|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 LINKING SOCIAL-ECOLOGICAL SYSTEM S AND LAND-USE LA ND-COVER CHANGE THROUGH A COMPLEX ADAPTIVE SYSTEMS APPROACH: A CROSS-BORDER STUDY OF SISAKET, THAILAND AN D ORDAR MEAN CHEY, CAMBODIA By LIN CASSIDY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007
2 2007 Lin Cassidy
3 To the memory of my father, Donn Cassidy I like to think he would have been proud of me
4 ACKNOWLEDGMENTS This dissertation would not have been possibl e without the interest and support of Mike Binford, my advisor, and the rest of my comm ittee: Jane Southworth, Grenville Barnes and Graeme Cumming. I was fortunate to have a co mmittee who encouraged me to push my limits, while still having fun with both th e ideas and the process. I am very appreciative of how each person helped me out exactly where their input was most needed. Thank you all. In Cambodia, I am indebted to the Center fo r Khmer Studies, in particular Ayrine Uk, for logistical support, an institutional base, and a place to work while in Siem Reap. Tonh and Chinda, and the University of Chicago Projec t field teams helped smooth out logistical arrangements and made me feel most welcom e. Without the support of APSARA, access to certain areas would not have been possible, and for their help in this regard I am very grateful. Mr Ith Sothar, director of the Cambodian Department of Geography was of great assistance with regard to GIS data and metadata. Jan-Pete r Mund at the Royal University of Phnom Penh generously provided me with information and data relating to Ordar Mean Chey province. I am further indebted to him and his students Nang Sinuon, Phieu Sayon, Sokny and Soveat for their assistance in deriving historical roads layers for my study area. Ian Thomas at the Cambodia Mine Action Agency also helped me immeasurably with regard to secondary data. Chan Sovann patiently and safely drove me along almost every tr ack that it was safe to travel in Ordar Mean Chey. His insights along the way gave me a fa r greater understanding th an I would otherwise have had. I am also grateful to HALO es pecially Chorn Nak and Tom Dibbs for providing me with additional data relating to roads and trac ks, and the location of landmines in Ordar Mean Chey. Staff at the Mekong River Commission, including the national committees in Thailand and Cambodia, generously gave of their time, knowledge and data. In particular, I thank Min Bunnara, Harald Kirsch, Huon Rath, Christoph Fe ldkoetter, and Khun Winai in this regard.
5 Toby Jackson and Chhim Phet in Siem Reap both provided enormous amounts of information, and gave me insights and a sense of context to my field experiences. In Thailand, the Thai Family Research Centre made sure that their na tions reputation for hospitality remained unblemished, with many of the staff opening their homes and families to me. My field work in Sisaket would not have been possible without the various efforts of Khun Sombat Sakuntasathien, Khun Pavisanat Pathomcharoensukchai, Khun Tipatavee Pathomcharoensukchai, Khun Natchayakorn San gkorntiraput, Khun Somkit Heetchana and the centres field staff in Sisaket and Buriram provi nces. Similarly, the University of ChicagoUTCC Research Centre at the University of Thai Chamber of Commerce and the UTCC-Chicago were most welcoming, and provided me with bot h accommodation and office space. I thank the director, Dr Vimut Vanichar earnth, as well as Khun Wich ian Kaeosombat, Teerachat Techapaisarnsaroenkit, Surayose Sricharoen, Pinyo Supakatisun, Kitti Chiewchan who helped with letters of introduc tion, translations, access to datasets, computer support and a sense of place. I thank Apiradee Bhatarakarnt for acting as an interpreter for several interviews with different government departments in Krungthep, not to mention the many government agencies and officers themselves who went out of their way to help me with data: Khun Sitichai from the Department of Highways; Anuchit Ratanasuwa n at the Thailand National Park Wildlife and Plant Conservation Department; Korn Manussr isuksi and Sakuna Wisuttiratanakun at the Thailand Royal Forestry Division; Dr Nipa Sujarae, director of the Department of Public Lands Management; staff at the Department of Rura l Roads; and last but by no means least, Khun Wanna and Ratchanee Dungchim at the Land Titling Project Office. I also thank Risa Patarasuk and her family Uthai and Waranya Patarasuk, a nd Waraporn Rewan for their hospitality. Risa
6 also helped me with fieldwork in Ordar Mean Chey, and with obtaining secondary data in Krungthep. During my stays in Ban Isay in Sisaket, Pha Kii and Ma y made sure I felt part of the family. I am very grateful for how they cut ac ross language barriers to ma ke sure I felt at home. My fieldwork expenses, as well as the purchas e of the satellite imagery used in this research, was largely covered by the NSF-funde d Economic Growth, Social Inequality, and Environmental Change in Thailand and Cambodia project (BCS-0433787). I am very grateful to Rob Townsend and Alan Kolata at the Univ ersity of Chicago, for their support both through this project and through their conne ctions in Thailand and Cambodia. I also thank John Felkner for all his efforts in making data available to me. At the University of Florida, I am very grateful to the Department of Geography especially to the Chair, Pete Waylen, and to Jane Southworth. During my time as a geography student they turned the depart ment around, and made it a base and home department to be proud of. In addition, Janes quick responses and help with my never-ending co mputer issues meant I never felt alone or in too much danger of losing my sanity. Tim Fiks help and advice with all things statistical turned several mysteries into strong arguments. I am thankful to Julia Williams, Desiree Price and Rhonda Black for always ensu ring that any administ rative concerns were immediately taken care of. I also thank th e Tropical Conservation and Development Program, both for funding support during my dissertation and for providi ng an unparalleled sense of community. Special thanks go to Jon Dain, Ka ren Kainer, Hannah Covert and Bob Buschbacher, who all not only get it, but give it too! The International Centre deserves the highest praise for understanding and accommodating the specific needs of international students. Maud Fraser made a lot of the challenges much easier to deal with.
7 The Land Use and Environmental Change Ins titute provided me with permanent access to state-of the-art computer facilities, a desk with a view of green trees and bl ue skies, financial and technical support needed for my research, and a community of fr iendship, brains, advice, life philosophy, support and a lot of good memories. This is where most of this dissertation happened but it would not have happened at all without Mark Brenner, Christian Russell, Joel Hartter, Andres Guhl, Amy Dani els, Gaby Stocks, Forrest Stev ens, Matt Marsik, Hector Castaneda, Natalia Hoyos and Mi riam Wyman. Equally important in this regard are the other members of my Gainesville family: Wendy-Lin Ba rtels, Franklin Paniagua, Alfredo Rios, David Buck, Ellie Harrison, John Engels, Christine Arch er, Paul Ghiotto, Camila Pizano, Joe Veldman, Anna Prizzia, Kelly Biedenweg, Katie Painter, and all the others who made my life rich and full. I can only hope that one day I can be around for ot hers the way these friends were there for me. The members of my extended family back home have no idea how important it was to me that they never lost faith even when I di d. For this I thank my mother, Penny Cassidy, and my sister, Kim Webber; as well as Robyn Sheldon, Charles Sheldon, Hugh and Eve MurrayHudson, and the rest of the Gabs crew; Birgit an d Reiner Khler; Marc Baar and Yvonne WardSmith; Hannelore and Helmut Bendsen; Ruth Blyther; and Amy Sullivan.
8 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ........11 LIST OF FIGURES................................................................................................................ .......13 ABSTRACT....................................................................................................................... ............16 CHAPTER 1 INTRODUCTION..................................................................................................................18 Statement of the Problem....................................................................................................... .18 Linking Research in Social-Ecological Sy stems and Land-Use Land-Cover Change: A Complex Adaptive Systems Approach...............................................................................20 Patterns of Land-use Land-Cover Diversity as Complex Adaptive System Components: A Cross-boundary Comparison of Two Pr ovinces in Thailand and Cambodia..................20 Social and Ecological Factors and Land-use Land-cover Diversity in Two Provinces in South-east Asia................................................................................................................ ...21 Importance of the Study........................................................................................................ ..22 2 LINKING RESEARCH IN SOCIAL-E COLOGICAL SYSTEMS AND LAND-USE LAND-COVER CHANGE: A COMPLEX ADAPTIVE SYSTEMS APPROACH............23 Introduction................................................................................................................... ..........23 Social-ecological Systems......................................................................................................25 Challenges to SESs Research.................................................................................................27 Land-use Land-cover Change.................................................................................................28 Challenges to LULCC Theory................................................................................................32 Integrating SES and LULCC Approaches..............................................................................33 Complex Adaptive Systems....................................................................................................35 Expressing LULCC in Terms of CAS Concepts.............................................................41 Focus on Diversity...........................................................................................................46 Testing LULC Diversity..................................................................................................47 Methodology for Evaluating LULC Diversity.......................................................................50 Spatially and Temporally Explicit Studies......................................................................51 Cross-border Studies.......................................................................................................52 Appropriate Levels and Scales of Analysis.....................................................................52 Exploring Patterns of LULC Diversity a nd its Response to Factors Influencing Change.........................................................................................................................54 A Preliminary Assessment of the LULC Diversity Concept..................................................61 Conclusions.................................................................................................................... .........65
9 3 PATTERNS OF LAND-USE LAND-COVE R DIVERSITY IN SISAKET, THAILAND AND ORDAR MEAN CHEY, CAMBODIA...................................................69 Introduction................................................................................................................... ..........69 LULC Diversity...............................................................................................................70 Broad Research Goal.......................................................................................................71 Specific Research Objective............................................................................................72 Study Area..................................................................................................................... ..74 Methods........................................................................................................................ ..........77 Study Design...................................................................................................................77 Determination of SES Components.................................................................................78 Calculation of Diversity fo r Fixed Points in Time..........................................................79 Change in Diversity over Time.......................................................................................81 Results and Discussion......................................................................................................... ..81 LULC Classifications......................................................................................................81 Frequency Distributions of LULC Diversity...................................................................83 Spatial Distribution of LULC Diversity..........................................................................91 LULC variety...........................................................................................................91 LULC relative abundance........................................................................................92 Change in Spatial Distribution of Diversity over Time...................................................95 Temporal Trends in LULC Variety and Variation in LULC Relative Abundance.........96 Further Studies...............................................................................................................103 Conclusion..................................................................................................................... .......105 4 SOCIAL AND ECOLOGICAL FACTORS AND LAND-USE LAND-COVER DIVERSITY IN TWO PROVINCES IN SOUTH-EAST ASIA.........................................107 Introduction................................................................................................................... ........107 Research questions........................................................................................................109 Study Area.....................................................................................................................111 Methods........................................................................................................................ ........115 Development of Datasets...............................................................................................115 Scale.......................................................................................................................... ....117 Analysis....................................................................................................................... ..118 Results........................................................................................................................ ...........120 Spatial Distribution of LULC Dive rsity and Explanatory Variables............................120 Distance to Roads and LULC Diversity........................................................................125 Distance to Markets and LULC Diversity.....................................................................129 Elevation and LULC Diversity......................................................................................133 Discussion..................................................................................................................... ........137 Conclusion..................................................................................................................... .......141 5 CONCLUSION.....................................................................................................................143 Significance of Findings.......................................................................................................144 LULC Diversity................................................................................................................. ...145 Complex Adaptive Systems, LULCC,and SESs..................................................................146
10 Further Work................................................................................................................... .....146 LIST OF REFERENCES.............................................................................................................148 BIOGRAPHICAL SKETCH.......................................................................................................162
11 LIST OF TABLES Table page 2-1 Selected examples of case studies su ggesting changes in LULC diversity and underlying factors............................................................................................................. .49 2-2 Quantifiable aspects of LULC dive rsity, indicating hypothesized responses....................58 3-1 Levels of analysis of LULC diversity................................................................................80 3-2 Error matrix and producers and user s accuracy for accuracy assessment of 2005 classification based on 172 reference points......................................................................82 3-3 Central tendencies of LULC variet y at the micro-scale (3x3 pixel window)....................85 3-4 Central tendencies of LU LC relative abundance (Simpsons index) at the micro-scale...85 3-5 Central tendencies of LULC variety at the meso-scale (33x33 pixel window).................86 3-6 Central tendencies of LU LC relative abundance (Simpsons index) at the meso-scale....87 3-7 Central tendencies of LULC variety at the macro-scale (303x303 pixel window)...........88 3-8 Central tendencies of LULC relative abundance (Simpsons index) at the macroscale.......................................................................................................................... ..........88 3-9 Comparison of mean LULC variety at three scales...........................................................89 3-10 Comparison of mean LULC rela tive abundance at three scales........................................89 4-1 Error matrix and producers and user s accuracy for accuracy assessment of 2005 classification based on 172 reference points....................................................................115 4-2 Scales of Analysis of LULC Diversity............................................................................117 4-3 Spearmans correlation coefficients showi ng LULC diversity in response to distance to roads at each time-step.................................................................................................125 4-4 Coefficients used to predict LULC divers ity at four time-steps in Sisaket and Ordar Mean Chey in response to distance to roads....................................................................127 4-5 Results of the models predicting LULC di versity in response to distance to roads........127 4-6 Spearmans correlation coefficients show ing LULC diversity in response to costweight distance to markets at each time-step...................................................................130 4-7 Coefficients used to predict LULC divers ity at four time-steps in Sisaket and Ordar Mean Chey in response to cost-weight distance to market..............................................131
12 4-8 Results of the models predicting LULC di versity in response to cost-weight distance to market...................................................................................................................... ....131 4-9 Spearmans correlation coefficients showi ng LULC diversity in response to elevation at each tim e-step..............................................................................................................133 4-10 Coefficients used to predict LULC diversity in response to elevation............................135 4-11 Results of the models predicting LULC diversity in response to elevation....................135
13 LIST OF FIGURES Figure page 2-1 Schematic diagram showing how the dive rgent SESs and LULCC approaches to the study of human-environment inte ractions can be linked...................................................34 2-2 Schematic diagram of Holling et al. s adaptive cycle (2002) showing path dependency..................................................................................................................... ...40 2-3 Representational diagram showing different temporal and spatial extents of selected SES processes (based on ideas in Holling, Gunderson, and Peterson 2002).....................53 2-4 Schematic diagrams representing a) d) the distribution of LULC diversity at different scales, and e) f) the interacti on of LULC diversity with slope and soil...........59 2-5 Schematic diagrams representing g) h) the interaction of LULC diversity with distance from roads and market, i) j) ch anges in LULC diversity over time, and k) l) changes in the relationships betw een LULC diversity and slope and soil..................60 2-6 Schematic diagrams representing m) n) changes in the relationships between LULC diversity and distance from roads and market........................................................61 2-7 Location of Trapeang Prasat village in northern Cambodia..............................................62 2-8 A test of LULC diversity on a) the 12-c lass classification of the Trapeang Prasat area, showing b) micro-scale divers ity and c) meso-scale diversity..................................63 2-9 Comparison of a) actual LULC diversity to null models of b) random diversity and c) a homogeneous landscape..................................................................................................65 3-1 Schematic diagram showing how the dive rgent SESs and LULCC approaches to the study of human-environment inte ractions can be linked...................................................72 3-2 Schematic diagram representing the dist ribution of LULC dive rsity at different scales......................................................................................................................... .........73 3-3 Study area map showing the different spectr al characteristics of Sisaket, Thailand and Ordar Mean Chey, Cambodia.....................................................................................75 3-4 LULC classifications for Sisaket, Thaila nd and Ordar Mean Chey, Cambodia at each time-step in the study period..............................................................................................83 3-5 Frequency distributions of LULC types in 1989, 1994, 2000 and 2005 at the microscale.......................................................................................................................... ..........85 3-6 Frequency distributions of LULC types in 1989, 1994, 2000 and 2005 at the mesoscale.......................................................................................................................... ..........86
143-7 Frequency distributions of LULC types in 1989, 1994, 2000 and 2005 at the macroscale.......................................................................................................................... ..........88 3-8 Spatial distribution of LULC variety in 2005, shown for the three different scales and in comparison to the initial classification..........................................................................91 3-9 Spatial distribution of LULC rela tive abundance in 2005, shown for the three different scales and in comparis on to the initial classification..........................................93 3-10 Spatial distribution of LULC variety in 1989, 1994, 2000 and 2005, shown at the meso-scale..................................................................................................................... .....95 3-11 Scatterplots of mean LU LC variety vs. variance of LULC variety for the combined study area, comparing values for actual and extensive traject ories against all theoretical potential trajectories.........................................................................................97 3-12 Spatial distribution of trajectories of cha nge in diversity derived for each scale from that scale's four time-steps of LULC variety.....................................................................99 3-13 Spatial distribution of tem poral variation in LULC relative abundance, calculated for each scale as the standard deviation of the values for each time-step at that pixel.........101 4-1 Schematic diagram showing how the dive rgent SESs and LULCC approaches to the study of human-environment inte ractions can be linked.................................................108 4-2 Hypothesized relationships between LULC diversity and a) distance to roads, b) distance to market, and c) elevation.................................................................................111 4-3 Study area map showing the different spectr al characteristics of Sisaket, Thailand and Ordar Mean Chey, Cambodia...................................................................................112 4-4 LULC classifications for Sisaket, Thaila nd and Ordar Mean Chey, Cambodia at each time-step in the study period............................................................................................113 4-5 Distribution of 1000 random sample points (500 per province) shown relative to the 2005 classification and microscale LULC diversity.......................................................119 4-6 LULC variety at the meso-scale for 1989, 1994, 2000 and 2005....................................121 4-7 Spatial distribution of LULC variety in 2005, shown for the three different scales of analysis and in comparison to the initial classification....................................................122 4-8 Distance to roads in 1989, 1994, 2000 and 2005.............................................................123 4-9 Cost-weight distance to market in 1989, 1994, 2000 and 2005.......................................124 4-10 Elevation in the study area shown together with meso-scale LULC diversity in 2005...125 4-11 Predicted LULC diversity at the microscale in response to distance to roads...............126
154-12 Residual map showing the difference be tween actual LULC diversity and that predicted by distance to roads..........................................................................................128 4-13 Predicted LULC diversity at the micro-sc ale in response to cost-weight distance to market......................................................................................................................... .....130 4-14 Residual map showing the difference be tween actual LULC diversity and that predicted by cost-weigh t distance to market....................................................................132 4-15 Predicted LULC diversity at the mi cro-scale in response to elevation............................134 4-16 Residual map showing the difference be tween actual LULC diversity and that predicted by elevation......................................................................................................136
16 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy LINKING SOCIAL-ECOLOGICAL SYSTEM S AND LAND-USE LA ND-COVER CHANGE THROUGH A COMPLEX ADAPTIVE SYSTEMS APPROACH: A CROSS-BORDER STUDY OF SISAKET, THAILAND AN D ORDAR MEAN CHEY, CAMBODIA By Lin Cassidy December 2007 Chair: Michael W. Binford Major: Geography The social-ecological systems (SESs) and la nd-use land-cover cha nge (LULCC) research programs represent two responses to our n eed to understand the complexity of humanenvironment interactions that are driving change at a global level. SESs research has, however, struggled to find ways to quantify broad-scale system processes, while LULCC research has been unable to express its patterns of change in a way that allows cross-site comparison. This dissertation argues that complex adaptive systems (CASs) theory can be used to link SESs and LULCC research. It shows that LULCC, when cons idered as the tangible expression of an SES, presents an opportunity to measure and quantify the effect of system processes. CASs theory, as used in SESs research, provides the abstrac tion necessary for extrap olating and predicting LULCC, even in highly dissimilar systems. The concept of diversity, among other CAS characteristics, can be applied to LULCC, allowi ng systems to be represented as landscapes of land-use land-cover (LULC) divers ity, which can then be mapped, measured and contrasted in different study sites. This research introduces the concept of LULC diversity, as embedded in the framework of CASs, and defines the methodology needed to show how LULC diversity has the potential to
17 allow researchers to find commonalities across a range of different social-ecological systems (SESs). A case study of the adjacent, but diffe rent landscapes of Sisaket province, Thailand, and Ordar Mean Chey province, Cambodia is used to test whether the spatial and temporal patterns, as well as the fre quency distributions of LULC diversity provide meaningful information of the landscape. Finally, known f actors influencing LULC change are used to predict change in LULC diversity instead of in actual LULC t ypes, to determine whether there are general trends in the response patterns of LULC diversity regardless of system type. Results show that not only is LULC diversity visible as distinct patt erns on the landscape, but also that these patterns rela te to known features and areas of change regardless of system type. While the two SESs do indeed exhibit the same directionality in the responses of their levels of LULC diversity to th e different factors, the differenc es in their biophysical and socioeconomic characteristics result in different strengths of relati onship. The findings of this research suggest that LULC diversity reflects the dynamic complexity of the human-dominated landscape. It responds in measurable ways to di fferent factors influenci ng LULCC. The spatial arrangement of LULC diversity can be used as a means to quantify change in SESs. As such, it provides a means of abstracting landscapes so th at comparisons can be made between landscapes and SESs that would otherwis e be context-dependent.
18 CHAPTER 1 INTRODUCTION This dissertation intr oduces and tests a theoretical framework and set of methods for uniting two existing research prog rams that address landscape-leve l change. Specifically, I draw on the theory of complex adaptive systems (CASs) that is used in social-ecological systems (SESs) research, and combine it with the methods used in land-use land-cover change (LULCC) studies. This allows the strengths of each prog ram to complement the other. The research is presented as three separate papers, presented in publication style fo r submission to academic journals. Each paper is therefore a stand-alon e document, addressing di fferent aspects of the research problem described below. The first, Quantifying Social-Ecological Systems through Land-Use Land-Cover Change: A Complex Adap tive Systems Approach introduces the two research programs, and shows the applicability of CASs theory. The second paper, Patterns of Land-use Land-Cover Diversity as Complex Ad aptive System Components: A Cross-boundary Comparison of Two Provinces in Thailand a nd Cambodia applies the theory of complex adaptive systems (CASs) generally, and the conc ept of LULC diversity specifically, to the description of landscape change in the cross-border provinces of Sisaket, Thailand, and Ordar Mean Chey, Cambodia. The third paper, S ocial and Ecological Fa ctors and Land-use Landcover Diversity in Two Provinces in South-east Asia tests the response of LULC diversity in Sisaket and Ordar Mean Chey to factors known to influence the distribution of different LULC types. Statement of the Problem The processes affecting the dynamics of soci al-ecological systems tend to be hard to describe, challenging to model, and even more di fficult to quantify and meas ure. This is in part because of the complexity of such systems (O'Neill 1999; Ludwig, Walker, and Holling 2002;
19 Chave and Levin 2002; Gibson, Ostrom, and Ahn 2000; Allen and Starr 1982). More particularly, some of the concepts are hard to translate into tang ible indicators. Resilience, for example, is difficult to quantify because to some extent it is a relativ e condition contingent upon earlier states and future pe rturbations (Carpenter et al 2001; Cumming et al. 2005). Consequently, researchers are struggling to de velop methods to quantif y the within-site and cross-site heterogeneity and complexity inherent in SESs. The challenge then, is to identify what quantitative data could provide a suitable representation of human -environment interactions in a given system. Whereas SESs research developed under the umbrella of a conceptual framework, LULCC studies first document case studies, and then se arch for theoretical models to explain the relationships they have found. Currently, howev er, researchers have yet to advance LULCC studies to a point where these theo ries explain a range of conditions This is largely because the complex nature of the social and ecological inte ractions underlying LULCC is a challenge to the post-hoc development of a universal deductive research approach (Brgi, Hersperger, and Schneeberger 2004). Even within the same LULC system, theories remain context-specific (Perz 2007). Researchers have found that the complexity of interactions makes it difficult to extract general characteristics from their case studies fo r comparison and application in other contexts (Briassoulis 2000; Rindfuss et al 2004). Consequently, LULCC st udies still lack a cohesive, trans-disciplinary framework on which to base th eir questions. With a more solid, generalized theoretical foundation and a bett er understanding of mechanisms, LULCC analyses would be able to provide planners and managers with more reliable predictions.
20 Linking Research in Social-Ecological Systems and Land-Use Land-Cover Change: A Complex Adaptive Systems Approach The first paper, presented in Chapter 2, provi des the conceptual linkages between SESs and LULCC, and details the characte ristics of CASs and how they apply to LULCC. The underlying assumption is that, by treating the patterns of LU LCC as tangible expressions of a SES, we can firstly transfer the concepts of CAS to LULCC, and secondly use this spatially explicit, areabased form of analysis as the basis for quantitative assessments of SESs. In this chapter I describe the current states of SESs and LULCC research, and detail the challenges that each face. I then introduce CASs highlighting key characteristics, and applying these conceptually to the study of LULC. Next one key CAS characteristic diversity is focused on for the development of the research fra mework necessary to test the effectiveness of this approach. Specifically, it addresses the uti lity of the concept of diversity, as embodied in CAS theory, as a way of generalizing the compon ents of different SES/LULC systems so that they can be compared. The framework and met hods detail how researchers can a) describe, map and measure the distribution of LULC diversity across landscapes, b) evaluate the patterns of LULC diversity in response to biophysical and socio-economic factors known to influence LULCC, and c) assess change over time in LU LCC diversity in respons e to those different processes. Patterns of Land-use Land-Cover Diversity as Complex Adaptive System Components: A Cross-boundary Comparison of Two Provinces in Thailand and Cambodia In Chapter 3, I use a cross-border comparison of the study area provinces to test whether LULC diversity has distinct fr equency distribution patterns, a nd whether spatial patterns of LULC diversity are visible in the landscape. I focus on diversity as a key CAS concept and as the main characteristic that underpins the comple xity of human-environment interactions central to both the LULCC and SESs research programs.
21 In order to test LULC diversity as a uni fying, non-ecosystem specific characteristic for cross-site comparison, we need to show it as non-random, presenting distinct patterns on the landscape at different scales, with different ma gnitudes of diversity dominating at different locations, and with meaningful variation in magnitudes of diversity over time. I use the following guiding questions to carry out this analysis: What is the distribution of LULC dive rsity across two different landscapes? What magnitudes of LULC diversit y dominate in each landscape? What happens to the LULC diversity of an area over time? How do the patterns of diversity co mpare between the two landscapes? The distribution of LULC diversity is different for both provinces and at different scales. While there is temporal variati on, the distributions for each province and scale maintain similar patterns over time. The spatial arrangement of the LULC diversity patt erns suggest that the spatial distribution is responding to the influence of drivers such as markets, roads, elevation and slope, and that the response of LULC diversity to these mechanisms should be tested. Social and Ecological Factors and Land-use Land-cover Diversity in Two Provinces in South-east Asia In order for the concept of LU LC diversity to prove itself meaningful in both LULCC and SESs research, it should respond to the same biophysical and socio-economic factors that are known to influence the distribution of individual LULC types. The purpose of Chapter 4 is to test those responses in two provi nces in south-east Asia to see a) whether LULC diversity changes with known factors influenc ing change in indivi dual LULC types, and b) how the actual relationships between LULC diversity and elevat ion, distance to roads and distance to market compare to those hypothesized in Chapter 1. Spearm ans rank correlations were run to establish general relationships for the en tire landscapes at three scal es. Because the explanatory mechanisms are scale-dependent, functions appr oximating the hypothesized distributions were
22 tested against micro-scale LULC diversity onl y. Distribution maps of LULC diversity as predicted by each factor were subtracted from observed LULC diversity, to create maps show where the models worked well or not. The study shows that for the most part, LULC di versity at all scales increases as elevation, distance to roads, and distance to market decreases. At the micro-scale, the distribution of LULC diversity in response to elevation and di stance to market bore a general resemblance to the schematic diagrams of the hypothesized rela tionships. However, LULC diversity did not match the predicted distribution shape in its resp onse to distance to road s, and this hypothesis should be modified. These findings confirm the utility of the LULC di versity concept as a generalization that supports cross-site compar isons between dissimilar landscapes, and a way of integrating SESs and LULCC research. Importance of the Study Together these three papers contribute to th e development of a research framework for the study of human-environment interactions that is founded on both theory and empirical measurement (Brgi, Hersperger, and Schn eeberger 2004; Kuhn 1996; Walker 2004). The approach outlined in this dissertation makes the link between SESs and LULC explicit, and supports the emergence of LULCC analysis as a science (Rindf uss et al. 2004). It also introduces techniques to quantify the generalized system char acteristics that define SESs, and shows how some of these can be measured (Carpe nter et al. 2001; Cummi ng et al. 2005). This research represents initial steps in evaluating CASs as a generalized, quantifiable theory of human-environment interactions. It is hoped that the work presented here will stimulate interest in other researchers to further test and expand on the ideas they contain.
23 CHAPTER 2 LINKING RESEARCH IN SOCIAL-E COLOGICAL SYSTEMS AND LAND-USE LAND-COVER CHANGE: A COMPLEX ADAPTIVE SYSTEMS APPROACH Introduction The purpose of this paper is to bridge tw o existing research programs that address landscape-level change under a single theoretical framework and approach. Specifically, the aim is to draw on the theory of complex adaptive systems (CASs) that is used in social-ecological systems (SES s) research, and to combine it with the methodology used in land-use land-cover change (LULCC) studies. This will allow the strengths of each program to complement the other. By linking the two research programs, LULCC analysis will gain access to the conceptual st rengths used in SESs research, while SESs analysis will be able to use the quantitat ive methods that underpin LULCC research. CAS theory is generalized, and its context-i ndependent characteristics can be used to understand landscape dynamics in different places. Just as species are all different at the level of the organism, but comparable at th e level of organ, tissue, cell or DNA, so too are landscapes. Landscapes differ from each other because the specific combination of components (including people), ecosystems, clim ate, culture, etc. is unique for each but all landscapes will contain these components. The LULCC and SESs research programs can be seen as two views addressing the same issue: both aim to understand the pr ocesses underlying environmental change, and both emphasize the role of humans as agents and recipients of that change. However, each of these views takes a different appro ach and provides a different perspective on human-environment interactions. While LULCC analysis has focused on mapping, describing and explaining changes in the Ea rths cover, SESs research addresses the flows of energy and information between system components (Gunderson and Holling
24 2002; Turner et al. 1995). As a result, each research program tends to comprise a different community of researchers, each us ing different framewor ks and vocabularies, and the collaborative efforts al ong these closely parallel path s have to date been few (Lambin et al. 2001; Schweik, Evans, and Morgan 2005 represent the exceptions). A system can be defined as a set of com ponents and the interactions between those components. A SES considers the way humans in teract with their e nvironment, not only in terms of direct use of food and resources, but also in terms of the policies, cultural practices, and institutions that influence wher e and how we make use of different aspects of the environment. Most of our interac tions occur on the surface of the earth, and our activities leave distinct spatial patterns, such as the geometric blocks of fields or the node-spokes of settlement-roads on the landscape. The changes in these spatial landscape patterns are one of the key things that LULCC researchers measure to determine the extent of human influence on the environment. This paper proposes that by treating the pa tterns of LULCC as tangible expressions of a SES, we can firstly tran sfer the concepts of CAS to LULCC, and secondly use this spatially explicit, area-based form of analysis as the basis for quantitative assessments of SESs. Linking the two approaches is a key step in the search for generalization at fundamental landscape levels. The underlyi ng assumption of this proposed linkage is that, by viewing LULCC as a physical manifesta tion of the system, we should be able to represent change in the cond ition or state of the SES. In this paper I first describe the current states of SESs and LULCC research, and detail the challenges that each face. I then introduce CASs, highlighting key characteristics, and applying these conceptua lly to the study of LULC. I next focus on
25 diversity as the key element of CASs, and as the starting poin t of generalizing landscapes. Finally, I show how patterns of LULC diversity can be mapped and measured, and suggest how the distribution of LULC dive rsity might respond to biophysical and socioeconomic drivers known to affect LULCC. Social-ecological Systems SESs are by nature complex. Even where humans have become separated from direct production and consumption of natura l resources, we effect and are affected by, global climate change and the rapid spread of chemical compounds (Folke et al. 2004; Vitousek et al. 1997). Wherever we live, we are dependent on the services that the environment provides (Costanza et al. 1998; Swift, Izac, and van Noordwijk 2004). The theoretical foundation of the SESs resear ch program builds directly on systems approaches as used by ecologists, but spec ifically includes humans and their sociocultural processes into the system definition. At any one point in time, researchers are studying property rights, land tenure systems, world views, ethics, economics, climate, soils, abundance of organisms, and how all th ese factors affect each other (Berkes and Folke 1998). A key area of research is variability in sy stems. This focus is built on a concern for sustainability as humans, we have an intere st in seeing desired c onditions of certain landscapes continue or persist in the long-term. Over the ye ars, scientists and managers have come to appreciate that short-term fluc tuations form part of a systems long-term identity, and affect not only humans, but also plants and animals. For example, we understand the annual shift from hot, dry summer to cold, wet winter to be part of the identity of a Mediterranean climate. While seasonal variation represents an example of predictable change, the concern for managers is also how to accommodate unpredictable
26 change, and whether a system can remain the same after an external shock or perturbation, or whether it changes so much th at it should now be cons idered to be a new system (Berkes and Folke 1998; Holling and Gunderson 2002). The ability of a SES to persist and maintain its general structure and same functioning in the face of external perturbations is known as its resilience (Carpenter et al. 2001; Holling 1973). It is largely through th is concept that SESs research addresses the issue of sustainable resource use. A simp lified example of this would be the case of the tropical forests in the Brazilian Amazon. Whereas the extraction of a few trees to build homes for indigenous people living in the forest has little impact on the forest, the large-scale clearing for commer cial farming and logging for export markets represents a perturbation that has seen large parts of the forest converted to farmland. Overall, much of the Amazon retains its identity as a tropical forest, but in many areas at a more local level, the identity of fores t has given way. The noti on of resilience is important because it draws attention to the fact that SESs move th rough a range of conditions or multiple stable states and are s ubject to a range of internal and external disturbances. Any management of a SES to maintain a desired state in the long-term needs an understanding of how dynamic and complex the interactions are in the SES under investigation (Berkes and Folke 1998; Folke et al. 2004). Recognition of SESs as CASs is implicit in the concept of resilience, and explicit in the SESs research program (Holling 2001; Janssen, Anderies, and Ostrom 2007; Levin 1999; Walker and Abel 2002). This is because many of the key char acteristics of CASs (as elaborated below) are important factors determining whether a SE S is sustainable and able to persist or not. Some of the factors that might affect whether a system is able to
27 maintain its functioning might be how abrup tly change occurs, whether that change occurs at a local level or a more regional leve l, whether several factors occur at the same time (such as fire, drought and over-grazing), the history of land use over the past few decades, and the individual land-use decisions made by people might lead to a distinct, collective agricultural landscape (Cummi ng and Collier 2005; Holling and Gunderson 2002; Janssen, Anderies, and Ostrom 2007; Levin 2003; Odum 1975). The use of CAS theory to understand SESs has been applie d in a range of settings, with notable case studies in lake districts, ocean fisheries, and savanna rangelands (Carpenter et al. 2001; Gross et al. 2006; Janssen, Anderies, a nd Walker 2004; Walters 1986; Wilson 2006). Challenges to SESs Research Although many researchers w ithin the SESs research program already draw on CAS theory, much of its use has remained c onceptual, and the measurement of change in SESs has remained elusive (Brock and Carp enter 2006; Carpenter et al. 2001; Cumming et al. 2005). This is in pa rt because the processes that affect the dynamics of socialecological systems tend to be hard to descri be, challenging to model, and even more difficult to quantify and measure. With most SESs being so very complex, it has been difficult to identify concrete system component s that can represent sufficient aspects of the SESs in ways that allows the critical, but somewhat abstract, CAS characteristics to be measured. Processes have synergistic effects (O'Neill 1999), or occur at a scale that makes them hard to detect at the scale of observation whether spatial or temporal (Chave and Levin 2002; Gibson, Ostrom, a nd Ahn 2000; Allen and Starr 1982), or the system will have multiple stable states (Ludwig, Walker, and Holling 2002). More particularly, some of the concepts, such as re silience, are hard to translate into tangible indicators, as to some extent this is a relative condition continge nt upon earlier states
28 (Carpenter et al. 2001; Cummi ng et al. 2005). Consequentl y, researchers are struggling to develop methods to quantify the within -site and cross-site heterogeneity and complexity inherent in SESs. The first cha llenge then, is to identify what quantitative data could provide a suitable representation of human-environment interactions in any or all SESs. Furthermore, SESs research tends to have a localized geographical focus. This is a consequence of its origins at the ecosystem le vel. The generalized abstractions (such as sustainability and resilience) used to descri be SESs have been applied to systems of limited extent even where these might be referred to as large-scale (Gunderson and Pritchard Jr 2002). Many of the processe s underlying complex SESs occur at even broader scales, across landscapes comp rising several ecosystems and several communities (Crawford 2005). The implication is that is that some of the higher-order, regional and global impacts of human-environm ent interactions are being missed. This second challenge to SESs research is one th at most clearly highli ghts the opportunity for drawing on existing landscape-level approach es to human-environment interactions: LULCC research. Land-use Land-cover Change LULCC analysis is an emerging science that aims to document and explain the localand regional-level change s to the earths surface, which, through their cumulative effects, also have global-leve l consequences (Lambin et al 1999; Rindfuss et al. 2004; Vitousek et al. 1997; Ojima, Galvin, and Turner 1994; Houghton 1994). The LULCC research program arises directly from th e realization of the potentially dramatic consequences of global-level changes, a nd the need to understand the dynamism and spatial and temporal variability inherent in human-environment systems (Arrow et al.
29 1995; Lambin et al. 1999; Foley et al 2005; Houghton 1994; McMi chael et al. 1999; Ojima, Galvin, and Turner 1994; Turner et al. 1995). LULCC research focuses on the dynamic interactions between human uses of landscapes that is, land use and the Eart hs biophysical conditions, as reflected by land-cover. The stated object ives of LULCC research has been to document and measure change in land cover, and to explain the coupled human-environment system dynamics that generate these changes (R indfuss et al. 2004 p 13978). LULC is far more than just a way of de scribing human activities it represents a dimension of system functioning, and can be seen as a tangible expression of humanenvironment interactions. In a human-domin ated landscape, the type, shape and location of land cover often correlate with how the la nd is or is not being used. LULCC is commonly understood to be the result of a range of biophysical and socio-economic drivers, with proximate drivers having gr eater influence than distant ones (Wood and Porro 2002). In other words, the patterns of LULCC that we see across the landscape are largely the result of the interplay between socio-economic and biophysical processes. Some examples of these interactions include scrub encroachment due to the suppression of fire and increase in grazing (Roques, OConnor, and Watkinson 2001; Wiegand, Saltz, and Ward 2006), changes in hydrology due to montane afforestati on (Farley, Jobbgy, and Jackson 2005; Gallart and Llorens 2004), mi gration due to land degradation (Markos and Gebre-Egziabher 2001; Henry et al. 2004), and tropical deforestation due to market demands for hardwood timber or agricultura l produce (Geist and Lambin 2001; Lambin, Geist, and Lepers 2003).
30 As with SESs research (and befitting such relatively young research approaches), most LULCC analyses have focused on cas e studies first documenting, and then modeling various socio-economic and biophysical interactions. Simulated modeling has moved from describing current interactions to predicting fu ture states as the body of knowledge has grown (Briassoulis 2000; Parker et al. 2003). With sufficient case studies now documented, LULCC researchers are beginn ing to put forward generalizations that can be used in deductive research. This is in contrast to the SESs approach, which started with theoretical generalizations, which were then explored through di fferent case studies. Among LULCC researchers, one of the primar y generalizations that has emerged is that of accessibility, that is, how easily humans can get themselves, and their resources, from place to place. This emphasis on accessibility within the LULCC research community indicates the magnitude of human impact that overshadows all other socialecological interactions. Based originally on Von Thnens 1826 rent-bid model of access to markets, accessibility has been tested as an explanatory variable for deforestation, agricultural intensification and urbanization in a growi ng number of studies; see for example Kim, Mizuno, and Kobayashi (2003); Southworth and Tucker (2001); Verburg et al. (2002); Walker (2004b); Liu (1999); Wa lker and Soleki (2004). Von Thnens work also forms the basis for LULCC studies relating to the economics of land values and distance to market (Liu 1999; Walk er and Soleki 2004). Economic geography, which is founded on location theory, has provi ded the strongest contribution to LULCC theory (Irwin and Geoghegan 2001; Walker 2004a). It tends to be empirically based, and draws on ideas of utility, uneven accumulation, technological evolution and institutional economics (Briassoulis 2000; Martin 1999). Economic geography has also been central
31 in the development of models, perhaps because of its empirical focus, and because economic characteristics and decisions tend to be more readily qua ntifiable and captured as indicators (Parker, Berger, and Manson 2001; Parker et al. 2003). Although economic geography has been the source of much of the theoretical development of the LULCC research progr am, accessibility and economics alone do not explain enough of the observed changes in land use in all cases (Cromley 1982). There has been a growing recognition that individual behavior, po licy and institutions, culture, demography and level of technol ogical development al so affect the type and intensity of LULCC (Briassoulis 2000; Ge ist and Lambin 2001). A dditional theories based on human-environment interactions, as observed by different disciplines such as economics, ecology, sociology, have been called into use at the case study level. These theories focus on population growth (e.g. Malthus), rural development (e.g. Boserup), and political ecology, inter alia; see for example (Briassoulis 2000; Entwistle et al. 1998; Kummer and Turner 1994; Ostrom et al. 1999; Pan et al. 2004). The empirical basis of its analyses has been one of the underlying strengths of LULCC research. Not only have most case st udies consisted of quantitative assessments of change, but most also tend to be spatially explicit. This allows for the location of the system in space and time to be taken into c onsideration and for the patterns and processes of that system to be linked to conditions re lated to a given geogra phic position. It also means that change over time can be linked to spatial variability in large, slow-moving regional processes (Lausch and Herzog 2002; Turner, Gardner, and ONeill 2001). In addition, this spatial approach means that LULCC research has been able to map, measure and model environmental change us ing geographic information systems (GIS)
32 and remote sensing, whose birds-eye view le nds itself to the detection of landscape patterning (Jensen 2000). Challenges to LULCC Theory Whereas SESs research developed under th e umbrella of a conceptual framework, LULCC studies have first documented the prob lem, and then searched for theoretical models to explain the relati onships they have found. This has hampered the post-hoc development of a universal deductive rese arch approach (Brgi, Hersperger, and Schneeberger 2004). Researchers have found that the variabil ity and complex nature of the social and ecological in teractions underlying LULCC ha ve provided a challenge to their ability to develop generalizations, and to predict accurately the extent, nature and even direction, of change (Briassoulis 2000; Ri ndfuss et al. 2004). As such, even within the same LULC system, theories remain context-specific (Perz 2007), or one-sided, reflecting the discip linary origin of the research er (Irwin and Geoghegan 2001). Consequently, LULCC studies still lack a cohesive, multi-disciplinary framework on which to base their questions (Perz 2007; Rindfuss et al. 2004). With a more solid, generalized theoretical foundati on, LULCC analyses would be able to provide planners and managers with more reliable predictions. For LULCC researchers then, the next step is to move away from viewing the complexity of LULCC as an obstacle, and to embrace it as SESs researchers have done. In order to generalize acro ss regions, LULCC researchers need to move beyond the specifics of a given system, such as how ha rdwood demand in the North drives tropical deforestation. Instead they should focus on describing system-neutral characteristics of systems such as the number of processes, ra ther than the individua l types of processes, or the degree of fragmentation, rather than what LULC types are being fragmented.
33 These underlying, fundamental, system-neutr al characteristics are the source of generalizations. That is, LULCC researchers al ready have a clear idea of what drives or limits change (roads, access to market, eleva tion, etc.) and these factors are important in most systems. However, an assessment of the response to these drivers is currently limited to being based on the sp ecific LULC type in an area and this might vary from place to place. If LULCC res earchers continue to work with the same drivers, but focus on a more generalized response characteristic, they can then compare the effects of these drivers even where the LULC types are highly dissimilar. System characteristics might vary in quantity (e.g. tropical forests sequest er more carbon annually than deserts do), but not in the nature, or quality, of the variable (car bon inputs and outputs) It is through these generalizations that an explicitly syst ems-based approach has the potential to give LULCC researchers a unified framework to ove rcome the current theoretical challenges this new field is facing. Integrating SES and LULCC Approaches Even though LULCC is concerned with re gional and global level change, it has, like SESs research, focused on lo cal case studies. This is largely through an appreciation of the level of decision-making th at drives change often bei ng at the level of individual households. To some extent, the levels of an alysis of the two research programs are quite well matched and both are able to explore and describe variability at a range of scales. Both have to deal with c onceptually separated component s, that is social / landuse vs. ecological / land-cover that pose pr oblems precisely because they do not always represent true separations. One example is the issue of scale mismatch, where management is often hampered by trying to apply a blanket, nati onal-level policy to landscapes that in fact comprise many lo cal-level ecosystems (Cumming, Cumming, and
34 Redman 2006; Walker 2004b). As Redman et al (2004) point out, the separation into social and ecological systems is conceptual only, resulting from disc iplinary divisions, more than from actual divisions in the real world. Another example is how disciplinary separation has resulted in different frami ng and foci of research such as how preferences, production costs, utility and valu e are dealt with differently by ecologists and economists (Daily et al. 2000). In a way, SESs and LULCC can be seen as two sides of the same humanenvironment coin (Figure 2-1). One descri bes the processes, th e other captures the patterns. We can therefore view LULCC as a physical manifestation (the patterns) of the Figure 2-1. Schematic diagram showi ng how the divergent SESs and LULCC approaches to the study of human-envir onment interactions can be linked. The quantitative analyses used in LULCC research call for measurable drivers, or mechanisms, while those used to describe change in SESs draw explanations from generalized abstrac tions. LULC diversity is the proposed link, since it combines a generalized CAS concept with a quantifiable landscape characteristic.
35 human-environment interactions (the processe s) that are encompassed by the SES. That is, the landscape patterns of LULCC are a tangible expression of, and therefore an opportunity to measure, SESs. Integrating th e two approaches seems logical. However, for this integration to work, we need to test whether the theory used to describe SESs can be applied with equal validity to LULCC. Certainly, we can apply the metaphors of resilience and sustainability to LULCC: We can evaluate a given landscape to see whether it has retained the same structur al (types, number of types and spatial arrangement of types of landcovers) and functional (the goods and services provided by the different land-uses) identity, despite shocks and perturbations such as market crashes, or prolonged drought. But can the CAS characteris tics that confer re silience be expressed through LULCC, so that we select appropriate indicators? Can LU LCC be explained in terms of a CAS? Complex Adaptive Systems Well-articulated theories of complexity ha ve emerged in a range of disciplines over the past century, including in the social, biological and physical sciences (Arthur 1999; Levin 1999; Limburg et al 2002; Manson 2001; Simon 1962 ; Werner 1999). At the same time, studies from across these disciplin es suggest that a diverse range of systems exhibit similar complex properties (Gunde rson and Pritchard Jr 2002; Levin 2003). Importantly, the need to understand and inco rporate change and va riability has given researchers the impetus to develop the idea of the CAS. A complex system is one that is made up of a large number of parts that interact in a non-simple way (Simon 1962 p. 468). The whol e that emerges from the interactions of these parts is more than the sum of its parts (Kauffman 1995; Koestler 1968). The definition of a complex adaptive system em phasizes two additional features: diversity,
36 and of course, adaptation. A CAS, accordi ng to Levin, is composed of a heterogeneous assemblage of types, in which structur e and functioning emerge from the balance between the constant production of diversity, due to various forces and the winnowing of that diversity through a selection process me diated by local interactions (Levin 1999 p. 231). The idea that systems evolve and take on di fferent states is centr al to the idea of a CAS (Hartvigsen, Kinzig, and Peterson 1998; Holland 1995; Holling, Gunderson, and Peterson 2002; Levin 2003). The ability of a system to adapt is explained by the interactions, or connectivity, between its component s. It is also defi ned by its resilience: its ability to absorb disturbance, self-organ ize, and persist in the same state over time (Folke et al. 2004; Holling 1973). However, in order to understand and manage for a systems resilience, one has to understand a nd be able to describe and measure the characteristics that confer resilience. CASs theory suggests that diversity/ heterogeneity, evolu tion, aggregation, hierarchy, emergence, localized interact ions, connectivity, path-dependency and nonlinearity, are all factors that define a sy stems structure and functioning, (Arthur 1999; Holland 1995; Kauffman 1995; Lansing 2003; Levin 1998) and hence resilience (Holling 1973). By understanding what these characteri stics represent, and the degree to which they are present, we can determine the ability of the system to pe rsist in the face of perturbations. Diversity or heterogeneity is the most f undamental component of a CAS, because it provides the range of components needed fo r adaptation (Levin 1999). Diversity means variety, at all levels of or ganization. Diversity is ne ither accidental nor random (Holland 1995 p. 27). It is dependent on lo cal conditions and interactions such as
37 mutation, recombination, and selection (Levin 1999). These factors, taken in their broader meanings, are the mechanisms that in troduce slight varia tions, and then allow those variations that best suit local conditions to persist. Diversity creates and preserves opportunities for adaptation by providing a ra nge of responses that cover most fluctuations in a systems conditions. The ra nge of responses contri butes to the systems resilience (Folke et al. 2004; Ho lling and Gunderson 2002; Levin 1998). Certain conditions will favor groups of indi vidual components adapted to them. If conditions change, other groups th at did not thrive before no w become those best suited to the prevailing environment. The pers istence of a given co mponent or group of components is dependent on the context provid ed by other components within the system (Holland 1995). While some individual componen ts may not persist, the CAS itself does. However, in any CAS some individual co mponents have stronger influences on the systems functioning and identity. Bala ncing not only the number of types of components, but also the density of each of the components is critic al to maintaining the resilience of a CAS. Evolution is the specific way in which ch ange occurs. Based on the principle of natural selection, evolution draws on the pe rsistence and variabil ity of a systems components. Variability pr ovides the opportunity for new components to develop, while persistence provides the system with continu ity (Moran 1979). Change in a CAS is not arbitrary. It occurs as conditions favor the persistence or emergen ce of certain groups of components over that of others (Levin 1998) These components are then able to replicate and expand, shifting the diversity within the system. Selection for persistence is tightly linked to the size of the pool of compone nt types available, and to the nature or
38 quality of the change (Levin 2003). Evolu tion is the consequence of adaptation, and provides the mechanism whereby complexity and hence resilience are conferred on a system (Holland 1995). Systems are hierarchical, and can be seen as comprising different levels where components form groups. Aggregation relates to the emergence of system-level patterns resulting from lower level in teractions (Holland 1995; Kauf fman 1995). Aggregation is defined as the combination of groups of co mponents, that together function as a functional unit at a higher organizational leve l (Levin 1998). It is through aggregation that hierarchies are created (Ahl and Allen 1996). For exam ple, Koestlers holons are aggregations, where humans are viewed as asse mblages of various cell-aggregated organs and society as an assemblage of organ-aggr egated humans (Koestler 1968). It is these different levels of subassemblies that a dd structural complexity to a system (Simon 1962). In the face of perturbation, subassemblies allow parts of the system to persist, and the system can therefore return more quickly to its earlier stat e or identity. Without these subassemblies, the system would break down into a collection of unrelated components. Reorganization would take much longer, and mi ght lead to the emergence of a different state or identity altogether. When the Khme r Rouge in Cambodia tried to do away with the familiar aggregations of family, nei ghborhood and village, social order rapidly descended into chaos. Although the notions of aggr egation and hierarchy s uggest discrete levels, hierarchical levels are in f act convenient heuristic devices that allow us to represent progression along a continuum (Allen and Starr 1982). It is convenient to examine three points along the continuum that comprises ou r system of interest. At the top, higher
39 level, units or aggregations are broad and slow moving. They determine the context in which the middle level units operate. The middl e level is in turn the context of finer scale, fast moving units. The behavior of fine scale and fa st moving units contribute to the identity of the middle leve l (Allen and Starr 1982). Wher eas climate, as a slow broad process determines the average annual precip itation a given location might receive, the actual amount of rain in any given year is determined by daily and monthly variation in seasonal atmospheric circulation patterns. By examining the system at the level above and the level below the level of interest, we ha ve a more complete picture of its structure and functioning (Ahl and Allen 1996). Localized interactions are the flows that transfer energy and information between the components. These are what make th e system dynamic. Interactions between components at the same level are much st ronger than those between components at different levels (Simon 1962). The number and type of interactions that flow from one component determine how connected the CA S is. Interactions and flows are the processes contributing to the systems connectivity. Proce sses that seem to be random at the level of the individual become predictabl e at the system level (Lansing 2003). Thus, the ways in which the components connect at th e local level will determ ine the identity of the system. The extent to which the syst em is connected determines the possible alternate states of the system Connectivity affects the re silience of a CA S by mediating disturbance due to external variability (H olling and Gunderson 2002). Up to a certain point, the greater the degree of connectivity, the more the sy stem is able to withstand external influences. However, systems that become over-connected lose variability and hence adaptability (Low et al. 2003; McClanahan, Polunin, and Done 2002).
40 If linearity is an outcome equal to the sum of the cont ributing components, then it is easy to understand no n-linearity as an outcome that does not reflect a simple sum of the parts (Holland 1995). Non-linear ity in CASs arises because of the way the components of the system interact with each other, creating multiplicative products and not simple summations. As components interact in comp lex ways, they also adapt and react to the patterns they create (Arthur 1999) They become locked onto a given trajectory or path by feedbacks the way their behavior influenc es the system of which they are a part (Figure 2-2). The path is unpredictable b ecause at each point along the trajectory the non-linear interactions have themselves influen ced the system. They have created a new, different direction for the path (Arthur 1999; Kauffman 1995; Lansing 2003). The Figure 2-2. Schematic diagram of Holling et al.s adaptive cycle (2002) showing a) that as the cycle moves from exploitation to conservation phase, it becomes locked onto a single trajectory, as a result of b) negative feedback removing certain paths and positive interactions reinforcing others. The width of cycle ribbon indicates the number of potenti al paths a system can follow.
41 particular path, with all its variations, defines the identity of the CAS itself. When a system becomes locked onto a given path, it is resistant to m odification, increasing system resilience until it reaches a point of criticality (Holland 1995; Levin 1998). Expressing LULCC in Terms of CAS Concepts CAS characteristics are already proving us eful in case studies evaluating humanenvironment interactions.1 For example, they are frequently used to explore SESs qualitatively (Abel and Stepp 2003; Berkes, Colding, and Folke 2003; Berkes and Folke 1998; Davidson-Hunt and Berkes 2003; Jansse n 1998; Lansing 2003). Such system-level analyses provide an important compass for regional science because they provide a metaphor for directing ideas. However, as noted above, ways of quantifying socialecological processes have remained elusive (Carpenter et al. 2 001; Chave and Levin 2002; Gibson, Ostrom, and Ahn 2000; Ludwig, Walker, and Holling 2002; Turner et al. 1995). To some extent this issue is wors e for large-scale, landscape-level systems because these tend to be even more highly complex and difficult to capture, let alone predict future states for (Cumming and Collier 2005; Gunderson and Pritchard Jr 2002). This has been a contributing reason for why CA S characteristics are not widely used by the LULCC research community, in spite of the early recognition of the complex system nature of LULCC (Turner et al. 1995). Nevertheless, the CAS characteristics discus sed above are readily translatable into concepts already used in LULCC researc h. Diversity is reflected in landscape heterogeneity. We can assess the landscape in terms of not only the absolute number of types of LULC, but also the pr evalence of those LULC types, particularly those that are 1 A March 2007 search on Academic Search Premier/ EBSCO retrieved more than 300 articles with key words complexity and human -environment interactions.
42 most critical to sustained func tioning of the SES. Spatial me trics of diversity derived by landscape ecologists are already becoming a common feature of LULCC analysis (Lausch and Herzog 2002). The distribution a nd spatial arrangement of the different LULC types are part of the landscapes dive rsity. As CAS theory would suggest, the number and relative distribution of LULC type s is not arbitrary. We know that spatial contiguity is the ordering principle for landscapes (Alle n and Hoekstra 1992 p. 47) and that there is spatial autocorrelation in the distribution of components, such as LULC types, across the landscape (Legendre 1993). LULC types, their locat ion and their extent are the result of discerni ble socio-economic and biophysic al conditions and drivers (Wood and Porro 2002). The view of different socio-economic a nd bio-physical driver s as selectors of LULC is akin to the notion of evolution. Looking at how LULC changes over time is essentially a description of how a landscape ev olves. The evolution of LULC arises out of the persistence and variab ility of the LULC types w ithin a given landscape. Variability provides the opportuni ty for new LULC types, wh ile persistence provides the LULC system with continuity. The disappear ance of LULC types suggests that they are no longer suited to current conditions within a given SES. The examination of change in LULC over time is an assessmen t of evolution within the SES. By determining the successional stages of change in LULC divers ity, we can describe the nature of that change in terms of emergence, persistence or disappearance of di fferent cover types. Aggregation, too, is a property of LULCC. We know that the actual interactions are often taking place at the household and land parc el level. It is he re that the decision to clear forest patches or plan t a particular crop is made. Ho wever, at this scale the range
43 of LULC types and their distri bution are not easily visible. For LULCC researchers, in their role as external observe rs, the individual cha nges occurring at th e level of household resource use or small-holder farm activities a ppear as noise. Patterns emerge at a higher level of aggregation, our level of interest (Ahl and Allen 1996) It is here that we can group the gradations of LULC into classes. These LULC classes reveal the patterns arising from the local-level interactions. They show where, for example, deforestation is concentrated, or how the spatial arrangement of forest loss is rela ted to other landscape features. In turn, our LULC classes are delimited and identified by the broad scale categories to which they contri bute. In terms of LULC patte rns, these categories appear to be homogeneous and undiffere ntiated (although still subject to change). For example, much LULCC research examines trends such as deforestation or de sertification that relate to a single, consolidated LULC type. At this level, where researchers aim to capture a specific process, studies are inte rested only in general categories of, for example, forest and non-forest (Caldas et al. 2007; Ngigi and Tateishi 2004; Stibig, Beuchle, and Achard 2003; Tansey et al. 2004). The importance of connectivity in LULCC is shown in the well-documented role of roads as both links and divisions. Roads connect the resources of previously remote areas to markets, and are a key determinan t in changing accessibility (Kim, Mizuno, and Kobayashi 2003; Laurance et al. 2002; Nagendra, Southworth, and Tucker 2003; Overmars and Verburg 2005; Southworth a nd Tucker 2001; Verbur g et al. 2002; Xu 2004). However, roads also fragment lands capes and reduce inte ractions and flows between areas of the same LULC type. Because fragmentation reduces landscape connectivity, it has also received much attention in LULCC research (Hong 1999;
44 Munroe, Croissant, and York 2005; Nage ndra, Munroe, and Southworth 2004). Landscape connectivity controls variability in land-use systems (Bogaert, Farina, and Ceulemans 2005; Burel and Baudry 2005), and ther efore influences the resilience of the LULC system. Over-connected LULC systems su ch as crop monocultures lead to critical dependencies on a narrow range of environmen tal conditions (Low et al. 2003). Some loosely connected LULC components need to be maintained in order to accommodate change (Pan and Bilsborrow 2005). Other exampl es of flows are agricu ltural inputs, such as labor or fertilizer, and financial remittances (Liu 1999; Walker and Soleki 2004). These provide less tangible, but nonetheless real, connections. Path-dependency in LULCC is created when a certain type of LULC is determined by the previous LULC at that specific loca tion. The non-reversibil ity of certain LULCC sequences is an obvious example. Forest ca n be converted directly to grassland, but grassland can only revert to fo rest if it passes thro ugh intermediate stages such as scrub (see for example the decision rules in Verburg et al. 2002). At each successional stage, a new set of LULCC options presents itself. This set of options depends entirely on the circumstances leading to the current LU LC. Non-linearities in LULC manifest themselves as the synergistic effects of inte racting drivers, which when combined, push a system onto an entirely different trajectory th an each of the indivi dual drivers would. For example, the combination of fire suppressi on and increased grazing in savanna can cause a permanent shift from grassland to woodland (Folke et al. 2004), or the presence of invasive species can alter wet forest recove ry after hurricanes (Lugo et al. 2002). This means that the future state of an area of LU LC cannot easily be predicted. In a forest, for example, future states may be entirely diffe rent, depending on whether it will be steadily
45 thinned, leading first to sparse forest, and then to scrub, or wh ether it will be clear-cut for grazing or crops. Importantly, CAS properties are independent of the specific biophysical and socioeconomic conditions of different study areas As such, they give us points of commonality, or generalizations, for cross-site comparison in the study of LULCC. They provide a way to compartmentalize our appro ach to the complex whole, without favoring one part of the system over another. They al so allow us to view LULC abstractly, as a system. As such, because they contribute to structure and functioning, these characteristics can be seen as determini ng a given LULC systems resilience. By evaluating them, we should be able to assess the ability of that LULC system to persist in the face of perturbations, and whether or not it will continue in a recognizable form for the foreseeable future. Exactly as for any SE S, the desirability of a given LULC system persisting will depend on how that system is valued by inhabitants, planners, managers and policy-makers. Strategies that ta rget the mechanisms underlying the CAS characteristics will allow the decision-makers to have greater influence over the ways in which humans influence environmental change. Clearly, CASs can provide the metaphorical link between LULCC to SESs. While some SESs do not lend themselves completely to spatial, land-based analyses (e.g. fisheries) for the most part the CAS charac teristics highlight points of commonality in both research programs. What remains is to detail the way forward from metaphor to measurement (Carpenter et al 2001). The CASs approach it self is complex, with many identifiers that could be used as a starting point for selecting indi cators. However, one
46 characteristic appears to contribute more to the definition of a CAS than any other: diversity Focus on Diversity Diversity is a good characteristic for m easuring LULCC as a CAS, because it is considered the most fundamental CAS character istic. Diversity of traits provides the range of components and responses for adaptation (Levin 1999). Because adaptability and adaptation are dependent on variability, dive rsity determines the potential for change in any system (Levin 1999; Moran 1979). Th is is true for diversity of languages, organisms, skin color, beak-siz e, leaf area, and in this ca se, LULC. For SESs, as with any other system, diversity serves as a buffe r to local variation (Low et al. 2003). To give a LULC example, studies show that eco logical heterogeneity is closely linked to long-term land-use practices th at lead to landscape comp lexity (Naveh 1994). As discussed earlier, diversity is not ra ndom (Holland 199). It is very much context-dependent. This is important becau se if we understand the factors affecting diversity, we can predict and plan for desired levels and patterns of that diversity. This serves as a reminder that diversity is i nherently neither good nor bad. That kind of subjective evaluation depends on whether or not the observer has an interest in seeing the system maintain its current st ructure and functioning. In orde r to fully capture diversitys contribution to a CAS, it is important to think not only as the number of types of components, but also the relative abundance of these types (Magu rran 1988). This is because it is the distribution of the total popu lation across the range of types that gives the system its structure (MacArthur 1960). Diversity is readily quantifiable, and exampl es from a range of disciplines show not only that measurement is possible, but also that diversity is an impor tant reflection of the
47 state of the system in question. Biodiversi ty, including species and ecosystem diversity, is perhaps the most well-known (Dauber et al. 2003; Peet 1974), but research in sociology, economics and anthropology shows that livelihood diversity (Kruseman, Ruben, and Tesfay 2006; Perz 2005) and cultural diversity (Maffi 2005; Stepp, Castaneda, and Cervone 2005) are considered suc cessful indicators of global conditions. To be an effective measurement, LULC mu st: exhibit itself in terms of diversity, have some order to its arrangement in space and time, and of course, be expressible in quantifiable units. If the LULC of a given ar ea exhibits the property of diversity, we can expect to observe three things. Firstly, and most obviously, that there will be diversity of LULC types, because heterogeneity is the most fundamental property of a CAS (Levin 1999). Secondly, that there will be some orde r, or non-random pattern to the distribution of LULC types across the landscape, because local interactions between the underlying biophysical and socio-economic conditions are sp atially explicit, and will influence what LULC types are able to emerge where (Leg endre 1993). Thirdly, we should be able to detect heterogeneity at all spatial scales (Levin 1999) This assumes that the classification system used to determine LULC diversity is an accurate representation of the full range of LULC types in a study area, and an understanding of how as with other types of diversity (e.g. ecologi cal, linguistic) the level to which the classification is taken will affect the level of diversity detectable in the landscape, in much the same analysis of linguistic diversity depends on whet her one is studying this at the family level (Indo-european) or at the dialect level (Catalan Spanish). Testing LULC Diversity There are three main steps to take in or der to test whether LULC diversity is a useful way to express change in land-based SESs. To start, we must describe the
48 distribution of LULC diversity across the landscape with conc epts of diversity already in use in other disciplines. For example, doe s LULC richness (the total number of LULC types) vary in a non-random way? Is ther e variation in the re lative abundance of the different LULC types, with some types domin ating, and other types be ing relatively rare? How do the patterns of LULC richness and re lative abundance change over space? How do these patterns change at different scales of analysis, from local through to broad scale? Next, we must interpret LULC diversity in terms of its response to different processes. We should be able to study how areas of differi ng LULC diversity are related to biophysical and socio-economic factors know n to influence LULCC. There are two reasons for doing this: firstly if we unders tand diversity to be non-random, we must be able to show some causation, and secondly, to support the argument that diversity will allow researchers to compare LULCC in different systems, we must show that the drivers they use to evaluate change affect LULC dive rsity in similar ways. Finally, we must be able to evaluate change in LULC diversity in response to those different processes. Only by measuring how diversity changes over ti me, and by showing that change as a response to different drivers, can we evaluate the ex tent to which diversity is a measure of the systems adaptive capacity. Working with drivers that are shown in th e literature to be si gnificant factors in LULCC strengthens the chance for results to reflect the validity of diversity as the response variable, and puts the focus on diversity as a response variable, instead of on the drivers as explanatories. LULCC studies ha ve repeatedly identified the following factors as drivers of LULCC: soils, slope, roads, and markets, inter alia (Caldas et al. 2007; Arima et al. 2005; Etter et al. 2006; Lambi n, Geist, and Lepers 2003; Manson 2005; Mas
49Table 2-1. Selected examples of case studies sugges ting changes in LULC divers ity and underlying factors Paper Location Observed Changes Related to LULC Diversity Selected Factors Alados et al. 2004 Spain Temporal and spatial variation in LULC types Elevation;, slope; distance from town; population density Bassett and Zueli 2000 Cte dIvoireTemporal and spat ial variation in LULC types Fire; grazing; rainfall Brgi, Hersperger, and Schneeberger 2004 Switzerland Landscape change and persistence Topography; distance from centre; accessibility Caldas et al. 2007 Brazil Conversion of one LULC type to several others Soil quality; distance to highway; demographic characteristics; wealth Chomitz and Gray 1996 Belize Conversion of one LULC type to several others Road density; distance from market, soil quality, rainfall Crews-Meyer 2004 Thailand Landscape change and persistence, LULC heterogeneity Elevation; distance from rivers, roads, villages Cropper, Puri, and Griffiths 2001 Thailand Conversion of one LULC type to several others Slope; elevation; protected areas; population density Douglas 2006 S-E Asia Land degradation as consequence of new and varied LULC types Roads; market access; off-farm labor; conflict Erenstein, Oswald, and Mahaman 2006 West Africa An agro-ecological gradient of land use, diversity of land use highest at intermediate point on gradient Distance from urban market Etter et al. 2006 Colombia Conversion of one LULC type to several others Soil; cost-weight distan ces to roads, rivers, towns; neighboring LULC type Geoghegan, Wainger, and Bockstael 1997 USA Assesses land use diversity in a study of spatial metrics diversity as a factor contributing toward land values Diversity interacting with distance from capital city Laurance et al. 2002 Brazil A single LULC type forest giving way to several LULC types along gradients of roads and settlements Population density; road density Munroe, Southworth, and Tucker 2002 Honduras Conversion of one LULC type to several others Distance to roads, towns; slope; elevation Overmars and Verburg 2005 Philippines A gradient, with few LULC types at each extreme, and a higher diversity of types in the middle of the gradient Slope; elevation; distance to: road, market, village, river; population density, ethnicity Pontius, Shusas, and McEachern 2004 USA Explores persistence vs. temporal variation in LULC types n/a Stefanov and Netzband 2005 USA Scale dependence, aggregation of LULC types n/a Van Gils and Loza Armand Ugon 2006 Bolivia Conversion of one LULC type to several others Land tenure; distance from roads, settlements
50 et al. 2004; Messina et al. 2006; Overmars and Verburg 2005; Soares-Filho et al. 2004; Southworth, Munroe, and Nagendra 2004). In these works, and in others, there is already an indication of how the observed change relates to the number of types, partly because each proximate driver of change referred to repres ents the introduction of at least one new LULC type. For example, most defore station studies essentially addr ess the replacement of a single LULC type (forest) with a patc hwork of several types, containi ng remnants of the old-growth forest, cash crops, subsistence cr ops, pastures, settlements and s econdary growth forests (Geist and Lambin 2001; Wood and Porro 2002; Van Gils and Loza Armand Ugon 2006). A list of recent studies, interpreted in te rms of change in the number or distribution of LULC types, and the cause of that change, are presented in Tabl e 2-1. While some papers qualitatively discuss proximate drivers and broad mechanisms (Gei st and Lambin 2002; Wood and Porro 2002), no cases studies were found that specifically test the scale-dependency of the relationship between LULCC and these factors. These papers prov ide a list of known driv ers influencing LULC diversity, and the nature of the relationships. Methodology for Evaluating LULC Diversity The test of whether LULC di versity provides a conceptual abstraction that allows for cross-site comparison, as well as an acceptabl e measurement of change in complex adaptive SESs has certain key elements. Because indi vidual systems are bounded in space and time, studies must have spatial and temporal extents that are described in terms of the system extents. Studies would need to compare at least two land scapes, to see how LULC diversity is expressed in different SESs. Because systems are multi-sc alar, LULC diversity should be evaluated at different spatial scales, in order to identif y the effects of scale on diversity responses.
51 Spatially and Temporally Explicit Studies Any attempt to quantify change can be done only if appropriate units of analysis are identified. Spatial studies typical ly use two units, where the unit of interest is the type (patch, class or category), and the unit of measurement is a single cell or pixel with defined areal extent in a row-column grid. Because geographic st udies are often spatially explicit, the effect of location can be taken into consideration. One can treat either patch or pixel as a case, or sample, in a dataset, to which a range of vari ables can be attribute d. Landscape ecology has developed indices of landscape pa tterns in general and environmenta l heterogeneity in particular (Gustafson 1998). Indeed, the idea of using LULC types is really no different from the patch concept used by landscape ecologists it simply adds the larger contribution of the socioeconomic influence in the definition of landscape patches. Traditionally, landscape patches have referred specifically to ecologi cal systems, but increasingly LU LCC researchers have adopted the idea and applied it to units on the landscap e that are categorized on the basis of human influence through land use (see for ex ample Southworth et al. 2004). It is important to define the spatial extent or boundary of ones system, so that the distinction between internal variation and external perturbation can be made clear. Similarly, the temporal grain and extent of the system should be defined. The extent includes both the expected durat ion of the system, and the to tal period of the study. The relationship between lifespan and study length dictates what one can infer from the changes observed over the study period. The interval for analysis of the dyna mics of change must also be defined, and whether the time-step being used cap tures change represents an actual trend, or simply fluctuations below the level of interest.
52 Cross-border Studies Cross-border studies provide an opportunity to explore the differe ntial impacts of the processes behind LULCC (Brgi, Hersperger, and Schneeberger 2004) By selecting cases with similar biophysical characteristics, they provide research conditions where a range of factors is held constant. This allows other particularly socio-economic aspects to stand out. If SESs have experienced different histor ies, they will have highly dissi milar landscapes, even if their initial biophysical conditions were comparable. Such contrasts provide informative comparisons to the application of the CASs framework. An ideal situation would be similar-sized, and similarly-formed, islands, or adjacent countries that are separated by a watershed or large river, that have similar biophysical characteristics at least initially, but which are socio-economically discrete. In such a case, one would seek to emphasize and explore the differe nt social, cultural, political and economic influences on LULC. Alternatively, one c ould find places that fall under the same socioeconomic system, but which cross some kind of bio-physical boundary, such as a mountain range that creates a rain-shadow, or islands belonging to the same nation but which were formed by different processes (e.g. volcan ic vs. carbonate platform). Appropriate Levels and Scales of Analysis Any case study to test LULCC as a CAS shoul d describe and analyze three levels the level of interest, and the cons training and contributing leve ls (Ahl and Allen 1996). The definition of these levels should not be arbitrary, but based on an informed assessment of the spatial and temporal extent of the processes in the system (Figure 2-3). LULCC research generally looks at patterns across large landscapes. Likewise, where it is necessary to represent change across an entire SES, br oad or regional-scale studies are the appropriate level to use in capturing socialecological heterogeneity and dynamics (Turner, Gardner, and ONeill 2001).
53 However, the processes underlying landscape pa tterns range from loca lized, household-level decisions, to national policy implementation, and fr om issues of farm-level soil fertility to geologic or climatol ogic gradients. Figure 2-3. Representational diagra m showing different temporal a nd spatial extents of selected SES processes (based on ideas in Holli ng, Gunderson, and Peterson 2002). Different system types are color-coded as follows: green ecological, yellow social, pink cultural, blue climatic. In rural tropical developing countries, the main level of interest w ould be a meso-level corresponding to the most tangib le level of human influence the community. Contributing to the identity of the community are its constituent members householders, whose individual choices and decisions combine (at the contributing level) to create the emergent identify of the community. National, provincial and county policies are put into effect at the macro-level and provide the constraining level, or context.
54 Exploring Patterns of LULC Diversity and it s Response to Factors Influencing Change LULC diversity offers not only a point of comp arison between different landscapes. It also can be used to compare the same landscape at differe nt moments in time, or at different scales of analysis. For example, the frequency distributi on of the variety and re lative abundance of LULC types at 3 scales, will differ significantly from each other (item i in Tabl e 2-2, Figure 2-4-a). At all three scales, the frequency distribution of LULC diversity (both variety and relative abundance) differ from a homogeneous distributi on of a single LULC type, and a completely random distribution of the maximum number of types (ii in Table 2-2, Figure 2-4Figure-b, -c, and -d). At the micro-scale, where mechanistic intera ctions are taking place, at any one location individual agents will be able to select for only a few types of LULC from the large range of LULC types. This means that though the individual types might vary between locations, at the micro-scale low magnitudes of LULC diversity w ill dominate (iii in Table 2-2, Figure2-4-b). At the meso-scale, which reflects the level of interest intermediate levels of diversity will dominate (iii in Table 2-2, Figure 2-4-c) The collective selections re sulting from the human-environment interactions occurring at the level below will produce a wide range of LULC types across the entire study area. This means that higher magni tudes of diversity will be present in some locations, but not others, as the presence of some types will be constrained in other locations by socio-economic and environmental factors. The br oad macro-scale represents the context, and as such contains the full range of LULC types. At this scale, high magnitude s of LULC variety will dominate, although there will still be spatial variation in LULC relative abundance (iii in Table 2-2, Figure 2-4-d). While the list of factors that could potentia lly influence LULC dive rsity is long, only a few key ones are selected here. As noted above, th ese have been chosen because they are known to
55 affect LULCC (see Table 2-1). This is done to shift the emphasis from what is causing change, to how the system is responding to change. Sinc e the understanding of the e ffect of these factors is known for individual LULC type s, the response of LULC divers ity can be better interpreted. The patterns of LULC diversity relating to key bi ophysical and socio-economic factors, such as slope, soil, distance to roads, and distan ce to market, can then be evaluated. LULC diversity will be relati vely low where slope is low, increasing initially as slope increases, until reaching a point of steepness wh ere LULC diversity will drop to very low magnitudes (iv in Table 2-2, Figur e 2-4-e). High LULC diversity will be associated with soil types with intermediate levels of fertility, a nd with good drainage, as many different kinds of LULC will be possible. Lowe r magnitudes of LULC diversit y will be found on nutrient poor soil types, and on very clayey soils because only a few uses will be possible, as well as on soil types that are extremely fertile because high-value crops might be the most profitable use (v in Table 2-2, Figure 2-4-f). Soil quality will be str ongly associated with slope, and the interaction of these variables will need to be controlled for. High LULC diversity is correlated with an in termediate distance from road (Nagendra, Southworth, and Tucker 2003). Close to roads, a few types will dominate, with diversity increasing sharply within accessible distance from the road, and then decreasing steadily as limited access reduces human influence on the la ndscape (vii in Table 2-2, Figure 2-5-g). A similarly ordered relationship, but with different intensity of relationship, will exist between LULC diversity and distance to market (vii in Ta ble 2-2, Figure 2-5-h). Closer to the market, diversity will be low, as only LULC types reflect ing goods that can be traded on the open market will be present. At some intermediate distance, as returns for effort become marginal, LULC diversity will be highest, as t ypes dominated by humans are inters persed with those with little
56 human influence. Importantly, most access to mark ets is along roads, and some considerations should be given to testing LULC diversity agai nst cost-weighted distan ce to market surfaces. The magnitudes of LULC diversity across a social-ecological landscape are constantly subjected to forces of change as that la ndscape dynamically adap ts to on-going humanenvironment interactions. At the same time, th e magnitude of LULC dive rsity at any one point in time will affect the potential magnitude in the future. By creating a sequential set of successional states, a trajectory of change in LU LC diversity for any location can be evaluated.2 Certain trajectories of LULC diversity magnit udes will dominate in a given landscape. This allows for an assessment of what magnitudes of LULC diversity persist over time, as well as providing information on where, and between whic h time-steps, the greatest change in the magnitude of LULC diversity has taken place. The long term trend will be toward s intermediate magnitudes of LULC diversity at all scales (viii in Table 2-2, Fi gure 2-5-i). At the meso-scale level of interest, fluctuations can be expected in response to variab ility in drivers operating at this scale, such as inter-annual variability in rainfall, or volatile market conditi ons. At the fine-scale, a gradual increase in the magnitude of LULC divers ity will occur, as humans introduce a larger range of LULC types into areas previously beyon d their immediate influence, while at the broadscale, there will be a gradual decline as certa in favored LULC types begin to dominate the landscape. The occurrence of a given magnitude of LULC diversity at any location is not random, but depends on the preceding magnitude at th at location. In addition, certain trajectories of change will dominate because some LULC type s are converted more easily (Verburg et al. 2002). Certain trajectories of change in diversit y magnitude are more likel y to occur than others, 2 In LULCC studies, the term trajectory refers to a cate gory of change that groups toge ther all areas that have the same specific LULC conditions at each of a series of timesteps. In terms of actual types an example would be forest-woodland-pasture-crop, and in terms of number of LULC types, 2-2-3-9, reflecting increasing LULC variety.
57 and only a few of the potential traj ectories will cover extensive areas. There are many potential LULC trajectories that will not ha ppen (e.g. forest to city to fore st to city across four 5-year time-steps), and likewise, there ar e many diversity trajectories that are unlikely (e.g. 10-1-10-1 in a place with a maximum of 10 LULC types). As scale increases, so too does the overall magnitude of diversity, since more types of LULC will be found as the area of analysis increases. This increase in magnitude means that the range of LULC variety levels increases, thereby increasing the number of both poten tial trajectories and those covering extensive areas on the landscape (viii in Table 2-2, Figure 2-5-j). In additi on, the spatial extent of a given trajectory is dependent on the initial exte nt of its initial class. Just as there is continual va riation in magnitude of LULC diversity, some of the factors influencing LULC diversity themse lves are undergoing change at clear ly discernible rates. This means that the relationship between LULC diversity and these factors will change over time, and that this change provides important information on the state of the SES. Other factors, however, remain effectively constant. Over time, the threshold for a decrease in LULC diversity will become associated with steeper slopes, while at the same time LULC diversity in flatter areas will decrease (ix in Table 2-2, Figure 2-5-k). As demand for agricultural land increases, areas that are flatter lose LULC diversity as agricultu re replaces other, less human-dominated, cover types. Initially, there may be some increased hetero geneity, but ultimately all land tends towards conversion to the LULC represented by the dominan t crop (Soini 2005). Sim ilarly, over time the mean magnitude, as well as the variance, of LULC diversity will increase on certain soil types, but not necessarily all (x in Ta ble 2-2, Figure 2-5 -l). Over ti me, LULC diversity will decrease closer to the road, while higher magnitudes of di versity will be found further and further away
58 Table 2-2. Quantifiable aspects of LULC diversity, indicating hypothesized responses Measurable Response Predicted Response Attributes Diagram in 2-4 to 2-6 i Magnitude of LULC diversity Scale dependent a) ii Magnitude of real-world LULC diversity compared to null and random models A null model of one type would have no diversity, whereas a random model would represent all magnitudes of diversity equally. Actual LULC diversity would have measurable mean and variance. b), c), d) iii Magnitude of LULC diversity Distinct distribution patterns, peaking at intermediate magnitudes which vary according to scale b), c), d) iv Interaction of LULC diversity and slope 3rd order polynomial relationship e) v Interaction of LULC diversity and soil type Difference in range and mean magnitude of diversity according to soil type / fertility f) vii Interaction of LULC diversity and distance to roads, distance to market 2nd order polynomial relationship g), h) viii Change over time in magnitude of LULC diversity Scale dependent, but at all scales a tendency towards intermediate magnitudes of diversity persisting i), j) ix Change over time in relationship between LULC diversity and slope A weakening relationship at later points in time, with peak magnitudes of diversity associated with increasingly steep slopes k) x Differential changes over time in magnitudes of LULC diversity on different soil types Initially, increasing differences in mean magnitudes of LULC diversity, until demand for land is high. Then, increasing magnitudes of diversity on poorer soil types. l) xi Change over time in magnitude of LULC diversity with changes in distance to roads A weakening relationship at later points in time, with peak magnitudes of diversity associated with increasingly greater distances to roads m) xii Change over time in magnitude of LULC diversity with changes in distance to market Strong relationship maintained, but peak magnitudes of diversity associated with shorter distances to market n)
59 a) b) c) d) e) f) Figure 2-4. Schematic diagrams representing a) d) the distribution of LULC diversity at different scales, and e) f) the interacti on of LULC diversity with slope and soil
60 g) h) i) j) k) l) Figure 2-5. Schematic diagrams representing g) h) the interaction of LULC diversity with distance from roads and market, i) j) chan ges in LULC diversity over time, and k) l) changes in the relationships betwee n LULC diversity and slope and soil.
61 m) n) Figure 2-6. Schematic diagrams representing m) n) changes in the relationships between LULC diversity and distance from roads and market. (xi in Table 2-2, Figure 2-6-m). With increa sing demand for land, the conversion of LULC to agriculture will occur further and further from the road, pushing areas of high LULC diversity further away. In highly connected SESs with dense road networks, the relationship between LULC diversity and distance from roads becomes w eak. These trends will be different for the changing relationships with distan ce to market. Over the trajec tory period, LULC specialization in response to economic forces will lead to incr easing LULC diversity closer to markets, while lower magnitudes of LULC diversity will still be found further and further away (xii in Table 22, Figure 2-6-n). A Preliminary Assessment of the LULC Diversity Concept The village of Trapeang Prasat lies in the hilly east of Ordar Mean Chey province in northern Cambodia (Figure 2-7). It is a frontier town that has e xperienced rapid change since the end of fighting between the Khmer Rouge guerril las and the Vietnamese-backed government in the early 1990s (Gottesman 2004). In the ensuing years it has grown rapidly, and with this growth, its forested areas have gi ven away to a range of different human-dominated LULC types.
62 According to the hypotheses in Table 2-2, we should be able to see detectable patterns of LULC diversity distributed across the landscape. This distribution would differ significantly from the LULC diversity of a null model of total ra ndomness, which would comprise a landscape containing all the potential ma gnitudes of LULC diversity, di stributed randomly across the landscape (Figure 2-4a). The dist ribution would also differ from LULC diversity calculated on a null model of total homogeneity, which would cons ist of a landscape containing only one LULC type that is, all pixels in the landscape will ca rry the same value. According to the hypotheses, LULC diversity at different scales would reveal different patterns on the landscape. Close to the roads leading to Trapeang Prasat, as well as close to the town itself (where the market is located at the intersection of the two ro ads) diversity should be higher, decreasing further and further Figure 2-7. Location of Trapeang Pras at village in northern Cambodia.
63 away from these features. As one moves south into the elevated area to the south of Trapeang Prasat, LULC diversity should decrease (Table 2-2). A 12-class classification map of the 30 m resolu tion subset of the sate llite image shown in Figure 2-7 was created to assess the distribution of LULC dive rsity in the landscape surrounding Trapeang Prasat (Figure 2-8a). Two differe nt moving window sizes (3x3 pixels and 33x33 pixels) were run across the classified raster to measure LULC diversity (a s a simple measure of variety of LULC types) at a micro-scale of arou nd 1 ha and a meso-scale of about 100 ha (Figure Figure 2-8. A test of LULC dive rsity on a) the 12-class classification of the Trapeang Prasat area, showing b) micro-scale diversity a nd c) meso-scale diversity. Micro-scale diversity of a random lands cape is shown in d) while that of a completely homogeneous landscape is shown in e).
64 2-8b and Figure 2-8c). A 3x3 pixel moving wi ndow was then run across the random landscape and the homogeneous landscape (Figure 2-8d and Figure 2-8e). As Figure 2-8 shows, the actual distribution of LULC diversity differs considerably from the random and homogeneous null models, and shows the key landscape features visible in the classification. At the micro-scale (Figure 2-8b) the patterns closely mi rror those of the actual classification, but it the magnitudes of diversity at this scale are quite fragmented, so that the matrix of crop and woodland visible in a) begins to look almost random. At the meso-scale in c) key processes are emphasized, so that the roads and village node stand out clearly as areas of high diversity. The undisturbed forest on the hills to the south of Trapeang Prasat appears as an area of low LULC diversity, while the matrix of crops and woodland show intermediate magnitudes of diversity. The landscape is now generalized and repr esented as quantitative integer data. The frequency distribution for LULC divers ity shows the expected left-skewed shape (Figure 2-9) and resembles the hypothesized dist ribution curve, and differs from the two null models. Since the data are quant itative, we can calculate means a nd standard deviations for the landscape. The micro-scale has a low mean magnitude of LULC diversity (1.97 types), with a standard deviation of 1.15, while the meso-scale s mean is 5.28 with a standard deviation of 3.11. This difference is that anticipated by th e shift along the x-axis anticipated by the hypothesized distribution in Figure 24c. With these measures alone, it is clear that to contrast with other landscapes even, for example, to the west of the same province in Ordar Mean Chey, there would be distinct comm onalities, but also differences relating to the extent of human influence on the landscape.
65 Micro-scale LULC Diversity (Actual) 0 5 10 15 20 25 30 35 40 45 50 123456789101112 Magnitude of LULC diversity (no. of LULC types)Percentage of landscape a) Micro-scale LULC Diversity (Random) 0 5 10 15 20 25 30 35 40 45 50 123456789101112 Magnitude of LULC diversity (no. of LULC types)Percentage of landscapeb) Micro-scale LULC Diversity (Homogeneous) 0 10 20 30 40 50 60 70 80 90 100 123456789101112 Magnitude of LULC diversity (not of LULC types)Percentage of landscape c) d) Figure 2-9. Comparison of a) act ual LULC diversity to null models of b) random diversity and c) a homogeneous landscape. Conclusions At the same time as researchers have begun to embrace the dynamic and multi-scalar nature of the world (Folke et al. 2002; Holling 2001; Lambin, Geist, and Lepers 2003; Munroe, Southworth, and Tucker 2004; Young et al. 2006), the uncertainties and complexities that abound in human-environment interactions presen t distinct challenges to both land-use land-
66 cover change (LULCC) and social-ecological syst ems (SESs) research. Because of the unique combinations of landscape characteristics, land cha nge scientists have struggled to find sufficient generalizations that allow them to theorize be yond immediate case studies, hindering their ability to extrapolate and predict using a hypothetico -deductive approach (Bri assoulis 2000; Brgi, Hersperger, and Schneeberger 2004; Perz 2007; Rindfuss et al. 2004). For SESs researchers, on the other hand, the difficulty has be en to identify ways to evaluate system change quantitatively (Carpenter et al. 2001; Cumming et al. 2005). The theoretical framework and methodological a pproach outlined here explores both the usefulness, and the feasibility, of explicitly linking LULCC and SESs research. By using the CASs paradigm, we can allow the strengths of each research program to fill in gaps in the other. Firstly, the CASs theoretical framework generall y, and the concept of diversity in particular, provides researchers with measur able points for comparison, extrapolation and prediction. For researchers focusing on LULCC, a CASs approach should lead to more coherent and systematic analyses that are not context-sp ecific or discipline-dependent (Perz 2007). LULC diversity can be tested and evaluated in any landscape, indepe ndent of the prevailing environmental and social characteristics. As such, CASs theory allows LULCC research to move from inductive towards deductive analysis, by providing the theoretical basis from which to posit predictive hypotheses. Trends in magnitudes of diversity can be in ferred, based on the CASs ideas of selection, emergent properties, non-linearit y, path dependency and scale, inter alia (Ahl and Allen 1996; Arthur 1999; Holland 1995; Kauffman 1995; Lansing 2003; Levin 1998). Secondly, by explicitly treating LULCC as th e expression of the va rious components of a SES, CASs theory provides the opportunity to m easure that systems processes. Quantifying a systems resilience has to date largely been an elusive target, with implications for effective
67 management. This framework suggests that in order to assess the lik elihood of a system changing or persisting in a give n form, one needs to work with the properties that confer resilience. Conceptually, eval uating change in LULC diversity does precisely that, since any change in the range of potential responses will aff ect the ability of a system to adapt and persist (Holland 1995; Levin 2003). By using LULC as an indicator, we can now spatially represent change in the condition or st ate of different SESs. The focus on diversity as opposed to specific LULC types provides a generalized unit of measurement that can be tested and compared in disparate SESs, whose diversity is built of different sets of LULC types. Even where sy stems might have differe nt magnitudes of LULC diversity, drivers of change are still likely to cause similar response pa tterns, due to the ongoing selection process and its interplay with size of the pool of components. For example, the magnitude of diversity in a given landscape will have a non-linear respon se to the development of a new road, or the opening of a new market, with magnitudes of diversity increasing up to some intermediate distance, and then decreasing again. The nature of the relationship should hold constant for different systems, even t hough the exact coefficients representing that relationship might vary, allowi ng for extrapolation and predic tion. Yet it is precisely the difference in those coefficients that provides with a meaningful measure of how each system is faring relative to the other. Analysis at multiple scales is important, as LULC diversity patterns will respond to different factors at diffe rent scales. For example, different broad-scale factors such as national policies and climatic conditions provide the cont ext that determines the maximum magnitude of diversity. Likewise, variation in so il type or quality at the local le vel will influence both the type
68 and range of different LULC types, leading to different magnitudes of diversity emerging at the level of interest. Because they are complex and adaptive, SESs are constantly changing. If we conduct analyses across multiple points in time, we can pr edict that changes (or not) in LULC diversity will reveal where the SES identity is changing or persisting in terms of its structure and function. Over time, we can expect to see a trend towards intermediate magnitudes of diversity in response to local human-environment intera ctions (Levin 1999). For example, places with low LULC diversity will experience the emergence of higher magnitudes as people convert land in undeveloped areas, or develop spec ialty activities to meet market niches in over-simplified systems. In areas with very high LULC diversit y, market forces will favor some activities over others, leading to a sele ction process that reduces magnitudes somewhat. The holistic approach provided by CASs analysis can accommodate the range of social and environmental facets of LULCC change, and deal with contrasting cross-site comparisons. As such it should prove useful to researchers tr ying to develop LULCC research as a science (Rindfuss et al. 2004; Walker 2004a). Because LU LC is spatially and temporally explicit, it should be useful to consider it as the physical manifestation of an underlying SES, thus providing an opportunity for SESs research to progress from using CASs mainly as a qualitative evaluation to using it for quantitative assessments. By fr aming human-environment interactions as CASs, we can link the patterns of LU LCC to the processes of SESs and so quantify change and resilience across different landscapes.
69 CHAPTER 3 PATTERNS OF LAND-USE LAND-COVER DIVE RSITY IN SISAKET, THAILAND AND ORDAR MEAN CHEY, CAMBODIA Introduction The full development of a scie nce or discipline re quires both strong theoretical foundations and empirical measurements. Land-use land-co ver change (LULCC) researchers are seeking theoretical generalizations that would increase their ability to compare different regions and contexts (Perz 2007, Rindfuss et al. 2004). The complex nature of the social and ecological interactions underlying LULCC is a challenge to the developm ent of a deductive research approach (Brgi, Hersperger, a nd Schneeberger 2004). The variabil ity and complexity of locallevel interactions have confounded researchers ability to predict accurately the extent, nature and even direction of change beyond the im mediate location being studied. Theoretical frameworks currently used to guide the rela tively young LULCC research program tend to have a one-sided focus that depends on the disciplinar y origin of the researcher (Irwin and Geoghegan 2001; Walker 2004). Consequently researchers have yet to adva nce LULCC studies to a point where existing theories can be tested under a range of conditions. This paper tests the application of the theo ry of complex adaptive systems (CAS s) generally, and the concept of diversity specifically, to the study of land-use and land-cover (L ULC), as a way of generalizing landscapes (so that they are describable withou t reference to specific LULC types) to allow prediction and cross-site comparis on (as detailed in Chapter 2). Increasingly, researchers are aware of the need for an integrated and interdisciplinary approach that incorporates eco logical, social and economic sciences (Holling, Gunderson, and Ludwig 2002). There is also a growing body of research treating the human-environment interactions that underpin LULC C as social-ecological systems (SESs) (Berkes and Folke 1998; Janssen, Anderies, and Ostrom 2007; Redman, Grove, and Kuby 2004) SES research builds on
70 the understanding of human-dominated systems as being complex, diverse, multi-scalar, nonlinear and self-organizing that is, as CASs (Cumming and Collier 2005; Holling and Gunderson 2002; Janssen, Anderies, and Ostrom 2007; Levin 2003; Odum 1975). Since LULCC and SESs are both examinations of human-environm ent interactions at the landscape level, their research programs can be seen as complement ary. LULCC addresses th e spatial pattern, while SESs treat the process, behind th ese interactions. As such, LU LCC is a tangible expression of SESs (as detailed in Chapter 2). Because CASs theory has been used successfully to give qualitative and quantitative evaluations of dispar ate SESs (Berkes, Colding, and Folke 2003), we propose that by framing LULCC as the manifestati on of the activities of a SES, we should be able to evaluate different landscapes in terms of SESs, to see how changes in the structure and functioning of generalized SES ch aracteristics, such as divers ity and emergent properties, respond to the processes already known to a ffect LULCC. A CAS approach should allow researchers to target the very complexity of human-environment interactions that has hampered the ability of researchers of LULCC to move beyond context-spec ificity, while simultaneously providing the means for SESs researchers to quantify processe s underpinning SESs. LULC Diversity Diversity is one of the most fundamental CA S characteristic in any system (Holland 1995; Levin 1999). It provides the range of components which form the basis for the responses that lead to adaptation. The concept of diversit y is familiar, and has been applied widely, from cultural anthropology to livelihood systems, sociology, economi cs and ecology (Dauber et al. 2003; Kruseman, Ruben, and Tesfay 2006; Maffi 2005; Peet 1974; Pe rz 2005; Stepp, Castaneda, and Cervone 2005). Diversity is also measurab le, which means that it can be quantified and evaluated for any dynamic system. It can ther efore provide a point of comparison between systems with dissimilar states of structure and functioning. As such, it is a good characteristic
71 for testing the applicability of CASs theory to LULCC, while also being informative as a standalone concept. The spatial analysis of diversity is not new; landscape ecologists re gularly use diversity metrics to analyze ecological change at broad scales (Lausch and Herz og 2002). It is a small step from the way landscape ecologists divide thei r landscapes into ecosystems, to the categories of LULC where land-change scientists add th e human dimension. For this reason, we move beyond analyses of individual LULC types to lo oking at the whole lands cape and examining all the types together. Landscapes can be evaluate d both in terms of the ab solute number of LULC types, as well as the extent to which LULC t ypes are sustained, especially dominant types or those that underpin the functioning of the SES. A focus on the diversity of LULC types, will allow studies to be independent of the actual nature of the LULC categories themselves, and so allow for extrapolation and prediction (as detailed in Chapter 2). Broad Research Goal This paper is part of a broader program de veloping CASs approaches for linking LULCC analyses to SESs as a way of qua ntifying generalized processes (as detailed in Chapter 2). This entails combining CASs the framework used by th e SESs research program to describe process with the spatially exp licit methods used in LULCC research to measure pattern (Figure 3-1). The assumption that LULC represents a tangible expression of the components of a SES allows the strength of each research program to fill the gap of the other. If we want to use LULC to represent the com ponents of a SES, we need to be able to identify the range of LULC types in a landscape and we would need to be able to detect different LULC patterns at different scales. If we e xpress these patterns th rough a CAS framework, we should be able to use LULC diversity responses to various drivers of change as a way of
72 studying a given social-ecologica l landscape in terms of its connectedness, dynamism, stability and likelihood of persistence. Figure 3-1. Schematic diagram showing how the divergent SESs and LUL CC approaches to the study of human-environment interactions can be linked. The quantitative analyses used in LULCC research call for measurable drivers, or mechanisms, while those used to describe change in SESs draw e xplanations from generalized abstractions. LULC diversity is the proposed link, sin ce it combines a generalized CAS concept with a quantifiable landscape characteristic. Specific Research Objective In this paper, we focus on diversity as a key CAS concept and as the main characteristic that underpins the complexity of human-environmen t interactions central to both the LULCC and SESs research programs. In order to test LU LC diversity as a unifyi ng, non-ecosystem specific characteristic for cross-site comparison, we need to show it as non-rando m, presenting distinct patterns on the landscape at different scales, with different magnitudes of diversity dominating at
73 different locations, and with meaningful vari ation in magnitudes of diversity over time (as detailed in Chapter 2). We use the following questions to guide our research: What is the distribution of LULC divers ity across two landscapes separated by an international boundary and with different sociopolitical histories, but otherwise similar? What magnitudes of LULC diversity are found in each landscape? What happens to the LULC diversity of each area over time? How do the patterns of diversity co mpare between the two landscapes? Can the patterns be interpreted at l east qualitatively in terms of underlying mechanisms? As stated in Chapter 2, we hypothesize th at both landscapes will have left-skewed distributions at the micro-scale, approximately normal distributions at th e meso-scale, and rightskewed distributions at the macro-scale (Figure 3-2). However, we expect that the specific distribution shapes for each landscape will differ in terms of the degree of skewness and kurtosis in the distribution, and that there will be variation in the precise shape of the di stributions over time. At the micro-scale at a ny one location, individual agents w ill be able to select for only a few types of LULC from the larg e range of LULC types. Though individual types might vary Figure 3-2. Schematic diagram representing the distribution of LULC di versity at different scales.
74 between locations, at the micro-scale low magnit udes of LULC diversity will be found. At the meso-scale, the collective selections resul ting from the human-environment interactions occurring at the level below will produce a wide range of LULC types across the entire study area. This means that higher magnitudes of dive rsity will be present in some locations, but not others, as the presence of some types will be constrained in other locations by socio-economic and environmental factors. The broad macro-scale represents the context, and as such contains the full range of LULC types. At this scale, high magnitudes of LULC diversity will dominate. Study Area Large, regional, broad-scale comparative studies are useful for investigating the heterogeneity and dynamics of entire SESs (Tur ner, Gardner, and ONeill 2001). This study examines two SESs as defined by the political boundaries of the crossborder provinces of Sisaket, Thailand and Ordar Mean Chey, Cambod ia (but excluding the south-east corner of Sisaket which falls beyond the footprint of the satellite imagery). The study area was chosen because the two provinces have experienced very different political histories. The provinces currently exhibit highly dissimilar vegetation and socio-economic ch aracteristics (Figure 3-3), in spite of similar underlying climatic biophysical characteristics. This provides an ideal opportunity to test whether LULC diversity as a concept can tran scend context-specificity. The SESs are taken to be manifested by the LULC pa tterns distributed across th e provinces that is, the distribution of LULC types acro ss the landscapes are a result of activities within the SESs (as detailed in Chapter 2). Cross-border studies provide an opportunity to explore differences in the processes behind LULCC change (Brgi, Hersperger, and Schneeb erger 2004). Sisaket and Ordar Mean Chey, being spatially adjacent to each other, have the same edaphic, geomorphological and climatic contexts. Both are in the Asian monsoonal tropi cs, mostly flat, and have a few large rivers
75 crossing them. They have pronounced, fiveto seven-month, rainy seasons. Both were originally covered with semi-deciduous moist trop ical forest. Geographi cally, the two provinces are separated by a low escarpment that drops from Thailand into Cambodia. Uplifting tilts the plateau back into Thailand so that the two pr ovinces are separated hydr ologically. Although the last couple of years have seen so me cross-border trade, the provin ces are for the most part socioeconomically separate. They have been subjec t to different social, cultural, political and economic influences, partly because of the cen turies-long antagonistic relationship between the two nations (Chandler 2000; Wyatt 1984). Figure 3-3. Study area map showing the different spectral characte ristics of Sisaket, Thailand and Ordar Mean Chey, Cambodia
76 Sisaket is a predominantly agrarian provin ce situated in northeast Thailand, at the southern edge of the Khorat Plateau. The north-eas t is drier and poorer than the rest of Thailand. Until recently, it lagged behind in economic and infr astructural development, as well as in agricultural expansion (Parnw ell 1988). The province, measuring approximately 8860 km2, is home to about 1.5 million people, with a population density of approximately 160 people/km2. In spite of the regions relati ve poverty, nearly all rural vill ages are (as of 2005-2006) linked by paved roads and have electricity, clean water supplies, schools and h ealth facilities. For the last several hundred years, the SES of this part of Thailand has centered on ri ce cultivation (Semthiti 1951; Wyatt 1984). Households are for the most part small-scale farmers, producing crops both for their own consumption and for sale. Sisakets lo ss of forest cover mirrors that of the rest of the country (Sponsel 1998). Rice fields now do minate the landscape (Felkner 2000). Large tracts of forested areas are re stricted to the extreme south, on the border with Cambodia, where the steep slopes of the escarpm ent mountains and border security policies and activities limit agricultural expansion. Ordar Mean Chey shares the eastern third of its northern border with the western half of Sisakets southern border (see Figure 3-3) It comprises an area of about 6630 km2, with a population of about 70,000. Its popu lation density, ~ 10 people/km2, is an order of magnitude smaller than Sisakets. It is one of Cambodi as poorest districts. The province was separated from Siem Reap province in th e 1990s as the area it covered wa s still under the power of Pol Pots Khmer Rouge, and as such, beyond the cont rol of the national government. All roads are unpaved and seasonally impassable. Most settlements do not have electricity or formal water supplies. Forests and woodlands still dominate the landscape, alt hough less so to the west of the province. The SES here can be defined as bein g one of subsistence farming and heavily reliant
77 on natural resources extraction. Forest cover is rapidly be ing removed; both for timber harvesting and for slash and burn agriculture (C han, Tep, and Sarthi 2001). Rice production is limited mainly to the western side of the province, closer to the district capital of Samraong. The eastern area is hilly, and less su ited to rice production. Where fi elds are emerging in the east, these tend to be for upland crops. Methods Study Design Both CASs and SESs are understood to be dynamic and changing, and yet functioning within defined spatial and temporal boundaries. The SESs of the prov inces are represented spatially as gridded landscape models. Each gr id cell, or pixel, cont ains a single value and represents a unit of sampling and an alysis. Each pixel represents the LULC type present at that geographic location for a given point in time, and t ogether all pixels provide a snapshot in time of each SESs population of LULC types. Th e term landscape refe rs specifically to the spatial extent of each province and its constitu ent pixels. The study examines how diversity changes over a temporal extent of 16 years, fr om 1989 to 2005, and captured across 4 time-steps. This paper addresses the research questions sepa rately for the two provinces, and then compares the results. In order to study the influen ce of spatial scale on landscape patterning, LULC diversity is measured at three different spatial resolutions (A llen and Starr 1982). In terms of SESs, we note that decision-making is often made at the hous ehold level, contributing to the emergence of pattern at the community or vi llage level. Broader political influences at the national and provincial level provide the contex t and constraints. Observations in the field, average district and commune sizes calculated from GIS data, and other studies suggest that the spatial extent of these different scalar influences would range from around 1 Ha (t he order of magnitude of the
78 average household land-holding), thr ough the level of interest of a typical commune or district, with order of magnitude of 100 Ha, to about 100 km2 a scale that captures the cumulative effects of about 10 communes while still acco mmodating variation with in the province (Chan, Tep, and Sarthi 2001; van Wey 2005; Mekong Ri ver Commission 2003; National Statistics Office 2003). Determination of SES Components The landscapes of the study area were populated with LULC types through the interpretation and classification of the six re flectance bands of Landsat 5 Thematic Mapper and Landsat 7 Enhanced Thematic Mapper + satellite imagery, with 30 m pixel resolution3. The study area falls across two scenes: WRS-2 Path 127 Rows 49-50. Cloud-free, seasonally comparable, paired scenes (sequentially acquired) were selected for four points in time across the study period: 22nd January 1989, 28th January 1994, 25th March 2000 and 27th February 2005 all in the dry season of this north-east monsoon region. LULC t ypes were identified during an initial visit to the study area in 2005, and after a second visit in 2006, a final set of 14 classes was determined to be a representative model of th e two provinces SESs. These classes were based on extensive field visits in 2005 and 2006, comparis ons with related projects, and earlier maps (Blasco, Bellan, and Lacaze 1997; Mekong Rive r Commission 2003; Nang Rong Projects 2004) All images were radiometrically calibrat ed (Green, Schweik, and Hanson 2000; Markham and Barker 1986; Teillet and Fe dosejeus 1995), and georectified using the 2000 image as the base. Each pair of scenes was mosiacked t ogether and then subset to the study area. Classification of the 2005 image was done using a hybrid supervised-unsupervised iterative selforganizing data approach based on 276 randomly located training samples collected in 2006. 3 The thermal band was omitted so that the classifications could be used in a separate study on within-class variation in temperature.
79 The accuracy of this classification was tested using 172 randomly located points collected in 2005, on a grouping of the classes into 7 categorie s, because the 14-class classification was fragmented into many 1-pixel patches. The spectra l signatures generated by the classification of the 2005 image were then used to cla ssify the earlier image subsets. The resultant LULC classifications are consid ered the starting point for this study. The different measures of diversity at the three scales were calculate d directly from these images. The 30 m pixel resolution, or grain, was consider ed appropriate for expl oring patterns of LULC diversity, since the smallest possible window-siz e (3x3 pixels) corresponds roughly to the 1 Ha size of the lowest level of analys is. At this resolution and with this smallest window-size, nine (i.e. two-thirds of the maximum number) of the types can potentially be observed at the contributing, household level of analysis, and at the community level of interest of 100 Ha, enough pixels are present to captu re all 14 possible LULC types a nd to reveal detailed spatial patterns of diversity. Calculation of Diversity for Fixed Points in Time The two measures of diversity assessed in th is paper are based on an analogy with species diversity in ecosystems. This does not mean th at each LULC type is a species in a SES. However, by using this theoretical parallel we can apply the anal yses that have been developed for biological diversity to hypothesis tests for LULC diversity. We therefore assess both variety, a count of the total number of types in a given area, and relative abundance, the frequency of occurrence of each of the types. Since the classi fication process resulted in a total of 14 different LULC types, this value represents the maximum total variety possible in the study area, with 1 being the lowest possible value. In this study the Simpsons i ndex of relative abundance is used, in the following form:
80 S = 1 (p2/p) (1) where: S = index value p = proportion (McGarigal et al. 2002). The Simpsons index runs from 0 to 1. In the form used here, 0 represents complete dominance by one type within th e neighborhood of analysis, and 1 represents equal distribution of all types in the neighborhood. Variety and relative abundance values for each pixel in the landscapes were determined by calculating the number and type of classes in a specified area surr ounding the pixel through a moving window analysis. In order to capture the sp atial extents appropriate to the three different levels of analysis, the size of the moving window was adjusted to include the different neighborhood sizes corresponding to each leve l: 3x3, 33x33 and 303x303 pixels (Table 3-1). For each time-step, three differe nt variety, and three different relative abundance landscape models were generated based on the three different scales. The moving wi ndow analysis yields a series of maps that allows in terpretation of the distribution of variety and relative abundance across the landscapes. The frequenc y distributions of the values across the pixels are further summarized through histograms and summary statis tics, allowing comparisons between the two provinces and across time. Table 3-1. Levels of anal ysis of LULC diversity Micro-scale Contributing level Meso-scale Level of interest Macro-scale Constraining level SES level Household / farm Community District / province Spatial extent ~ 1 Ha ~ 100 Ha ~ 100 Km2 Analysis size 3x3 pixel window 33x33 pixel window 303x303 pixel window
81 Change in Diversity over Time Because variety is an integer measure, trajecto ries of change in this measure of diversity can be calculated.4 This was done for all three scales, by creating spatial out puts that represent the sequence of variety values in each pixel over the 4 time-steps. With 14 magnitudes of variety and four time-steps, there are an unwieldy 38416 potential trajectorie s; however, most trajectories will not be filled or to cover significan t areas. This is because not all trajectories are plausible within the time-frame of the study. For example, it is unlikely for an extensive area of very high diversity to change to very low diversity, and then back to high, and then low, again. Importantly, since the variety values are di screte and not categor ical, for any given trajectory class the mean and variance of va riety over time can be calculated, while still providing information on the direc tionality of change in the de gree of diversity. Simpsons index analyses yields continuous data outputs, removing the option to calculate trajectories. For this reason, for each level of anal ysis, the standard deviation ove r time in Simpsons index was calculated for each pixel. Results and Discussion LULC Classifications Although the classifications are th e starting point of the diversity analysis, the results of the classifications are shown in Figur e 3-4, and discussed here briefl y. Several of the classes fall along a continuum of dense forest less dens e forest dense canopy woodland sparse canopy woodland rice under sparse trees treeless rice areas. Some of th e classification errors reflect the effect of arbitrary cut-off points imposed al ong this range (Table 32). The most difficult class to separate from others was scrub / upland crops. This is a result of the classifications 4 The term trajectory refers to a categ ory of change that groups together all areas that have the same specific LULC conditions at each of a series of time-steps. With continuou s data, the variation is such that the likelihood of several pixels following the exact same sequence is minimal.
82 being based solely on the spectral characteristi cs of the satellite imagery, which in some instances are very similar for different LULC cl asses. The accuracy assessment for the grouped classes resulted in an overall accuracy of 85.5% and Kappa statistic of 0.779. Table 3-2. Error matrix and producers and users accuracy for accuracy assessment of 2005 classification based on 172 reference points Reference Data Classified Data Water Forest Woodland Rice Scrub / Crop Rubber Built Water 4 0 0 0 0 0 0 Forest 0 49 4 1 3 0 0 Woodland 0 0 6 0 2 0 0 Rice 0 1 2 77 7 0 0 Scrub / crop 0 3 1 0 6 1 0 Rubber 0 0 0 0 0 4 0 Built 0 0 0 0 0 0 1 % Producers Accuracy 100.0 92.5 46.2 98.7 33.3 80.0 100.0 % Users Accuracy 100.0 86.0 75.0 88.5 54.6 100.0 100.0 In Sisaket, the extent to which rice cultiv ation dominates the landscape is striking. Over the time period of the study, large areas of this SES have remained unchanged not only the rice production areas, but also the fore sted border escarpment and riparian corridors. Some key changes have taken place, however. The extrem e north-east, the escarp ment foothills and the three hilly areas to the south-east have seen th e emergence of upland crops, woodland and most recently rubber plantations. On the other side of the border, Ordar Mean Cheys landscape reflects its troubled political conditions. From 1989 to 1994 there is a reduction in area under rice, an d large increase in woodland, as the province experienced continued c onflict, as landmines were laid down by both the new government and the retreating Vietnamese -backed factions. After 1994 the area began to stabilize, particularly in the west. This led to the reestablishm ent of rice fields, and increased
83 settlement in the east. Associated with this settlement is a steady decline in forest cover. Most notable is the development of what was in 1994 a narrow track through the dense forest in the south-east of Ordar Mean Chey, into a dramatic wedge of deforestation by 2005. Figure 3-4. LULC classifications for Sisaket, Thailand and Ordar Mean Chey, Cambodia at each time-step in the study period. The circle on the 2005 classification shows the location of the wedge of deforestation in Ordar M ean Chey, while the rectangles highlight Sisakets extreme north-east, escarpment foothills and three hilly areas, where woodland has given way to upla nd crops and rubber plantations. Frequency Distributions of LULC Diversity At the micro-scale the distributions are le ft-skewed, as hypothesized in Figure 3-2. The range of LULC variety is smaller, limited by the moving-window size. A high magnitude of variety would suggest a landscape so fragmented that at this scal e, distribution of LULC types
84 would appear almost random. In fact, as Figur e 3-5 and Table 3-3 show, the mean magnitude of LULC variety is much lower, reflecting the ordered sp atial structure that is vi sible in Figure 3-8. Even though the mean values of LULC variet y at the micro-scale are not substantially different for the two landscapes (Table 3-3), thei r frequency distributions have different shapes (Figure 3-5). While both are le ft-skewed, the distribution across the different magnitudes is more even for Ordar Mean Chey. The shape of the frequency distribution is roughly the same for each year in the Sisaket landscape, but in Ordar Mean Chey, the right-hand tail grows longer in 2005, as is reflected in the mean value for that year which at 3.2 is noticeably higher than in previous years. This indicates that the Ordar Mean Chey landscape has experienced a considerable increase in the numbe r of its LULC types for that tim e-step. The differences in the distribution patterns for each land scape suggest that there are f undamental differences in the driving mechanisms in each SES. At this micro-scale, the distri bution of relative abundance clos ely follows that of variety. For example, over the study period, both variety a nd relative abundance peak in 1994 in Sisaket, and in 2005 in Ordar Mean Chey. Note too the large increase in both the mean and median values for LULC relative abundance for Ordar Mean Chey for that year (Table 3-4). An examination of the original classifications a nd imagery shows that in 1994, there was a large increase in the amount of shallow water and wetl ands across Sisaket whic h is attributed, on the basis of rainfall records, to the heavier rains associ ated with the previous years El Nio event. Not only do these areas of water themselves add to diversity at the local level, but they also provide farmers an opportunity to grow an extra crop. While the Sisaket variation appears to be due to climatic fluctuations, the spatial arrangeme nt of the patterns of diversity in Ordar Mean Chey suggests that in this SES, the increase in di versity is a response to settlement expansion.
85 Sisaket Micro-scale0 100000 200000 300000 400000 500000 600000 123456789No. of LULC TypesArea in Hectares 1989 1994 2000 2005 Ordar Mean Chey Micro-scale0 100000 200000 300000 400000 500000 600000 123456789No. of LULC TypesArea in Hectares 1989 1994 2000 2005 Figure 3-5. Frequency distri butions of LULC types in 1989, 1994, 2000 and 2005 at the microscale Table 3-3. Central tendencies of LULC va riety at the micro-s cale (3x3 pixel window) Sisaket Ordar Mean Chey Median Mean Std Dev Median Mean Std Dev 1989 2 2.72 1.30 3 2.85 1.34 1994 3 3.00 1.34 3 2.88 1.33 2000 2 2.79 1.28 3 2.68 1.21 2005 3 2.92 1.29 3 3.20 1.43 Table 3-4. Central tendencies of LULC relative abundance (Simpsons inde x) at the micro-scale Sisaket Ordar Mean Chey Median Mean Std Dev Median Mean Std Dev 1989 0.493 0.448 0.228 0.493 0.445 0.263 1994 0.493 0.494 0.215 0.493 0.450 0.253 2000 0.493 0.457 0.224 0.493 0.423 0.249 2005 0.493 0.471 0.231 0.590 0.498 0.256 As people have moved into the eastern parts of the province that was previously solely forest, they have cleared land using swidden techniqu es, creating patches of woodland, upland crops and rice. Repeated plowing has reduced the numbe r of trees in some of the first-established fields. Houses and stores have been built. Each of these activities has increased the number of LULC types present. The meso-scale is the scale at which patterns we re anticipated to be most evident, because the 30 m grain-size allows enough pixels to be present in the 100 ha window to show where
86 patterns emerge. Indeed, the fre quency distributions and mean va lues of diversity for the two landscapes differ more from each other at this sc ale than they do at the micro-scale. At the meso-scale, the distributions are slightly more right-skewed than predicted, and the distribution of LULC variety in Ordar Mean Chey is bimodal in all years (Figure 3-6). This suggests that parts of the landscape, as detected at this scale, are more strongl y influenced by social-ecological interactions, while others remain relatively untouched such as the large ar eas of dense forest to the east where conflict between the Khmer Rouge and the national government restricted settlement. Due to political in stability, the dense forests in th e east have had almost no human population and have consequently ex perienced very little human act ivity. In the more stable Sisaket Meso-scale0 50000 100000 150000 200000 250000 300000 350000 400000 1234567891011121314No. of LULC TypesArea in Hectares 1989 1994 2000 2005 Ordar Mean Chey Meso-scale0 50000 100000 150000 200000 250000 300000 350000 400000 1234567891011121314No. of LULC TypesArea in Hectares 1989 1994 2000 2005 Figure 3-6. Frequency distri butions of LULC types in 1989, 1994, 2000 and 2005 at the mesoscale Table 3-5. Central tendencies of LULC vari ety at the meso-scale (33x33 pixel window) Sisaket Ordar Mean Chey Median Mean Std Dev Median Mean Std Dev 1989 10 9.58 2.49 9 8.10 2.60 1994 11 10.28 2.18 9 8.02 2.57 2000 10 9.68 2.34 8 7.36 2.25 2005 10 9.57 2.31 9 8.18 2.33
87 Table 3-6. Central tendencies of LULC relative abundance (Simpsons index) at the meso-scale Sisaket Ordar Mean Chey Median Mean Std Dev Median Mean Std Dev 1989 0.613 0.609 0.182 0.741 0.634 0.254 1994 0.671 0.650 0.164 0.726 0.633 0.230 2000 0.642 0.632 0.160 0.706 0.619 0.235 2005 0.673 0.646 0.173 0.777 0.677 0.228 west, people have implemented a range of la nd uses, including rice production, upland crops farming, grazing and hunting in woodlands, and ha rvesting forest products from different stages of forest succession. Apart from this bimodality, Ordar Mean Ch ey as a whole also appears to exhibit considerable variation over time. At this scale, we are able to observe the strong decrease in both variety and relative abundance in Ordar Me an Chey in 2000 (Figure 3-6, Table 3-5 and Table 3-6), a trend that also exists at the mi cro-scale but which is somewhat overshadowed by the 2005 increases. Sisaket, on the other hand, app ears to have a much more stable distribution across time again with the ex ception of the 1994 increase in mean variety (Figure 3-6 and Table 3-5). At the macro-scale, the distributions for bot h provinces are, as hypothesized, right-skewed (Figure 3-7). If the macro-scale is intended to represent the context, the differences in LULC variety between the two landscapes certainly sup port this. Firstly, the range of LULC types observed in Ordar Mean Chey was smaller than in Sisaket. Secondly, the median value of LULC variety in Sisaket has remained at the maximum for all four years, while in Ordar Mean Chey this value has dropped over time (Table 3-7). In addition, the Sisaket la ndscape shows much less variability at this scale, although this appears to be increasing slightly over time. The differences in spatial variability can also be seen graphically in the 2005 example given in Figure 3-8 (below). The patterns of the mean values for both variety and relative ab undance at this macroscale do not suggest any particular trend (Table 3-7 and Table 38). This makes sense if one
88 considers this to be the scale of broader, slower-moving proce sses reflecting the provincial and national level. Sisaket Macro-scale0 100000 200000 300000 400000 500000 600000 700000 800000 1234567891011121314No. of LULC TypesArea in Hectares 1989 1994 2000 2005 Ordar Mean Chey Macro-scale0 100000 200000 300000 400000 500000 600000 700000 800000 1234567891011121314No. of LULC TypesArea in Hectares 1989 1994 2000 2005 Figure 3-7. Frequency distri butions of LULC types in 1989, 1994, 2000 and 2005 at the macroscale Table 3-7. Central tendencies of LULC va riety at the macro-scale (303x303 pixel window) Sisaket Ordar Mean Chey Median Mean Std Dev Median Mean Std Dev 1989 14 13.91 0.31 12 11.58 1.05 1994 14 13.54 0.55 12 11.70 0.99 2000 14 13.84 0.56 11 11.15 1.37 2005 14 13.68 0.68 11 11.37 1.06 Table 3-8. Central tendencies of LULC relativ e abundance (Simpsons inde x) at the macro-scale Sisaket Ordar Mean Chey Median Mean Std Dev Median Mean Std Dev 1989 0.658 0.659 0.130 0.791 0.708 0.214 1994 0.711 0.697 0.114 0.793 0.711 0.191 2000 0.684 0.685 0.109 0.780 0.706 0.197 2005 0.714 0.704 0.119 0.820 0.749 0.175 Not surprisingly, as we increase the scale, the mean values for LULC variety and relative abundance increase (Table 3-9 and Ta ble 3-10). This is because the scaling up process used is similar to the ecological species-area curve, wh ere the larger the area sampled, the greater the likelihood of capturing all LULC types. A comp arison of the different mean values at the
89 different scales, however, shows no consistent relationship between the scales over time, nor does there appear to be a strong link between variety and relative abundance. If the two measures were to follow the same trends in sp ace and time, this would suggest that all LULC types were changing in the same manner with resp ect to their initial extent. However, socioeconomic and environmental factors tend to select preferentially for some types more than others. For example, built areas emerge where settlements become established in the latter stages of the study period, and onl y account for very small proportions of changed areas. At the wedge of deforestation in Ordar Mean Chey, rice, crops and plantations do not replace forest in equal proportions. In Sisaket, some LULC type s, such as rubber, that existed only in some localized parts of the landscape have emerged at more and more locations, whereas woodlands have been steadily declining. Table 3-9. Comparison of mean LULC variety at three scales Sisaket Ordar Mean Chey Micro Meso Macro Micro Meso Macro 1989 2.72 9.58 13.91 2.85 8.10 11.58 1994 3.00 10.28 13.54 2.88 8.02 11.70 2000 2.79 9.68 13.84 2.68 7.36 11.15 2005 2.92 9.57 13.68 3.20 8.18 11.37 Table 3-10. Comparison of mean LULC relative abundance at three scales Sisaket Ordar Mean Chey Micro Meso Macro Micro Meso Macro 1989 0.448 0.609 0.659 0.445 0.634 0.708 1994 0.494 0.650 0.697 0.450 0.633 0.711 2000 0.457 0.632 0.685 0.423 0.619 0.706 2005 0.471 0.646 0.704 0.498 0.677 0.749 The frequency distributions and measures of cen tral tendency show that LULC diversity is a useful, quantifiable SES charac teristic. They show that there is a defined, non-random, structure to the distribution of both LULC variety and relative abundance. There are two main strengths to using LULC diversity as a lands cape generalization. The first is that the
90 distributions are shaped differen tly for the two provinces at a ll three scales, with the same general shape being held for each of the landscape s at different points of time. This general distribution is important, because it suggests that, at least in the short-term, the landscapes are maintaining their identity over time (Cumming and Collier 2005). These landscapes therefore provide information on the state of the two SESs, and show that LULC can indeed be used to represent SESs. The generalization to LULC divers ity allows us to express spatial and temporal variation in the landscape as a whole, and to compare this overall vari ation in a quantitative manner, even where the types of LULC contributing to the vari ation might be different. For example, the SES in Sisaket is more diverse th an that of Ordar Mean Chey. Since the two landscapes have the same climate and geophysical characteristics, the differences in overall magnitudes of diversity can be attributed to differences in socio-economic conditions. Even though Sisaket might have a greater range of activiti es within its SES, its landscape is dominated by only a few of those activities, whereas in Orda r Mean Chey, there is a much more equitable representation of the different activities on th e landscape. If increasing dominance is an indication of path-dependency rela ted to stage of adap tive cycle (Chapter 2), this would suggest that during the study-period, Sisaket wa s some way along the conservation ( k ) phase, while Ordar Mean Chey was still em erging from the exploitation ( r ) phase (see Gunderson and Holling 2002). The second strength of LULC diversity as a landscape generalization is in the way in which within each SESs general LULC diversity distribution shape the measures of diversity vary over time. These variations contain important information. For example, across the study period, Sisaket had a peak of change in 1994, whereas for Ordar Mean Chey this was in 2005. This shows that each landscape, as a whole, is responding to different processes. Additionally,
91 we can explore at which scale the variation is strongest. Since different underlying processes or mechanisms are scale-dependent, this informa tion provides a guide to what those processes might be. Spatial Distribution of LULC Diversity LULC variety A visual inspection of LULC variety shows that while patterns are discernable at all three scales of analysis, they are clearest at the meso-scale of in terest (Figure 3-8). It is at this scale Figure 3-8. Spatial distribution of LULC variet y in 2005, shown for the three different scales and in comparison to the initial classifica tion. At the micro-sc ale, patterns appear more random, while those at the macro-sc ale appear more uniform. The insets, centered on the same riparian area, each cover a spatial extent representing roughly ten times the moving window size used for each scale, and are intended to show that these patterns are not solely due to the ex tent of the map, but also hold constant relative to the moving window size.
92 that one can clearly discern the effects of nodes and connectivity provided by settlements, roads and rivers. At this scale in Sisaket, which has been densely populated and intensely cultivated for centuries, high diversity follows the riparian co rridors as well as the hilly areas already noted to the south-east and north-east of the province, and along the foothills of the escarpment. In Ordar Mean Chey, on the other hand, there are ve ry few areas of high LULC variety. These areas are associated with the villa ge of Anlong Veng, and with the ro ad in the west that links the district capital of Samraong to the border in the north and the rest of the country to the south. At all scales, areas with less hu man-environment interactions ge nerally have the lowest LULC diversity. Notable exceptions are the areas of extensive rice fields in Sisaket, which show how much this crop dominates the landscape at some locations. The macro-scale strongly reflects the national-level differences in the two SESs. At this scale patterns do not follow topographical features; instead they show the overall condition for each landscape. LULC relative abundance The maps of the Simpsons index show how equitably the different LULC types are represented at any given location. All three scales show that, in terms of relative abundance, Ordar Mean Cheys diversity appe ars to be spatially separated into areas of dominance by one type, and other areas where distribution of types is more even (Figure 3-9). This abrupt gradient of change suggests a frontier region, and is a cons equence of the political conflict that has only recently abated to a level that allo ws development within the SES. In Sisaket, there is much more of a smoot h gradient between areas of dominance by one type, and those where most types are present. The north-eastern area, escarpment foothills and three hilly areas highlighted in Figure 3-4 al so display much higher magnitudes of relative abundance. The well-established provincial capit al of Sisaket town, which is located in the
93 north-centre at the fork in the rivers, is clearly visible as a bright circle of high relative abundance of LULC types. Figure 3-9. Spatial distributi on of LULC relative abundance in 2005, shown for the three different scales and in comparis on to the initial classification. Even with a generalized landscap e attribute such as LULC di versity, the information about the how the underlying mechanisms differ in each landscape is not lost. From even the most cursory examination of the diversity maps, it is clear that there is defi nite structure to the distribution of LULC diversity, bot h in terms of variety and relati ve abundance. One can trace the patterns of diversity relativ e to social and environmental f eatures, such as roads, market villages, rivers and hilly areas. At each scale we can see differences between the two SESs in terms of the spatial distribution of their LULC di versity. At the micro-scale, Sisaket has a more
94 spatially ordered structure to the distribution of th e number of LULC types. At the micro-scale, the Sisaket SES is so highly connected by its networ k of roads and settlements, that these appear to have less of an influence on LULC variety than environmental factors do. Much of its landscape has areas of intermediate LULC dive rsity, with high diversity located along river systems and in gently sloped hills. Areas of low LULC variety are restricted to steeper, forested escarpment of the protected border area. At this scale, in contrast, Ordar Mean Cheys landscape appears to be a complex mix of high a nd intermediate LULC variety, with the hilly forests in the south-east being th e most discernible exception. At the meso-scale, the patterns in Sisaket do not change relative to the micro-scale, but become more marked. In Ordar Mean Chey, however, we begin to see roads and villag es stand out as nodes a nd linear features of higher diversity. In this SES, accessibility is st ill key in determining LULC variety distribution. Perhaps unsurprisingly, it is at the macro-scale of LULC variet y that the contrast between the two provinces becomes most evident, althoug h the degree to which the two SESs differ is remarkable. This difference underscores the exte nt to which their diffe rent national identities have subjected the provincial-level SESs to diffe rent contexts and differe nt sets of constraints (Allen and Starr 1982). Access and economy, as part of these contextual factors, determine the level of infrastructural development. The spat ial arrangement of LULC variety suggests that infrastructure is well-developed all over Sisaket, while in Ordar Mean Chey, it is limited to localized areas surrounding the few major towns. LULC relative abundance is a more nuanced m easure of diversity than LULC variety, and this is evident in the maps, particularly at th e macro-scale. Nevertheless, the Simpsons index maps also show strong differences between the two provinces. While Sisaket shows a good range in degree of dominance by some LULC types, in Ordar Mean Chey, there appears to be
95 either complete dominance by one type (dense forest ), or a relatively equitable distribution of all present types. In Ordar Mean Ch ey these patterns relate to rapi d recolonization of an area that was inaccessible for a long period. In Sisaket, on the other hand, access is less of a limiting factor because there are so many roads. The re lative abundance patterns al so suggest that LULC diversity is linked to environmental conditions, such as areas of higher soil fertility that are found along the rivers or in the escarpment foothills. Change in Spatial Distribut ion of Diversity over Time Apart from minor fluctuations, the Sisake t landscape shows no substantial change in overall magnitude of diversity over time, suggesti ng that at least for the period of study this landscape as a whole is stable (Fi gure 3-10). Temporal variation se ems to be limited to a slight Figure 3-10. Spatial distribution of LULC variety in 1989, 1994, 2000 and 2005, shown at the meso-scale. The circle on the 2005 ma p shows the wedge of deforestation.
96 increase in diversity in 1994 before returning to a more intermediate overall magnitude in 2000 and 2005. This fluctuation is possibly due to the effect of increased ra ins from the previous years El Nio conditions. In 1994 there was an increase in the area of shallow water and wetlands, and in the areas of upland crops and ot her plantations. In Ordar Mean Chey, the changes are more dramatic. Most notable is that the wedge of deforestation visible in the southeast in Figure 3-4 is now visible as a wedge of in creasing LULC diversity. The effect of access on LULC diversity is also visibl e as ribbons of high diversity cut through areas of lower diversity as a consequence of the developm ent of roads and settlements. Temporal Trends in LULC Variety and Variation in LULC Relative Abundance The trajectories of change in LULC variety ac ross the four time-steps were created for all three scales. These were done across the entire study area, treating the two provinces as a single surface to highlight the extent to which certain trends were taki ng place only in one of the two landscapes. As anticipated, not all of the poten tial trajectories were filled, and only a small fraction 19 trajectories (0.28% of those possible) at the micr o-scale, 370 (0.96%) at the mesoscale, and 143 (0.37%) at the macro-scal e covered areas of 10,000 pixels (9 km2) or more, hereinafter called extensive trajectories. By plotting out mean LULC variety over time against the variance in LULC variety over time for each trajectory, we can see what types of trajectories are missing. Figure 3-11 shows the potential, actual and extensive trajectories, wi th each point representing a specific trajectory. The first row shows the plots for every possible combination of LULC va riety across the four time-steps, whether that combination was found in the study area or not. Most of the points are concentrated in areas of the graphs that relate to intermediate magnitudes of diversity, and low variance. A comparison of the actual-occurring tr ajectories with all potential ones shows what combinations of LULC variety magnitudes are li kely in the real world, and what types of
97 trajectories account for the dominant patterns on the landscape. The graphs in Figure 3-11 show that large swings from high to low to high to low diversity at any location are conceptually unlikely most of the missing trajectories, at all three scales, are t hose that have both high means and variances. Indeed, when considering only those trajectories that cover extensive areas, there are none which have th e highest theoretically possible variance. The circles on the Micro-scale Meso-scale Macro-scale Potential Trajectories Actual Trajectories Extensive Area Trajectories (> 9 km 2) Figure 3-11. Scatterplots of mean LULC variety vs. variance of LULC variety for the combined study area, comparing values for actual and extensive trajec tories against all theoretical potential trajectories. Extens ive areas with the greatest variance are circled. Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time Diversity mean over time Diversity variance over time
98 graphs representing extensive areas identify those trajectories th at both cover large areas, and have the highest observed variance. These trajec tories are the ones that represent the greatest change in magnitude of diversity. By focusing on these extensive trajectories with the highest variance over time, we can identify the most dynamic parts of the landscape. At the micro-scale, the trajecto ries of diversity th at cover the greatest extent are all areas that remained at low magnitudes of diversity ac ross the entire time period (Figure 3-12). The most extensive trajectory is, in fact, that with only one type of LULC, corresponding to undeveloped forested areas, occurring in the hilly regions of both provinces landscapes. The second most extensive trajectory is that where there are only 2 types of LULC types at all four time-steps. Note, however, that in Sisaket, this relates to areas of rice production, while in Ordar Mean Chey, it relates to riparian forest. The third most extensive trajectory at the microscale shows a slight increase at the end of th e trajectory period, and occurs far more extensively in the Sisaket landscape than that of Ordar Mean Chey. The fact that the most extensive trajectories at this scale have such little variance, and such low variety at all points of time, might seem to be imply that at the micro-scale, the entire landscapes is homogeneous and unchanging. However, these three trajectories together account for only 14.9% of Sisaket and 9.3% of Ordar Mean Chey. Since the rest of the landscape is covered by other much smaller trajectories, in fact both provinces are both spatially and temporally variable at this scale. As with the static analyses shown in Figure 3-5 to Figure 3-7, the dominant trajectories reflect the influence of scale, increasing in mean value (Figure 3-11) as th e scale shifts upwards. At the meso-scale, the spatial distribution of eac h of these trajectories is once again location
99 specific, in spite of appearing somewhat patchy (Fi gure 3-12). As with the micro-scale, all of the most extensive trajectories have ve ry little variance. At this scal e, the most extensive trajectories occur mainly in Ordar Mean Chey. The bimodality for this provinces lands cape that is evident in Figure 3-6 is picked up again in the trajectori es and reflects path-dependency that is, areas that were initially extensive in 1989 will determine which extensiv e trajectories are possible. While the most extensive trajectory is at an intermediate magnitude of LULC variety, the second-most extensive is at a low magnitude. This second-largest trajecto ry is associated with the forested areas to the south-ea st, while the other most extensive trajectories are all located in the western part of the province wh ere internal conflict remained relatively lower. The fact that Figure 3-12. Spatial distribution of trajectories of change in dive rsity derived for each scale from that scale's four time-step s of LULC variety. Each number in the legend name represents the number of LU LC types in the relevant pixel-window neighborhood at each time-step.
100 such small areas (less than 0.9 % of Sisakets la ndscape) of the extensive trajectories are found in Sisaket suggests that in this province there is great variability in magnit udes of diversity at the community level. The influence of the nationallevel context is clearl y evident at the macroscale. The extensive tr ajectories dominate the Si saket landscape (80.6%), but are entirely absent from Ordar Mean Chey (Figure 3-12). The trajec tories are all of persistently high magnitudes of LULC variety, in strong cont rast to the micro-scale. Turning to the extensive trajectories with the greatest variance in LULC variety over time, the most striking fact is that, at all three scales, the greatest va riance occurs in the wedge of deforestation in the far west of Ordar Mean Chey and almost ex clusively at that wedge (Figure 3-12). Dramatic change is not occurring anywhe re in Sisaket, and only at the macro-scale do places in Ordar Mean Chey other than the defo restation wedge show considerable variation. Note that for each high-variance trajectory at this broad scale, a single contiguous area is affected. The northern-most block of change (1 2-13-11-9) relates to th e opening of a border-post with the Thai province adjacent to Sisaket. S outh and east of that is the town of Anlong Veng (12-13-9-11). The similarity of th ese two trajectories is a reflecti on of the opening up of this part of the province with the declin ing influence of the anti-govern ment Khmer Rouge party, leading to increased settlement, clearing of land, and dive rsification of land-uses in response to economic demands. From the sequence of the values of diversity in the trajectories, we can also see the directionality behind that variance. At all three scales most of the area under the high-variance trajectories shows a trend of increasing LULC vari ety over time. At the microand meso-scales, it is clear that most of the change in Or dar Mean Chey occurred between 2000 and 2005. Although the values for the most variable macroscale trajectory, and th e timing of the increase
101 in magnitude of LULC variety, differ from those in the other scales, the lo cation at the core of the deforestation wedge is exactly the same. Due to the continuous nature of the Simp sons index, it is not possible to produce meaningful trajectories of change in LULC re lative abundance. Neve rtheless, the standard deviations over time are still revealing in spit e of the non-direc tional nature of these data. The maps of standard deviation in LULC relative abundance over time (Figure 3-13) show that this measure, like that of LULC variet y (Figure 3-12), is most variable at the wedge of deforestation in Ordar Mean Chey. Again, this dramatic change is visible at all three scales. Figure 3-13. Spatial distribution of temporal va riation in LULC relative abundance, calculated for each scale as the standard deviation of th e values for each time-step at that pixel. The inset map shows that not all change is an increase over time, as some places (with negative values) expe rienced a net decrease in re lative abundance between 1989 and 2005.
102 At the micro-scale, which shows the eff ects of household-level decision-making, the landscapes in both provinces show that some areas have a fair amount of variation in relative abundance. Sisaket has a mean value of 0.147 in standard deviation for its whole landscape at this scale, while in Ordar Mean Chey this va lue is 0.152. This suggests that in both provinces farmers engage in different activit ies at the same location from year to year. At the meso-scale, which relates to the community, considerably le ss variation is expre ssed in both landscapes (Sisaket 0.062, Ordar Mean Chey 0.073), even though in the latter province there is extensive change. When we consider the mean magnitudes of di versity for each SES, the overall variation over time is minimal. This makes sense because the study period represents only a short segment of each SESs history. Indeed, the shor t duration adds significa nce where some places have experienced significant change. If we look at both diversity measures and all scales at the same time, there do not appear to be any definite trends in the mean magnitudes of diversity for the landscapes as whole SESs over the 16-year pe riod. Instead, there appear to be fluctuations within essentially stable (in this short-term) ma gnitudes of LULC diversit y. Within each of the whole landscapes, however, considerable change is taking place in specific, more localized areas. If we express these landscape changes in the la nguage of CASs, we can see the graphs of the means and variances of the LULC variety trajectories as evidence of non-random structure leading to the emergence of patte rns of change. Not only have some of the unlikely potential trajectories been winnowed awa y, but only a few trajectories cover extensive parts of the landscape (see Chapter 2). This means it is easy to identify dominant trends within each SES, not only in terms of what those trends are, but also where they are located. The most extensive trajectories suggest that there is some dire ctionality and path-dep endency, where initial
103 conditions appear to be dicta ting what follows afterwards, an d whether a given magnitude of LULC variety will persist over time. It is the high variance traject ories that most support the idea that LULC diversity is a useful generalization that allows cross-site comparison while still capturing important change in a landscape. The fact that Ordar Mean Cheys area of deforestation is prominent in each scale of both measures of change in diversity show s clearly that diversity has meaning on the SES landscape. If the assumption of LULC as an expr ession of a SES is corre ct, the location of the most extensive trajectories in th e more politically stable west of Ordar Mean Chey suggest that complexity in this area of the SES is being developed. Further Studies Having established that diversity as a SES char acteristic is detectable in LULC landscapes, and that landscapes can be compared and contrasted in terms of this concept, the next step is to test how LULC diversity responds to known drivers of LULCC, si nce typically it is through an understanding of the causes of ch ange that researchers are able to predict and plan for that change. In this paper we have qualitatively de scribed some of the factors that might underlie existing diversity patterns, or l ead to changes in the distributi on of diversity. If a formal analysis of the response of diversity to these f actors is as revealing as exploring how actual LULC types respond, then this would provide a pa th out of the context-specificity that is currently beleaguering research ers. From the patterns observed on the landscape, we hypothesize that LULC diversity is generally higher closer to roads and market settlements, factors that influence accessibility, and that are kn own to lead to change in specific LULC types in other places (Southworth and Tucker 2001; Ver burg et al. 2002; Walker and Soleki 2004). In addition, in both Sisaket and Ordar Mean Chey, low LULC diversity appears to be located in higher areas with steeper slopes, factors that have been associ ated with the distribution of
104 specific LULC types (Mas et al. 2004; Overmars and Verburg 2005). In Sisaket, areas of high diversity are located in the kind of low-lying hills that are associated with more fertile soils which are associated with high ra tes of LULCC in the lowlands of Colombia (Etter et al. 2006). The combination of stability ove r time of diversity magnitudes with high population density in Sisaket suggests that land tenure plays an im portant role in regulat ing magnitudes of LULC diversity, much as land rights affect agricu ltural expansion and other forms of LULCC (McConnell, Sweeney, and Mulley 2004). These, a nd other, hypothesized relationships can form the basis for comparing the response patterns of different diversity landscapes to underlying mechanisms. Other methods of accommodating the scale effect could be usefully explored, such as trying different grain sizes. Sli ght adjustments might increase the emergenc e of patterns, while larger grain sizes could reduce pr ocessing time while still revealing the same information. It would also be useful to experiment with merged or grouped LULC categories, as this represents another way of scaling up. In theory, a grouped classification of around 6 cl asses, such is more commonly used by land change scientists, should yield results with similar means but greater variances. Finally, additional consideration could be gi ven to increasing the time period. Palaeoecological studies of historical societies show that successful SESs persist for hundreds of years (Binford et al. 1997; Janssen, Anderies, and Ostr om 2007). To adequately pick up trends over this temporal extent, studies s hould span at least three generati ons of the dominant organism humans. This would suggest a period of approxima tely 75 to 100 years. At present this would require turning to alternative sources than remote sensing to provide maps of the earlier LULC states particularly in developing countries. However, new techniques are being tried, and as
105 time passes, the aerial photograph and satellite imagery datasets are providing longer windows of analysis. Conclusion This paper represents a necessary first step in evaluating the utility of LULC diversity in linking LULCC to SESs research: th e verification that it is indeed possible to use the concept as a generalization for comparing a nd contrasting different SESs. That human activities generally lead to greater LULC diversity is perhaps obviou s. What is less obvious is that these landscape patterns can be used by SESs researchers to evalua te the condition of a SES of interest. Framing LULCC in concepts used in that field makes the spatial, quantifiable information founded on LULCC available to them. This research shows th at LULC diversity is di stributed in distinct patterns across the landscapes of the two study area SESs. Thes e spatial patterns show that different underlying processes linke d to the roles of rivers, soil, elevation, roads and settlement are visible at different scales. LULC diversit y provides a way to quantify SESs as expressed on the landscape, and as such can be used to measure system characteristics such as change and persistence (Carpenter et al. 2001; Cumming et al. 2005). Because the LULCC approach is spatial, a focus on changes (or not) in LULC di versity reveal where system identity is changing or persisting. By treating landscapes as SESs, LULCC res earchers can draw on generalizations from CASs theory such as diversity to compare ch ange in different landsca pes in their entirety even where the nature of that change might differ because the landscapes comprise different LULC types. The research described here leads us to conclude that LULC diversity serves as a unifying, non-ecosystem specific characteristic for cross-site comparison. As such it offers researchers the potential to overc ome context-specificity (Perz 2007). As a quantifiable concept, LULC diversity offers an additional advantage over the standard classification approach to
106 LULCC. Because the measures of diversity ar e discrete (variety) a nd continuous (relative abundance) as opposed to categorical, this incr eases the options for exploring spatial and temporal variability in a quantitative manner. As a theoretical construct, LULC diversity provides additi onal insight into the humanenvironment interactions that are shaping our world, particularly when considered through the framework of a CAS. Diversity can be evaluate d at different scales, reflecting the multi-scalar nature of SES processes (Ahl and Allen 1996). The spatial arrangement of LULC diversity suggests that the SESs are self -organizing, with patterns emergi ng from interactions at lower levels (Holland 1995; Levin 2005). Magnitudes of diversity at any give n location within the landscape can decrease as some LULC types ar e winnowed away, or increase, as new types evolve such as the conversion of rice fields to upland crops and r ubber plantations (Levin 1999). Change in LULC diversity is non-linear, as evidenced by the change trajectories with high variance. This threshold-type response is another characteristic of a complex, adaptive SES (Folke et al. 2004; Holland 1995; Holling and Gunderson 2002). By linking LULCC to SES research through th e CAS framework, we are able to provide the former with the necessary concepts to deve lop an umbrella deducti ve approach for guiding the emerging discipline of land ch ange science (Rindfuss et al. 2004). At the same time, by using LULC as the physical expres sion or representation of a SES, we provide the latter with the means to start quantifying certain types of SE Ss (Carpenter et al. 2001; Cumming et al. 2005).
107 CHAPTER 4 SOCIAL AND ECOLOGICAL FACTORS AND LAND-USE LAND-COVER DIVERSITY IN TWO PROVINCES IN SOUTH-EAST ASIA Introduction Studying human-environment interactions th rough the lenses of land-use land-cover change (LULCC) and social-ecological systems (SES s) research offers a broader perspective, but each approach faces certain challenges. LU LCC research has excellent tools to measure landscape-level change, but is str uggling to develop a theoretical framework that is generalized enough to surmount disciplinary boundaries or the complexities of comparisons beyond specific locations (Brgi, Hersperger, and Schneeberge r 2004; Carpenter et al. 2001; Kuhn 1996; Perz 2007; Walker 2004). On the other hand, SESs re search that is, research conducted through programs such as the Resilience Alliance that sp ecifically focus on incorporating both social and ecological aspects through a systems approach was developed under the umbrella of a broad conceptual framework that of complex adaptiv e systems (CASs) but, with a few exceptions, has been unable to find ways to quantify change (Chapter 2; Cumming et al. 2005; Carpenter et al. 2001, Anderies et al. 2006). Since these research program s both address the dynamics of human-environment interactions, it seems apposite to explore whet her the strengths of each can address the weaknesses of th e other (see Figure 4-1). This study explores empirically how LULC dive rsity, as a higher level CAS characteristic, responds to the biophysical and socio-economic fact ors that are known to influence change at a lower level in individual LULC types (Alados et al. 2004; Brgi, Hers perger, and Schneeberger 2004; Chomitz and Gray 1996; Crews-Meyer 200 4; Douglas 2006; Erenstein, Oswald, and Mahaman 2006; Etter et al. 2006; Munroe, S outhworth, and Tucker 2002; Overmars and Verburg 2005; Van Gils and Lo za Armand Ugon 2006). It explores the distribution of LULC
108 diversity in response to elevation, distance to ro ads, and distance to markets, and evaluates how the resultant distributions compare to hypothesized relationships (Chapter 2). Figure 4-1. Schematic diagram showing how the divergent SESs and LUL CC approaches to the study of human-environment interactions can be linked. The quantitative analyses used in LULCC research call for measurable drivers, or mechanisms, while those used to describe change in SESs draw e xplanations from generalized abstractions. LULC diversity is the proposed link, sin ce it combines a generalized CAS concept with a quantifiable landscape characteristic. The CASs approach underpinning SESs res earch emphasizes the dynamism and change inherent in human-environment interactions (Berkes and Folke 1998). The key focus is on system resilience that is, how systems persist and maintain identity as they move through a range of conditions and are subjecte d to a range of internal and ex ternal perturbati ons (Carpenter et al. 2001; Holling 1973). The concept of SES resilience is built on the explicit recognition of such systems as being complex and adaptive (Holling 2001; Janssen, Anderies, and Ostrom
109 2007; Levin 1999; Walker and Abel 2002). The empirical, spatially explicit approach of LULCC research allows for the spatial and temporal location of the system to be analyzed (see Figure 4-1). With the increasi ng abilities of the applications of technologies such as remote sensing and geographic information systems (GIS), researchers can not only measure, but also map and model change and dynamism in LULC (Briassoulis 2000; Irwin and Geoghegan 2001; Manson 2005; Mas et al. 2004; Parker et al. 2003; Rogan and Chen 2004; Stibig, Beuchle, and Achard 2003; Verburg et al. 2002; Wessels et al. 2004; Wulder et al. 2007). The responses of landscape patterns to both local, proximate driv ers and larger, more regional factors help researchers understand spatial va riability (Lausch and Herzog 2002; Turner, Gardner, and ONeill 2001; Wood and Porro 2002). Here we test whether LULC diversity responds to these drivers, so that this generalized concept can be used to compare conditions in a range of landscapes. Diversity is a good place to start, since it is the most fundamental CAS characteristic, providing the complexity of compon ents and range of responses required for adaptation (Levin 1999). Diversity is also readil y applicable to LULC research The spatial arrangement and distribution of different LULC types define a landscapes diversity. Further, landscape heterogeneity is a familiar concept, and the spatial metrics of diversity derived by landscape ecologists are increasingly being adopted by LU LCC researchers (Cadenasso, Pickett, and Schwartz 2007; Lausch and Herzog 2002; Nagendra, Munroe, and Southworth 2004). Research questions The patterns of LULC diversity are useful to both LULCC and SESs research because they describe the system as expresse d on the landscape and provide ge neralized information to allow cross-site comparison (Chapter 3). LULC di versity patterns should respond to the same mechanisms that are known to affect spatial and temporal variation in the distribution of
110 individual LULC types. The way in which LULC diversity relates to these mechanisms provides important information on the state of the SES. As suggested in Chapter 2, we can hypothesi ze about the relationships between LULC diversity and different mechanisms by assessing how the range of LULC types varies spatially. For example, we expect LULC diversity to be lo w very close to roads, then increasing sharply within accessible distance to the road, before d ecreasing steadily (Figure 4-2). We base this expectation on how household-level strategies play out collectively. For example, beyond easy foot access, human influence decreases and mainly natural vegetation is found. This means that the dominant crop (which in most subsistence communities is the most important LULC type) would be found along rural roads to the general exclusion of other types. Further from the road, subsistence crops are interspe rsed with less frequently visite d areas such as long-term tree crops, pastures, as well as a ra nge of disturbed and undisturbed natural vegetation. Beyond easy foot access, human influence decreases and mainly natural vegetation is found. Likewise, right in the vicinity of the market, households activitie s will relate mainly to residence and the most important crop, and the number of LULC types will be low. Diversity will increase strongly with distance, since immediately beyond the resi dential areas householders will have a range of activities that are support ed by market demands. At some point access becomes a limiting factor to the viability of market-inf luenced LULC types, so that LULC diversity tails off with increasing distance. We hypothesize that cl ose to markets built ar eas will dominate the landscape, surrounded immediately by only the most important commercial crop, while further out commercial crops become interspersed with increasing areas of subsistence crops and a range of natural vegetation types. As with roads, as human influence drops off, natural vegetation will dominate the landscape (Figure 42). We suggest that the distri bution of LULC diversity with
111 increasing elevation will start off with intermedia te magnitudes in low areas, increasing steadily with increasing elevation until steepness and rockiness causes a sudden drop-off in the number of LULC types. Lower areas are more accessible, and will be dominated by the main subsistence crop. With increasing elevation, changes in soil and slope make other crops more viable, so subsistence crops are interspersed with a range of tree crops and other commercial crops, as well as natural vegetation. At higher altitudes limite d access and thin soils make agriculture less viable, so that only undis turbed natural vegetatio n is found (Figure 4-2). a) b) c) Figure 4-2. Hypothesized relations hips between LULC diversity a nd a) distance to roads, b) distance to market, and c) elevation. We pose the following questions to test whet her these hypotheses correctly identify the nature of these relationships: How does LULC diversity change with dist ance to roads, distance to markets, and elevation? How do the general relationships between LU LC diversity and these factors vary at different scales? To what extent does the actual distribution of LULC diversity agree with that predicted by models describing the hy pothesized relationship? Study Area Cross-border studies provide an excellent opportunity for the comparative study of differences in the processes behind LULCC change (Brgi, Hersperger, and Schneeberger 2004). This study tests the utility of the LULC dive rsity concept in the neighboring provinces of
112 Sisaket, Thailand and Ordar Mean Chey, Cambodia in an analysis of change over a 16-year period from 1989 to 2005. The two provinces cu rrently have highly dissimilar landscape configurations (Figure 4-3). We consider the two landscapes to represent the SESs of the two provinces, with the system processes repres ented by the patterns of four 14-category classifications (Figure 4-4) of Landsat TM and ETM+ imagery from seasonally comparable dates (see Chapter 3). The 14 classes, as described in Chapter 3, were identifie d using data collected on extensive field visits in 2005 and 2006, comparis ons with related projects, and earlier maps (Blasco, Bellan, and Lacaze 1997; Mekong Rive r Commission 2003; Nang Rong Projects 2004). A low escarpment, dropping from Thailand into Cambodia, separates the two provinces. As a result of uplift, the plateau drains north into Thailand, resulting in hydrological separation. Figure 4-3. Study area map showing the different spectral characte ristics of Sisaket, Thailand and Ordar Mean Chey, Cambodia
113 Sisaket and Ordar Mean Chey share genera l edaphic, geomorphological and climatic characteristics. Both are for the most part fl at. A few large rivers cross them, and both were originally covered with semi-deciduous moist trop ical forest. They both lie within the Asian monsoonal tropics, and have pronounced, fiveto seven-month, rainy seasons. Currently, however, the provinces have ve ry different landscape configur ations, the consequence of different social, cultural, po litical and economic histories (Chandler 2000; Wyatt 1984). In addition, Cambodias internal strugg les since the 1970s have deepened its isolation from the rest of the world. Figure 4-4. LULC classifications for Sisaket, Thailand and Ordar Mean Chey, Cambodia at each time-step in the study period. The circle on the 2005 classification shows the location of the wedge of deforestation in Ordar M ean Chey, while the rectangles highlight Sisakets extreme north-east, escarpment foothills and three hilly areas, where woodland has given way to upla nd crops and rubber plantations.
114 Sisakets predominantly agrarian landscape is defined by a dense hub-and-spoke network of mostly paved roads that radi ate through a vast expanse of rice fields. Services are good, and all villages have electricity, piped water, and access to health and educat ion facilities. Ordar Mean Chey, in contrast, had no pa ved roads during the study period.5 Because this province was the last holdout of the Khmer Rouge guerrill as, it remained inaccessible to the national government until only recently. As a result, living conditions are very poor. There is no piped water, no mains electricity, and onl y a few larger villages have schools and clinics (Chapter 3). As Figure 4-4 shows, Sisakets landscape is dominated by extensive areas of rice fields traversed by a few large rivers. The rice cultivation areas, and th e forested southern escarpment, have changed little during the study period. More locally, some dynamism in the SES is evident in the change in LULC in the far north-east, th e three hilly areas to the south-east, and in the foot-hills of the forested escarpment. Thes e areas have transitioned between woodland, upland crops, and increasingly, rubber plantations. The influence of intensive agriculture eviden t in Sisakets landscap e provides a marked contrast to that of Ordar Mean Chey. Much of the rice production areas visible in the west of the Cambodian province in 1989 are, according to oral histories, collective farms that collapsed somewhat by 1994 due to ongoing internal conflict between the Khmer Rouge guerrillas and the new government. Areas of woodland expanded with land abandonment, presumably in part due to the presence of land mines. After 1994 the we stern part of Ordar Mean Chey stabilized, leading to the re-emergence of rice areas. By 2000, the village of Anlong Veng (visible as a small zone of rice around a reservoi r due south of Sisakets western border) expanded as this part of the province became the last hol d-out of the Khmer Rouge. The eastern half of this province 5 The first road was paved in late 2005 / early 2006.
115 then began to experience stea dy deforestation. A key development was the creation of a track through the south-eastern forest block in 1994 that opened up the area and led to the expansion of agriculture around the village of Trapeang Prasat. This is visible as a dramatic wedge of rapid deforestation between 2000 and 2005. Methods Development of Datasets This spatially explicit study compares th e landscapes underlying the two SESs, as described by the classifications shown in Figure 4-4 and discussed above. The 14 classes include several that fall on a continuum ranging from dense forest through less dense forest, dense canopy woodland, sparse canopy woodland, rice under sp arse trees to treeless rice areas. The classification errors are in part a reflection of the somewhat arbitrar y cut-off points imposed along this range (Table 4-1). The class scrub / upland crops was most difficult to separate from others. This is a result of the cla ssifications being based solely on the spectral characteristics of the satellite imagery, which in some instances are very similar for different LULC classes. The accuracy assessment for the gr ouped classes resulted in an overall accuracy Table 4-1. Error matrix and producers and users accuracy for accuracy assessment of 2005 classification based on 172 reference points Reference Data Classified Data Water Forest Woodland Rice Scrub / Crop Rubber Built Water 4 0 0 0 0 0 0 Forest 0 49 4 1 3 0 0 Woodland 0 0 6 0 2 0 0 Rice 0 1 2 77 7 0 0 Scrub / crop 0 3 1 0 6 1 0 Rubber 0 0 0 0 0 4 0 Built 0 0 0 0 0 0 1 % Producers Accuracy 100.0 92.5 46.2 98.7 33.3 80.0 100.0 % Users Accuracy 100.0 86.0 75.0 88.5 54.6 100.0 100.0
116 of 85.5% and Kappa statistic of 0.779. These classifi cations are taken to be the starting point of this study. We created raster-based maps of LULC di versity at three diffe rent scales by running moving windows of 3x3, 33x33, 303x303 pixels (Table 42) across the four in itial classification images. The possible values of LULC variety ra nge from a low of 1 to a high of 14 (because the classifications have 14 LULC types). The el evation map was derived from the 90 m Shuttle Radar Topography Mission digital topographic data, and resample d to 30 m using a bilinear interpolation approach to limit steppiness in the dataset. Because road networks change over time, di fferent distance to roads maps were generated for each time-step. In both Sisaket and Ordar Mean Chey, most householders travel and transport their produce by motorcycle. Pave d roads and unpaved tracks appeared to be equally accessible for most people, and were ther efore treated as equal for this analysis. A substantial number of tracks for Ordar Mean Chey were recorded using a GPS receiver during fieldwork in 2005 and 2006. These were then overla id over the satellite images for each year, and additional tracks were added or removed accord ing to what was visible in each time-step, to create roads layers for each time-step. In Sisaket, highly detailed roads layers for 2004 were obtained from the Department of Highways and De partment of Rural Roads. These were then adjusted according to the satellite imagery. Each time-steps roads layers were then buffered to create the distance to roads maps. Towns with markets were also identified dur ing fieldwork. Although informal trade may also occur, for this study market refers to permanent, managed infras tructure established for formal trade in locally grown produce. In Si saket, a list was obtai ned from the provincial administration, while in Ordar Mean Chey, this information was obtained by visiting all three
117 large villages. Each market was visited, so that geographic coordinates and dates of establishment could be verified. For each timestep, point files were made containing the markets that had been establishe d by that time-step, and buffered to create distance to market maps. Although straight-line di stance does have some effect, most access to markets is along roads. Since both roads and markets showed ch ange over time, we combined roads and markets to create cost-weight distance to market maps by assigning 1 to roads a nd tracks, and 10 to all other surfaces as a simple order of magnitude weighting to constraining the distance to markets buffers. Scale The 30 m pixel resolution of the classifi cations is small enough to observe both the patterns in LULC diversity and the factors influe ncing LULCC, even at localized spatial extents (Chapter 3). At this resolution and with th e smallest possible window -size (3x3 pixels) as described below variation at the micro-scale can be measured. In addition, this smallest window size corresponds roughly to 1 ha the size of the finest scal e of analysis (Table 4-2). At 100 ha (corresponding to the community leve l of interest) enough pi xels are present to capture all 14 possible LULC types and to reveal detailed spatial patterns of diversity. Because many social-ecological processes are scale-dependent (Allen a nd Starr 1982), we considered three scales of analysis, with the meso-scale of interest corres ponding to the sphere of influence of the community (Table 4-2). For the purposes of this study, LULC diversity is assessed on the basis of variety the number of type s of LULC present at each scale. Table 4-2. Scales of Analysis of LULC Diversity Micro-scale Contributing level Meso-scale Level of interest Macro-scale Constraining level SES level Household / farm Community District / province Spatial extent ~ 1 Ha ~ 100 Ha ~ 100 Km2 Analysis size 3x3 pixel window 33x33 pixel window 303x303 pixel window
118 Analysis Data exploration showed that all of the va riables had strongly non-normal distributions, and for this reason Spearmans rank correlations was used for initial assessment of the relationships. Because the datasets were very la rge (over 7 million cases), the statistical power is so strong that all relationships ar e significant with P-values less than 0.0001. In order to assess significance without usin g the P-values, we compared th e correlation and F-test values themselves to the results of th e same tests run on 20 different ra ndomizations of the datasets. Since the randomizations had the same distribution as the actual tests, this process showed that the actual test values were not due to chance (Cumming 2004, Ef ron and Tibshirani 1993). The Spearmans rank correlation is a good firs t step in determining whether LULC diversity responds to the same factors that affect individu al LULC types, however, the relationship that the test measur es is linear. The relationships predicted by the hypotheses are non-linear, and the correlation coefficients do no t capture how the relati onship varies with elevation or distance to roads and markets. Th e shape of the hypothesize d curves (Figure 4-2) suggests instead a multiple-order polynomial rela tionship, which can be expressed in a general form for all three factors as: dx ax) cx (bx2 (1) where: y = predicted LULC diversity x = the explanatory factor of interest (distan ce to roads, distance to market, or elevation) and where the coefficients ha ve the following effects: a determines the initial values of the LULC diversity response,
119 b influences the range of values in th e first phase of the response curve, c influences the range of values in the second phase of the response curve, and d controls the functions sigmoidality by determining the angle of the tail. This function was used to create maps of predicted micro-scale LULC diversity as a response to each of the three local -level explanatory factors. Thes e predicted surfaces were then compared to the actual spatial distribu tion of micro-scale LULC diversity. Random samples of 500 pixels per province (Figure 4-5) were used to generate coefficients for each explanatory factor in each province a nd at each time-step. The means, variances and distribution shapes for the samples were compared to those for the entire province datasets to verify that they were representative. The coe fficients were identified by running the samples in LAB Fit, a curve-fitting analyt ical package (Silva and Silva 1999-2007) to establish how the Figure 4-5. Distribution of 1000 random sample points (500 per pr ovince) shown relative to the 2005 classification and micr o-scale LULC diversity.
120 actual LULC diversity values in the random sample s responded to each explanatory factor within this function. The resultant coefficients were th en applied to the maps of the explanatory factors using the function described above, to create ma ps predicting where LULC diversity would be located if that factor alone w ould influence its distribution. Th e predictive maps were then subtracted from the actual LULC diversity maps to determine: a) how much of LULC diversity was explained by that factor, b) in which locations the predic tions were most accurate, and c) whether or not the hypothe sized shape of the relationshi p was generally correct. Results Spatial Distribution of LULC Div ersity and Explanatory Variables When compared to the classifications in Figur e 4-4, the maps in Figure 4-6 show that the distribution of LULC diversity imitates the general patterns evident in the changing landscapes of the two provinces. In Sisake t, areas of high LULC variety fo llow the rivers, and depict the changing agriculture, along the escarpment foot-h ills, and in the three hi lly areas in the southeast. The provincial capital, Sisa ket town, appears as a circle of high LULC variety in the northcentre of the map, just west of the large river junction. In Ordar Mean Chey, the dense forests on the s outh-eastern hills stand out as areas of low LULC variety. The village of Anlong Veng is visible from 1994 onwards as a node of high diversity, with decreasing magnitude s radiating out from the settlement. The development of the road into the dramatic wedge of deforestation is clearly evident as first a line, and then a triangle of higher LULC variety cutting into the low LULC variety of the forested areas.
121 Figure 4-6. LULC variety at the meso-scal e for 1989, 1994, 2000 and 2005. Distinct patterns can be followed across the landscape (see Chapter 3). The circle on the 2005 map shows the location of the wedge of defore station in Ordar Mean Chey, while the freeform shows the low LULC diversity of the dense forests, and the rectangles highlight Sisakets escarpment foothills and three hilly areas. A comparison of LULC variety at the three different scales shows that the responses of LULC diversity to underlying biophysical and socioeconomic mechanisms present a smoother response at the meso-scale, allowing the dominant f eatures to stand out (Fi gure 4-7). In addition, the influence of national context is clearly evid ent in the differences in LULC variety between the two provinces at the macr o-scale, with the stronger global economic links in Thailand supporting the existence of a gr eater range of LULC types.
122 Figure 4-7. Spatial distribution of LULC variety in 2005, shown for the three different scales of analysis and in comparison to the initial classification. Each inset, centered on the same riparian area, shows a spatial extent representing roughly ten times the moving window size used for each scale, to show that the apparent differences are not simply a result of spatial zoom. At the microscale, patterns appear more random, while those at the macro-scale appear more uniform. Infrastructural conditions are markedly different in the two SESs. For instance, the road network in Sisaket makes the entire landscape highly connected (Figure 4-8). In 1989, the maximum distance to any road was 9.2 km, with a median distance of 200 m. With little room for further road expansion, this median dist ance remained unchanged through to 2005. In Ordar Mean Chey, not only the number of roads, but al so the location of some tracks changed, as is common of dirt tracks in unde veloped areas with low input from government. In 1989, the
123 median distance to road was 1.2 km, with the gr eatest distance being 18.2 km. By 2005 the median and maximum distances had roughly ha lved to 0.8 km and 9.7 km respectively. Figure 4-8. Distance to roads in 1989, 1994, 2000 and 2005 In Sisaket, based on date of establishment, the number of markets se lling farmers produce increased from four in 1989 to fourteen by 2005. Since the road network was well-established throughout the study period, it was the markets them selves that most changed the cost-weight distance to market across time (F igure 4-9). In Ordar Mean Chey, even though no new markets had been established by 1994, access was improved through the development of new tracks. With the establishment of a market in An long Veng in 1999, and another in Trapeang Prasat in 2004, the cost-weight di stance to market was greatly redu ced in the eastern part of the province. However, to some extent the eastern and western halves of the province each retain
124 stronger links to the rest of the country to the so uth than they do to each other. The roads from Samraong to Siem Reap (a major town south of the study area), and Anlong Veng to Siem Reap, are better established than that linking Samraong to Anlong Veng. Figure 4-9. Cost-weight distan ce to market in 1989, 1994, 2000 and 2005 Although Sisaket sits on a plat eau above Ordar Mean Chey, the difference between their mean elevations is only about 90 m (Figure 4-10). Ordar Mean Chey has a median elevation of 70 m, while that of Sisaket is 141.5 m. There is so me overlap in the spatial distributions of all of the explanatory factors discussed here. In Si saket, roads and markets are located only on the lower, flatter areas, and not in the higher, mountainous areas. In Ordar Mean Chey, there has been more change in the relationships. At the start of the study period, settlement and related
125 infrastructure were located in the lower western pa rt of the province, with hills (other than the steep escarpment) found in the east. However, over time roads and settlements have expanded into this slightly more elevated area. Figure 4-10. Elevation in the study area shown together with meso-scal e LULC diversity in 2005. Distance to Roads and LULC Diversity A comparison of the Spearmans correlation coe fficients for LULC diversity and distance to road shows that, overall, Ordar Mean Chey ha s a much stronger response to distance to roads at all scales, reflecting the difference in road de nsities between the two provinces (Table 4-3 and Figure 4-8). The negative signs indi cate that increasing distance to roads is associated with a decreasing magnitude of LULC di versity at all scales. In bot h provinces there is a general increase in the strength of the relationship over time at all scales although this increase is not Table 4-3. Spearmans correlation coefficients showing LULC dive rsity in response to distance to roads at each time-step Sisaket Ordar Mean Chey 1989 1994 2000 2005 1989 1994 2000 2005 Micro-scale -0.118 -0.158 -0.191 -0 .194 -0.275 -0.203 -0.237 -0.271 Meso-scale -0.149 -0.238 -0.289 -0. 162 -0.263 -0.318 -0.356 -0.400 Macro-scale -0.098 -0.047 -0.022 -0 .167 -0.182 -0.122 -0.392 -0.277
126 always constant. In both provin ces, the correlations are strongest at the meso-scale, suggesting that it is at this scale that the dominant patterns emerge on the landscape (Figure 4-7). However, at this broader scale interactio ns with other mechanisms are stronger, making it more difficult to model the specific nature of th e relationship between LULC dive rsity and distance to roads. The extent to which the actual relationships match the hypothesized shape in Figure 4-2a can be seen in Figure 4-11, which shows the pred icted response of micr o-scale LULC diversity to distance to roads for each time-step for the random sample of 500 points. In Sisaket, only 1989 shows the predicted initial increase, before LULC diversity magnitudes drop off with increasing distance. From 1994 onwards, conditions have changed substantially, suggesting that the actual relationship differs fr om that hypothesized, with a more linear decrease in diversity with increasing distance to roads. In Ordar Mean Chey a similar shift is observed, but only from 2000 onwards. This delay relative to Sisaket show s the extent to which Ordar Mean Chey lags behind that province in terms of infrastructural development. That both provinces experience this same change toward a more linear relati onship suggests that over time, the system is Sisaket0 1 2 3 4 012345 Distance to roads (km)Predicted LULC diversity 1989 1994 2000 2005 a) Ordar Mean Chey0 1 2 3 4 03691215 Distance to roads (km)Predicted LULC diversity 1989 1994 2000 2005 b) Figure 4-11. Predicted LULC diversity at the micr o-scale in response to di stance to roads in the function dx ax) cx (bx2for four different points in time in a) Sisaket, and b) Ordar Mean Chey, based on the random sample with n=500, and using the coefficients generated from modeling the relationship with actual LULC diversity values. Coefficients for the models are given in Table 4-4.
127 Table 4-4. Coefficients used to predict LULC di versity at four time-step s in Sisaket and Ordar Mean Chey in response to dist ance to roads in the function dx ax) cx (bx2. Sisaket a b c d 1989 2.717 -3.828E-01 6.080E-03 -2.879E-01 1994 3.291 3.356E-02 2.364E-03 -7.883E-01 2000 2.913 1.053E-01 -8.528E-03 -6.072E-01 2005 3.097 5.607E-02 1.319E-04 -6.996E-01 Ordar Mean Chey a b c d 1989 3.217 -8.988E-02 3.291E-03 1.256E-01 1994 3.329 4.197E-02 -2.140E-03 -3.231E-01 2000 3.099 2.335E-02 -7.881E-04 -2.692E-01 2005 3.385 -7.609E-02 5.548E-03 -3.469E-02 Table 4-5. Results of the models predicting LU LC diversity in response to distance to roads showing goodness-of-fit, proportion of lands cape variance explained by the model, mean difference and standard deviation betw een predicted and actual LULC diversity for the entire landscapes, and proporti on of landscape where predicted LULC diversity is equal to or less than 1 ty pe different from actu al LULC diversity. r 2 Adj. R2 Mean difference between actual and predicted LULC diversity Standard deviation of difference between actual and predicted LULC diversity % of landscape with difference less than 1 between actual and predicted LULC diversity Sisaket 1989 1.370 0.022 0.04 1.44 51.51 1994 1.692 0.062 0.15 1.42 48.46 2000 1.289 0.035 0.27 1.35 49.67 2005 1.660 0.034 0.21 1.38 56.62 Ordar Mean Chey 1989 1.637 0.098 -0.09 1.30 48.92 1994 1.853 0.022 -0.04 1.33 50.22 2000 1.528 0.041 -0.03 1.19 55.81 2005 1.943 0.070 0.05 1.41 49.13 adapting as decreasing distance to roads allows households to cha nge their livelihood strategies. Although the good fits for all of th e province/years show that LU LC diversity does respond to distance to roads, as expected this factor alone does not explain much of the variance in microscale LULC diversity when the f unction is applied to the entire landscapes (Table 4-5 and Figure 4-12) with the adjusted R-squared values for Sisa ket being particularly low. This implies that while LULC diversity does res pond to distance to roads, in th e province, roads are no longer a major determinant in the distribution of LULC di versity. In both provinces however, the models
128 predict LULC diversity to within one type acro ss about 50% of the landscape (Table 4-5). In Sisaket the models marginally under-predicted the overall LULC diversity, as is shown by the mean differences between actua l and predicted, whereas in Ordar Mean Chey excepting 2005 the models slightly over-predicted LULC diversity (Table 4-5). The overand underpredicted ar eas on the maps provide important information on what the other factors contributing to th e distribution of LULC diversity ar e. For example, it is clear Figure 4-12. Residual map showing the differenc e between actual LULC diversity and that predicted by distance to roads in the function dx ax) cx (bx2. Orange to red areas show where the model over-predicts LULC di versity, and pale to dark green areas show where the model under-predicts LULC diversity. Insets show arbitrary locations of roughly 4 km x 4 km to give an indication of the distribution of the residuals when displayed at a finer map-scale.
129 that at all time-steps in Sisaket, rivers are areas of high LULC divers ity independent of the influence of distance to roads (Figure 4-12). The inset for Sisaket in 1989 shows that while diversity is correctly predicted immediately next to the road, from about 100 ms distance it is over-predicted for the next 250 m as is shown by th e distinctive bands paralleling the roads. The residual maps for 2000 and 2005 suggest that the more linear relationship depicted for these years in Figure 4-11 is a better description of this provinces relation to diversity than that hypothesized in Figure 4-2a. Give n the density of the road networ k in Sisaket (Figure 4-8), the fact that distance to roads does not play a major role in explaining the total variation in LULC diversity across the provinces landscape is not surprising. For Ordar Mean Chey, the most striking featur e of the residual maps is how the distanceto-roads models over-predict for more elevated areas, adding weight to the hypothesis that elevation is also an important f actor influencing LULC diversity. Over time, however, as roads penetrate into the hillier east of the province, the spatial exte nt of this over-prediction decreases (Figure 4-12). Rivers do not appear to have the same influence on LU LC diversity in Ordar Mean Chey as they do in Sisaket. Distance to Markets and LULC Diversity The Spearmans rank correlation coefficients show that cost-weight distance to market is an important determination of the overall magnit ude of LULC diversity (Table 4-6). This is particularly true in Ordar Mean Chey. As with distance to ro ads, the general trend is for a decrease in magnitude of LULC diversity with increasing cost-weight distance to market. The anomalous positive correlations for Sisaket in 1989 can be attributed to the lack of a market to the south-east of the province until 1994 (Figure 4-9) in spite of the presen ce of the 3 hilly areas of fertile soil and high LULC diversity highlighted in Figure 4-4. Unlike distance to roads, there is no single scale at which the co rrelations are strongest instead th is varies from year to year.
130 In Ordar Mean Chey, there is a considerable decrease in correlation strength in 2000 at the microand meso-scales, but a strong increase at th e macro-scale. This relates to the addition of the provinces second market, in Anlong Veng (Fi gure 4-9), which reduced the influence of the Samraong market at the microand meso-scales, and created more evenly distributed access for the entire landscape. Table 4-6. Spearmans correlation coefficients showing LULC diversity in response to costweight distance to markets at each time-step Sisaket Ordar Mean Chey 1989 1994 2000 2005 1989 1994 2000 2005 Micro-scale 0.028 -0.067 -0.262 -0. 086 -0.426 -0.422 -0.262 -0.264 Meso-scale 0.026 -0.147 -0.126 -0. 130 -0.361 -0.598 -0.290 -0.386 Macro-scale -0.068 -0.083 -0.271 -0 .144 -0.120 -0.098 -0.394 -0.267 In Sisaket there has been great er variability over time in cost-weight distance to market compared to distance to roads, and this is refl ected in the changing shapes of the models using this factor to predict LULC diversity (Figure 4-13). In all years there is an initial increase before magnitudes drop off with increasing distance, yet not to the degree suggested by the hypothesized curve (Figure 4-2b). The change in Ordar Mean Chey has been more dramatic. Sisake t 0 1 2 3 4 5 050100 Cost-weight distance to market (km)Predicted LULC diversity 1989 1994 2000 2005 a) Ordar Mean Chey0 1 2 3 4 5 075150225300 Cost-weight distance to market (km)Predicted LULC diversity 1989 1994 2000 2005 b) Figure 4-13. Predicted LULC diversity at the micr o-scale in response to cost-weight distance to market in the function dx ax) cx (bx2for four different points in time in a) Sisaket, and b) Ordar Mean Chey, based on the rando m sample with n=500, and using the coefficients generated from modeling the re lationship with actual LULC diversity values. Coefficients for the models are given in Table 4-7.
131 Table 4-7. Coefficients used to predict LULC di versity at four time-step s in Sisaket and Ordar Mean Chey in response to cost-weigh t distance to market in the function dx ax) cx (bx2. Sisaket a b c d 1989 2.377 6.037E-03 -5.834E-05 -3.114E-02 1994 2.164 8.572E-03 -1.784E-04 8.439E-03 2000 2.307 3.280E-03 -8.693E-05 9.842E-03 2005 2.511 6.005E-03 -2.485E-04 1.714E-02 Ordar Mean Chey a b c d 1989 3.695 -3.440E-04 -4.613E-06 4.189E-03 1994 3.173 6.445E-04 -1.655E-05 8.457E-03 2000 2.643 7.095E-04 -3.604E-05 1.428E-02 2005 4.282 1.627E-03 -4.586E-06 -5.288E-02 Table 4-8. Results of the models predicting LULC diversity in response to cost-weight distance to market showing goodness-of-fit, proporti on of landscape variance explained by the model, mean difference and standard devi ation between predicted and actual LULC diversity for the entire landscapes, and proportion of landscape where predicted LULC diversity is equal to or less than 1 type different from actual LULC diversity. r 2 Adj. R2 Mean difference between actual and predicted LULC diversity Standard deviation of difference between actual and predicted LULC diversity % of landscape with difference less than 1 between actual and predicted LULC diversity Sisaket 1989 1.335 0.044 0.18 1.34 63.47 1994 1.647 0.087 0.06 1.33 50.32 2000 1.289 0.027 0.19 1.27 63.97 2005 1.647 0.042 0.07 1.26 56.30 Ordar Mean Chey 1989 1.410 0.222 -0.09 1.21 56.03 1994 1.584 0.162 -0.05 1.21 55.88 2000 1.522 0.044 -0.04 1.16 56.88 2005 1.912 0.085 0.01 1.37 49.21 Since there was only one market in the provinc e in 1989 and 1994, the addition of a yet another by 2000 almost halved the effective distance to market for the eastern part of the province (Figure 4-9 and Figure 4-13b). By 2005 both the incr ease in road density and the addition of a third market have seen the response of LULC di versity change to being almost linear. These shifts could be qualitatively inte rpreted as representing adaptations in the SES state, particularly at the local level.
132 As with those based on distan ce to roads, the models pred icting LULC diversity based on cost-weight distance to market generally under-predicted LULC diversity in the Sisaket landscape and over-predicted for Ordar Mean Chey (Table 4-8). For the most part, cost-weight distance to market better predicts diversity th an distance to roads al one, producing models for which 50 to 64% percent of the landscape is predic ted to within one LULC type (Table 4-8 and Figure 4-14). The adjusted R-squared values show that in Ordar Mean Chey in 1989 and 1994, market proximity, as a single variable, explaine d much of the variati on in LULC diversity Figure 4-14. Residual map showing the differenc e between actual LULC diversity and that predicted by cost-weight distan ce to market in the functiondx ax) cx (bx2. Orange to red areas show where the model over-predicts LULC diversity, and pale to dark green areas show where the model under-predicts LULC diversity. Inse ts show arbitrary locations of roughly 4 km x 4 km to give an indication of the distribution of the residuals when displayed at a finer map-scale.
133 magnitudes across that provinces landscape (Table 4-8), and the spatial extent of the overprediction associated with the hilly south-east is reduced, relative to the models for distance to roads (Figure 4-14). These market models have tended to over-predict LULC diversity in the west of the province for all but 2000 (Figure 4-14 ). In Sisaket, the low adjusted R-squared values again suggest that this province is so hi ghly connected in infrastr uctural terms that other factors now play a larger role in explaining the spatial distri bution of LULC diversity. For example, rivers are again visible as areas for wh ich LULC diversity is under-predicted, as is the area around Sisaket town from 1994 onwards. The 1994 Sisaket map shows large areas of overpredicted diversity, and this model had the lowest total predicted area within 1 type of the actual LULC diversity (Figure 4-14 and Table 4-8). Elevation and LULC Diversity Spearmans rank correlation coefficients show that for the whole landscapes, there is a general decrease in the number of LULC types with increasing elevatio n at all scales (Table 4-9). Again, the relationships are str onger in the less-devel oped province of Ordar Mean Chey than they are in Sisaket. There is so me variability in the scale at whic h the relationship is strongest at the different time-steps, however, in Sisaket for the most part it appears weakest at the microscale, possibly because this linear measurement ma sks the fact that, at th is fine scale, higher LULC diversity is associated with intermediate elevation with lowest areas dominated by rice, and highest areas dominated by fore st (Figure 4-4). In contrast, strong relationships are found at Table 4-9. Spearmans correlation coefficients showing LULC dive rsity in response to elevation at each time-step Sisaket Ordar Mean Chey 1989 1994 2000 2005 1989 1994 2000 2005 Micro-scale -0.023 -0.028 -0.033 -0 .037 -0.422 -0.427 -0.335 -0.378 Meso-scale -0.088 -0.243 -0.260 -0. 127 -0.196 -0.473 -0.358 -0.329 Macro-scale -0.165 -0.090 -0.306 -0 .194 -0.067 -0.035 -0.321 -0.373
134 the micro-scale in Ordar Mean Chey at all time-s teps (Table 4-9), possibly because much of the human-induced change in this prov ince is still taking place in the lower areas to the west of the province (Figure 4-4). Because elevation does not change over time we expect less vari ation in the relation between this factor and microscale LULC diversity compared to the relation to roads and markets, and the graphs in Figure 415 show that this is indeed th e case. For all time-steps, the shapes of the Sisaket LULC diversity models based on elevation resemble the hypothesized shape (Figure 4-2c) more closely than those of the other factors. The model shapes for Ordar Mean Chey present more abrupt changes with increasing elevation, a nd also show a second increase in LULC diversity afte r the initial strong decrease, fr om about 200 m elevation onwards. This is probably because the eastern half of th e province is higher than the western half while still being relatively flat, so that multiple LULC types can emerge (Figure 4-10). Overall LULC diversity in the eastern half is generally lower because this area has had limited settlement because of the areas political history. Sisaket0 1 2 3 4 100200300400500600 Elevation (m)Predicted LULC diversity 1989 1994 2000 2005 a) Ordar Mean Chey0 1 2 3 4 0100200300400 Elevation (m)Predicted LULC diversity 1989 1994 2000 2005 b) Figure 4-15. Predicted LULC diversity at the micr o-scale in response to el evation in the function dx ax) cx (bx2for four different points in time in a) Sisaket, and b) Ordar Mean Chey, based on the random sample with n=500, and using the coefficients generated from modeling the relationship with actual LULC diversity values. Coefficients for the models are given in Table 4-10.
135 Table 4-10. Coefficients used to predict LULC di versity in response to el evation in the function dx ax) cx (bx2. Sisaket a b c d 1989 1.112 1.626E-03 -3.947E-06 2.058E-03 1994 1.155 1.926E-03 -4.704E-06 1.588E-03 2000 1.116 1.647E-03 -3.961E-06 1.875E-03 2005 0.798 2.526E-03 -5.885E-06 1.891E-03 Ordar Mean Chey a b c d 1989 2.407 3.782E-03 -4.440E-05 5.932E-03 1994 3.028 1.408E-03 -2.426E-05 5.935E-03 2000 1.631 5.275E-03 -5.103E-05 6.475E-03 2005 2.377 4.163E-03 -4.281E-05 5.008E-03 Table 4-11. Results of the models predicting LU LC diversity in response to elevation showing goodness-of-fit, proportion of landscape varian ce explained by the model, mean difference and standard deviation between predicted and actual LULC diversity for the entire landscapes, and pr oportion of landscape where predicted LULC diversity is equal to or less than 1 type diffe rent from actual LULC diversity r 2 Adj. R2 Mean difference between actual and predicted LULC diversity Standard deviation of difference between actual and predicted LULC diversity % of landscape with difference less than 1 between actual and predicted LULC diversity Sisaket 1989 1.322 0.056 0.11 1.26 66.50 1994 1.632 0.096 0.05 1.29 56.11 2000 1.239 0.066 0.17 1.27 64.28 2005 1.515 0.116 0.06 1.25 59.66 Ordar Mean Chey 1989 1.384 0.226 -0.12 1.18 59.63 1994 1.600 0.141 0.70 1.24 58.68 2000 1.422 0.112 -0.08 1.11 61.10 2005 1.615 0.228 -0.02 1.26 55.61 Not only does elevation most closely match th e hypothesized response of LULC diversity, but it also predicts most accurately the spatial distribution of distribut ion of LULC diversity, with between 55 and 66% of the la ndscape being predicted to with in one LULC type in all timesteps and both provinces (Table 4-11 and Figur e 4-16). In Ordar Mean Chey, elevation as a single factor accounts for a lot of the overall variance in the landscape, as is shown by the adjusted R-squared values. As with distance to roads and cost-w eight distance to market, most of the under-predicted areas in Sisaket relate to riparian features, again highlighting their
136 importance in determining the distribution of LULC diversity in this province (Figure 4-16). In Ordar Mean Chey in 1994 extensive areas in the north-western part of the province, which are slightly elevated, are under-predict ed, and this is reflected in the mean difference between actual and predicted LULC diversity (Table 4-11), wh ich shows that this ye ars model performed poorly relative to the other years. In contra st, the under-predictions in 2005 relate to the dramatic wedge of deforestation described in Fi gure 4-4. The residual maps of the elevationbased models for Ordar Mean Chey (Figure 4-16) s how that the hilly south-eastern area is wellFigure 4-16. Residual map showing the differenc e between actual LULC diversity and that predicted by elevati on in the functiondx ax) cx (bx2. Orange to red areas show where the model over-predicts LULC diversity, and pa le to dark green areas show where the model under-predicts LULC diversity. Inse ts show arbitrary locations of roughly 4 km x 4 km to give an indication of the dist ribution of the residuals when displayed at a finer map-scale.
137 predicted, with errors to the north and west. This is in contrast to the roadsand market-based models, which performed better in the north and south, and over-p redicted for the hilly southeast (Figure 4-12 and Figure 4-14). This sugge sts that a model that combined cost-weight distance to market (which includes the influence of roads) and elevation might explain much of the variance in Ordar Mean Chey. Discussion Overall, magnitudes of LULC diversity are high er in Sisaket than in Ordar Mean Chey. Likewise, this province clearly ha s a greater degree of infrastructu ral development. This means that not only do the SESs have different potential responses, but also differences in the factors affecting the responses. Variati ons at different time-steps in th e relationships with the distance to roads and market variables are to some extent due to changes in these explanatory variables. This does not diminish the importance of the changing LULC diversity values, but instead highlights the importance of other system attributes such as the connectivity of the SESs. For example, the weaker relationships in Sisaket between LULC diversity and roads and markets suggest a higher degree of connec tivity and rigidity to that SE S, where human influence has reached a point where it limits the system ability to respond to these factors (Carpenter, Ludwig, and Brock 1999; Holling and Gunderson 2002). Th is suggests that the strength of the relationship between LULC divers ity and various underlying mechan isms could be explored as an indicator of the sustainability of the SES. Roads clearly play a key role in determin ing the distribution of micro-scale LULC diversity however, not in the form hypothe sized. Although the qual itative shape of the hypothesis was not supported by the models, there is nevertheless an overall decline in LULC diversity with increasing distance to roads, a nd LULC diversity is shown to be affected by accessibility in the same way as change in individu al LULC types. In Sisaket, the road network
138 is so dense that in this small-hol der agrarian system, one is never more than three or four fields away from a road. With the exception of th e border escarpment, the low LULC diversity associated with greatest distance from road is re lated to the dominance of the main subsistence crop rice, and not undisturbed na tural vegetation. Given that ne arly all land is within ready access by road, it appears that close to roads, more economically attractive land uses that tend to cover smaller areas, such as residences, cash cr ops and plantations, combin ed with the diversity associated with road verges (scrub and ephemeral surface water) mean that in Sisaket LULC diversity is highest close to th e road. In Ordar Mean Chey, ev en though people may mainly be planting the dominant subsistence crop close to the road, the linea r settlement pattern means that here too, diversity is high close to the roa d, as people tend to have cash and tree crops around their homes, while far from the road, as expected LULC types tend to be restricted to the dominant natural vegetation. This means that the relationship between distance to roads and LULC diversity may be better e xpressed as a linear de cline with increasi ng distance. Although the specific nature of the hypothe sized relationship was not borne out, nevertheless the data show that LULC diversity is affected by the spatial distribution of roads. While the hypothesized graph for changes in LU LC diversity with cost-weight distance to market suggested large and abrupt change with initial in creases in distance, the data show that this trend is more gradual. Ne vertheless, the models still show the initial increase in LULC diversity with increasing distance, followed by a decrease, described by the hypothesized leftskewed, inverted U distribution. In Sisaket, the uneven spatial distribution of markets in the earlier years influenced the nature of their relati onship to LULC diversity, drawing attention to the issue of feedbacks. As markets became more evenly distributed across the province, the
139 relation to LULC diversity took on a shape closer to that hypoth esized, raising the question of how much markets affect LULC diversity a nd how much they are affected by it. While road densities in Sisaket change littl e over the study period so that access to market was influenced more by changes in the distributi on in markets, in Ordar Mean Chey, changes in the roads and tracks played a large role in ch anging access to markets. The shape of the 1989 model reflects how many roads linked to a single market, and with an additional market in 2000, the model more closely resembles the hypothesized shape. By 2005, however, the addition of a third market in the frontier area of the eastern pa rt of the province where abrupt change is taking place, means that the relationship in the last ti me-step was almost linear, with high diversity close to the newly established markets, declinin g sharply as human influence gave way to the densely forested areas. Encompassed in the effect of elevation on LULC diversity are related factors such as slope and soil. This is most evident in Sisaket, where the three hilly areas of intermediate elevation in the south-east have high LULC di versity primarily because of the richer soils there, whereas the much higher escarpment mountains are very steeply sloped and rocky, limiting human access and use. The combination of these factors cont ributes to the way in which the models conform to the hypothesized shape. In Ordar Mean Chey politics have limited the ability of humans to expand their influence into the gently sloped hills of the eastern part of the province, so that at all time-steps in the study period there is an abrupt increase and then decrease in the number of different LULC types within a relatively small elev ation range. In the futu re, as the east of the province opens up, we might expect this transition to be more gradual. The areas of over-prediction in the residual maps for cost-weight distance to market (which incorporates the effect of roads) and elevation clearly show that complementary information is
140 contained in the prediction models based on each of these variables. It would therefore be useful to identify a model that combined cost-weight dist ance to market (which incorporates the effect of roads) with elevation, as toge ther they would likely explain mu ch of the varian ce in the spatial distribution of LULC diversity. While the models explored here describe the na ture of the responses for LULC diversity at the micro-scale only, it is clear fr om the Spearmans rank correlati ons that the effect of these local-level drivers can nevertheless also be felt at broader scales. However, at each higher scale of analysis, a larger area is incorporated in the assessment of diversity, which increases the influences and interactions of other processes, making it hard to identify how each variable individually affects LULC divers ity at these broader scales. To understand better the distribution of LULC diversity at the mesoand macro-scal es, however, it is more important to identify factors that contribute to patterns at these broader scales (Levin 1992). For example, the maps suggest that rivers may be a major underl ying mechanism controlling LULC diversity distribution at the meso-scale. The more fertile soils and access to water in riparian areas not only create a greater range of natural LULC t ypes (Naiman and Decamps 1997), but also provide the conditions for a range of LULC types more than individual households could account for, but which collectively a community or village w ould likely generate. Li kewise, a broad-scale mechanism such as degree of in tegration in the global economy might influence the range of human-dominated LULC types and so bette r explain macro-scale LULC diversity. This study shows that at the micro-scale th e hypothesized relationships between LULC diversity and these factors were supported, with some modificati on to the exact nature of the relationship in the case of dist ance to roads. Even in the dissimilar systems that the study provinces represented, LULC diversity had the same general responses to factors known to
141 influence LULCC. That is, increasing accessibi lity and infrastructural development led to increasing LULC diversity in much the same wa y that they lead to forest conversion and agricultural expansion; and highe r elevations constrained LULC diversity just as they do the conversion of one specific type of LULC to se veral others (Chapter 2; Cropper, Puri, and Griffiths 2001; Brgi, Hersperger, and Schneeberge r 2004; Caldas et al. 2007; Chomitz and Gray 1996; Crews-Meyer 2004; Erenstein, Oswald, a nd Mahaman 2006; Etter et al. 2006; Messina et al. 2006). The way in which LULC diversity relates to these mechanisms provides important information on the state of the SES. Variations in the responses of LULC diversity show that the SESs are adaptive, responding to the influen ce of the factors tested here. Weakening relationships suggest that a system is becoming more connected. The processes leading to the emergence of patterns on the landscape result from household-level decisions to national policies. Just as these pro cesses are scale-dependent (Levin 1992), so too are the response patterns (Chapter 3). While the individual choi ces of farmers might appear as random when examined at the level at which they occur, thei r aggregate effect at the community level emerges as a pattern at a broader level. At the same time, the context provi des a set of conditions that will impose limitations on the potential range of f actors (Ahl and Allen 1996). For example, the degree to which a country is integrated into the global economy will determine the potential range of LULC types the provincial SES can include. Conclusion While LULC diversity provides a simple but useful generalization for cross-site comparison, predicting its response to different factors requires furthe r development of the theory and the identification of factors more likel y to influence spatial and temporal variation at the mesoand macro-scales. The use of CASs to frame the analysis of landscape change
142 provides a means to integrate SESs research with LULCC research. LULC diversity links the metaphors that guide SES research to the more pr actical reality of how humans, as the dominant species, structure of the landscape. The effect of drivers of change on different landscapes can be compared by measuring the response of LULC di versity instead of the responses of individual LULC types which might otherwis e be context-specific (Perz 2007) The condition of SESs can be expressed quantitatively by using LULC types as an expression of SES components. Linking LULCC and SES research allows a systems location in space and time, and at multiple scales, to be included in the eval uation of its past, present and future states.
143 CHAPTER 5 CONCLUSION The broad goal of this dissertation has been to address the question of whether CASs theory and the concept of LULC diversity can help link SESs research and LULCC research, and in so doing provide the former with a way of meas uring system change and the latter with generalizations for cross-site comparisons a nd deductive analysis. To this end, in this dissertation I set out to test whet her it was possible to detect va riation in LULC diversity across landscapes, whether this va riation displayed some structure or organization in its distribution at different scales, and whether it wa s possible to attribute the dist ribution of LULC diversity to underlying biophysical and socio-economic mech anisms. These research problems were answered in the three separate resear ch papers comprisi ng this dissertation. The first paper in this dissertation shows that linking LULCC and SESs research under the same theoretical framework is possible. LULCC, as expressed across landscapes, can be seen as a tangible expression of SES processes. The CASs c oncepts such as diversity that are explicit in the study of SESs have been implicit in many LU LCC policy statements and case studies. The novel application of the CASs approach build s on known relationships between change in individual LULC types and various factors influencing that change. These interactions are used to predict the distribu tion patterns and responses of LULC diversity to these underlying system mechanisms. The paper explains how a spatially explicit approach using landscapes provides a way for quantifying components of SESs. The second paper proves that LULC diversity does have spatial and distributional structure. LULC diversity patterns form distinct patterns on the landscape, which suggest the influence of rivers, mountains, roads and settlements. LULC diversity maintains similar frequency distributions across time, wh ile the shape of the distri butions is scale-dependent.
144 The third and final paper demonstrates that even in dissimilar SES landscapes, LULC diversity has the same general responses to unde rlying mechanisms. These responses vary with spatial scale, but there are distinct trends in the direct ion of the relationships such as increasing magnitudes of LUC diversity with decreasing elev ation, distance from roads and distance from markets. At the micro-scale, LULC diversity is shown to respond to el evation and distance to market in ways similar to those hypothesized that is, an initial increase to higher magnitudes followed by decrease to lower magnitudes. Howe ver, the relationship to distance to roads was found to be more linear. Significance of Findings Together, these three papers contribute to the development of two emerging research programs LULCC, and SESs. Firstly, the novel manner in which CASs theory and the concept of diversity are applied in this dissertation provides LULCC resear chers with generalizations for comparison, extrapolation and prediction. This appr oach allows us to look at landscapes as LULC diversity can be tested and evaluated in any landscape, independe nt of the specific environmental and social characteristics. CASs theory should prove useful to researchers trying to move LULCC research from inductive toward s deductive analysis (Rindfuss et al. 2004; Walker 2004), because it can be the source of predictive hypotheses th at are neither contextspecific nor discipline-dependent (Perz 2007). Trends in magnitude s of diversity can be inferred by drawing on the CASs ideas of selection, emer gent properties, non-linearity, path dependency and scale, inter alia (Ahl and Allen 1996; Ar thur 1999; Holland 1995; Ka uffman 1995; Lansing 2003; Levin 1998). Secondly, by explicitly treating LULC types as the expression of SES components, the approach developed here provides SESs research ers with the means to measure change in the system (Carpenter et al. 2001; Cu mming et al. 2005). By using LU LC diversity as a generalized
145 unit of analysis in a SES, we can spatially repres ent change in the condition or state of any SES. We can then use change in LULC diversity to assess the likelihood of a SES changing or persisting in terms of its stru cture and function, since any cha nge in the range of potential responses will affect the ability of a system to adapt and persist (Holland 1995; Levin 2003). The landscape approach also allows SESs res earch access to the well-developed methods used by LULCC scientists an d landscape ecologists. LULC Diversity This research verifies the utility of LULC diversity as a generalization for comparing and contrasting different SESs. LULC diversity patterns not only in corporate and mirror the actual LULC types from which they are derived, but they also provide an indicat ion of the degrees of intensification, fragmentation a nd diversification of the SES landscape (Bogaert, Farina, and Ceulemans 2005; Crawford et al. 2005). LULC diversity may have different overall magnitudes in different landscapes, and respond in different ways a nd at different points in time to similar biophysical and socioeconomic fact ors. However, the shapes of LULC frequency distributions are similar enough that these differences can be compared and contrasted, and lead to the conclusion that LULC diversity is a means to ove rcoming context-specificity (Perz 2007; Brgi, Hersperger, and Schneeberger 2004). As a quantifiable concept, LULC diversity o ffers distinct advantages over the standard classification approach to LULCC. Because the measures of dive rsity are discrete (variety) and continuous (relative abundance) as opposed to categorical, th is increases the options for exploring spatial and temporal va riability in a quantitative manne r. LULC diversity provides a way to quantify SESs as expressed on the land scape, and can be used to measure system characteristics such as change and persistence (C arpenter et al. 2001; Cumming et al. 2005). In
146 addition, because the LULCC approach is spatial, a focus on changes (or not) in LULC diversity reveal where system identity is changing or persisting. Complex Adaptive Systems, LULCC, and SESs As a theoretical construct, LULC diversity provides additi onal insight into the humanenvironment interactions that are shaping our world, particularly when considered through the framework of CASs. For example, LULC diversity shows different responses at different scales, reflecting the multi-scalar nature of SES pr ocesses (Ahl and Allen 1996). The spatial arrangement of LULC diversity s uggests that SESs are self-org anizing, with landscape patterns emerging from interactions with a complex a rray of mechanisms (Holland 1995; Levin 2005). Locally, magnitudes of diversity in some cases decrease as LULC type s are winnowed away, and in other cases increase, as new types evolve (Levin 1999). The threshold-type response of LULC diversity change trajectories in places of hi gh temporal variability is evidence that change in LULC diversity is non-linear, another characteristic of a complex, adaptive SES (Folke et al. 2004; Holland 1995; Holling and Gunderson 2002). Further Work The next step in the application of CASs to the study of human-environment interactions would be to test the concept of LULC diversity in other landscapes It would be useful to have sufficient case studies to develop a general dist ribution curve for LULC diversity at different scales, against which individual scenarios can be assessed. Similarly, we need to explore whether the responses of LULC diversity to various underlying mechanisms hold up under other social-ecological conditions, such as in urban areas or savanna rangelands (Cadenasso, Pickett, and Schwartz 2007; Walker and Abel 2002). The range of explanatory factors should be e xplored further to identify the best models explaining the distribution of LULC diversity (f or each scale of analysis) in one type of
147 landscape, and then to test whet her the same model would work in different landscapes. This dissertation suggests that the relationships between LULC diversity and most underlying mechanisms are likely to be non-linear. Furthe r attention is required to try to understand how and why the nature of the relationships varies. Because the underlying processes appear to be scale-dependent, different variables will need to be identified for inclusion in the models for each of the scales of analysis. Additional CAS character istics, such as connectivity, aggr egation and path-dependency are regularly referred to in LULCC st udies. Studies that specifically examine how change in LULC types can be captured by these characteristics sh ould be pursued. The large differences in road density, combined with the differenc es in response of LU LC diversity to distance to roads, in the two study provinces suggests that the role of connectivity in dete rmining LULC diversity should be explored further. Like wise, the effects of aggrega tion require additional study, by experimenting with merged or grouped LULC cate gories. The effect of scale of analysis on scale of process would also benefit from further a ttention. For example, slight adjustments might increase the emergence of patterns, while larg er grain sizes could re duce processing time while still revealing the same information. Although not always framed in the same terminology, many LULCC modelers recognize path-dependency in their systems of interest, and this work could be extended to focus on whether the probab ilities of conversion of individual LULC types can be applied in a similar fashion to predic t changes in magnitudes of LULC diversity.
148 LIST OF REFERENCES Abel, T., and J. R. Stepp. 2003. A new ecosystems ecology for anthropology. Ecology and Society 7 (3):art12. Ahl, V., and T. F. H. Allen. 1996. Hierarchy Theory: a Vision, Vocabulary and Epistemology. New York: Columbia University Press. Alados, C., C. Pueyo, O. Barrantes, J. Escs, L. Giner, and A. Robles. 2004. Variations in landscape patterns and vegetation cove r between 1957 and 1994 in a semiarid Mediterranean ecosystem. Landscape Ecology 19:543-559. Allen, T. F. H., and T. W. Hoekstra. 1992. Toward a Unified Ecology. New York: Columbia University Press. Allen, T. F. H., and T. B. Starr. 1982. Hierarchy: Perspectives for Ecological Complexity. Chicago: University of Chicago Press. Anderies, J.M., B. Walker, and A. Kinzig. 2006 Fifteen weddings and a funeral: case studies and resilience-based management. Ecology and Society, 11(1):art21. Arima, E., R. Walker, S. Perz, and M. Cald as. 2005. Loggers and fore st fragmentation: Behavioral models of road building in the Amazon Basin. Annals of the Association of American Geographers 95 (3):525-541. Arrow, K., B. Bolin, R. Costanza, P. Dasgupta, C. Folke, C. S. Holling, B.-E. Jansson, S. Levin, K.-G. Mler, C. Perrings, and D. Piment el. 1995. Economic growt h, carrying capacity and the environment. Science 268:520-521. Arthur, W. B. 1999. Complexity and the economy. Science 284:107-109. Bassett, T. J., and K. B. Zueli. 2000. Envir onmental discourses and the Ivorian Savanna. Annals of the Association of American Geographers 90 (1):67. Berkes, F., J. Colding, and C. Folke, eds. 2003. Navigating Social-Ecological Systems: Building Resilience for Complexity and Change. Cambridge, UK: Cambridge University Press. Berkes, F., and C. Folke. 1998. Linking social and ecological system s for resilience and sustainability. In Linking Social and Ecological Syst ems: Management Practices and Social Mechanisms for Building Resilience eds. F. Berkes and C. Folke, 1-25. Cambridge, UK: Cambridge University Press. Binford, M., A. Kolata, M. Brenner, J. Ja nusek, M. Seddon, M. Abbott, and J. Curtis. 1997. Climate Variation and the Rise and Fall of an Andean Civilization. Quaternary Research 47:235-248.
149 Blasco, F., M. F. Bellan, and D. Lacaze. 1997. Vegetation Map of Cambodia. Toulouse, France: Laboratoire d'Ecologie Te rrestre de Toulouse. Bogaert, J., A. Farina, and R. Ceulemans. 2005. Entropy increase of fragmented habitats: as sign of human impact? Ecological Indicators 5 (207-212). Briassoulis, H. 2000. Analysis of Land Use Ch ange: Theoretical and Modeling Approaches. Regional Research Institute, West Virginia University. Brock, W. A., and S. R. Carpenter. 2006. Varian ce as a leading indicato r of regime shift in ecosystem services. Ecology and Society 11 (2):art9. Burel, F., and J. Baudry. 2005. Habitat quality and connectivity in agricult ural landscapes: the role of land use systems at various scales in time. Ecological Indicators 5:305-313. Brgi, M., A. M. Hersperger, and N. Schneeber ger. 2004. Driving forces of landscape change current and new directions. Landscape Ecology 19:857-868. Cadenasso, M. L., S. T. A. Pickett, and K. Schwartz. 2007. Spatial heterogeneity in urban ecosystems: reconceptualizing land c over and a framework for classification. Frontiers in Ecology and the Environment 5 (2):80-88. Caldas, M., R. Walker, E. Arima, S. Perz, S. Aldrich, and C. Simm ons. 2007. Theorizing land cover and land use change: The peasan t economy of Amazonian deforestation. Annals of the Association of American Geographers 97 (1):86-110. Carpenter, S. R., D. Ludwig, and W. A. Br ock. 1999. Management of eutrophication for lakes subject to potentially irreversible change. Ecological Applications 9 (3):751-771. Carpenter, S. R., B. Walker, J. M. Anderies and N. Abel. 2001. From metaphor to measurement: resilience of what to what? Ecosystems 4:765-781. Chan, S., S. Tep, and A. Sarthi. 2001. Land Te nure in Cambodia a Data Update: Cambodia Development Resource Institute. Chandler, D. 2000. A History of Cambodia. 3rd ed. Boulder, CO, USA: Westview Press. Chave, J., and S. Levin. 2002. Scale and scaling in ecological and economic systems. Stockholm, Sweden: Beijer International Institute of Ecological Economics, The Royal Swedish Academy of Science. Chomitz, K. M., and D. Gray. 1996. Roads, land use, and deforestation: A spatial model applied to Belize. The World bank Economic Review 10 (3):487-512. Costanza, R., R. d'Arge, R. de Groot, S. Fa rber, M. Grasso, B. Hannon, K. E. Limburg, S. Naeem, R. V. O'Neill, J. Paruelo, R. G. Raskin, P. Sutton, and M. van den Belt. 1998.
150 The value of the world's ecosystem services and natural capital. Ecological Economics 25:3-15. Crawford, T. W. 2005. Spatial fluctuations as sign atures of self-organization: a complex systems approach to landscape dyna mics in Rondnia, Brazil. Environment and Planning B 32 (6):857-875. Crawford, T. W., J. P. Messina, S. M. Ma nson, and D. O'Sullivan. 2005. Guest editorial: Complexity science, complex systems, and land-use research. Environment and Planning B 32 (6):792-798. Crews-Meyer, K. A. 2004. Agricultural landscape change and stability in northeast Thailand: historical patch-level analysis. Agriculture, Ecosystems and Environment 101:155-169. Cromley, R. G. 1982. The Von Thnen model and environmental uncertainty. Annals of the Association of American Geographers 72 (3):404-410. Cropper, M., J. Puri, and C. Griffiths. 2001. Pred icting the Location of De forestation: The Role of Roads and Protected Ar eas in North Thailand. Land Economics 77 (2):172-186. Cumming, G. S. 2004. The impact of low-head dams on fish species richness in Wisconsin, USA. Ecological Applications 14 (5):1495-1506. Cumming, G. S., G. Barnes, S. Perz, M. Schmink, K. E. Sieving, J. Southworth, M. Binford, R. D. Holt, C. Stickler, and T. Van Holt. 2005. An exploratory framework for the empirical measurement of resilience. Ecosystems 8:975-987. Cumming, G. S., and J. Collier. 2005. Change and identity in complex systems. Ecology and Society 10 (1):art. 29. Cumming, G. S., D. H. M. Cu mming, and C. L. Redman. 2006. Scale mismatches in socialecological systems: Causes, consequences and solutions. Ecology and Society 11 (1):art14. Daily, G. C., T. Sderqvist, S. Aniyar, K. Arrow, P. Dasgupta, P. R. Ehrlich, C. Folke, A.-M. Jansson, B.-E. Jansson, N. Kautsky, S. Levin, J. Lubchenco, K.-G. Mler, D. Simpson, D. Starrett, D. Tilman, and B. Walker. 2000. The va lue of nature and th e nature of value. Science 289 (5478):395-396. Dauber, J., M. Hirsch, D. Simmering, R. Wa ldhart, A. Otte, and V. Wolters. 2003. Landscape structure as an indicator or biodiversity: Matrix effects on species richness. Agriculture, Ecosystems and Environment 98:321-329. Davidson-Hunt, I. J., and F. Berkes. 2003. Nature and society through the le ns of resilience: toward a human-in-ecosystem perspective. In Navigating Social-Eco logical Systems:
151 Building Resilience for Complexity and Change. eds. F. Berkes, J. Colding and C. Folke, 53-82. Cambridge, UK: Cambridge University Press. Douglas, I. 2006. The local drivers of land degradation in south-east Asia. Geographical Research 44 (2):123-134. Efron, B. and R.J. Tibshirani. 1993 An Introduction to the Bootstrap. New York, Chapman and Hill. Entwistle, B., S. J. Walsh, R. R. Rindfuss, and A. Chamratrithirong. 1998. Land-Use/Land-Cover and Population Dynamics. In People and Pixels: Linking Remote Sensing and Social Science. eds. D. Liverman, E. F. Moran, R. R. Rindfuss and P. C. Stern, 121-144. Washington DC: National Academy Press. Erenstein, O., A. Oswald, and M. Mahaman. 2006. Determinants of lowland use close to urban markets along an agro-ecological gradient in West Africa. Agriculture, Ecosystems & Environment 117 (2/3):205-217. Etter, A., C. McAlpine, D. Pullar, and H. Possingham. 2006. Modeling the conversion of Colombian lowland ecosystems since 1940: Drivers, patterns and rates. Journal of Environmental Management 79:74-87. Farley, K. A., E. G. Jobbgy, and R. B. Jacks on. 2005. Effects of afforestation on water yield: a global synthesis with implications for policy. Global Change Biology 11 (10):1565-1576. Felkner, J. S. 2000. A Digital Spatial Predic tive Model of Land-Use Change Using Economic and Environmental Inputs and a Statistical Tree Classification Approach: Thailand, 1970s-1990s., Graduate School of Design, Harvard University, Cambridge MA. Foley, J. A., R. DeFries, G. P. Asner, C. Barford, G. Bonan, S. R. Carpenter, F. S. Chapin, M. T. Coe, G. C. Daily, H. K. Gibbs, J. H. Kelkowski, T. Holloway, E. A. Howard, C. J. Kucharik, C. Monfreda, J. A. Patz, I. C. Prentice, N. Ramankutty, and P. K. Snyder. 2005. Global consequences of land use. Science 309:570-574. Folke, C., S. R. Carpenter, T. Elmqvist, L. H. Gunderson, C. S. Holling, and B. Walker. 2002. Resilience and sustainable development: Buil ding adaptive capacity in a world of transformations. Ambio 31 (5):437-440. Folke, C., S. R. Carpenter, B. Walker, M. Scheffer, T. Elmqvist, L. H. Gunderson, and C. S. Holling. 2004. Regime shifts, resilience, and biodiversity in ecosystem management. Annual Review of Ecology, Evolution and Systematics 35:557-581. Gallart, F., and P. Llorens. 2004. Observations on land cover changes and water resources in the headwaters of the Ebro catchment, Iberian Peninsula. Physics and Chemistry of the Earth 29:769-773.
152 Geist, H. J., and E. F. Lambin. 2001. What Dr ives Tropical Deforestation? Louvain-la-Neuve, Belgium: LUCC International Projec t Office, University of Louvain. Geist, H. J., and E. F. Lambin. 20021. Proximate causes and underlying driv ing forces of tropical deforestation. BioScience 52 (2):143-150. Geoghegan, J., L. A. Wainger, and N. E. Bockst ael. 1997. Spatial landscape indices in a hedonic framework: an ecological economics analysis using GIS. Ecological Economics 23 (3):251-264. Gibson, C. C., E. Ostrom, and T. K. Ahn. 2000. Th e concept of scale and the human dimensions of global change: a survey. Ecological Economics 32:217-239. Gottesman, E. 2004. Cambodia after the Khmer Rouge New Haven, CT, Yale University Press. Green, G., C. M. Schweik, and M. Hans on. 2000. Radiometric calibration of Landsat Multispectral Scanner and Th ematic Mapper images: guidelines for the global changes community. Working Paper. Bloomington, Indian a: Center for the Study of Institutions, Population, and Environmental Change. Gross, J. E., R. R. J. McAllister, N. Abel, D. M. Stafford-Smith, and Y. Maru. 2006. Australian rangelands as complex adaptive systems: A conceptual model and preliminary results. Environmental Modeling & Software 21 (9):1264-1272. Gunderson, L. H., and L. Pritchard Jr, eds. 2002. Resilience and the Behavior of Large-scale Systems. Washington DC: Island Press. Gustafson, E. J. 1998. Quantifying la ndscape spatial patter n: What is the state of the art? Ecosystems 1:143-156. Hartvigsen, G., A. Kinzig, a nd G. D. Peterson. 1998. Use and analysis of complex adaptive systems in ecosystem science: overview of special section. Ecosystems 1:427-430. Henry, S., P. V., D. Oudraogo, and E. F. Lamb in. 2004. Descriptive analysis of the individual migratory pathways according to environmental typologies. Population and Environment 25 (5):397-422. Holland, J. H. 1995. Hidden Order: How Adapta tion Builds Complexity. Reading, MA: Helix Books. Holling, C. S. 1973. Resilience and st ability of ecological systems. Annual Review of Ecology and Systemics 4:1-23. 2001. Understanding the complexity of ec onomic, ecological, and social systems. Ecosystems 4:390-405.
153 Holling, C. S., and L. H. Gunderson. 2002. Resilience and Adaptive Cycles. In Panarchy: Understanding Transformations in Human and Natural Systems eds. L. H. Gunderson and C. S. Holling, 25-62. Washington DC: Island Press. Holling, C. S., L. H. Gunderson, and D. Ludwig. 2002. In Quest of a Theory of Adaptive Change. In Panarchy: Understanding Transformati ons in Human and Natural Systems eds. L. H. Gunderson and C. S. Holling. Washington DC: Island Press. Holling, C. S., L. H. Gunderson, and G. D. Pe terson. 2002. Sustainability and Panarchies. In Panarchy: Understanding Transformati ons in Human and Natural Systems eds. L. H. Gunderson and C. S. Holling. Washington DC: Island Press. Hong, S. K. 1999. Cause and consequence of lands cape fragmentation and changing disturbance by socio-economic development. Journal of Environmental Sciences 11 (2):181-187. Houghton, R. A. 1994. The worldwide extent of land-use change. BioScience 44 (5):305-313. Huang, Q.-H., and Y.-L. Cai. 2007. Simulation of land use change using GIS-based stochastic model: the case study of Shiqia n County, Southwestern China. Stochastic Environmental Research & Risk Assessment 21 (4):419-426. Irwin, E. G., and J. Geoghegan. 2001. Theory, da ta, methods: developing spatially explicit economic models of land use change. Agriculture, Ecosystems and Environment 85:7-23. Janssen, M. 1998. Use of complex adaptiv e systems for modeling global change. Ecosystems 1:457-463. Janssen, M., J. M. Anderies, and E. Ostrom. 200 7. Robustness of Social-Ecological Systems to Spatial and Temporal Variability. Society and Natural Resources 20:307-322. Janssen, M., J. M. Anderies, and B. Walker. 2004. Robust strategies for managing rangelands with multiple stable attractors. Journal of Environmental Economics and Management 47:140-162. Jensen, J. R. 2000. Remote Sensing of the Environment: An Earth Resource Perspective. Edited by K. C. Clarke, Prentice Hall Series in Geographic Information Science Upper Saddle River, NJ: Prentice Hall. Kauffman, S. 1995. At Home in the Universe: The Search for the Laws of Self-Organization and Complexity. New York: Oxford University Press. Kim, D.-S., K. Mizuno, and S. Kobayashi. 2003. M odeling urbanization by accessibility in rapidgrowth areas. Journal of Urban Planning and Development 129 (1):45-63. Koestler, A. 1968. The Ghost in the Machine. New York: Macmillan.
154 Kruseman, G., R. Ruben, and G. Tesfay. 2006. Diversity and development domains in the Ethiopian highlands. Agricultural Systems 88:75-91. Kuhn, T. 1996. The Structure of Scie ntific Revolutions. 3rd ed. Chicago: University of Chicago Press. Kummer, D. M., and B. L. Turner II. 1994. The human causes of deforestation in Southeast Asia. BioScience 44 (5):323-328. Lambin, E. F., X. Baulies, N. Bockstael, G. Fisc her, T. Krug, R. Leemans, E. F. Moran, R. R. Rindfuss, Y. Sator, D. Skole, B. L. Turn er II, and C. Vogel. 1999. Land-Use and LandCover Change Implementation Plan. Louvain -la-Neuve, Belgium: LUCC International Project Office, University of Louvain. Lambin, E. F., H. J. Geist, and E. Lepers. 2003. Dynamics of Land-Use and Land-Cover Change in Tropical Regions. Annual Review of Environment & Resources 28 (1):205. Lambin, E. F., B. L. Turner II, H. J. Geist, S. B. Agbola, A. Angelsen, J. W. Bruce, O. T. Coomes, R. Dirzo, G. Fischer, C. Folke, P. S. George, K. Homewood, J. Imbernon, R. Leemans, X. Li, E. F. Moran, M. Mortimore, P. S. Ramakrishnan, J. F. Richards, H. Skaness, W. Steffen, G. D. Stone, U. Svedi n, T. A. Veldkamp, C. Vogel, and J. Xu. 2001. The causes of land-use and land-cove r change: moving beyond the myths. Global Environmental Change 11:261. Lansing, J. S. 2003. Comple x adaptive systems. Annual Review of Anthropology 32:183-204. Laurance, W. F., A. K. M. Alern az, G. Schroth, P. M. Fearnside, S. Bergen, E. M. Venticinque, and C. Da Costa. 2002. Predictors of deforestation in the Brazilian Amazon. Journal of Biogeography 29:737-748. Lausch, A., and F. Herzog. 2002. Applicability of landscape metrics for the monitoring of landscape change: issues of scale, resolution and interpretability. Ecological Indicators 2:3-15. Legendre, P. 1993. Spatial autocorrela tion: trouble or new paradigm? Ecology 74 (6):16591673. Levin, S. 1992. The problem of pa ttern and scale in ecology. Ecology 73 (6) 1943-1967. 1998. Ecosystems and the biosphere as complex adaptive systems. Ecosystems 1:431436. 1999. Fragile Dominion: Complexity and the Commons. Reading, MA, USA: Helix Books.
155 2003. Complex adaptive systems: expl oring the known, the unknown and the unknowable. Bulletin of the American Mathematical Society 40 (1):3-19. 2005. Self-organization and the emergence of complexity in ecological systems. BioScience 55 (12):1075-1079. Limburg, K. E., R. V. ONeill, R. Costanza, and S. Farber. 2002. Complex systems and valuation. Ecological Economics 41:409-420. Liu, L. 1999. Labor location, conservation and la nd quality: the case of West Jilin, China. Annals of the Association of American Geographers 89 (4):633-657. Low, B., E. Ostrom, C. Simon, and J. A. Wilson. 2003. Redundancy and diversity: do they influence optimal management? In Navigating Social-Ecologi cal Systems: Building Resilience for Complexity and Change eds. F. Berkes, J. Colding and C. Folke. Cambridge, UK: Cambridge University Press. Ludwig, D., B. Walker, and C. S. Holling. 2002. Models and metaphors of sustainability, stability and resilience. In Resilience and the Behavior of Large-scale Systems eds. L. H. Gunderson and L. J. Pritchard, 21-48 Washington DC: Island Press. Lugo, A. E., F. N. Scatena, W. Silver, S. M. Coln, and P. G. Murphy. 2002. Resilience of Tropical Wet and Dry Forests in Puerto Rico. In Resilience and the Behavior of Largescale Systems eds. L. H. Gunderson and L. Pritchard Jr, 21-48. Washington DC: Island Press. MacArthur, R. 1960. On the Relative Abundance of Species. The American Naturalist 94 (874):25-36. Maffi, L. 2005. Linguistic, cultu ral and biological diversity. Annual Review of Anthropology 29:599-617. Magurran, A. E. 1988. Ecological Diversity and Its Measurement. Princeton, NJ, Princeton University Press. Manson, S. M. 2001. Simplifying complexity: a review of complexity theory. Geoforum 32:405415. 2005. Agent-based modeling and genetic progr amming for modeling land change in the Southern Yucatn Penins ular Region of Mexico. Agriculture, Ecosystems and Environment 111:47-62. Markham, B., and J. Barker. 1986. Landsat M SS and TM post-calibration dynamic ranges, exoatmospheric reflectances and at-s atellite temperatures. Lands at Technical Notes, no. 1. In EOSAT Landsat Data User Notes Lanham, Maryland.
156 Markos, E., and K. Gebre-Egziabher. 2001. Rural out-migration in the drought prone areas of Ethiopia: a multilevel analysis. International Migration Review 35 (3):749-771. Martin, R. 1999. The new geographical turn in economics: some critical reflections. Cambridge Journal of Economics 23 (1):65-91. Mas, J. F., H. Puig, J. L. Palacio, and A. Sosa-Lpez. 2004. Modeling deforestation using GIS and artificial neural networks. Environmental Mo deling & Software 19 (5):461-471. McClanahan, T. R., N. V. C. Polunin, and T. J. Done. 2002. Resilience of Coral Reefs. In Resilience and the Behavior of Large-scale Systems eds. L. H. Gunderson and L. Pritchard Jr, 111-163. Washington DC: Island Press. McConnell, W. J., S. P. Sweeney, and B. Mulle y. 2004. Physical and soci al access to land: spatio-temporal patterns of agricu ltural expansion in Madagascar. Agriculture, Ecosystems & Environment 101:171-184. FRAGSTATS: Spatial Pattern Analysis Progr am for Categorical Maps. University of Massachusetts, Amherst. McMichael, A. J., B. Bolin, R. Costanza, G. C. Daily, C. Folke, K. Lindahl-Kiessling, E. Lindgren, and B. Niklasson. 1999. Globalization a nd sustainability of human health: an ecological perspective. BioScience 49 (3):205-210. Mekong River Commission. 2003. People and the Environment Atlas of the Lower Mekong Basin Vientiane: Mekong River Commissi on. CD-ROM Interactive Atlas. Messina, J. P., S. J. Walsh, C. F. Mena and P. L. Delamater. 2006. Land tenure and deforestation patterns in th e Ecuadorian Amazon: Conflicts in land conservation in frontier settings. Applied Geography 26:113-128. Moran, E. F. 1979. Human Adaptability: An Introduc tion to Ecological Anthropology. North Scituate, Massachusetts. : Duxbury Press. Munroe, D. K., C. Croissant, and A. M. York. 2005. Land use policy and landscape fragmentation in an urba nizing region: assessing the impact of zoning. Applied Geography 25 (121-141.). Munroe, D. K., J. Southworth, and C. M. Tu cker. 2002. The dynamics of land-cover change in Western Honduras: exploring spatial and temporal complexity. Agricultural Economics 27 (3):355-369. 2004. Modeling spatially and temporally co mplex land-cover change: The case of western Honduras. Professional Geographer 56 (4):544-559.
157 Nagendra, H., D. K. Munroe, and J. Southwort h. 2004. From pattern to process: landscape fragmentation and the analysis of land use/land cover change. In Agriculture, Ecosystems and Environment 101:111-115. Nagendra, H., J. Southworth, and C. M. Tu cker. 2003. Accessibility as a determinant of landscape transformation in Western H onduras: linking pattern and process. Landscape Ecology 18: 141-158. Naiman, R. J., and H. Decamps. 1997. The ecology of interfaces: Riparian Zones. Annual Review of Ecology & Systematics 28:621. Nang Rong Projects. 2004. LULC Classification Scheme: Ca rolina Population Center, University of North Carolina. National Statistics Office. 2003. 2003 Agricultural Census Northeastern Region, ed. N. S. Office: Ministry of Informa tion and Communication Technology. Naveh, Z. 1994. From biodiversity to ecodivers ity: a landscape-eco logy approach to conservation and restoration. Restoration Ecology 2 (3):180-189. Ngigi, T. G., and R. Tateishi. 2004. Monitoring deforestation in Kenya. International Journal of Environmental Studies 61 (3):281-291. O'Neill, R. V. 1999. Recovery in complex ecosystems. Journal of Aquatic Ecosystem Stress and Recovery 6:181-187. Odum, E. P. 1975. Ecology, the Link Between th e Natural and the Social Sciences New York. Holt, Rinehart and Winston. Ojima, D. S., K. A. Galvin, and B. L. Turner II. 1994. The global impact of land-use change. BioScience 44 (5):300-304. Ostrom, E., J. Burger, C. B. Field, R. B. Norgaard, and D. Policansky. 1999. Revisiting the commons: local lessons, global challenges. Science 284 (5412):278-282. Overmars, K. P., and P. H. Verburg. 2005. Analys is of land use drivers at the watershed and household level: Linking two paradigms at the Philippine forest fringe. International Journal of Geographical Information Science 19 (2):125-152. Pan, W. K. Y., and R. E. Bilsborrow. 2005. The us e of a multilevel statistical model to analyze factors influencing land use: a study of the Ecuadorian Amazon. Global and Planetary Change 47:232-252. Pan, W. K. Y., S. J. Walsh, R. E. Bilsborrow, B. G. Frizzelle, C. M. Erlien, and F. Baquero. 2004. Farm-level models of spatial patterns of land use and land cover dynamics in the Ecuadorian Amazon. Agriculture, Ecosystems and Environment 101: 117-134.
158 Parker, D. C., T. Berger, and S. M. Mans on. 2001. Agent-Based Models of Land-Use and LandCover Change: LUCC Report Series No. 6. Report and Review of an International Workshop, October 4 2001, Irvine, Californi a, USA. Paper read at Agent-Based Models of Land-Use and Land-C over Change., at Irvine, CA. Parker, D. C., S. M. Manson, M. Janssen, M. J. Hoffmann, and P. Deadman. 2003. Multi-agent systems for the simulation of land-us e and land-cover change: A review. Annals of the Association of American Geographers 93 (2):314-337. Parnwell, M. J. G. 1988. Rural poverty, developm ent and the environment: the case of NorthEast Thailand. Journal of Biogeography 15:199-208. Peet, R. K. 1974. The measurement of species diversity. Annual Review of Ecology & Systematics 5:285-307. Perz, S. 2005. The importance of household asset diversity for livelihood diversity and welfare among small farm colonists in the Amazon. The Journal of Development Studies 41 (7):1193-1220. 2007. Grand theory and context-specificity in the study of forest dynamics: Forest transition theory and other directions. Professional Geographer 59 (1):105-114. Pontius, R. G., E. Shusas, and M. McEach ern. 2004. Detecting important categorical land changes while accounting for persistence. Agriculture, Ecosystems & Environment 101 (2-3):251-268. Redman, C. L., J. M. Grove, and L. H. Kuby. 2004. Integrating social science into the LongTerm Ecological Research (LTER) network: Social dimensions of ecological change and ecological dimensions of social change. Ecosystems 7:161-171. Rindfuss, R. R., S. J. Walsh, B. L. Turner II J. Fox, and V. Mishra. 2004. Developing a science of land change: challenges and methodological issues. Proceedings of the National Academy of Science 101 (39):13976-13981. Rogan, J., and D. M. Chen. 2004. Remote sensi ng technology for mappi ng and monitoring landcover and land-use change. Progress in Planning 61:301-325. Roques, K. G., T. G. OConnor, and A. R. Watkinson. 2001. Dynamics of shrub encroachment in an African savanna: relative influences of fire, herbivory, rainfall and density dependence. Journal of Applied Ecology 38 (2):268-280. Schweik, C. M., T. P. Evans, and J. G. Morgan. 2005. Open source and open content: A framework for global collaboration in social-ecological research. Ecology and Society 10 (1):1-25.
159 Semthiti, T. 1951. Co-operation, Credit, and Capital in the Agriculture of Thailand. Paper read at International Conference on Land Tenure and Related Problems in World Agriculture, 1956, at Madison, WI, USA. LAB Fit Curve Fitting Software (Nonlinear Re gression and Treatment of Data Program) V 7.2.34. Simon, H. A. 1962. The architecture of complexity. Proceedings of the American Philosophical Society 106 (6): 467-482. Soares-Filho, B., A. Alencar, D. Nepstad, C. Di az, S. Rivero, L. Solrzano, and E. Voll. 2004. Simulating the response of land-cover cha nges to road paving and governance along a major Amazon highway: The Santarm-Cuiab corridor. Global Change Biology 10:745-764. Soini, E. 2005. Land use change patterns and livelihood dynamics on the slopes of Mount Kilimanjaro, Tanzania. Agricultural Systems 85 (3): 306-323. Southworth, J., D. K. Munroe, and H. Na gendra. 2004. Land cover change and landscape fragmentation comparing the utility of continuous and disc rete analyses for a western Honduras region. Agriculture, Ecosystems and Environment 101:185-205. Southworth, J., H. Nagendra, L. A. Carlson, and C. M. Tuck er. 2004. Assessing the impact of Celaque National Park on forest fragmentation in western Honduras. Applied Geography 24 (4):303-322. Southworth, J., and C. M. Tucker. 2001. The infl uence of accessibility, local institutions, and socio-economic factors on forest cover cha nge in the mountains of western Honduras. Mountain Research and Development 21 (3):276-283. Sponsel, L. E. 1998. Historical Ecology of Thailand. In Advances in Historical Ecology ed. W. Bale. New York, USA: Columbia University Press. Stefanov, W.L. and M. Netzband. 2005. Assess ment of ASTER land cover and MODIS NDVI data at multiple scales for ecological characterization of an arid urban center. Remote Sensing of Environment. 99 (1/2):31-43. Stepp, J. R., H. Castaneda, and S. Cervone 2005. Mountains and bi ocultural diversity. Mountain Research and Development 25 (3):223-227. Stibig, H.-J., R. Beuchle, and F. Achard. 2003. Ma pping of the tropical fo rest cover of insular Southeast Asia from SPOT4-Vegetation images. International Journal of Remote Sensing 24 (18):3651.
160 Swift, M. J., A.-M. N. Izac, and M. van Noor dwijk. 2004. Biodiversity and ecosystem services in agricultural landscapes are we asking the right questions? Agriculture, Ecosystems and Environment 104:113-134. Tansey, K. J., A. J. Luckman, L. Skinner, H. Balzter, T. Strozzi, and W. Wagner. 2004. Classification of forest volume resources using ERS tandem coherence and JERS backscatter data. International Journal of Environmental Studies 25 (4):751-768. Teillet, P. M., and G. Fedosejeus. 1995. On the dark target approach to a tmospheric correction of remotely sensed data. Canadian Journal of Remote Sensing 21 (4):374-387. Turner II, B. L., D. Skole, S. Sanderson, G. Fi scher, L. Fresco, and R. Leemans. 1995. Land-Use and Land-Cover Change Science/Research Plan. Louvain-la-Neuve, Belgium: LUCC International Project Office, University of Louvain. Turner, M. D. 2003. Methodological reflections on the use of remote sensing and GIScience in human ecological research. Human Ecology 31 (2): 255-279. Turner, M. G., R. H. Gardner, and R. V. ONeill. 2001. Landscape Ecology in Theory and Practice: Pattern and Process. New York: Springer-Verlag. Vgen, T.-G. 2006. Remote sensing of complex land use change tr ajectories: A case study from the highlands of Madagascar. Agriculture, Ecosystems & Environment 115 (1-4):219228. van Wey, L. K. 2005. Land Ownership as a Determin ant of International a nd Internal Migration in Mexico and Internal Migration in Thailand. International Migration Review 39 (1):141-172. van Gils, H., and A. Loza Armand Ugon. 2006. What Drives Conversion of Tropical Forest in Carrasco Province, Bolivia? AMBIO A Journal of the Human Environment 35 (2):81. Verburg, P. H., W. Soepboer, T. A. Veldkamp, R. Limpiada, V. Es paldon, and S. S. A. Mastura. 2002. Modeling the spatial dynamics of regi onal land use: the CLUE-S model. Environmental Management 30 (3):391-405. Vitousek, P. M., H. A. Mooney, J. Lubchenc o, and J. M. Melillo. 1997. Human domination of Earths ecosystems. Science 277 (5323):494-499. Walker, B., and N. Abel. 2002. Resilient Rangelands Adaptation in Complex Systems. In Panarchy: Understanding Transformati ons in Human and Natural Systems eds. L. H. Gunderson and C. S. Holling, 293-313. Washington DC: Island Press. Walker, R. 2004a. Introduction [to special issue on land-use land-cover change]. International Regional Science Review 27 (3):243-246.
161 2004b. Theorizing land-cover and land-use cha nge: the case of tr opical deforestation. International Regional Science Review 27 (3):247. Walker, R., and W. Soleki. 2004. Theorizing land -cover and land-use change: the case of the Florida Everglades and its degradation. Annals of the Association of American Geographers 94 (2): 311-328. Walters, C. J. 1986. Adaptive Management of Renewable Resources. New York: Macmillan. Werner, B. T. 1999. Complexity in natural landform patterns. Science 284:102-104. Wessels, K. J., R. DeFries, J. Dempewolf, L. O. Anderson, A. J. Hansen, S. L. Powell, and E. F. Moran. 2004. Mapping regional la nd cover with MODIS data for biological conservation: Examples from the Greater Yellowstone Ec osystem, USA and Par State, Brazil. Remote Sensing of Environment 92 (1):67-83. Wiegand, K., D. Saltz, and D. Ward. 2006. A patchdynamics approach to savanna dynamics and woody plant encroachment Insights from an arid savanna. Perspectives in Plant Ecology, Evolution and Systematics 7:229-242. Wilson, J. A. 2006. Matching social and ecologi cal systems in complex ocean fisheries. Ecology & Society 11 (1):263. Wood, C. H., and R. Porro, eds. 2002. Deforestation and Land Use in the Amazon. Gainesville, FL: University Press of Florida. Wulder, M. A., J. C. White, S. Magnussen, a nd S. McDonald. 2007. Valida tion of a large area land cover product using purpose -acquired airborne video. Remote Sensing of Environment 106 (4):480-491. Wyatt, D. K. 1984. Thailand: A Short History. Chang Mai, Thailand: Silkworm Books. Xu, W. 2004. The changing dynamics of land-use ch ange in rural China: a case study of Yuhang, Zhejiang Province. Environment and Planning A 36 (9):1595-1615. Young, O. R., E. F. Lambin, F. Alcock, H. Haberl, S. I. Karlsson, W. J. McConnell, T. Myint, C. Pahl-Wostl, C. Polsky, P. S. Ramakrishnan, H. Schroeder, M. Scouvart, and P. H. Verburg. 2006. A portfolio approach to analyzing complex human-environment interactions: Institutions and land change. Ecology and Society 11 (2):art31.
162 BIOGRAPHICAL SKETCH Lin Cassidy was born in South Africa. She gained her Bachelor of Arts at the University of Cape Town, with a double major in French a nd sociology. She then completed a 1-year honors degree in French literature, before moving to Maun, Botswana in 1987. After working in the wildlife-based eco-tourism industry in the Okavango Delta for several years, she attended the University of Zimbabwe, where she obtained a Bachelor of Science special honors degree in sociology. Lin returned to Botswana where she ha s been working as a freelance consultant in rural development, mainly on short-term proj ects for donor organizations, NGOs, or government departments. Increasingly she has focused her work on the community-based natural resources management sector, with a part icular interest in the areas surrounding th e Okavango. In 2001 she became a naturalized Botswana citizen. Lin came to the University of Florida to learn more about the ecological asp ects of natural resources management and to pursue the links between social and ecological systems, and in 2003 obt ained her Master of Science degree in interdisciplinary ecology. She completed her PhD in geography, with a minor in political ecology, through the University of Floridas Land Use and Environmental Change Institute and Department of Geography, before returning home to Botswana to continue working on issues relating to conservation and development, and peoples reliance on natu ral resources and the environment.