei, A., 1992. Multi-temporal Aerial Remote Sensing Dynamic Research on the Zonal
Urbanization Process of Zhenru Town, Shanghai City. Journal of East China Normal
University, Special Issue on Remote Sensing, pp84-92.

Murphey, R., 1953. Shanghai: Key to Modern China. Cambridge, MA: Harvard
University Press.

Ning, Y.M. and Yan, Z.M., 1995. Changing Industrial and Spatial Structure in Shanghai.
Urban Geography, Vol.16, p577—594.

Ping, X., 1997. The Primary Research on Cultivated Land Change and Its Impacts on
Grain Production in China. Paper Presented at the IGU-LUCC conference, Brisbane,

1997.

Singh, A. 1989. Digital change detection techniques using remotely-sensed data.
International Journal of Remote Sensing 10: p989-1003.

Skole, D.L.; Salas, WA. and Silapathong, C. 1998. Interannual variation in the terrestrial
carbon cycle: significance of Asian tropical forest conversion to imbalances in the global
carbon budget. From “Asian Change in the Context of Global Change”, J .N. Galloway
and J .M. Melillo (editors). Cambridge University Press, 1998.

108

Smil, 1995. Who Will Feed China? China Quarterly, 143, p801-813.

Tjallingii, S. P., 2000. Ecology on the edge: Landscape and ecology between town and
country. Landscape and Urban Planning, 48(3-4): p103-l 19.

United Press International, 2000. Wildlife species threatened by urban sprawl in China.
Tuesday, November 28, 2000.

Verburg, P.H.; Koning, G.H,J.; Kok, K; Veldkamp, Aland Bouma, J ., 1999. A Spatial
Explicit Allocation Procedure for Modeling the Pattern of Land Use Change Based upon
Actual Land Use. Ecological Modeling, 116: p45-6l.

Verburg, P.H., Veldkamp, A. and Fresco, L.O., 1999. Simulation of Changes in the
Spatial Pattern of Land Use Change in China. Applied Geography 19 (1999), p211-233.

UNESCAP, 2001. Modalities for Environmental Assessment: Urban Development and
Environment Protection in Shanghai (Working Paper).

Ward, D., Phinn, SR. and Murray, A.T., 2000. Monitoring Growth in Rapidly
Urbanizing Areas Using Remotely Sensed Data. Professional Geographer, Vol.52, No.3,
p371-386.

Wood, CH. and Skole, D. 1998. Linking Satellite, Census and Survey Data to Study
Deforestation in the Brazilian Amazon. From: Pixel and Pixel: Linking Remote Sensing
and Social Science, Diana Liverman, Emilio Moran, Ronald Rinfuss, and Paul C. Stem
(eds); National Academy Press, Washing DC.

WRI, 1997. World Resources: A Guide to the Global Environment —- Urban Environment.
World Resources Institute, Washington, DC.

Yeh, AG and Li, X., 1997. An Integrated Remote Sensing and GIS approach in the
Monitoring and Evaluation of Rapid Urban Growth for Sustainable Development in the
Pearl River Delta, China. International Planning Studies, 2(2): pl93-210.

Yeh, AG. and Li, X. 1998. Sustainable land development model for rapid growth areas
using GIS. International Journal of Geographical Information Science, 12(2): pl69-l 89,
Mar 1998.

Yeung, Y.M. and Sung, Y.W., 1996. Shanghai: Transformation and Modernization under
China’s Open Policy. Hong Kong, China: The Chinese University of Hong Kong Press,
p321-340.

Yusuf, S. and Wu, WP, 1997. The Dynamics of Urban Growth in Three Chinese Cities.
New York, NY: Oxford University Press.

109

Zhang, L. and Zhao, S.X.B., 1998. “Re-examining China’s ‘Urban’ Concept and the
Level of Urbanization.” The China Quarterly 154, p330-381.

Zhou, D., 1997. “On Rural Urbanization in China.” In G.E. Guildin, ed., Farewell to
Peasant China: Rural Urbanization and Social Change in the late 20th Century. New
York: ME. Sharpe, pl3-46.

Zhou, Y. and Ma, L.J. C., 2000. Economic Restructure and Suburbanization in China.
Urban Geography, 2000, Vol. 21, No.3, p205-236.

110

CHAPTER 4
Land Use/Land Cover Change

in the Brazilian Amazon and the State of Rondonia

The Brazilian Amazon is an agricultural frontier. It is chosen to be a representative case
to study the consequences of agricultural activities on a primarily forested landscape. The
main objective of this chapter is to understand the complex factors of human socio-
economic activities as they control processes of land use/land cover change (LUCC).
Based on this analysis, projections of future land use/land cover change in the region are
made using a process-based systematic model that links remotely sensed observations of

LUCC to the demographic and economic data.

This chapter will proceed through the following sections: the first section describes the
background and main themes of this study, starting with the environmental significance
and global attentions to land use/land cover change in the Brazilian Amazon, followed by
the importance of LUCC monitoring, process understanding as well as LUCC modeling.
The second section outlines a systematic modeling scheme, with elaborations on the
advantages of dynamic system modeling method and the importance to integrate
remotely sensed land cover information with census data. The third through fifth sections
construct a systems model of land use/land cover change by integrating land cover
change with the region’s demographic and socioeconomic structures, as well as land use
behaviors. The LUCC systematic model is comprised of a land cover change model, a

socio—economic model and a population model, with multi-interactions between each

111

modeling component. The sixth section performs sensitivity analyses to examine the
model behavior, and also to determine some critical model parameters that are subject to
high uncertainties. Section seven presents major modeling results and model validation.
The last section addresses the implication of the study and possible improvements for

future modeling work.

4.1 Introduction

Research on tropical deforestation is important because of its significant role in the global
carbon cycle, its relation to ecological complexity, and its impact on changes in
biodiversity. As the world’s largest contiguous tropical forest biome covering an area of
approximately 500 million hectares and containing 40 percent of the world's remaining
tropical rainforest, the Brazilian Amazon has come into the international spotlight over
the past two decades because of concerns for its high deforestation rate, currently at an

average of nearly 2 million hectares per year (Houghton et al, 2000).

Land use/land cover change (LUCC) is well known as an important driver of ecological
change in Amazonia, with large influence on hydrology, climate, and global
biogeochemical cycles (Houghton 1991; Crutzen and Andreae 1990; Houghton and Skole
1990, Salati and Vose 1984, Shukla et al. 1990). As the results, there has been rapid

growth in empirical social science research on LUCC in Amazon.

This study aims to understand the impacts of human land use activities on land cover

Changes and to project future land use/land cover change in Amazonia based on the

112

understanding. Three related questions are raised within this content: What changes have
happened to the land? How did these changes occur? And what will happen in the firture?
In order to address these questions, this study proposed three theme activities with the
applications of different research tools. Theme one is to monitor land cover change with
remote sensing satellite images, which offers an objective quantitative approach to
measure the magnitude and rate of land cover change. Theme two is understand the land
use/land cover change process and underlying driving forces. In a human-dominate
landscape, it is the human actions rather than natural forces that are the sources of most
contemporary change in land. Econonric and demographic theories are applied to explain
land use/land cover change processes. Theme three is to develop a land use/land cover
change model that links remotely sensed land cover change with the demographic and
socio-economic understanding of land use processes. The model is then used to project

what changes would happen to the land in the Brazilian Amazon.

4.1.1 LUCC Monitoring

The general interest in land use/land cover change is stimulated by current global change
concerns, yet empowered by the new capacity of landscape monitoring with the earth
observing systems. Satellite remote sensing has been used extensively to map the
magnitude and rates of deforestation in Amazonia (e.g. Tucker et al. 1984, Malingreau
and Tucker 1988, and Skole and Tucker 1993). These studies have been successful at
documenting deforestation rates using remote sensing analyses that measure changes in
forest versus non-forest covers. However, we should consider more than only forest

versus non-forest when considering land use/land cover change in the Amazonia. The

113

simplistic view of forest versus non-forest overshadows some interesting dynamics which

are only recently being observed.

New research in Amazonia has Shown land use/ land cover change as a highly dynamic
process of clearing, abandonment and re-clearing (Alve and Skole, 1996; Steininger,
1996; Houghton et al, 2000). It has documented high frequency of secondary vegetation
regeneration following deforestation, generally vigorous regrowth of forests on former
pastures and agricultural land. Recent studies have mapped the extent and temporal
dynamics of secondary vegetation at landscape level (e. g. Alves and Skole 1996, Mausel
et al. 1993, Skole et a1. 1994, Adams et al. 1995, Steininger 1996, Foody et al. 1996). A
satellite analysis in Rondonia by Alves and Skole (1996) showed that secondary growth
accounted for over 40% of the deforested areas. Almost 60% of the areas in secondary
growth in 1986 were re-cleared at least once by 1992, and over 55% of the deforested
areas in 1986 were abandoned into secondary grth by1992. This analysis and other
similar satellite based analyses (Brondizio et al. 1994, Moran et a1, 1994; Steininger,
1996, F oody et a1. 1996) are useful for characterizing the dynamics of the landscape
change resulting from tropical land use. By the year 1986, around 6% of the primary
forest in the Brazilian Amazon had been deforested, an area of 23.8 million ha (including
16.8 million ha of deforestation and 7.0 million ha of secondary growth). In 1992, more
than 9% of the trees were deforested, representing an area of 34.7 million hectares
(including 23.9 million hectares in deforestation and 10.8 million hectares in secondary
growth). By 1999, the overall deforested area had reached 47.3 million hectares, of which

secondary growth was 15.9 million hectares. Given the widespread vegetation succession

ll4

and its carbon sequestration potential, secondary growth has become a large and rapidly
changing carbon pool. The balance of atmospheric carbon in Amazonia depends nearly as

much on succession in the wake of disturbance as on deforestation.

4.1.2 Process Understanding

Observing the magnitudes and patterns of landscape change can provide detailed
information on “what changes have happened at when and where”, but it does not lead
much insight into the question of “why do the changes occur”. We need to move from
pattern description to process understanding, to find the mechanisms which govern the
land use/land cover process. It is very important to understanding the factors that
influence land use/land cover changes. Controlling deforestation is difficult until we

know more about the causes and processes being involved.

It is important to distinguish between two categories of causes for land cover change: the
proximate sources of changes and underlying driving forces of changes (Turner et al,
1990). Proximate sources are human activities that make use of, and hence change or
maintain, attributes of land cover dynamics, such as agricultural expansion and cattle
ranching. They are activities directly responsible for the physical changes in the land.
Large-scale deforestation in Amazonia did not begin until the early 1970s, when the
Brazilian government introduced ambitious road construction and colonization programs
intended to develop the remote Amazonia, which stimulated massive movement of

ranchers, farmers, loggers and miners into the region and caused dramatic changes in

115

land cover as results of their land use activities such as agricultural development, logging

and mining (Schmink and Wood 1992).

Underlying driving forces contribute to the changes of land indirectly by motivating and
constraining production and consumption, therefore give rise to proximate sources of
land changes. The identification of driving forces is a more complex subject than that of
the proximate causes (Lambin, 1994; Angelsen and Kaimowitz, 1999). There are many
factors, including social-economic, physical-ecological and political-institutional factors
that may influence the magnitude of the proximate source activities and help generate or
release the environmental pressures on the process of land use/land cover change.
Candidate forces driving land use/land cover changes include population growth, level of
affluence, technology development, property right regime, political economy and
government policies, international trade and debt, and attitudes and values (Turner and
Meyer 1991; Stem et al. 1992; Burgess, 1993). The challenge is to design a manageable
yet realistic model that can incorporate the effects of the key driving variables of land

use/land cover change.

4.1.3 LUCC Modeling
From pattern observation and process understanding, the next step is to build an LUCC
model, an important endeavor for projection of “what would happen in the future” which

are beyond the reach of direct observations from satellite remote sensing.

116

Little attention was paid to LUCC modeling until the late 1980s, since when various
models have been developed to answer such questions as how changes in land use
influence changes in land cover, how different configurations of economic development
and technological level influence deforestation, and what is the relative importance of
these factors in driving deforestation (Lambin, 1994). Answering these questions requires
an understanding of the mechanisms of land use/land cover change, and the design of
models that are able to incorporate both the proximate sources and the associated driving
forces. A range of LUCC models has already existed in literature. Extensive reviews can
be found in Lambin (1994), Angelsen and Kaimowitz (1999) and Agarwal and Green et

al (2000).

Modeling has become the key to project long-term trends of tropical deforestation.
However, most LUCC models concerned about the fate of tropical forest. Few of them
addressed the issue of secondary growth. Walker (1999) developed an LUCC model to
predict the fallow times of secondary regrowth as well as the areas of regrowth associated
with each regrth cohort as a function of family size. But the model was focused only
on the subsistence economy of shifting cultivators, therefore did not provide a complete

accounting for economic activities of other major land use agents in the Amazonia.

The existing LUCC models are focused either on field-level understanding of land use
change in small areas, or on basin-wide land cover projection without the detailed
accounting for land use change. Without basin-scale land use and land cover change

projection, the future of the Amazonia remains uncertain. A recent research effort led by

117

Laurance et a1 (2001) was to develop a basin-wide GIS model to project future land cover
change in the Brazilian Amazon. By integrating GIS data layers on deforestation,
logging, mining, highways and roads, navigable rivers, vulnerability to wildfires,
protected areas, and existing and planned infrastructure projects, the GIS model was used
to predict the pattern and pace of forest degradation and what road construction would do
to Brazil's rainforest over the next 20 years. Modeling results Showed strikingly extensive
deforestation even under the optimistic scenario, therefore, sounded alarms for
development plans in the Amazonia. However, like in most land use/land cover change
models to date, the driving forces in this model (population growth, industrial logging
and mining, changing Spatial patterns of deforestation and wildfire) were treated as
simple explanatory variable layers that are exogenous from outside of the land use/land
cover change process. There was no explicit internal interaction or feedback built into the
models (Lawrance et a1, 2001 ). Therefore the models did not adequately account for the

mechanism complexity of land use/land cover change.

This study adopts a systems modeling method to simulate the dynamic structure and
behavior of land use/land cover change as complex systems with interactions, feedback
loops and reciprocal effects between the proximate sources and driving forces of land
use/land cover change. It aims to model the dynamic processes of how forests become
deforested, how deforested areas grow into secondary vegetation and what is the fate of
this secondary growth. Land cover changes are largely caused by human being’s land use

practices for socio-economic purposes. In a human-dominate landscape, the influence of

118

all economic forces must be recognized for a model to be scientifically predictive

(Turner, 1990; Braat and Lierop, 1987; Parks, 1991).

4.1.4 Study Areas

Two studies were undertaken in this chapter: a region-wide analysis for the Amazonia
and a smaller-scale case study for the state of Rondonia, the region’s newest frontier (see
figure 4.1). By definition, the Brazilian Amazon include the states of Acre, Amapa, Para,
Rondonia, and Roraima, plus part of Mato Grosso, Maranhao, and Tocantins. The State
of Rondonia is located in the western portion of the Brazilian Amazon with an area of
243,044 kmz. It has developed an economy that is centered on agriculture, timber
extraction and mining (Dale et al, 1993). Rondonia is selected to be a case study site
because it has patches of soil, vegetation and climatic conditions that represent large
areas of the Amazon Basin, but it has the most intensive deforestation among all the
states in the Brazilian Amazon. It is the most recent state to undergo the transformation

from a remote intact forest to a nearly developed agricultural frontier.

 

   

\ N . I"
‘\J/_' »’%-E,

State of Rondonia Brazilian Amazon

 

 

 

4'

 

Figure 4.1: Study Areas: the Brazilian Amazon and the State of Rondonia

119

4.2 Modeling Method and Data

4.2.1 Conceptual Model

To understand the dynamics of land cover change in a particular region, we have to
regard the region with the context of its population structure, socio-economic structure
and land use structure, treating these components as a coherent system that cannot really

be understood from its separate components.

In this study, the Brazilian Amazon is assumed to be a coherent system in three parts: a
land subsystem, a population subsystem, and a socio-economic subsystem which consists
of proximate sources and driving forces (see Figure 4.2).

Import/Export

f
o

I‘-
-<. {-

 

«:7

{ Proximate
Sources

Land Use Socio—Economic Population 3:15;. ,
System System System .43....3

Driving
\ Forces

Figure 4.2: A system of LUCC with a land use system, a socio-economy system (with
sources and causes) and a population system. Interactions between the system and its
environment include both population migration and goods import/export.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Five major proximate sources of land use/land cover change will be analyzed in this

study: agricultural development (including farming and cattle ranching), logging, mining,

120

fire, and road construction. The driving forces of land use change are categorized into
variables that affect the demands for land (population growth, consumption behavior and
income) and variables that control the intensity of land exploitation (technology and soil
fertility). Population migration and agricultural product import/export are two other

variables that indicate the interactions between the system and its environment.

AS the landscape is changed by human activities through various socio-economic
processes, the land use system is highly dependent on changes taking place in the
population system and the socio—economic system in general. It is the dynamic
interactions between the proximate sources and driving forces that lead to the dynamics
of land use/land cover change process. This study adopts a systematic modeling method
to explicate these interactions and feedbacks behaviors. It develops a systematic model
of land use/land cover change based on regional economic development and demographic

dynamics.

4.2.2 Modeling Method

Traditional economic models are limited in their power to adequately explain or predict
nonlinear, dynamic complexity of the land use system in the Brazilian Amazon. Instead,
dynamic system modeling is a useful tool to simulate the dynamic structure and behavior
of complex systems with interactions, feedback loops and reciprocal effects. The concept
of dynamic feedback has a rich history in ecological modeling that dates back to the early
20th century when Volterra used now-classic differential equations to show how the

interaction between predator and prey could drive sustained oscillations (Harmon and

121

Ruth, 1997). Ecological science has had a profound influence on economists’
understanding of the concepts of system stability and the complex relationship between
economics and the environment (Perman et a1, 1998). Since the 19708, the systems
approach has been applied in economic modeling and later in integrative economic-

ecological modeling (Braat and Lierop, 1987).

A systems—based modeling approach can yield more accurate projections of changes in
land by taking into account the multidimensional interactions between humans and their
environment by clarifying how the economy operates and how the environment reacts
therefore allowing for better decision-making and policy optimization. However, the
prerequisite for developing of a dynamic systems model is that the processes and
mechanisms of land use/land cover change for a given situation are well understood
(Lambin, 1994). Building a systematic model is essentially a process of formalizing and
articulating, in a comprehensive manner, the cause/effect relationships that have been
described in prior studies. The main point of dynamic system modeling is to understand
the land use/land cover change process within a system, and to express the structure of a
system as mathematical relationships. But no single modeling approach can be expected
to elucidate the full range of socio-economic processes affecting land use/land cover
change. Statistical methods will be used as a supplement to the systematic modeling
method to parameterize the model and quantify the empirical relationship between
different variables. Simulation was developed within the STELLA modeling
environment, a graphic-based dynamic modeling method developed by High

Performance Systems, Inc.

122

4.2.3 Modeling Data

There are two major types of data in this study: remotely sensed data on land cover
change and census data on agricultural land use, demographic and other socio-economic
situations. When large areas are studied, census and statistic data are virtually the only
regional wide sources that contain extensive sets of data on socio-economic and
demographic characteristics of population (Wood and Skole, 1998). Integration of census
and statistical socio-economic data with remotely sensed land cover data have been found
useful to some modeling work regarding land use/land cover change (e.g., Reis and

Margulis, 1992; Wood and Skole, 1998; Moran et a1, 1994).

Information regarding land types is derived from LAN DSAT satellite data. TM imageries
of the Brazilian Amazon Basin for the years 1986, 1992 and 1999 are available fi'om the
Tropical Forest Information Center (www.trfic.msu.edu). An iterative, self-organizing
(ISO) unsupervised clustering technique was applied to the geo-referenced imageries
with a knowledge-based classifier. Seven thematic features of land classes are extracted
for each scene: forest, deforestation, secondary growth, water, cloud, cloud shadow and
cerrado. The classification is based on a series of thresholds used to isolate each class. In
general, visible bands are used to discriminate between forest and non-forest, whereas,
the near infrared bands are used to separate forest from early secondary growth forests.
The classification is then converted from raster to vector format and input into the project
GIS. In the GIS the vectors are checked using visual interpretation by overlaying the
vectors on the 1:250,000 scale colorfire photoproducts and on the digital data. The

vectors are edited until they match the interpreter's visual assessment. The results are then

123

registered to the tile system to facilitate data manipulation and analysis. Scenes were
edge-matched and merged to form a seamless land cover data set for the entire Brazilian
Amazon basin (see figure 4.3). In this study, water, cloud, cloud shadow and cerrado are
grouped into one category “other” to yield four categories of features (forest,

deforestation, secondary growth and other).

 

Figure 4.3: Land Cover Data in the Brazilian Amazon Based on Landsat Satellite Image
Classification for Year 1986 and 1992

The socio-economic data set was fetched in ASCII format from the anonymous FTP site
(www.2rc.utexas.edu) at the Population Research Center at the University of Texas at

Austin. It contains indicators of the social, demographic and economic characteristics of

124

the Legal Amazon and the State of Rondonia in 1980 and 1991. Data for year 1986 were
calculated by geometric interpolation. Indicators for demographic characteristics were
derived by aggregating the individual-level characteristics of the population reported in
the 25 percent sample of the 1980 demographic census of Brazil, and the 10 percent
sample of the 1991 demographic census. Data for agricultural production came from
1980 agricultural census of Brazil, 1989 agricultural production annual and 1990 cattle
and extractive production annuals of Brazil. An important value of agricultural census
reports is that they provide data not available from remote sensing, such as crop type,
production and other useful information related to land use and management. Skole et al
(1994) showed a comparability of census-derived and satellite-derived estimates of total
area deforested, suggesting an important role of agricultural census data in land use/land

cover change modeling.

The LUCC systems model consists of a land cover change model, a socio-economic
model and a population model. The next three sections will explore each model

component in more details.

4.3 Land Cover Change Model

Most land cover change models to date have been focused on the conversion between
forest and non-forest covers. However, we should consider more than only forest versus
non-forest when considering land use/land cover change in the Brazilian Amazon. There
are three types of land cover of interest in this study: forest, deforestation and secondary

growth. Forest is tropical rainforest with a closed canopy; deforestation is forest clearing

125

and conversion to agricultural and other uses while secondary grth refers to vegetation

regeneration following the reduced use of agricultural land.

Associated with these three types of land cover are four processes of change that form a
sequential loop: deforestation, agricultural land use reduction, secondary growth re-
clearing, and forest and vegetation regeneration. The land cover change model aims to
capture the dynamics of the processes of forest conversion, regeneration, and re-cycling

or re-clearing of regeneration vegetation (see Figure 4.4).

Regrth

 

l ' J‘Reclearance r
[ Forest }———>l Deforestation 4 2nd Growth l
J Deforest [

* * Secondary
Succession

 

 

 

 

 

 

 

 

Figure 4.4: Diagram of Land Cover Change Model with Four LUCC processes:
deforestation, secondary succession, forest regrowth and secondary growth reclearance.

Upon the arrival of agricultural colonists, parcels of forest are cut and burned. Crops and
pastures are grown on the cleared land until the fertility of the land is exhausted, after that
the colonists will move forward to clear forested area in new locations, while the land
with exhausted fertility will be left fallow for several years. The fallow land will grow
into secondary vegetation with critical contributions from factors such as soils, nutrients
and land use history (Salomao et al 1996; Moran et al. 1996). Most likely, secondary

grth will be recleared again for agricultural use once the fallow land regains fertility,

126

therefore will not revert back to forest (Scatena et a1, 1996). Alves and Skole (1996) used
multi-temporal Spot XS imagery for a site in Rondonia, Brazil to Show that only 42% of
the areas in secondary vegetation in 1986 remained in secondary vegetation over the
following 6 years. But if left alone, secondary growth will increase its biomass and
canopy density, and finally succeed to forest. The reflectance characteristics of secondary
growth will eventually become indistinguishable from those of mature, undisturbed forest

(Mausel et al. 1993, Boyd et al. 1996).

The land cover change model is simulated in STELLA (figure 4.5). The mathematical
expression is formulated using linear algebra in equations 1.01 — 1.05 as shown on table
4.1. Initial areas for the forest, deforestation and secondary grth pools are set at
360,308,500 hectares, 168,041,200 hectares and 7,024,400 hectares respectively in the
Brazilian Amazon basin, 20,338,140 hectares, 1,633,760 hectares and 363,310 hectares in
the state of Rondonia. These are land cover data derived from satellite images analysis of
the year 1986. The annual rate of deforestation (Deforest), abandonment
(total_abandonment) and the rate at which secondary forests are converted back to
pasture and agriculture (total_reclearance), therefore, the rate of vegetation regeneration
(regrowth_rate1) are left undefined in this part of the LUCC model. The following text
will introduce the factors and processes that control the relative magnitude of forest
clearing and agricultural land abandonment, which will finally lead to the development of

a LUCC model that can track the land use/land cover change on an annual basis.

127

  

 

  

FOREST

total abandonment total reclearanoe

Figure 4.5: A Diagram for the Land Cover Change Model. Regrowth] is a biflow between
two state variables DEF ORESTED and SECONDGROWTH. If total_abandonment is larger
than total_reclearance, the flow goes into SECONDGROWTH; else the flow goes into

DEF ORESTED.

Table 4.1 Equations for the Land Cover Change Loop

 

No.

Equation Specification

 

1.01

1.02

1.03

1.04

1.05

 

FOREST(t) = F OREST(t - dt) + (Regrowth2 - Deforest) x dt

DEF ORESTED(t) = DEF ORESTED(t - dt) + (Deforest — Regrowthl) x dt
SECONDGROWTH(t)

= SECONDGROWTH(t - dt) + (Regrowthl —- Regrowth2) x dt
Regrowth_l = total_abandonment — total_reclearance

Regrowth_2 = SECONDGROWTH x regrowth_rate2

 

128

 

4.4 Socio—Economic Model

The socio-economic model links between the land cover change model and the
population model. The influence of socio-economic forces must be recognized for a
model to be scientifically predictive in a human-dominate landscape (Turner, 1990; Braat
and Lierop, 1987; Parks, 1991; Kaimowitz et a1, 1997). Unlike most systematic models
that view land as an ecological system that will eventually produce some economic
benefits, the socio—economic model in this project will explicitly account for the socio-

economic processes as well as their impacts on land use/land cover change. Figure 4.6

 

and equation 2.01 illustrate how the magnitude of deforestation is determined by the five
major sources of land use/land cover change, including agricultural development

(farming and cattle ranching), logging, fire, mining and road construction.

Deforest

  
  

agro development road

logging mining
fire

Figure 4.6: Sources of Deforestation in the Brazilian Amazon: agricultural
development, logging, fire, mining and road construction

Table 4.2 Equations for Source Activities of Deforestation

 

No. Equation Specification

 

2.01 Defor = agro_development + logging + fire + mining + road

 

 

129

4.4.1 Importance of Agricultural Development

The major cause of deforestation in the Brazilian Amazon is agricultural development,
especially large-scale cattle ranching operations (Hecht, 1985; Moran et al, 1994; Walker
et al, 2000). Feamside (1993) attributed 70 percent of deforestation in Amazonia to cattle
ranching. Estimates of Walker et al (2000) in the State of Para supported the general view
that cattle ranching might be held accountable for a large proportion of forest clearing.
While the links between cattle ranching and tropical deforestation have been widely
acknowledged, there is a growing realization for the role of small farmers, the largest
socio-economic group who produced 80 percent of the total food production in Amazonia
(Scatena et a1, 1996). Faminow (1998) estimated that small farmers created at least half

of the deforestation area in Amazonia.

In this study, eleven agricultural products are analyzed in the agricultural developing
system, including ten types of crop cultivation (rice, sugarcane, soybean, corn, manioc,
beans, coffee, banana, cacao, black pepper) and cattle ranching. It is the conversion of
closed forest canopy to pasture and cropland that accounts for most of the landscape
dynamics (Feamside, 1993; Southgage, 1994; Moran et al, 1994; Parayil and Tong,

1998)

4.4.2 Agricultural Economic Processes

Figure 4.7 illustrates three dynamic agricultural economic processes (demand expansion,

inventory correction and expectation adjustment) which interact to generate the

130

equilibrium between the supply and demand for each agricultural product, and a Cobb-
Douglas production function that determines the amount of land needed as input to the

agricultural production system.

J

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Population

 

 

 

 

 

 

Figure 4.7: Dynamic Agricultural Economic Growth Processes (demand expansion,
inventory correction and expectation adjustment) and agricultural land demand
through a Cobb-Douglas Production Function

If the demand for agricultural products of an economy increases for reasons such as
population growth, agricultural producers will increase the production, which will lead to
increased incomes. On one hand, increased income will increase the demand for

agricultural goods. In economic terms, this is known as induced consumption that causes

131

an expansion in demand. On the other hand, increased income will increase the further
ability of agricultural supply. In macro economy, an initiative that leads to an increase in
demand will ripple through the economy. By the time the ripples die out, total expansion

in demand can greatly exceed the initial increase.

Expectation adjustment is a process that seeks to eliminate the discrepancies between
expected demand and actual demand. Agricultural producers will adjust their
expectations for agricultural demand based on the difference between expected demand

and actual demand so as to avoid oversupply.

The difference between agricultural supply and demand is agricultural inventory. An
inventory correction process is to adjust the production behaviors so as to eliminate the

discrepancies between actual and desired inventory.

A Cobb-Douglas production function links the agricultural economy system with
agricultural land uses. It describes an empirical relationship between specified inputs and
outputs. The agricultural economy is conceived as a machine that transforms the input
factors into agricultural outputs through a Cobb-Douglas function, which determines the

amount of land needed to produce a certain amount of agricultural supply.

The difference between needed agricultural land and existing agricultural land is

agricultural land discrepancy. People will alter their land use behaviors as responses to

the discrepancy level. More land will be cleared if the actual land area is short of the

132

needed land area; on the contrary, more land will be abandoned if less land is needed for

production.

4.4.3 Modeling Agricultural Land Use Processes
Figure 4.8 shows the model architecture for these components. The arrays in Figure 4.8

have a dimension of agricultural type with the eleven major agricultural products.

Equations 2.01 - 2.18 on table 4.3 are the mathematical expressions for the diagram.

 

 

  
 

v ‘I s
33 ./ .
/

agro “b“ marginal consumption I

rural population auto spending rate "'93" 909018000

Figure 4.8: Diagram of Agricultural Development and Land Use Change

133

Table 4.3 Equations of Agricultural Economic Processes
for the Socio-Economic Model

 

No. Equation Specification

 

(1) Demand expansion

2.01 Agro_Demandp = in_demandp + ex_demandp

2.02 in_demandp = auto_demandp + induced_demandp

2.03 auto__demandp = rural _pop x auto_demand_ratep

2.04 induced_demandp = incomep x marginal_consumptionp
2.05 incomep = AGRO_SUPPLYp

2.06 ex_demandp = urban_demandp + exportp

2.07 urban_demandp = urban _pop x auto_demand_ratep
2.08 exportp= Agro_supply * export__ratio

(2) Expectation adjustment

2.09 EXPECTED_DEMANDp(t)

= EXPECTED_DEMANDp(t - dt) + (Expectation_Changep) x dt
2.10 INIT EXPECTED_DEMANDp

= auto_demandp/( l - marginal_consumptionp)
2.1 l Expectation_Changep

= (AgroDemandp - EXPECTED_DEMANDP) x fractional_adjustmentp
(3) Inventory Correction

2.12 INVENTORYp(t)

134

= INVENTORYPO - dt) + (Agro_Supplyp- Agro_Demandp) x (it

 

2.13 INIT INVENTORYp = desired_inventoryp

2.14 desired_inventoryp = EXPECTED_DEMANDp x inventory_coveragep
2.15 inventory_discrepancyp = desired_inventoryp - INVENTORYp

2.16 inventory__correctionp = inventory_discrepancyp x fractional_correctionp
2.17 Agro_Supplyp = EXPECTED_DEMANDp + inventory_correctionp

(4) Cobb-Douglas Production

2.18

land_inputp = exp [(ln(Agro_Supplyp) — ln(cp) — bp x ln(agro_laborp))/ap]

(5) Agricultural Land Use Change

2.19

2.20

2.21

2.22.1

2.23.1

2.22.2

2.23.2

land_discripancyp = land_inputp — AGROLANDp

AGROLANDp(t)
= AGROLANDp(t - dt)
+ (Clearancep — Auto_abandonmentp — In_abandonmentp) x dt

Auto_abandonmentp = AGROLANDp x auto_abandon_ratep
If land_discrepancyp >= 0
Clearancep = land_discrepancyp + Auto__abandonment
In_abandonment = 0
If land_discrepancyp < 0
Clearancep = Auto__abandonmentp

In_abandonmentp = -(land_discrepancyp) - Auto_abandorimentp

135

 

2.24 New_clearancep = Clearancep x clearance_ratep

2.25 Reclearancep = Clearancep x (l-clearance_ratep)

 

Note: p is the dimension for the array of agricultural type, representing rice, sugarcane,
soybean, corn, manioc, beans, coffee, banana, cacao, black pepper and cattle (in order).
(1) Demand Expansion

Demand for each agricultural product comes from two sources: an internal demand and
an external demand (equation 2.01). The internal demand is to satisfy the needs for
farmers living within the agricultural production system, while the external demand

comes fiom outside of the system.

The volume of the internal demand is a firnction of the internal population size and
consumption behavior. It is comprised of an autonomous demand and an induced
demand (equation 2.02). Autonomous demand is independent of the consumers’ income
(equation 2.03), while induced demand is dependent on consumer’s income and marginal
propensity to consume (equation 2.04). Farmers allocate some fraction of their income to
the purchase of goods and service. Since they derive their income from agricultural
activities, their income is a function of agricultural production (equation 2.05). When
incomes change, consumers will vary their spending in the proportion to the change,
called the marginal propensity to consume. Increased agricultural supply will lead to an

increased income so as to increase the internal demand for agricultural goods.

External demand came from outside of the agricultural production system. It consists of

demand from urban residents in Amazonia and demand from people outside of the

136

region (equation 2.06). Agricultural demand from urban residents is dependent on the
size of urban population (equation 2.07). There is an active market for agriculture goods
in areas outside of the Legal Amazon area, where not much land is available for
agricultural use. The semi-open access status of the vast amount of tropical forestry in
the Amazon basin makes it more subject to be converted to agricultural use. Since no
reliable data source for agricultural commodity export exists, export ratio is assumed to

be 10% of total agricultural supply in this model (equation 2.08).

(2) Expectation Adjustment

Expectation adjustment is a process that seeks to eliminate the discrepancies between
expected demand and actual demand. The expected demand is adjusted by expectation
change (equation 2.09), with the initial expected demand set to be a function of the
autonomous demand and the marginal propensity to consume (equation 2.10). The
expectation change is the fractional adjustment times the difference between actual

agricultural demand and expected demand (equation 2.11).

(3) Inventory Adjustment

The difference between agricultural supply and demand is agricultural inventory
(equation 2.12). Initially, inventories for all agricultural products are set at their desired
inventory levels (equation 2.13). Desired inventory is the coverage fi'action of expected
demand (equation 2.14), and inventory discrepancy is the difference between desired
inventory and the actual inventory (equation 2.15). The inventory correction process is to

adjust the production behaviors so as to eliminate inventory discrepancies (equation

137

2.16). Producers will regulate their production activities to make supply conform to the
volume of demand. A negative feedback loop uses the discrepancy from equilibrium
agricultural production to control supply responses, which equals the expected

agricultural demand adjusted by inventory correction (equation 2.17).

(4) Cobb-Douglas Production Function

The economic processes of demand expansion, expectation adjustment and inventory
correction work together to decide how much agricultural products will be produced each
year in order to meet the changing demand. Then a transformed Cobb-Douglas firnction
is used to calculate how much land is needed to produce the amount of agricultural

products (equation 2.18).

For simplicity, we consider agricultural production with only two inputs: land and labor.
The agricultural economy is conceived as a machine that converts the input factors into
agricultural outputs through a Cobb-Douglas function, which is defined as the following:

Y=cf{N,L}=c xN’x Lb

Where Y is agricultural output, N is agricultural land, and L is agricultural labor. c is a
constant scaling parameter that accounts for technological progress, while the exponents
a and b are elasticity ratios that give estimates of the percentage change in agricultural
output for one percent change in the corresponding agricultural input (land and labor).
Given data on quantities of agricultural inputs and outputs, we can estimate the

parameters of the Cobb-Douglas production function by taking logarithms to transform it

138

into a linear relationship, then regressing the log of the quantity produced on the logs of

the input quantities and a constant term.

Take logs of both side of the function, the Cobb-Douglas functions is transformed into:

ln(Y) = ln(c) + a x ln(N) + b x 1n(L)

The Ordinary Least Squares (OLS) method is employed to estimate the parameters
(constant c and exponents a, b) of the Cobb-Douglas production firnction for each
agricultural product. With known agricultural output and labor, we can calculate needed

land as input to agricultural production based on the following equation:

N= exp [(ln(Y) — ln(c) — bx ln(L))/a]

(5) Agricultural Land Use Change

The difference between needed agricultural land and existing agricultural land is
agricultural land discrepancy (equation 2.19). There are three processes that change the
stock of agricultural land area: autonomous abandonment, induced abandonment as well

as clearance (equation 2.20).

Land abandonment is the combined results of both autonomous and induced
abandonment. The largest part of the Amazonian soils has a low fertility as far as
nutrients for plants are concerned. Autonomous abandonment is the land taken away
from agricultural production as the result of soil fertility exhaustion, no matter what is the

level of agricultural land discrepancy (equation 2.21). Induced abandonment and

139

clearance amount are determined by the discrepancy level of agricultural land demand.
When the land discrepancy is larger than zero, that is, when more agricultural land is in
need, more land will be cleared and added to the agricultural land stock to make up the
land deficiency cause by land discrepancy and autonomous abandonment (equation
2.22.1), and no land will be induced to be abandoned (equation 2.23.1). On the contrary,
if the discrepancy is less than zero, in another word, existing agricultural land exceeds the
needed land, people tend to reduce the land stock. No new land will be cleared (equation
2.22.2), and more land will be abandoned (equation 2.23.2). Once the amount of annual
clearance is determined, equations 2.24 and 2.25 are used to divide between new
clearance of primary forest and reclearance of secondary growth, two important sources

of new agricultural land.

4.4.4 Other Proximate Sources

In addition to agricultural development, logging, fire, mining and road construction are
four other proximate sources of deforestation. Figure 4.9 illustrates the interactions
among these source activities, and the interactions between these sources and agricultural

development as well as population.

140

 

 

 

Deforestation

T

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Fire fl——4 Logging <—> Road <—> Mining Agricultural
Development

1 4

Migration ............. D Population .................... 5

 

 

 

 

 

 

Figure 4.9: Interaction Loops between Deforestation Source Activities

(1) Logging

The Amazon basin is an enormous warehouse of tropical timber with an estimated 50
billion cubic meters of standing wood. Logging is becoming an increasingly important
activity in the Brazilian Amazon; it is by far the most extensive forms of human activity
in Amazonia (Parayil and Tong, 1998). Basin-wide estimates of logging area vary
widely. Nepstad et al. (1999) used survey data to estimate a total logged area of 9,000-
15,000 km2 in 1996-1997, nearly equal the area of forest converted to pasture or
agriculture. Krug (2000) estimated a logging rate of 1,561 ka/year using visually
interpreted Landsat prints. Matricardi et a1 (2001) developed an automatic remote sensing
analysis along with visual interpretation and quantified the selective logging area in the
Amazonia to be 5,627 km2 by 1992, which increased to 9,449 km2 by 1996 and 23,379

km2 by 1999.

141

Logging was considered as forest degradation as opposed to a source of deforestation
until recently (Panayotou and Sungsuwan, 1994; Houghton et al, 1991). With only a few
marketable species, logging in the Brazilian Amazon is highly selective. Current logging
practice generally removes timbers in a volume range of around 60 cubic meters per ha (~
2-10 trees per ha) out of a total of up to 300 trees above 10cm diameter. But as many as
27 trees can be damaged for each tree harvested as a high level of collateral damage

(V erissimo et al. 1992). Besides damage to nearby trees, selective logging have far more
detrimental environmental consequences by increasing fire susceptibility and stimulating

road construction.

(2) Road Construction

Road construction that provides access to remote forested areas is a critical source of
deforestation, but with more serious effects than the road construction activities
themselves. It is a main factor driving the vicious feedback cycles that threaten Amazon
ecosystems (Cochrane et al, 1999). A major component of the colonization programs in
the Amazonia was road construction and improvement (Laurance et al, 2001; Dale et al,
1993; Pfaff, 1997; Frohn et al, 1990). Highways were paved amid much controversy
regarding the potential environmental impacts. Financed road construction programs have
stimulated a massive shift in investment from southern Brazil to the Amazonia. Besides
facilitating timber transportation, roads opened for logging stimulated immigration,
which in turn brings more logging, more agricultural development and subsequently more

deforestation to the area. Uncontrolled exploration by loggers catalyzes deforestation by

142

opening roads into unoccupied forest area that are subsequently colonized by ranchers

and farmers (V erissimo et al. 1995).

(3) Forest Fire

Forest fire is an important deforestation source that is closely related with logging.
Logging changes forest structure and increases forest flammability (Holdsworth and Uhl,
1997). Fires are frequent because of the accidental spread of fire from nearby pastures.
The fire susceptibility is subject to human disturbance. According to Uhl and Kaufman
(1990), the flammability of forest ecosystems is highly dependent on the structure of the
vegetation that determines the microclimate of the stands. A primary forest with
undisturbed canopies does not reach flammable levels even in periods of more than 30
rainless days. A lS-year-old secondary forest need a period of 8-10 rainless days to have
flammable conditions, and a pasture needs only 24 rainless hours. The selectively logged
forest, with artificially open canopies, becomes flammable in 5-6 rainless days. Research
of Cochrane et a1 (1999) Shows that fire in highly degraded areas are significantly more

severe in all respects of flame height, intensity, depth, residence time and rate of Spread.

(4) Mining

Mining is also an active source of deforestation. The Brazilian Amazon has significant
mineral resources including coal, gold and the ores of iron, aluminum, copper, tin, kaolin
and some rare minerals such as niobium. There were 530 mineral extraction
establishments and 3,161 mineral processing establishments in 1980. Like logging

establishments, these mining establishments are responsible for road constructions

143

leading to mineral-rich areas, providing a rapidly growing economic impetus for road
building in the Amazon (Laurance et a1, 2001). While mineral extraction may destroy
hundreds to thousands of hectares of forest directly, mineral processing can consume
even larger areas of forest. Fearnside (1989) explored substantial destruction of forested
areas in the eastern Amazon to provide raw materials for charcoal to be used in smelting

pig iron.

(5) Modeling

 

Figure 4.10 shows the structure of this part of the model. Equations 2.26 to 2.31 are the

mathematical representations of the model.

tire

  

I“ a «.
. l

road construction rate

-". 3 lb. 3’

IMMIGRATION

- a:
O

LOGGINGAREA
, Logging

Migrating Road Construction

Figure 4.10: Modeling Other Proximate Sources of Deforestation - Logging, Road
Construction, Mining and Forest Fire

144

Table 4.4 Equations of Other Proximate Sources of the Socio-Economic Model

 

No. Equation Specification

 

2.26 ROAD(t) = ROAD(t - dt) + (Road_Construction) * dt

2.27 Road_Construction = ROAD * Road_Construction_rate

2.28 Road_Construction_rate = a + b* LOGGINGAREA +c* MININGAREA
2.29 LOGGINGAREA(t) = LOGGINGAREA(t - dt) + (Logging) * dt

2.30 Logging = LOGGINGAREA * Road_Construction_rate

 

2.31 fire = LOGGINGAREA * random (min, max)

2.32 MININGAREA(t) = MININGAREAO - dt) + (Mining) * (it

2.33 Mining = MININGAREA * Road_Construction_rate

2.34 IMMIGRATION(t) = IMMIGRATION(t - dt) + (Migrating) * dt

2.35 Migrating = IMMIGRATION * Road_Construction_rate

 

Equations 2.26 to 2.28 model road construction. New road constructions extend the road
network in Amazonia by adding more roads (equation 2.26 and 2.27). The annual rate for
road construction is determined by logging and mining activities (equation 2.28). The
parameters a, b, and c in equation 2.28 are estimated using municipal-level data on
logging, mining and road construction. Annual logging area, mining area and population
migration are all defined as functions of road construction (equations 2.29 — 2.34). Fire is
a firnction of logging (equation 2.35). Areas with more logging are subject to a higher

probability of catching fire. A random function with a minimum probability of 0.1 and a

145

maximum probability of 0.5 is used to defined the magnitude of forest fire caused by

logging.

4.5 Population Model

The population model estimates changes in population and associated demographic
characteristics based on historic records and projections. There is a popular notion in
literature that associates land use/land cover change with demographic pressure
(Bilsborrow and Okoth-Ogendo, 1992; Mather, 1996; Wood and Perz, 1996). Most of
these studies showed a significant correlation between population size (or density) and
land use/land cover change, and used the statistical evidence to theorize the argument that
demographic pressure is the most important driving force for land use/land cover change

(Meyer and Turner 11, 1992).

Intuitively, there is a positive correlation between population and deforestation. Fulfilling
the resource requirements of a growing population ultimately requires some form of land-
use change to provide enough land for food production. In a dynamic perspective, this
implies a positive relationship between population growth and deforestation. However,
land use/land cover change is a more complex problem than simply because of too many
people (Skole et al, 1994). The intensity of the relationship between population and land
use/land cover change can vary considerably because it is mediated by a series of socio-
cultural, economic and ecological factors (Andersen and Reis, 1997). Malthusian and
neo-Malthusian demographic theory offers an over-simplistic understanding of the

complex relationship of resources and population. Population in Amazonia should be

146

analyzed in a more comprehensive way that extends beyond a sheer size of population
size, which represents only one important variable in the complex relationship between

population and resources.

This study proposes a dynamic population system model that considers at least the
following demographic factors: (1) age structure and gender composition, (2) population
distribution (urban vs. rural), (3) occupation, and (4) population growth mode (natural

population growth vs. migration).

4.5.1 Age Structure and Gender Composition

Population age structure, along with the gender composition, matters in determining the
impacts of population on land use/land cover change. The effect of population growth on
land use/land cover change, via increases in food consumption, is maximum when
children reach adolescence, i.e., with a time lag of approximately 15 years (Lambin,
1994). In the same way, age structure decides the degree of participation in agricultural
production. Population within the age range of 15-65 is the main labor pool that is
economically active. Together with population’s gender composition, age structure can
influences the region’s population growth. Women of age 15-60 are the main fertility

group that contributes to the natural growth of population.

The population pyramids (Figure 4.11) Show the relative population size of each age

group and gender category. The youngest age group is at the bottom and the oldest is at

the top; males are on the left and females are on the right. Each horizontal bar represents

147

the number of people in a 5-year age group as shown by the labels on the left. The
triangle-shaped pyramid of the Brazilian Amazon is typical of many developing
countries. The wide bottom of the pyramid indicates a high population growth rate as
well as a gradual shift in the age structure in favor of the economically active population.
Besides this, young people are more likely than their older counterparts to migrate,
primarily as they leave the parental home in search of new opportunities. As a result,
given the relatively large younger generation, we might anticipate an increasing level of

migration and urbanization.

 

 

1500 1000 500 0 500 1000 1500
Population (In thousands)

Figure 4.11: Population Pyramid in the Brazilian Amazon, 1990. Data source:
[BGE (Brazilian Geographic and Statistical Institute)

4.5.2 Population Distribution: Urban and Rural
The general picture of Amazonia is rural and wildness. However, more than half of its

population is living in urban areas. Total population in the Brazilian Amazon grew at an

148

average annual rate of 4.0% during the period 1970-1991. Urban population grew much
faster than that of rural, leading to a dramatic change in the composition of the population
(Andersen and Reis, 1997). According to demographic census, urban population

percentage had increased from 27.7% in 1940 to 55.6% in 1989 (Figure 4.12).

 

 

 

 

 

 

40i

Urban%

 

 

 

 

 

1940 1950 1960 1970 1980 1989

Figure 4.12: Growth of Urban Population Percentage (1940 — 1989). Data source:
IBGE (Brazilian Geographic and Statistical Institute) census.

Urban and rural population has different production and consumption behaviors,
therefore the ways in which population is distributed among urban and rural areas can
affect land use change (Andersen and Reis, 1997; Moran, 1999). Therefore it is important
to differentiate these two groups of population and to model their behaviors separately. A
traditional rural family lives on a farm, where the demand for land comes both fi'om
agricultural production and consumption, while an urban family typically lives in urban

areas and works in the service sector with no direct impact on deforestation. However,

149

 

urban residents will have an indirect effect through the demand for agricultural goods in

the form of both raw materials and food consumption.

4.5.3 Occupation

Occupation is an important linkage between the population system and the economic
system. Economically active population work with different occupations, therefore have
different production functions and different requirements for land as an input to
production. It is the human economic actions undertaken by different population groups
or agents (such as small farmers, cattle ranchers and loggers etc) that are directly
responsible for land use/land cover change. The overall human impacts on land use/land
cover change will depend on the relative size of each occupational group and its
production behavior. Cattle ranchers account for 6-8% of the rural population, roughly
10% of small farmers, who are the predominant majority of rural working force. But
cattle ranchers have far more impacts on the process of land use/land cover change than
small farmers (Hecht, 1985; Moran et al, 1994; Fearnside,l993). As seen from figure

4.13, cattle ranching area is much larger than small farm cultivation area.

150

Small Farm
\ '.

Cattle Ranchirrg

 

Figure 4.13: Different Land Use Patterns for Different Occupational Groups.
LANDSAT TM image (path 230, row 69, 05/15/99)

4.5.2 Population Growth: natural population growth vs migration

The population in the Brazilian Amazon increased from 7.3 million in 1970 to 16.6
million in. About 40% of the increase was caused by migration into the area, so the
number of immigrants in the period is 3-4 million (Andersen and Reis, 1997). Both
natural population grth (the excess of births over deaths) and migration will lead to an
increase in population size, but with different mechanisms. Migration is facilitated by
road construction, while natural population grth is more a function of the region’s
existing population. In order to have a better picture of the future population size in the
Brazilian Amazon, we need to examine these two components of population growth

separately.

151

Migration has been a driving force for extensive land use/land cover change. Large
numbers of migrants in search of land and employment entered the area, as well as firms
that wanted to establish cattle ranches. Early in the 1960s, the Brazilian government
decided to initiate a huge development program that Should integrate the Amazon region
into the rest of the economy. Since then, around 60,000 kilometers of roads were
constructed in the region, several hundred thousand people were helped to settle along
these roads, and millions of others followed without official help (Andersen and Reis,

1997).

4.5.5 Modeling Population Growth and Distribution

The population growth model is a discrete-time model that describes four ecological
processes: age growth, reproduction, mortality and migration. It is used to generate the
region’s intrinsic rate of overall population growth and the proportion of each age class.
All population is disaggregated into six age cohorts: (1) below 15; (2) 15-29; (3) 30 —
44; (4) 45-59; (5) 60 — 74; (6) 75 and up. As shown on figure 4.14, population of each
age cohort is an array with a gender dimension. Equations 3.01 — 3.05 are the
mathematical expressions of the population growth model, where i represents the
population cohort number (i = l, 2, 3, 4, 5, 6), and j is the population gender dimension (j

=1 for male; j = 2 for female).

152

. . birth rate 1529

' l

‘ o + «4%
«011529

surviv rate 1529 . “M‘s” ./'

   
  

  
 
 
    
 
   
 
       

Deatht SBeiow

3_7T— . r
em .
/

surviv3044

 

 
 
      

survivi 529 Miorationi

      

“M30“ unriv rate 3044

63.]

Migrationfii

 

a! immigration surviv3044

   
  
 
  
   
 
   

 

2‘ r}. 3 .95
urnmmmtt 3 5 iii
AGE1 SBELOW g g; AGE4559

. ..rs§1mszr'!éi

total population

J , .,... 33; death rate 4559
milérgiggii Grow4559

"nu“,

 

S5
5 5::
r :3 6: J
: ................ J 5 5g
"t'ifi!!!!i§£5§ii...£ D :: instrumental}
AGE75UP 5g; AGE4559
:3
weatherman-‘1
AG £6074

 

 
      
 

’“7 -
Death75up AGE 5U Groweou

death rate 0074
DoathRateTSUp

Figure 4.14: Population Growth Model with Four Ecological Processes: age growth,
age-specific reproduction and mortality, and migration.

153

Table 4.5 Equations of the Population Growth Model

 

No. Equation Specification

 

3.01 AGE; 3(1) = AGE” (l-dt)+ (GIOWH 3+ Migration“ - GTOWiJ - Death”) X (II

3.02 Deathw- = AGE” x death_rateiJ
3.03 GrOWiJ = AGEiJ x (l- death_rateiJ) x (l/xi) ( for all i > 0)
4
3-04 GTOWOJ = _§(AGELZ x birth_rate; x surviv_ratet)/2 I!“
if
6 2 E:

3-05 total _population = (Z1 ZlAGEtJ)
1= J:

 

Note: i represents the population cohort number (i = l, 2, 3, 4, 5, 6), and j is the
population gender dimension (j =1 for male; j = 2 for female).

Population in one cohort moves up into the next cohortby the growing process as part of
the progress through the life cycle. Equation 3.01 shows that the number of population in
age cohort i at time t+1 equals to the number of population in the same cohort of the
previous time t, plus the population grown from the prior age cohort i-l and the migrated
population into the age group, minus the population grown into the next age cohort i+1

and the population that die within the age cohort i.

Equation 3.02 calculates the number of population who dies within the age group. The
mortality rate varies across all age cohorts. The last cohort is moved out of the population

system at above age 75.

154

For an age cohort with an x-year interval, nearly the whole cohort is moved to the next
stage within x years. Since the time step for the model simulation is one-year, we assume
the transfer coefficient of the age development process to be l/cohort size, that is, l/x of

the survived population in an x-year cohort will grow to the next cohort (equation 3.03).

In equation 3.04, we assume that only female population in the age range of 15 to 59
years old (cohort number 2, 3 and 4) have reproduce capability that varies among
different age groups. We assume that the birth rate, death rate, and mortality rate will
remain at the current level. We need to model the reproductive rule such that each fertile
cohort reproduces at the age-specific birth rate multiplied by the size of the
corresponding female population. Statistics shows the survival rate for a new-bom is
dependent on the mother’s age, so the actual number of population adding to the age
cohort i = 0 (below 15 years old) is adjusted by new—born survival rate. We also assume
half of the survived newborns will add to the female population and the other half will

add to the male population.

The total number of population is summed over all age cohorts and gender dimension in

equation 3.05.

The population number of each age group as well as the corresponding birth rate, death
rate and new-bom survival rate is available from the census data. Yet the number of
population migrated into the region is undefined at this point. We will project the annual
population migration based on road construction and economic development in the socio-

economic model.

155

 

The following part of the population model deals with the region’s urban/rural
distribution and occupational compositions (see figure 4.15 for model structure and table
4.6 for model specification). The number of people living in either urban or rural area is
determined by the urbanization level (equation 3.06 and 3.07). Only people within the
above age 15 and below age 75 are considered as economically active (equation 3.08),
based on which we calculate the active employment of both urban and rural areas
(equation 3.09 and 3.10). Equation 3.1 l generates the number of agricultural labors that

works as farmers (k=l), as cattle ranchers (k = 2) or as day labors (k = 3).

 

total population

9 z z .
'._ J. .

A E75UP ' " AGEtSBELOW

   

daylabor

  

rural population

Others

total population

Figure 4.15: Population Urban/Rural Distribution and Occupational Composition

156

Table 4.6 Equations of Population Distribution Model

 

 

No. Equation Specification

3.06 urban _population = total _population x urbanization

3.07 rural _population = total _population x (1- urbanization)
2

3.08 active_ratio = [total _population — 21(AGEIJ + AGEfiJ)] / total _population
J:

3.09 urban_employee = urban _population x active_ratio

3.10 rural_employee = rural _population x active_ratio

3.11 agro_labork = rural_employee x labor__ratiok

 

4.6 Sensitivity Analysis

Sensitivity analysis is a useful tool for model evaluation. It helps build confidence in the

model by studying the uncertainties that are often associated with parameters in models.

There are three critical parameters subject to high uncertainties in the model: the

succession rate of undisturbed secondary grth to forest (regrowth_rate2), the

autonomous abandonment rate of agricultural land with exhausted fertility

(auto__abandon_rate) and the ratio of new clearance on primary forest (clearance_rate).

Parameter sensitivity analyses are performed to determine how sensitive the model is to

changes in the value of these parameters, and to indicate which parameter values are

reasonable to be used in the model.

157

4.6.1 Secondary Growth Regrowth Rate

If secondary growth is not disturbed, its reflectance characteristics will eventually
become indistinguishable from those of mature, undisturbed forest (Mausel et al. 1993,
Boyd et al. 1996). But the question of how long it takes secondary growth to revert to
mature forest remains quite uncertain. Brown and Lugo (1990) showed that secondary
forest recovered most of its biomass during the first thirty years, after which biomass
accumulation slows considerably. Analysis of Vieira (1996) concluded that secondary
forest recover only thirty five percent of mature forest biomass after twenty years.
Saldarriaga et al. (1988) found a steady biomass increase for older secondary vegetation,
but even after eighty years biomass remained substantially less than primary forest. In a
similar analysis. Houghton and Skole (2000) indicated that abandoned agricultural land
recovered 70 percent of their original biomass in 25 years and the remaining 30% over
the next 50 years. But they realized that the assumption that “forests are fully regrown
after as little as 75 years are probably not valid”. In this analysis, we selected four
possible forest regrowth rates for sensitivity analysis: 1/25, 1/50, 1/75 and l/ 100,
implying the assumptions that undisturbed secondary growth will recover most of its
biomass stock and take similar spectral reflectance and Structural quality like that of a

primary forest canopy in 25, 50, 75 or100 years.

4.6.2 Autonomous Abandonment Rate
The largest part of the soil in Amazonian has a low fertility. Autonomous abandonment is
the land taken away from agricultural production as the result of soil fertility exhaustion.

is independent of the agricultural economic process. The autonomous abandonment rate

158

is also subject to high variability. According to a Landsat TM image analysis by Skole et
al (1994), approximately 11% of the active agricultural land was abandoned each year
between 1986 and 1988. However, between 1988 and 1989 when there was more than a
twofold increase in forest clearing, 22% of the agricultural land was abandoned annually.
The amount of land being abandoned is the combined results of both autonomous and
induced abandonment. It is not easy to determine how much land is abandoned
autonomously because of ecological constraints, and how much abandonment is induced
by economic concerns. In this analysis, we assume the autonomous abandonment rate has
a lower bound of 10% and an upper bound of 20%. Five runs of sensitivity analyses were
performed on an incremental basis, where the autonomous abandonment rate was set to

be 0.1, 0.125, 0.15, 0.175 and 0.2 respectively.

4.6.3 Clearance Rate

Clearance of primary forest and reclearance of secondary growth are two important
sources of new agricultural land. The agricultural economic processes help to determine
how much more land is needed to put into production, but they couldn’t provide
information on the percentage of land coming from cutting primary forest and percentage
of land from reclearing secondary growth. Satellite TM image analyses by Skole et a1
(1994) Showed that 42% of the new agricultural land in the Amazonia was created from
clearing of secondary growth between 1988 and 1989. The clearance rate in this study is
set to be 0.50, 0.55, 0.60 and 0.65 for each sensitivity analysis run, assuming that 50%,

55%, 60% or 65% of the clearance will come from deforestation on primary forest.

159

4.6.4 Sensitivity Analysis Results

Running sensitivity analyses on the specified model parameters show that specific
parameter values can change the magnitude of deforestation and secondary growth, but
not at a highly significant level. Results of sensitivity analyses are compared with
remotely sensed land cover data to choose which parameter values are reasonable to use
in the model. If the model behaves as expected from real world observations, it gives
some indication that the parameter values reflect the real system behavior, so it is
reasonable to use the estimated parameter value in the model. Based on this rule, the
secondary growth regrowth rate in the final model was set to be 1/75, the autonomous
abandonment rate was 0.125, and the ratio of new agricultural land cleared from primary

forest was 0.60.

Despite the magnitude difference of LUCC, there were no significant changes in the
overall system behavior. Dynamic system models are general insensitive to many
parameter changes. It is the structure of the system, not the parameter values, that has
most influence on the behavior of the system. By showing that the system does not react
greatly to changes in parameter value, it reduces the modeler’s uncertainty in model
behavior. In addition, it gives an opportunity for a better understanding of the dynamic

behavior of the system.

4.7 LUCC Modeling Results and Validation

The LUCC model is applied to the Brazilian Amazon and to the State of Rondonia as

well. The base year for model Simulation is set to be 1986. The model projects the

160

 

region’s population, economic growth as well as land use/land cover change into year
2020 at a one-year time step. Remotely sensed land use/land cover data for years of 1992
and 1999, agricultural census data for year 1990 and demographic census data for year

1991 are used to validate model results.

4.7.1 Population Modeling Results and Validation

Figure 4.16 shows the exponential population grth pattern for the Brazilian Amazon
basin and for the State of Rondonia as well. Based on model projection, population of the
Brazilian Amazon will reach 31.2 million by year 2020, nearly twice the population size
of year 1991. During 1986 — 2020, the average annual rate of population growth in the
Brazilian Amazon basin is 3.87%. According to IBGE census data, population growth
rate during1980 — 1990 was 4.69%. Therefore we can expect a Slower population growth
rate in the Amazonia in the next two decades. For the State of Rondonia, the population
size will grow from 1.1 million in 1991 to 3.8 million in 2020, with an annual population
growth rate of 10.3%. The population growth pattern for the State of Rondonia resembles

that of the whole basin, but with a faster pace.

161

 

40

30 ,.___ +Amazon Basin , _ L L
+ Rondonia

 
   

 

 

10

Population (1 .000.000 Persons)

1
l
l
l
a
l
i
l
l
'1
l
l
l
l
l
l
l
l

 

 

 

0 YYYYYYYYYYY fir T fir f I I I f

19861988 19901992 19941996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020

Figure 4.16: Population Growth in the Brazilian Amazon and the State of Rondonia
(1986 - 2020)

According to the demographic census, 52% of the population lived in the urban areas of
the Brazilian Amazon in 1991. Urbanization level in Rondonia is higher, with 58% of
population lives in urban area in 1991. Figure 4.17.1 and figure 4.17.2 illustrate the
different growth pattern for urban and rural population in the Brazilian Amazon and in
Rondonia. On both figures, urban population keeps an exponential growing pattern like
that of the total population, while rural population grows in a linear fashion. Therefore
urban population tends to play a more and more important role in the region’s economic
development. Compare the urban/rural distribution figure of the basin with that of
Rondonia, we can see the rural population in Rondonia grows faster than that of the
whole basin. So rural economy in Rondonia might have more important impacts on the

landscape change process.

162

 

1"‘ _u_
—+—

20.0 t_l URBAN ._ L, _ L- f,,_,._________‘_~______ __ J
1:393:81;

150 "'—""—""" ““— " * *'#"“"“"—“—‘“ ____ 7"“—”‘—"‘— ”"—” “"’”

 

 

10.0

population (million)

 

 

 

5.0 “......— _._. ._ -__.. __ ._.____—.- i -. _..

 

 

 

0 0 t ' I i T I T V T r l l
I

b at o '1 h b ‘b o '1 9 ‘b o '1; b: b a: o
1 ‘5 °9 9 q q o o o \ \ x x x r).
i {ll \9 a“ \q \q 3” \01 A9 '19 ’19 qJo°b '19 '19 .19 '19 .19 '19 go

L- .—

7 l T I I

 

 

Figure 4.17.1: Urban vs. Rural Population in the Brazilian Amazon (1986 — 2020)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

r,,______ ___i__.___ L“- _ _
3.0

E?

.9.

"8'

V

E

.2

E

5

O-

O

Q-
OOTI HTTT TI IIIIIIIII III ITrI

lb‘bb'vb-b‘b '1, b‘bQ'lab-bQ-ro
‘b‘bq 9919 echo \xxxxn

! \01 9 \01 @931 \ \9 "19° '19 ,1, n9 WQQ ,19 '19 ’19 '19 '19 ”or

 

 

Figure 4.17.2: Urban vs. Rural Population in the State of Rondonia, Brazil (1986 — 2050)

163

 

Table 4.7 shows the model validation for projected population of different age group. It
compares the model-estimated population size of 1991 with the observed population data

from demographic census in 1991 to examine the fitness of the model.

Table 4.7 Validation Table for Population of Different Age/Gender Groups

 

<15 15-30 30—45 45—60 60-75 >75 Total

 

Census 3,309.9 2,179.2 1,281.4 650.5 303.5 100.9 7825.4

Mm MM“ 3180.1 2193.7 1301.9 662.4 306.2 96.6 7740.9

 

3mm“ E""'% -3.92 0.67 1.60 1.83 0.89 -4.26 -1.08
Amazon Census 3,235.8 2,172.3 1,250.3 612.2 285.5 91.9 7648.1

 

(1991) Fem“ MM“ 3136.3 2180.4 1,273.9 624.4 2862 92.1 7593.3

0
Em” /' -307 0.37 1.89 1.99 0.25 0.22 -0.72

Census 233.7 172.4 105.1 50.4 21.4 3.4 586.3

 

 

Mm MM“ 234.1 168.0 112.3 47.8 19.4 3.4 585.0

State of 3"" % 0.17 -255 6.85 -5.16 -935 0.00 -022

R°“"°'““ Census 224.7 167.2 95.8 41.3 14.8 3.2 547.0
(1991)

Fem" MN“ 229.0 162.9 105 40.6 14.6 2.6 554.7

Error %

 

 

 

1.91 -2.57 9.60 -l.69 -l.35 18.75 1.41

 

Unit: 1000 persons. Validation data source: 1991 Demographic Census

Error percentages for the Brazilian Amazon range from 0.22% (female, age older than
75) to —4.26% (male, age older than 15). The overall population size for year 1991 was
15.4 million, while the model projected population size is 15.3 million, an estimate with a
marginal error of -0.9%. In the State of Rondonia, error percentages of population

estimates range from -0.00% (Male, age older than 75) to —18.75% (Female, age older

164

than 75). The total population of Rondonia was 1.133 million in 1991 while and model
estimate was 1.139 million. The overall error percentage of population model estimate in

Rondonia is as low as 0.57%.

4.7.2 Socio—Economic Model Results and Validation

Figure 4.18.1 and figure 4.18.2 show the expansion of agricultural land area in the
Brazilian Amazon and in the State of Rondonia during the period of 1986 - 2020. The
land areas of rice, sugarcane, soybean, corn, manioc, beans, coffee, banana, cacao, black

pepper are grouped into one class called cultivation area.

According to the model projection, total agricultural land area in the Brazilian Amazon
will increase from 16.7 million hectares in 1986 to 44.8 million hectares in 2020, among
which cattle ranching area will increase from 8.5 million ha to 20.8 million hectares. As
seen from figure 4.18.1, basin-wide agricultural cultivation area and cattle ranching area
will experience similar level of grth during the study period. For the State of Rondonia
alone, the total agricultural area will grow from 1.63 million hectares in 1986 to 7.33
million hectares in 2020, with cattle ranching area growing from 1.13 million hectares to
5.28 million hectares. Compare figure 4.18.1 with figure 4.18.2, we can tell that
Rondonia will experience faster speed of agricultural expansion, especially for cattle
ranching, than the whole region of Amazonia basin. Figure 4.18.2 shows that cattle
ranching area in Rondonia grows as much as twice faster than the agricultural cultivation
area. It is the cattle ranching activities that account for most of the area expansion of

agricultural land in Rondonia.

165

 

 

5000 iii ;’; 7;,4‘? -..“,L ,L 7 74 A 7— 7‘»--— V v
i IDRanching ‘

 

Am (mllllon ha)

10.00 '

 

 

 

 

 

 

Area (million ha)

 

 

Figure 4.18.2: Agricultural Land Area Projection for the State of Rondonia (1986 — 2020)

166

Table 4.8 Shows the model validation for different agricultural land areas using
agricultural census data in 1990. Comparison Shows the generally, the model estimated
agricultural land areas have a very good fitness with the agricultural land census data.
The error percentage ranges from -1.62% (area for manioc) to 10.90% (area for
sugarcane) for the Brazilian Amazon, and fi'om 0.00% (sugarcane) to —10.25% (coffee)
for the State of Rondonia. The marginal error for the all agricultural land projection is -—

7.09% for the Brazilian Amazon and only —0.69% for the State of Rondonia.

 

 

 

 

 

 

Table 4.8 Validation Table for Agricultural Land Areas
Brazilian Amazon (1990) State of Rondonia (1990)
Census Model Error % Census Model Error %
Rice 2,351.0 2,470.6 5.09 149.6 149.8 0.13
Sugarcane 4,565.7 4,068.2 -10.90 0.4 0.4 0.00
Beans 341.3 334.9 -1.88 122.1 112.9 -7.53
Manioc 493.4 485.4 -1.62 28.6 28.4 -0.70
Corn 1,236.1 1,209.1 -2.18 158.0 151.6 -4.05
Soybeans 1,747.0 1,587.1 -9.15 9.2 9.1 -1.09
Banana 102.8 100.5 -2.24 21.9 21.7 -0.91
Cacau 82.5 77.4 -6.18 38.5 38.3 -0.52
Coffee 247.3 229.5 -7.20 138.5 124.3 -10.25
Black Pepper 29.4 29.9 1.70 0.0 0.0 n/a
Cattle 12,6998 1 1,609.1 -8.59 1541.0 1556.0 0.97
TOTAL 23,896.3 22,201.7 -7.09 2207.8 2192.5 -0.69

 

 

 

Agricultural land area unit: 1,000 ha. Validation data source: 1990 agricultural census

167

 

 

4.7.3 Land Cover Change Modeling Results and Validation

Based on the demographic model and the socio-economic economic model, future land
use/land cover changes are projected for the whole basin as well as for the State of
Rondonia only. Figure 4.19.1 and figure 4.19.2 present the model results for different
land types. As shown on the figures, forest area in both cases will decline steadily while
deforestation and secondary grth areas will increase Significantly, indicating massive

deforestation as well as extensive growth of secondary vegetation in the years to come.

According to remote sensing TM image analysis, the basin-wide deforestation area in
1986 was 15.8 million hectares, and a secondary growth area was 7.02 million ha.

The LUCC model projected that, by year 2020, deforestation area in the Brazilian
Amazon will grow to 50.5 million hectares, and secondary growth will reach 60.5 million
hectares. Total deforested area (including areas under deforestation and secondary
growth) will account for 29% of the previously forested area of the Brazilian Amazon in
2020. The basin-wide estimates are compatible with the results from Laurance et al’s
(2001) GIS model, which showed that 28% of the Brazilian Amazon will be deforested or
heavily degraded by year 2020 under the optimistic scenario, and 42% of the region will
be deforested under their non-optimistic scenario. The situation is worse in the State of
Rondonia, where there were 1.6 million hectares of deforestation area and 0.63 million
hectares of secondary growth in year 1986, but model results Show that deforestation area
will rise to 9.4 million hectares; and secondary growth will be 7.0 million hectares in year
2020. According to this projection, 73% of the primary forest area will disappear from

Rondonia by 2020 (figure 4.19.2).

168

 

 

 

 

3. 70e+008 _- ----

5.500+007
6.50e+007 ... /
\JJ /2/3
/2 //
3.208+008

’ 1: FOREST 2: DEFORESTED 3~ SECONDGROWTH
1:
2
3

 

 

 

 

 

 

 

 

 

 

1:
2: 3.50e+007 2 \N ‘3
3: 3.50e+007 / 1
///3/ \1
1 2 70 008 ,/—2——3/
I . 6+
2: 1.508+007
3: 500000000
1986.00 1994.50 2003.00 2011.50 2020.00
3 @1- Land Cover (Untitled) Time 11:03 PM Thu. Nov 22. 2001

 

 

 

Figure 4.19.1: LUCC for the Brazilian Amazon (1986 — 2020)

 

 

’ 1: FOREST 2: DEFORESTED 33 SECONDGROWTH
12 2.508+007
21 950000000
32 700000000
2 3
"1\ /

 

.r \JJ /
1.SOe+007 \/2 /

 

 

 

 

 

 

 

 

 

1:
2: 550000000 N r
3: 350000000 / 1 3
1/ // \1_
/3
/

1: 500000000 /——2 3
2: 150000000 /
3: 0.00

1986.00 1994.50 2003.00 2011.50 2020.00

8 a] Land Cover (Untitled) time 11:04 PM Thu. Nov 22. 2001

 

 

Figure 4.19.2: LUCC for the State of Rondonia (1986 - 2020)

169

 

 

Though both deforestation and secondary growth areas will experience significant
increase during 1986-2020, it seems clear that the magnitudes and rates of their changes
are quite different. Here we define the annual growth rate for deforestation grth rate

and secondary growth in the following manner:

AnnualDeforRatet = (DEFORESTED — DEFORESTED 1-1)/DEFORESTED 1-1 * 100
AnnualRegrowRatea = (SECONDGROWTHt— SECONDGROWTH 1-1)

/ S ECONDGROWTH 1-1 * 100

As shown on figure 4.20.1, deforestation area grth for the Amazonia starts with a high
growth rate of 11% in 1988, which lowers down to a bottom threshold rate of 0.92% in
2009, after that the deforestation growth rate starts to head up again to the level of
approximately 2%. The curve of secondary growth exhibits a very different pattern. It
starts with a low rate of 2%, which quickly rises to nearly 10%, after that its growth rate
began to decline to the level of 3%. The annual growth rate for deforestation fells below
that of secondary growth for the first time in year 1992, after that regrowth rate remains

higher than deforestation rate throughout the temporal horizon of model projection.

Similar patterns are found in the case of Rondonia (figure 4.20.2), where deforestation
growth rate declined from 13.38% in 1988 to 2.75% in 2020, and secondary growth
increases by a percentage of 6.48% in 1988, which rises to 12.44% in 1996 and later
drops down to 4.85% in 2020. However, unlike that of the basin-wide curve, the
deforestation grth rate in Rondonia keeps declining throughout the model period, and

the annual growth rate of secondary growth starts to exceed that of deforestation in 1992.

170

 

 

 

P
c>

.Annualeounh(90)
01
o

E"
o

 

 

 

.1

 

1 + AnnualDeforRate J”
ff— AnnualRegrowRate

 

 

 

 

 

 

0.0 T r
1988

T- - - _ -_ - _ ---- - .. ..---- e-

I T—r T

2020

T fj fTT

2012 2016

T T T T T

2008

T T j T T I T T T T T I T

1992 1996 2000 2004

 

 

Figure 4.20.1: Annual Deforestation Rate and Secondary Growth Rate in the
Brazilian Amazon (1986 -— 2020)

 

 

 

 

.... __ __ ._.____ .1

j,+ AnnualDeforRate T“
-..... ..._T+ AnnualRegrowRate .-

 

 

 

 

Annual Growth (%)

4.0 ii

2.0 “"‘ " ‘—_—‘—

 

 

 

 

 

 

 

 

 

 

1988

1
T
T
T00
T
T

f T T T T T f ‘- T T T T T T . T T T T T T 7 T T : T T T

1992 1996 2000 2004 2008 2012 2016 2020

 

T

Figure 4.20.2: Annual Deforestation Rate and Secondary Growth Rate in the State of
Rondonia (1986 — 2020)

171

Land use/land cover data derived from remote sensing image classification of year 1992
and 1999 are used to validate the model results. Remote sensing data show that
deforestation area in the Brazilian Amazon was 23.9 million hectares in 1992 and 31.4
million hectares in 1999; secondary growth was 10.8 million in 1992, which grew to
and15.9 million hectares in 1999. In comparison, modeling results shows that
deforestation area is 26.7 million hectares in 1992 and 36.9 million hectares in 1999,
while secondary growth area is 10.4 million hectares in 1992, and 19.6 million hectares
in 1999. The difference between the observed and model estimated deforestation areas

was 2.8 million hectares in 1992, and 5.5 million hectares in 1999 (see Figure 4.21.1).

For the State of Rondonia, there were 2.2 million hectares of deforestation and 1.0
million hectares of secondary growth in the year 1992 based on remote sensing image
classification. In 1999, deforestation area grew to 4.3 million hectares and secondary
growth area increased to 1.3 million hectares. Modeling results for the state of Rondonia
shows that deforestation area is 2.9 million hectares in 1992 and 4.7 million hectares in
1999, while secondary growth area is 0.7 million hectares in 1992, and 1.5 million
hectares in 1999. The difference between the observed and model estimated deforestation
areas was 0.7 million hectares in 1992, and 0.3 million hectares in 1999 (see Figure

4.21.2).

172

 

 

 

 

 

 

 

 

 

 

 

 

 

40.0
35.9 _
D Remote Sensmg
31.4 I Model Estimate
30.0 - 5 - —41' 7 7 .-. .. .-- -- ~-
207 : i
E 23.9
§
0' 20 0 _
8.
V?
10.0 1 ’
0.0 ' "
1992 Deforest 1999 Deforest 1992 Regrowth 1999 Regrowth

 

 

 

Figure 4.21.1: Validation Chart for Land Types of the Brazilian Amazon, 1992 and 1999

 

 

6.0
'3 Remote Sensing
I Model Estimate

  

4.0"

1,000.000 ha

2.0 5'

 

 

 

 

0.0 r

1992 Deforest 1999 Deforest 1992 Regrowth 1999 Regrowth

 

 

 

Figure 4.21.2: Validation Chart for Land Types of the State of Rondonia, 1992 and 1999

173

The differences between the observed deforestation and model estimates may originate
from the fact that a lot of deforestation caused by activities such as logging or mining
may not be easily detectable from satellite images. The difference might also result from
the uncertainty caused by cloud and cloud shadows in satellite image classification. Land
areas under the classes of cloud and cloud shadows may actually be deforested areas, but
those areas are excluded from this project. The LUCC model only considers three types
of land: forest, deforestation and secondary growth. It assumes active deforestation only
takes place on forested land or secondary growth. But the fact is that some agricultural
land is cleared from other types of land such as cerrado, which may further complicate

the land use/land cover change process.

4.8 Conclusions and Discussions

Land cover changes associated with land use practices have exerted great influences on
global environmental change (Houghton and Skole, 1990; Houghton, 1991). In addition
to a need for refinements in measurements of the rate, extent and spatial pattern of
deforestation, it is important to understand the processes of human land use and other
socio-economic behaviors as well as their consequent impacts on land cover changes; and

based on these, to make projections for future land use/land cover change in Amazonia.

Most LUCC models are mainly focus on deforestation issues rather than the land use/land

cover change dynamics. The forest versus non-forest view overshadows some very

interesting dynamics which are only recently being observed and discussed. This analysis

174

not only addressed the problem of how forests become deforested, it also studied the
process of how deforested areas grow into secondary vegetation and what is the fate of
this secondary growth. By analyzing the dynamics of forest clearing and abandonment in
the Brazilian Amazon, this study will significantly enhance the understanding of land

use/land cover change process

Instead of using traditional statistical modeling method, this analysis adopts a systems
modeling approach to simulate the dynamic structure and behavior of land use/land cover
change as a complex system with interactions, feedback loops and reciprocal effects
between population groups, economy sectors and land use system. Most land use/land
cover change view land as an ecological system without explicitly incorporating the
socio-economic process; a few have incorporated the socio-economic process, but in a
limited way (Parks, 1991). Though the extent and frequency of land use/land cover
change are the results of a multitude of human decisions and physical constraints, it is
possible to identify the key role-players and driving forces for land use/land cover change
processes, to understand how the major determinants interact to drive land cover change
and how projections about them could be used to project future patterns of land use,
future rates of land cover change, and future states of land cover. Unlike any black-box or
abstract LUCC models that do not clearly identify the agents and actions being modeled,
the systematic LUCC model in this project is built upon theoretical and empirical
understanding of the factors and processes that control the relative magnitude of forest
clearing and agricultural land abandonment, leading to the development of a diagnostic

LUCC model that tracks the perturbations and responses of agricultural economy on an

175

annual basis. The process-based LUCC model can be linked to other carbon or
ecosystem process models to account for carbon emission or ecosystem disturbance in

Amazonia.

Remotely sensing data collected at multiple times are used as initial inputs to the LUCC
model and also as data for modeling results validation. Remote sensing provides a great
opportunity to characterize and map land use/land cover distribution and change patterns
with detail that is not possible using census data or field surveys alone. This unique
advantage of remote sensing has stimulated a growing interest in making scientific
progress through the use of remotely sensed data in social science research community,
so-called “Pixelize the Social” (Turner II, 1997). Linking remote—sensing data sets to the
human driving forces of land use and cover change permits the development of
empirically-based diagnostic models of important land use/land cover changes such as

tropical deforestation and secondary growth.

Model validations show that the systematic LUCC model is a powerful tool for projecting
future land use/land cover change for the Brazilian Amazon as a whole region, and for
the State of Rondonia as well. This will help to move toward an integrated modeling
scheme at various scales, where a global land use/land cover change model can be able to
scale down to a landscape model paradigm to account for the local socio-economic
dynamics yet retain its global focus, while a landscape model can be able to scale up to
global level to address the geobiophysical issues such as global change but still be able to

account for local spatial dynamics (Clark, 1985; Gallopin, 1991; Mather and Sdasyuk,

176

 

1991; Turner 11 et al, 1993). Developing models of land use/land cover change at the
local-to-regional scales allows for better estimations of future rates, pattern and locations
of LUCC change, which are very important to integrated regional and global models of
land use in the future. Since land use/land cover changes are local in domain, but are
replicable globally (Turner et al, 1990), the dynamic system modeling approach in the
Brazilian Amazon may provide insights into modeling land use/land cover changes that

are taking place in other regions around the world.

The basin-scale or state-level analysis elucidates the interannual variations in
deforestation and secondary growth, but it is not able to draw a detailed map to provide
information on where the land cover changes are mostly likely going to happen. The
limitation of little spatial articulation is attributed to the reason that most socio-economic
data are only available on aggregated geographic units, which restricts the opportunity for
behavioral interpretations of land cover change. Such aggregations conceal the spatial
variations of land use practices. The rates at which land is cleared or abandoned are
highly dependent on local land use practice. In some cases, deforestation and secondary
succession exist as a tightly coupled system in which secondary growth is continually
recycled back into farmland. In other cases, active land management maintains the land
in agriculture, or the lack of active land management or population displacement results
in long-term succession. Aggregations can also conceal the dynamic balance between
gross deforestation and regrowth, which together define net forest loss. Finer scale study
at the level of the farming household overcomes certain of these inherent limitations to

more aggregate analyses (Walker and Homma, 1996; Leinbach and Smith 1994). Future

177

research can apply this model to finer scales so as to include local variations and produce

temporal-spatial maps of land use/land cover change.

4.6 References:

Adams, J.B., Sabol, D.E., Kapos, V., F ilho, Roberts, D.A., Smith, M.O., and Gillespie,
AR. 1995. Classification of multispectral images based on fractions of end members:
Application to land-cover change in the Brazilian Amazon. Remote Sensing of
Environment 52:p137-154.

Alves, D. S. and Skole, D.L., 1996. Characterizing Land Cover Dynamics Using Multi-
Temporal Imagery. International Journal of Remote Sensing 17(4): p835-839.

Andersen, L. and Reis, EL, 1997. Deforestation, Development and Government Policy
in the Brazilian Amazon: an econometric analysis. IPEA-TEXTO Para Discussao N0 513.
Arthur, W.B. 1990. Positive Feedbacks in the Economy. Scientific American 262: p92-
99.

Angelsen, A. and Kaimowitz, D., 1999. Rethinking the Causes of Deforestation: Lessons
from Economic Models," World Bank, 1999.

Bilsborrow, RE. and Okoth-Ogendo, H.W.O. 1992. Population Driven Changes in Land
Use in Developing Countries. Ambio 21: p37-45.

Boyd, D. S., Foody, G. M., Curran, P. J., Lucas, R. M., and Honza, M. 1996. An
assessment of radiance in Landsat TM middle and thermal infrared wavebands for the
detection of tropical forest regeneration. International Journal of Remote Sensing
17(2):p249-261.

Braat, LC. and Lierop, W.F.J., 1987. Economic-Ecological Modeling. Studies in
Regional Science and Urban Economics Series Vol. 16, North-Holland.

Brondizio, E. S., Moran, E.F. et al., 1994. Land Use Change in the Amazon Estuary -
Patterns of Caboclo Settlement and Landscape Management. Human Ecology 22(3):
p249-278.

Brown, S. and Lugo, A. 1990. Tropical secondary forests. Journal of Tropical Ecology
6:p1-32.

Burgess, J .C., 1993. Timber production, timber trade and tropical deforestation. Ambio,
22(2-3): p136-l43.

Clark, WC. 1985. Scales of Climate Impacts. Climatic Change 7: p5-27.

178

Cochrane, M. A., Alencar, A., Schulze, M. D, Souza Jr., C. M., Nepstad, D. C., Lefebvre,
P., & Davidson, B, 1999. Positive feedbacks in the fire dynamic of fire in tropical
ecosys-closed canopy tropical forests. Science 284: p1832-1835.

Crutzen, P.J. and MO. Anreae. 1990. Biomass Burning in the Tropics: Impact on
Atmospheric Chemistry and Biogeochemical Cycles. Science. 250: 1669-1678.

Dale, V. H., R. V. O'Neill, and Southworth, F. 1993. Causes and effects of land-use
change in central Rondé‘mia, Brazil. Photogrammetric Engineering & Remote Sensing
59zp997-1005.

Deal, B and Farello, C. et al, 1999. A dynamic model of the spatial spread of an
infectious disease. Working paper at Univesity of Illinois, Urbana Champaign.

Faminow, M.D., 1998. Cattle, deforestation and development in the Amazon: an
economic, agronomic and environmental perspective. Oxford: CAB International.

F eamside, P. M., 1989. Extractive reserves in Brazilian Amazonia", Biocience, American
Institute of Biological Sciences, June, 1989b.

Fearnside, P.M., 1993. Deforestation in the Brazilian Amazon: the effect of population
and land tenure. Ambio, 22(8): p537-545.

Fearnside, P.M. 1989b. "Deforestation and Agricultural Development in Brazilian
Amazonia." Interciencia. Vol. 14, No. 6. Nov/Dec. p. 291-297.

Foody, G. M., Palubinskas, G. et al., 1996. Identifying Terrestrial Carbon Sinks -
Classification of Successional Stages in Regenerating Tropical Forest from Landsat Tm
Data.” Remote Sensing of Environment 55(3): p205-216.

F oody, G., Palubinskas, G., Lucas, R., Curran, P.J. and Honzak, M. 1996. Identifying
Terrestrial Carbon Sinks: Classification of Successional Stages in Regenerating Tropical
Forest from Landsat TM Data. Remote Sensing of the Environment 552p205-2126.

Frohn, R.C., Dale, V.H. and Jimenez, B.D., 1990. Colonization, road development and
deforestation in the Brazilian Amazon Basin of Rondonia. ORNLTM-11470, Oak Ridge,

Tennessee.

Gallopin, G.C. 1991. Human dimensions of global change: Linking the global and local
processes. International Social Sciences Journal 130: p707-7l8.

Grainger A., 1995. National land use morphology: patterns and possibilities. Geography
20: p235-45.

Hannon, B. and Ruth, M. 1997. Modeling Dynamic Biological Systems, Springer-Verlag,
New York.

Hecht, S., 1985. Environment, development and politics: capital accumulation and the
livestock sector in Eastern Amazonia. World Development, 13(6),p663-684.

Holdsworth, A.R., Uhl, C. 1997. Fire in Amazonian selectively logged rain forest
and the potential for fire reduction. Ecological Applications 7(2):p7l3-725.

179

Houghton, R. A., 1991. “Tropical Deforestation and Atmospheric Carbon Dioxide.”
Climatic Change 19(1-2): p99-l 18.

Houghton, R. A., Skole, D.L., Nobre, G.A., Hackler, J .L., Lawrence, K.T. and
Chomentowski, W.H., 2000. Annual fluxes or carbon from deforestation and regrowth in
the Brazilian Amazon. Nature 403(6767): p301-304.

Houghton, RA. and Skole, D.L.l990. Carbon. In W. B. Meyer and B. L. Turner H, (eds).
Changes in Land Use and Land Cover: A Global Perspective. Cambridge University
Press, Cambridge, p393-408.

Houghton, R.A.; Leflcowitz, D.S. and Skole, D.L. 1991. Changes in the landscape of
Latin America between 1850 and 1985. Forest Ecology and Management, 38: p143-l72.

Kaimowitz, D. and Angelsen, A. 1997. Economic models of tropical deforestation: a
review. Bogor (Indonesia) Centre for International Forestry Research (CIFOR).

Kaimowitz, D. and Angelsen, A., 1999. Rethinking the Causes of Deforestation: Lessons
from Economic Models. The World Bank Research Observer, 14(1):p73-98.

Krug, T. 2000. Deforestation and Fire in the Brazilian Amazonia: a historical perspective.
In. LBA First Scientific Conference. Belem, Para. June 26-30, 2000. p3.

Lambin, ER 1994. Modelling deforestation processes: a review. TREES Publications
Series B: Research Report 1. European Commission Joint Research Centre, Ispra.

Laurance, W.F., Cochrane, M.A, Bergen, S., Fearnside, P.M., Delaménica, P., Barber, C.,
D'Angelo, S., Fernandes, T., 2001. The Future of the Brazilian Amazon, Science, Jan 23,
2001

Leinbach, TR. and Smith, A. 1994. Off-Farm Employment, Land, and Life Cycle:
Transmigrant Households in South Sumatra, Indonesia. Economic Geography 70(3):273-
295.

Malingreau, JP. and Tucker, C.J. 1988. Large-scale deforestation in the southeastern
Amazon basin of Brazil. Ambio l7:p49-55.

Mather, A. 1996. The Human Drivers of Land-Cover change: The Case of Forests. Paper
presented at the Open IGBP/BACH-LUCC Joint Inter-Core Projects Symposium on
Interactions between the Hydrological Cycle and Land Use/Cover. Kyoto, Japan,
November 4-7.

Mather, J.R. and Sdasyuk, G.V. 1991. Global change: geographical approaches. Tucson:
University of Arizona Press.

Matricardi, E.A.T., Skole, D.L., Chomentowski, W., Cochrane, M.A., 2001. Multi-
temporal detection and measurement of selective logging in the Brazilian Amazon using

180

Landsat data. Working paper at Basic Science and Remote Sensing Initiative, Michigan
State University.

Mausel, P., Wu, Y., Li, Y., Moran, BF, and Brondizio, ES. 1993. Spectral identification
of successional stages following deforestation in the Amazon. Geocarto International
4:p6l-71.

Meyer, W.B. and Turner II, B.L. (eds) 1994. Changes in Land Use and Land Cover: A
Global Perspective. Cambridge University Press, Cambridge.

Meyer, W.B. and Turner 11, BL. 1992. Human Population Growth and Global Land-
Use/Cover Change. Annual Reviews in Ecology and Systematics 23: p39-61.

Moran, E. F. (1999). “Rainforest cities: Urbanization, development, and globalization of
the Brazilian Amazon, (1997) by J .O. Browder, B.J. Godfrey.” American Anthropologist
101(3): p645-647.

 

Moran, E.; Brondizio, E.; Mausel, P. and Wu, Y. 1994. Integrating Amazénian
vegetation, land use, and satellite data. BioScience 44(5): p329-338.

Moran, E.F., Packer, A., Brondizio, E., and Tucker, J. 1996. Restoration of Vegetation
Cover in the Eastern Amazon. Ecological Economics 18: p41-54.

Nepstad, D. C., A. Verssimo, A. Alencar, C. Nobre, E. Lima, P. Lefebvre, P. Schlesinger,
C. Potter, P. Moutinho, E. Mendoza, M. Cochrane, and V. Brooks. 1999. Large-scale
impoverishment of Amazonian forests by logging and fire. Nature 398: p505-508.

Panayotou, T. and S. Sungsuwan. 1994. “An Econometric Analysis of the Causes of
Tropical Deforestation: The Case of Northeast Thailand.” In Causes of Tropical
Deforestation. (Pearce and Brown, eds.), p192-210, London: UCL Press.

Parayil, G. and Tong, F ., 1998. Pasture-led to logging led deforestation in the Brazilian
Amazon: the dynamics of socio-economic change. Global Environmental Change, 8(1):
p63-79.

Parks, P. 1991. Models of Forested and Agricultural Landscapes: Integrating Economics.
In: Turner, M. G. and Garden, R. H. (eds) 1991. Quantitative Methods in Landscape
Ecology. Springer-Verlag, New York.

Perman, R., Ma, Y. and McGilvray J ., 1996. Natural Resource and Environmental
Economics. Addison Wesley Longman Publishing, New York.

Pfaff, A.S.P., 1997. What drives deforestation in the Brazilian Amazon? Evidence from
satellite and socioeconomic Data. The Environment, Infrastructure, and Agriculture
Division, Policy Research Department, World Bank.

 

Pritchard, Jr. L.; Geoghegan, J .; Sanderson, S. and Turner 11. B. L. 1996. Socialising The
Pixel and Scalar Themes in Land-use/Cover Change (LUCC). Paper resented at the
Workshop on "People and Pixels: Using Remotely Sensed Data in Social Science

181

Research", Committee on the Human Dimensions of Global Environmental Change,
National Research Council (U 8), November 12th-l3th, 1996.

Reis, EL and Margulis, S. 1991. Options for Slowing Amazon Jungle Clearing. In
Dombusch, R. and Poterba, J .M. Global Warming: economic Policy Responses, the MIT
Press, Cambridge.

Reis, J .R. and Guzman, RM. 1992 An Econometric Model of Amazon Deforestation.
Texto Para Discussao No 265.

Salati, E. and RB. Vose. 1984. Amazon Basin: a System in Equilibrium. Science.
225:129-138.

Saldarriaga, J. G., West, D. C., Tharp, M. L., and Uhl, C. 1988. Long-Term
Chronosequence of Forest Succession in the Upper Rio Negro of Colombia and
Venezuela. Journal of Ecology 76:p938-958.

Salomao, R., Nepstad, D., and Vieira, I. C., 1996, Biomassa de florestas tropicais e efeito
estufa. Ciéncia Hoje 21(123).

Scatena, F.N., Walker, R.T., Homma, A.K.O., Conto, A.J.D., Ferreira, C.A.P., Carvalho,
R.D.A., Rocha, A.C.P.N.D., Santos, A.I.M.D. and Oliveira, P.M.D., 1996. Cropping and
fallowing sequence of small farmers in the “terra firme” landscape of the Brazilian: a

case study from Santarem, Para. Ecological Economics, 18 (1996): p29-40.

Schweik, CM. and Green, G.M., 1999. The Use of Spectral Mixture Analysis to Study
Human Incentives, Actions and Environmental Outcomes. Social Science Computer
Review, 17(1): p40-63.

Schmink, M. and Wood, C. H., 1992. Contested Frontiers in Amazonia. New York:
Columbia University Press.

Shukla, J ., C. Nobre, and P. Sellers. 1990. Amazon Deforestation and Climate Change.
Science. 247:1322-1325.

Skole, D.L. and Tucker, C. 1993. Tropical Deforestation and Habitat Fragmentation in
the Amazon: Satellite Data from 1978 to 1988. Science 260: pl905-1910.

Skole, D.L.; Chomentowski, W.H.;Salas, WA. and Nobre, AD. 1994. Physical and
Human Dimensions of Deforestation in Amazon. BioScience, 44(5):p314-321.

Skole, D.S. and Tucker, G]. 1993. Tropical deforestation and habitat fragmentation in
the Amazon: Satellite data from 1978 to 1988. Science 260:p1905-l910.

Southgate, D., 1994. Tropical deforestation and agricultural development in Latin
America,” in Katrina Brown and David W. Pearce (eds.), The Causes of Tropical
Deforestation. London: University College London Press, 1994.

Steininger, M. K. 1996. Tropical secondary forest regrowth in the Amazon: age, area, and
change estimation with Thematic Mapper data. International Journal of Remote Sensing
172p9-27.

182

Steininger, M. K., 1996. Tropical Secondary Forest Regrowth in the Amazon - Age, Area
and Change Estimation with Thematic Mapper Data. International Journal of Remote
Sensing 17(1): p9-27.

Stern, D.I., Common, MS, and Barbier, ES, 1996. Economic growth and
environmental degradation: The environmental Kuznets curve and sustainable
development. World Development 24(7), p1 151-1 160.

Tucker, C.J., Holben, B.N., and Goff, TE. 1984. Intensive forest clearing in Rondonia,
Brazil as detected by satellite remote sensing. Remote Sensing of Environment 15:p255-
261.

Tumer H, B.L.; Skole, D.; Sanderson, S.; Fischer, G.; Fresco, L. and Leemans, R. 1995.
Land-Use and Land-Cover Change Science/Research Plan. IGBP Report No. 35 and
HDP Report No. 7. International Geosphere-Biosphere Programme and the Human
Dimensions of Global Environmental Change Programme, Stockholm and Geneva.

Turner H , B. L., R. H. Moss, and D. L. Skole, eds. 1993. Relating land use and global
land-cover change: A proposal for an IGBP-HDP core project. Report from the IGBP-
HDP Working Group on Land-Use/Land—Cover Change. Joint publication of the
International Geosphere-Biosphere Programme (Report No. 24) and the Human
Dimensions of Global Environmental Change Programme (Report No. 5). Stockholm:
Royal Swedish Academy of Sciences.

Turner, B. L., 1997: Socializing the Pixel in LUCC, LUCC Newsletter, No. l, p. 10-1 1.

Turner, M. 1990. Spatial and temporal analysis of landscape patterns. Landscape Ecology
4: p21-30.

Turner, M.G.; Dale, V.H. and Gardner, RH. 1989. Predicting Across Scales: Theory
Development and Testing. Landscape Ecology 3: p245-52.

Uhl, C.; Kaufinan, J. B. 1990. Deforestation, fires susceptibility, and potential tree
responses to fire in the Eastern Amazon. Ecology, 71(2): 437-449.

Verissimo, A., Barreto, P. et al., 1992. Logging Impacts and Prospects for Sustainable
Forest Management in an Old Amazonian Frontier - the Case of Paragominas. Forest
Ecology and Management 55(1-4): 169-199.

Vieira, I., Saloméo, R., Rosa, N., Nepstad, D., and Roma J. 1996. O Resascimento da
Floresta no Rastro da Agricultura. Ciéncia Hoje 20(119).

Walker, R. 1999. The structure of uncultivated wilderness: Land use beyond the
extensive margin. Journal of Regional Science 33:p3 87-410.

Walker, RT. and Homma, A. 1996. Land Use and Land Cover Dynamics in the Brazilian
Amazon: An Overview. Ecological Economics, Vol. 18 (1), p. 67-80.

183

Walker, R., Moran, E. and Anselin, L., 2000. Deforestation and cattle ranching in the
Brazilian Amazon: external capital and household processes. World Development, 28(4):
p683-699.

Walker, RT. 1993. Deforestation and Economic Development. Canadian Journal of
Regional Science, Vol. 16(3), p. 481-497.

Wood, C .H. and Perz, S. 1996. Population and land use change in the Brazilian Amazon.
In: S. Ramphal, and S. Sindig (eds). Population Growth and Environmental Issues.
Westport, Conn. Praeger.

Wood, CH. and Skole, D., 1998. Linking satellite, census and survey data to study

deforestation in the Brazilian Amazon. In “People and Pixels: linking remote sensing and
social science’. National Academy Press, Washing DC, 1998.

184

CHAPTER 5

Conclusions

This dissertation work aims to monitor and model the interlinkages of land use/land cover
change as well as its relationship with the human socio-economic driving forces using
GIS, remote sensing and various modeling techniques. It is designed to better understand
the complexity of causes and mechanisms governing land use/land cover changing
processes, and to investigate the environmental impacts of socio-economic activities and

to quantify the built-up process of landscape change.

The interactions between different land use types as shown on the LUCC triangle call for
a need to conceive a general modeling framework of land use/land cover change to
integrate several model components which account for the main categories of land
use/land cover change at different stages of development, for example, deforestation
driven by agricultural expansion, forest regrowth due to land abandonment, deforestation
at urban fringe, farmland loss as the result of urban sprawl etc. In each geographical
context at a certain timing stage, one or more components of the model would take
prominence on the others, to reflect the dominant driving forces and resulted land

use/land cover change types in that region (Lambin, 1994).

The East Region and the Shanghai Municipality of China are selected to be the cases for
studying the change of agricultural landscape in highly urbanized areas, while the
Brazilian Amazon Basin as well as the State of Rondonia in Brazil is a representative

case to study the consequences of agricultural activities on a primarily forested

185

landscape. Remote sensing technique is applied to detect the magnitude and pattern of
land use/land cover change, and firture land use/land cover change are projected by taking
into account the factors that cause or control these changes, with an ultimate objective to
provide scientific projections of land use/land cover change as reference to future policy

design in developing countries.

There are various types of model that have already existed in land use/land cover change
research literature. We need to decide what type of model is the best to simulate the
pattern and processes of land use/land cover change, and how to incorporate both the
spatial and temporal dynamics for a model to be predictive across space and over time. In
terms of modeling method choice, the necessary properties and requiem for a good model
have been known as transparency, robustness, reasonable data needs and appropriate
spatial-temporal resolution and inclusion of key policy variables as outlined by Lee

(1973)

There are three basic types of land use/land cover change models being used in this
dissertation work: analytical model, statistical/empirical model and systematic model.
Each modeling method has its own advantages and limitations. Analytical models are
abstract, theoretical constructs with no empirical data; but they can provide an important
supplement to abstract propositions because it has more ability to map the complex
causal connections, including feedback loops and reciprocal effects etc. (Turner et a1
1991). Therefore analytical models have been developed in each case study of this

dissertation work.

186

Statistical/empirical models quantify the relations between variables based on empirical
historical or cross-sectional data. This approach has standardized methods and is easy to
apply and interpret. It is a powerful tool to identify the key driving force variables with
the strongest association with the LUCC process under study. A multinomial logit model
is constructed for the agricultural land use change study in the East Region of China, with
both socioeconomic and biophysical variables included in the model. In the case of urban
sprawl study in Shanghai, China, a principal component regression method is applied to
explore the multidimensional relationships between the demographic and socio-economic
driving forces of land use change. No biophysical variables are included in the model
because of the relatively homogenous landscape characteristics of a small plain region

like Shanghai.

However, most statistical/empirical models are intrinsically descriptive. They can only
provide insights into the empirical relationships over a system’s history or at a particular
point in time, but is of limited use for analyses of a system’s future development under
alternative management schemes (Agarwal et al, 2000). So we have to be very cautious to
extrapolate beyond the range of the original observations. Another limitation of the
statistical/empirical approach is that it does not adequately account for the complexity of,
and relationships among, socioeconomic and environmental factors; its explanatory
variables are treated as exogenous factors fi'om outside of the land use/land cover change

process, and rarely is there any internal interaction or feedback explicitly included in the

187

 

 

models. Therefore, this type of models is limited in the power to adequately explain or

predict the nonlinear, dynamic complexity of the land use/land cover change processes.

Instead, a systematic model is developed for the case study of Brazilian Amazon basin
and the State of Rondonia. Systematic models are useful tools to simulate the dynamic
structures and behaviors of complex systems with interactions, feedback loops and
reciprocal effects. Unlike the empirical statistical models that focus on specific process
affected by a defined set of variable, systematic models examines land use/land cover
change as one component of an economic-ecological system. Models are built on the
belief that the processes and mechanisms by which a system operates in a given situation
have been well understood; building a systematic model is essentially to formalize and
articulate, in a comprehensive manner, the causal/effect relationships and the system’s
operating mechanism (Lambin, 1994). The systematic modeling approach has an
advantage over statistical/empirical modeling method because it builds into the
representation of a phenomenon those aspects of a system that we know actually exists. It
doesn’t rely on historic data to reveal statistical or empirical relationships. It is desirable
to introduce the mechanisms of human social economic activities directly into a model of
land use/land cover change, and to build an integral systematic model. This advantage
also allows dynamic models to be used in more applications than empirical models.
Systematic models are often transferable to new applications because the fundamental
concepts on which they are built are present in many other systems as well (Agarwal et
al, 2000). However, rich data sets are often required for a systematic model to deal with

multidimensional feedbacks among system components. Another challenge of systematic

188

modeling approach is the difficulty in deciding how to incorporate model complexity

without jeopardizing either model feasibility or model reality.

Like all other environmental problems, land use/land cover changes are deep-rooted in
development. While it is development that causes most of the environmental problems, it
is also development that can solve them. The only possible answer is sustainable
development. The challenge of achieving sustainable development in the context of
global environmental change is particularly acute for developing countries such as Brazil
and China, two of the world’s largest nations under increasing pressure to modify their
development strategies so as to reduce the adverse impacts of global environmental
change. Scientific modeling projections of land use/land cover change can provide
important reference to future policy design and land management in these countries. Land
use model of the East Region of China show a rapid pace of cultivated land loss in the
next half century, therefore, raise serious concerns for the Malthusian type question of
whether China can feed itself. Urban sprawl in Shanghai will take over more agricultural
and forest area in Shanghai municipality, threatening the region’s food security and
ecological stability. The problem of tropical deforestation in the Brazilian Amazon has
been a primary focus of the international environmental research community. Results of
the systematic model of land use/land cover change in the Brazilian Amazon demonstrate
an accelerating speed of deforestation in the coming decades. Nearly half of the standing
forest will be cleared within next fifty years; all the primary forest in the state of
Rondonia, a frontier state of agricultural development, will disappear from sight within

less than forty year. Projections of annual deforestation and secondary growth in the

189

Brazilian Amazon can bring more insights into the missing carbon sink of the global

carbon cycle.

References:

Agarwal, C., Green, G.L., Grove, M., Evans, T. and Schweik, C., 2000. A Review and
Assessment of Land-Use Change Models: dynamics of space, time and human choice.
Center of the Study of Institutions, Indiana University.

Lambin, ER 1997. Modelling and monitoring land-cover change processes in tropical
regions. Progress in Physical Geography, 21: (3), p375-393.

Lee, J .D.B., 1973. Requiem for Large-Scale Models. AIP Journal, May 1973. p163-77.

190

 

k3\1