LAND-COVER CHANGE, FRAGMENTATION, AND AGRICULTURE IN SOUTHWEST PUERTO RICO: 1982-2002 By ROBERT DANIEL LOPEZ A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2006
Copyright 2006 by Robert Daniel Lopez
To my parents, Mary and Lenord Lopez, who nurtured me to wonder and be curious about everything; to family and friends near and far; to my wonderful wife Amanda; and to our three felines Bears, Gabe and Tigro.
ACKNOWLEDGMENTS The roster of people that I need to thank and recognize for both putting me in the position of achieving an M.S. degree in geography and actually keeping the mental and physical will to complete the research is very long. I hope to leave no one absent. I want to first thank my committee chair, Dr. Jane Southworth, by whom I came to UF to study land use and land cover change and geographic information science under. Dr Southworthâ€™s greatest contribution to my graduate career was encouraging me to figure things out for myself, although she was always there to offer guidance. I thank Dr. Michael W. Binford and Dr. Stephen R. Perz of the University of Florida and Dr. Tom R. Allen of Old Dominion University for their interdisciplinary insight into my project. I thank my fellow graduate students at UF for the very genial and intriguing chats about everything from Geodatabases to our favorite brands of Caribbean rum. I thank Lin Cassidy, Cerian Gibbes, and Alexis Villejas, graduate colleagues in the Geography department, for their assistance during fieldwork in Puerto Rico. I thank Dr. Graeme Cumming in UFWEC for providing the financial assistance needed to do my fieldwork in Puerto Rico and making available to me decadesâ€™ worth of satellite data covering the island. I thank the Puerto Rico Planning Board for the GIS data and other information. To leave no one out, I thank all my other acquaintances at UF and elsewhere who contributed ideas and encouragement to my research and knowledge. Lastly, I thank my darling wife Amanda; her parents, Ron and Brenda; my parents, Mary and Lenord; my stepfather, Steve Ball; my brothers, Jeff and Jon; my lifelong iv
friends, Trey, Jason, and Mike; and my cats for their endless support and fellowship during my graduate career and life. v
TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iv LIST OF TABLES ...........................................................................................................viii LIST OF FIGURES ...........................................................................................................ix ABSTRACT .........................................................................................................................x CHAPTER 1 INTRODUCTION........................................................................................................1 Study Area....................................................................................................................7 Situation of the Caribbean Basin Islands...............................................................7 Puerto Rico: Site and Situation..............................................................................8 The Spatial Component of LUCC Research...............................................................11 Landscape Fragmentation and Landscape Ecology....................................................13 Research Objectives....................................................................................................15 Previous Research.......................................................................................................16 2 FOREST COVER CHANGE, FRAGMENTATION, AND AGRICULTURE IN SOUTHWEST PUERTO RICO.................................................................................21 Introduction.................................................................................................................21 Research Objectives and Questions............................................................................24 Study Area..................................................................................................................25 Research Methods.......................................................................................................28 Data Acquisition and Processing.........................................................................28 Satellite imagery...........................................................................................28 GIS data........................................................................................................30 Land-cover training samples........................................................................30 Census data...................................................................................................31 Image Classification and Manipulation...............................................................32 Single time-step land-cover classification....................................................32 Accuracy assessment....................................................................................33 Post-classification change detection.............................................................33 Landscape metrics (LM)..............................................................................34 Results.........................................................................................................................35 vi
Land-Cover Condition and Dynamics.................................................................35 Regional results............................................................................................35 Results across municipalities.......................................................................36 Landscape fragmentation.............................................................................37 Human Dynamics: Population, Urban Growth, and Agriculture........................38 Population and urban expansion..................................................................38 Agriculture...................................................................................................38 Discussion and Conclusion.........................................................................................39 3 SUMMARY AND THESIS CONCLUSIONS..........................................................66 LIST OF REFERENCES...................................................................................................69 BIOGRAPHICAL SKETCH.............................................................................................75 vii
LIST OF TABLES Table page 2-1 Landsat-platform satellite imagery used in this study: raw characteristics..............46 2-2 Human dynamic variables used in the research.......................................................47 2-3 Accuracy assessment for 2002 and 1992 land cover classifications........................48 2-4 Regional land-cover classifications and change trajectory results...........................49 2-5 Characteristics of forest cover across the southwestern municipalities...................50 2-6 Class-level landscape metrics for the region............................................................50 2-7 Land cover change outcomes and associated landscape metrics in municipalities..51 2-8 Agriculture dynamics for 1978-2002.......................................................................53 viii
LIST OF FIGURES Figure page 1-1 Puerto Rico in the Caribbean Basin.........................................................................20 2-1 Job sector characteristics of Puerto Rico: 1940-2000..............................................54 2-2 Study area characteristics.........................................................................................55 2-3 Topography and physical regions of the southwest.................................................56 2-4 Population characteristics of the southwest.............................................................57 2-5 Results of the forest and non-forest land cover classifications: 1982-2002.............58 2-6 Land cover change trajectory extent and results for single time steps.....................59 2-7 Twenty-year land cover change outcome for southwest Puerto Rico: 1982-2002..60 2-8 Land cover change trajectory proportions across counties......................................61 2-9 Land cover change outcomes in counties: 1982-2002.............................................62 2-10 Forest dynamics across the six southwestern counties: 1982-2002.........................63 2-11 Distribution of 20-year LCCT outcomes and their associated metrics at the class level..........................................................................................................................64 2-12 Agriculture trends across the study area: 1978-2002...............................................65 ix
Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science LAND COVER CHANGE, FRAGMENTATION, AND AGRICULTURE IN SOUTHWEST PUERTO RICO: 1982-2002 By Robert Daniel Lopez May 2006 Chair: Jane Southworth Cochair: Michael Binford Major Department: Geography In an effort to assess forest-cover dynamics and the effect of human activity on forest change in southwest Puerto Rico, this research used Landsat satellite imagery and a geographic information system to first map forest or non-forest land cover for 1982, 1992, and 2002 and then derived land-cover change trajectories, or outcomes, to quantify and spatially identify these changes across this mostly tropical dry landscape. In addition, landscape metrics were used to infer the state of forest fragmentation over time. And lastly, it incorporated independent variables from National Agriculture Statistics Service (NASS) data to track the changes in agricultural activity in the six municipalities concurrent with the land-cover change data. Changes in agriculture activity were compared to land-cover change outcomes in the counties. Results found that as a region, forest cover did not experience large percentages of re-growth as found in other areas of the island, although forest stability and re-growth were abundant in the northern, higher elevations. The most extensive outcome of land x
cover change was older clearing of forest cover not to be re-grown to forest; counties in higher elevations experienced continuous forest re-growth and decreasing forest fragmentation. Shifting and highly fragmented forest change outcomes were significant in extent and suggest intense small farm activity, possibly tenant-operated as agriculture data trends show. Besides declines from 1978-1982, data suggest few linear decreases in agriculture since 1982 and a complex relationship with the landscape. The research established adequate datasets of general forest cover and its change dynamics over 20 years, but additional classifications are needed to infer change at finer temporal resolution. Agriculture data proved insightful as to ascertaining the human effect on forest cover in the region, but without an explicit land-cover division between non-forest and actual cropland classes, precise and spatially explicit causal relationships are difficult to conclude. Nevertheless, the research contributed to the relatively thin library of case studies focusing on forest research in the Caribbean Basin, the phenomenon of reforestation, and remote sensing derived land cover change trajectories. xi
CHAPTER 1 INTRODUCTION George Perkins Marsh (1801-1882) was of the first to recognize, scientifically observe, quantify, debate, and write about human impacts on the physical landscape and environment (Man and Nature, 1864). Marsh's view of land was an all-encompassing one that considered the natural environment altered by human action. He was well ahead of his time as his work fell decades backstage to industrial, fossil-fuel-driven progress, empire building, world wars and impending globalization. Earlier in 1825, Johann Heinrich von Thnen (1780-1850) had published The Isolated State in which he argued the principle that land was put into the use from which the highest rent could be obtained, thus arranging land uses around a central market place. This perspective held a narrow view of the strict economic function of land. These two early, contrasting theories became the branches from which most modern-day theories of land cover and land use change progressed (Briassoulis, 2000). The industrial first half of the twentieth century in the U.S. saw the sociologists of the Chicago School establish theories of human ecology to the human-land relationship to explain land-use changes and structure of urban land. Christaller's Central Place Theory (1933), the Radial Sector Theory of Hoyt (1939) and Harris and Ullman's Multi-nuclei concept (1945) brought urban and regional planning to the forefront of the land-use change phenomenon. The 1956 International Symposium and book titled Man's Role in Changing the Face of the Earth set forth by leading scientists from geography, anthropology, and history. The attention given to this body of work began progress on 1
2 environmental and natural resource sustainability awareness from a multi-disciplinary perspective. These attentions became of interest in the general public forum in the 1960s and 1970s when U.S. government agencies like the Environmental Protection Agency (EPA) were established and began to implement legislation regarding the health of forests, wetlands, water bodies, as well as the atmosphere and the Earth's subsurface. Towards the end of the 1980s and early 1990s, focus on land-use and land-cover change turned to its effect on global environmental changes. Pioneering works, such as Land Transformation in Agriculture (Wolman and Fournier, 1987), The Earth as Transformed by Human Action (Turner et al., 1990) and Changes in Land Use and Land Cover (Turner and Meyer, 1994) and research agendas set forth by Land Use Land Cover Change (LUCC), International Geosphere-Biosphere Program (IGBP), Human Dimensions of Global Change (HDGC) and the National Aeronautic and Space Administration (NASA) developed concerted multi-disciplinary agencies to address particularly the issue of global environmental change, its component causes and effects and the human agency involved. These entities established research framework and needs of which numerous subsequent research papers answered. The call for interdisciplinary cooperation of the physical and life sciences with the social and behavioral sciences spawned a new paradigm of human dimensions of physical landscape change. The discipline of geography became an instrumental factor in joining the remaining disciplines, given their inclusion of both the human and physical environments. The LUCC research agenda continues strong today. Spatially-explicit and theoretically-complex models are being developed to analyze the input of thousands of
3 land-cover change case studies and output scenarios for use in planning more sustainable outcomes for the environment (Evans et al., 2005). As such, the transformation and change of landscapes over time has been determined to be a major factor involved in altering global, regional and local environmental conditions and the potential sustainable development of natural resources (Turner and Meyer, 1994; Roberts, 1994; Turner et al., 1995). Land-use and land-cover change (LUCC), particularly involving tropical forest cover, has been cited as one of the dominant driving forces affecting global carbon storage change, soil conditions, biodiversity and changing of local and regional economies and cultures (Douglas, 1994; Silver et al., 2000; Tucker and Southworth, 2004). Land-cover is defined as the biophysical surface and near-surface attributes of the earth and includes examples such as forest, bare rock, open water, and grassland. Land-use is how humans use it for their activities, such as for agriculture, settlement, recreation, transportation, and the like (Briassoulis, 2000). It is often challenging to infer land use from land cover; additional ground observations are needed as well as an understanding of why, how and to what extent changes in one lead to changes in the other (Turner & Meyer, 1994). Tropical forest is the most altered forest cover worldwide, losing extent and homogeneity to human activities such as agriculture, natural resource extraction and urban development (Randolph et al., 2005). The Food and Agriculture Organization (FAO) estimate of tropical forest loss throughout the 1990s was 15.2 million ha. (152,000 km 2 ) each year (Lambin et al., 2003). Although the most common change involves clearing of forest, reforestation is also occurring and receives less attention. Forests in
4 developing states need critical attention in the realm of human to physical change research as these states often coincide with being in the tropics and in social and economic instability (Williams, 1994). Houghton (1994), Turner and Meyer (1994) state that the single most important input to the change of global environmental conditions is the change and depletion of forest cover in the Tropics. Forests have the greatest biomass and thus are the largest terrestrial source of carbon sequestration, which essentially stores and filters excess carbon dioxide (CO 2 ) gas and mitigates its release into the atmosphere (Silver et al., 2000; Rudel, 2001). Upon loss of this vital biomass, more carbon is released and less sequestered by forest cover. Forests are the habitat for the greatest number of species of both flora and fauna and when their integrity is decreased, habitat is lost (Randolph et al., 2004). Forests are the source of income and livelihood for many human settlements in both industrial and agricultural state economies. Their unsustainable use results in degradation of both physical and cultural landscapes (Perz, 2003). Most LUCC research focuses on deforestation and its socioeconomic causes and biophysical impacts. In LUCC research, the term 'drivers' is used to indicate the human action that causes change to land cover. A proximate driver is the immediate, physical action of land-cover clearing, like slash and burn, logging, and the like. Distal drivers are the social, economic and political phenomena that cause any given physical transformation of land cover taking place. These can 'drive' from the local, regional or global scale. The most often cited distal drivers of tropical deforestation are human population dynamics, technological change, institutional structure and economic factors (Geist and Lambin, 2002). Amazonian deforestation is now a popular and epitomizing
5 example of seeing temporal loss of forest due to both proximate and distal human action. The "fishbone" pattern of forest clearing was a result of a distant government decision to open frontier land to smallholder farmers; the road built through the area determined the pattern of forest loss over time. Reforestation is a much less well-developed research area of LUCC. Some tropical lands are experiencing significant forest re-growth rates as marginal agriculture declines and industrialization occurs (del Mar Lopez et al., 2001). More often than not, this regrowth is unmanaged; regrowth may also consist of a totally different forest type than what was cleared in the past. Also typical in Latin America and the Caribbean is the conversion from cropland to more lucrative cattle grazing and pasture land (FAO, 2000). These types of change have both dramatic and very subtle but equally important implications for the character of a given ecosystem. LUCC is by nature a multiand interdisciplinary endeavor and thus theories from many research arenas ground the attempt to link human actions to the physical landscape. Briassoulis (2000) and Van Wey et al. (2005) review some of the most relevant and cited of these theories. There rarely is a one-to-one relationship of cause and effect in landcover change. Most contemporary researchers in the LUCC field have found that complex interrelationships between theories and drivers are more prevalent than a single cause (Lambin et al., 2000; Lambin et al., 2001; Briassoulis, 2000). Lambin (2001) suggested that many of these causal relationships are but myths, unsupported by empirical evidence of case study. Census data on demographic, economic and agricultural characteristics are employed in the research of forest change. These data are often the best and most
6 economical source to represent and determine elements of the human sources (both distal and proximate) of landscape change. Issues arise, however, with census data as its methods of survey and sampling vary across time. Census data also tends to be aggregated (usually into levels of administrative boundaries) and inherits error and uncertainty as the aggregation scales upward towards larger geographic extents (Evans and Moran, 2002; Perz, 2003). Agriculture was the first major impact that humans had on land cover change worldwide and began around 10,000 years ago; the sedentary human then built settlements and infrastructure which further altered natural landscapes (Wolman and Fournier, 1987; Richards, 1990). Today, although variable around the globe, forest conversion to agricultural land is the most common land cover transformation. Houghton (1994) stated that cropland in tropical regions doubled in just the last 50 years. The dynamics most often cited in conjunction with forest change involve land in agriculture expanding into and fragmenting forests (agriculture extensification); abandoned cropland reverting back to natural forest cover (Lambin et al., 2000; Keys and McConnell, 2005); and fallowing--the process by which land-cover alternates between crops and forest regrowth every few years (Abizaid and Coomes, 2004). A combined knowledge of the patterns and processes associated with these agriculture dynamics, the human context in which they occur, and the rates and patterns of land-cover change accompanying them is needed to better manage agriculture and forestry and model future scenarios of their combined sustainable existence within the landscape. Research needs in the forum of tropical forest change are noted in Randolph et al. (2005). Beyond just looking at static land-cover classes captured in single time
7 observations, the analyses of land-cover change trajectories in conjunction with landscape fragmentation indices provide the ability to monitor both the temporal and spatial composition of change across a landscape and to link these to elements of social change (Mertens and Lambin, 2000; Southworth et al., 2002). Study Area The geographic context of this thesis is the Caribbean Basin. This region contains a diverse range of cultures, economies, and land-cover histories. The Caribbean Basin is relatively understudied in the LUCC record but offers many unique geographic contexts in which to apply it. Situation of the Caribbean Basin Islands The Caribbean Basin's hot, tropical, and predictable climate, accessibility, and relative proximity to Europe were prime determinants of its colonial take-over. The entire region reflects, in elements of society, politics and culture, the intense foreign colonization that occurred starting in the late 1400s. Both the large and smaller islands have deep roots in the global market economy in that most of them, following "post-discovery," conformed to being the source of tropical and exotic agricultural goods that people in temperate, industrialized countries discovered and craved thereafter (Pico, 1974). Expansive primary forest was cleared to grow sugar, coffee, bananas, and the like on large commercial plantations on many of the islands. Peasant farmers not working on the hot and humid, lowland plantations lived impoverished on the margins of good land, usually in high elevations and steep slopes. Across the Basin, environmental degradation, mainly in the form of deforestation and soil erosion, was severe by the 1930s, and remains the case on some islands today.
8 Since the mid-twentieth century, the agriculture-based economy in the region has gradually yielded its prominence to industries of tourism and to some extent, manufacturing. Relying exclusively on tourism puts these islands in a vulnerable spot, as tourists and their dollars are not a constant or a given at any time. Tourism also brings with it the environmental pressures of development, including forest clearing, loss of prime agriculture land, ecosystem pollution, and social problems, such as crime (del Mar Lopez, 2001). Greater regional problems of the contemporary CB include the need for establishing alternative sources of income and the production of food for domestic consumption and local trade. Regional terrestrial ecosystems exist mostly on either volcanic-derived or uplifted fossilized coral materials; most of which, throughout geologic history, have cycled through periods of connection and isolation. The recent work of Areces-Mallea et al. (1999) provides a comprehensive inventory, classification, and description of the terrestrial vegetation ecosystems of the CB. Broadly, these consist of tropical wet and dry forest formations, savannah grasslands, cloud forests, mangroves, and coastal beaches. In 2000, the FAO reported 5,974,000 ha. (59,740 km 2 ) of forest, only covering about 26% of the land area. From 1990-2005, the Basin experienced a net forest gain of nearly 12%, mainly in Cuba and Puerto Rico. Puerto Rico: Site and Situation The island of Puerto Rico (PR) is centered at N. Lat. 18 o 15' W. Long. 66 o 30' in the Caribbean Basin (CB) tropics at the eastern end of the string of larger islands designated the Greater Antilles and the western end of the smaller Lesser Antilles (Figure 1-1). Geologically, the majority of the island arc derives from submerged mountains and provides the barrier between the Atlantic Ocean to the north and the CB to the south
9 (Pico, 1974). The total area of Puerto Rico is 9,104 km 2 , including coastal and inland water bodies. The island takes a rectangular shape stretching 182 km east-west and 66 km north-south, and its landscape consists of mostly mountains and hills (about 75%) with extensive coastal plain areas in the north and east (about 25%). The highest point is 1,338m in the rugged south central mountain range that spans nearly the entire east to west extent of the island. Pico (1974) summarizes some major climatic controls of PR including 1) its tropical latitude totally within the area of maximum annual insolation; 2) its situation of being a small island in an oceanic basin thus moderating extreme temperatures; 3) the northeast to southwest direction of the trade winds which absorb moisture in the northern coastal plains and drop it as an orographic rain towards the central mountain ranges and 4) the islandâ€™s topography creating humid moist north lowlands, cooler wet central mountains and a much drier and hotter southern landscape. This holds true for many land areas in the CB. The island is susceptible to tropical cyclones: Hurricane Georges in 1998 was the last to give the island a direct hit resulting in significant damage (Ayala-Silva and Twumasi, 2004). Although seasonal climatic variation is minimal across PR, the months of June through November tend to show more precipitation and temperature rise than in December to May. Ewel and Whitmore (1973) established that Puerto Rico contained six Holdridge Life Zones. Holdridge defined these as bio-climatic mapping units, each of which may encompass a variety of soils, vegetation, microclimates, and land-use patterns (Ewel and Whitmore, 1973). It was considered that temperature and rainfall prevailed over other environmental factors in determining vegetation. These dictate the presence of certain
10 regimes of natural vegetation across the island. In Puerto Rico, Ewel and Whitmore mapped the following life-zones: subtropical moist forest, subtropical wet forest, subtropical dry forest, subtropical rain forest, subtropical lower montane rain forest, and subtropical lower montane wet forest. A Spanish colony by 1509, it took nearly 200 years for major land-cover transformations to present themselves on the island. By 1820, it is known that nearly two-thirds of the island had been stripped of its natural forest cover, yielding its terra to Crown commercial mega-crops of sugarcane and coffee (Franco et al., 1997). The eighteenth century saw much development for the island as roads, rails and port cities were established; the inland mountain areas were settled as smallholder farmers were pushed into the marginal uplands of the interior. Tremendous environmental problems followed in the rural areas and resulted in impoverished socio-agricultural conditions (Pico, 1974; Rudel and Lugo, 2000). Puerto Rico became a U.S. territory, as a treaty stipulation following the Spanish-American War in 1898, and by 1952 a U.S. commonwealth. Population at the turn of the nineteenth century was 1 million people, a great majority of them rural. Agricultural-economy, environmental degradation and a rural poor population dominated the Puerto Rican culture until a series of socioeconomic fixes were implemented under then Governor Lus Muoz Marn (1898-1980) in the mid-1940s. Known as Operation Bootstrap, this government-led initiative to reduce poverty via expansion of industry resulted in the rapid decline of agriculture and the pulling away of rural farm workers into factory, commercial, and government jobs in the cities. Population across the twentieth and into the twenty-first century has grown over 400%, now concentrated
11 around urban centers, giving Puerto Rico the status of one of the most densely populated places on the earth. In the forty years since Operation Bootstrap was initiated, agriculture land declined from encompassing 85% of the island to only 37%. This abandoned land has been transformed to either forest or built cover (Grau et al., 2003). The Spatial Component of LUCC Research The scientific paradigm of quantification and statistics set on course its own revolution in the multi-disciplinary environmental sciences by the latter half of the twentieth century. As computer and digital technology rapidly developed in the 1960s, applications were immediately adapted to assist in the investigative efforts of the environmental sciences. Ian McHarg's Design with Nature (1969) introduced the concept of using data overlays for spatial analysis of smart planning of man-altered land covers (Briassoulis, 2000). Significant advances in computer technology and satellite remote sensing (RS) have established the tools of geographic information science (GIS) which provide the digital data sets, storage, manipulation, query, and mapping needed for assessing changes of the earth's biosphere (Jensen 2000; de Sherbinin et al., 2002; Rogan and Chen, 2004). The Earth Resource Technology Satellite, which became Landsat, was the first RS platform launched by NASA in 1972. Landsat is the longest data-producing satellite and thus can provide data for up to thirty-three years, depending on the region under study (Jensen, 1996). The Landsat Satellite Program was launched in 1972 and started as a means to collect a global inventory of land cover, environmental conditions and subsequently, the human alteration of these. Development of Geographic Information Systems (GIS) in the
12 1970s and 80s provided the perfect tool for analyzing, synthesizing, modeling and displaying the quantifiable changes made to the physical environment; it became and remains an omnipresent and powerful mechanism for informing policy-making decisions about land, economics, politics and numerous other human agencies For the application of land-cover mapping, RS satellite sensors record the reflected and emitted solar radiance from materials at the earth's surface in the various electromagnetic wavelengths, depending on the radiometric capability of the sensor. The Landsat Thematic Mapper 4 and 5 (TM) sensors, for example, record a gapped range of reflected light from 0.45m (band 1 in the visible spectrum) to 2.35m (band 7 in the middle-infrared spectrum); an additional band records emitted radiance from 10.5m to 12.0m (the thermal band). The digital RS data becomes a representative grid of the area of interest mapped out at the pixel resolution of the particular sensor, 30 m 2 blocks for Landsat TM (Jensen, 2000; Binford et al., 2004). With empirical knowledge of the reflectance signature of various earth materials, the on-the-ground reality of land cover can be ascertained from the pixel values in association with ground-truthing. A plethora of different image enhancement techniques and ancillary GIS data allow for more precise interpretation and classification of the data, but detailed knowledge of the study area and ancillary ground data must also be used to assist the computerized classification algorithms. Classification groups pixels by similarity of their reflectance values under restraint of user-defined parameters; the resulting dataset is a digital land-cover grid (Rogan and Chen, 2004). Remote sensing science provides the methods for ascertaining and mapping land cover. Change in land cover is detected quantitatively by areal extent and occurs as
13 either a conversion from one cover type (e.g. forest) to another (bare earth) or by modification, like selective logging. Analysis of the same pixels from two or more points in time provides the means to detect change; these are referred to as land cover change trajectories (Mertens and Lambin, 2000). Besides defining descriptive change to the physical landscape, a more comprehensive aspect of LUCC research is to investigate and understand the linkages between human action and the physical results on their landscape (Geist and Lambin, 2001; Green et al., 2005; Hagendra, et al., 2004). The first step in any type of LUCC research is the creation of digital datasets representing the accurate and precise identification, quantification and extent of land cover classes of interest at different points in time and geographic scales; this provides the subsequent datasets for analyses of temporal change and the spatial character of landscape fragmentation. Secondly, elements from social science, often provided by census data or field surveys, can be integrated in the analyses of biophysical data to infer human impact on landscapes. Documented case studies become scalable and comparable within various historical, socioeconomic and geographical contexts (NRC, 1998). Armed with the information that these types of analyses provide, management and policy decisions regarding natural resources and their use, sustainability and conservation can become better informed, modeled and hopefully more applicable. Landscape Fragmentation and Landscape Ecology A landscape is defined as: â€œa heterogeneous land area composed of a cluster of interacting ecosystems that is repeated in similar form throughoutâ€ (Forman and Gordon  cited in McGarigal and Marks ." Ecosystems are defined as biophysical landscape units, which possess some common (spatial or structural) characteristics
14 (Haines-Young, 2005). Landscape ecology researches the relationships between patterns of land cover and that of ecosystem function and process at a given geographic scale (Gustafson, 1998; Van Wey et al., 2005). The joining of landscape ecology and the LUCC discipline shared the common denominator regarding the concern of 'behind the scenes' ramifications of the loss of habitat to flora and fauna due to fragmentation, particularly happening to tropical forest landscapes (McGarigal and Marks, 1995). Fragmentation of a landscape occurs as large, contiguous patches of a land cover type are broken up by another type over space and time. This has measurable effect on ecological, environmental and human characteristics. Landscape ecology concerns itself with the spatial arrangement and configuration of a landscape including patchiness, density, edge effect, biologic connectivity (corridors) and many others, in order to assess the state of an ecosystem's health and sustainability (McGarigal and Marks, 1995; Lugo, 2000). Land cover classification maps model patches of landscape components (vegetation cover, topography, watersheds) and their position within a matrix (the whole landscape under study). In an attempt to quantify the characteristics and relationships of these eco-patches, various metrics and computer software programs have been developed over the years. The software FRAGSTATS (McGarigal and Marks, 1995) is a comprehensive program that allows spatially explicit data input for landscape metrics computation. The metrics chosen by the user are dependent on the research question, but a few basic metrics can assess general landscape integrity and the state of fragmentation (McGarigal and Marks, 1995; Nagendra, 2003). These measures are of valuable interest to vegetation and wildlife scientists and other natural resource managers. The amalgam of land cover
15 change and landscape fragmentation in research is a relatively new and pioneering concept. Research Objectives As mentioned earlier, Puerto Rico holds a unique situation in that: 1) it is an island and therefore, a relatively closed system, 2) its colonial and commonwealth linkage with the U.S. has bestowed upon it better socio-economic opportunities, and 3) it has already experienced a cycle of extreme land degradation and deforestation, industrialization and urbanization and a subsequent decline in agriculture leading to rapid forest re-growth of abandoned agriculture land. This thesis continues and contributes to research in the LUCC and HDGC initiative for Puerto Rico, to quantify extent of land cover and trajectories of its change over time using geo-spatial techniques, assess the associated dynamics of landscape fragmentation, to evaluate effects of geographic scale in LUCC dynamics and assess the relation of human activity, specifically trends in agriculture, on land cover change. Although both forest and non-forest cover is mapped and addressed in this research, emphasis is on forest change and how agriculture affects this change. This research strives to further understand the forests in Puerto Rico and their spatial dynamics and to contribute data for use in their preservation. The research is undertaken in the mainly tropical dry southwest region of Puerto Rico, an area comprised of six municipalities and traditionally rich in agriculture and socioeconomic diversity. To facilitate and itemize the research objectives, a number of research questions and associated hypotheses were generated. They are as follows: Q1. What was the extent of forest cover in the years 1982, 1992 and 2002 for the study area? Was there net forest gain or loss over this time?
16 H1. The region will show a net increase in forest cover. Q2. What are the trajectory outcomes of land cover change, their percentage area of occurrence and their spatial distribution in the study area? How do these vary across the six counties? H2. he most prevalent LCCT will indicate older and recent forest re-growth (NF-F-F and NF-NF-F classes). H3. he most change classes will occur in lower elevation regimes. Q3. How did forest dynamics compare across the six counties? What percent of forest cover classified in 1982 remained forest throughout the 20 years? H4. Forest cover will be maintained equally across all six counties since 1982. Q4. How has the extent and homogeneity of forest cover patches changed over time? Did forests become more or less fragmented at the region level? H5. Forest cover will show less fragmentation, particularly in the northern reaches. Q5. Are there discernable patterns or trends in the way agriculture dynamics have resulted in land cover change outcomes and pattern across municipalities? H6. Decrease in number of farms and land reported in agriculture use will result in more areal extent of forest re-growth outcomes over time. H7. Rise in tenant farms will result in more shifting forest cover change outcomes with high NP and small LPI. Previous Research Some significant studies researching the linkages between LUCC and human activity in Puerto Rico will be reviewed here. Thomlinson et al. (1996) undertook an analysis of LUCC from 1936-1988 in a northeastern municipality of Puerto Rico close to the San Juan metropolis area. Interpretation of aerial photographs determined large-scale agriculture abandonment and subsequent transfer of these lands to pasture in the lowlands and dense forest in the uplands. It was found that patches of remaining dense forest cover in 1936 spatially seeded larger dense forest patches in future observations. This confirms Brown and Lugo's (1990) suggestion that remnant forest patches are of great
17 ecological importance. Pascarella et al. (2000) found similar dynamics in their study in the Cayey Mountains of the southeastern island. They reported a 62% increase in forest cover from 1937-1995 due to abandoned crop fields. Higher elevations were indicative of older re-growth and pasture in proximity to older forest patches was likely to "seed" and grow back. This study as well as the research of Dietz (1986), Wallach (1989), Suarez (1998) and others alludes to the fact that nearly 90% of Puerto Rican land was in agricultural use around 1900 and forest cover existed in the form of small patches; this has entirely reversed today as secondary forests are the most prominent land cover island-wide. Rudel and Perez-Lugo (2000) attempted an island-wide inventory of forest change. Their research found a 9-37% increase in forest across the island from 1950-1990, unsurpassed proportionate to land size by anywhere else on earth. This paper incorporates biophysical factors and human dynamics as independent variables to understanding Puerto Rico's reforestation. This research assumes the forest transition theory put forth by Rudel (1998) and Mather and Needle (1998) is occurring on the island. Census data for the study is aggregated to the municipality level and determines both demographic and agricultural dynamics that could predict re-growth occurrence. Significant findings include a confirmation with Thomlinson (1996) and Pascarella et al. (2000) in that forest re-growth is most concentrated in the higher elevations; reforestation occurs mostly in the island's interior. The authors found that the once dominant crops of sugarcane, coffee and tobacco had the sharpest decline in area and that forests took over these abandoned crops.
18 The studies of Ramos-Gonzalez (2001) and Helmer et al. (2002) were the first on record to systematically map and assess change in land-cover using satellite imagery and GIS techniques; the former focused on the northeast and the latter island-wide. Ramos-Gonzalez addresses land cover change from 1978-1995 and found 50% loss of agriculture land, significant vegetation re-growth in higher elevations and large areas of new built-up land mostly in the coastal lowlands. Nearly all changes occurred on relatively flat land (0-10 slope). This study brings up interesting points about the state of land change in Puerto Rico that warrants further study: 1) the fact that the island is small and densely populated yet showing dramatic decline in agriculture confirms the island's problem with feeding itself now and in the future, 2) the idea that what little prime agriculture land the island possesses is being built on and un-used for crops and 3) the reforestation occurring island-wide is largely unmanaged and in private ownership. The detailed island-wide land-cover mapping of Helmer et al. (2002) provided a snapshot of cover in 1992. Closed forest comprised 41.6%. Pasture and grassland comprised 36.7%. Agriculture comprised only 5.9%, of which coffee was 2.4% of this total. Built cover comprised 10.5%. A significant finding of this research was that areas formerly used for sugarcane and pineapple crops were shifting to hay and pasture use. This project achieved the best digital dataset of land cover from which future studies could establish a baseline. Lugo (2002) uses the Puerto Rican experience of forest re-growth as an example of managing tropical landscape in the future. He sites Beller et al. (1990) in establishing that islands are exemplar for investigating the relationship between change in physical landscape and socioeconomic human drivers because of the sufficient control of outside
19 variables. Lugo addresses landscape fragmentation in association with forest change and how even with its extreme population density, the forested landscape appears to be decreasing in patchiness. This research, along with Grau et al. (2003), Helmer (2004), and previous others, calls for further focus on the dynamics of forests in current agricultural lands (those not abandoned) and management strategies which will mitigate the competition from encroaching commercial and residential development. The overall results of contemporary research concerning land cover change in Puerto Rico show that the island as a whole is experiencing reforestation as a result of abandoned agriculture, although not all regions of the island are represented by empirical findings in this assumption. Forests are becoming more homogenous, especially in higher elevations and with increased distance from urban areas. Socioeconomic factors quantitatively related to forest cover change are highly supportive of forest transition theory in which the pull of the rural population and smallholder farmers to cities for better economic opportunity results in abandoned fields which naturally revert back to secondary forest growth. The future of forest dynamics in Puerto Rico partly rests with the phasing out of tax leniency and other incentives for U.S.-based industry. With much of the forest re-growth having occurred on privately-owned land and given the bleak situation of food security to an ever-increasing population, another period of deforestation could occur as agriculture again becomes a primary land use. This thesis is divided into three chapters. Chapter One, serves as an introduction to the concepts of land cover change, particularly forest cover, and associated research fields and literature comprising its study. It establishes the research objectives for the thesis and introduces the area under study. Chapter Two is formatted as a stand-alone
20 paper for publication detailing the LUCC research of southwest Puerto Rico and Chapter Three integrates the results of Chapter Two into a broader discussion of regional and global comparisons of findings and concludes the thesis. Figure 1-1. Puerto Rico in the Caribbean Basin.
CHAPTER 2 FOREST COVER CHANGE, FRAGMENTATION, AND AGRICULTURE IN SOUTHWEST PUERTO RICO Introduction Land use and land cover change (LUCC) research is an essential component in understanding local, regional and global environmental change and developing strategies for the sustainable use and conservation of natural resources, particularly forest extent and integrity. This is especially a concern for the Tropics of Latin America where rates of deforestation have risen in unison with population increase, urbanization and other socioeconomic factors (Geist and Lambin, 2002). The change and juxtaposition of forests and other land cover types within a landscape is often the direct result of human dynamics, be it factors of demographic, economic or political nature (Thomlinson et al., 1996; Tucker and Ostrom, 2005). Changing landscape composition and pattern, traditionally studied in landscape ecology, has ramifications on ecosystems and their components, flora and fauna habitats and environmental responses (Van Wey et al., 2005). Landscape ecology focuses on assessing how landscape parameters, or metrics, affect integrity of the overall ecosystem while integrating in the methods, tools and concepts of geographic science adds a valuable spatial perspective to the discipline (Evans and Moran, 2002; Griffiths and Mather, 2000). Forests are generally defined as continuous stands of woody tree species or a tree-dominated ecosystem. Forest cover comprises about 25% of the Earth's surface 21
22 and accounts for >80% of its total biomass (Randolph et al., 2005). Composition and structure of forests depend upon ecosystem variables such as climate, precipitation, soil characteristics and the like. Forests provide numerous ecosystem services, particularly acting as habitat to a variety of flora and fauna and as the prime source of atmospheric carbon dioxide (CO 2 ) storage or sequestration (Silver et al., 2000; Randolph, et al., 2005). The tangible services (wood material, pulp, charcoal, medicine among many others) that forests provide to humans result in their vulnerability to being cleared and degraded. Agriculture is the primary proximal driver in which humans have altered forested land cover throughout history and is still true today, followed closely in more recent times by urbanization (Wolman and Fournier, 1987, Marcano-Vega et al., 2002). When forest cover is altered or destroyed, both the ecosystem and human services suffer. Although a comprehensive and interdisciplinary body of research has addressed the human-environment component of understanding forest cover change, more case studies are needed in less represented sites, situations and landscape contexts (Lambin et al., 2003). In addition, investigation of both the pattern and process of forest change warrants more research (Nagendra et al., 2004). The conceptual and digital labor involved for quantifying and analyzing landcover change trajectories (LCCT), landscape fragmentation (LF) and proximal and distal land use processes is assisted by the spatial component provided by remote sensing, GIS technology and methods of both the geography and landscape ecology disciplines. Numerous studies have approached the combination of these methodologies for a wide breadth of LUCC applications (Gustafson, 1998; Mertens
23 and Lambin, 2000; Southworth et al., 2002; Southworth et al., 2004). Land management entities and schemes need this interdisciplinary knowledge to monitor, query, update and distribute information for better-informed policy-making decisions The geographic scale at which environmental and human processes occur affect, among other things, the perspective from which conservation and land use management play out. For example, what might be apparent at a regional scale hides significant variation within the human-induced borders that comprise it (de Sherbinin et al., 2002). There are relatively few studies incorporating these research objectives in the Caribbean Tropics; there are fewer addressing tropical dry forests there or elsewhere (Abizaid and Coomes, 2004). Small island developing states, or SIDS, are historically and continuously at the mercy of their larger continental counterparts and more prone to degradation of natural resources. Sustainability of natural resources of SIDS is vital to their future opportunity to become more self-sufficient, especially in the areas of land management and food security. The Caribbean Sea Ecosystem Assessment (CARSEA) extension of the Millennium Ecosystem Assessment Project (MA) has targeted land use and cover change research to be a priority for Caribbean SIDS. Studying the interface of human activity and changes in the physical environment is well applied to islands because they are relatively closed systems in the context of both. Outside influencing factors are minimized and the control of variables is facilitated better (Lugo, 2002). The island of Puerto Rico is unique to its neighbors in the CB in that one, it is a commonwealth of the U.S., thus having the advantage of economic strength and
24 opportunity and two, where most land in Latin America and the Caribbean is experiencing deforestation, studies of Puerto Rico have showed continuous net forest gain for the last five decades (Dietz, 1986). A direct result of the socio-economic improvements and industrialization of Puerto Rico implemented under then Governor Luiz Munoz-Marin from the late 1940s (known as Operation Bootstrap) was the immediate and steady decline in agricultural economy and lifestyle (Morales-Carrion, 1983 and Rudel and Lugo, 2000) (Figure 2-1). The most physical manifestation of these changes was the steady abandonment of marginal cropland and re-growth of forests (Grau et al., 2003). A few case studies target these dynamics island-wide or in individual municipalities in the northeastern island, (Thomlinson, J.;1996; Pascarella et al.; 2000; Ramos-Gonzalez, 2001) but no published research investigating these dynamics in the southwest--a region of mostly tropical dry landscape and traditionally rich in agriculture, was found. Are the municipalities here experiencing the same net re-forestation as in the central and eastern regions? Is agriculture in decline here at all? Research Objectives and Questions The overall objective for this research was to assess the dynamics of forest cover, fragmentation and the human impact, mainly agriculture activity, across the southwest municipalities of Puerto Rico from 1982-2002. To achieve this, satellite imagery data, GIS and landscape metrics were used to map forest cover and its change in quantity and pattern over time; National Agriculture Statistics Service (NASS) data tracked agricultural characteristics in the region. Specific research questions and their associated hypotheses are listed below:
25 Q1. What was the extent of forest cover in the years 1982, 1992 and 2002 for the study area? Was there net forest gain or loss over this time? H1. The region will show a net increase in forest cover. Q2. What are the trajectory outcomes of land cover change, their percentage area of occurrence and their spatial distribution in the study area? How do these vary across the six counties? H2. The most prevalent LCCT will indicate older and recent forest re-growth (NF-F-F and NF-NF-F classes). H3. The most change classes will occur in lower elevation regimes. Q3. How did forest dynamics compare across the six counties? What percent of forest cover classified in 1982 remained forest throughout the 20 years? H4. Forest cover will be maintained equally across all six counties since 1982. Q4. How has the extent and homogeneity of forest cover patches changed over time? Did forests become more or less fragmented at the region level? H5. Forest cover will show less fragmentation, particularly in the northern reaches. Q5. Has agriculture activity decreased steadily across the 20-year study span? Are there discernable links between agricultural changes and land cover change outcomes across municipalities? H6. Linear decrease in number of farms and land reported in agriculture use will result in more areal extent of forest re-growth outcomes over time. H7. Rise in tenant farms will result in more shifting forest cover change outcomes with high NP and small LPI. Study Area This research investigates an area of about 700 km 2 (69,900 ha.) in southwest Puerto Rico comprised of six municipalities or counties: Cabo Rojo, Guanica, Hormigueros, Lajas, Sabana Grande, and San German. The study area is approximately the size of the Okefenokee Swamp in southeastern Georgia and northeastern Florida, U.S. or Charlotte County, FL, U.S. (Fig 2-2).
26 Topography ranges from sea level up to 900 m above. The southwest is the driest region of the island, due to the orographic effect the central mountain ranges impose on it, stopping and draining the saturated westerly tropical cloud-masses. Average annual rainfall for the southwest is less than 35 inches (890 cm) but this varies across the physiographic units that comprise the area (Fig 2-3). Ewel and Whitmore (1973) in classifying Holdridge Life Zones of the island included most of the study area in the Tropical Dry zone. The physiographic regions of Puerto Rico that Pico (1974) describes as encapsulating these six counties are the rainy west-central mountains; semi-arid southern foothills; wet-dry western coastal valleys and dry southern coastal lowlands. In the majority of the study area, rainfall and flooding are prevalent from May to November; December to March are very dry. Soils are described as mainly fertile, poorly drained alluvium-based material; this changes to a more volcanic-derived soil base in the higher elevations of San German and Sabana Grande. Natural vegetation ranges from tropical moist deciduous in the north to a dry chaparral, scrub-shrub forest and savanna in the south. The Guanajibo River Watershed is the most significant in the region but many smaller drainage basins catch water from the northern highlands. Four major rivers drain the region: the Guanajibo, Rosario, Grande and Cupeyes. An irrigation canal originates from the Rio Loco in the east and provides the Lajas Valley its agricultural blood. The study area contains six large population centers, which coincide with the municipal seats. The combined human population of the region in 2000 was 174,714 (U.S. Census) making an average population density of 250 persons per km 2 , relatively
27 rural when compared to the islandâ€™s average of around 400 people per km 2 . Most people are situated around the population centers of San German, Guanica and Hormigueros. Population distribution at the census block level is mapped in Figure 2-4. Grau et al. (2000) measured the migration characteristics for counties in Puerto Rico and found that these six counties experienced only moderate immigration and emigration between 1940 and 1990. Figure 2-4 graphs each municipality's population growth over time. The regional economy has traditionally been dominated by agriculture, mainly sugarcane and in the highlands coffee. Smallholder farms tend to be in the higher elevations and larger commercial farming occurs in the Lajas Valley and flat coastal plain (Rudel and Perez-Lugo, 2000; Grau et al., 2000). The Lajas Valley Irrigation Project and canal (LVIP) completed in the 1950s strengthened the agricultural-economy here. Land not irrigated for crops and once used for sugarcane is now used for pasture and hay, which is a prevalent use (Helmer et al., 2002). Crops grown across the region today include bananas, plantains, coffee, sugarcane, rice, pineapple, and various fruits and vegetables; fishing is a major food commodity. The port cities and ancillary economic activities of Guanica in the south and Mayaguez (just north of the study area boundary and Puerto Rico's second largest city) have contributed additional industrial and commercial economy to the region (Pico, 1974; Sotomayor, 2004). The Industrial Incentives Act of 1954 allowed generous tax exemptions to US-based manufacturing and hotel firms; it was further extended in 1963, but only to industry that located outside what were then the major metropolitan areas (Pico, 1974; Sotomayor). As a result, several petroleum refineries
28 and other manufacturing ventures were established in the southwest in the 1960s and 70s, drawing population and bringing development that inevitably encroached on prime agricultural land. Tourism activity increased in the 1980's, putting further pressure on coastal and other 'scenic' ecosystems. In an effort to diminish urban sprawl and promote agriculture on the region's most fertile land, the Puerto Rico Planning Board, Department of Agriculture and other stakeholders initiated the establishment the Lajas Valley Agriculture Reserve in 1999, creating an agriculture-friendly zoning and land use scheme. The local universities of San German and University of Puerto Rico in Mayaguez (UPRM) contribute to the agriculture experiment station in Lajas where technicians investigate and test new technologies for improving agricultural techniques and crops. Environmental concerns in the region include soil erosion, degradation of prime cropland, pollution of waterways from fertilizer runoff, pollution of mangroves, urban and coastal development pressures and fragmentation of habitats due to increased expansion of road and infrastructure networks. Ten parks/reserve lands including state forests and wildlife refuges comprise about 7,770 ha. or 11% of the study area. Research Methods Data Acquisition and Processing Satellite imagery Landsat satellite imagery was used to create the land cover datasets; it was readily available and is best suited for longitudinal land cover change comparisons (Jensen, 1996; Ramos-Gonzalez, 2001). Land cover observation intervals of ten years were chosen primarily because of data availability, but it was also assumed to be an adequate temporal scale for determining general forest cover change. Three
29 different Landsat sensor platforms were incorporated given the length of the study: Multi-Spectral Scanner (MSS), Thematic Mapper 4 (TM4), and Enhanced Thematic Mapper (ETM+). More in-depth information about these sensors can be found in (de Sherbinin et al., 2004; Green et al., 2005). Erdas Imagine 8.4, ESRI's ArcGIS 8.3 and ArcView 3.2 were used for image processing and analysis. The study area in the dry southwest required one satellite scene for each observation with the exception of the 1982 MSS images, which required a mosaic with a sliver of p005r048 to the south; a histogram matching technique standardized reflectance values between these scenes. Preference was to use scenes from December to March (which corresponds to the region's drier and fallow season) but cloud-free and high quality imagery was the priority. For 2002, a February scene was used; for 1992, an August scene; and for 1982, a February scene mosaicked with a sliver of a March scene (Table 2-1). The first step in processing the imagery was to subset out the land area of interest from the vast amount of water pixels. Each subset scene was then calibrated following the Center for the Study of Population and Environmental Change (CIPEC) -published protocol to standardize the correction of radiometric and atmospheric distortions in satellite data. Theoretical calibration procedure is found in (Jensen, 1996). The calibrated scenes were geo-registered to UTM zone 19N WGS84 projection using ground control reference points (GCP) from both a global positioning system (GPS) unit in the field and existing GIS data (roads and water features) derived from 1:24,000 scale topographic maps; then scenes were coregistered to one another, with the 2002 image serving as the base. Image to image
30 registration ensures pixel for pixel â€˜line-upâ€™ when multi-scene land cover classification change detection is the objective (Jensen, 1996; Rogan, 2004). All scenes achieved a RMSE of < 0.5 pixel value (< 15m) and were re-sampled to 30 m resolution in Imagine. Re-sampling the 80 m resolution of the MSS scene to 30 m inevitably decreased accuracy, but was considered adequate for general forest identification. GIS data GIS data were acquired from the Planning Board of Puerto Rico in San Juan. Layers used in this study included island and municipality boundaries, roads, waterways and watersheds, urban expansion boundaries, park and reserve boundaries and soil types and characteristics. The data was transformed to UTM zone19 WGS84 projection. GPS points from fieldwork included land cover training samples (TS) and GCP. Digital elevation datasets (DEM's) were acquired from the Geo-community Internet website ( http://www.geocomm.com ) and transformed into both Erdas Imagine (.img) and ESRI ArcGRIDs (GRID) formats for processing in the GIS software. The DEM's were mosaicked and clipped to the six county study area. They were then reclassified to four elevation regimes. Land-cover training samples Fieldwork in 2004 provided the GPS coordinate collection and attribute data of land cover classes of interest, following the TS protocol and worksheet developed by CIPEC scientists (Kauneckis et al, 1998). A Garmin GPSMap76S unit was used and each point was taken by averaging and achieving an accuracy reading of < 5m. TS coordinates and data were transformed into GIS features and projected into WGS84 UTM zone 19, creating a spatially explicit database capable of overlaying with the
31 satellite imagery and the other GIS data. Since classification of the images was done by an unsupervised algorithm method, field training samples were used in the assessment of classification accuracy for the 2002 scene. Census data Census data are often the only source of standardized social and agricultural characteristics on record, especially at finer geographic scales, like U.S. counties. Human-environment interaction is usually assessed at the regional or national scale; further scaling down can reveal relationships overlooked at smaller geographic scales (Radeloff et al., 2000; Perz, 2003). A major element to investigate in this research was the variation in human activities' effect on forest cover change outcomes between the six municipalities of the study area. Humans often sanction their activity and decision-making regarding the environment in the context of these superficial boundaries. In addition to population change, agricultural activity was chosen because of its establishment as a primary driver of forest change in other studies of Puerto Rico and elsewhere in the tropics. The variables in this research were chosen to measure increase or decrease in agriculture activity as well as for ascertaining land use and land quality characteristics. These variables and their assumed ties to land cover change are summarized in Table 2-2. Although prone to under-representation and occasional bias, social, economic and political dynamics can be assumed with some confidence using census statistics (Radeloff et al., 2000). Data on population and agricultural variables were collected for each municipality across the temporal scale of the research; data from 1978-1982 offered insight to conditions leading up to the first land cover observation in 1982. Data was mostly accessed via the Internet sites of the U.S. Census Bureau and the National
32 Agriculture Statistics Service (NASS) excluding the older census books, which were in library government documents archives. Agriculture data were recorded for 1978, 1982, 1987, 1992, 1998 and 2002. Population and Housing data were recorded for 1980, 1990 and 2000. Documentation of these data, their collection methods, uses and error can be found at their respective Internet websites: (http://www.census.gov) and (http://www.nass.usda.gov/Census_of_agriculture/index.asp) Image Classification and Manipulation Single time-step land-cover classification For classifying land cover in the imagery, the unsupervised classification technique of ISODATA was used. This technique uses software-intuitive algorithms to cluster like brightness values into classes (ERDAS, 1996; Binford et al., 2004). Classifying scenes from the past without actual field training samples from these times made this option more logical and allows all scenes to be classified in the same manner. In addition, having all images calibrated to a standardized protocol further enhanced their cross comparability. Before executing the ISODATA clustering, spectral signature patterns were evaluated for forest plots in order to decide which spectral bands in each scene provided the most information for classification. Using TM and ETM combinations of band 7 (2.08-2.35m), band 4 (0.76-0.90m) and band 3 (0.63-0.69m) and MSS band 4 (0.8-1.1m), band 2 (0.6-0.7), for the respective time steps, 60 classes were derived for each scene. A random stratified (equalized by thematic class) array of points was generated across each scene resulting in a table of X,Y coordinates, thematic class value and an empty field for reference class; these tables were converted into GIS layers. For 2002, a combination of fieldwork observations and aerial photos from 2001 (GIS Online
33 Portal and Data Viewer: Puerto Rico Planning Board http://gis.jp.gobierno.pr/) were used to plot each random X,Y point and record the cover class for that area (forest or non-forest) in the GIS table. For 1992, Terraserver ( http://www.terraserver.com) , an online source of older aerial photographs (from 1991), was used to plot 1992 points and gather land-cover class. For 1982, there were no aerial photos available; GIS soils layer was used, digitized at 1:24,000 scale and based on the Puerto Rico Soil Survey (Roberts, 1943) to ascertain land cover. An attribute of each soil layer was its use; forest or non-forest cover was assessed from this information. In this dataset, an attribute of each thematic class in the original unsupervised classified scenes now had an information class to be assigned to it. Forest or non-forest classes were recoded to two values (1 and 2 respectively) and No Data (0) comprised leftover background and water areas. The scenes were then transformed to thematic format and attribute fields of area in hectares and class names added to the data table. These final datasets were then clipped to the six-municipality study area spatial extent. Accuracy assessment Separate sets of random generated points were used to perform accuracy assessment on the 2002 and 1992 classified scenes. These points were located using the same aerial photo data used to create training samples. Overall accuracy for the 2002 classification was 83.33% with a Kappa statistic of 0.67; for 1992, the overall accuracy was 89.74% with a Kappa statistic of 0.79 (Table 2-3). No accuracy assessment was performed on the 1982 classification. Post-classification change detection Image addition (IA) was used to create datasets of land-cover change trajectories (change of class between observations) for the study area (Southworth et
34 al., 2002; Nagendra et al., 2003). This and other post-classification methods can be further investigated in (Singh, 1989). Essentially, eight trajectory classes are formed from two-class, three time-step classifications (2 3 ). Three LCC trajectory (LCCT) datasets were created: 1) 1982-1992, 2) 1992-2002 and 3) 1982-1992-2002. The resulting datasets were recoded with class names and area in hectares and saved as an 8-bit unsigned raster format (ERDAS, 1996). Landscape metrics (LM) Landscape fragmentation analysis was undertaken using the software FRAGSTATS version 3.3 (McGarigal and Marks, 1995), which is capable of quantifying numerous landscape metrics at the landscape, class and patch level of measurement. For part of this study, evaluating temporal landscape fragmentation in conjunction with land cover change was undertaken by tracking the number of patches, average patch size and largest patch index at the class and landscape level and thus derived the metrics: CA= total area in hectares of each input land cover class PLAND= percent of entire landscape that each class occupies NP= total number of patches for each class or entire landscape LPI= largest patch index is the percentage of the total landscape under study that a class's largest patch takes up MPS= the mean size in ha. of all the patches of a class combined Metrics were calculated for the three individual time step LC classifications and the three LCCT datasets for the entire study are and the separate municipalities.
35 Results Land-Cover Condition and Dynamics Regional results The six southwestern municipalities of the study area comprise 69,900 ha (about 700 km 2 ). In 1982, 31,927 ha. (45.9%) of the region was forested and 37,604 ha. (54.1%) was non-forest. In 1992, 23,100 ha. (33.3%) was forested while 46,283 ha. (66.7%) was non-forest. In 2002, 30,843 ha. (44.3%) was forested and 38,814 ha. (55.7%) was non-forest (Figure 2-5). The region had a 27.6% decline in forest area from 1982 to 1992 and then regained 33.5% forest cover from 1992 to 2002. This resulted in a total 3.4% net loss of forest and 3.2% increase in non-forest cover. Land-cover change trajectories (LCCT) represent the outcomes of land cover change over designated ranges of time and in doing so, delineate patches of unique ecosystems. These data quantify and map areas in stable, alternating or changing covers. From 1982-1992, 17,690 ha. (25.6%) of the region remained in forest while 32,161 ha. (46.4%) remained deforestedâ€”a total of 49,851 ha. (72%) in static land cover. 14,082 ha. (20.3%) of land was involved in deforestation while only 5,320 ha. (7.7%) was reforested. From 1992-2002, 19,217 ha. (27.7%) remained in forest while 34,828 ha. (50.2%) remained clearedâ€”a total of 54,045 ha. (77.9%) in stable land cover. 3,868 ha. (5.6%) of land was involved in deforestation while 11,425 ha. (16.5%) was reforested (Figure 2-6). The 20-year LCCT analysis (Figure 2-7) determined that 15,851 ha. (22.9%) had remained in stable forest cover and 25,919 ha. (37.5%) remained cleared across 20 yearsâ€”a total area of 41,770 ha. (60.4%) that displayed no discernable change from the satellite-based analysis. 8,886 ha. (12.8% and the most extensive trajectory)
36 of the land experienced deforestation by 1992 and remained so while 3,290 ha. (4.8%) experienced reforestation by 1992 and remained forest. 6,225 ha. (9.0%) of land was in recent reforestation while only 1,833 ha. (2.6%) recently deforested. Shifting forest cover is significant in this region as 7,205 ha. (10.4%) of land was alternating between forest and non-forest across 20 years. In all, 27,439 ha. (39.6%) of the land in the southwest was in a dynamic state. Results for all regional land cover change are summarized in Table 2-4. Results across municipalities The distribution of land cover change across all component municipalities is charted in Figure 2-8 and mapped in Figure 2-9. Scaling down the larger region to its six political boundaries revealed that Guanica, Lajas and Cabo Rojo accounted for 23.3%, 18.3% and 13.3% net area deforested respectively; Hormigueros, San German and Cabo Rojo for 26.6%, 21.2% and 15.2% forest re-growth. All municipalities, with the exception of Hormigueros, experienced more deforestation than re-growth in the 1980's; this trend was reversed, with the exception of Guanica, in the 1990s. All six municipalities contributed from 16-25% of total land cover change from 1982-1992 with Cabo Rojo being the most dynamic. In the 1990's, Guanica and Hormigueros were responsible for the highest rates of deforestation and concurrently, Hormigueros along with Sabana Grande had the highest reforestation rates. Spatial distribution shows major permanent forest clearing (F-NF-NF) was patchy and concentrated in the Lajas Valley, around the urban corridor near Cabo Rojo city and southwest Cabo Rojo County. San German municipality had the most stable forest cover (mostly concentrated in higher and steeper elevations) over 20 years and Lajas, the most stable non-forest cover. Lajas was the most stable unit in total and
37 Hormigueros, the most dynamic. The change trajectories that occurred the least over 20 years were recent clearing (F-F-NF) and shifting cover with recent clearing (NF-F-NF), both indicators of recent deforestation. Figure 2-10 graphs the variables assessing the forest dynamics across the regional municipalities. Assuming that stable, undisturbed forest cover provides the best quality ecological services like carbon storing, habitat and biodiversity, it is important to report and monitor temporal and spatial attributes of stability. Only the counties of Sabana Grande and San German retained more than 50% of the original forested area in 1982; Cabo Rojo and Lajas retained the least with only 30.9% and 21.2% respectively. Tracking re-growth of forest, the land cover change outcome classes indicating this dynamic (NF-F-F; NF-NF-F) showed that Hormigueros had the greatest area of re-forestation (26.6%) after 20 years; other counties fell roughly between 7-21%. Table 2-5 summarizes forest characteristics across municipalities. Landscape fragmentation Landscape metrics for the region and separate counties are summarized in Table 2-6 and Table 2-7 respectively; extent for each LCCT class and its associated metrics are mapped in Figure 2-11. Landscape-wide, the number of patches decreased in the 1980s as 12.8% of forest was cleared into other land uses; this dynamic was inverted in the next decade as patchiness increased again by 8% as 11% of forest grew back in the areas of the north. Class-level metrics for the individual land cover observations showed that in 1980, forest cover comprised less space, was patchier on the landscape than was non-forest and its most extensive continuous patch (LPI) accounted for only 18% of the landscape. From 1992-2002, forest gained back 11% while becoming less fragmented and achieving a 55% larger LPI, its then largest
38 patch comprising nearly 30% of the region. The least fragmented and most homogenous LCCT outcomes over 20 years were the stable forest and non-forest areas, which concurrently accounted for the greatest percentages of the regional landscape. Areas of shifting forest cover (F-NF-F and NF-F-NF) contributed to 28% of the total number of outcome patches with small LPI and MPS; the largest patches comprising only 0.09% and 0.1% of the landscape respectively. Human Dynamics: Population, Urban Growth, and Agriculture Population and urban expansion Population in the region grew steadily from 1970-2000 with the greatest spike occurring in the 1970s (18-31% increase in the counties). Cabo Rojo had the greatest population during all years, but all counties displayed the same linear growth during the period; this is reflective of the island-wide trend. This research did not evaluate the rural to urban distribution of this population but did quantify urban expansion in area as having doubled in area (from 12% of the study area in 1980 to 25% in 2000). Agriculture Agriculture characteristics and change for the study are summarized in Table 2-8 and graphed in Figure 2-12. From 1978-1982 there was a very steep decline in the number of farms in every municipality; most numbers were cut at least in half in those four years. However, land reported in agricultural use increased in Guanica, Lajas and Sabana Grande by 72%, 17% and 86% respectively. The loss of farms in this period occurred largely to small farms of 20 ha. or less. Large farms (>20 ha.) however, increased in all municipalities except Cabo Rojo and San German. Farms in both full and part ownership situations showed heavy decline in all counties while tenant farm numbers rocketed in Guanica and Lajas and to a lesser extent, Sabana
39 Grande. By 1982, San German, Sabana Grande and Guanica all had greater than 50% forest cover. In the years from 1982-1992, the number of farms continued to show net decline except for in San German where farms increased by 59%. All counties experienced significant losses in reported agriculture land, and small farms again declined except for in San German, where they spiked with an 86% increase. Large farms declined except for in Hormigueros. San German experienced a >100% increase in full owner farms in these years; Cabo Rojo and Sabana Grande saw smaller increases while Guanica, Hormigueros and Lajas numbers declined. There were significant increases in the number of tenant farms in the 1980's; every municipality except Guanica and Sabana Grande had tenant farm fluxes well above 100%. Forest cover change outcome by 1992 included major forest losses in all counties except for Hormigueros. Lajas and Cabo Rojo cleared 55.8% and 48.7% respectively of their forest cover extent in 1982. From 1992-2002, Cabo Rojo, Guanica and Lajas gained back farms of both large and small classes. Land reported in agriculture use in Guanica spiked nearly 100% in the 1990s. Number of full-owner farms increased by >200% in Guanica and declined in all other counties. Tenant farms in Hormigueros and San German dropped drastically. Forest cover change outcome from 1992-2002 saw significant re-growth in all the municipalities except for Guanica, which had a 5% loss. Discussion and Conclusion The region of southwest Puerto Rico was very dynamic in forest cover change, pattern and unlike many areas of the island, agricultural activity in the years between 1982-2002. The small (3.4%) net loss of forest found for the aggregated region
40 immediately contrasted with the significant forest re-growth found throughout the 1980's and 1990's in other research (Thomlinson, 1996; Helmer, 2002). Forest cover was the minority throughout the time period. The clearing of 27.6% forest extent in the 1980's and subsequent re-growth of 33.5% the next decade suggests that this landscape is experiencing decadal scale forest dynamics, although spatial distribution of this change shows very different areas of clearing and re-growth in each decade. As such, Hypothesis 1, that net forest re-growth would be seen at the region level, is rejected. Twenty-year outcomes of land cover change determined that nearly 40% of the landscape in the southwest had some sort of change, similar to the proportions of land cover change found in Southworth et al., 2002 and Nagendra et al., 2003. The most extensive land cover change outcome was indicative of permanent forest clearing (F-NF-NF) thus tentatively rejecting Hypothesis 2 that the most extensive LCCT class would involve forest re-growth. Looking at the map of stable non-forest with the appendage of the permanent forest clearing extent suggests agricultural extensification; these could be areas of crop conversion to cattle or pasture land typical in Latin America and the Caribbean or the result of recent government incentives to promote agriculture in the Lajas Valley. Most of the land area (70%) is below 100m in fairly flat terrain; 86% of total area deforested (either F-F-NF or F-NF-NF classes) in 20 years was in this elevation regime and 75% of all cover change occurred here. This is supportive of Hypothesis 3, that most change would occur in lower elevations, and other studies that found these areas "easier" to either develop on or convert to pasture if the infrastructure and market is there to do so (Ramos
41 Gonzalez, 2001; Lopez et al., 2001). As expected, forest cover within the boundary of parks and preserves showed very little change in the last 20 years. This shows reserve areas are functioning well although closer investigation of what is occurring in the perimeter of parks boundaries would be of interest. Great variability in forest cover dynamics was found between the region's counties thus unsupportive of Hypothesis 4 which assumed relatively equal forest recovery trajectories for all municipalities. The only counties that showed linear land cover change from 1982-2002 were Hormigueros, with re-growth of forest and Guanica, with loss of forest; the latter situation should be carefully monitored as a look at the mapped out extent of forest clearing shows close vicinity to coastal and forest reserve boundaries. The decrease to increase trend of forest cover shown in Cabo Rojo, Lajas, San German and Sabana Grande over 20 years, with a predominant forest loss in 1992, suggests that much forest land here is in transition, possibly decadal in scale. There was little forest clearing, with the exception of Guanica, from 1992-2002. This is a positive outlook for forest regeneration here in the future. Sabana Grande and San German kept much of their original forest and regained forest on much of the older clearings. However, the older forest re-growth filled in cleared areas in the higher elevations of both counties, corroborating the activity of rural, fringe farm abandonment found in similar studies in Puerto Rico (Lopez, 2001; Grau et al., 2003). Newly cleared areas in Guanica and Lajas could be a result of the opening of the LVAR in 1999 given the rise in farm numbers of both size classes and land reported in agricultural use there from 1992-2002.
42 Findings of high proportions of stable and re-growth forest classes with small NP in Sabana Grande and San German confirmed Hypothesis 5 that these northern counties would exhibit this character over time. This outcome was expected to be accompanied by a linear loss of small farms as these abandoned farms have been shown to grow back into secondary succession faster; this relation was confirmed by the census data in Sabana Grande up to 1987 but failed to correlate in San German as number of farms, both small and large, increased from 1982 to 1992 and then decreased again by 2002. Situations here could be concurrent with the findings of Pascarella (2000) in that smallholder, fringe farms have been abandoned and these areas have converted to secondary forest succession. The decrease in small farms in these counties support this even more. Hypothesis 6, which states that a decrease in farms and farm area would result in prevalent forest re-growth trajectories, was only supported in Hormigueros and Sabana Grande. A look at the spatial distribution of re-growth classes in these counties shows mostly large patches in the former, possibly re-growth of abandoned crop areas and filling in of stable forest areas in the latter, also the result of small farm vacancy. Full owner and tenant farms showed few linear increases or decreases and could not support Hypothesis 7 that linear increase would result in patchier, shifting forest outcomes. Population density has risen steadily across all counties but is still far below the average for the island except for in Hormigueros--the most dynamic county regarding forest cover change and the closest in vicinity to the urban agglomeration of Mayagez to the northwest. This research did not measure within-study area rural to
43 urban migration so this affect on forest cover change in the area cannot be ascertained here. However, looking at the spatial distribution of permanent forest clearing around major population centers suggests expansion of these areas at the expense of forest. Using only forest and non-forest land cover classifications in this study limited the ability to comment conclusively on forest change as a direct result of agriculture land; in addition, tree-like crops (bananas, coffee, etc.) were indistinguishable from other forest cover. The agriculture census data, however, gave some insight into agriculture's role in forest cover change outcome. The classification of forest cover over the three time periods was adequate in representing each time observation and thus in creation of the derived trajectory classes. However, the ten year interval between land cover classifications in what was found to be a very dynamic area would benefit from a higher temporal resolution of study dates for improved land cover observations, perhaps in sync with the NASS report years. The integration of simple landscape class metrics into land cover and land cover change extents added more explanatory power to understanding the nature of forest cover change in the area. The measures of NP and LPI effectively answered general questions about the degree of fragmentation for each land cover change outcome. Future research for this study area should include spatially-explicit pattern analyses where forest patches are assessed for their potential clearing probability as was performed in western Honduras by Nagendra et al. (2003). In addition, research investigating the structural metrics of just stable forest cover patches would be of ecological interest.
44 In comparing trends in agricultural activity to forest cover change outcome at the municipality level, use of the NASS data was insightful but proved inconclusive in this research for showing definitive causal relationships; as this was not a goal of the paper, the tables offered probably the best quantitative representation of agriculture trends over a long period of time and at the finer scale of counties. Potential variables to include in future research would be: indicators of type of farm or dominant farming item; number of years an operator was on the same farm and predominant income of farm operator. Measuring rural to urban migration within the region is a potential study to undertake as well as a survey of U.S.-based industry that might, as a result of the phasing out of industrial tax incentives, cease operations in the southwest counties. This could have enormous ramifications of land cover in the area. With the available data collected and analyzed via the established research questions, conclusions drawn for this study included As a region, forest cover did not experience large percentages of re-growth as found in other areas of the island, although forest stability and re-growth were abundant in the northern, higher elevation reaches of Sabana Grande and San German. The most extensive outcome of land cover change was older clearing of forest cover not to be re-grown to forest (F-NF-NF). Given the spatial distribution of this dynamic, it is assumed that these areas were converted to cattle/pasture use from tree-like crops post-1982. Most land cover change occurred in the elevational below 100m above sea level. This is easier land to develop or farm on and tends to contain the major transportation networks. LCC outcomes suggesting shifting type agriculture were found to be significant in extent in all counties. These areas were very patchy and averaged very small MPS and LPI values. It is assumed these represent small farms, possibly tenant-operated.
45 The agriculture variables used to help infer land use processes from the land cover change outcomes at a county-by-county analysis did an adequate job, but it was found an additional classification division of explicit cropland from non-forest would explain more about these relationships. In addition, the suggestion of large-scale cropland conversion to pasture in the southern counties warrants adding these data to the agriculture database. Although Puerto Rico is often cited as an exemplar case of successful re-forestation, this research found the area in the southwest to have a net loss of forest cover and very dynamic in non-linear, temporal landscape change. The growing built development in the southwest is encroaching on prime agricultural land that may, in the future, be called upon to be a domestic food source. Agricultural activity over the last 20 years has significantly affected the extent and homogeneity of forests; further investigation of socioeconomic drivers of these changes is warranted and would be facilitated by a combined analysis of U.S. Census Population and Housing and NASS statistics. The data and findings of this research are important to assigning focus to areas that need better forest management techniques as well as pinpointing areas of stable forest outside parks and reserves for potential conservation. The general findings of landscape structure and pattern, along with rates and directions of forest cover change will be useful to wildlife managers in researching specific habitats. Lastly, this research contributes to a relatively sparse collection of HDGC/LUCC studies on small tropical islands.
46 Table 2-1. Landsat-platform satellite imagery used in this study: raw characteristics Sensor/Date Path/row Spatial Resolution Correction Format Time-step MSS3 2.10.1982 005/047 80m Systematic NLAPS 1982-T1 MSS3 3.26.1981 005/048 80m Systematic NLAPS 1982-T1 TM4 8.19.1992 005/048 30m Systematic NLAPS 1992-T2 ETM+ 2.20.2002 005/048 30m Systematic HDF Fast 2002-T3
47 Table 2-2. Human dynamic variables used in the research Variable Source Time range Measured Hypothesized tie to land cover change Population 1 U.S. Census Data Table SF1 1970, 1980, 1990, 2000 Total numbers Population density; % population change Total farms 2 NASS Table 16 1978, 1982, 1987, 1992, 1998, 2002 Total number of farms for region and municipality Total number of farms will indicate intensity of agricultural activity Land in agriculture use 2 NASS Table16 1978, 1982, 1987, 1992, 1998, 2002 Hectares reported in ag-use Farm expansion Farm size 2 NASS Table 17 1978, 1982, 1987, 1992, 1998, 2002 Small farms=up to 20 ha. Large farms=>20 ha. Total number of farms for each size class in each municipality Farm size indicates: 1) small-holder vs. large-holder farms 2) farm influence on local economy 3) shifting vs. more permanent agriculture Farm ownership situation 2 NASS Table 20 1978, 1982, 1987, 1992, 1998, 2002 FOF=full own farm TEF=tenant Total number of farms reported in each operation class in each municipality Farm tenure indicates: 1) FOF are more permanent agriculture 2) TEF will represent shifting crops and will be the least maintained 1 U.S. Census Bureau Population and Housing 2 National Agriculture Statistics Service
48 Table 2-3. Accuracy assessment for 2002 and 1992 land cover classifications Time-step year Class name Ref totals Class. totals No. correct Prod acc.% Users acc.% Overall Accuracy % Kappa Statistics 2002 F 30 26 23 76.67 88.46 NF 30 34 27 90.00 79.41 83.33 0.77 1992 F 33 28 27 81.82 96.43 89.74 0.79 NF 45 48 43 95.56 89.58
49 Table 2-4. Regional land-cover classifications and change trajectory results Dataset Area in hectares % region or surface Assumption for study Ind. Observation 1982 F 31927 45.9 NF 37604 54.1 1992 F 23100 33.3 NF 46283 66.7 2002 F 30843 44.3 NF 38814 55.7 forest or crops, pasture, grass, built, other LCC Trajectories 1982-1992 F-F 17690 25.6 stable forest; protected; abandoned agriculture NF-F 5320 7.7 recent re-growth F-NF 14082 20.3 recent clearing NF-NF 32161 46.4 stable agriculture or built land 1992-2002 F-F 19217 27.7 same as 1982-1992 dataset NF-F 11425 16.5 F-NF 3868 5.6 NF-NF 34828 50.2 1982-2002 F-F-F 15851 22.9 stable forest; protected; abandoned agriculture F-F-NF 1833 2.6 recent clearing of older, stable forest F-NF-F 5183 7.5 shifting cover with recent re-growth F-NF-NF 8868 12.8 older clearing; built; permanent agriculture NF-F-F 3290 4.8 older re-growth of abandoned agriculture NF-F-NF 2022 2.9 shifting cover with recent clearing NF-NF-F 6225 9.0 recent re-growth of older agriculture NF-NF-NF 25919 37.5 stable agriculture or built land
50 Table 2-5. Characteristics of forest cover across the southwestern municipalities Land-cover dynamics over 20 years Municipality % Net F change (+,-) % area of 1982 F % original F left from 1982 % area of F re-growth % area of F clearing % area of stable F % area of stable NF % area of shifting LC Cabo Rojo -4.0 37.5 30.9 15.2 13.3 11.6 48.1 11.8 Guanica -16.2 54.8 48.7 7.1 23.3 26.7 33.9 9.0 Hormigueros +14.0 37.9 48.0 26.6 12.7 18.2 28.5 14.0 Lajas -10.3 29.7 21.2 8.2 18.3 6.3 59.4 7.8 Sabana Grande +2.5 60.4 65.6 13.4 10.9 39.6 24.6 11.5 San German +14.1 60.5 71.9 21.2 7.1 43.5 16.2 12.0 Table 2-6. Summary of class-level landscape metrics for the region Dataset LC AREAin ha. PLAND NP PD LPI MPS 1982 F 31927 45.9 1610 2.32 18.57 19.83 NF 37604 54.1 1315 1.89 48.66 28.60 1992 F 23100 33.2 1929 2.78 16.02 11.98 NF 46283 66.7 807 1.16 63.43 57.35 2002 F 30843 44.3 1639 2.35 29.42 18.81 NF 38814 55.7 1324 1.9 49.31 29.32 1982-1992 F Stable 17690 25.5 2411 3.48 12.45 7.34 NF-F 5320 7.7 5055 7.3 0.19 1.05 F-NF 14082 20.3 4937 7.13 1.27 2.85 NF Stable 32161 46.4 1940 2.8 40.81 16.58 1992-2002 F Stable 19217 27.7 2533 3.65 14.78 7.59 NF-F 11425 16.5 5472 7.89 2.19 2.09 F-NF 3868 5.6 4908 7.08 0.37 0.79 NF Stable 34828 50.2 1971 2.84 37.40 17.67 1982-2002 F Stable 15851 22.9 2408 3.48 11.64 6.58 F-F-NF 1833 2.7 3550 5.13 0.27 0.52 F-NF-F 5183 7.5 6466 9.34 0.09 0.80 F-NF-NF 8886 12.8 6066 8.76 0.38 1.46 NF-F-F 3290 4.8 4598 6.64 0.06 0.72 NF-F-NF 2022 2.9 3847 5.56 0.10 0.53 NF-NF-F 6225 9.0 6531 9.44 1.35 0.95 NF Stable 25919 37.5 2921 4.22 16.42 8.87
51 Table 2-7. Land cover change outcomes and associated landscape metrics in municipalities Landscape Class Metrics Municipality LCC outcome % land NP PD LPI MPS Cabo Rojo F-F-F 11.6 712 1.95 0.72 2.75 F-F-NF 1.8 824 2.26 0.03 0.37 F-NF-F 8.6 1708 4.67 0.17 0.85 F-NF-NF 11.5 1601 4.38 0.71 1.87 NF-F-F 3.7 1071 2.93 0.04 0.58 NF-F-NF 3.2 870 2.38 0.18 0.63 NF-NF-F 11.5 1798 4.92 1.05 1.09 NF-NF-NF 48.1 747 2.04 6.35 10.93 Guanica F-F-F 26.7 259 1.50 7.32 9.79 F-F-NF 7.7 618 3.58 1.07 1.18 F-NF-F 4.8 679 3.94 0.13 0.67 F-NF-NF 15.6 737 4.27 1.06 2.01 NF-F-F 2.1 428 2.48 0.06 0.46 NF-F-NF 4.2 636 3.69 0.08 0.61 NF-NF-F 5.0 533 3.09 0.32 0.89 NF-NF-NF 33.9 353 2.05 10.82 9.12 Hormigueros F-F-F 18.2 204 2.90 3.71 2.60 F-F-NF 3.6 310 4.40 0.15 0.34 F-NF-F 7.0 405 5.75 0.31 0.50 F-NF-NF 9.1 329 4.67 0.48 0.81 NF-F-F 10.9 365 5.19 0.24 0.87 NF-F-NF 7.0 372 5.29 0.28 0.55 NF-NF-F 15.7 418 5.94 1.59 1.10 NF-NF-NF 28.5 278 3.95 2.44 2.98 Lajas F-F-F 6.3 424 1.72 0.45 2.29 F-F-NF 2.1 632 2.57 0.04 0.52 F-NF-F 4.9 1084 4.41 0.16 0.70 F-NF-NF 16.2 1309 5.32 0.39 1.90 NF-F-F 1.7 624 2.54 0.04 0.42 NF-F-NF 2.9 737 3.00 0.16 0.60 NF-NF-F 6.5 1231 5.00 0.12 0.81 NF-NF-NF 59.4 406 1.65 33.09 22.53
52 Table 2-7. Continued. Landscape Class Metrics Municipality LCC outcome % land NP PD LPI MPS Sabana Grande F-F-F 39.6 342 2.00 17.99 10.69 F-F-NF 1.5 422 2.47 0.04 0.33 F-NF-F 9.9 967 5.66 0.22 0.95 F-NF-NF 9.4 864 5.05 0.16 1.00 NF-F-F 6.5 699 4.09 0.18 0.86 NF-F-NF 1.6 429 2.51 0.04 0.34 NF-NF-F 6.9 1006 5.89 0.10 0.63 NF-NF-NF 24.6 427 2.50 4.03 5.34 San German F-F-F 43.5 560 2.45 21.55 10.94 F-F-NF 1.5 749 3.28 0.02 0.29 F-NF-F 9.9 1725 7.56 0.11 0.81 F-NF-NF 5.6 1309 5.74 0.10 0.60 NF-F-F 9.1 1478 6.48 0.10 0.87 NF-F-NF 2.1 819 3.59 0.02 0.35 NF-NF-F 12.1 1710 7.49 1.71 1.00 NF-NF-NF 16.2 844 3.70 1.17 2.70
53 Table 2-8. Agriculture dynamics for 1978-2002 Agriculture variables: showing reported NASS numbers in 1978 to 1982 County # farms Land in agriculture use (hectares) # small farms < 20 ha. # large farms > 20 ha. #Full-owner farms #Tenant farms Forest cover extent in 1982 1978 1982 1978 1982 1978 1982 1978 1982 1978 1982 1978 1982 ha. % Cabo Rojo 445 183 12227 9918 372 124 73 59 144 55 24 16 6800 37.5 Guanica 211 80 4326 7440 191 55 20 25 29 18 1 15 5228 54.8 Hormigueros 139 75 n/a 1995 123 63 16 12 96 47 5 3 1107 37.9 Lajas 332 248 10712 12546 279 182 53 66 145 64 7 19 4592 29.7 SabanaGrande 435 204 4725 8830 390 158 45 46 174 111 7 11 5582 60.4 San German 568 308 8329 6421 482 233 86 75 369 196 23 16 8516 60.5 1982-1992 F in 1992 1982 1992 1982 1992 1982 1992 1982 1992 1982 1992 1982 1992 ha. % Cabo Rojo 183 167 9918 7160 124 108 59 59 55 72 16 58 3490 19.3 Guanica 80 33 7440 1213 55 22 25 11 18 7 15 21 3860 40.6 Hormigueros 75 67 1995 1369 63 50 12 17 47 29 3 28 1158 40.0 Lajas 248 150 12546 9276 182 104 66 46 64 59 19 46 2031 13.2 SabanaGrande 204 162 8830 3624 158 120 46 42 111 129 11 12 4543 49.2 San German 183 167 9918 7160 124 108 59 59 55 72 16 58 3490 19.3 1992-2002 F in 2002 1992 2002 1992 2002 1992 2002 1992 2002 1992 2002 1992 2002 ha. % Cabo Rojo 167 178 7160 6072 108 113 59 65 72 51 58 63 6050 33.5 Guanica 33 43 1213 2377 22 31 11 12 7 22 21 16 3666 38.6 Hormigueros 67 40 1369 685 50 34 17 6 29 12 28 7 1512 51.9 Lajas 150 160 9276 9634 104 91 46 69 59 50 46 49 2994 19.4 SabanaGrande 162 134 3624 2412 120 111 42 23 129 97 12 16 5807 62.9 San German 489 344 4970 4967 443 302 46 42 404 273 58 12 10509 74.6
54 Figure 2-1. Job sector characteristics of Puerto Rico: 1940-2000
55 A B Figure 2-2. Study area characteristics. A) Southwest study area. B) Study area extent relative to Florida
56 A B Figure 2-3. Topography and physical regions of the southwest. A) Elevation. B) Percent slope
57 Figure 2-4. Population characteristics of the southwest. A) Population map. B) County trends in population from 1940-2000 A B
58 C Figure 2-5. Results of the forest and non-forest land cover classifications: 1982-2002. A) 1982. B) 1992. C) 2002 B A
59 Figure 2-6. Land-cover change trajectory extent and results for single time steps. A) 1982-1992 change. B) 1992-2002 change B A
60 Figure 2-7. Twenty-year land-cover change outcome for southwest Puerto Rico: 1982-2002
61 A B C Figure 2-8. Land cover change trajectory proportions across counties. A) 1982-1992. B) 1992-2002. C) 1982-2002
62 Figure 2-9. Land-cover change outcomes in counties: 1982-2002
63 Figure 2-10. Forest dynamics across the six southwestern counties: 1982-2002. A) Selected forest dynamics across counties. B) County extent of forest cover at all time observations B A
64 Figure 2-11. Distribution of 20-year LCCT outcomes and their associated metrics at the class level
65 #farms= total number of farms reported for each county #small farms= total number of small farms ( < 20 ha.) reported for each county #large farms= total number of small farms ( > 20 ha.) reported for each county #fof farms= total number of full-owned farms reported for each county #tef farms= total number of tenant farms reported for each county Source: National Agriculture Statistics Service Figure 2-12. Agriculture trends across the study area: 1978-2002
CHAPTER 3 SUMMARY AND THESIS CONCLUSIONS Contrary to the existing research findings in Puerto Rico (Thomlinson et al., 1996; Rudel and Lugo, 2000; Grau et al., 2003), this research found significant forest-cover instability and a 3.4% net loss of forest-cover in the region of the southwest island from 1982-2002; the six municipalities that comprise the region experienced quite different dynamics in forest change outcomes across the 20 years. Land-cover change trajectory outcomes determined that the most extensive dynamics regionwide was indicative of older and permanent deforestation, probably land going into either pasture or expanding urbanization. Northern counties, consequently in higher elevations, experienced more forest re-growth while their southern counterparts had more clearing. Most land-cover change activity happened in elevations below 100 m. Landscape metrics showed forest patchiness decreasing along with increasing LPI in higher elevations and increasing in counties with high forest clearing events. Agriculture variables used in this research to assess the human tie to forest cover change suggested a much more complex interaction than expected; although the number of farms experienced a steep decline from 1978-1982 in all municipalities, few additional linear trends are seen in the data. It appears agriculture in the southwest remains a dominant factor on the landscape and is causing forest clearing in some places. Specific findings from this research are As a region, forest cover did not experience large percentages of re-growth as found in other areas of the island, although forest stability and re-growth were abundant in the northern, higher elevation reaches of Sabana Grande and San German. 66
67 The most extensive outcome of land cover change was older clearing of forest cover not to be re-grown to forest (F-NF-NF). Given the spatial distribution of this dynamic, it is assumed that these areas were converted to cattle/pasture use from tree-like crops post-1982. Most land cover change occurred in elevations < 100 m above sea level. This is easier land to develop or farm on and tends to contain the major transportation networks. LCC outcomes suggesting shifting type agriculture were found to be significant in extent in all counties. These areas were very patchy and averaged very small MPS and LPI values. It is assumed these represent small farms, possibly tenant-operated. The agriculture variables used to help infer land use processes from the land cover change outcomes at a county-by-county analysis did an adequate job, but it was found an additional classification division of explicit cropland from non-forest would explain more about these relationships. In addition, the suggestion of large-scale cropland conversion to pasture in the southern counties warrants adding these data to the agriculture database. This research used remote sensing, GIS and methods of geographic science to achieve a reasonable quantification and digital map of forest cover for 1982, 1992 and 2002; in addition, it quantified and mapped outcomes of change between forest and all non-forest land cover classes. It used simple landscape metrics to better infer forest dynamics and evaluated forest transformation in topographic gradients. The study derived a 24-year database of a few agriculture variables chosen to measure the level of agricultural strength across the municipalities of the southwest island and evaluated the trends in agricultural change in relation to forest cover change. The research findings were compared to other similar, contemporary research on Puerto Rico land cover change. The findings of this research address some major issues encompassing the LUCC-HDGC initiative to relate biophysical land cover change to the activity of humans. The discipline of geography is centered on place, space and time and in this, lends great asset to joining and scaling interdisciplinary data regarding landscapes. The data created in
68 this study: 1) established a base from which future analyses can use as input and 2) can be used in numerous practical applications of GIS for local and regional land managers. Further directions of research for the study area include 1) performing an accuracy assessment on the land cover change trajectories, 2) spatial identification and sampling of farm size and tenure classes in each county, 3) further classification to distinguish forest-like crops from secondary forests, 4) land cover classifications to fill in interval dates, shrinking the temporal resolution of the analyses and 5) investigation of how each municipality in the study area contributes to the control and management of its land resources. In summary, this research adds to the thin library of case studies looking at forest cover dynamics on the islands of the Caribbean Basin and islands in general, tropical land cover change where documented re-forestation is more predominant than deforestation and the use of remotely-sensed derived land cover classifications, census data and landscape measures to better infer pattern and process of forest change. The last decade has seen tremendous advances in the conceptualization of both land cover change impacts on the environment and the human activity that drives changing of land use and cover. Technology of geographic science has offered exceptional data, tools and methods for LUCC investigation. LUCC research agendas have helped guide the content and protocol of case study but the variance of this goes along with the scores of different geographies of the world and many are under-represented, like islands and tropical dry forests. As more results come in from these types of areas, we can begin to make generalizations that help guide local, regional and world policy regarding these minute but very complex gems of the globe.
LIST OF REFERENCES Abizaid, C. and Coomes, O., 2004. Land use and forest fallowing dynamics in seasonally dry tropical forests of the southern Yucatan Peninsula, Mexico. Land Use Policy 21, 71-84. Aide, M., Zimmerman, J., Herrera, L., Rosario, M., Serrano, M., 1995. Forest recovery in abandoned tropical pastures in Puerto Rico. Forest Ecology and Management 77, 77-86. Areces-Mallea, A., Weakly, A., Li, X., Sayre, R., Parrish, J., Tipton, C., Boucher, T., 1999. A Guide to Caribbean Vegetation Types: Preliminary Classification Systems and Descriptions. Washington D.C., The Nature Conservancy. Ayala-Silva, T. and Twumasi, Y., 2004. Hurricane Georges and vegetation change in Puerto Rico using AVHRR satellite data. International Journal of Remote Sensing 25 (9), 1629-1640. Binford, M., Lee, T., Townsend, R., 2004. Sampling design for an integrated socioeconomic and ecological survey by using satellite remote sensing and ordination. Proceedings of the National Academy of Sciences of the United States of America. 101 (31), 11,517-11522. Briassoulis, H., 2000. Analysis of Land Use Change: Theoretical and Modeling Approaches. In: Loveridge S (Ed), The Web Book of Regional Science (www.rri.wvu.edu/regscweb.htm). Morgantown, WV, Regional Research Institute, West Virginia University. Brown, S. and Lugo, A., 1990. Tropical secondary forests. Journal of Tropical Ecology 6 (1), 1-32. del Mar Lpez, T., Aide, M., Thomlinson, J., 2001. Urban expansion and the loss of prime agricultural lands in Puerto Rico. Ambio 30 (1), 49-54. de Sherbinin, A., Balk, D., Yager, K., Jaiteh, M., Pozzi, F., Giri, C., Wannebo, A., 2002. A CIESIN Thematic Guide to Social Science Applications of Remote Sensing. Center for International Earth Science Information Network of Columbia Univeristy. Palisades, NY, Columbia University Press. Dietz, J., 1986. Economic History of Puerto Rico. Princeton, NJ, Princeton University Press. 69
70 Douglas, I., 1994. Land degradation in the humid tropics. In: Roberts N. (Ed), The Changing Global Environment. Cambridge, Blackwell. ERDAS, Inc. 1996: Earth Resources Data Analysis System Field Guide, Fifth Edition. Atlanta, ERDAS. Evans, T. and Moran, E., 2002. Spatial integration of social and biophysical factors related to land-cover change. Population and Development Review 28, Supplement: Population and Environment: Methods of Analysis, 165-186. Evans, T., Munroe, D., Parker, D., 2005. Modeling land-use/land-cover change: exploring the dynamics of human-environment relationships. In: Moran E and Ostrom E (Eds), Seeing the Forest and the Trees: Human-Environment Interactions in Forest Ecosystems. Cambridge, MA, MIT Press, pp.187-213. Ewel, J. and Whitmore, J., 1973. The ecological lifezones of Puerto Rico and the U.S. Virgin Islands. U.S. Forest Service. Research Paper ITF-18. Rio Piedras (Puerto Rico) Food and Agriculture Organization [FAO]. Global Forest Resources Assessment 2000: Main Report. Rome. Franco, P., Weaver, P., Eggen-McIntosh, S., 1997. Forest Resources of Puerto Rico, 1990 Resource Bulletin SRS-22. Asheville, NC: U.S. Department of Agriculture, Forest Services, Southern Research Station. Geist, H. and Lambin, E., 2001. What drives tropical deforestation? LUCC Series No. 4. Louvain la Neuve, Belgium, LUCC International Project Office. Geist, H. and Lambin, E, 2002. Proximate causes and underlying driving forces of tropical deforestation. Bioscience 52, 143-150. Grau, H., Aide, M., Zimmerman, J., Thomlinson, J., Helmer, E., Zou, X., 2003. The ecological consequences of socioeconomic and land-use changes in post-agricultural Puerto Rico. Bioscience 53 (12), 1159-1168. Green, G., Schweik, C., Randolph, J., 2005. Useful concepts and approaches for HDGC research: scales of space, time and human decision-making. In: Moran, E. and Ostrom, E. (Eds), Seeing the Forest and the Trees: Human-Environment Interactions in Forest Ecosystems. Cambridge, MA, MIT Press, pp.131-160. Griffiths, G. and Mather, P., 2000. Remote sensing and landscape ecology: landscape patterns and landscape change. International Journal of Remote Sensing 21 (13/14), 2537-2539. Gustafson, E., 1998. Quantifying landscape spatial pattern: what is the state of the art? Ecosystems 1, 143-156.
71 Haines-Young, R., 2005. Landscape pattern: context and process. In: Wiens J and Moss M (Eds), Issues and Perspectives in Ecology. Cambridge (UK): Cambridge University Press. Helmer, E., 2004. Forest conservation and land development in Puerto Rico. Landscape Ecology 19, 29-40. Helmer, E., Ramos, O., del Mar Lpez, T., Quiones, M., Diaz, W., 2002. Mapping the forest type and land cover of Puerto Rico, a component of the Caribbean biodiversity hotspot. Caribbean Journal of Science 38 (3-4), 165-183. Houghton, R., 1994. The worldwide extent of land-use change. Bioscience 44 (5), 305-309. Jensen, J., 1996. Introductory Digital Image Processing: A Remote Sensing Perspective. Upper Saddle River, NJ, Prentice-Hall. Kauneckis, D., Brondizio, E., Greene, G., Schweik, C., Carlson, L., Slusher, B., Molnar, M., Summers, K., Shields, S., Ivie, S., Day, A., 1998. CIPEC Training Sample Protocol. CIPEC Internet Website: http://cipec.org/research/methods/tsinstructions.pdf . Accessed March, 2006. Keys, E. and McConnell, W., 2005. Global change and the intensification of agriculture in the tropics. Global Environmental Change 15, 320-337. Lambin, E., Geist, H., Lepers, E., 2003. Dynamics of land-use and land-cover change in tropical regions. Annual Review of Environmental Resources 28, 205-41. Lambin, E., Turner, B., Geist, H., Agbola, S., Angelsen, A., Bruce, J., Coomes, O., Dirzo, R., Fischer, G., Folke, C., George, P., Homewood, K., Imbernon, J., Leemans, R., Li, X., Moran, E., Mortimore, M., Ramakrishnan, P., Richards, J., Skanes, H., Steffen, W., Stone, G., Svedin, U., Veldkamp, T., Vogel, C., Xu, J., 2001. The causes of land-use and land-cover change: moving beyond the myths. Global Environmental Change 11 (4), 261-269. Lugo, A., 2002. Can we manage tropical landscapes?An answer from the Caribbean perspective. Landscape Ecology 17, 601-615. Lugo, A., Schmidt, R., Brown, S., 1981. Tropical Forests in the Caribbean. Ambio 10, 318-324. Marcano-Vega, H., Aide, M., Baez, D., 2002. Forest regeneration in abandoned coffee plantations and pastures in the Cordillera Central of Puerto Rico. Plant Ecology 161, 75-87. Mather, A. and Needle, C., 1998. The forest transition: a theoretical basis. Area 30 (2), 117-124.
72 McGarigal, K. and Marks, B., 1995. FRAGSTATS: Spatial pattern analysis program for quantifying landscape structure, Vol. 2.0. Forest Science Lab, Oregon State University, Corvalis. Portland, USDA Forest Service, Pacific Northwest Research Station; General Technical Report PNW-GTR-351. Mertens, B., Lambin, E., 2000. Land-cover change trajectories in Southern Cameroon. Annals of the Association of American Geographers 90 (3), 467-494. Morales-Carrion, A., 1983. Puerto Rico, a political and social history. New York, Norton. Nagendra, H., Southworth, J., Tucker, C., 2003. Accessibility as a determinant of landscape transformation in western Honduras: linking pattern and process. Landscape Ecology 18, 141-158. Nagendra, H., Monroe, D., Southworth, J., 2004. From pattern to process: landscape fragmentation and the analysis of land use/land cover change. Agriculture, Ecosystems and Environment 101, 111-115. National Research Council [NRC], 1998. Human Dimensions of Global Environmental Change. In Global Environmental Change : Pathways for the Next Decade. Washington, D.C.: Academy Press. Pascarella, J., Aide, M., Serrano, M., Zimmerman, J., 2000. Land-use history and forest regeneration in the Cayey Mountains, Puerto Rico. Ecosystems 3, 217-228. Perz, S. and Skole, D., 2003. Social determinants of secondary forests in the Brazilian Amazon. Social Science Research 32, 25-60. Pico, R., 1974. The Geography of Puerto Rico. Chicago, Aldine. Ramos-Gonzlez, O., 2001. Assessing vegetation and land cover changes in northeastern Puerto Rico: 1978-1995. Caribbean Journal of Science 37 (1-2), 95-106. Randolph, J., Green, G., Belmont, J., Burcsu, T., Carlson, L., Vogt, N., Welch, D., 2005. Forest ecosystems, ecology and the human dimensions. In: Moran, E. and Ostrom, E., (Eds), Seeing the Forest and the Trees. Cambridge, MA, MIT Press, pp.105-125. Richards, J., 1990. Land transformation. In: Clark, W., Kates, R., Richards, J., Mathews, M., Meyer, W., Turner, B., (Eds), The Earth as Transformed by Human Action. Cambridge (UK), Cambridge University Press, pp.163-178. Roberts, N., 1994. The global environmental future. In: Roberts, N., (Ed), The Changing Global Environment. Cambridge, MA, Blackwell. Roberts, R., 1942. Soil Survey of Puerto Rico. USDA Series 1936, No. 8 U.S. Printing Office, Washington, D.C.
73 Rogan, J. and Chen, D., 2004. Remote sensing technology for mapping and monitoring land-cover and land-use change. Progress in Planning 61, 301-325. Rudel, T., 1998. Is there a forest transition? Deforestation, reforestation and development. Rural Sociology 63 (4), 533-552. Rudel, T., 2001. Sequestering carbon in tropical forests: experiments, policy implications and climate change. Society of Natural Resources 14, 525-531. Rudel, T., Perez-Lugo, M., Zichal, H., 2000. When fields revert to forest: development and spontaneous reforestation in post-war Puerto Rico. Professional Geographer 52, 186-397. Silver, W., Osterag, R., Lugo, A., 2000. The potential for carbon sequestration through reforestation of abandoned tropical agricultural and pasture lands. Restoration Ecology 8 (4), 394-407. Singh, A., 1989. Digital change detection techniques using remotely-sensed data. International Journal of Remote Sensing 10 (6), 989-1003. Sotomayor, O., 2004. Development and income distribution: the case of Puerto Rico. World Development 32 (8), 1395-1406. Southworth, J., Monroe, D., Nagendra, H., 2004. Land cover change and landscape fragmentation-comparing the utility of continuous and discrete analyses for a western Honduras region. Agriculture, Ecosystems and Environment 101, 185-205. Southworth, J., Nagendra, H., Tucker, C., 2002. Fragmentation of a landscape: incorporating landscape metrics into satellite analyses of land-cover change. Landscape Research 27 (3), 253-269. Suarez, N., 1998. The rise and decline of Puerto Rico's sugar economy. Washington D.C., Economic Research Service/USDA. Thomlinson, J., Serrano, T., Lpez, M., Aide, M., Zimmerman, J., 1996. Land-use dynamics in a post-agricultural Puerto Rican landscape (1936-1988). Biotropica 28, 525-536. Tucker, C. and Ostrom, E., 2005. Institutional analysis in multidisciplinary research on human dimensions of global environmental change. In: Moran, E. and Ostrom, E., (Eds), Seeing the Forest and the Trees. Cambridge, MA, MIT Press, pp.81-103. Tucker, C. and Southworth, J., 2005. Processes of forest change at the local and landscape levels in Honduras and Guatemala. In: Moran, E. and Ostrom, E., (Eds), Seeing the Forest and the Trees. Cambridge, MA, MIT Press, pp.253-277.
74 Turner, B., Clark, W., Kates, W., Richards, J., Mathews, J., Meyer, W., (Eds), 1990. The Earth as Transformed by Human Action. Cambridge (UK), Cambridge University Press. Turner, B. and Meyer, W., 1994. Global land-use and land cover change: an overview. In: Turner, B. and Meyer, W., (Eds), Changes in Land Use and Land Cover: A Global Perspective. Cambridge, MA, University Press, pp.1-3. Turner, B., Skole, D., Sanderson, S., Fischer, G., Fresco, L., Leemans, R., 1995. Land-use and Land cover change. Science/Research Plan. Stockholm and Geneva, IHDP Report No. 35 and HDP Report No. 7. Van Wey, L., Meretsky, V., Ostrom, E., 2005. Theories underlying the study of human dimensions of global environmental change. In: Moran, E. and Ostrom, E., (Eds), Seeing the Forest and the Trees. Cambridge, MA, MIT Press, pp. 23-56. Wallach, B., 1989. Puerto Rico: growth, change, progress, development. Focus 39 (2), 27-33. Williams, M., 1994. Forests and tree cover. In: Turner B. and Meyer W., (Eds), Changes in Land Use and Land Cover: A Global Perspective. Cambridge, University Press, pp.97-124. Wolman, M. and Fournier, F., (Eds), 1987. Land Transformation in Agriculture. Chichester, (UK), John Wiley and Sons.
BIOGRAPHICAL SKETCH Robert D. Lopez was born in Williamsburg, VA and grew up in a small Chesapeake Bay county east of the York River. Robert received a B.S. in geography from Old Dominion University, Norfolk, Virginia and a M.S. in geography from University of Florida in 2006. Robert currently lives with his wife Amanda in Ohio, where they are busy pursuing daily work, dreams, good health, and travel. 75