|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 PARKS AND POLICIES: INTEGRATING GEOSPATIAL TOOLS AND MODELING TO EVALUATE CONSERVATION INTERVENTIONS IN CENTRAL INDIA By PINKI MONDAL A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011
2 2011 Pinki Mondal
3 To my parents
4 ACKNOWLEDGMENTS I would like to take this opportunity to thank my committee members who made this dissertation possible with their continued support My sincere gratitude must be given to my adviser Dr. Jane Southworth, whose guidance, support, and sense of humor helped m e to surviv e graduate school. In addition, I would like to thank Dr. Tim Fik and Dr. Harini Nagendra for their valuable input and constant enthusiasm towards my research. I would also like to thank Dr. Brian Child and Dr. Mel Sunquist for their support thr oughout the degree program. Much of the data for this dissertation was generously shared by Dr. Michael Binford and Dr. Peter Waylen. I would also like to acknowledge the continuous motivation provided by my peers in the University of Florida geography dep artment. I must thank Desiree Price and Julia Williams who managed all the paper works, so that I can concentrate on my research. I would like to thank the University of Florida Alumni Fellowship and Dr. Andy Tatem for funding. This research was partially funded by a field research grant from Tropical Conservation and Development Program at the University of Florida. The fieldtrip required for this study would not have been possible without the help and support from the members of SHODH, especially Dr. Ruc ha Ghate, and my field assistant Abhishek Durshetwar. I would like to thank the Maharashtra State Forest Department, India, especially Dr. Mohan Jha for the help provided during the field work for this study. Finally, I would like to acknowledge my parent s unconditional love and support which helped me to remain focused during trying times. My heartfelt gratitude goes out to my husband, Chandranath Basak, whose continued criticism and incessant culinary creations improved the quality of this dissertation.
5 TABLE OF CONTENTS ACKNOWLEDGMENTS .................................................................................................. 4 page LIST OF TABLES ............................................................................................................ 8 LIST OF FIGURES .......................................................................................................... 9 LIST OF ABBREVIATIONS ........................................................................................... 11 ABSTRACT ................................................................................................................... 12 CHAPTER 1 INTRODUCTION .................................................................................................... 14 Outline and Scope of the Research ........................................................................ 14 Study Area .............................................................................................................. 15 Dissertation Structure ............................................................................................. 16 2 TRENDS OF FOREST DYNAMICS IN TIGER LANDSCAPES ACROSS ASIA ..... 19 Introduction ............................................................................................................. 19 Methods .................................................................................................................. 23 Results .................................................................................................................... 24 Summary Characteristics of the CaseStudies ................................................. 24 Spectrum of Forest Dynamics: Deforestation to Reforestation ......................... 25 Proximate Causes of Changes in Forested Landscapes: Country Level Issues ............................................................................................................ 27 Discussion .............................................................................................................. 29 Forest Dynamics in Individual Countries: Trends, Challenges and Implications for Tiger Conservation ............................................................... 29 Bangladesh ................................................................................................ 29 China ......................................................................................................... 30 India ........................................................................................................... 31 Indonesia ................................................................................................... 32 Myanmar .................................................................................................... 33 Nepal ......................................................................................................... 34 Russia ........................................................................................................ 35 Forest Dynamics in Tiger Landscapes: Common Needs for Tiger Conservation ................................................................................................. 35 Conclusion ........................................................................................................ 37 3 PROTECTION VS. COMMERCIAL MANAGEMENT: SPATIAL AND TEMPORAL ANALYSIS OF LAND COVER CHANGES IN THE TROPICAL FORESTS OF CENTRAL INDIA ............................................................................. 48
6 Introduction ............................................................................................................. 48 Indian Forestry Policies and Practices ................................................................. 51 Methods .................................................................................................................. 53 Satellite Imagery and GIS ................................................................................. 53 Image Pre processing ...................................................................................... 53 Image Classification ......................................................................................... 54 Change Detection ............................................................................................. 54 NDVI Standard Normal Deviates and Image Differencing ................................ 55 Results .................................................................................................................... 56 Spatial and Temporal Land Cover Change Trajectories ................................... 56 Temporal Changes in Vegetation Productivity .................................................. 57 Discussion .............................................................................................................. 57 Conclusion .............................................................................................................. 63 4 EVALUATION OF CONSERVATION INTERVENTIONS USING A CELLULAR AUTOMATAMARKOV MODEL ............................................................................. 74 Introduction ............................................................................................................. 74 Methods .................................................................................................................. 79 Land Cover Maps ............................................................................................. 79 CAMarkov Model ............................................................................................. 79 Fragmentation Indices ...................................................................................... 83 Results .................................................................................................................... 84 Changes in Land Covers and Forest Transitions ............................................. 84 Changes in Fragmentation Patterns ................................................................. 85 Model Validation ............................................................................................... 86 Discussion .............................................................................................................. 87 Conclusion .............................................................................................................. 91 5 QUANTIFYING SURFACE GRADIENTS WITH A 2 BAND ENHANCED VEGETATION INDEX (EVI2) .................................................................................. 98 Introduction ............................................................................................................. 98 Methods ................................................................................................................ 102 Study Area ...................................................................................................... 102 Spatial Data .................................................................................................... 102 Statistical Analysis .......................................................................................... 104 Surface Metrics .............................................................................................. 104 Results .................................................................................................................. 106 Statistical Analysis .......................................................................................... 106 Surface Metrics .............................................................................................. 107 Discussion ............................................................................................................ 108 Conclusion ............................................................................................................ 111 6 COMPARISON OF ASTER DERIVED THERMAL INDICES AND TASSELED CAP NORMALIZED INDICES IN DETECTING TROPICAL FOREST SUCCESSIONS .................................................................................................... 118
7 Introduction ........................................................................................................... 118 Methods ................................................................................................................ 119 Study Area ...................................................................................................... 119 Data Pre processing ....................................................................................... 120 Successional Forest Subsets ......................................................................... 121 Emissivity and Surface Temperature Extraction ............................................. 121 Normalized Tasseled Cap Indices .................................................................. 122 Statistical Analysis .......................................................................................... 122 Results .................................................................................................................. 123 Discussion ............................................................................................................ 123 7 CONCLUSION ...................................................................................................... 130 Contribution of the Study ...................................................................................... 130 Implication for PTR ............................................................................................... 131 LIST OF REFERENCES ............................................................................................. 133 BIOGRAPHICAL SKETCH .......................................................................................... 151
8 LIST OF TABLES Table page 1 1 Research objectives and research questions ..................................................... 17 2 1 Individual description of the casestudies ........................................................... 39 2 2 Country wise summary of proximate causes behind forest dynamics and priority issues for tiger conservation ................................................................... 43 3 1 Description of satellite imagery used in the study ............................................... 65 3 2 Change trajectory classes created for the four date change analysis ................ 65 3 3 Two date land cover change trajectories for park and buffers ............................ 66 3 4 Four date land cover change trajectories for park and surrounding buffers ....... 68 4 1 Forest conservation acts/policies of India from 19471990.. ............................... 92 5 1 Characteristics of ASTER visible near infrared, short wave infrared and thermal infrared bands. ..................................................................................... 112 5 2 Correlation matrix showing the level of interrelationships among the remote sensing variables .. ............................................................................................ 113 5 3 Discriminant analysis results showing the relative capability of the remote sensing variables.. ............................................................................................ 114 5 4 Surface metrics calculated from EVI2 for four sample landscapes from the study area. ........................................................................................................ 114 6 1 Calculation for newly developed emissivity indices. ......................................... 126 6 2 The level of interrelationships of forest successions with remotely sensed thermal and vegetation indices.. ....................................................................... 126 6 3 Discriminant analysis results showing the relative capability of input variables in detecting tropical forest successions.. .......................................................... 1 27
9 LIST OF FIGURES Figure page 1 1 Location of the study area .................................................................................. 18 2 1 Spatial distribution of the case studies within tiger ranging countries ................. 45 2 2 Summary characteristics of the case studies ...................................................... 46 2 3 Forest dynamics and temporal span of the case studies .................................... 46 2 4 Country wise proportion of proximate causes driving forest dynamics ............... 47 3 1 Study area showing the boundary of the study area, and locations of the visited villages. ................................................................................................... 69 3 2 Series of two date change trajectories of the park and 10 km buffer. ................. 70 3 3 Four date change trajectory (1977 1989 2000 2007) of the park and 10 km buffer. ................................................................................................................. 71 3 4 Series of two date NDVI standard normal deviate image subtractions ............... 72 3 5 Two date change trajectories of forest cover showing changes in forest cover transitions. .......................................................................................................... 73 3 6 Spatial and temporal changes in land cover dynamics within the study area during 19772007.. ............................................................................................. 73 4 1 Map of the study area showing PTR and surrounding buffer up to 10 km, and visited villages in the area. ................................................................................. 93 4 2 Land cover maps showing actual landscapes, and CA Markov model generated landscapes. ....................................................................................... 94 4 3 Actual and predicted land covers for the park and the buffer showing changes in forest and bare soil. .......................................................................... 95 4 4 Two date change trajectories for the park and the buffer showing actual vs. predicted forest transitions. ................................................................................ 95 4 5 Series of landscape metrics for forest cover within the park.. ............................. 96 4 6 Series of landscape metrics for forest cover in the buffer.. ................................. 97 5 1 Land cover map show ing locations of four sample landscapes in reference to the Pench Tiger Reserve. ................................................................................. 115
10 5 2 EVI2 histograms for four sample landscapes. .................................................. 116 5 3 Spatial patterns in the study area demonstrated by various remote sensing variables .......................................................................................................... 117 6 1 Study area showing four date change trajectory, stable forest mask, and regenerated forest mask. .................................................................................. 128 6 2 Study area showing estimated blackbody surface temperature of stable forest. ............................................................................................................... 129
11 LIST OF ABBREVIATIONS ASTER Advanced Spaceborne Thermal Emission and Reflection Radiometer GIS Geographic Information Science IGBP International GeosphereBiosphere programme IHDP International Human Dimensions Programme on Global Environmental Change NASA National Aeronautics and Space Administration SMD Science Mission Directorate
12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy PARKS AND POLICIES: INTEGRATING GEOSPATIAL TOOLS AND MODELING TO EVALUATE CONSERVATION INTERVENTIONS IN CENTRAL INDIA By Pinki Mondal May 2011 Chair: Jane Southworth Major: Geography Tropical forests worldwide are undergoing rapid changes due to increasing human populations and varied land use practices. In an effort to protect these forests and their species, the number of protected areas has increased exponentially in recent decades. While the amount of money invested into conservation has been increasing along with the number of protected areas across the globe in the past few decades, few well designed empirical studies try to show what could have happened in the absence of the conservation efforts. This study combi nes tools and techniques from remote sensing, GIS, and landscape ecology to evaluate conservation interventions in a tropical landscape in Central India. The study area, namely Pench tiger reserve, Maharashtra is embedded within a matrix of other land cov er types, including nonprotected forests, agri cultural fields, or urban areas. It is critical to monitor this landscape for changing forest dynamics as this region can host viable tiger population in presence of adequate prey base, which in turn is direct ly dependent on the forest for food and cover. This study utilizes spatiotemporal forest cover transitions, and simulated landscapes assuming absence of the conservation intervention as proxies for the effectiveness of this protected area. This
13 study also explores the role of continuous vegetation indices, and surface temperature derived from multiple satellite images in describing and quantifying changes in forest composition and structure. Methods used in this study are robust and can be used in any ecol ogical settings. Findings of this study show that while national level policies establishing protected areas play an important role in conservation, alternative approaches including favorable management policies could also be effective to extend conservat ion over larger areas. This study also suggests that combination of methods from multiple disciplines can be effectively used for spatiotemporal assessment of changing forest cover, structure and composition as well as empirical evaluation of conservation intervention, particularly in developing countries.
14 CHAPTER 1 INTRODUCTION Outline and Scope of the Research Earths carbon cycle and ecosystems are being subjected to human intervention and climate changes on an unprecedented scale, both in terms of rate and extent (NASA SMD, 2007). A better understanding of the implications of these changes is required for food production, biodiversity, sustainable resource management, and the maintenance of a healthy, product ive environment (PinstrupAnderson and PandyaLorch, 1998). The scientific community recognizes the importance of frequent, repeat observations from space, at both moderate and high spatial resolutions, to address the heterogeneity of living systems. In addition, research projects that integrate an understanding of the coupled humanenvironment system, such as the IGBP/IHDP Global Land Project, are considered an important component of global change research. This dissertation contributes to such fields of r esearch by integrating advanced techniques from remote sensing, GIS and landscape ecology that contribute to the better understanding of landscape dynamics in humandominated socioecological systems. Specifically, this dissertation documents spatial and t emporal landscape patterns emerged from differential conservation strategies in a humanmodified landscape through analysis of land cover conversion and modification as well as empirical evaluation of conservation interventions. This dissertation addresses issues related to biodiversity conservation, forest management strategies, and national level policies in and around a tiger reserve in Central India, which I am using as a template for tropical ecosystems. This research is based on quantitative analysis of multi date and multi platform satellite images of the
15 study area. There are five research objectives, which along with the corresponding research questions are listed under Table 11. Study Area Pench Tiger Reserve ( PTR) is situated in the Nagpur distr ict of the northern part of the state of Maharashtra (Fig. 1 1). The forest type is southern tropical dry deciduous forest (Champion and Seth, 1968). The floral diversity is dominated by teak ( Tectona grandis ) and includes other species such as Indian laur el ( Terminalia tomentosa), white marudah ( Terminalia arjuna) and bamboo ( Bambusa arundinacea ). The faunal diversity includes 33 species of mammals including tiger, leopard, spotted deer and over 160 species of birds, reptiles, and fishes. Overall topography of the region is undulating, the highest peak inside the park being 583 m. Most of the annual precipitation (1050 mm) is received between June and September, with temperatures varying from 14 to 43 C. PTR, Maharashtra shares its name and the northern boundary with PTR, Madhya Pradesh (Fig. 1 1). These two reserves form together the first inter state tiger reserve of India. However, they fall under the jurisdiction of two different states, and thus two state forest departments. PTR, Maharashtra (area: 257.2 km2) was declared a national park by the Government of Maharashtra in 1975 and a tiger reserve by the Government of India in 1999. Over 40 villages within 10 km from the park boundary (Fig. 1 1) host significant human (mostly of Gond and Korku tribe s and agriculturists by occupation) and cattle population. The administration and management of PTR, Maharashtra comes under the state government. The forests outside the park are commercially managed by the state government (Nagpur Forest Division) and the Forest Development Corporation of
16 Maharashtra (FDCM) a public sector unit. Previously the focus of the commercial management was to replace natural mixed forests with monoculture plantations of commercially important species (e.g. teak). This changed dur ing the 1990s, with stakeholders participating in conservation efforts, such as the World Bank assisted Maharashtra Forestry Project (World Bank, 1991). Dissertation S tructure The dissertation consists of seven chapters with five of them structured as indi vidual publishable papers. This introductory chapter with outlines of research objectives is followed by chapter two (Paper 1) that situates this research in its wider context by synthesizing results from published quantitative or qualitative data on forest dynamics from 24 protected areas in 7 tiger ranging countries. Chapter three (Paper 2) integrates remote sensing, GIS and field observations to examine the effects of different management strategies on land cover changes in the study area. Chapter four (Paper 3) combines a cellular automataMarkov modeling approach and a counterfactual approach showing what may have happened in this particul ar landscape in the absence of the national level conservation interventions. Chapter five (Paper 4) explores the utility of 2 band Enhanced Vegetation Index (EVI2) in quantifying landscape composition and configuration. Chapter six (Paper 5) compares the performance of remote sensing indices and surface temperature in differentiating forest successions. Finally, chapter seven summarizes the findings of this study and highlights the contribution of this study to land change science in general and the environmental change in c entral India in particular.
17 Table 11. Research objectives and research questions Research o bjective Research q uestion Chapter RO1: To synthesize the current body of knowledge on forest dynamics in tiger landscapes across Asia and to situate the present study in its wider context What is the current status of knowledge on forest dynamics trends in and around tiger reserves/wildlife sanctuaries/national parks across Asia ? Chapter two RO2: To compare and contrast land cover changes in and outside PTR resulting from differential management strategies Does exclusionary management of PTR result in the conservation of forest cover and/or regeneration of forest cover? Chapter three Do the spatial and temporal land cover trajectories vary in areas of commercial extraction outside the park boundaries from those found within the park? Are the changes shown by land cover change trajectories also supported by the changes in NDVI values, reflecting vegetation productivity? RO3: To evaluate the conservation intervention in PTR by comparing a modeled scenario with the actual landscape What is the likely role of the conservation interventions in guiding land cover change and landscape fragmentation in and around PTR? Chapter four Can spatial model be combined with counterfactual approach to empirically evaluate conservation intervention? RO4: To explore the utility of satellite-derived vegetation index as an indicator of landscape heterogeneity as reflected in different land covers in Pench landscape Can EVI2 be used as a gradient surface to represent land cover diversity in a humanmodified tropical landscape? Chapter five How effective are the surface metrics, in terms of describing landscape composition and configuration? RO5: To assess the utility of satellite-derived thermal indices and vegetation indices in differentiating tropical forest successions How effective are the surface temperature, emissivity, and normalized tasseledcap indices in differentiating tropical forest successions? Chapter six
18 Figure 11. Location of the study area
19 CHAPTER 2 TRENDS OF FOREST DYNAMICS IN TIGER LANDSCAPES ACROSS ASIA Introduction The geographical range of the wild tigers has been greatly reduced over the last 50 years, with these large carnivores currently occupying a mere 7% of their historical range (Dinerstein et al., 2006). While tigers are still found in a wi de variety of forest types, including the coniferous deciduous forests of eastern Russia, the tall grass habitats south of the Himalayas, the swamps and mangrove forests of Sundarbans, the moist/dry deciduous forests of Central Indian Highlands and the tro pical rainforests of Sumatra and Malaysia, frequent reports of humantiger conflicts are not uncommon in these heavily humanaltered landscapes. One of the conservation strategies to save the wild tigers is to create tiger landscapes that are anchored on prey rich protected areas (PA), i.e. national park/tiger reserve/wildlife sanctuary, and are linked to greater ecological systems through habitat corridors (Karanth and Stith, 1999). However, in humandominated landscapes this strategy requires direct or indirect involvement of the local inhabitants to be effective. PAs are often surrounded by competing land uses, and have been traditional sources of livelihoods for rural people. Hence, trends in forest dynamics in a protected landscape could be used to understand the prospects for sustainable humannature coexistence, which is a prerequisite for longterm and effective tiger conservation. Forest dynamics are often used to measure PA effectiveness (DeFries et al., 2005; NaughtonTreves et al., 2005; Hayes, 2 006; Nagendra, 2008). Avoiding forest cover loss cannot be the ultimate test for PA effectiveness, as biodiversity can be significantly compromised by other threats, such as subsistence hunting or commercial
20 hunting/poaching ( Redford, 1992), in an otherwis e ecologically sustainable PA landscape. Nonetheless, intact or regenerated forest cover is a useful proxy for conservation success, probably because forest dynamics can be linked to many factors relevant to conservation evaluations. Forest composition, quality, and structure provide a fair representation of wild herbivore distribution as they show preferences for certain forest types (Kuijper et al., 2009; Millington et al., 2010). Additionally, forest degradation through livestock grazing suggests communi ty dependence on the forest products, thus indicating lack of alternative means for sustainable livelihood. Intense resource competition between grazing livestock and wild herbivores might result in shrinking prey base (Madhusudan, 2004), which can have deleterious effects on tiger population (Karanth and Stith, 1999). It is also possible to measure social sustainability of a conservation effort through documenting park people relationships, which is likely to reflect in the anthropogenic forest disturbance (e.g. destruction of forest, deliberate fire) of the region. Forest cover, along with substantial prey occurrence and water, are the primary requirements for supporting viable tiger population and the prey base (Sunquist and Sunquist, 2002). While tigers can thrive in a wide variety of ecosystems, they have been reported to exhibit strong habitat preferences, at least for some of their activities. For example, studies have shown tiger preferences for better covered Mangrove woodlands and less disturbed grasslands in Sundarbans, Bangladesh (Khan and Chivers, 2007), less open moist deciduous forest in Nagarahole, India (Karanth and Sunquist, 2000), less degraded or primary forest in Indonesia (Linkie et al., 2008), and mixed deciduous open forest and mixed tall grasslands in Nepal (Smith et al., 1998). Evidently, tigers
21 seem to prefer undisturbed, less degraded forested habitat for better cover and higher prey density. However, tigers have also shown high adaptability in selectively logged forests of Malaysia (Rayan and Mohamad, 2009) and are expected to benefit from forest type conversion in Russia (Cushman and Wallin, 2000). Nonetheless, lack of habitat corridors between isolated forest patches (or between two PAs ), such as human settlements, extensive agricultural fields, and other land uses breaks the continuity of genetic exchanges between two isolated populations, eventually causing biodiversity loss ( Heywood, 1995). Most importantly, forested habitats are essential in order to support prey abundance and distribution, since prey depletion is often considered the most important factor driving decline of tigers throughout their range ( Karanth and Stith, 1999). One of the reasons behind the extinction of the Javan tiger is the conversion of the forested habitats of the island to teak during the 1900s, resulting in rapid declination of the prey base with eventual extinction of Javan tigers (Seidensticker 1987). Hence, information on current status of forest cover dynamics is an important component not only in designing movement corridors and mixeduse transition zones surrounding PAs but also to take timely decision to protect forest dependent prey species. Identifying trends in forest dynamics requires time series analysis of spatial patterns of forest cover Remote sensing provides particularly useful tool in landscapelevel spatial analysis by providing satellite snapshots at multiple spatial and temporal scales (DeFries et al., 2005; Nagendra, 2008). Satellite image analysis is probably the most widely used technique in global level studies analyzing transformations in the biophysical and ecological attributes of the earths surface (DeFries et al., 2005; Nepstad et al., 2006). However, such analyses are sometimes considered to be less
22 capable of capturing degradation (i.e. changes within a single land cover), which also is an important indicator of forest dynamics. Fieldbased ecologic al study can be particularly useful in such cases in identif ying changes in forest structure, composition and diversity ( Kumar and Shahabuddin, 2005; Yadav and Gupta, 2006; Mehta et al., 2008). Such studies are expensive and timeconsuming compared to remote sensing studies; however, these methods can complement each other to provide perspectives on both forest conversion and modification. Much attention has been given at national and international levels to establish a common platform for tiger conservation strategies in the 13 tiger ranging countries (Global Tiger Initiative 2010). These 13 countries, namely Bangladesh, Bhutan, Cambodia, China, India, Indonesia, Laos, Malaysia, Myanmar, Nepal, Russia, Thailand, and Vietnam, represent a wide spectrum of socioecological factors that impact landscape dynamics. In or der to propose feasible conservation strategies, the socioecological factors governing the spatial landscape patterns must be understood and taken into account One way to achieve this goal is by looking at landscape dynamics and drivers especially within and outside PAs Here I synthesize results from published quantitative or qualitative data on forest dynamics from 24 PAs in 7 tiger ranging countries (Fig. 2 1). The trends of forest dynamics were identified as the temporal changes in forest covers/com position/density, mostly within the entire landscape in which the PA is embedded, with occasional exclusion of neighboring landscapes. Additional data from these 7 countries were used to understand the nature of the institutional regime and the proximate f actors governing the landscape dynamics. This study uses both landscapelevel and species level published data on forest dynamics in
23 tiger landscapes to identify the dominant trend and most discussed drivers of change in these landscapes, which could further be used to locate priority landscapes for habitat restoration. I expect to see a wide spectrum of forest cover dynamics, some showing severe deforestation, with some with reforesting trend. This expectation is consistent with the wide variety in this database in terms of population density, location of the PA, and socioeconomic structure of the society. Methods Two academic databases Web of Science and Google Scholar were used to locate peer reviewed articles containing quantitative information about forest cover within and around PA s (national parks, tiger reserves or wildlife sanctuaries) with reported tiger populations in 13 tiger countries. I did not include PAs which, even though located within these countries, do not host wild tiger population. I considered the articles for analysis only if they have used remote sensing techniques or fieldbased ecological methods to report spatial or temporal changes in habitat quality or forest cover in general. Most of the remote sensing studies have reported forest cover change over decades, sometimes before and after protection. Some of the studies, however, only provide single date measurement s and/or maps of forest cover, and have used either satellite images or various fieldbased methods. These studies w ere included in this analysis only if they provided important information on spatial pattern of forest cover/composition/density, such as within or outside PA boundaries. Information was primarily selected from 27 journal articles from 7 countries Bangladesh, China, India, Indonesia, Myanmar, Nepal, and Russia. While considerable amount of research has been conducted in rest of the tiger ranging countries, most of them report tiger
24 population density (commonly using cameratraps), and do not provide infor mation on changes in forest cover or habitat, hence these were not included in this study. Results Summa ry Characteristics of the CaseS tudies A total of 24 case studies were found that met all the criteria for inclusion in this meta analysis. A m ajority of the case studies are from India (~54 %), with 3 each from China and Indonesia (~12.5%), 2 from Nepal (~ 8 %), and 1 each from Bangladesh, Myanmar and Russia. Since socio economic factors often vary from one country to the other, there would be some bias in the findings due to the over representation of the Indian case studies. However, this also reflects the great number of tiger reserves in India, and emphasizes the need for such research in the tiger ranging countries that are not represented in t his dataset. The size of the study area ranges from less than 500 km2 to over 2000 km2 (Fig. 2 2a), with the largest covering about 7000 km2. The institutional management regime for most of the cases is centralized, with PAs under government management, except for China and Nepal. In China, collective forests total about 60% of the total forest area, and provide legal resource extraction and management rights to the communities, in spite of being government controlled (Miao and West 2004). In Nepal, communities are partially involved in buffer zone forests in some PAs (Nagendra et al., 2005). In India, some of the study areas participated in ecodevelopment projects (Fig. 2 2b) which were expected to be participatory efforts to reduce extraction and graz ing pressure from the core as well as the surrounding landscape by reducing forest dependence for fuelwood (through supply of biogas) and controlled grazing on marked pasture. These efforts had mixed results in meeting social or ecological goals, and are considered by
25 many scholars to have been a failure to introduce a participatory management regime in in habited landscapes (Shahabuddin, 2010). The number of human settlements within the core of the PA varied from 0 (Bandipur, Corbett in India; SikhoteA lin in Russia) to 27 (Sariska in India) (Fig. 2 2c). Almost all the Indian PAs host forest villages/settlements within the core. However, many Indian PAs have either relocated most of the forest villages or are in the process of relocation (e.g. Bhadra and Pench in India ). The number of villages/settlements in the surrounding landscape (mostly within 10 km), however, is much higher (Fig. 2 2d). Most of the case stud ies have over 100 villages within 10 km from the park boundary, while only one has less than 5 villages in the vicinity of the PA (Namdapha in India). Spectrum of Forest Dynamics: Deforestation to Reforestation A major portion of the case studies (4 6 %) reported forest dynamics based on multi date remote sensing method, while 33 % report forest cover status using fieldbased ecological methods. Only 21 % of the studies employed a singledate remote sensing approach. As expected, these regions experienced very different trajector ies of forest change, from degradation and deforestation to isolation of the PAs through changing land uses in the surroundings, to even reforestation in some landscapes (Fig. 2 3). A detailed description of the forest dynamics is provided in Table 2 1. India exhibits all of the forest cover trajectories mentioned above. W hile degradation is prevalent in Indian PAs ( Kumar and Shahabuddin, 2005; Nath et al., 2005; Karanth et al., 2006; Giriraj et al., 2008; Mehta et al., 2008), deforestation is often either dominant in the surroundings (Nagendra et al., 2006; Imam et al., 2009) or is followed by reforestation as a consequence of village relocation (Imam et al., 2009; DeFries et al., 2010) or change in degree of protection ( Mondal and Southworth, 2010).
26 What is alarming here is that some PAs experienced deforestation within the core itself, in spite of the highly protected status (Sarma et al., 2008; Mondal and Southworth, 2010). On a positive note, PAs with maintained or regenerated forest cover are not uncommon (Singh et al., 2009; DeFries et al., 2010). However, PAs with den se to moderate forest cover in the core can be isolated if surrounded by other land uses ( DeFries et al., 2010), thus resulting in restricted animal movement. The Bangladesh case study suggests forest disturbance generated by both natural and anthropogeni c factors (Giri et al., 2007). Soil erosion poses a major threat in the estuarine ecosystem of the Sundarbans, which is further aggravated by agricultural and aquacultural conversion of forested land. The study covers both the sides of Sundarbans (i.e. in India and Bangladesh) and suggests forest degradation while the net area of mangrove forests over the study period remained within the error margin. All of the Chinese and Indonesian case studies suggest severe deforestation, both within and outside the PA s ( Coggins 2000; Kinnaird et al., 2003; Linkie et al., 2004; Gaveau et al., 2007). The case stud y from Myanmar exhibits deforestation both within and outside the PA, with higher rates outside, followed by reforestation (Songer et al., 2009). Interestingly the degree of protection did not appear to contribute to the forest dynamics in these case studies, as protected and nonprotected forests experienced similar trajectories of forest change. On the other hand, the role that the nature of management regime plays is quite evident in the forest dynamics of Nepal. Studies have reported reforestation in previously degraded forests in Nepal, which is often linked to institutional decentralization (Nagendra et al., 2005; Panta et al., 2008). The Russian case study suggests varied degrees of disturbance in nonprotected areas,
27 with higher disturbance in easily accessible forests (Cushman and Wallin, 2000). However, the reported change in forest type (conversion to hardwood forest) is thought to be beneficial for habitat generalist species such as Siberian tiger. Proximate Causes of Changes in Forested Landscapes: Country L evel Issues Given the biophysical, socioeconomic and institutional differences in the case studies, a wide range in the proximate drivers of fores t change that shape these landscapes was expected. A ll the factors cited as probable causes behind the changes can be grouped under 7 broad categories (Fig. 2 4), consistent with other global level studies of drivers of land use/land cover change (Lambin e t al., 2001; Geist and Lambin, 2002). Table 2 2 provides a comprehensive list of proximate drivers of forest change grouped by countries. The two most cited drivers of forest change in the Indian case studies were fuelwood/fodder extraction and grazing ( K umar and Shahabuddin, 2005; Nath et al., 2005; Karanth et al., 2006; Davidar et al., 2007; Mehta et al., 2008; Imam et al., 2009; DeFries et al., 2010). This is consistent with the Indian scenario where most of the PAs are surrounded by forest villages, and people are dependent on forest products for subsistent living and cattleraising. Since Indian PAs are exclusionary in terms of institutional regime, forest villages are often relocated outside the PA boundary, thus exerting more pressure on the surrounding landscape (Nagendra et al., 2006) Illegal/commercial timber collection through selective logging is also not uncommon (Imam et al., 2009; Mondal and Southworth, 2010), sometimes even from the core forest. Agricultural conversion, including a shift tow ards tea/coffee plantations have also been cited as a common cause for clear felling (Giriraj et al., 2008). Post independence India experienced rapid infrastructural growth, as a result of which considerable
28 amounts of forested regions were cleared and l ater submerged for hydroelectric projects (Bhat et al., 2001). Hence, controlled flooding, along with natural floods, has been mentioned as a common cause of deforestation in many Indian PAs (Nath et al., 2005; Sarma et al., 2008; Mondal and Southworth, 2 010). Wildfire and road development leading to higher tourism pressure and improved accessibility to local markets are among the other causes that have been identified in the Indian case studies (Nagendra et al., 2006; DeFries et al., 2010). Even though cited in most of the Indian case studies, fuelwood/fodder collection was not mentioned as a proximate cause of forest cover change in other countries, except Nepal (Nagendra et al., 2005). The other cause cited for Nepal was logging (Panta et al., 2008), w hich is also prevalent in Indonesia and Russia ( Cushman and Wallin 2000; Linkie et al., 2004). Wildfire is another common contributing factor of deforestation and/or degradation in Indonesia and Russia ( Cushman and Wallin, 2000). However, the most important driving factor in Indonesia is probably agricultural conversion; as even the core forests are under direct threat of agricultural encroachments, especially along the boundaries (Kinnaird et al., 2003; Gaveau et al., 2007). Agricultural expansion and fl ooding were the only t wo causes mentioned in the case study from Myanmar (Songer et al., 2009). These two factors were also held responsible for mangrove forest degradation in Bangladesh (Giri et al., 2007), along with salinity change and shrimp farming. T he three PAs from China included in this study mentioned plantation, especially bamboo forestry, as the major driver of change in forest composition within and outside the PAs (Coggins 2000). However, it is difficult to reach any conclusion on the factors responsible for forest cover change from the very
29 few case studies from Bangladesh, China, Myanmar, Nepal, and Russia. The list of the factors from these countries is likely to expand with inclusion of more case studies. Discussion Forest Dynamics in Individual Countries: Trends, Challenges and Implications for Tiger Conservation Bangladesh The mangrove swamps of the Sundarbans, Bangladesh provide the last remaining refuge to the Royal Bengal Tiger. This mangrove ecosystem, shared between Bangladesh (60% ) and India (40%), is the worlds largest coastal wetland. Although this dynamic ecosystem frequently suffers from massive soil erosion due to its geographic location, the resulting forest loss is often offset by the accreted land and new mangrove plantati ons ( Giri et al., 2007). In addition to the natural disturbances, mangrove forests of Bangladesh have a long history of human exploitations as well ( Seidensticker et al., 1991). A vast proportion of mangrove forests was cleared and converted to agriculture during Turk and British colonization. Later, in spite of having a PA network, clear felling and timber extraction were allowed within the protected and reserved forests, leading to major mangrove forest loss in three decades since independence in 1947 ( Ri chards 1990). Todays densely populated Sundarbans is threatened not only by agriculture, but also rapidly growing aquaculture. Local people depend on forest products for subsistence living, and exert additional pressure in the mangrove forests by harvest ing fish and shrimp/prawn larvae ( Gopal and Chauhan, 2006). Besides forest degradation, human activities (such as forest clearance and conversion to other land uses) also directly influence fresh water flow and sediment accretion in this ecosystem. Such
30 ch anges in salinity can have deleterious effects on vegetation types, herbivorous species, and the predators ( Hussain and Acharya, 1994). To minimize forest loss the Government of Bangladesh adopted its second forest policy in 1994 in order to expand the PA network, to design and implement multiple usages of Sundarbans, and to involve local people through afforestation of encroached forests and other activities ( GoB 1994). However, none of the existing forest policy identified the importance of buffer zones surrounding the PAs, which are often embedded within an already degraded reserve forest ( Sharma et al., 2008). While multiple commitments have been included in an amendment of the Bangladesh Wildlife (Preservation) Order to better protect/manage the forests and the species, challenges remain in terms of taking trans boundary initiatives with India for Sundarbans management, limiting allowable resource extraction from the PAs, providing economically viable alternatives to shrimp farming, and implementi ng effective participatory management ( Iftekhar 2006; Chowdhury et al., 2009). China China is distinguished from other countries included in this study by its formalized land tenure system that started during collectivization and communization, and resulted in collective forests. The management regime of collective forests is complex and h as changed frequently (Miao and West, 2004). During the post revolution period (post 1949), China witnessed changes in the forest institutional regime, with many of its private, communal and stateowned forested land changing to elementary cooperativemanaged forest in order to promote sustainable forestry (Liu, 2001). However, further amalgamation of cooperative lands into communes and associated corruption, along with the government promoted iron and steel production program are believed to have
31 result ed in large scale deforestation (Liu, 2001). The post revolution era also witnessed a rapid increase in PAs with over 2000 nationally protected nature reserves by 2004 (Xu and Melick, 2007). P ark people conflict has increased consequently, as traditional forest management practices are affected by conservation restrictions. It is often argued that the PAs will be more effective in conserving flagship species under traditional community management, rather than constantly changing, and confusing insti tutional re gimes (Xu and Melick, 2007). However, the small size of many PAs, increasing PA isolation as a result of fragmentation, and highlevel of human disturbance within the core forest are often held responsible for decline or complete absence of tiger and its p rey species in South China tiger reserves (Tilson et al., 2004). India India, hosting onethird of the wild tiger population is a major player in the global tiger conservation scenario. With Project Tiger launched in 1973, Indian government assigned maximum protection to the tiger reserves in the country. Many national parks were promoted to the tiger reserve status. However, conservation practices in India continued to be a reflection of the colonial legacy of commercial extraction for a long time even aft er the independence (post 1947), resulting in alarming forest loss, especially in regions with limited protection (Bhat et al., 2001). In addition, other factors including frequent relocation of forest villages outside the PAs, national level emphasis on agricultural expansion, and industrial activities (such as road network development, hydroelectric projects) contributed to post independence forest degradation (Bhat et al., 2001; Nagendra et al., 2006; Mondal and Southworth, 2010). Fortunately, various r emedial measures have been implemented nationwide to restrict deforestation and degradation including complete ban on tree felling within any national park, and
32 implementation of plantation programs in the surrounding areas. Resource collection, grazing and hunting activities are strictly banned in the core reserves, while these activities are often restricted in the surroundings as well. The World Bank supported India Eco development Project (19962004) aimed at reducing forest dependence among communities through livelihood support activities (World Bank 1996). This ambitious project was not primarily a community based project. However, one of the main objectives of this project was to increase opportunities for local participation in PA management. Unfortunately, it was never implemented as a participatory project and, given these flaws in implementation, many conservationists are skeptical about its impact on reducing anthropogenic pressure on the PA landscapes (Shahabuddin, 2010). As a result, many of the reserves are becoming increasingly isolated because of rapidly changing land uses in the surroundings (DeFries et al., 2010). With negligible participation from the forest fringe communities in the park management and little benefit received by the communities, humantiger conflict is increasingly obvious (Madhusudan, 2003; Ogra and Badola, 2008). Indonesia The link between Asian financial crisis and massive deforestation in biodiversity rich and ecologically complex ecosystems of Indonesia is highly contested and is often political (Sunderlin, 1999; Robertson and van Schaik 2001). However, this rapid deforestation is more alarming as Indonesia has an extensive system of PAs, detailed land use plans for commercial logging, and abundant donor assistance (Kinnaird et al., 2003). Multiple studies have documented legal/illegal logging, agricultural conversion by new settlers arriving through Indonesias transmigration program, development of oil palm and pulpwood plantations as driving factors for Indones ias massive deforestation
33 (Barr, 2001; Robertson and van Schaik 2001; Holmes 2002). Since tigers show a tendency to move away from forest edges to avoid disturbances, such as hunting and other human activities (Griffiths and van Schaik 1993), only a fr action of the core forest is available to these wideranging species. As a result, humantiger conflict is very high in disturbed areas where the probability of humantiger interaction is maximum (Nyhus and Tilson, 2004). Declaration of more PAs that have suitable habitat and prey base and that can serve as corridors between isolated forested patches, such as Batang Gadis National Park (Wibisono et al., 2009), can have positive impacts on tiger conservation. However, to preserve the dwindling Sumatran tiger population, the extensive agroforestry matrix surrounding the PAs must be wisely used (Nyhus and Tilson 2004). Myanmar Myanmars complex political history has shaped changes in protected landscapes (Aung et al., 2004). Post independence (post 1948) Myan mar had other priorities than wildlife protection, focusing on economic and institutional changes Post war social unrest aggravated hunting, with no provision for wildlife habitat protection in the 1954 amendment of the Burma Wildlife Protection Act (Gutter, 2001). Following the Asian economic crisis in the mid 1980s and increasing pressure for natural resources from the neighboring countries the sever ity of the threat to Myanmars forest and wildlife became apparent and conservation strategies began to emerge in national policies. However, tiger populations continued to decline as a result of hunting, prey depletion, habitat fragmentation, inadequate PA coverage, lack of effective management, and the flourishing trade of tiger parts in global market (Lynam et al., 2006). Myanmars first tiger reserve, also the worlds largest (Hukaung valley) was recently proposed to be
34 created as part of a national tiger conservation strategy (Lynam et al., 2009). While extensive surveys have confirmed tiger population i n extensive and continuous forested tracks of farther north and farther south of Myanmar, it would be particularly challenging to establish a sound conservation plan, especially stopping tiger habitat loss in trans border areas and cross border trafficking of tiger parts (Lynam et al., 2006). Nepal Tigers used to range across most of lowland Nepal until settlement programs in the 1960s converted most of this region to agricultural land (Gurung, 1983). Further decline in the tiger population was prevented through several government initiated programs, such as creation of a lowland park system, community forestry in the PA surrounding landscapes, and the Terai Arc Landscape (TAL) project (Gurung e t al., 2008). These projects together contributed significantly to improve habitat quality not only within the PA but also in the neighboring landscape, to reduce anthropogenic pressure on forest resources through benefit sharing with the local communities and to link distant lowland forests that can act as breeding habitat. As a result, the Terai arc landscape is now considered to be one of the best tiger habitats in the world (Barlow et al., 2009), which underscores the need to reestablish the dispersal corridors across this landscape (Wikramanayake et al., 2004). However, during the Maoist insurgency (20002007), conservation efforts were severely affected both in national parks and community managed forests, resulting in a rapid increase in wildlife (es pecially rhino) poaching in and around Chitwan and Bardia national parks (Martin et al., 2008). While no study to my knowledge has identified the political instability as a direct factor affecting tiger conservation, Baral et al., (2006) have documented the adverse effects of
35 such political turmoil on the wildlife population, habitat, and overall conservation efforts in Nepal. Russia Unlike other forested areas in temperate zone, the Russian Far East (RFE) was able to maintain its primary forests until the last decades of the last century, since when political and economic changes led to urbanization and industrialization of this area (Cushman and Wallin, 2000; Kondrashov, 2004). Although wildfire shapes most of the landscapelevel changes in this landscape, increasing road density is rapidly coming to constitute a severe threat to the tiger population through expanding tourism in PAs and increasing hunting, logging in unprotected areas (Kerley et al., 2002; Kondrashov, 2004). Kerley et al., (2002) found t hat road access severely hinders PAs from functioning as source populations of tigers, and indicate that it would be difficult to sustain a viable tiger population with increased human access in unprotected areas. Additional pressure from local communities for forest products will only further complicate the situation. While some efforts have been made to engage indigenous communit ies in participatory management in PA surrounding landscapes (Shmatkov and Brigham, 2003), such efforts should focus more in areas with suitable tiger habitats. Forest Dynamics in Tiger Landscapes: Common Needs for Tiger Conservation PAs are established to protect biodiversity; specifically in tiger landscapes the first priority is, and should be, to provide adequate protection to this endangered charismatic species. However, irrespective of countries, conservation goals are often hindered by conflicting institutional regimes and management strategies, pressing need for alternate livelihood for forest dependent communities, lack of infrastructure for effective PA management, and humanwildlife conflict (Table 2 2). A ll countries that harbor viable in -
36 situ tiger populations have identified restoring habitats and establishing habitat corridors as one of the priority issues for tiger c onservation (Global Tiger Initiative, 2010). While maintaining the core forest is the primary objective of any conservation effort, the surrounding landscape should be given similar attention in order to maintain a greater ecosystem (DeFries et al., 2007). An intact or minimally disturbed greater ecosystem is a prerequisite to host a thriving prey base, without which it is almost impossible to support a viable tiger population (Karanth and Stith, 1999). As identified by Laurance (1999), it is not enough to conserve isolated and fragmented PAs; the quality of the peripheral forest matrix must also be protected and regenerated in order to decrease the risk of mortality for wide ranging mammals. Unfortunately, demands for maintaining such a landscape oft en severely underestimat es the regular needs of the local communities who live in and around tiger ranging PAs It has long been identified that development needs of the local communities that share the ecosystem with tiger and its prey species must be addressed to further conservation goals (MacKinnon et al., 1999). However, from the case studies discussed here, these issues are evidently not being addressed sufficiently According to this meta analysis, tiger reserves are often subjected to agricultural, infrastructural and commercial developments (Table 2 2), which are reflected in deforestation/degradation in the surrounding landscape, if not in the core forest (Table 2 1). Further, almost none of the countries have a participatory management regime balancing the needs of the forest fringe communities and the conservation goals. Decentralized management has attracted much attention over the last decades, accused by conservationists of not being conservationcentered, and by critics of
37 coercive conserv ation as not being sufficiently development oriented. The balance between conservation and development is society specific and context specific and a complex task that is not easily achievable, requir ing sustained financial, social, institutional and political commitment at all levels from national to local Of the countries discussed here, only Nepal has a community forest and park buffer zone management program that incorporates community participation in the forest management, albeit with negligible p ower to introduce/modify rules of the government authority managing PAs. While India has experimented with initiatives towards involving local communities in ecodevelopment programs, these have been widely critiqued as being far less participatory in prac tice, and not having achieved their stated goals of providing local communities with a stake in conservation goals (Ostrom and Nagendra, 2006; Shahabuddin, 2010). Yet, much of the deforestation and forest fragmentation in and around tiger embedded PAs can be traced to human impact. There is a clear and imperative need to involve local forest dependent communities as partners in conservation initiatives on the ground, not just on paper. Conclusion This metaanalysis synthesized published quantitative data on forest dynamics from 24 tiger hosting protected areas from 7 countries. F indings indicate that a majority of the protected area landscapes are undergoing rapid deforestation, degradation and forest fragmentation, across all countries surveyed. While prox imate causes cited as impacting forest cover change include both natural and anthropogenic factors, the drivers of change are largely human. Many of these changes can be traced to a lack of participation of the forest fringe communities in forest management and conservation, which appears critical given that most remaining tiger habitats are in areas inhabited by
38 poor, rural, forest dependent communities. A critical and alarming finding of this study is the lack of information on forest dynamics in and around a majority of tiger reserves across Asia. In an era where remotely sensed information on forest cover change is easily available, and can be interpreted with reference to field data to understand changes in tiger habitat, it is disturbing to note that s uch information is not more widely available.
39 Table 21. Individual description of the casestudies Type of study Study Period Forest dynamics Source Bangladesh Sundarbans National Park Multi temporal remote sensing study 1977 1989 2000 Afforestation during the first half, followed by deforestation in the later half net change within margin of error. Degradation is common. Giri et al., 2007 China 3 Nature Reserves in Southeast Uplands: Meihuashan, Wuyishan, and Longxishan Field based ecological study 1994 1995 Deforestation and degradation as most of the community managed forests are being replaced by anthropogenic bamboo monoculture. Evidence of high-level human and cattle disturbance within the core of the Meihuashan reserve. Coggins 2000 India Bandipur National Park, Karnataka Field based ecological study 2005 2006 Differential degradation within the park itself. Evidence of change in species assemblages from herbaceous, large woody trees in less disturbed zones to small woody trees, more shrub species in more disturbed zones. Mehta et al., 2008 Bhadra Wildlife Sanctuary, Karnataka Field based ecological study 2002 Evidence of degradation close to villages. Degree of disturbance depends on size of the village, or in other words human and cattle population. Karanth et al., 2006 Chandoli National Park, Maharashtra Single-date remote sensing study 2005 High fragmentation within the park; severe deforestation outside the park. Evidence of regeneration, especially in grasslands, after relocation of forest villages. Imam et al., 2009
40 Table 21 continued Type of study Study Period Forest dynamics Source Corbett Tiger Reserve, Uttarakhand Single date remote sensing study 2005 2006 Dense to moderate forest cover for a major part of the reserve. Singh et al., 2009 Ka lakadMundanthurai Tiger Reserve, Tamil Nadu Multi -temporal remote sensing study 1973-1990-2004 Degradation in the evergreen forest; forest type is continuously being changed to the semi -evergreen forest and grassland, with most changes during pre-establishment period. Giriraj et al., 2008 Kanha Tiger Reserve, Madhya Pradesh Single date remote sensing study 2007 Considerable amount of continuous surrounding forest cover, especially in higher elevation. Core forest is in better condition than buffer forest. Evidence of reforestation after village relocations. Ravan et al., 2005; DeFries et al., 2010 Manas Tiger Reserve, Assam Multi temporal remote sensing study 1977 1998 2006 Deforestation within the park even after establishment; drastic reduction in alluvial grassland. Sarma et al., 2008 Nagarhole Tiger Reserve, Karnataka Single-date remote sensing study 2009 Good amount of intact forest cover within the reserve; evidence of forest corridors connecting the reserve with the surrounding. However, the reserve is embedded within a matrix of agricultural fields and tea/coffee plantation. Mahanty 2002; DeFries et al., 2010 Namdapha Tiger Reserve, Arunachal Pradesh Field based ecological study --Degradation especially in surrounding areas. Nath et al., 2005 Pench Tiger Reserve, Maharashtra Multi temporal remote sensing study 1977 1989 2000-2007 Deforestation followed by reforestation both in the park and the surroundings. Mondal and Southworth, 2010
41 Table 21 continued Type of study Study Period Forest dynamics Source Ranthambore Tiger Reserve, Rajasthan Single date remote sensing study 2008 Little forest cover outside the park boundary; concrete wall surrounding the park and agricultural mosaic isolate the park from the surroundings. DeFries et al., 2010 Sariska Tiger Reserve, Rajasthan Field based ecological study 2004 Degradation mainly changes in vegetation structure and composition in disturbed areas. Kumar and Shahabuddin, 2005; Yadav and Gupta, 2006; Shahabuddin and Kumar 2007 Tadoba-Andhari Tiger Reserve, Maharashtra Multi -temporal remote sensing study 1989-2001 Reforestation/maintenance in the reserve; primarily deforestation/degradation in the surroundings with reforestation at places. Nagendra et al., 2006 Indonesia Bukit Barisan Selatan National Park Multi temporal remote sensing study 1972 1985 2002 Deforestation both within the park and within 10 km from the park boundary with higher rate of deforestation in the buffer, however total area of forest is higher in the park. Evidence of reforestation in inactive encroachments within the park. Kinnaird et al., 2003; G aveau et al., 2007 Gunung Raya Wildlife Sanctuary Multi temporal remote sensing study 1972 1985 2002 Severe deforestation within and outside the sanctuary with high fragmentation. Gaveau et al., 2007 Tapan valley containing a part of Kerinci Seblat National Park Multi -temporal remote sensing study 1985-1992-1999 Deforestation Linkie et al., 2004
42 Table 21 continued Type of study Study Period Forest dynamics Source Myanmar Chatthin Wildlife Sanctuary Multi temporal remote sensing study 1973 1989 1992-20012005 Deforestation both within and outside the sanctuary till 2001, with higher rates outside. Evidence of reforestation during 2001-2005 both within and outside the sanctuary. Songer et al., 2009 Nepal Chitwan district covering Royal Chitwan National Park Multi temporal remote sensing study 1976 1989 2001 Reforestation in buffer zone forests and deforestation within community forests. Varying deforestation rate depending on forest types. Deteriorating forest condition outside the park. Nagendra et al., 2004, 2005; Panta et al., 2008 Royal Bardia National Park Field based ecological study 2005 Degradation in forests closer to villages. Human disturbance impacted forest vegetation structure and diversity, specifically species richness, tree density, diversity, evenness and basal area of trees. Evidence of lack of alternative resource collection areas and liveli hood need for the communities leading to more disturbances. Thapa and Chapman 2010 Russia Sikhote Alin range, covering a part of Sikhote-alinsly MAB Biosphere Reserve Multi temporal remote sensing study 1972 1992 Deforestation rate varied depending on the forest type with higher disturbance in accessible forests. Cushman and Wallin 2000; Kerley et al., 2002 Source includes both primary and supplementary articles used in this study
43 Table 2 2. Country wise summary of proximate causes driving changes in forest structure and composition mentioned in the casestudies, and priority issues for tiger conservation Country Proximate causes of deforestation/degradation Priority issues for tiger conservation Bangladesh Agricultural and aquacultural conversion, flood/soil erosion, illegal/legal logging Improving law enforcement, PA infrastructure/ management, capacity building for wildlife crime detection and habitat management, mitigating human-wildlife conflict, averting developmental threats, establishing transboundary cooperation, increasing public awareness China Bamboo plantations National level policy for wildlife protection, improving PA infrastructure and management, restricting land use changes and establishing corridor connectivity, limiting poaching, trafficking and trading tiger parts, establishing transboundary cooperation, increasing public awareness India Fuelwood/fodder extraction, grazing, agricultural conversion, tea/coffee plantation, illegal/legal logging, flood, fire, tourism/development Strengthening national level policy for wildlife protection, improving PA infrastructure and management, mitigating human-wildlife conflict, restoring habitats for tigers and pr ey, restricting land use changes and establishing corridor connectivity, averting developmental threats, increasing public awareness Indonesia Illegal/legal logging, agricultural conversion, coffee/pepper plantation, fire Strengthening national level poli cy for wildlife protection, improving wildlife crime control, improving PA infrastructure and management, mitigating human-wildlife conflict, establishing corridor connectivity, averting developmental threats, increasing public awareness Myanmar Agricultu ral conversion, plantation, flood Improving wildlife crime control, improving PA infrastructure and management, restricting land use changes and establishing corridor connectivity, establishing transboundary cooperation, averting developmental threats, inc reasing public awareness Nepal Fuelwood/fodder extraction, illegal/legal logging Strengthening national level policy for wildlife protection, improving PA infrastructure and management, mitigating human-wildlife conflict, restoring habitats for tigers and prey, strengthening community -based management, establishing corridor connectivity, establishing transboundary cooperation, increasing public awareness
44 Table 22 continued Country Proximate causes of deforestation/degradation Priority issues for tiger conservation Russia Illegal/legal logging, fire Establishing new PAs, eliminating gaps in institutional mandates, improving wildlife crime control, restoring prey -base, averting developmental threats, improving logging practices, encouraging community based management, substituting forest-based livelihoods, managing forest fires adapted from Global Tiger Initiative website
45 Figure 21. Spatial distribution of the case studies within tiger ranging countries
46 Figure 2-2. Summary characteristics of the case studies Figure 23. Forest dynamics and temporal span of the case studies
47 Figure 2-4. Country wise proportion of proximate causes driving forest dynamics
48 CHAPTER 3 PROTECTION VS. COMMERCIAL MANAGEMENT: SPATIAL AND TEMPORAL ANALYSIS OF LAND COVER CHANGES IN THE TROPICAL FORESTS OF CENTRAL INDIA Introduction Humandriven land cover change is considered as the single most important variable affecting ecological systems (Vitousek, 1994) and can significantl y change the amount, type and successional state of forests. These changes in forest ecosystems are considered to have global conservation importance since the effects of significant forest cover loss or within forest modification may include massive soil erosion (Sidle et al., 2006), destabilization of watersheds (Rai and Sharma, 1998), a loss of sustainable forest uses and threats to indigenous people on a regional scale (Samal et al., 2003) and large scale release of atmospheric carbon dioxide through the burning of fossil fuels contributing more to the greenhouse effect (Houghton, 1994). Another alarming aspect of this crisis is the threat to biological diversity (Shukla et al., 1990; Laurance, 1999). Even if the target of 1012% land protected and reser ved across the globe is achieved, up to 50% of tropical species are predicted to go extinct in the next few decades (Soule and Sanjayan, 1998). While the first priority for conservation is still increasing the number of protected areas globally, it is also important to address issues related to sustainable forest management outside protected areas (Putz et al., 2001). Parks are believed to protect forest from changes and are established to maintain carbon sinks, protect biodiversity, and to help stabilize global climate (Munroe et al., 2007). Many protected areas in tropical forests have been reported to be successful in slowing the rates of forest loss (Bruner et al., 2001). However, current controversies show skepticism in the effectiveness of parks as management regimes, which is further
49 complicated by the evaluation of the regional impact of parks over time (Ostrom and Nagendra, 2006). Protected areas or parks have witnessed a 500% increase in land over the past three decades (Wittemyer et al., 2008); how ever, most of them are under severe threat from encroachment, agriculture, ranching, urban development, illegal and legal logging, and collection of nontimber forest products (IUCN, 1999; WWF, 2004). Isolation of ecosystems due to the forest cover decline in the matrix surrounding protected areas is not uncommon in developing countries of Latin America, Asia and Africa (DeFries et al., 2005). Especially in India, where the parks frequently contain forest villages located within or outside the administrativ e boundary (Nagendra et al., 2006), it might be particularly challenging for the resource managers to maintain a park and its surrounding matrix as a whole ecosystem. Several studies have challenged the idea that conservation is best served by sustainabl e timber or forest management (Rice et al., 1997; Bowles et al., 1998) mainly based on the argument that conservation can only be served by strict protection (Bowles et al., 1998). However, with high cost for protection, pressures for profitable forest res ource extraction and human population growth, sustainable forest management is often the only chance left for maintaining forest cover and biodiversity (Whitmore, 1999). Since forest management practices in many ways depend on the countrys developmental s tage and politically determined economic incentives (GarciaFernandez et al., 2008), more smaller scale case study type approaches are needed to understand the effect of different management strategies on forest cover transitions. Tropical dry forests, even though more threatened and less protected than tropical moist forests (Murphy and Lugo, 1986; Janzen, 1988), are particularly under studied in India (Sagar and Singh, 2004). Dry forests have often been preferred by humans due to
50 their fertile soils (more suitable for agriculture) and more favorable climate for livestock (Murphy and Lugo, 1986). They have also been subjected to unsustainable forest resource extraction pressures (Bullock et al., 1995) to support high human population densities. Resource managers are often confronted with the daunting task of managing natural resources sustainably while meeting the needs of the human population. This can be a specific problem in developing countries (Gadgil, 1991; Rodgers, 1991), e.g. India, where approximately 50 million people depend directly on forests for their livelihoods (Hegde et al., 1996), and where substantial loss of forest area might result from continued grazing, fuelwood collection and nontimber forest product extraction (Kothari et al., 1989). More studies involving longterm monitoring of tropical dry ecosystems are thus required to develop effective forest management policies (Kumar and Shahabuddin, 2005). This study contributes to the land change science and conservation literature by examini ng the effects of different management strategies on land cover changes, within and outside Pench Tiger Reserve (PTR), Maharashtra in Central India, using a host of spatial data, GIS tools, and field observations. Among different forest management practices within India (such as commercial timber extraction, participatory management, and community based management) the present study deals with two opposing strategies strict protection and commercial extraction. While I expect to see different forest cover change trajectories resulting from different management practices, a balance should be maintained between forest covers within and outside the park to avoid isolation. PTR, Maharashtra was chosen for this study as it is one of the reserves in the Indian s ubcontinent which could potentially host a large tiger population (Ranganathan et al., 2008); thus maintaining the forest cover and ecological health of
51 the park and its surrounding matrix can contribute to the local chapter of the global tiger conservatio n effort. This study uses both discrete land cover maps and continuous Normalized Difference Vegetation Index (NDVI) data sets for a spatial and temporal land cover change analysis, specifically asking the following questions: Does exclusionary management of PTR, Maharashtra result in the conservation of forest cover and/or regeneration of forest cover? Do the spatial and temporal land cover trajectories vary in areas of commercial extraction outside the park boundaries from those found within the park? Are the changes shown by land cover change trajectories also supported by the changes in NDVI values, reflecting vegetation productivity? Indian Forestry Policies and P ractices Pre and post independence forest management practices in India have been mostly g uided by national level policies (Bhat et al., 2001). Forest categories such as Reserved Forest (RF) and Protected Forest (PF) were first introduced during British rule (Forest Protection Act, 1927), and were retained by the Government of India after i ndependence (post 1947). Presently, RF enjoys higher degree of protection than PF, since no hunting, and/or grazing are allowed in the former except under specific orders. Continued conservation efforts have led to pockets of RF (often upgraded to national parks and wildlife sanctuary) embedded within a PF matrix. Besides different management strategies, other factors including frequent relocation of forest villages outside protected areas, national level emphasis on agricultural expansion, and industrial a ctivities (such as road network development, hydroelectric projects) contributed to post independence forest degradation (Bhat et al., 2001; Nagendra et al., 2006).
52 The focus of colonial period policies was to extract commercial timber under several work ing circles (which is an area allocated to timber extraction under a working plan and which has a 30year extraction cycle), which is still being practiced in nonreserve forests. Post independence priority changes of the Indian government introduced needbased forest management where vast areas of forest were cleared and selectively cut (Pandey, 1992; Prasad, 2000) to supply timber for farm building construction, industry, and hydroelectric projects. However, continued forest loss set off an alarm and various plantation programs were implemented (1980 onwards) to increase forested areas, although most resulted in failure (Pandey, 1992). In 1988 the forest policy was revised (MoEF, 1988) with priority given to forest and biodiversity conservation rather than financial benefits from forests and acknowledging the need of community participation in the comanagement of degraded forest. As a result of this amendment, the post 1988 period witnessed a complete ban on tree felling within any national park contrary to the usual practice of controlled timber extraction through working circles within national parks. In the following two decades several other changes in government policies have continued to identify the importance of community involvement in conservation, such as the National Afforestation and Ecodevelopment Board (NAEB) set up in 1992, participatory management programs such as Joint Forest Management (JFM) and Village Eco development (VED), the World Bank funded India Ecodevelopment Project (19962001) and the Wildlife Amendment Act (2002). While it is particularly difficult to find one optimal management regime in the tropics (Pearce et al., 2003), comparative analyses of different regimes might be an effective tool to find regional solutions
53 Methods Satellite Imagery and GIS A multitemporal and multispectral dataset were used for this study (Table 3 1). While the sensors offer different spatial and spectral resolutions, such multispectral datasets are often unavoidable in studies spanning over several decades and have been successfully applied in other regions (Zoran and Anderson, 2006; Ahmed et al., 2009). The images used for this study are from the post monsoon season (October to January) to minimize the cloud cover and phenological variations. The park boundary along with a 10 km buffer (which is required to be treated as an ecosensitive zone under the Environment Protection Act, 1986 and thus is required to receive strict protection) was used for this study. Multiple zones were then cre ated starting from the park boundary for every 2 km up to the required buffer limit of 10 km, and henceforth would be mentioned as zone 1 through 5 for convenience (Fig. 3 1 ). I used this 2km exclusive buffer zone approach to understand the role of distance from the park boundary in the land cover changes. Since this study involves comparisons of land cover changes of a park and its surrounding matrix, and is not a comparison of two parks, I excluded PTR, Madhya Pradesh from the buffer zones. Image Preprocessing Image preprocessing was carried out in ERDAS Imagine 9.3 and ENVI 4.3 software packages. The ASTER images, being Level 1B data, were generated by applying coefficients for radiometric calibration and geometric rectification (Abrams et al., 2002). A mosaic of the two ASTER images was created to cover the study area and was then reprojected to the UTM WGS84 coordinate system (Zone 44 North) with a spatial resampling of 30 m. Geometric rectification was carried out on all the Landsat
54 images using the mosaicked ASTER as a basemap and nearest neighbor resampling algorithm, with root mean square (RMS) error of less than 0.5 pixels (<15 m) via imageto image registration. Radiometric calibration and atmospheric correction were carried out to correct for s ensor drift, differences due to variations in the solar angle, and atmospheric effects (Green et al., 2005). Image Classification A hybrid approach was used to classify the images. First, an unsupervised classification method with ISODATA clustering was us ed to generate preliminary classes. Then similar classes were merged based on spectral signatures and training samples collected during May June 2008. Finally, I defined three land cover classes forest (both natural forests and plantations), bare soil (agricultural land and cleared forest) and water (reservoir, small ponds, and rivers) based on distinct spectral signatures and used those signatures for supervised classification with the minimum distance to means classification algorithm. Accuracy testing was performed using 143 training samples. The overall classification accuracy was over 94% with a kappa value of 0.89. Each image was classified following the same method, rather than applying signatures from one date back in time. Change D etection Post classification change detection is a widely used pixel based change detection method (Jensen et al., 1995), where two (or more) classified images are compared using a change detection matrix. For this study, twodate change trajectory images (19771989, 19892000, and 20002007) were generated for the park and for each of the buffers (Fig. 3 2). Also a four date change trajectory image (Fig. 3 3) was generated
55 to capture the overall land cover changes since 1977 (19771989 20002007). In order to maintain the spatial compatibility, the classified MSS image (originally of 60 m resolution) was artificially down scaled to 30 m before performing the change trajectory analysis. The overall change trajectory image from four dates, with three landcover classes fo r each date resulted in 81 possible change trajectories. Since interpretation of these 81 trajectories can be confusing, I further collapsed these trajectories into eight categories (Table 3 2 ), partly following the classification scheme suggested by Mertens and Lambin (2000). NDVI Standard Normal Deviates and Image Differencing NDVI was used in this study to provide estimation for vegetation productivity or biomass (Gamon et al., 1995; Sims et al., 2006). Due to the usage of a multi sensor dataset and known correlation of precipitation and NDVI, I standardized NDVI values across dates using the following formula: (NDVIi NDVI)/NDVI, (3 1) where NDVIi is the NDVI value of a particular pixel in a particular year, NDVI and NDVI are the mean NDVI value and standard deviation respectively for that year. These NDVI standard normal deviates record NDVI variations across an image with respect to the mean. In other words, higher NDVI standard normal deviate values show regions with higher vegetation productivi ty, while lower NDVI standard normal deviate values show nonforest regions. Following the standardizations, twodate image differencing was performed to identify the changes in NDVI standard normal deviate values over time. Three stable forest masks cre ated from each twodate change trajectory were used to extract within-
5 6 class ecological changes in the stable forest class. Fig. 3 4 shows differences between NDVI standard normal deviate values for only those areas which were forest on all dates and so in a traditional classification scheme are considered areas of no change Results Spatial and Temporal Land Cover Change Trajectories The twodate change trajectories reveal a distinct spatiotemporal pattern of forest cover changes (Fig. 3 5). Classes that converted from nonforest to forest were grouped as reforestation and those converted from forest to nonforest were grouped as deforestation. As outlined in Fig. 3 5, the park suffered notable deforestation before 1989, which became negligible in later years. While the proportion of stable forest consistently increased throughout the study period, post 1989 years witnessed greater reforestation than deforestation, both within and outside the park (Fig. 3 5). A detailed description of t he twodate change trajectories is provided in Table 33 While the twodate change trajectories show decadal changes in forest covers, a four date change trajectory draws an overall change scenario experienced by the study area in the last 30 years (Fig. 3 6). In addition to stable forest, deforestation and reforestation classes, four date change trajectory provides information on unique forest transition dynamics, namely old forest clearing with regrowth and forest regrowth with new clearing (Table 3 2), which mainly represent rotational forests showing various forest transition pathways. A detailed description of the four date change trajectory can be found in Table 34
57 Temporal Changes in Vegetation P roductivity Fig. 3 4 captures subtle variat ions in NDVI standard normal deviate values, thus in productivity or vegetation health, within socalled stable land cover classes (forest on all dates according to traditional classification). The stableforest NDVI standard normal deviate subtraction i mage for 19771989 (Fig. 3 4a) suggests similar or lower vegetation productivity in the later year, leading to lower NDVI standard normal deviate values, for most parts of the study area (differences ranging from 0 to 2). During 1989 2000, vegetation productivity seemed to increase (Fig. 3 4b), especially within the park resulting in higher NDVI standard normal deviate values in 2000 (differences ranging from 0.01 to 1). According to Fig. 3 4c, vegetation productivity in 2007 either remained same or decreased when compared to that from 2000, resulting in lower NDVI standard normal deviate values in 2007, especially within the park. Discussion Preceded by a global trend of tropical deforestation in the 1980s, many tropical countries witnessed a decline in deforestation rates (FAO, 1993, 2006) with a select few even exhibiting net reforestation (Lugo and Helmer, 2004; Rudel, 2005; ArroyoMora et al., 2005). With considerable research advances in recent years, many socioeconomic, demographic, and biotic facto rs have been identified promoting forest regeneration (de Jong et al., 2001; Perz and Skole, 2003). In the Indian subcontinent in particular, the expansion of forest plantations (in itself a result of earlier forest scarcity stimulating demand for forest products) have led to reforestation (Salam et al., 2000; Lamb and Gilmour, 2003). Properly planned and conducted logging is an integral part of sustainable forest management (Putz et al., 2001) and should not lead to forest cover loss when accompanied by replantation or natural regeneration (Kauppi et al., 2006).
58 Unfortunately, commercial logging in tropical co untries often represents timber mining activity without any consideration of the renewability of natural resources (Putz et al., 2000). The present study provides a direct comparison of the effects of strict protection and commercial extraction on forest and landcover changes and thus contributes to the forest management literature. The study area exemplifies a typical case of contrasting management regimes. This condition enabled me to examine the role of different management approaches in preserving forest cover. While I expected to see maintained and/or regenerated forest cover within the park as a result of protection, I also expected greater for est loss over time outside the park, as suggested by many casestudies in the tropics (see DeFries et al., 2005). Part of the findings supports other studies showing the effectiveness of parks in maintaining forest cover (Bruner et al., 2001), while this s tudy also supports the argument that commercial extraction can lead to sustainable forestry (Whitmore, 1999), at least within the time period of this study. Indian forestry has undergone several changes from the 1980 onwards, as discussed earlier The findings of the present study have captured the dynamics of those government policies well. While the protected status of the park became effective from 1975, it still continued to suffer notable forest loss until 1989. Especially, in the southwestern boundary of the park, forested areas were cleared during 1977 1989 under scientific management (source: PTR management plans and personal communication with the office of the Field Director, PTR), as reflected in Fig. 3 2. Although the boundary of the park is under continued pressure from the surrounding villages (mainly grazing), this region of the park showed a forest gain in the 19892000
59 change trajectory. This pre1989 forest loss and post 1989 forest cover recovery within the national park could be attributed to the 1988 notification by which felling became banned in national parks, exhibiting the importance of national level policies in favor of conservation. The twodate change trajectory of 19771989 also captured the disturbances created during t he Totladoh hydro electric dam construction in the northern part of the park. A considerable amount of forested area ( ~ 20 km2, source: PTR management plan) was submerged contributing to the forest loss during this time period. Uncontrolled forest resource extraction near the construction site also contributed to the already deteriorating forest conditions. Resettlement and relocation issues also played their part in changing forest cover and quality. At the time of notification of the park, there were two f orest villages inside PTR, namely Bodalzira and Fulzari Bodalzira was submerged during the construction of the Totladoh dam and the population was resettled in Fulzari with compensation. Fulzari is still located within the park ( Fig. 3 1 ), the relocation procedure of which is in progress. The dominance of nonforest land covers in the northwestern and southeastern boundaries of the park relates to the presence of Fulzari and other villages in vicinity of the park which are dependent on the forest for cat tle grazing and other minor forest products. While two date change trajectories show decadal changes in forest covers as a result of changing forest policies, a four date change trajectory draws an overall change scenario experienced by the study area in t he last 30 years. The stable forest class within the park is notably greater than that in any of the buffer zones (Fig. 3 6). With an exception in zone 2, stable forest decreases with increasing distance from the park. The stable bare soil class mainly den otes agricultural land in this area, which is much lower
60 within the park than the buffers (Fig. 3 6). The stable water class has negligible percentage (0.31.4%; refer to Table 3 3 for details), related to the temporal changes of water body locations. This might be attributed partly to the presence of the reservoir in the later images and partly to the precipitation variability resulting in more water in some images. The effect of the dam construction is also evident in the deforestation class, especially within the park (Fig. 3 3). While deforestation is considerably lower in the buffer zones, reforestation trends are quite interesting. The park gained 15.3% forest cover during the study period, which is understandable due to its protected status. However, the buffer zones which are commercially managed also show notable reforestation (2035.2%; see supplementary material for details). It is also interesting to note that reforestation increases with increasing distance from the park boundary (Fig. 3 6). This study supports several other studies reporting reforestation trends (Perz and Skole, 2003; ArroyoMora et al., 2005). However, these regenerated forests are often not ecologically equivalent to the original forest (Lugo and Helmer, 2004). The present study does not compare species level forest quality between dates; instead it uses NDVI standard normal deviate as a proxy of vegetation health and productivity. NDVI standard normal deviate image differences for the stable forest class show a decrease in vegetation productivity (Fig. 3 4) in later years in spite of an increase in forest cover suggested by discrete classification. This degradation could potentially be due to several factors, e.g. grazing, changes in tree species richness, stand density, and canopy cover as a result of plantations (Pelkey et al., 2000; Gillespie et al., 2009). While my study design does not allow me to identify the causal factors of this degradation, this
61 difference in forest quality is important, and will hopefully lead to more detailed ecological studies in this region. The differences in forest coverage in and around PTR, Maharashtra most likely reflect the difference in degree of protection. Particularly, the amount of stable forest is considerably lower right outside the park (Fig. 3 5), suggesting over exploitation and consequent degradation of the matrix. Tracts of PF outside the park are mainly managed for commercial logging and are huge sources of revenue for both the government and private sectors. These forests are also subject to severe pressure from the forest dependent communities for cattle grazing and fuelwood collection. All of these factors might collectively contribute to the significantly lower stable forest coverage surrounding the park. However, an increasing r ate of reforestation along with a decreasing rate of deforestation in the matrix is encouraging for management practices other than strict protection, particularly sustainable resource extraction. Yet, we must be cautious in giving credit solely to the com mercial management practices, since a share can potentially also be attributed to, at least for the present case study, several other changes in Indian forestry as discussed below. The India Ecodevelopment Project (19962001), as mentioned earlier aimed at involving forest dependent communities in sustainable forest management. Among all the villages that fall within the 10 km buffer around PTR, Maharashtra only two villages, namely Sillari and Tuyapar from Maharashtra and several other villages f rom Madhya Pradesh were targeted under the Ecodevelopment Project. This new participatory management regime i s expected to reduce extraction and grazing pressure on the protected area as well as the surrounding buffer by reducing forest dependence for fuelwood (through supply of biogas) and controlled grazing on marked pasture. The
62 overall increase in forest cover, especially forest gain in the northwestern, northeastern and southeastern parts of the 10 km buffer, shown in the 20002007 change trajector y speaks for the importance of these policy changes in forest regeneration (Fig. 3 2). Yet, researchers have continually been skeptical about the degree of true participation of the communities and intention of forest officials and public sectors to engage the local people in conservation efforts ( Woodman, 2004; Nayak and Berkes, 2008). Nevertheless, the concept of involving local people in conserving forest is appealing and with true implementation holds the promise of successful and potentially sustainabl e forest management. Maharashtra forestry project, as mentioned earlier, was implemented with a loan from the state government and approval from the World Bank for certain models of plantation for different agroclimatic zones and species, with benefit sharing by local communities (World Bank, 1991; FDCM website). Such changes in forest management policies are well reflected in the change trajectories showing forest cover increase in the buffers over time. However, the lower percentage of stable forest right outside the park, i.e. in zone 1 (Fig. 3 5) suggests that further efforts must be made, from both government and private sectors, to engage park forest fringe communities in forest conservation without compromising their regular needs. Alternative liveli hood options should be created through participatory management practices such as Eco development Projects. A demand for such projects in the villages has also been identified (source: Management Plan of PTR, Maharashtra). It can be hoped, therefore, that with increasing awareness for conservation and reduced forest dependence through alternate livelihood options, ecological restoration of the surrounding matrix is possible.
63 In a landscape such as my present study area that supports scattered agriculturists or large nonpopulated areas of agriculture it might not always be possible or desirable to declare a large tract of forest as a protected area. Forest resource management in such a landscape involves complex dynamics of resource economy, rural livelihoo ds, and conservation initiatives. While it is challenging to find a balance, stakeholders managing the surrounding matrix play a major role in limiting a park s isolation and facilitating connectivity through effective forest management. Conclusion This study investigates the patterns and processes of landand forest cover change trajectories in and around PTR, Maharashtra. It targets a landscape with different management regimes, which is representative of numerous similar scenarios in developing c ountries, as well as elsewhere in India. R emote sensing and GIS techniques were used to map and quantify land cover conversions. S tandardized NDVI was also used to detect temporal changes in vegetation productivity. India, as with many other countries in A sia, has witnessed forest regrowth patterns in last two decades. In the presence of many management regimes it might be particularly difficult to identify a single optimal strategy. While many argue for strict protection to conserve biodiversity, proponents for sustainable commercial extraction are no less vocal. This study identifies the importance of protection as a measure of forest conservation, and also acknowledges the capability of achieving conservation goals through other participatory management practices. In the face of increasing human population growth and consequent need for resource extraction (both for personal and commercial profit) in developing countries, this research leads to a recommendation to maintain a balance between both strict protection and sustainable
64 extraction. T hat this study hopefully will provide insight to local forest resource managers interested in developing ecologically sustainable forest management strategies that can be applied to other tropical settings as well.
65 Table 3 1. Description of satellite imagery used in the study Sensor Type Acquisition date Path/Row Data source Landsat MSS January 1977 155/45 Global Land Cover Facility (GLCF) website hosted by University of Maryland Landsat TM November 1989 144/45 Global Land Cover Facility (GLCF) website hosted by University of Maryland Landsat ETM+ November 2000 144/45 Global Land Cover Facility (GLCF) website hosted by University of Maryland ASTER October 2007 144/45 Land Processes Distributed Active Archive Center (LP DAAC) located at USGS Earth Resources Observation and Science (EROS) Center Table 3 2. Change trajectory classes created for the four date change analysis Land cover categories Description Stable forest Forest on all four dates Stable bare soil Bare soil on all four dates Stable water Water on all four dates Reforestation Non -forest in 1977 and forest in 2007 irrespective of 1989 and 2000 Deforestation Forest in 1977 and non forest in 2007 irrespective of 1989 and 2000 Old forest clearing with regrowth Forest in 1977 and 2007, but non-forest in 1989 or 2000 or both Forest regrowth with new clearing Non -forest in 1977 and 2007, but forest in 1989 or 2000 or both Stable non-forest Non forest on all four dates, but changed from bare soil to water or vice versa Non forest = bare soil or water
66 Table 33. Two date land cover change trajectories for park and buffers with area in km2 and proportion of total area affected by each trajectory (%) Land cover categories Park Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 (a) 19771989 Stable forest 149 .7 (60.0 %) 51.4 (37.8%) 62.9 (43.4%) 63.1 (37.8%) 64.9 (34.8%) 72.1 (32.9%) Stable bare soil 19.5 (7.8%) 45.7 (33.5%) 44.5 (30.7%) 56.6 (33.9%) 74.2 (39.8%) 94.5 (43.2%) Stable water 1.9 (< 1 .0 %) 1.6 (1.2%) 1.0 (<1 .0 %) 1.2 (<1 .0 %) 2.7 (1.5%) 0.7 (<1 .0 %) Forest to bare soil 41.1 (16.5%) 21.4 (15.7%) 14.9 (10.3%) 21.2 (12.7%) 16.6 (8.9%) 19.4 (8.9%) Forest to water 5.6 (2.2%) 0.1 (<1 .0 %) -0.1 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) Bare soil to forest 9.9 (4.0%) 9.3 (6.8%) 13.9 (9.6%) 18.8 (11.3%) 21.4 (11.4%) 25.9 (11.8%) Bare soil to water 3.2 (1.3%) 0.1 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) 0.2 (<1 .0 %) 0.2 (<1 .0 %) Water to forest 16.4 (6.6%) 4.1 (3.0%) 4.8 (3.3%) 4.1 (2.4%) 4.5 (2.4%) 4.6 (2.1%) Water to bare soil 2.3 (<1 .0 %) 2.5 (1.9%) 2.9 (2.0%) 1.9 (1.2%) 2.0 (1.1%) 1.5 (<1 .0 %) (b) 1989 2000 Stable forest 166.4 (66.7%) 56.6 (41.5%) 72.1 (49.8%) 72.4 (43.3%) 76.1 (40.8%) 84.4 (38.7%) Stable bare soil 22.6 (9.1%) 46.7 (34.3%) 41.0 (28.3%) 55.2 (33.0%) 66.7 (35.7%) 89.0 (40.8%) Stable water 8.4 (3.4%) 1.7 (1.2%) 1.0 (<1 .0 %) 1.3 (<1 .0 %) 2.9 (1.6%) 0.9 (<1 .0 %) Forest to bare soil 8.3 (3.3%) 7.7 (5.7%) 9.1 (6.3%) 13.1 (7.9%) 14.0 (7.5%) 17.5 (8.0%) Forest to water 0.9 (<1 .0 %) 0.7 (<1 .0 %) 0.5 (<1 .0 %) 0.5 (<1 .0 %) 0.6 (<1 .0 %) 0.4 (<1 .0 %) Bare soil to forest 39.6 (15.9%) 21.6 (15.9%) 20.1 (13.9%) 23.6 (14.1%) 23.6 (12.6%) 25.5 (11.7%) Bare soil to water 0.7 (<1 .0 %) 1.3 (<1 .0 %) 0.9 (<1 .0 %) 0.9 (<1 .0 %) 2.7 (1.4 %) 0.5 (<1 .0 %) Water to forest 0.3 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) Water to bare soil 2.2 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) 0.1 (<1 .0 %) (c) 2000 2007 Stable forest 203.0 (81.4%) 70.5 (51.8%) 81.4 (56.2%) 87.7 (52.5%) 90.9 (48.7%) 101.1 (46.2%) Stable bare soil 7.1 (2.9%) 29.8 (21.8%) 26.3 (18.2%) 27.7 (16.6%) 32.9 (17.6%) 42.9 (19.6%) Stable water 9.1 (3.7%) 3.3 (2.4%) 1.7 (1.2%) 2.2 (1.3%) 5.6 (3.0%) 1.3 (<1 .0 %) Forest to bare soil 2.2 (<1 .0 %) 7.1 (5.2%) 10.4 (7.2%) 7.8 (4.7%) 8.0 (4.3%) 8.5 (3.9%)
67 Table 33.c ontinued Land cover categories Park Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Forest to water 1.2 (<1 .0 %) 0.6 (<1 .0 %) 0.4 (<1 .0 %) 0.6 (<1 .0 %) 0.9 (<1 .0 %) 0.6 (<1 .0 %) Bare soil to forest 14.8 (5.9%) 23.9 (17.6%) 23.2 (16.0%) 39.3 (23.5%) 46.1 (24.7%) 62.7 (28.6%) Bare soil to water 11.2 (4.5%) 0.8 (<1 .0 %) 0.7 (<1 .0 %) 1.4 (<1 .0 %) 1.8 (<1 .0 %) 1.2 (<1 .0 %) Water to forest 0.7 (<1 .0 %) 0.1 (<1 .0 %) 0.2 (<1 .0 %) 0.2 (<1 .0 %) 0.3 (<1 .0 %) 0.2 (<1 .0 %) Water to bare soil 0.1 (<1 .0 %) 0.2 (<1 .0 %) 0.5 (<1 .0 %) 0.3 (<1 .0 %) 0.3 (<1 .0 %) 0.3 (<1 .0 %) Zone 1 = 02 km, total area = 136.2 km2; Zone 2 = 24 km, total area = 144.8 km2; Zone 3 = 46 km, total area = 167.1 km2; Zone 4 = 6 8 km, total area = 186.7 km2; Zone 5 = 810 km; total area = 218.9 km2. Distances of the buffers are calculated from the park boundary.
68 Table 34. Four date land cover change trajectories for park and surrounding buffers with area in km2 and proportion of total area affected by each trajectory (%) Land cov er categories Park Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Stable forest 141.8 (56.9%) 46.7 (34.3%) 57.1 (39.5%) 55.5 (33.2%) 57.2 (30.6%) 63.0 (28.9%) Stable bare soil 4.8 (1.9%) 24.0 (17.6%) 21.3 (14.7%) 22.7 (13.6%) 28.1 (15.1%) 37.9 (17.4%) Stable water 1.4 (0.6%) 1.5 (1.1%) 1.0 (0.7%) 1.1 (0.7%) 2.7 (1.4%) 0.6 (0.3%) Reforestation 38.1 (15.3%) 27.2 (20%) 30.7 (21.2%) 47.3 (28.3%) 60.7 (32.5%) 76.8 (35.2%) Deforestation 15.5 (6.2%) 5.6 (4.1%) 3.6 (2.5%) 4.6 (2.7%) 5.1 (2.8%) 4.8 (2.2%) Old forest clearing with regrowth 38.5 (15.5%) 20.6 (15.1%) 16.9 (11.7%) 24.3 (14.6%) 19.3 (10.4%) 23.6 (10.8%) Forest regrowth with new clearing 3.1 (1.2%) 8.4 (6.1%) 12.1 (8.4%) 9.4 (5.6%) 9.3 (5%) 9.7 (4.4%) Stable non-forest 6.1 (2.4%) 2.2 (1.6%) 2.1 (1.5%) 2.1 (1.3%) 4.3 (2.3%) 1.8 (0.8%)
69 Figure 31. Study area showing the boundary of the Pench Tiger Reserve, Maharashtra and the buffer zones created for this study (from the park boundary up to 10 km), and locations of the visited villages.
70 Figure 32. Series of t wo date change trajectories of the park and 10 km buffer (a) 19771989, (b) 19892000, and (c) 20002007.
71 Figure 33. Four date change trajectory (197719892000 2007) of the park and 10 km buffer.
72 Figure 34. Series of t wo date NDVI standard normal deviate image subtrac tions (color coded) showing within class changes in NDVI standard normal deviate values for stable forest class (a) 19771989, (b) 19892000, and (c) 2000 2007.
73 Figure 35. Two date change trajectories of forest cover within the park and 10 km buffer showing changes in forest cover transitions related to distance from the park boundary. Figure 36. Spatial and temporal changes in land cover dynamics within the park and 10 km buffer in las t three decades (1977 2007). Only classes showing notable changes (>3%) are included.
74 CHAPTER 4 EVALUATION OF CONSERVATION INTERVENTIONS USING A CELLULAR AUTOMATAMARKOV MODEL Introduction Global environmental change has many interacting components, land use/land cover change probably representing the single most important factor affecting ecological systems (Vitousek, 1994). Changes in forest ecosystems particularly, are of global conservat ion importance since deforestation or degradation can have several ecological impacts, including massive soil erosion (Sidle et al., 2006), destabilization of watersheds (Rai and Sharma, 1998), large scale release of atmospheric carbon dioxide through the burning of fossil fuels contributing to the greenhouse effect (Houghton, 1994), and decline of biodiversity (Shukla et al., 1990; Laurance, 1999). Protected areas or parks are established globally to conserve ecosystems, maintain carbon sinks, protect biodiversity, and to help stabilize global climate (Munroe et al., 2007). However, current controversies show skepticism in the effectiveness of parks as management regimes, which is further complicated by the evaluation of the regional impact of parks over t ime (Ostrom and Nagendra, 2006) and a lack of well designed empirical studies to estimate the effectiveness of conservation intervention (Ferraro and Pattanayak, 2006). Positive conclusions on protected area effectiveness have been derived based on improv ed ecosystem conditions inside the protected areas compared to outside ( Sanchez Azofeifa et al., 2003) or perceptions of success by protected area managers (Bruner et al., 2001). However protected areas are often established in regions with minimum ecolog ical perturbations, and hence would most likely experience little degradation over time even without the protection (Green and Sussman, 1990). Proper evaluation of conservation initiatives, therefore, should estimate what protected areas
75 would look like wi thout the policy/law enforcement (Ferraro and Pattanayak, 2006). Conservation practitioners and policymakers increasingly recognize the importance of empirical evaluations of their interventions (Balmford et al., 2005). As suggested by Ferraro and Pattanay ak (2006), a counterfactual approach asking a fundamental question what would have happened had there been no intervention? could be effective in this matter. A counterfactual scenario is little different from a predicted scenario, since for the latt er it is possible to observe the scenario at the predicted time step and validate the model, but for the former the scenario is nonexisting and thus it is not possible to directly compare the modeled scenario with the actual one. A conventional counterfac tual approach generally compares treatment and control units which are affected and unaffected by a certain policy intervention respectively, since it is not possible in reality to know what would have happened without a certain policy. However in the real world such units are often unavailable because of other confounding factors simultaneously affecting one or the other unit. In this paper, I proposed a cellular automataMarkov Chain model (CA Markov) as an alternative to simulate a counterfactual scenario. A Markov chain model is commonly used to quantify transition probabilities of multiple land cover categories from discrete time steps. These probabilities are then used with a CA model to predict spatially explicit changes over a certain period of time. The footstone of a CA Markov model is an initial distribution and a transition matrix, which assumes that the drivers that produce the detectable patterns of land cover categories will continue to act in the future as they had been in the past (Bri assoulis, 2000). This very assumption makes a CA Markov model suitable for a counterfactual approach, since I was interested to
76 extrapolate the preintervention landscape into the future assuming no change in the form of intervention. A combination of rem ote sensing and GIS has been utilized effectively both for the CAMarkov modeling approach (Peterson et al., 2009) and the counterfactual approach (Gaveau et al., 2009). The present study integrates these two approaches using remote sensing and GIS techniques. The availability of preand post intervention satellite images makes this tool particularly useful for the present study. Integrating these approaches would be ideal for study areas from developing countries, such as India, where land use land cover change around protected areas often reflect a complex interaction between law enforcement, local livelihood and different forest management activities, thus making it difficult to adapt a conventional counterfactual approach. India provides an ideal example of a country under increasing demographic pressure and economic growth. It is also a country with mixed conservation legacy, resulting from almost 200 years of British rule. The colonial legacy of centralized forest management is clearly demonstrated by the m ajor post independence national level acts/policies (Table 4 1) which have been identified as playing a pivotal role in Indian forestry and conservation (Kulkarni, 1987; Ghate, 1992; Bhat et al., 2001). However, there have been several attempts made by the Indian Government to protect forests for wildlife and local communities, rather than using them just for revenue generation, as evidenced by the forest acts/policies from the 1970s and 1980s (Table 4 1). Indian forestry, as a result, is often guided by complex interactions between national level bureaucracies, statelevel interpretations of national level policies, local livelihoods, conservation efforts of nongovernmental organizations, and commercial interests of
77 private sector units and state gov ernments (Woodman, 2004; Barve et al., 2005; Karanth et al., 2006; Robbins et al., 2006; Robbins et al., 2007; Mondal and Southworth, 2010). While the number of protected areas is growing fast, it covers less than 5% of the total land area in India (National Wildlife Database, 2009). Beyond the protected area boundaries, the landscapes are typically comanaged by several institutions and are often subjected to park focused integrated conservation and rural development efforts (Woodman, 2004). State forestry projects and ecodevelopment projects, funded by both national and international agencies, are among such efforts that promote forest regeneration, ecological restoration and ecodevelopment activities, especially in degraded forests surrounding protected areas. It is therefore difficult to choose control and treatment units, since control units, even though apparently unaffected by the protected area establishment, are subjected to other, simultaneous, conservation efforts. Like most of the protect ed areas in India, PTR wa s subjected to various national level policies/acts throughout the study period (19772007). PTR was declared a national park by the state government in 1975 and later a tiger reserve by the central government in 1999. While many conservation efforts (both national and international) govern the landscape dynamics in and around PTR (Mondal and Southworth, 2010), the Revised Forest Policy in 1988 (an amendment to the National Forest Policy, 1980) has been identified by the academics (Ghate, 1992) and park managers (the Office of the Field Director, personal communication; PTR management plans) as probably the most important one for the study area. This policy was notably different from its predecessors as it banned tree felling in any national park in India and also acknowledged community
78 involvement in protection and regeneration of forests surrounding protected areas. As suggested by the park managers (the Office of the Field Director, personal communication) this policy increased the degree of protection within PTR, resulting in increased forest coverage (Mondal and Southworth, 2010). I chose the 1988 Forest Policy as the conservation intervention in my study as I w as interested to see the changing patterns of forest transitions wi thin the park in the absence of this intervention. Adapting the conventional control treatment approach is not feasible for this study, since the implementation times of the Revised Forest Policy (1988) and afforestation and plantation strategies in the surroundings (early 1990s) overlap. While this overlapping makes it difficult to choose ecologically similar areas from the surroundings as control units for a counterfactual approach, it also provides a unique opportunity to generate a counterfactual s cenario for the buffer as well in absence of the afforestation and plantation strategies. Here I adapt a CA Markov modeling approach as an alternative to the conventional counterfactual approach to estimate the landscape level dynamics in absence of the c onservation initiatives both for the park and the surrounding landscape. The research questions addressed in this paper are: Can CAMarkov model be combined with counterfactual approach to empirically evaluate conservation intervention? What is the l ikely role of the conservation interventions in guiding ( a) land cover change and ( b) landscape fragmentation in and around protected areas? Specifically I used classified satellite images prior to the 1988 Forest Policy to predict the probability of changing forest covers and fragmentation patterns assuming no intervention afterwards and then compared the predicted results to actual scenarios as an effect of the intervention. I concluded with discussing the importance of national -
79 level policies in forest conservation in the landscape under study and the ability of the proposed method to evaluate conservation intervention in any other socioecological setting. Methods Land Cover M aps I used Landsat images from 1977, 1989 and 2000 and a mosaic of ASTER images from 2007 to generate land cover maps of the park and surrounding buffer areas up to 10 km (Fig. 41) The 1989 image was considered to represent pre1988 policy change scenario because of the lag effect of policy creation and implementations to its effects being seen on the ground. PTR, Madhya Pradesh (which is a part of the 10 km buffer) was excluded from the study, since it has a different degree of protection than the rest of the buffer. Classification accuracy of the most recent land cover map is approximately 94% and was determined using training samples collected during summer, 2008. Three land cover classes were defined in the final maps forest (both natural forests and plantations), bare soil (bare agricultural land and cleared forest) and water (reservoir, small ponds, and rivers). For this study, twodate change trajectory images for 19892000 and 19892007 were generated for the park and the buffer in order to compare them with the modeled predicted changes in the absence of the intervention. These land cover change trajectories show areas of stable forest, stable nonforest, deforestation and reforestation, across the time periods studied. CAMarkov M odel A Markov chain is a stochastic process that consists of a finite number of states of a system in discrete time steps and some known transition probabilities pij, where pij is the probability of that particular system moving from time step i to time step j. For
80 example, for a system composed of multiple land covers, the state of a particular cell at time step i denotes the type of land cover of that particular cell at time step i, which might change (or re main the same) in the next time step j. The transition probability denotes the probability of each class changing to every other class (or remaining the same) from time step i to time step j. With Markov chain analysis, future land cover can be modeled on the basis of the preceding state; that is, a matrix of actual transition probabilities between states can be used to predict future changes in the landscape from current patterns (Brown et al., 2000). Cellular automata is a dynamic and spatially explicit modeling approach that encompasses f ive components (a) a space composed of discrete cells, (b) a finite set of possible states associated to every cell, (c) a neighborhood of adjacent cells whose state influences the central cell, (d) uniform transition rules through time and space, and (e ) a discrete time step to which the system is updated (Wolfram, 1984). A c ombination of Markov and c ellular a utomata approaches has been shown to improve models describing complex natural patterns (Marshall and Randhir, 2008; Fan et al., 2008; Peterson et al., 2009). The modeled landscapes were generated using the software IDRISI, the Taiga version. The MARKOV module o f the software analyzes a pair of land cover images and outputs a transition probability matrix, a transition areas matrix, and a collection of conditional probability images. The transition probability matrix is a text file that records the probability of each land cover category changing to every other category. The transition areas matrix is a text file that records the number of pixels that are expected to change from each land cover type to each other land cover type over the specified
81 number of time units. The conditional probability images report the probability of each land cover type to be found at each pixel after the specified number of time units. These images are calculated as projections from the later of the two input land cover images. The CA _MARKOV module then utilizes the transition area matrix and a base image to model future landscapes. In addition, the CA_MARKOV module integrates the spatial information based on the conditional probability images, where the CA model changes a particular pixel from one land cover class in time step i to another class in time step j based on the state of the local neighborhood. In other words, the CA model utilizes a collection of the conditional probability images (also known as the suitability images) to start an iterative process of relocating the pixels to the proximity of the same land cover class until it meets the area predicted by the Markov model for each land cover class. For this study, I used land cover maps (spatial resolution: 30 m x 30 m) fro m 1977 and 1989 (Fig. 4 2a and 4 2b) as input maps in a CA Markov model to predict changes for 2000 and 2007 (Fig. 4 2e and 4 2f) with the assumption that the pre1989 drivers acting on the landscape are the only drivers guiding post 1989 changes. This ass umption obviously was not met in reality as tree felling was banned within the park after 1989 as a result of changes in national level policy and other conservation strategies were implemented in the surrounding areas during the 1990s. However, this assum ption is essential in my current counterfactual approach reconstructing an alternative scenario with an absence of the national level policy intervention. Both of the transition probabilities for 2000 and 2007 were calculated by the MARKOV module based on the transitions between 1977 and 1989. The CA_MARKOV module then used these transitions to predict spatial patterns for 2000 and 2007 using
82 the 1989 image as the baseimage. Then I calculated predicted land cover areas from the transition matrices generat ed for 2000 and 2007. I also created twodate change trajectories for forest transitions for comparison purposes. Since I was interested in forest transitions of the area (i.e. classes changing to or from forest class or stable forest class) only, I did not report the stable nonforest class in the change trajectories. Pixels estimated to convert from nonforest to forest were grouped as reforestation and those estimated to convert from forest to nonforest were grouped as deforestation. Pixels estimate d to remain as forest in both dates (i.e. in 1989 and 2000 for 2000 transition probabilities and in 1989 and 2007 for 2007 transition probabilities) were grouped as stable forest. I 2 test to compare the actual changes with the predicted changes, 2 values indicating higher levels of dissimilarity between compared land covers across time and space. Since it was not possible to directly validate the counterfactual scenarios, a land cover map for the year 2007 was predicted based on the maps from 1977 and 2000 for the validation purpose. Choosing the input maps from 1977 (well before the intervention) and 2000 (well after the intervention) enabled me to disregard the deforestation trend in the 1980s, and also to capture the current reforestation trend in the area in the predictive model. I tested this modeled landscape for accuracy using the training samples collected during summer 2008. I believed that this alternate modeled landscape would better represent the actual land cover if the assumption that the 1988 policy change is important in changing land cover trends in the study area, is in fact true.
83 Fragmentation I ndices The longterm impacts of policy intervention on forest fragmentation, along with forest cover, are also critical i ndicators of a parks success (Nagendra et al., 2006; Hartter and Southworth, 2009). Identifying trends in landscape fragmentation over time can provide us with insights on the conservation strategies impacting land cover change. I wanted to examine if lan d cover maps generated for the counterfactual scenario can be used to infer fragmentation trends. I ran multiple instances of the CA_MARKOV module with different numbers of CA iterations to generate a range of possible fragmentation scenarios. I generated 8 different scenarios resulting from the following iteration numbers: 5, 7, 9, 10, 13, 15, 20, and 25. Then I used FRAGSTATS 3.3 to calculate a set of landscape metrics for each of these scenarios to describe the spatial patterns of the forest class. The f ollowing landscape metrics (see McGarigal and Marks, 1995 for detailed description) were selected based on their capability to describe the spatial pattern without being redundant (Haines Young and Chopping, 1996). Number of patches: the total number of patches. Edge density: total edge per unit area for the class, in meters per hectare. Mean patch area: average area for the class, in hectares. Mean shape index: average complexity of patch shape compared to a square patch of identical area. For a single patch, the range varies from 1 when square to increasing numbers without limit with increasing complexity. Mean nearest neighbor distance: average nearest neighbor distance in meters. Interspersion juxtaposition index: measurement of the degree of interspersion of patches of a class with other categories. This index ranges between 0 and 100, with lower values for increasingly uneven distribution of patches of different classes. The above patchand class level indice s were used to compare descriptive metrics of forest cover pattern across the four observed dates. These metrics were also
84 used to compare actual patterns from the years 2000 and 2007 with those predicted for the same years. I obtained an estimated distribution of the predicted values for each of these indices for both 2000 and 2007, and then tested the hypothesis if the observed values fall within 95% confidence interval around the predicted mean values. Results Changes in Land Covers and Forest Transitions Forest covers as seen in actual changes and predicted changes, both within the park and in the surrounding buffer, are notably different in amount and show opposite trends (Fig. 4 2 and Fig. 4 3). Predicted forest cover for 19772007, based on the CA m odel, shows a deforestation trend for the park and an almost stable trend for the buffer (Fig. 4 3a). However, the actual changes show reforestation, both for the park and the buffer. Bare soil class shows exactly opposite trends. An increase in the soil c over was predicted for the park, while the buffer was predicted with almost no change. The actual changes show a declining trend for both (Fig. 4 3b). A 2 test (p=0.05) confirms statistically significant dissimilarity between actual and predicted changes across time and space. Actual and predicted forest transitions based on twodate change trajectory images (19892000 and 19892007) are shown in Fig. 4 4. The model predicts approximately 20% more deforestation in 2000, only within the park. For 2007, t he predicted deforestation is even higher, at 26%. Deforestation in the buffer was predicted to be 10% (2000) up to 16% (2007) higher than what actually occurred. Predicted reforestation within the park was approximately 4% lower than the actual reforestat ion, both in 2000 and 2007. Predicted reforestation in the buffer was approximately 3% higher for 2000 and 7% lower for 2007 than the actual reforestation. The modeled
85 landscape was predicted with 20% (2000) to 26% (2007) less stable forest within the park when compared to the actual landscape. The model also predicts 9% (2000) to 17% (2007) less stable forest within the buffer compared to the actual landscape. Changes in Fragmentation P atterns The overall descriptive changes in selected patchand class le vel metrics for the forest class differ considerably for actual and predicted scenario (Fig. 4 5 and Fig. 4 6). If I consider only the actual changes across four dates, the year 1989 shows the maximum fragmentation in terms of the landscape metrics, within the park in particular. Both for the park and the buffer, the number of forest patches was lowest in 1977, which reached to its highest in 1989, and is showing a declining trend afterwards. According to the patch area metric, forest patches were largest i n 1977, and the smallest in 1989, both for the park and the buffer. Both the increasing number and decreasing size of the forest patches indicate forest habitat loss and increased fragmentation during 19771989. The graphs for interspersionjuxtaposition index for the park and the buffer also indicate that forest patches were distributed more evenly in 1977 with respect to the other land cover classes, while the distribution was most uneven in 1989 representing highly fragmented landscape. As a result of t his fragmentation, mean distance between forest patches was lowest in 1989, especially in the park. Evidently, establishment of the park did not prevent the landscape from being more fragmented. Post intervention landscapes, however, seem to recover having larger forest patches with more uniform shape, decreasing edge density, and more uniform distribution with respect to the other land cover classes. Multiple predicted scenarios, as generated by running multiple CA_MARKOV modules, resulted in a range of predictions for each of the landscape metrics (Fig. 4 5
86 and Fig. 4 6). I tested whether the observed value for a particular index could have come from the corresponding estimated distribution (at 95% confidence interval).The tests reveal that most of the fr agmentation indices derived from the predicted land cover maps differ significantly from the actual patterns from the years 2000 and 2007, the predicted scenarios having larger and more complex patches with greater inter patch distance, with a few exceptions (Fig. 4 5 and Fig. 4 6). In other words, the post 1989 real landscape consists of smaller evenshaped patches those are closer to each other, while the simulated landscape consists of bigger patches with higher edge densities, which are more complex shapewise and farther away from each other. The predicted landscapes, both for the park and the buffer, also indicate towards an uneven distribution of forest patches with respect to the other land covers. Model V alidation The overall accuracy of the alternate predicted map for 2007, based on the transitions between 1977 and 2000 in an attempt to validate the methodology, is 80%. While the model performed quite well in predicting the overall trend, areas of mismatch between actual and predicted landscapes include areas with post 2000 forest regeneration (northeastern and southwestern corners of the buffer), and seasonally flooded areas surrounding the reservoir in the northern part of the park (Fig. 4 2). Since, the predict ion is based on transitions between 1977 and 2000, the model failed to accurately predict areas with post 2000 forest regeneration. Also, the model predicted the areas surrounding the reservoir within the park as bare soil, based on the 2000 image which was collected well after monsoon season. These areas are classified as water in the actual 2007 image which was acquired right after the monsoon season, preventing the model having greater accuracy. Overall however, this level of accuracy
87 reflects well o n the use of this modeling technique to evaluate conservation interventions in this landscape. Discussion India has a network of 660 protected areas that includes 99 national parks covering an area of 39,048 km2 (1.19% of the country) (National Wildlife Database, 2009). This network was created to help conserve a significant part of the countrys biodiversity. The basic approach of park management in India has been exclusionary based on the assumption that permanent human settlement within or in near vicini ty of the park degrades the ecosystem through resource extraction, which has been supported by several case studies (Barve et al., 2005; Karanth et al., 2006; Davidar et al., 2007). However, in the present case study, the park suffered from deforestation e ven after it was declared a national park, despite the fact that only one village is located within its boundaries. It was also suggested that the reason behind this forest cover loss within the park was not related to resource extracting communities, but rather is due to the regular forest management activities conducted by the state government (Mondal and Southworth, 2010). The counterfactual scenario, as simulated by the CA Markov model, is notably different from the actual one, both in terms of individ ual land covers (Fig. 4 2 and Fig. 4 3) and forest transitions (Fig. 4 4). The modeled scenario predicts 20% 26% more deforestation within the park itself than actually occurred. Clearly, if there had there been no intervention in 1988 in the form of the policy amendment, the park would probably have continued to suffer from extensive deforestation, challenging the very idea of national park creation for conservation. However, we must use these numbers with caution and rather focus on the pattern. The actual scenario, even in the absence
88 of the intervention, could have been something different than that predicted in this study. The rotational forests have a cycle of 30 years and it is likely that there could have been other spots of forest clearings, rather than aggregated cleared lands in the southwestern corner of the park as predicted (Fig. 4 2d). I attempted to capture the variability in the model predictions by generating multiple counterfactual scenarios with different CA iteration numbers, which migh t have an effect on the spatial patterns of the modeled landscapes (Fig. 4 5 and Fig. 4 6). As mentioned earlier all types of Indian forests used to be managed under several working circles with a 30year extraction cycle until the 1988 forest policy amendment (MoEF 1988). The findings of this study points out that in the absence of this amendment the park would have suffered continued deforestation, contrary to the purpose of its establishment. It could therefore be suggest ed that the intervention was indeed successful in limiting deforestation. What is interesting to note here is the intervention I chose was the revised policy in 1988 that banned tree felling within any national park, rather than the establishment of the park itself in 1975. This corr oborates my argument that mere establishment of a national park, without favorable national level policies and proper law enforcement, might not be effective from the conservation point of view. The surrounding buffer, despite being managed for commercial extraction shows an encouraging trend of reforestation. At a broader scale, this reforesting trend has been credited to several factors including degree of protection, commercial plantation, radical policy changes, and social forestry (Mather, 2007; Nagendra, 2009). Particularly, the implementation of various stateand World Bank funded forestry projects in the
89 1990s has been suggested to make the reforesting trend possible in my study area (Mondal and Southworth, 2010). my findings suggest that in the absence of these forestry projects achieving the reforesting trend might not have been possible. Hence it is important to have favorable forest management practices surrounding a national park in order to conserve a greater ecosystem and its biodiversity, ot herwise, increasing deforestation in the surroundings would result in an islandized park (Hartter and Southworth, 2009). While the predicted land cover maps suggest a positive outcome of the intervention in terms of forest conservation, the landscape metrics suggest something quite contrary to this. The predicted land cover maps show larger forest patches when compared to the actual maps, except for the year 2007 (Fig. 5 and Fig. 6). This might be an artifact of this approach, since the CA model predicti on is solely based on the input images and is a neighborhood based analysis. Since the forested area at the southwestern boundary of the park in the 1977 image was selectively cut between 1977 and 1989 and was classified as bare soil in the 1989 image, the CAMarkov model has assigned maximum bare soil pixels surrounding this area based on its higher probability. This resulted in similar land cover patches clumped together, which in turn resulted in larger patches. As already mentioned, the actual scenar io could have been different, even in the absence of the intervention, and so the fragmentation pattern could have been different. I have tried to capture this variability by reporting an estimated distribution for each of the metrics resulting from differ ent counterfactual scenarios (Fig. 4 5 and Fig. 4 6). Nevertheless, the positive influence of the intervention within the park is indirectly supported by the actual patterns (Fig. 4 5). The actual
90 patterns show increased landscape fragmentation till 1989, even after the establishment of the park that not only ceased after the policy enforcement in 1988 but also continued to recover with time. Similarly, the predicted patterns for the buffer (Fig. 4 6) suggest that without the stateand World Bank funded af forestation programs it would have been difficult to achieve the current reforesting trend necessary to maintain the greater ecological system. CAMarkov models are extensively used to identify and describe a myriad of processes including land cover change (Fan et al., 2008), rangeland dynamics (Li and Reynolds, 1997), species composition (Silvertown et al., 1992), forest succession (Alonso and Sole, 2000), and watershed response (Marshall and Randhir, 2008). Although a recent study has utilized CA Markov m odels to identify the role of different management eras of the former Soviet Union in changing forest cover patterns (Peterson et al., 2009), my study is unique in utilizing CA Markov models to generate a counterfactual scenario. Moreover, this simple appr oach is easy to use with remote sensing data and is capable of empirical evaluation of protected area management. This approach is particularly useful in areas where conventional methods of measuring protected area effectiveness (Andam et al., 2008; Gaveau et al., 2009) are difficult to adapt. Model validation suggests the predicted changes to be governed by the input maps, and in turn the policy implementation, rather than an artifact of the approach. Since the module used in this study requires two land cover maps as inputs, the method could potentially simulate counterfactual landscapes in any other scenario. However, future research on different ecological and social settings is needed to establish this as a robust method.
91 Conclusion I proposed an empirical tool that integrates counterfactual approach and CA Markov modeling to evaluate conservation intervention. I tested this method to understand landscape dynamics in a humandominated tropical landscape using a case study from central India. I explored the importance of national level policies in conservation efforts asking a fundamental question: what would have happened had there been no intervention? I utilized the proposed method to simulate alternate scenarios for 2000 and 2007 for the study area, assuming the absence of the 1988 forest policy intervention. Findings of this study suggest there would have been more deforestation, both within the park and in the surroundings, in absence of the intervention. This study also suggests t hat integrating CA Markov models and counterfactual approach can be effective in evaluating conservation initiatives and law enforcement. This approach can be applied to any socioecological settings, and is particularly useful in developing countries wher e landscape dynamics is often governed by a group of social, ecological, economical and political factors.
92 Table 4 1. Forest conservation acts/policies of India from 19471990. The list has been restricted to acts/policies that have been identified by academics and park managers to govern landscape dynamics in the study area. Forest Acts/Policies Salient Features National Forest Policy, 1952 An extension of the colonial British Policy that the claims of communities living in and around forests should not override national interests; in other words the policy justified destruction of forests for the construction of roads, building of irrigation and hydroelectricity projects Discouraged communities (mainly tribal) living in and around forests from using the forests Focused on obtaining more and more revenue from the forests Wildlife Protection Act, 1972 Provided increased degree of protection to wildlife and their habitat, as hunting and harvesting scheduled plant or animal species became banned. Helped increase protected areas (national park and wildlife sanctuary) throughout the country Had different provisions for communities claiming land rights within sanctuaries, including a) exclusion of such areas from the protected area boundary, and b) compensation and relocation outside the protected area boundary, with the consent of the land holders. However, national parks were given more protection with almost no boundary alteration and more restrictions on entry in the park Banned grazing and c ollecting forest products within national parks Forest Conservation Act, 1980 Conversion of government owned forested land to non forest areas, such as cultivation or other land uses was highly restricted. However, this provision excludes conversion necessary for conservation, development and management of forests and wildlife, suc h as establishment of check posts, fire lines, wireless communications and construction of fencing, bridges and culverts, dams, waterholes, trench marks, boundary marks, pipelines etc. Had provisions for tree felling under a scientifically prepared management plan within national park or sanctuary for improvement of wildlife and its habitat. Suggested that removed forest produce should be used to meet the needs of the people living in or around the park/sanctuary and not for commercial purposes. However, in case of large scale timber/nontimber removal disposal through sales is allowed on approval of the Central Government. Revised Forest Policy, 1988 an amendment of 1980s Forest Conservation Act Banned tree felling in any national park Identified t he tendency to consider forests as source of revenue to be the reason behind depleted forest resources Identified need for afforestation programs Suggested involvement of tribal communities in protection, regeneration and development of forests.
93 Figure 41. Map of the study area showing PTR and surrounding buffer up to 10 km, and visited villages in the area.
94 Figure 42. Land cover maps showing actual landscapes for (a) 1977, (b) 1989, (c) 2000, and (d) 2007, and CA Markov model generated landscapes for (e) 2000 and (f) 2007 based on (a) and (b).
95 Figure 43. Actual and predicted land covers for the park and the buffer showing changes in (a) forest and (b) bare soil from 1977 to 2007. Figure 44. Two date change trajectories for the park and the buffer showing actual vs. predicted forest transitions for (a) 1989 2000 and (b) 1989 2007.
96 Figure 45. Series of l andscape metrics for forest cover within the park The graphs show actual and predicted changes in (a) patch number, (b) mean patch area, (c) edge density, (d) mean nearest neighbor distance, (e) mean shape index, and (f) interspersionjuxtaposition index. Multiple CA Markov models with different numbers of CA iterations were run to predict a range of forest fragmentation scenarios. Mean values are plotted along with 1 standard deviation for each of the predicted metrics. Actual values that fall within 95% confidence interval around the mean predicted values are marked by asterisks.
97 Figure 46. Series of l andscape metrics for forest cover in the buffer. The graphs show actual and predicted changes in (a) patch number, (b) mean patch area, (c) edge density, (d) mean nearest neighbor distance, (e) mean shape index, and (f) interspersionjuxtaposition index. Multiple CA Markov models with different numbers of CA iterations were run to predict a range of forest fragmentation scenarios. Mean values are plotted along with 1 standard deviation for each of the predicted metrics. For some of the metrics, the variability is smaller than the symbols. Actual values that fall within 95% confidence interval around the mean predicted values are marked by asterisks.
98 CHAPTER 5 QUANTIFYING SURFACE GRADIENTS WITH A 2 BAND ENHANCED VEGETATION INDEX (EVI2) Introduction Since its inception as a discipline, landscape ecology has maintained a strong focus in quantifying spatial and temporal heterogeneity, in order to link landscape patterns to organisms, communities or ecological processes (Turner et al., 2001). Ecological patterns could be an important key to identifying the processes that shape a landscape. Especially, in human dominated landscapes, ecological patterns could result from a complex interaction between several social, ecological, economic and political factor s (Foster et al., 1998; Turner et al., 2001; Gardner and Engelhardt 2008; Walsh et al., 2008), and hence can be used to identify underlying processes. With the patchmosaic model as the operational paradigm, rapid advances have been made in quantitative l andscape ecology in the past few decades. Most of these quantitative methods require a discrete patchbased representation of the landscape. While these methods have been quite useful in quantifying landscape patterns, error propagation through subjectivit y and nonrealistic assumptions of discrete landscape structure can not be completely ruled out (McGarigal and Cushman, 2005). The patch mosaic model, also referred to as the patchcorridor matrix model (Forman and Godron, 1986; Forman, 1995), views the s urface as an aggregate of discrete patches and analyzes the surface properties through categorical representation of the landscape. The concept of the patch mosaic model is appealing to human intuition for its simplicity, consistency with welldeveloped and widely used quantitative techniques designed for discrete dataset and the availability of user friendly software packages such as FRAGSTATS. While this model provides clear and concise end
99 products and has significantly contributed in understanding ecological patternprocess relationships (Turner, 2005), it is nonrealistic in capturing surface heterogeneity and may result in a loss of important ecological information. Most surface processes are inherently continuous in nature, hence account for a rather continuous representation of the surface properties. Another limitation of the patchmosaic model is that it does not consider aspects of the third dimension, such as surface roughness, landform or relief variability (Hoechstetter et al., 2008) or intens ity of the predictor variable (McGarigal and Cushman, 2005). While FRAGSTATS now offer a moving window analysis that generates continuous surfaces at different scales, which can be linked to organisms perspectives, the analysis is based on categorical maps and hence inherits all the limitations of the patch mosaic model. A conceptual shift has been proposed suggesting the surface gradient model as a supplement for the patch mosaic model to incorporate hierarchical and multi scale ecological systems (McGari gal and Cushman, 2005). However, few studies test this new model in quantifying surface pattern. In surface gradient model it is possible to retain the inherent variability of a continuous predictor variable, without introducing subjectivity through defining sharp boundaries needed for categorization. Since the underlying heterogeneity of the predictor variable is preserved through space and across scale, it is possible to test a single predictor variable against multi scalar response variables. Most im portantly, continuous spatial and temporal patterns can be directly linked to ecological processes acting at different scales, without compromising organisms perspectives (Muller, 1998). The surface spatial heterogeneity strongly depends on the predictor variable used to analyze it (Turner et al., 2001). Spatial and temporal remote sensing data have been
100 widely used to build ecosystem models about vegetation characteristics (Foody and Curran, 1994; Franklin, 1995; Kustas and Norman, 1996; Melesse et al., 2007; Stockli et al., 2008). The Enhanced Vegetation Index (EVI) is one such continuous variable that correlates well with several components of ecosystem dynamics such as leaf area index, biomass, canopy cover, and the fraction of absorbed photosynthetical ly active radiation (Boegh et al., 2002; Huete et al., 2006). EVI has been shown to be less prone to saturation in temperate and tropical forests making it more efficient over other widely used vegetation indices such as NDVI (Huete et al., 2006). Due to t hese properties EVI has been more effective for monitoring seasonal, inter annual, and longterm variation in vegetation structure. EVI could be particularly useful in tropical regions where forest degradation or thinning is more prevalent than forest conv ersion and especially in places where discrete remote sensing classification techniques fail to capture subtle changes within a single class. Recently, a 2band variation of EVI (EVI2) was developed for sensors with no blue band, such as MODIS and ASTER (J iang et al., 2008; Stevens, 2009). EVI2 has been reported to correspond well with the original EVI (Jiang et al., 2008) and is expected to contribute towards a multi sensor seamless vegetation index record. However, before using EVI2 to describe surface gr adients, this variable needs to be tested for its capability to represent land cover diversity without loosing important information about the landscape. To adopt the surface gradient models, landscape ecologists need to utilize a new suite of tools to analyze landscape structure. There are many promising techniques widely used in other disciplines, such as fractal analysis, spectral and wavelet analysis and surface metrology, which are theoretically suitable to analyze continuous variables
101 (McGarigal and Cushman, 2005; Cushman et al., 2010). Surface metrology, primarily a field developed for microscopy and molecular physics, is one such technique that holds promise for landscape surface analysis. In the past decade, several families of surface metrics hav e been developed and widely used in surface metrology. Scanning probe Image Processor (SPIP) software package has implemented such surface metrics that can quantify measures of surface amplitude in terms of its overall roughness, skewness and kurtosis, and total and relative amplitude (McGarigal et al., 2009). These techniques have made notable progress in the last decade, where they have been applied to different fields of study, such as in pharmaceutical science quantifying the intrinsic adhesion forces of lubricants and other pharmaceutical materials to a steel surface (Lee, 2004), in cell research quantifying the effect of certain hormone in kidney dysfunction (Oberleithner et al., 2004), and in polymer science mapping various surface properties of polym er based materials (Munz et al., 2003). Theoretically these techniques hold promise for surface gradient approach in landscape ecology as well. Yet few studies employ these tools to analyze continuous surface gradients. This research attempts to adopt the surface gradient model as the conceptual basis and applies some relatively less known quantitative techniques for measuring surface gradients. This study explores EVI2 as an indicator of landscape heterogeneity as reflected in different land covers in a humanmodified tropical landscape in Central India. Specific research questions addressed in this study are: Can EVI2 be used as a gradient surface to represent land cover diversity in a humanmodified tropical landscape? How effective are the surface metri cs, in terms of describing landscape composition and configuration?
102 Advantages and disadvantages of using such metrics are then discussed in the context of the current study. Methods Study Area Four sample landscapes were delineated to capture the varying landscape composition and configuration (Fig. 51). These sample landscapes were chosen based on their land use history and different landscape compositions as shown by the EVI2 histograms (Fig. 5 2). Subsets 1 is located at the reserve boundary and is composed of protected forest, commercially managed forest and some agricultural lands. The only forest village in the reserve is located within subset 1. Subset 2 contains commercially managed forest and residential area with agricultural land. Subset 3 is l ocated within the reserve and is almost entirely composed of protected forest. Subset 4 is composed of agricultural lands with sparse vegetation. Both the subsets 2 and 4 are located far from the reserve, and are different from each other regarding landscape composition and configuration. Spatial Data A mosaic of two ASTER images was used to calculate EVI2 for this study. Both of the images are from the same date to minimize variability and noise. Images were geometrically rectified, radiometrically calibrated, and atmospherically corrected (Adler Golden et al., 1999; Abrams et al., 2002). EVI2 was calculated as: EVI2 = 2.5*[(NIR RED)/(NIR+2.4*RED+1)] (Jiang et al., 2008; Stevens, 2009), where NIR and RED are estimated surface reflectance values for near infrared and visible red bands (ASTER bands 3 and 2, respectively). EVI2, like NDVI, is a unitless continuous variable and ranges between 1 and 1. Remote sensing variables, other than EVI2, were also
103 extracted from the images in order to include them in the statistical analysis. These variables include estimated surface reflectance values derived from each of 9 ASTER nonthermal bands (Table 51), one temperature image with blackbody surface temperature values calculated from all five ASTER thermal bands, and NDVI values calculated from estimated surface reflectance values for ASTER bands 2 and 3. A discrete categorical map (accuracy 94%, see Chapter 3 for details) was included as a proxy for landscape heterogeneity to check the utility of EVI2 to represent land cover diversity. The landscape shows disti nct characteristics in terms of the remote sensing variables used in this study (Fig. 5 3). Discrete classification of this landscape identified 3 broad land cover classes: forest, bare soil and water (see Mondal and Southworth, 2010 for details). The land cover map suggests somewhat equal proportion of forest and nonforest (cleared forest, bare agricultural land, and water) components in the entire landscape. However, forest is the dominant land cover within the tiger reserve. The land cover map also suggests extensive agricultural land and cleared forest (identified as bare soil in the map) surrounding the reserve. Forested areas, within or outside the reserve, are associated with lower blackbody surface temperature, and higher EVI2 and NDVI values. Areas designated as bare soil in the land cover map, are readily distinguishable for their higher blackbody surface temperature values (Fig. 5 3). These are also the areas with lower EVI2 and NDVI values. Water bodies also have lower EVI2 and NDVI values; howev er have lower blackbody surface temperatures unlike bare soil.
104 Statistical Analysis Prior to any statistical analysis all images were rescaled to 30 m x 30 m pixel size. A representative landscape was chosen as a subset from the image for statistical analy ses to reduce computing time (Fig. 51). Then the variables were converted in an ASCII dataset to be imported in the statistical software package PASW Statistics 18. Since, this analysis included both categorical and continuous variables, and linearity among these variables was not out of question, eta ( ) coefficient was used to measure the correlation. Eta is equal to Pearsons r for linear relationship, but is greater than r for nonlinear relationship. Eta2 describes the percent of variation in the dependent variable described linearly or nonlinearly by the independent variable. A correlation matrix was calculated, which examines the level of interrelationships among the reflective bands, estimated surface temperature, vegetation indices and discrete land cover classes (Table 5 2). Eta varies from 0 to 1 and has no sign. Although the eta coefficient can not be used to establish causal direction, it is a useful coefficient to examine correlations among mixed variables of nonlinear association (Southworth, 2004; Garson, 2008). A discriminant function analysis was used to identify the relative capability of EVI2 to differentiate between different land covers. The Wilks lambda coefficient was used to determi ne the relative importance of input variables in discriminating land covers (Table 5 3). The magnitude of the lambda coefficient (0.0 1.0) is inversely proportional to the ability of a variable to discriminate between the classes (Boyd et al., 1996). Surfa ce Metrics This study utilized some of the metrics generated by the software SPIP and examined their utility in quantifying surface properties. SPIP generated metrics can be
105 of both spatial and nonspatial in nature. While nonspatial metrics can explain l andscape composition, spatial metrics can be used to describe landscape configuration. Spatial metrics incorporate spatial information by describing variability in the overall distribution of the gradient surface as well as the spatial arrangement, locatio n or distribution of peaks/valleys in the gradient surface. Nonspatial metrics are sensitive to overall height distribution in the gradient surface, but not the spatial arrangement, location or distribution of peaks/valleys in the gradient surface. In thi s study both spatial and nonspatial metrics were included. Specifically, the following metrics were chosen for their relevance to landscape ecology and to avoid redundancy (McGarigal et al., 2009). Selected metrics are described here briefly, for a detail ed description please refer to McGarigal et al., 2009 and SPIP documentation (SPIP). Surface skewness (Ssk) : describes the asymmetry of surface height distribution about the mean. A surface with dominantly higher values (or peaks) and occasional lower values would have negative Ssk values, while a surface with dominantly lower values (or valleys) and occasional higher values would have positive Ssk values. High skewness (positive or negative) denotes a landscape with dominant surface heights, i.e. ex treme values in the surface properties. Landscapes with somewhat even distribution of peaks and valleys would have Ssk values near to zero. For this study, landscapes with dominant vegetation would have mostly higher EVI2 values with occasional lower values for nonvegetated areas, hence are expected to have negative Ssk values. On the contrary, landscapes with mostly nonforested areas would have lower EVI2 values with occasional peaks for vegetation, and hence are expected to have positive Ssk values. S urface kurtosis (Sku) : describes the shape of the surface distribution. Like Ssk Sku is sensitive to high peaks or deep valleys. A landscape with relatively even distribution of surface heights about the mean is said to be platykurtic, with Sku valu es less than 3. A landscape with areas well above or below the mean (dominantly high peaks or deep valleys) has Sku values greater than 3 and is called leptokurtic. For example, landscapes with dominant land covers (forest or nonforest) are expected to be leptokurtic, while landscapes with somewhat even proportion of land cover components are expected to be platykurtic. Dominant texture direction (Std) : returns the angle of the dominating texture in the image as calculated from the Fourier spectrum. St d values range between 0 and 180 degrees. This metric is only meaningful when there is a dominant texture in the
106 landscape; otherwise it returns a value of zero (i.e. for flat landscapes). The distribution of Std is given by a graph called angular spectrum, which represents relative amplitude for different angles as calculated by summation of the amplitudes along M equiangularly separated radial lines. This spatial metric could be particularly useful to locate parallel ridges (e.g. forest corridors) embedded within other land cover components (e.g. matrix of nonforest), where texture direction would be parallel to the direction of the ridges. Texture direction index (Stdi) : measures the relative dominance of the dominating texture direction and is calc ulated as the average amplitude sum divided by the amplitude sum of the dominating direction. Stdi values range between 0 and 1. Landscapes with dominant texture direction have values closer to 0, while landscapes with similar amplitude sum for all directi ons have a value equal to 1. Ssk and Sku are called amplitude parameters and are nonspatial in nature. Std and Stdi are called spatial parameters as they provide spatial information about the landscape. The software requires the landscapes to be square i n shape; hence all the four sample landscapes are 5 km x 5 km in extent. Results Statistical Analysis EVI2 was identified as the most suitable continuous variable, among the ones used in this study, to represent the land cover diversity as suggested by the correlation and discriminant analysis results. EVI2 was highly correlated with land cover classes lso had the minimum value for the lambda coefficient (0.28), meaning that it has a high capability of differentiating between different land covers (Table 53). While both the vegetation indices (EVI2 and NDVI) performed quite well in terms of representing land cover diversity, EVI2 has an advantage over NDVI in tropical regions with high biomass, as in the present study area. However, the land cover map used in this study has very
107 low thematic resolution. Hence, correlation of EVI2 to land cover maps with higher thematic resolution should be further tested. Surface Metrics Ssk values are negative for all the landscapes except subset 4 (Table 54). Negatively skewed surface distribution, for this particular study, means dominance of high EVI2 values, which in turn means dominance of vegetation in the sample landscapes. However, the magnitude of Ssk values for these landscapes vary greatly. Subsets 1 and 2 have comparatively higher negative values denoting dominance of vegetation with some deep valleys, i.e. nonvegetated area such as bare agricultural land, water and forest vil lages. Subset 3 has a very low negative value denoting mostly vegetated area with very little nonvegetated area. Subset 4 has a low positive Ssk value, representing a landscape of dominantly nonvegetated area with some vegetation. Sku values for these f our landscapes also support the above findings. All the landscapes except subset 4 have high (>3) Sku values (Table 54). Among these four landscapes, subset 1 has the highest kurtosis, which can be linked to the location of a forest village represented by very low EVI2 values (Fig. 53). Subsets 2 and 3 have moderately high Sku values, suggesting nonuniform distribution of vegetated and nonvegetated components in landscape composition. In addition, if we consider Sku and Ssk values for subset 3 together, it informs not only about dominance of forest cover but also about the lack of dominant nonforest land cover (i.e. deep valley). Subset 4 has Sku value less than 3, representing a landscape with somewhat uniform distribution of vegetated and nonvegetated surface. This typically represents agricultural fields with surrounding trees.
108 Spatial metrics for the sample landscapes suggest lack of any dominant texture direction. Most of the Stdi values are near to 1 (Table 5 4), meaning almost even distribution of peaks and valleys in this landscape. Subsets 2 and 3 have moderate Stdi values (0.6 0.7) suggesting weak to moderate texture in directions given by corresponding Std values (Table 5 4). Subsets 1 and 4 lack any dominant texture direction, as suggested by Stdi values (Table 5 4). Discussion Remote sensing variables have frequently been identified to be particularly suitable for monitoring dynamic environmental processes, such as the exchange of mass and energy among soil, vegetation, and atmosphere (Sellers et al., 1997; Mecikalski et al., 1999; Turner et al., 2003; Leuning et al., 2005). Monitoring rapid environmental changes, especially vegetation, under differing natural and anthropogenic influence has become increasingly efficient in the past few decades due to the availability of consistent remote sensing datasets. Vegetation indices, especially NDVI, are datasets which can be directly linked to the continuous surface processes, and are more likely to reflect the actual landscape with their inherent gradients. This study suggests that EVI2 is another important variable that can be used for environmental monitoring as a proxy for land cover diversity (Tables 5 2 and 5 3). Moreover, EVI2 has the capability to capture subtle changes in vegetation condition and structure, especially to differentiate between surface greenness and leaf area index (LAI) for vegetation with different background soil reflectance (Rocha and Shaver, 2009), and thus can be directly linked to land cover modifications. Designed for sensors without a blue band, EVI2 can not only be used for moderate to largescale studies involving ASTER, MODIS and AVHRR datasets, but can also be utilized in multi sensor
109 time series analysis. The findings show ed strong correlation between EVI2 and discrete land cover classes, suggesting the robustness of this variable as a representative for land cover diversity. Correlation of EVI2 and NDVI enabled confident use of EVI2 in this study, as NDVI has been previous ly used in many studies to successfully describe spatial heterogeneity (Kawabata et al., 2001; Garrigues et al., 2006; Gillespie et al., 2006; Garrigues et al., 2008). Given their varied composition and configuration, the sample landscapes were expected t o have varied surface metrics. Apparently, only nonspatial metrics performed well in bringing out the different characteristics of these landscapes (Table 5 4). Both Ssk and Sku values for subsets 1 and 2 indicate dominant surface heights for the landscapes, which can be interpreted as landscape dominance or evenness, analogous to the patchbased evenness metrics (McGarigal et al., 2009). The tiger reserve in the study area has maintained its forest cover in the last decade after showing some initial defor estation trend in the early 1980s (Mondal and Southworth, 2010). Hence, a uniform EVI2 distribution was expected for subset 3 which is a part of the reserve. The surface metrics not only identified dominance of vegetation in this subset, but also gave an i dea of the relative proportion of vegetation and nonvegetation in the subset through its negative skewness and moderate kurtosis values (Table 5 4). The nonspatial metrics also accurately identified dominance of bare agricultural land (with low EVI2 values) with substantial presence of vegetation (peaks in EVI2 values) in subset 4, as reflected in its positive skewness and low kurtosis values (Table 4). Spatial metrics, on the other hand, did not provide with any additional ecological information. According to the spatial metrics the sample landscapes lack any dominant
110 texture (Table 5 4). This is partly expected as subsets 3 and 4 visually lack strong texture orientation, being mostly dominated by vegetation and agricultural land respectively (Fig. 5 3 ). However, texture direction index failed to pick up any signal for sudden drop in EVI2 value due to the location of a forest village within the reserve in subset 1, returning a high Stdi value of 0.82 (Table 5 4). Stdi value of 0.60 for subset 2 somewhat supports the sharp boundary between forest and agricultural lands (Fig. 5 3). However, quite contrary to the expectation, subset 3 which is mainly composed of protected forest has moderate texture direction index ( Stdi=0.70). Further examination of the l andscape revealed the presence of fine scale linear nonvegetated elements (low EVI2), identified as the tributaries to the Pench River. It was interesting to note that this metric could identify such fine scale elements in a landscape, yet was not very sensitive to sharp boundaries between different land covers. This discrepancy could be explained by any of the following: a) the landscapes are visually misleading and actually lack strong texture or b) the metrics are not designed to capture texture orientation defined by sharp boundary between two land covers or sudden decrease in continuous variable at a landscape level. Also the texture orientation metric could be particularly sensitive to linear elements and only pick up signals when landscapes have such features. Whichever is the case, there is a strong need to test these metrics in other landscapes to evaluate their compatibility in landscape studies. While some of the surface metrics used in this study were able to describe the landscape characterist ics, there are definitely some questions about applicability of such metrics in landscapelevel studies and robustness of these techniques. Even though there is a plethora of metrics available in SPIP, this study includes only four
111 metrics for specific rea sons. Most of the SPIP generated metrics are sensitive to the unit of the predictor variable, thus cannot be used with unitless variables. Many of the SPIP generated metrics have theoretical applicability in landscape level studies, such as summit density ( Sds ), surface area ratio ( Sdr ), root mean square slope ( Sdq ), dominant radial wavelength ( Srw) (refer to SPIP documentation for detailed description of these metrics). However, SPIP did not generate any interpretable results for these metrics for the pres ent study. This can be attributed to the fact that the software used in this study was developed for microscopy and molecular physics, which deal with very different spatial scales than landscape ecology studies. While theoretically these metrics hold promise to be efficient in ecological studies, the scale compatibility issue further underscores the serious need to develop user friendly and easily available software compatible for landscape studies. Conclusion This study utilized the surface gradient m odel as a conceptual basis to quantify EVI2 as a surface gradient. This study explored the utility of some relatively less studied tools in a tropical landscape. EVI2, used as a proxy for land cover diversity in this study, correlated well with the low res olution thematic land cover map. This study also suggests that there is potential for the surface metrics in ecological studies, provided availability of user friendly tools and utilization of effective predictor variable. Currently, there are several avai lable tools that can handle continuous variables to generate surface metrics. However, the challenge would be to tailor those tools for landscape ecology and/or develop new tools meant for landscape level studies.
112 Table 51. Characteristics of ASTER vis ible near infrared (VNIR), shortwave infrared (SWIR) and thermal infrared (TIR) bands. Band Spectral Resolution (m) Spatial Resolution (m) VNIR 1 0.520 0.600 15 x 15 2 0.630 0.690 3 0.760 0.860 SWIR 4 1.600 1.700 30 x 30 5 2.145 2.185 6 2.185 2.225 7 2.235 2.285 8 2.295 2.365 9 2.360 2.430 TIR 10 8.125 8.475 90 x 90 11 8.475 8.825 12 8.925 9.275 13 10.250 10.950 14 10.950 11.650
113 Table 5 2 Correlation matrix showing the level of interrelationships among the reflected bands (Band 1 through 9), one blackbody surface temperature image (Temp) calculated from five thermal bands, vegetation indices (EVI2 and NDVI) and discrete land cover classes The table lists eta coefficients ( ) as a measure of nonlinear association. Band1 Band2 Band3 Band4 Band5 Band6 Band7 Band8 Band9 Temp EVI2 NDVI Class Band1 1.00 Band2 .97 1.00 Band3 .72 .69 1.00 Band4 .88 .88 .87 1.00 Band5 .89 .91 .82 .97 1.00 Band6 .90 .91 .83 .98 .99 1.00 Band7 .89 .90 .82 .97 .99 .99 1.00 Band8 .90 .92 .82 .97 .98 .99 .99 1.00 Band9 .87 .90 .80 .95 .98 .98 .98 .98 1.00 Temp .69 .73 .39 .60 .73 .72 .74 .75 .74 1.00 EVI2 .81 .86 .88 .83 .81 .82 .82 .82 .81 .86 1.00 NDVI .84 .87 .84 .85 .85 .86 .86 .86 .85 .93 .95 1.00 Class .62 .67 .59 .62 .62 .62 .63 .63 .61 .74 .85 .82 1.00
114 Table 53 Discriminant analysis results showing the relative capability of input variables in differentiating between different land cover classes. The magnitude of the Wilks lambda coefficient is inversely proportional to the ability of a variable to discriminat e between the classes and is sorted in ascending order in the table. Variables Wilks Lambda EVI2 .28 NDVI .33 Temp .45 Band2 .55 Band4 .61 Band5 .61 Band6 .61 Band7 .61 Band8 .61 Band1 .62 Band9 .62 Band3 .65 Table 5 4 Surface metrics calculated from EVI2 for four sample landscapes from the study area. Ssk and Sku are nonspatial metrics denoting landscape composition, while Std and Stdi are spatial metrics describing landscape configuration. All of these are unitless m etrics, except Std which is measured in degrees. Surface Metrics Subset 1 Subset 2 Subset 3 Subset 4 Surface Skewness (Ssk) -1.73 -1.61 -0.25 0.33 Surface Kurtosis (Sku) 8.44 5.24 4.37 2.84 Dominant Texture Direction (Std) 74.71 89.53 78.99 83.81 Texture Direction Index (Stdi) 0.82 0.60 0.70 0.88
115 Fi gure 5 1. Land cover map shows locations of four sample landscapes used in this study in reference to the Pench Tiger Reserve.
116 Figure 52. EVI2 histograms for four sample landscapes.
117 Figure 5 3. Spatial patterns in the study area as demonstrated by (a) land cover map, (b) blackbody surface temperature ( C) image, (c) NDVI image, and (d) EVI2 image.
118 CHAPTER 6 COMPARISON OF ASTERDERIVED THERMAL INDICES AND TASSELED CAP NORMALIZED INDICES IN DETECTING TROPICAL FOREST SUCCESSIONS Introduction Successional forests differ in their average height and basal area, the stand volume and basal density, and physiognomic characteristics (Moran and Brondizio, 1998). Quantification o f specific forest parameters is important to better understand the impact of different successional stages on ecosystem changes, specifically water and nutrient cycling and carbon sequestration (Lu et al., 2003; Song et al., 2007). The first step to this quantification would need detailed mapping of different successional stages. Moderate resolution satellite images, such as those from Landsat, have been successfully used for such mapping purposes. However, technical issues related to longevity and quality of Landsat sensors warrant for assessment of alternate moderate resolution satellite images for such mapping purposes. While ASTER has comparable spatial, spectral and temporal resolutions making it effective replacement for Landsat, few studies report the capability of ASTERderived vegetation indices or thermal indices in detecting tropical forest successions. The tasseled cap indices, primarily developed for Landsat sensors, have been widely used for mapping and monitoring land cover changes, especially subtle changes in vegetation condition ( Cohen and Spies, 1992; Cohen et al, 1995). These indices, when normalized, often highlight a certain land cover characteristics (Healey et al., 2005). For example, brightness, greenness, and wetness indices normalized for a dense forest class would record spectral distance of a given pixel from the dense forest centroid of the corresponding indices for a particular scene (Masek et al., 2008), thus highlighting disturbed forest pixels. These indices require linear combinations of
119 individual channels specific to each sensor. While coefficients for tasseled cap transformations have been proposed (Yarbrough et al., 2005) for ASTER, the utility of ASTERderived tasseled cap indices in identifying vegetation structure or c omposition is yet to be tested. Use of thermal infrared emissivities as an indicator of land degradation is a fairly new approach within remote sensing realm and is advantageous in many ways. Emissivity is defined as the ratio between actual emitted radiation and emitted radiation from a blackbody at the same temperature. Emissivity is independent of temperature itself and varies spectrally according to surface composition and geometry. These latter properties make emissivity observations potentially useful for land cover characterization. Along with emissivities, surface temperature also holds promise to effectively detect forest successions. This study compares ASTER derived tasseled cap indices, surface temperature and emissivitity in detecting tropical forest successions. Specifically, this study seeks to identify the degree of association of the remote sensing indices to different forest successions derived from historical land cover maps. Methods Study Area This study utilized a heterogeneous landscape located in Central India as a template for tropical ecosystem (Fig. 6 1). The study area is dominated by tropical deciduous forest, especially teak. Some of these forests are protected as the landscape hosts Pench Tiger Reserve. The rest of the forests ar e commercially managed and supports subsistence livelihood of the local people. Interplay of changing forestry policies, various degrees of protectionism and local livelihood has resulted in
120 successional forests in this landscape, which is representative of many developing countries. Data Pre processing Two scenes of ASTER L1B data, acquired on October 30, 2007, were downloaded from LPDAAC for path 144/row 45. These images were generated by applying coefficients for radiometric calibration and geometric rectification (Abrams et al., 2002). Because ASTER images come with three spatial resolutions, visible infrared bands (bands 1 through 3) and shortwave infrared bands (bands 4 through 9) were fused to 30m pixel resolution. To reduce betweenscenes variabilit y, the at sensor reflectance was calculated based on the following equation: p 2 L. (Esz))1 (6 1) where d is the astronomical distance of the earth to the sun, L is the at sensor sz is the solar zenith angle in radians, and Eis the solar spectral irradiance as calculated by World Radiation Center (Yarbrough et al., 2005). The two at sensor reflectance images were then mosaicked and later used to calculate tasseled cap indices. Thermal atmospheric correction is a necessary step to approximate and remove atmospheric contributions from thermal infrared radiance data. Thermal bands (bands 10 through 14) were atmospherically corrected using the inbuilt function of ENVI 4.5 software package, which used the InScene Atmospheric Compensation (ISAC) algorithm (Johnson and Young, 1998). Assumptions of the algorithm include uniformity of atmosphere over the data scene, presence of a near blackbody surface within the scene and absence of r eflected downwelling radiance. Upwelling atmospheric radiance and atmospheric transmission are approximated using the Planck function and
121 assumed emissivity of 1. The two atmospherically corrected scenes were then mosaicked to cover the study area. Successional Forest Subsets I extracted two forest subsets for this study, for which I used a change detection map generated for a previous study (Mondal and Southworth, 2010) (Fig. 6 1). The first subset is for mature forests that have been stable for last 30 years (1977 onwards, just for park). For this group, I included areas those were forests on all four dates (1977, 1989, 2000 and 2007) in the four date change detection map. However, forested areas on these four dates do not necessarily represent mature forest, as it is possible for these areas to witness land cover changes between any two years. To overcome this limitation, I extracted data only for the park, hopefully increasing the probability of the stable forest subset truly representing mature forest be cause of the protected status of the park since 1975. The other subset is for regenerated forests with age less than 7 years. For this group I included areas those converted from any nonforest land cover in 2000 to forest in 2007. Emissivity and Surface Temperature Extraction This study used emissivity normalization method to separate emissivity and temperature values from atmospherically corrected thermal infrared radiance data. This technique calculated the temperature for every pixel and band in the dat a using a fixed emissivity value (1 for this study). The highest temperature for each pixel is then used to calculate the emissivity values using the Planck function. While multiple methods for separating temperature and emissivity data exist, emissivity normalization method has been reported to be slightly superior to the others in a heterogeneous landscape dominated by gray body targets, such as vegetation (French et al., 2008). The fixed
122 emissivity value for this study was chosen as 1 to minimize the correlation with any particular land cover, as I w as interested to compare the emissivity values derived from all of the 5 ASTER thermal bands in identifying different forest successions. Besides, at the 90 m spatial resolution of the thermal bands, vegetati on is viewed as a canopy and multi scattering effects would increase maximum emissivity to over 0.99 ( Salisbusy and DAria, 1992). In addition to the emissivity values for all the 5 thermal bands, 3 new emissivity indices were developed using combinations of emissivity values for bands 10, 11 and 12. The formulas for these indices are listed under Table 6 1. Normalized Tasseled Cap Indices The at sensor reflectance values for visible infrared and shortwave infrared bands were utilized for tasseled cap transformation using the transform coefficients developed for ASTER at sensor reflectance (Yarbrough et al., 2005). The three principal components in the direction of brightness, greenness, and wetness were further used to develop normalized tasseled cap indices as described below. I used the stable forest mask to generate the normalized tasseled cap indices, such that (for example): B = (B BB (6 2) where B is the tasseled cap brightness index for any pixel, B is the mean tasseled cap B is the standard deviation of brightness within the stable forest class. Statistical Analysis The tw o subsets were used to extract data for each of the vegetation and thermal indices used in this study. A smaller data set was chosen for statistical analysis through random sampling. Then the variables were converted in an ASCII dataset to be imported in t he statistical software package PASW Statistics 18. As a measure of
123 correlation eta ( ) coefficient was used, since, unlike Pearsons r eta can be used for both linear and nonlinear relationships. Eta2 describes the percent of variation in the dependent variable described linearly or nonlinearly by the independent variable. Eta varies from 0 to 1 and has no sign. Although the eta coefficient can not be used to establish causal direction, it is a useful coefficient to examine correlations among mixed vari ables of nonlinear association (Southworth, 2004; Garson, 2008). (Table 6 2). A discriminant function analysis was used to identify the relative capability of the thermal and vegetation indices included in this study to detect the forest successions. The Wilks lambda coefficient was used to determine the relative importance of input variables in discriminating two forest classes (Table 6 3). The magnitude of the lambda coefficient (0.0 1.0) is inversely proportional to the ability of a variable to discri minate between the classes (Boyd et al., 1996). Results Eta coefficient suggests strongest correlation between surface temperature and forest successions closely followed by two of the normalized emissivity indices (Table 6 2). Individual emissivities for bands 10, 11, and 12 performed better than the normalized tasseled cap indices and emissivities for bands 13 and 14. Discriminant function analysis suggests similar results with minimum Wilks lambda coefficient for surface temperature, followed by two of the new emissivity indices (12by10 and 11by10). Normalized greenness and wetness indices were least capable to detect forest successions (Table 6 3). Discussion Landsat tasseled cap indices have been used to detect successional stages in various forest ec osystems, such as conifer forest (Cohen and Spies, 1992), hardwood
124 and softwood forests (Jin and Sader, 2005), and tropical forests (Lu et al., 2004). Normalization of these indices is expected to highlight deviation of spectral signatures of any given pix el from that of a stable forest. However, for the present study normalized greenness and wetness indices failed to correlate with forest masks with different ages. Unlike vegetation indices, emissivity responds to plant canopy geometry and patterns between plant canopies. Hence it was expected to detect any successional changes in the present landscape. Previous study has reported capability of ASTER band 11 emissivity to detect degradation in sparsely vegetated terrain (French et al., 2008). This was possi ble as high emissivity vegetation is clearly distinguished from low emissivity soil in a moderate resolution image. However, in a vegetation dominated region, such as my study area, it is difficult to distinguish between different successional stages by em issivity alone because of the very small emissivity ranges. The new emissivity indices slightly improved the performance, as they tend to highlight the small differences between emissivity values for different bands and different vegetation structure. The land surface temperature performed best among the used variables to detect successional changes. Surface temperature reflects canopy temperature, rather than near surface temperature, in tropical forests. Hence mature and old forests often have a lower te mperature than young opencanopy forests (Southworth, 2004). In this study, however, some areas classified as the stable forest have higher temperature (Fig. 6 2 ). These areas most probably do not represent mature forest, but are classified as forest in al l four dates resulting in being classified as stable forest in the four date change trajectory image. To address this issue I included stable forest only from the park.
125 This study highlights the need of further research to investigate the capability of ASTERderived vegetation and thermal indices in detecting forest composition, structure, and density. Landsat derived indices have been widely used in mapping and monitoring forest characteristics. Applicability of ASTER derived indices in such purposes would ensure longterm record of forest characteristics.
126 Table 61. Calculation for newly developed emissivity indices. The numbers 10, 11 and 12 represent emissivity values for ASTER bands 10, 11, and 12 respectively New Emissivity Index Calculation 11b y10 11 10/11 + 10 12by10 12 10/12 + 10 11by12 11 12/11 + 12 Table 62. The level of interrelationships of forest successions derived from change detection map with remotely sensed thermal and vegetation indices. The table lists eta coefficients linear association and is sorted in ascending order in the table. Remote sensing variables Eta Eta2 Surface temperature 0.566 0.321 Index 12by10 0.536 0.288 Index 11by10 0.525 0.276 Band 12 emissivity 0.470 0.221 Band 10 emissivity 0.456 0.208 Band 11 emissivity 0.376 0.141 Index 11by12 0.318 0.101 Normalized tasseled cap brightness 0.296 0.088 Band 14 emissivity 0.094 0.009 Band 13 emissivity 0.058 0.003 Normalized tasseled cap greenness 0.014 0.000 Normalized tasseled cap wetness 0.004 0.000
127 Table 6 3. Discriminant analysis results showing the relative capability of input variables in detecting tropical forest successions. The magnitude of the Wilks lambda coefficient is inversely proportional to the ability of a variable to discriminate between the classes and is sorted in ascending order in the table. Remote sensing variables Wilks lambda Surface temperature 0.679 Index 12by10 0.712 Index 11by10 0.724 Band 12 emissivity 0.779 Band 10 emissivity 0.792 Band 11 emissivity 0.859 Index 11by12 0.899 Normalized tasseled cap brightness 0.912 Band 14 emissivity 0.991 Band 13 emissivity 0.997 Normalized tasseled cap greenness 1.000 Normalized tasseled cap wetness 1.000
128 Figure.61. Study area showing (a) four date change trajectory, (b) stable forest mask, and (c) regenerated forest mask.
129 Figure.62. Study area showing e stimated blackbody surface temperature of stable forest.
130 CHAPTER 7 CONCLUSION Contribution of the Study Tropical forests provide important ecosystem services. In order to protect those forests and forest dependent species, protected areas have been established across the globe. While the field of biodiversity conservation increasingly recognizes the need for empirical evaluations of conservation interventions, few well designed empirical studies try to show what could have happened in the absence of the conservation efforts. In addition, landscape dynamics in and around a protected area need to be identified in order to inform natural resource manager s. This dissertation targets a tropical protected landscape located in Central India. This decadal study documents changes in land cover in and around PTR during 19772007, wit h emphasis on forest transitions and productivity. It has been suggested by the park managers and the academics that a particular national level policy change was particularly effective in increasing forest cover in national parks countrywide. This study t ests the effectiveness of that particular policy in limiting deforestation in PTR by generating modeled landscapes and then comparing them with the actual landscapes. Since tropical forests tend to saturate in terms of the most widely used remote sensing v egetation index NDVI, this study utilizes EVI2 in quantifying land cover diversity in the study area. This study also reports the capability of ASTER derived surface temperature, emissivity and tasseled cap indices in detecting forest successions. While reporting regional changes through interdisciplinary tools and techniques, this study also highlights the applicability and potential of some of the newly developed indices and methods in landscape level studies. The overall focus of this research is environmental geography with specific
131 contribution towards the better understanding of spatial distribution of humanenvironmental interactions. Broadly this research contributes to two related fields of study land change science and biodiversity conservation/natural resource management. Implication s for PTR PTR, as one of the tiger reserves of India, holds potential to sustain viable tiger population, in presence of thriving prey base, which in turn depends on healthy vegetation cover. While PTR is the smallest tiger reserve in India, it is contiguous to another tiger reserve with the same name in the state of Madh ya Pradesh. In addition, PTR is connected to Kanha Tiger Reserve in Madhya Pradesh. In order to maintain the connectivity with other suitable tiger habitats it was therefore necessary to gain knowledge on the changing landscape dynamics after its establishment as a national park. PTR witnessed several changes in degree of protectionism after its declaration as a national park in 1975 and as a tiger reserve in 1999. Relocation of core villages and settlements outside PTR, and presence of over 40 villages within 10 km of park boundary have resulted in a highly human modified m atrix in which PTR is embedded. Identifying the spatial patterns of changes in land cover would facilitate better understanding of the underlying processes and mitigating them. A new management plan has been approved by the Government of India, where the surrounding has been declared as the buffer area. This declaration along with alternate ways of subsistence livelihood for the local people c ould ensure decline in anthropogenic pressure in the core and buffer forest. As this study has showed, presence of villages in the surroundings has resulted in a patchy mosaic of land uses, which c ould be limiting
132 animal movements Further efforts should be made at national and subnational administrative level to ameliorate park people relationship and to introduce a truly participatory management approach. This study is the building block of a larger scale study that I intend to pursue in future. The Pench landscape is fascinating for a land change scientist in many ways, and will play a pivotal role in the local chapter of the global tiger conservation effort. Future directions of research would include (a) identifying subtle withinclass changes using advanced remote sensing techniques, which will contribute to land change science, (b) identifying degradation in nonprotected matrix, which will have cruci al implications for biodiversity conservation, and (c) developing and testing robust methods from different subdisciplines applicable to any ecosystem, which will require an inter disciplinary approach to study humanenvironmental interactions.
133 LIST OF RE FERENCES Abrams, M., Hook, S., Ramachandran, B., 2002. ASTER User Handbook: Advanced Spaceborne Thermal Emission and Reflection Radiometer. Version 2, Jet Propulsion Laboratory/California Institute of Technology. Adler Golden, S.M., Matthew, M.W., Bernstei n, L.S., Levine, R.Y., Berk, A., Richtsmeier, S.C., Acharya, P.K., Anderson, G.P., Felde, G., Gardner, J., Hoke, M., Jeong, L.S., Pukall, B., Ratkowski, A., Burke, H.H., 1999. Atmospheric Correction for Short wave Spectral Imagery Based on MODTRAN4. In, S PIE Proceedings on Imaging Spectrometry, pp. 6169. Ahmed, M.H., Leithy, B.M., Thompson, J.R., Flower, R.J., Ramdani, M., Ayache, F., Hassan, S.M., 2009. Application of remote sensing to site characterisation and environmental change analysis of North Afri can coastal lagoons. Hydrobiologia 622, 147171. Alonso, D., Sole, R.V., 2000. The DivGame Simulator: a stochastic cellular automata model of rainforest dynamics. Ecological Modelling 133, 131141. Andam, K.S., Ferraro, P.J., Pfaff, A., Sanchez Azofeifa, G .A., Robalino, J.A., 2008. Measuring the effectiveness of protected area networks in reducing deforestation. Proceedings of the National Academy of Sciences of the United States of America 105, 16089 16094. Arroyo Mora, J.P., Sanchez Azofeifa, G.A., Rivard B., Calvo, J.C., Janzen, D.H., 2005. Dynamics in landscape structure and composition for the Chorotega region, Costa Rica from 1960 to 2000. Agriculture Ecosystems & Environment 106, 2739. Aung, M., Swe, K.K., Oo, T., Moe, K.K., Leimgruber, P., Allendorf, T., Duncan, C., Wemmer, C., 2004. The environmental history of Chatthin Wildlife Sanctuary, a protected area in Myanmar (Burma). Journal of Environmental Management 72, 205216. Balmford, A., Bennun, L., ten Brink, B., Cooper, D., Cote, I.M., Crane, P., Dobson, A., Dudley, N., Dutton, I., Green, R.E., Gregory, R.D., Harrison, J., Kennedy, E.T., Kremen, C., Leader Williams, N., Lovejoy, T.E., Mace, G., May, R., Mayaux, P., Morling, P., Phillips, J., Redford, K., Ricketts, T.H., Rodriguez, J.P., Sanjayan, M., Schei, P.J., van Jaarsveld, A.S., Walther, B.A., 2005. The convention on biological diversity's 2010 target. Science 307, 212213. Baral, N., Heinen, J.T., 2006. The Maoist peoples war and conservation in Nepal. Politics and the Life Sciences 24. Barlow, A.C.D., McDougal, C., Smith, J.L.D., Gurung, B., Bhatta, S.R., Kumal, S., Mahato, B., Tamang, D.B., 2009. Temporal variation in tiger (Panthera tigris) populations and its implications for monitoring. Journal of Mammalogy 90, 472478.
134 Barr, C., 2001. P rofits on paper: the political economy of fiber, finance, and debt in Indonesias pulp and paper industries. In: Barr, C. (Ed.), Banking on sustainability: structural adjustment and forestry reform in post Suharto Indonesia. Center for International Forest ry and World Wildlife Fund International, Macroeconomics Program Office, Washington, D.C., pp. 6099. Barve, N., Kiran, M.C., Vanaraj, G., Aravind, N.A., Rao, D., Shaanker, R.U., Ganeshaiah, K.N., Poulsen, J.G., 2005. Measuring and mapping threats to a wil dlife sanctuary in southern India. Conservation Biology 19, 122130. Bhat, D.M., Murali, K.S., Ravindranath, N.H., 2001. Formation and recovery of secondary forests in India: a particular reference to Western Ghats in South India. Journal of Tropical Fores t Science 13, 601620. Boegh, E., Soegaard, H., Broge, N., Hasager, C.B., Jensen, N.O., Schelde, K., Thomsen, A., 2002. Airborne multispectral data for quantifying leaf area index, nitrogen concentration, and photosynthetic efficiency in agriculture. Remote Sensing of Environment 81, 179193. Bowles, I.A., Rice, R.E., Mittermeier, R.A., da Fonseca, G.A.B., 1998. Logging and tropical forest conservation. Science 280, 18991900. Boyd, D.S., Foody, G.M., Curran, P.J., Lucas, R.M., Honzak, M., 1996. An ass essment of radiance in Landsat TM middle and thermal infrared wavebands for the detection of tropical forest regeneration. International Journal of Remote Sensing 17, 249261. Briassoulis, H., 2000. Analysis of land use change: Theoretical and modeling approaches. http://www.rri.wvu.edu/WebBook/Briassoulis/chapter4(models6).htm#4.7.2. Brown, D.G., Pijanowski, B.C., Duh, J.D., 2000. Modeling the relationships between la nd use and land cover on private lands in the Upper Midwest, USA. Journal of Environmental Management 59, 247263. Bruner, A.G., Gullison, R.E., Rice, R.E., da Fonseca, G.A.B., 2001. Effectiveness of Parks in Protecting Tropical Biodiversity. Science 291, 125128. Bullock, S.H., Mooney, H.A., Medina, E., 1995. Seasonally dry tropical forests. Cambridge University Press, Cambridge. Champion, H.G., Seth, S.K., 1968. A Revised Survey of the Forest Types of India. Government of India Publications, New Delhi, India. Chowdhury, M.S.H., Koike, M., Muhammed, N., 2009. Embracing collaborative protected area management for conservation: an analysis of the development of the forest policy of Bangladesh. International Forestry Review 11, 359374.
135 Coggins, C.R., 2000. Wi ldlife conservation and bamboo management in China's Southeast uplands. Geogr. Rev. 90, 83111. Cohen, W.B., Spies, T.A., 1992. Estimating structural attributes of douglas fir western hemlock forest stands from Landsat and SPOT imagery. Remote Sensing of E nvironment 41, 117. Cohen, W.B., Spies, T.A., Fiorella, M., 1995. Estimating the age and structure of forests in a multi ownership landscape of western Oregon, USA. International Journal of Remote Sensing 16, 721746. Cushman, S.A., Wallin, D.O., 2000. Ra tes and patterns of landscape change in the Central Sikhotealin Mountains, Russian Far East. Landscape Ecology 15, 643659. Cushman, S.A., Gutzweiler, K., Evans, J.S., McGarigal, K., 2010. The Gradient Paradigm: A Conceptual and Analytical Framework for Landscape Ecology. In: Cushman, S.A., Huettmann, F. (Eds.), Spatial Complexity, Informatics, and Wildife Conservation. Springer, pp. 83108. Davidar, P., Arjunan, M., Mammen, P.C., Garrigues, J.P., Puyravaud, I.P., Roessingh, K., 2007. Forest degradation in the Western Ghats biodiversity hotspot: Resource collection, livelihood concerns and sustainability. Current Science 93, 15731578. de Jong, W., Freitas, L., Baluarte, J., van de Kop, P., Salazar, A., Inga, E., Melendez, W., Germana, C., 2001. Secondary f orest dynamics in the Amazon floodplain in Peru. Forest Ecology and Management 150, 135146. DeFries, R., Hansen, A., Newton, A.C., Hansen, M.C., 2005. Increasing isolation of protected areas in tropical forests over the past twenty years. Ecological Applications 15, 1926. DeFries, R., Hansen, A., Turner, B.L., Reid, R., Liu, J., 2007. Land use change around protected areas: Management to balance human needs and ecological function. Ecological Applications 17, 10311038. DeFries, R., Karanth, K.K., Pa reeth, S., 2010. Interactions between protected areas and their surroundings in humandominated tropical landscapes. Biological Conservation 143, 28702880. Dinerstein, E., Loucks, C., Heydlauff, A., Wikramanayake, E., Bryja, G., Forrest, J., Ginsberg, J. Klenzendorf, S., Leimgruber, P., OBrien, T., Sanderson, E., Seidensticker, J., Songer, M., 2006. Setting Priorities for the Conservation and Recovery of Wild Tigers: 2005 2015. A Users Guide. In. WWF, WCS, Smithsonian, and NFWFSTF, Washington, D.C. New York.
136 Fan, F.L., Wang, Y.P., Wang, Z.S., 2008. Temporal and spatial change detecting (19982003) and predicting of land use and land cover in Core corridor of Pearl River Delta (China) by using TM and ETM+ images. Environmental Monitoring and Assessmen t 137, 127147. FAO, 1993. Tropical Forest Resource Assessment 1990. In, FAO Forestry Paper No. 112. Food and Agricultural Organization of the United Nations, Rome.
137 Garca Fernndez, C., Ruiz Prez, M., Wunder, S., 2008. Is multipleuse forest management widely implementable in the tropics? Forest Ecology and Management 256, 14681476. Gardner, R.H., Engelhardt, K.A.M., 2008. Spatial processes that maintain biodiversity in plant communities. Perspectives in Plant Ecology Evolution and Systematics 9, 211228. Garrigues, S., Allard, D., Baret, F., Weiss, M., 2006. Quantifying spatial heterogeneity at the landscape scale using variograrn models. Remote Sensing of Environment 103, 8196. Garrigues, S., Allard, D., Baret, F., Morisette, J., 2008. Multivariate quantification of landscape spatial heterogeneity using variogram models. Remote Sensing of Environment 112, 216230. Garson, G.D., 2008. Statnotes: Topics in multivariate analysis. http://faculty.chass.ncsu.edu/garson/PA765/statnote.htm Gaveau, D.L.A., Wandono, H., Setiabudi, F., 2007. Three decades of deforestation in southwest Sumatra: Have protected areas halted forest loss and logging, and promoted regrowth? Biological Conservation 134, 495504. Gaveau, D.L.A., Linkie, M., Suyadi, Levang, P., Leader Williams, N., 2009. Three decades of deforestation in southwest Sumatra: Effects of coffee prices, law enforcement and rural poverty. Biological Conservation 142, 597605. Geist, H.J., Lambin, E.F., 2002. Proximate causes and underlying driving forces of tropical deforestation Bioscience 52, 143150. Ghate, R., 1992. Forest Policy and tribal development: A study of Maharashtra. Concept Publishing Company, New Delhi. Gillespie, T.W., Zutta, B.R., Early, M.K., Saatchi, S., 2006. Predicting and quantifying the structure of tropical dry forests in South Florida and the Neotropics using spaceborne imagery. Global Ecology and Biogeography 15, 225236. Gillespie, T.W., Saatchi, S., Pau, S., Bohlman, S., Giorgi, A.P., Lewis, S., 2009. Towards quantifying tropical tree species richness in tropical forests. International Journal of Remote Sensing 30, 16291634. Giri, C., Pengra, B., Zhu, Z.L., Singh, A., Ti eszen, L.L., 2007. Monitoring mangrove forest dynamics of the Sundarbans in Bangladesh and India using multi temporal satellite data from 1973 to 2000. Estuarine Coastal and Shelf Science 73, 91100. Giriraj, A., Irfan Ullah, M., Murthy, M.S.R., Beierkuhnlein, C., 2008. Modelling Spatial and Temporal Forest Cover Change Patterns (19732020): A Case Study from South Western Ghats (India). Sensors 8, 61326153.
138 Global Tiger Initiative (GTI) website, 2010: last accessed Octo ber 20, 2010. http://www.globaltigerinitiative.org/ Gopal, B., Chauhan, M., 2006. Biodiversity and its conservation in the Sundarban Mangrove Ecosystem. Aquat. Sci. 68, 338354. Government of Bangladesh (GoB ), 1994. National for estry policy, 1994. In. Ministry of Environment and Forest, Government of Bangladesh. Green, G.M., Sussman, R.W., 1990. Deforestation history of the eastern rain forests of Madagascar from satellite images. Science 248, 212215. Green, G.M., Schweik, C.M., Randolf, J.C., 2005. Retrieving landcover change information from Landsat satellite images by minimizing other sources of reflectance variability. In: Moran, E.F., Ostrom, E. (Eds.), Seeing the forest and the trees: Human environment interactions in for est ecosystems. MIT Press, Cambridge, MA pp. 131160. Griffiths, M., van Schaik, C.P., 1993. The impact of human traffic on the abundance and activity periods of Sumatran rainforest wildlife. Conservation Biology 7, 623626. Gurung, K.K., 1983. Heart of t he Jungle. Andre Deutsch, London. Gurung, B., Smith, J.L.D., McDougal, C., Karki, J.B., Barlow, A., 2008. Factors associated with human killing tigers in Chitwan National Park, Nepal. Biological Conservation 141, 30693078. Gutter, P., 2001. Environment and law in Burma. Legal Issues on Burma Journal 9, 1 28. Haines Young, R., Chopping, M., 1996. Quantifying landscape structure: a review of landscape indices and their application to forested landscapes. Progress in Physical Geography 20, 418445. Hartter, J., Southworth, J., 2009. Dwindling resources and fragmentation of landscapes around parks: wetlands and forest patches around Kibale National Park, Uganda. Landscape Ecology 24, 643656. Hayes, T.M., 2006. Parks, People, and Forest Protection: An Institut ional Assessment of the Effectiveness of Protected Areas. World Development 34, 20642075. Healey, S.P., Cohen, W.B., Yang, Z.Q., Krankina, O.N., 2005. Comparison of Tasseled Cap based Landsat data structures for use in forest disturbance detection. Remote Sensing of Environment 97, 301310. Hegde, R., Suryaprakash, S., Achoth, L., Bawa, K.S., 1996. Extraction of nontimber forest products in the forests of Biligiri Rangan Hills, India .1. Contribution to rural income. Econ. Bot. 50, 243251.
139 Heywood, V.H., 1995. The global biodiversity assessment. UNEP, University Press, Cambridge. Hoechstetter, S., Walz, U., Dang, L.H., Thinh, N.X., 2008. Effects of topography and surface roughness in analyses of landscape structurea proposal to modify the existing set of landscape metrics. Landscape Online 1, 114. Holmes, D.A., 2002. The predicted extinction of lowland forests in Indonesia. In: Wickramanayake, E., Dinerstein, E., Loucks, C.J., Olson, D.M., Morrison, J., Lamoreux, J., McKnight, M., Hedao, P. (Eds.), Terre strial ecoregions of the IndoPacific: a conservation assessment. Island Press, Washington, D.C. Houghton, R.A., 1994. The Worldwide Extent of LandUse Change. Bioscience 44, 305313. Huete, A.R., Didan, K., Shimabukuro, Y.E., Ratana, P., Saleska, S.R., Hu tyra, L.R., Yang, W.Z., Nemani, R.R., Myneni, R., 2006. Amazon rainforests greenup with sunlight in dry season. Geophysical Research Letters 33. Hussain, Z., Acharya, G., 1994. Mangroves of the Sundarbans. Volume 2: Bangladesh. World Conservation Union, G land. Iftekhar, M.S., 2006. Forestry in Bangladesh: An overview. J. For. 104, 148153. Imam, E., Kushwaha, S.P.S., Singh, A., 2009. Evaluation of suitable tiger habitat in Chandoli National Park, India, using multiple logistic regression. Ecological Modelling 220, 36213629. IUCN, 1999. Threats to forest protected areas: A survey of 10 countries carried out in association with the World Commission on Protected Areas. In. Research Report from IUCN The World Conservation Union for The WWF/World Bank Al liance for Conservation and Sustainable Use. Janzen, D.H., 1988. Tropical dry forests: the most endangered major tropical ecosystem. In: Wilson, E.O., Peters, F.M. (Eds.), Biodiversity. National Academy Press, Washington. Jensen, J.R., Rutchey, K., Koch, M.S., Narumalani, S., 1995. Inland wetland change detection in the everglades water conservation area 2a using a timeseries of normalized remotely sensed data. Photogrammetric Engineering and Remote Sensing 61, 199209. Jiang, Z.Y., Huete, A.R., Didan, K. Miura, T., 2008. Development of a twoband enhanced vegetation index without a blue band. Remote Sensing of Environment 112, 38333845.
140 Jin, S.M., Sader, S.A., 2005. Comparison of time series tasseled cap wetness and the normalized difference moisture index in detecting forest disturbances. Remote Sensing of Environment 94, 364372. Johnson, B.R., Young, S.J., 1998. InScene Atmospheric Compensation: Application to SEBASS Data Collected at the ARM Site. Space and Environment Technology Center, The Aerospace Corporation. Karanth, K.U., Stith, B.M., 1999. Prey depletion as a critical determinant of tiger populations. In: Seidensticker, J., Christie, S., Jackson, P. (Eds.), Riding the Tiger: Tiger Conservation in Humandominated Landscapes. Cambridge Universi ty Press, Cambridge, UK, pp. 100113. Karanth, K.U., Sunquist, M.E., 2000. Behavioural correlates of predation by tiger (Panthera tigris), leopard (Panthera pardus) and dhole (Cuon alpinus) in Nagarahole, India. Journal of Zoology 250, 255265. Karanth, K. K., Curran, L.M., ReuningScherer, J.D., 2006. Village size and forest disturbance in Bhadra Wildlife Sanctuary, Western Ghats, India. Biological Conservation 128, 147157. Kauppi, P.E., Ausubel, J.H., Fang, J.Y., Mather, A.S., Sedjo, R.A., Waggoner, P.E., 2006. Returning forests analyzed with the forest identity. Proceedings of the National Academy of Sciences of the United States of America 103, 1757417579. Kawabata, A., Ichii, K., Yamaguchi, Y., 2001. Global monitoring of interannual changes in vegetati on activities using NDVI and its relationships to temperature and precipitation. International Journal of Remote Sensing 22, 13771382. Kerley, L.L., Goodrich, J.M., Miquelle, D.G., Smirnov, E.N., Quigley, H.B., Hornocker, N.G., 2002. Effects of roads and human disturbance on Amur tigers. Conservation Biology 16, 97108. Khan, M.M.H., Chivers, D.J., 2007. Habitat preferences of tigers Panthera tigris in the Sundarbans East Wildlife Sanctuary, Bangladesh, and management recommendations. Oryx 41, 463468. Kin naird, M.F., Sanderson, E.W., O'Brien, T.G., Wibisono, H.T., Woolmer, G., 2003. Deforestation Trends in a Tropical Landscape and Implications for Endangered Large Mammals. Conservation Biology 17, 245257. Kondrashov, L.G., 2004. Russian Far East forest di sturbances and socioeconomic problems of restoration. Forest Ecology and Management 201, 6574. Kothari, A., Pande, P., Singh, S., Variava, D., 1989. Management of national parks and sanctuaries in India: a status report. Indian Institute of Public Admini stration, Delhi, India.
141 Kuijper, D.P.J., Cromsigt, J., Churski, M., Adam, B., Jedrzejewska, B., Jedrzejewski, W., 2009. Do ungulates preferentially feed in forest gaps in European temperate forest? Forest Ecology and Management 258, 15281535. Kulkarni, S. 1987. Forest Legislation and Tribals: Comments on Forest Policy Resolution. Economic and Political Weekly 22, 21432148. Kumar, R., Shahabuddin, G., 2005. Effects of biomass extraction on vegetation structure, diversity and composition of forests in Sari ska Tiger Reserve, India. Environmental Conservation 32, 248259. Kustas, W.P., Norman, J.M., 1996. Use of remote sensing for evapotranspiration monitoring over land surfaces. Hydrol. Sci. J. J. Sci. Hydrol. 41, 495516. Lamb, D., Gilmour, D., 2003. Rehabi litation and Restoration of Degraded Forests. In. International Union for Conservation of Nature (IUCN) and the World Wildlife Fund (WWF), Gland, Switzerland and Cambridge, UK. http://app.iucn.org/dbtw wpd/edocs/FR IS005.pdf#search=%22rehabilitation%20and%20restoration%20of%20degraded% 20forests%22. Lambin, E.F., Turner, B.L.I., Geist, H.J., Agbola, S.B., Angelsen, A., Bruce, J.W., Coomes, O., Dirzo, R., Fischer, G., Folke, C., George, P.S., Homewodd, K., Imbernon, J., Leemans, R., Li, X., Moran, E.F., Mortimore, M., Ramakrishnan, P.S., Richards, J.F., Sknes, H., Steffen, W., Stone, G.D., Svedin, U., Veldkamp, T.A., Vog el, C., Xu, J., 2001. The Causes of LandUse and LandCover Change. Moving Beyond the Myths. Global Environmental Change: Human and Policy Dimensions 11, 261269. Laurance, W.F., 1999. Reflections on the tropical deforestation crisis. Biological Conservati on 91, 109117. Lee, J., 2004. Intrinsic adhesion force of lubricants to steel surface. Journal of Pharmaceutical Sciences 93, 23102318. Leuning, R., Cleugh, H.A., Zegelin, S.J., Hughes, D., 2005. Carbon and water fluxes over a temperate Eucalyptus forest and a tropical wet/dry savanna in Australia: measurements and comparison with MODIS remote sensing estimates. Agric. For. Meteorol. 129, 151173. Li, H., Reynolds, J.F., 1997. Modeling effects of spatial pattern, drought, and grazing on rates of rangeland degradation: A combined Markov and cellular automaton approach, In Scale in Remote Sensing and GIS. eds D.A. Quattrochi, M.F. Goodchild. Lewis Publishers, New York. Linkie, M., Smith, R.J., Leader Williams, N., 2004. Mapping and predicting deforestation p atterns in the lowlands of Sumatra. Biodiversity and Conservation 13, 18091818.
142 Linkie, M., Haidir, W.A., Nugroho, A., Dinata, Y., 2008. Conserving tigers Panthera tigris in selectively logged Sumatran forests. Biological Conservation 141, 24102415. Liu, D.C., 2001. Tenure and management of nonstate forests in China since 1950 A historical review. Environmental History 6, 239263. Lu, D., Mausel, P., Brondizio, E., Moran, E., 2003. Classification of successional forest stages in the Brazilian Amazon ba sin. Forest Ecology and Management 181, 301312. Lu, D.S., Mausel, P., Brondizio, E., Moran, E., 2004. Relationships between forest stand parameters and Landsat TM spectral responses in the Brazilian Amazon Basin. Forest Ecology and Management 198, 149167. Lugo, A.E., Helmer, E., 2004. Emerging forests on abandoned land: Puerto Rico's new forests. Forest Ecology and Management 190, 145161. Lynam, A.J., Khaing, S.T., Zaw, K.M., 2006. Developing a national tiger action plan for the union of Myanmar. Environmental Management 37, 3039. Lynam, A., Rabinowitz, A., Myint, T., Maung, M., Latt, K., Po, S., 2009. Estimating abundance with sparse data: tigers in northern Myanmar. Popul. Ecol. 51, 115121. MacKinnon, K., Mishra, H., Mott, J., 1999. Reconciling the needs of conservation and local communities: Global Environment Facility support for tiger conservation in India. In: Seidensticker, J., Christie, S., Jackson, P. (Eds.), Riding the Tiger: Tiger Conservation in Humandominated Landscapes. Cambridge University Press, Cambridge, UK, pp. 307315. Madhusudan, M.D., 2003. Living amidst large wildlife: Livestock and crop depredation by large mammals in the interior villages of Bhadra Tiger Reserve, south India. Environmental Management 31, 466475. Madhusudan, M.D., 2004. Recovery of wild large herbivores following livestock decline in a tropical Indian wildlife reserve. Journal of Applied Ecology 41, 858869. Mahanty, S., 2002. Conservation and development interventions as networks: The case of the India Ecodevelopment Project, Karnataka. World Development 30, 13691386. Marshall, E., Randhir, T.O., 2008. Spatial modeling of land cover change and watershed response using Markovian cellular automata and simulation. Water Resources Research 44. Martin, E., Martin, C., Vigne, L., 2008. Recent political disturbances in Nepal threaten rhinos: lessons to be learned. Pachyderm, 98107.
143 Masek, J.G., Huang, C.Q., Wolfe, R., Cohen, W., Hall, F., Kutler, J., Nelson, P., 2008. North American forest disturbance mapped from a decadal Landsat record. Remote Sensing of Environment 112, 29142926. Mather, A.S., 2007. Recent Asian forest transitions in relation to forest transition theory. International Forestry Review 9, 491502. McGarigal, K., Marks, B.J., 1995. FRAGSTATS: Sp atial pattern analysis program for quantifying landscape structure. U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, OR. McGarigal, K., Cushman, S.A., 2005. The gradient concept of landscape structure. In: Wiens J., Moss, M. (Eds.), Issues and perspectives in landscape ecology. Cambridge University Press, Cambridge, pp. 112119. McGarigal, K., Tagil, S., Cushman, S.A., 2009. Surface metrics: an alternative to patch metrics for the quantification of landscape str ucture. Landscape Ecology 24, 433450. Mecikalski, J.R., Diak, G.R., Anderson, M.C., Norman, J.M., 1999. Estimating fluxes on continental scales using remotely sensed data in an atmospheric land exchange model. J. Appl. Meteorol. 38, 13521369. Mehta, V.K., Sullivan, P.J., Walter, M.T., Krishnaswamy, J., DeGloria, S.D., 2008. Ecosystem impacts of disturbance in a dry tropical forest in southern India. Ecohydrology 1, 149160. Melesse, A.M., Weng, Q.H., Thenkabail, P.S., Senay, G.B., 2007. Remote sensing sensors and applications in environmental resources mapping and modelling. Sensors 7, 32093241. Mertens, B., Lambin, E.F., 2000. Landcover change trajectories in southern Cameroon. Annals of the Association of American Geographers 90, 467494. Miao, G.P., West, R.A., 2004. Chinese collective forestlands: contributions and constraints. International Forestry Review 6, 282298. Millington, J.D.A., Walters, M.B., Matonis, M.S., Liu, J.G., 2010. Effects of local and regional landscape characteristics on wildlif e distribution across managed forests. Forest Ecology and Management 259, 11021110. MoEF, 1988. National Forest Policy (1988). Ministry of Environment and Forests, Government of India New Delhi. Mondal, P., Southworth, J., 2010. Protection vs. commercial management: Spatial and temporal analysis of land cover changes in the tropical forests of Central India. Forest Ecology and Management 259, 10091017.
144 Moran, E.F., Brondizio, E.S., 1998. Land use change after deforestation in Amazonia. In: Liverman, D., Moran, E.F., Rindfuss, R.R., Stern, P.C. (Eds.), People and Pixels: Linking Remote Sensing and Social Science. National Academy Press, Washington, DC, pp. 94120. Muller, F., 1998. Gradients in ecological systems. Ecological Modelling 108, 321. Munroe, D.K ., Nagendra, H., Southworth, J., 2007. Monitoring landscape fragmentation in an inaccessible mountain area: Celaque National Park, Western Honduras. Landscape and Urban Planning 83, 154167. Munz, M., Cappella, B., Sturm, H., Geuss, M., Schulz, E., 2003. Materials contrasts and nanolithography techniques in scanning force Microscopy (SFM) and their application to polymers and polymer composites. In, Filler Reinforced Elastomers Scanning Force Microscopy. Springer, pp. 87210. Murphy, P.G., Lugo, A.E., 1986. Ecology of Tropical Dry Forest. Annual Review of Ecology and Systematics 17, 6788. NASA. 2007. Science Plan for NASAs Science Mission Directorate 20072016. http://nasascience.nasa.gov/ about us/science strategy Nagendra, H., 2009. Drivers of regrowth in South Asia's human impacted forests. Current Science 97, 15861592. Nagendra, H., 2008. Do parks work? Impact of protected areas on land cover clearing. Ambio 37, 330337. Nagendra, H., Karmacharya, M., Karna, B., 2005. Evaluating forest management in Nepal: Views across space and time. Ecology and Society 10. Nagendra, H., Pareeth, S., Ghate, R., 2006. People within parks forest villages, landcover change and landscape fragmentation in the Tadoba Andhari Tiger Reserve, India. Applied Geography 26, 96112. Nath, P.C., Arunachalam, A., Khan, M.L., Arunachalam, K., Barbhuiya, A.R., 2005. Vegetation analysis and tree population structure of tropical wet evergreen forests in and around Namda pha National Park, northeast India. Biodiversity and Conservation 14, 21092136. N ational Wildlife Database, 2009: last accessed September 13, 2010. http://oldwww.wii.gov.in/nwdc/ NaughtonTreves, L., Holland, M.B., Brandon, K., 2005. The Role of Protected Areas in Conserving Biodiversity and Sustaining Local Livelihoods. Annual review of environment and resources 30, 219252.
145 Nayak, P.K., Berkes, F., 2008. Politics of cooptation: Community forest management ve rsus joint forest management in Orissa, India. Environmental Management 41, 707718. Nepstad, D., Schwartzman, S., Bamberger, B., Santilli, M., Ray, D., Schlesinger, P., Lefebvre, P., Alencar, A., Prinz, E., Fiske, G., Rolla, A., 2006. Inhibition of Amazon deforestation and fire by parks and indigenous lands. Conservation Biology 20, 6573. Nyhus, P.J., Tilson, R., 2004. Characterizing humantiger conflict in Sumatra, Indonesia: implications for conservation. Oryx 38, 6874. Oberleithner, H., Ludwig, T., Ri ethmuller, C., Hillebrand, U., Albermann, L., Schafer, C., Shahin, V., Schillers, H., 2004. Human endothelium: Target for aldosterone. Hypertension 43, 952956. Ogra, M., Badola, R., 2008. Compensating HumanWildlife Conflict in Protected Area Communities: Ground Level Perspectives from Uttarakhand, India. Human Ecology 36, 717729. Ostrom, E., Nagendra, H., 2006. Insights on linking forests, trees, and people from the air, on the ground, and in the laboratory. Proceedings of the National Academy of Science s of the United States of America 103, 1922419231. Pandey, D., 1992. Assessment of tropical forest plantation resources. Unpublished report to Forestry Department, Food and Agriculture Organization of the United Nations, Rome. Department of Forest Survey, Swedish University of Agricultural Sciences. Panta, M., Kim, K., Joshi, C., 2008. Temporal mapping of deforestation and forest degradation in Nepal: Applications to forest conservation. Forest Ecology and Management 256, 15871595. Pearce, D., Putz, F.E., Vanclay, J.K., 2003. Sustainable forestry in the tropics: panacea or folly? Forest Ecology and Management 172, 229247. Pelkey, N.W., Stoner, C.J., Caro, T.M., 2000. Vegetation in Tanzania: assessing long term trends and effects of protection using satell ite imagery. Biological Conservation 94, 297309. Perz, S.G., Skole, D.L., 2003. Social determinants of secondary forests in the Brazilian Amazon. Social Science Research 32, 2560. Peterson, L.K., Bergen, K.M., Brown, D.G., Vashchuk, L., Blam, Y., 2009. F orested landcover patterns and trends over changing forest management eras in the Siberian Baikal region. Forest Ecology and Management 257, 911922.
146 Pinstrup Anderson, P., and R. PandyaLorch. 1998. Food security and sustainable use of natural resources: a 2020 vision. Ecological Economics 26 (1): 110. Prasad, R., 2000. Sustainable forest management for dry forests of South Asia. In: Cheng, T.L., Durst, P.B. (Eds.), Development of national level criteria and indicators for the sustainable management of dry forests of Asia: background papers. RAP Publishers/FAO, Thailand, pp. 2774. Putz, F.E., Dykstra, D.P., Heinrich, R., 2000. Why poor logging practices persist in the tropics. Conservation Biology 14, 951956. Putz, F.E., Blate, G.M., Redford, K.H., Fimb el, R., Robinson, J., 2001. Tropical forest management and conservation of biodiversity: an overview. Conservation Biology 15, 720. Rai, S.C., Sharma, E., 1998. Comparative assessment of runoff characteristics under different land use patterns within a Hi malayan watershed. Hydrological Processes 12, 22352248. Ranganathan, J., Chan, K.M.A., Karanth, K.U., Smith, J.L.D., 2008. Where can tigers persist in the future? A landscape scale, density based population model for the Indian subcontinent. Biological Co nservation 141, 6777. Ravan, S., Dixit, A.M., Mathur, V.B., 2005. Spatial analysis for identification and evaluation of forested corridors between two protected areas in Central India. Current Science 88, 14411448. Rayan, D.M., Mohamad, S.W., 2009. The importance of selectively logged forests for tiger Panthera tigris conservation: a population density estimate in Peninsular Malaysia. Oryx 43, 4851. Redford, K.H., 1992. The empty forest. Bioscience 42, 412422. Rice, R.E., Gullison, R.E., Reid, J.W., 19 97. Can sustainable management save tropical forests? Scientific American 276, 4449. Richards, J.F., 1990. Agricultural impacts in tropical wetlands: Rice paddies for mangroves in south and southeast Asia. In: Williams, M. (Ed.), Wetlands: A Threatened Landscape. Basil Blackwell, Oxford. Robbins, P., McSweeney, K., Waite, T., Rice, J., 2006. Even conservation rules are made to be broken: Implications for biodiversity. Environmental Management 37, 162169. Robbins, P.F., Chhangani, A.K., Rice, J., Trigosa, E., Mohnot, S.M., 2007. Enforcement authority and vegetation change at Kumbhalgarh wildlife sanctuary, Rajasthan, India. Environmental Management 40, 365378.
147 Robertson, J.M.Y., van Schaik, C.P., 2001. Causal factors underlying the dramatic decline of the Sumatran orangutan. Oryx 35, 2638. Rocha, A.V., Shaver, G.R., 2009. Advantages of a two band EVI calculated from solar and photosynthetically active radiation fluxes. Agric. For. Meteorol. 149, 15601563. Rodgers, W.A., 1991. Managing natural resources for the maintenance of biological diversity in India. In: Gopal, B. (Ed.), Ecology and Sustainable Development. National Institute of Ecology, New Delhi, India, pp. 8183. Rudel, T.K., 2005. Tropical Forests: Regional Paths of Destruction and Regeneration in the Late Twentieth Century. Columbia University Press, New York. Sagar, R., Singh, J.S., 2004. Local plant species depletion in a tropical dry deciduous forest of northern India. Environmental Conservation 31, 5562. Salam, M.A., Noguchi, T., Koike, M., 2000. Understanding why farmers plant trees in the homestead agroforestry in Bangladesh. Agroforestry Systems 50, 7793. Salisbury, J.W., D'Aria, D.M., 1992. Emissivity of terrestrial materials in the 8 14 [mu]m atmospheric window. Remote Sensing of Envir onment 42, 83106. Samal, P.K., Palni, L.M.S., Agarwal, D.K., 2003. Ecology, ecological poverty and sustainable development in Central Himalayan region of India. International Journal of Sustainable Development and World Ecology 10, 157168. Sanchez Azo feifa, G.A., Daily, G.C., Pfaff, A.S.P., Busch, C., 2003. Integrity and isolation of Costa Rica's national parks and biological reserves: examining the dynamics of landcover change. Biological Conservation 109, 123135. Sarma, P.K., Lahkar, B.P., Ghosh, S ., Rabha, A., Das, J.P., Nath, N.K., Dey, S., Brahma, N., 2008. Landuse and landcover change and future implication analysis in Manas National Park, India using multi temporal satellite data. Current Science 95, 223227. Seidensticker, J., 1987. Bearing witness: Observations on the extinction of Panthera tigris balica and Panthera tigris sondaica. In: Tilson, R.L., Seal, U.S. (Eds.), Tigers of the world: the biology, biopolitics, management, and conservation of an endangered species. Noyes Publications, P ark Ridge, New Jersey, pp. 18. Seidensticker, J., Kurin, R., Townsend, A.K., 1991. The Commons in South Asia: Societal Pressures and Environmental Integrity in the Sundarbans. The International Center, Smithsonian Institution, Washington, D.C.
148 Sellers, P.J., Dickinson, R.E., Randall, D.A., Betts, A.K., Hall, F.G., Berry, J.A., Collatz, G.J., Denning, A.S., Mooney, H.A., Nobre, C.A., Sato, N., Field, C.B., Henderson Sellers, A., 1997. Modeling the exchanges of energy, water, and carbon between continents and the atmosphere. Science 275, 502509. Shahabuddin, G., 2010. Conservation at the crossroads: Science, society, and the future of India's wildlife. Permanent Black, Delhi. Sharma, R., DeCosse, P., Roy, M., Khan, M., Mazumder, A., 2008. Comanagement of protected areas in south Asia with special reference to Bangladesh. In. Nishorgo Support Project, Bangladesh Forest Department (BFD), Ministry of Environment and Forest, Government of Bangladesh. Shmatkov, N., Brigham, T., 2003. Nontimber forest products in community development: lessons from the Russian Far East. Forestry Chronicle 79, 113118. Shukla, J., Nobre, C., Sellers, P., 1990. Amazon Deforestation and Climate Change. Science 247, 13221325. Sidle, R.C., Ziegler, A.D., Negishi, J.N., Nik, A.R., Siew, R., Turkelboom, F., 2006. Erosion processes in steep terrain Truths, myths, and uncertainties related to forest management in Southeast Asia. Forest Ecology and Management 224, 199225. Silvertown, J., Holtier, S., Johnson, J., Dale, P., 1992. Cell ular automaton models of interspecific competition for space: The effect of pattern on process. Journal of Ecology 80, 527534. Sims, D.A., Luo, H., Hastings, S., Oechel, W.C., Rahman, A.F., Gamon, J.A., 2006. Parallel adjustments in vegetation greenness and ecosystem CO2 exchange in response to drought in a Southern California chaparral ecosystem. Remote Sensing of Environment 103, 289303. Singh, G., Velmurugan, A., Dakhate, M.P., 2009. Geospatial approach for tiger habitat evaluation and distribution in Corbett Tiger reserve, India. J. Indian Soc. Remote Sens. 37, 573585. Smith, J.L.D., Ahern, S.C., McDougal, C., 1998. Landscape Analysis of Tiger Distribution and habitat Quality in Nepal. Conservation Biology 12, 13381346. Song, C., Schroeder, T.A., Cohen, W.B., 2007. Predicting temperate conifer forest successional stage distributions with multitemporal Landsat Thematic Mapper imagery. Remote Sensing of Environment 106, 228 237. Songer, M., Aung, M., Senior, B., DeFries, R., Leimgruber, P., 2009. Spatial and temporal deforestation dynamics in protected and unprotected dry forests: a case study from Myanmar (Burma). Biodiversity and Conservation 18, 10011018.
149 Soule, M.E., Sanjayan, M.A., 1998. Ecology conservation targets: Do they help? Science 279, 20602061. Southworth, J., 2004. An assessment of Landsat TM band 6 thermal data for analysing land cover in tropical dry forest regions. International Journal of Remote Sensing 25, 689706. SPIP The scanning probe image processor Image metrology, APS, Lyngby: l ast accessed September 13, 2010. http://www.imagemet.com/ Stevens, F.R. 2009. Bridging the Landsat data gap: evaluating ASTER as an alternative. Thesis, University of Florida. Stockli, R., Rutishauser, T., Dragoni, D., O'Keefe, J., Thornton, P.E., Jolly, M., Lu, L., Denning, A.S., 2008. Remote sensing data assimilation for a prognostic phenology model. J. Geophys. Res. Biogeosci. 113, 19.Shmatkov, N., Brigham, T., 2003. Non timber forest products in communit y development: lessons from the Russian Far East. Forestry Chronicle 79, 113118. Sunderlin, W.D., 1999. The effects of economic crisis and political change on Indonesias forest sector, 1997 99. In, Center for International Forestry Research, Bogor, Indonesia. Sunquist, M.E., Sunquist, F.C., 2002. Wild Cats of the World. University of Chicago Press, Chicago, USA. Tilson, R., Defu, H., Muntifering, J., Nyhus, P.J., 2004. Dramatic decline of wild South China tigers Panthera tigris amoyensis: field survey of priority tiger reserves. Oryx 38, 4047. Turner, M.G., 2005. Landscape ecology: What is the state of the science? Annual Review of Ecology Evolution and Systematics 36, 319344. Turner, M.G., Gardner, R.H., O'Neill, R.V., 2001. Landscape ecology in theory and practice: pattern and process. Springer Verlag, New York. Turner, D.P., Urbanski, S., Bremer, D., Wofsy, S.C., Meyers, T., Gower, S.T., Gregory, M., 2003. A cross biome comparison of daily light use efficiency for gross primary production. Global Change Biology 9, 383395. Vitousek, P.M., 1994. Beyond global warming: ecology and global change. Ecology 75, 18611876. Walsh, S.J., Shao, Y., Mena, C.F., McCleary, A.L., 2008. Integration of hyperion satellite data and a household social survey to character ize the causes and consequences of reforestation patterns in the northern Ecuadorian Amazon. Photogrammetric Engineering and Remote Sensing 74, 725735.
150 Whitmore, T.C., 1999. Arguments on the forest frontier. Biodiversity and Conservation 8, 865868. Wibis ono, H.T., Figel, J.J., Arif, S.M., Ario, A., Lubis, A.H., 2009. Assessing the Sumatran tiger Panthera tigris sumatrae population in Batang Gadis National Park, a new protected area in Indonesia. Oryx 43, 634638. Wikramanayake, E., McKnight, M., Dinerstei n, E., Joshi, A., Gurung, B., Smith, D., 2004. Designing a conservation landscape for tigers in humandominated environments. Conservation Biology 18, 839844. Wittemyer, G., Elsen, P., Bean, W.T., Burton, A.C.O., Brashares, J.S., 2008. Accelerated Human P opulation Growth at Protected Area Edges. Science 321. Wolfram, S., 1984. Cellular automata as models of complexity. Nature 311, 419424. Woodman, J., 2004. Between Bureaucrats and Beneficiaries: The Implementation of Ecodevelopment in Pench Tiger Reserves India. Ph. D. dissertation. University of Cambridge, UK. World Bank, 1991: last accessed September 13, 2010. http://web.worldbank.org/external/projects/main?pagePK=64283627&piPK=73230 &theSitePK=40941&menuPK=228424&Projectid=P010390. World Bank, 1996. India ecodevelopment: Project document. Washington, DC. WWF, 2004. How effective are protected areas? A report prepared for the Seventh Conference of Parties of the Convention on Biological Diversity. WWF International, Switzerland. Xu, J.C., Melick, D.R., 2007. Rethinking the effectiveness of public protected areas in southwestern China. Conservation Biology 21, 318328. Yadav, A.S., Gupta, S.K., 2006. Effect of micro environment and human disturbance on the diversity of woody species in the Sariska Tiger Project in India. Forest Ecology and Management 225, 178189. Yarbrough, L.D., Easson, G., Kuszmaul, J.S., 2005. U sing At Sensor Radiance and Reflectance Tasseled Cap Transforms Applied to Change Detection for the ASTER Sensor. In: King, R.L., Younan, N.H. (Eds.), Proceedings of the Third International Workshop on the Analysis of Multi Temporal Remote Sensing Images, Mississippi, USA. Zoran, M., Anderson, E., 2006. The use of multi temporal and multispectral satellite data for change detection analysis of the Romanian Black Sea coastal zone. Journal of Optoelectronics and Advanced Materials 8, 252256.
151 BIOGRAPHICAL SKETCH Pinki Mondal was born in 1981 in Calcutta, India. She graduated from high school in 1999. She went to the University of Calcutta to earn her Bachelor of Science degree in 2002 with geology major (honors), and minors in chemistry and mat hematics. In 2004, she earned her Master of Science degree from Jadavpur University, Calcutta in applied geology, with s pecialization in remote sensing. Her Masters thesis revolved around application of remote sensing in measuring land use and land cover c hange in the vulnerable estuarine ecosystems of Sundarbans, India. She chose to attend graduate school in the US to get better exposure to current developments in the fields of land change science, and remote sensing. While preparing for GRE, she completed the course works required to earn a post graduate diploma in computer application from the Department of Electronics and Accreditation of Computer Classes (DOEACC) under the Indian Ministry of Communications & Information Technology. In 2006, she started her graduate studies at the University of Florida to earn a PhD degree in geography with land change science concentration and multi disciplinary minor.