COFFEE AND LANDSCAPE CHANGE IN THE COLOMBIAN COUNTRYSIDE 1970-2002 By ANDRES GUHL A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2004
Copyright 2004 by Andres Guhl
ACKNOWLEDGMENTS I want to thank Dr. Gabriel Cadena (director of Cenicaf) for his unconditional support during the field work seasons I spent in Colombia in the summers of 2000, 2001, and the Fall semester of 2002. I also want to thank Dr. Jorge Botero, Oscar Orrego, Harold Castao, Andrs Echeverry, and Margarita Jaramillo of Cenicafâ€™s Conservation Biology group; and Dr. Orlando Guzmn and Dr. Alvaro Jaramillo of the Agroclimatology group for their valuable comments, support, and collaboration while doing my research in Colombia. I also want to thank Nancy Delgado for her hospitality while I stayed in Manizales. I want to thank the Tropical Conservation and Development program from the Center of Latin American Studies at the University of Florida for the financial support provided to carry out pre-dissertation field reconnaissance in 2001. I also want to thank Natalia Hoyos, my dissertation process partner, whose constant support helped me to finish my dissertation on time. Finally, I would like to thank all my friends here at the University of Florida, but specially Amy Daniels, Matt Marsik, Claudia Stickler, Tracy van Holt, Jane Southworth, and Lin Cassidy for all their help and support. iii
TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................iii LIST OF TABLES ...........................................................................................................viii LIST OF FIGURES .............................................................................................................x ABSTRACT .....................................................................................................................xiii CHAPTER 1 INTRODUCTION........................................................................................................1 Land-Use and Land-Cover Change (LUCC) and Global Environmental Change.......5 Human Activities and Landscape Change....................................................................7 Agricultural Intensification and Commercialization and LUCC................................11 Coffee and Landscape Change...................................................................................12 2 LITERATURE REVIEW...........................................................................................15 Land-Use and Land-Cover Change as a Research Topic...........................................15 Definitions of Land Use and Land Cover............................................................17 Land-Cover Change vs. Land-Use Change.........................................................18 The LUCC Initiative............................................................................................20 How Land-Use and Land-Cover Change Take Place..........................................22 Key Concepts of Land-Use and Land-Cover Change.........................................27 Spatial patterns are essential........................................................................27 Land-cover change is both a natural and human-induced process...............27 Natural landscapes may be the result of intentional management...............28 Land-cover dynamics...................................................................................29 Market involvement and land tenure: critical factors...................................30 The LUCC is the result of a few key processes shaping the landscape.......31 Spatial diversity of patterns of landscape change........................................32 Reconstructing land-use and land-cover change is challenging...................33 Recent Developments in Understanding the LUCC Process..............................34 The LUCC as a Multi-scale Process....................................................................36 Complexity of the Land-Use and Land-Cover Change Process..........................38 Modeling and LUCC...........................................................................................42 Agricultural Intensification and Commercialization..................................................46 iv
Measures of Agricultural Intensity......................................................................48 Agricultural Change Requires an Integrated Perspective....................................49 Conditions for Successful Intensification/Commercialization............................49 Pathways of Intensification and Commercialization...........................................51 Intensification, Commercialization, Agricultural Diversity, and Landscape Change.............................................................................................................52 Models of Agricultural Change: Intensification and Commercialization...........55 Boserupâ€™s model of agricultural intensification...........................................56 Chayanovâ€™s model of peasant farms............................................................57 Population cycle...........................................................................................59 Commercialization of agriculture.................................................................60 Induced intensification.................................................................................63 No Existing Single Model Can Adequately Explain Agricultural Development....................................................................................................65 Putting It All Together: Agriculture and Landscape Change.....................................67 Landscape Manifestations of Agricultural Intensification and Commercialization Conceptual Frameworks..................................................................................67 Landscapes and Agricultural Change..................................................................73 Research Questions and Hypotheses...................................................................75 Landscapes diversify as commercial intensification of agriculture takes place........................................................................................................76 In a region undergoing intensification, some land-use practices intensify while others disintensify.........................................................................77 Factors accompanying intensification vary with the context.......................78 3 COFFEE AS A CROP IN THE INTERNATIONAL MARKET...............................80 Coffee as a Crop.........................................................................................................80 Brief History of Coffee Production: Shifting Centers of Production.........................86 Early Beginnings in Africa and Arabia...............................................................86 To the East Indies and Europe.............................................................................88 Arrival to the New World....................................................................................89 Coffee and Its Transforming Power....................................................................91 Coffee Production Systems.........................................................................................94 Traditional Production System............................................................................97 The Intensive Coffee Production System..........................................................100 Preparing the Coffee Cherry for the Market......................................................104 Coffee in the International Market...........................................................................105 The Marketing Chain................................................................................................111 4 THE STAGE: COFFEE IN COLOMBIA................................................................115 Arrival and Spread of Coffee in Colombia...............................................................119 Models of Landscape Change 1850-1970: The Coffee Economy............................123 Contraction and Expansion in the Coffee Economy.................................................128 Model of Landscape Change 1970-2002: The Intensification of Production...........130 Shifting Centers of Production.................................................................................136 v
National Coffee Growers Federation: The Supporting Role of Juan Valdez...........136 5 METHODS AND DATA SOURCES......................................................................142 Data Sources.............................................................................................................142 Primary Data Sources........................................................................................142 Agricultural extension agent questionnaires..............................................142 Farmer interviews.......................................................................................146 Interviews with key informants..................................................................146 Secondary Data Sources....................................................................................147 The FNC coffee censuses...........................................................................147 Population census data...............................................................................150 Municipios y regiones de Colombia...........................................................151 Ministerio de agricultura. Estadsticas agrcolas por consenso..................152 Accessibility...............................................................................................153 Rainfall.......................................................................................................154 Ecotopes.....................................................................................................155 Location......................................................................................................155 Software Packages Used for Research.....................................................................156 Methods....................................................................................................................156 As Intensification Takes Place, Landscapes Diversify......................................156 Land-cover evolution.................................................................................156 Land-use system evolution.........................................................................157 Increasing agricultural diversity.................................................................158 Some Land Uses Intensify while Others Disintensify.......................................159 Evidence of coffee production intensification...........................................159 Evidence of simultaneous crop intensification and disintensification.......160 The Factors Accompanying Intensification Change from Region to Region...162 Exploratory analysis...................................................................................162 Principal components analysis...................................................................164 Multivariate regression...............................................................................166 Regional regression analysis......................................................................166 6 RESULTS.................................................................................................................168 Evidence of the Intensification of Coffee Production..............................................168 Evidence of Landscape Diversification....................................................................180 Simultaneous Intensification and Disintensification of Crops.................................192 Factors Accompanying Coffee Production Intensification in Different Areas of Colombia..............................................................................................................216 Principal Components Analysis........................................................................216 Multiple Linear Regression Analysis................................................................224 Regression Analysis at the Regional Level.......................................................227 7 DISCUSSION...........................................................................................................245 The LUCC and Coffee Production...........................................................................245 vi
Does Coffee Follow the Models of Commercial Intensification of Agriculture?....253 The Future of Coffee in Colombia: Future Landscapes and Prospects....................258 8 CONCLUSIONS, LIMITATIONS, AND FUTURE RESEARCH.........................270 Limitations of this Research.....................................................................................282 Further Questions and Future Research....................................................................285 APPENDIX A AGRICULTURAL EXTENSION AGENT QUESTIONNAIRE...........................289 B SPATIAL DISTRIBUTION OF THE VARIABLES USED IN THE REGRESSION ANALYSIS...............................................................................................................292 Original Variables.....................................................................................................292 Principal Components...............................................................................................307 LIST OF REFERENCES.................................................................................................314 BIOGRAPHICAL SKETCH...........................................................................................329 vii
LIST OF TABLES Table page 1. Land-cover conversion and modification.................................................................20 2. Coffee production systems characteristics...............................................................96 3. Traditional coffee production system advantages and disadvantages......................99 4. Intensive coffee production system advantages and disadvantages.......................102 5. Variables pre-selected for the regional analysis.....................................................163 6. Variables selected for multivariate regression analysis.........................................165 7. Summary of FNC coffee censuses.........................................................................171 8. Evolution of the area planted in coffee by time period..........................................175 9. Evidence of coffee production concentration in certain areas...............................177 10. Evolution of the proportion of the total area planted in coffee under the intensive production system..................................................................................................180 11. Descriptive statistics for the 1970 and 1993/97 coffee censuses ratios.................183 12. Evolution of the proportion of the area in coffee,pasture, and other land covers 1970-1993/97.........................................................................................................183 13. Major land-cover increase and decrease as a proportion 1970-1993/97................184 14. Standard deviations for each land cover around the cluster means of the K-Means Cluster Analysis for 1970.......................................................................186 15. Evolution of land-use system types, 1970-1993/97...............................................186 16. Crops with annual yield information (1988-2000) in the coffee growing departments............................................................................................................194 17. Cumulative year-to-year variations (regardless of sign) presented as a percentage for the coffee growing departments (1988-2000)................................203 viii
18. Correlation coefficients among different crop yields through time (1988-2000)............................................................................................................205 19. Number of statistically significant correlations among crop yields (yearly data)............................................................................................................207 20. Number and percentage of statistically significant correlations for each crop (annual yields)........................................................................................................208 21. Correlation coefficients among different crop yields through time (3-yr moving average), 1988-2000.........................................................................208 22. Number of statistically significant correlations among crop yields (3-yr moving average)............................................................................................213 23. Number and percentage of statistically significant correlations for each crop (3-yr moving average)............................................................................................213 24. Comparison of number of statistically significant crop relations for the annual yield time series and the 3-yr moving average time series.....................................214 25. Principal component analysis results on the factors accompanying agricultural intensification (Eigenvalues)..................................................................................215 26. Principal component analysis standardized Eigenvectors......................................217 27. Degree of correlation between variables and principal components......................218 28. Multiple linear regression models characteristics..................................................225 29. Possible landscape transformations associated with the commercial intensification of agricultural production...............................................................254 30. Variables selected for regression analyses and regional breakdowns....................292 ix
LIST OF FIGURES Figure page 1. Land-use and land-cover change cycle....................................................................23 2. Agricultural intensification and commercialization.................................................52 3. Coffee marketing chain..........................................................................................113 4. Location and general characteristics of the Colombian coffee lands.....................116 5. Coffee growing landscape......................................................................................117 6. Spread of coffee in Colombia.................................................................................120 7. Landscape evolution paths 1850-1970...................................................................123 8. Landscape evolution path 1970-2002....................................................................133 9. Erosion in a poorly managed intensive coffee plot................................................135 10. Evolution of the coffee growing area in Colombia................................................168 11. Evolution of total coffee production and yield 1970-2002....................................169 12. Price paid to the farmer for 125 kg of parchment coffee.......................................169 13. Evolution of the area planted in coffee 1970-1997 in the Colombian coffee growing municipalities...........................................................................................173 14. Evolution of the area planted in coffee by time period..........................................174 15. Evolution of coffee intensification in Colombia....................................................178 16. Evolution of land covers 1970-1993/97.................................................................181 17. Evolution of land covers as a percentage 1970-1993/97........................................181 18. Land-use systems...................................................................................................185 19. Evolution of land-use systems in Colombia 1970-1993/97 (n=503).....................187 x
20. Time of service of the agricultural extension agents by municipality...................189 21. Municipalities where crops are increasing in importance 1997-2002....................190 22. Municipalities where crops are decreasing in importance 1997-2002...................191 23. Recent trends in agricultural diversity...................................................................193 24. Evolution of crop yields from 1988 to 2000 in the coffee growing departments..195 25. Scatter plot of PC1 and PC2 values for the coffee growing municipalities...........219 26. Principal component multivariate linear regression residuals (no interactions among independent variables)............................................................................................220 27. Principal component multivariate linear regression residuals (interactions among independent variables)............................................................................................221 28. Multiple linear regression residuals.......................................................................226 29. Regional breakdown by mountain range for regression analysis...........................228 30. Regional breakdown by farm size for regression analysis.....................................229 31. Regional breakdown by Quality of Life Index (ICV) for regression analysis.......230 32. Box-Plot of the area in intensivecoffee per municipality by mountain range........231 33. Box-Plot of the area in intensive coffee per municipality by average farm size....233 34. Box-Plot of the area intensive coffee per municipality by Quality of Life Index (ICV)......................................................................................................................233 35. Multiple linear regression residuals. Regions by mountain range.........................235 36. Multiple linear regression residuals. Regions by farm size...................................236 37. Multiple linear regression residuals. Regions by Quality of Life Index (ICV)......237 38. Factors affecting the area planted in coffee in Colombia and their effects............266 39. Average cumulative rainfall in the two driest consecutive months.......................294 40. Average cumulative rainfall in the two wettest consecutive months.....................295 41. Rural population density........................................................................................296 42. Average age in rural areas......................................................................................298 xi
43. Number of dependents per rural household...........................................................299 44. Average size of coffee farms..................................................................................300 45. Number of financial institutions per municipality.................................................302 46. Number of state institutions per municipality........................................................303 47. Average distance to the closest town.....................................................................304 48. Municipal rural Quality of Life Index (ICV).........................................................305 49. Spatial distribution of the first principal component in the coffee growing municipalities.........................................................................................................309 50. Spatial distribution of the second principal component in the coffee growing municipalities.........................................................................................................310 51. Spatial distribution of the third principal component in the coffee growing municipalities.........................................................................................................311 52. Spatial distribution of the fourth principal component in the coffee growing municipalities.........................................................................................................312 xii
Abstract of dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy COFFEE AND LANDSCAPE CHANGE IN THE COLOMBIAN COUNTRYSIDE 1970-2002 By Andres Guhl May 2004 Chair: Michael W. Binford Major Department: Geography Land-use and land-cover change (LUCC) are important contributors to global environmental change. Landscape transformations lead to changes in nutrient cycling, hydrologic regulation, and climate among others at local, regional, and global scales. Presently, most landscape transformations are the result of human activities. Therefore, the study of how humans influence landscape patterns is an integral part of global environmental change research. Agriculture is one of the human activities that influences more landscape transformations. These changes can be the result of other land covers being replaced by agricultural fields (e.g., deforestation to open new land for cultivation) or changes in the management of agricultural fields, so as to increase productivity per unit of area (e.g., agricultural intensification). The first process has received more attention in the scientific literature than the second one. This dissertation contributes to this knowledge gap by xiii
analyzing the landscape impacts of commercial intensification of coffee agriculture in Colombia between 1970 and 2002. The results indicate there are widespread landscape transformations associated with agricultural intensification. For this case study, the total area planted in coffee decreased by more than 18% between 1970 and 2002. At the same time, total coffee production has increased about 50%. These two conditions are clear signs of agricultural intensification. The major landscape changes during the same period have been a sharp decrease in the area in pasture, and an increase in the area planted in other crops (e.g., fruit orchards, vegetables, cassava, plantains, and pineapple among others) . Therefore, the landscape is becoming more agriculturally diverse. There is evidence for both simultaneous intensification, and intensification/disintensification of crops. Finally, the variables usually associated with agricultural intensification at the household level do not exhibit a strong relationship with the area in intensive coffee at the municipal level. At this level, the variables that are more closely related to the area in intensive coffee are biophysical. In conclusion, the results of this dissertation suggest that agricultural intensification leads to widespread landscape transformation and agricultural diversification, emphasizing the importance of studying how different management practices change landscapes. xiv
CHAPTER 1 INTRODUCTION Any person revisiting a location after some time cannot help noticing that the appearance of the site has probably changed. For example, activities such as urban growth, agriculture, and logging leave their marks on the landscape for periods of time ranging from a few weeks to several decades, and landscapes that might have been familiar now look different. Although these kinds of transformations are more noticeable in areas directly influenced by human activities, they also occur in isolated regions and wilderness areas. It is not a surprise to find evidence of human activities even in the most remote places on the planet. During the past thousand years, as population increased, human activities took an increasingly significant role in changing the global environment, including the transformation of landscapes. This process of landscape change caused by the management practices associated with the exploitation of the planetâ€™s ecosystems is known in the scientific community as Land-Use and Land-cover Change (LUCC) (Turner et al. 1995; Lambin et al. 1999). These landscape and management transformations are so pervasive that, when aggregated globally, they affect key aspects of the Earth system functioning (Lambin, Rounsevell, and Geist 2000; Lambin et al. 2001). For example, agriculture, forestry, and other land-management practices modify entire landscapes, altering plant and animal communities, and change key aspects of nutrient cycling (Ojima, Galvin, and Turner 1994). Although landscape changes have significant impacts on the Earth system, it is not clear how the outcomes of human activities are affecting ecosystem functioning. 1
2 Research in recent years has tried to understand the complex linkages between human activities and management and ecosystem functioning. For example, research projects such as the Global Change and Terrestrial Ecosystems (GCTE) and Land-Ocean Interactions in the Coastal Zone (LOICZ) try to understand the linkages between human actions and ecosystems (IGBP 2003; Lambin et al. 1999). Despite the fact that scientific understanding of this process has improved dramatically, there are still enormous gaps in the current understanding of these ties. Research on the land-use and land-cover change process is of critical importance to understand how human activities affect the Earth system functioning. Agriculture is one of the most significant agents of environmental change (Reenberg and Lund 1998). Globally, agricultural production will intensify to satisfy the needs of increasing human population (Turner et al. 1995; Ojima, Galvin, and Turner 1994; Conway 2001). This will definitely result in major landscape transformations in many areas of the world, mostly those experiencing rapid population growth (i.e., the developing world). There are areas of the world that already experienced significant landscape transformations associated with the intensification of agricultural production. The Colombian coffee lands are one of these regions. Starting in the early 1970s there was a major push towards intensive coffee production caused by a desire for a more efficient production system that also provided an easier way of controlling a major coffee disease (Coffee Leaf Rust) (Gabriel Cadena personal communication). This disease, which was already in Brazil and Costa Rica at the time, was spreading rapidly to other countries. Adoption of this new production system resulted in dramatic landscape changes. For instance, the amount of land planted in coffee declined; and the change in
3 production system decreased forest cover, increased soil erosion, and greater pollution due to the increased use of chemical inputs (Errzuriz 1986; Rice 1996)Jorge Botero, personal communication; Gabriel Cadena, personal communication). Despite the fact that this process is relatively well documented, few studies have looked at the environmental and spatial characteristics of the intensification of coffee production in this South American country (Ortiz 1989). The main objective of this dissertation is to describe the landscape transformations associated with coffee production intensification in the Colombian coffee lands. There is a large body of scientific literature dealing with agricultural intensification and commercialization. There are many models conceptualizing these processes, and probably the best known are Boserupâ€™s model of intensification (Boserup 1965) and Chayanovâ€™s model of the evolution of peasant farms (Netting 1993). In these models, the intensification and or expansion of agriculture is related to population density in areas with little market involvement. Because in our contemporary world it is nearly impossible to find places not subject to market pressure, these models are somewhat unrealistic. Other models address the role of market forces. For example, Von Braunâ€™s (von Braun 1995) ideas about the commercialization of agriculture rely on accessibility and rural infrastructure, market reform, and trade policies as key elements determining the successful market involvement of peasant farms. Unfortunately, these models rely on only a few factors and fail to capture the complexity of the process of agricultural change itself, where different social, economic, political and environmental factors interact and feedback to produce a particular outcome (Goldman 1993). Furthermore, these conceptualizations were developed with the farm as its unit of analysis and do not
4 consider explicitly the landscape transformations associated with agricultural intensification and commercialization. Three basic ideas are analyzed in this Colombian coffee lands case study related to the landscape transformations associated with agricultural change. Conflicting evidence shows that while in certain areas of the world like Finland (Hietala-Koivu 2002) and Germany (Schuller et al. 2000) commercialization of agriculture has led to landscape homogenization; in other areas of the world like Kenya (Conelly and Chaiken 2000), Central Honduras (Southworth and Tucker 2001), and the Andean highlands (Bebbington 1996b) it has led to more landscape heterogeneity. The main hypothesis of this dissertation is that in areas where mechanization of agriculture is difficult, like the sloping terrain of the Colombian coffee lands, intensification and commercialization lead to a more heterogeneous and diverse landscape, as the intensification process allows farmers to produce the same or more of a given product in less land, opening land for other crops and land-use practices. The second idea is that in a particular region undergoing agricultural intensification and commercialization not all land uses intensify at the same time. Evidence from Peru (Wiegers et al. 1999) and Portugal (Pinto-Correia 1999) suggests this is true. Because farmers have limited labor and capital resources, they allocate them in a way to balance household needs and opportunities, disintensifying production in the land uses that under the current socioeconomic, political, and environmental conditions are less productive, and intensifying those production spaces that are more attractive. The third idea related to agricultural intensification and landscape transformation is that the socioeconomic, political, and environmental factors more strongly associated
5 with intensification and commercialization vary from region to region. While in certain parts of Central America the appearance of intensive commercial plantations is associated with government policies promoting export crops rather than population growth or other variables (Bilsborrow and Carr 2001), in Central Honduras coffee production intensification in certain areas and abandonment in other zones is mostly related to accessibility to road infrastructure (Southworth and Tucker 2001). The key concept that these two examples show is that different variables, or combinations of variables, are responsible for explaining agricultural intensification in different contexts and historical moments. Land-Use and Land-Cover Change (LUCC) and Global Environmental Change As stated above, land-use and land-cover change have an enormous influence on global environmental change (Riebsame et al. 1994; Turner et al. 1995). Changes at the local level can influence regional and global processes. There are many examples that highlight the role of human activities in global change, how effects at one scale influence other scales, and how changes in one location may influence other regions. One of the best examples is the production of greenhouse gases by burning fossil fuels, and released by certain land-management practices among others. These gases are produced at the local level, but have serious implications at the global level as their concentrations increase in the atmosphere (Houghton 1994; Mannion 2002; McKenzie 1998). Human manipulation of the landscape also results in changes in topography, climate, and hydrology. Irrigating and draining lands for agriculture dramatically changes the availability of water of many landscapes (Houghton 1994; FAO 2001). Large areas of terraces and raised fields transformed many mountain slopes and swampy lowlands, making them fit for agriculture (Denevan 1963; Knapp, Mathewson, and Denevan 1985).
6 Changes in vegetation patterns associated with land-cover change also have a significant impact on local and regional climate, potentially changing temperature and rainfall regimes (Richter 2000). The influence of human activities on landscapes also includes the extinction and introduction of certain species. The new species disrupt local ecosystem processes and may colonize large areas because of lack of natural enemies in their new environment (Ojima, Galvin, and Turner 1994). In a similar way, the extinction of indigenous species leaves certain niches and ecosystem functions empty. Finally, human activities also lead to widespread pollution of land, water, and air as a result of acid rain, fertilizer runoff, and waste disposal, among others (Houghton 1994; Mannion 2002). The examples presented so far show how changes in land-use and land-cover change have a multitude of different effects on ecosystems, both at the local and global level. There are cause-effect linkages within ecosystems, and also across ecosystems, and the effects of land-use and land-cover change on the ecosystem may not show up immediately. Therefore, it is not only important to know what covers occur where, but also how and why they are changing. Unfortunately, these questions (what, where, how, why) are currently poorly understood, and it will require the joint efforts of social and natural scientists to come up with explanations about the process of land-use and land-cover change (Turner et al. 1995) Natural vegetation integrates the influences of organisms, soils, climate, and topography, and other factors such as disturbance (Campbell 1996; Odum 1997; Ortiz 1989). It also summarizes the effects of human activities on natural ecosystems. Vegetation (and more generally, land cover) represents not only what is currently on the
7 ground, but the result of historic land-use practices. Thus, land cover can be regarded as an indicator of the environmental conditions of any given region, and an indicator of the impact that human activities have had in that area. Human Activities and Landscape Change Although land-cover change is not only the result of human activities, but also the outcome of natural processes, the pace, magnitude, and spatial reach of human alterations of the Earthâ€™s surface are unprecedented (Lambin and Geist 2001). Humans tend to create patchy landscapes (Odum 1997) that are the result of different management practices in different plots. The spatial and temporal configuration of this land-cover mosaic is the result of historic land use, which for a long time in human history was characterized by different farming systems coexisting together, and more recently also reflects the impacts of other human activities (Etter and Villa 2000; Mannion 2002) The question of land-use and land-cover change has attracted a wide variety of scientists concerned with modeling the spatial and temporal patterns of land conversion, and trying to understand the causes and consequences of those changes (Irwin and Geoghegan 2001). The study of human processes driving ecosystem change is critical for landscape management and planning (Etter and van Wyngaarden 2000). However, there is a need for quantitative, spatially-explicit data on how land cover has changed due to human activities in order to predict and understand future changes. Scientists interested in landscape transformations need to change from being only passive recorders of landscape changes to also being active in shaping future, more sustainable landscapes (Hobbs 1997). Unfortunately, in many instances there is very little historic information on the spatial evolution of landscapes, making it more difficult to understand the process of the land-use and land-cover change.
8 Landscapes are expressions of the spatial and temporal interactions between people and environment (Hll and Nilsson 1999). Although landscapes have a physical expression, it is how this physical reality is perceived by the land managers that determines how landscapes are used (Blaikie 1995). Usually, social and economic forces dictate how land is used (Ojima, Galvin, and Turner 1994; Rice 1997). Landscapes are not just the physical expression of the human-environment interface, but a cultural construct of a specific moment in space and time. In other words, analysis of landscapes and their evolution is not simply studying the spatial and temporal characteristics of the physical changes, but also the changes associated with the societies that shaped them. Landscapes are the stage where the permanent transformations take place as a result of changing values, policy, and economic conditions (Hll and Nilsson 1999). Changes in land use, which result in changes in land cover, reflect the history of mankind. Therefore, it should not come as a surprise that rates of land-use and land-cover change often parallel the rates of population growth (Houghton 1994). In addition to population growth, landscape changes are linked to economic development, technology, and environmental change (Houghton 1994). For much of human history, land managersâ€™ land-use decisions were mainly affected by the demand for products at the farm and village levels. In the last few decades, this pattern has changed, and the global interdependence of trade and markets has become the main driving force influencing land-use decisions at the local level (Nestel 1995). The human activities influencing landscape change arise from either need (subsistence) or greed (wealth generation) (Mannion 2002). According to Mannion, the most important of these activities are agriculture, forestry, urbanization, and mining.
9 They tend to operate all over the planet and their effects are easy to recognize. Mannion also lists other human actions that lead to landscape changes and have more localized effects. Among these are invasive species, waste disposal (e.g., sanitary landfills), the effects of war and terrorism (e.g., destruction by bombs), tourism (e.g., tourist complexes such as Disneyworld), and sports (e.g., construction of golf courses) (Mannion 2002). Of the whole range of human actions that have an effect on land-use and land-cover change, agriculture and urbanization are the most influential (Brown 2001; Mannion 2002). The effect of agriculture in changing landscapes is evident. According to the FAO, in the year 2000 more than 49 million square kilometers, about 37% of the land surface of the planet, were agricultural land (FAO 2001). The agricultural growth needed to supply food for the growing worldâ€™s population may take one of two forms (Perrings 2001): Extensive growth: Land conversion to agriculture. It is accompanied by habitat destruction, fragmentation, and biodiversity loss. Generally associated with areas of low population density and high population growth like Sub-Saharan Africa and Latin America Intensive growth: More efficient management (i.e., more output per unit of land) that might result in the alteration of the mix of species. It affects crops, livestock, symbiotic organisms, competitors, and predators, and also results in biodiversity loss. It characterizes regions with high population density and rapid growth, particularly regions in Asia. Today, the conversion to cropland, and perhaps intensification of agricultural production, are more rapid in the developing world, while in the developed world the area in cropland has decreased (Turner et al. 1995). This roughly corresponds to the patterns of population change in the 20 th century, characterized by rapid growth in the tropics, and by slow declines in population in some areas of the developed world. Another critical aspect in explaining agricultural change is the industrialization and mechanization of
10 agriculture. For example, starting in the 1950s, the agriculture in Europe switched from more traditional systems relying on animal and human power to more industrialized agriculture based on heavy machinery (Di Pietro 2001; Schuller et al. 2000). Landscapes changed dramatically, as more homogeneous landscapes are more suitable for heavy machinery, and pollution increased because of increased reliance on chemical inputs. The view of agricultural growth as being caused by either extensification or intensification is not adequate for areas where there is not a clear distinction between cultivated and non-cultivated landscapes (Boserup 1965). In many areas of the world, complex agroforestry systems exist, and are the result of human activities on existing forests. Examples can be found in the Amazon estuary, where aa groves have been carved out of the forest through careful management (Smith 1999b), or in areas of the Peruvian Amazon, where a wide range of extractive activities leave their permanent imprint on secondary growth in fallow plots through management (Coomes and Barham 1997). Urbanization is the other major influence on landscape change. Although urban areas occupy less than 2% of the Earthâ€™s surface, they have major land-use and land-cover effects through rural-urban linkages (Lambin et al. 2001). An increasingly significant proportion of land-use and land-cover change are the result of urban demands rather than rural subsistence needs (Turner et al. 1995). Urban residents need not only to satisfy their basic nutritional requirements, but also bring in construction materials and other supplies needed for industrial activities (fuel supply, etc.). Cities do not restrict themselves to importing food, fiber, and other materials that cannot be produced locally. Urban areas are also a source of waste materials that are usually shipped to the
11 surrounding areas, altering landscapes through sanitary landfills, and pollution due to industrial and human waste. Agricultural Intensification and Commercialization and LUCC Agricultural intensification and commercialization are integral parts of the process of agricultural change. Because of the prominent role of agriculture in explaining land-use and land-cover change, it is expected that intensification and commercialization of agriculture will also contribute to the transformation of rural landscapes. Intensification can be defined as the substitution of land for inputs of capital, labor, and skills, so as to gain more production from a given area (Brookfield 2001; Lambin, Rounsevell, and Geist 2000). Commercialization of agriculture refers to the process of increasing the market involvement of agricultural production, gradually switching from subsistence agriculture to production for market consumption (von Braun 1995). In todayâ€™s globalizing world, it is impossible to find pure subsistence or pure commercial small farms, and most farmers are engaged in both kinds of production (Brush and Turner 1987; Dorsey 1999; Brown and Shrestha 2000). For this reason, agricultural intensification and commercialization usually go hand in hand. Agricultural intensification and commercialization are some of the most significant forms of land-cover modification, mostly through complex management systems that result in higher yields (Lambin, Rounsevell, and Geist 2000). Their role as driving forces of landscape transformation is expected to increase. Although intensification already has transformed many landscapes, it will become more important as food demands keep increasing, and the availability of agricultural lands reaches a limit (Turner et al. 1995; Gregory and Ingram 2000). Commercialization will also spread as the demand and prices of cash crops rises in the international market, and creates incentives to increase
12 commercial crop production in the third world (Nestel 1995). Despite the fact that agricultural intensification and commercialization are increasing all over the world, there is still significant land-cover conversion into agriculture, mostly in the developing world (Bilsborrow and Carr 2001) The choice of different forms of agricultural intensification results in different landscape effects. In Europe, agricultural intensification has resulted in a decline of traditional agriculture and landscape simplification (Di Pietro 2001). Other areas of the world, like the mountains of Nepal, have experienced little change in area under different land covers (Brown and Shrestha 2000). However, many of those land-use practices have intensified. In general, landscape effects of agricultural intensification and commercialization are diverse. Because different households choose different livelihood strategies based on their differential access to certain key resources, family history and experiences, and local environmental conditions, their choices for intensification and commercialization change accordingly (Marquette 1998; Coomes and Barham 1997; Yapa 1977). Coffee and Landscape Change Like any other form of agriculture, coffee production implies drastic land-cover changes. Because coffee originally developed as an understory bush in the rainforests of East Africa, rainforest soils are still the land preferred for coffee (Wrigley 1988; Willson 1999). According to Wrigley, the trend is to use chainsaws and fire when opening a new coffee plot for cultivation from virgin forests. Although the tree canopy might not necessarily be removed, undesirable large trees are cut down. Willson (1999) suggests that this trend of leaving the rainforest canopy intact is becoming less common, and purposefully planted trees replace the rainforest trees. Only in the most traditional
13 systems, the canopy of the original rainforest is maintained (Gobbi 2000), and in other less-traditional systems the shade is provided by purposefully planted, useful trees. Therefore, in most instances, the establishment of a coffee plot is accompanied by a landscape transformation in which the original vegetation is replaced by another type of cover. Coffee is currently planted in nearly 11 million hectares (FAO 2001). Therefore, the transformations just described affected this area at some time in history. The recent trend towards intensification of coffee production, in which shade is removed (or, in the best situation, reduced) represents another major transformation that can change coffee-producing landscapes in large areas of the world. In Latin America alone, nearly 2.7 million hectares would be potentially transformed by the changes in production system associated with intensification of coffee agriculture (Rice 1996). In Colombia, the move toward intensive coffee agriculture started in the late 1960s (Palacios 1980; Parsons 1968), and has continued thanks to the support of the Colombian National Coffee Growers Federation (FNC). Availability of relatively detailed land-cover history in the coffee-growing area provides a unique opportunity to study the impacts of land-cover modification (i.e., intensification) on agricultural landscapes. The purpose of this dissertation is to study the landscape changes associated with coffee production intensification in the Colombian coffee lands. This introductory chapter is followed by a literature review chapter (Chapter 2) where the most relevant literature on LUCC and agricultural commercialization and intensification is introduced. This chapter ends with the main hypotheses introduced earlier in this chapter. Chapter 3 presents a brief history of coffee as a crop, and Chapter 4 presents how it became established in Colombia. Chapter 5 reviews the methods and data sources used to test the
14 hypotheses introduced in Chapter 2. Chapter 6 presents the results of the analyses. Chapter 7 discusses the implications of these results, and Chapter 8 closes up this dissertation with the discussion and conclusions.
CHAPTER 2 LITERATURE REVIEW The previous chapter presented clear evidence of the relationship between human activities and landscape evolution. Agriculture and urbanization are two of the most dominant processes linking human actions and landscape change. In particular, agriculture is a key force leading to landscape transformation. Despite the evident connection between the process of agricultural change and the evolution of agricultural landscapes, there is relatively little research on how agriculture leads to land-use and land-cover change, and it is mostly concerned with the transformation of forested areas into agricultural plots. The purpose of this chapter is to review the current scientific literature between agricultural intensification, land-use and land-cover change, and to develop a conceptual framework that links these two, somewhat isolated bodies of literature, into a conceptual framework that will be used for this dissertation. Land-Use and Land-Cover Change as a Research Topic Land-use and land-cover change result from the complex interaction of human and biophysical processes. Therefore, understanding the human-environment interaction is at the heart of research on human-induced landscape evolution. As a result, there is a need for an integrated social-natural sciences framework in studying the impact and feedback of land-use and land-cover change on the global environment. With this integrated framework in mind, there are three questions arise regarding land-use change (Ojima, Galvin, and Turner 1994): How are land-use changes contributing to global environmental change? 15
16 What socioeconomic factors determine land use, and how will they change? How does land use modify processes that influence global change? These questions clearly highlight the importance of understanding not only the impacts of the process of land-use and land-cover change, but also the process itself. Although understanding how this process takes place is critical for the analysis of landscape changes, it is not sufficient for comprehending the full scope of the effects of land-use and land-cover change on the global ecosystem. Scientists and policy makers need a reasonable understanding of the spatial patterns associated with human use of different landscapes, both historically and in present times, to come up with a reasonable explanation of how the interaction of socioeconomic and biophysical driving forces results in specific land-use patterns (Etter and van Wyngaarden 2000). In other words, the patterns of land-cover change need to be documented in order to derive an explanation of what socioeconomic and biophysical driving forces lead to land-use and land-cover change. Thus, it is essential to document which land covers are changing and where the change is occurring in order to answer how and why those covers are changing. Although there is nearly an infinite variety of landscapes based on the myriad of possible interactions and feedbacks between socioeconomic and biophysical variables that induce land-cover change, it is possible to group them into three broad categories based on the degree of influence of human or biophysical drivers in shaping the landscape (Odum 1997): Fabricated landscapes: developed sites such as cities and transportation corridors, among others. Changes of these sites are mostly influenced by human-induced variables. Intensive use of energy for human activities. Domesticated landscapes: examples include agricultural lands and managed forests. Environments in which the sun provides the basic energy source that is increased and harnessed by human efforts. Changes to these sites are also strongly
17 influenced by human-induced forces, but less than in the fabricated landscapes. Also, biophysical changes have a more important effect than in fabricated landscapes. Energy use is intensive, but less intensive than in fabricated landscapes. Natural landscapes: Self-supporting/self-sustaining environments. Mostly controlled by biophysical processes, but influenced by human activities. Natural does not mean not touched by humans, but landscapes where the structure of the ecosystem is still controlled by natural processes. Energy is used for ecosystem function. Based on this distinction, it is possible (at least in a general sense) to identify the forces responsible for landscape change. In a fabricated landscape, biophysical drivers still have an impact in the spatial arrangement of land covers, but human-induced drivers are more important in explaining land-cover and land-use change. This distinction also indicates that the key variables controlling and shaping landscape evolution change with different settings. This idea is discussed later in with more detail. Definitions of Land Use and Land Cover Land use and land cover are concepts that, although intimately related, have different meanings. Land cover refers to the biophysical state of the Earthâ€™s surface (Turner et al. 1995), or the attributes of the land (Lambin, Rounsevell, and Geist 2000). Land use, on the other hand, refers to how humans use the land cover (Lambin, Rounsevell, and Geist 2000), or to the functional role of land cover in economic activities (Campbell 1996). In any given landscape, land cover refers to what is on the ground at any given time, whereas land use is how those covers are used and for what purpose. Land cover can be observed directly, whereas land use needs to be inferred from knowledge of the state of land cover (Campbell 1996). Land cover is the result of the interaction of human land use with the natural biophysical conditions, including the history of the site (Turner et al. 1995).
18 Land-Cover Change vs. Land-Use Change In the same way that land use and land cover are different notions, land-use change and land-cover change also refer to different processes. While land-use change refers to changes in the management of the land and its attributes, land-cover change refers to what covers are changing in any given landscape. Because the land use of any given land cover is based on the intended outcome or benefits derived from its exploitation, changes in management practices (i.e., changes in land use) also have land-cover transformations (i.e., land-cover change). Understanding these two interrelated processes is essential for LUCC research. However, to understand why and how land covers change due to land-use transformations, it is necessary to analyze first what, when, and where land covers change. Once researchers have a grasp of land-cover changes, it is possible to relate them to socioeconomic, environmental, political, and institutional changes that result in land-use change. This dissertation focuses on land-cover change, and it provides the basic information necessary to come up with land use explanations associated with the observed landscape transformations. There are two distinct processes of land-cover change (Turner et al. 1995; Lambin et al. 2001; Lambin, Rounsevell, and Geist 2000; Jansen and Di Gregorio 2002): 1) land-cover conversion, and 2) land-cover modification. From a human point of view, land-cover conversion refers to the process of purposefully transforming one land cover into another, leading to major changes in the socioeconomic and environmental processes going on in any given area without necessarily changing the land-use practices. The best example of land-cover conversion is deforestation (Allen and Barnes 1985; Bilsborrow and Carr 2001; Perrings 2001). The land cover is transformed into something else, usually agricultural land. This drastically changes the land use associated with that parcel
19 of land, but also alters the local hydrological cycle, nutrient cycling, and other ecological processes. There are also examples of land-cover conversion resulting from natural causes. A volcanic eruption, major flood, or earthquake may wipe out many of the preexisting land covers, and change some of the ecological processes that characterized the landscape before the transforming event took place. Slower processes like climate change also lead to land-cover conversion. Land-cover modification is a process in which some changes in land cover result as management practices change, but the land cover remains the same. The impacts of land-cover modification are more subtle than the impacts of land-cover conversion, as the changes in land cover are less noticeable. Perhaps this is one of the main reasons why until very recently most LUCC research was mostly focused on land-cover conversion (Lambin et al. 2001; Lambin, Rounsevell, and Geist 2000). Land-cover modification is not well studied, and at the global level, often ignored (Jansen and Di Gregorio 2002). However, now scientists recognize that land-cover modification is widespread and needs to be studied in detail. One of the best examples of land-cover modification is the selective extraction of certain forest species in many areas of the world (Ochoa-Gaona 2001). As certain valuable trees are extracted, the floristic composition of the forest changes, but the land cover still remains as forest. The management of the forest for the exploitation of specific species leads to a new type of forest, characterized by a different community, while remaining a forested landscape. Another useful example is the landscape transformations associated with a change in city zoning. Suppose a residential area changes to a commercial area. Slowly, commercial activities will creep into the residential neighborhood, changing its appearance but still maintaining the urbanized
20 landscape. Table 1 presents the most important characteristics of land-cover conversion and modification. Table 1. Land-cover conversion and modification Land-cover conversion Land-cover modification Purposefully transforming one land cover into another Major disruption of ecological processes Can be the result of natural processes Not necessarily associated with changes in land management (i.e., plantation forestry) Direct transformation of land cover Unintended changes in land cover associated with changes in land management Changes in ecological processes are less noticeable It is the result of human manipulation Associated with changes in land management Indirect transformation of land cover The LUCC Initiative Because of its importance as an agent of global environmental change, the study of Land-use and land-cover change has become a priority research agenda. In 1993, the International Geosphere Biosphere Program (IGBP) and the Human Dimensions Program (HDP) launched the Land-use and Land-Cover Change (LUCC) initiative in order to improve the scientific understanding on landscape evolution and its relation to global change (Turner et al. 1995). LUCC research is multidisciplinary and interdisciplinary in nature because it requires an integrated approach where biophysical and socioeconomic variables are combined in order to explain landscape transformations (Jansen and Di Gregorio 2002). Additionally, it includes a spatially explicit framework to analyze landscape changes. With this integrated and spatially explicit character, the social and natural sciences should come together to explain which land covers are changing, where they are changing, and how and why those changes occur. According to the LUCC Science/Research Plan (Turner et al. 1995), the natural sciences should focus their attention on the states and flows of the biosphere, including the effects of land-cover
21 transformations. The social sciences, on the other hand, should emphasize the impacts of land-use practices and how they change as environmental, political, and environmental factors evolve. The LUCC research initiative has three main research foci (Turner et al. 1995): Land-use dynamics (Focus 1): To understand the process of land-use change. It addresses questions like what are the effects of particular land-use practices on land cover over time, what are the main driving forces that determine a land-use strategy and how it changes over time, and how are land-use strategies developed. Land-cover dynamics (Focus 2): To understand land-cover change. In particular, it addresses questions such as which land covers are changing and where, what are the rates of land-cover change, and what are the spatial and environmental attributes that contribute the most to the explanation of land-cover change. Regional and global modeling (Focus 3): To analyze and predict land-use and land-cover changes. It emphasizes the simulation of driving forces of land-use and land-cover change and their feedbacks with the idea of making projections about the process of landscape change. Up to this date, most of the LUCC research is based on case studies usually at the local and regional scales, and it is possible to find articles dealing with landscape transformations associated with human use from all over the world. When looking at the evolution of this subject in the scientific literature, there is no doubt that land-use and land-cover change are becoming more and more important research subjects. For this purpose, I used the Web of Knowledge (ISI 2003) to track the evolution of the literature on land-use and land-cover change from 1993 to 2002. Four keywords were used separately to search for articles published in peer-reviewed journals (LUCC, landscape change, land-use change, land-cover change). The results show that the number of articles dealing with these subjects has increased from 35 in 1993 to 290 in 2002, with the majority of the articles listed under the keyword â€œland-use changeâ€ (15 in 1993 and 186 in 2002).
22 How Land-Use and Land-Cover Change Take Place Conceptually, the process of land-use and land-cover change is relatively simple. Because most contemporary land-use and land-cover change are the result of human activities, presently, land-cover change is largely driven by land use (Turner et al. 1995). Land-use practices are the product of different combinations of environmental, political, institutional, and socioeconomic factors. The most likely determinants of land use are as follows (Turner et al. 1995): Demographic factors (population density, size) Technology Level of affluence Political structures Economic factors (land tenure, access to markets) Attitudes and values The evolution of land use is a complex process in which environmental, social, economic, politic, and institutional forces interact at different scales, influencing land managersâ€™ decisions and transforming the landscape. These factors offer possibilities and constraints to the land manager, changing their land-management practices accordingly (land-use change), which result in land-cover changes. According to the LUCC Science/Research Plan already mentioned, human driving forces, mediated by a socioeconomic setting, and influenced by existing environmental conditions lead to an intended land use of an existing land cover through the manipulation of the biophysical attributes of the land. This process is shown in Figure 1. In this figure, a land manager learns from its surroundings, including the current land cover in a specific area. The land manager can be a person, a corporation, agency, or any other actor interesting in managing the land. This manager is also embedded in a larger environmental, socioeconomic, and politic and institutional context that determines
23 Land Cover LandUse EnvironmentalSocioeconomicPolitic andInstitutional Land Cover LandUse EnvironmentalSocioeconomicPolitic andInstitutional EnvironmentalSocioeconomicPolitic andInstitutional Figure 1. Land-use and land-cover change cycle the acceptable and desirable behavior. Based on this knowledge of the larger context and the land cover, the land manager chooses certain land-use practices that alter the land cover. As a result of these land-use practices, the environmental, socioeconomic, and political and institutional conditions may be altered (dotted line in Figure 1). Land managers confront these changes on a daily basis, and the process just described starts all over again when the land manager notices the changes in land cover and the larger context, and makes land-management decisions accordingly. Current land-cover change is intimately related to land-use change, but may also result from natural processes (McKenzie 1998). The Land-use and Land-cover Change Science/Research plan also describes the possible ways in which land-cover change may take place. Land-cover change may be the result of independent changes in biophysical drivers. For example, a forest fire triggered by lightning results in major landscape transformations. Land cover may also change as the result of direct alteration by humans. The clear cutting of a parcel of forest is a good example of land-cover change as a result of direct alteration. There are also ways in which land-cover change results from indirect
24 influences. Human activities may change a biophysical attribute of a landscape that may lead to landscape transformations. A good illustration of this kind of land-cover change is the landscape impacts of soil erosion on water bodies. Human practices may increase soil erosion in dramatic ways. As sediment is washed out by rains, and transported to the oceans by rivers and streams, it changes the characteristics of these waterways. Rivers that once may have been clear become murky, altering the conditions for organisms in that environment. The final way in which land-cover changes occur is also more indirect, and requires the interaction of forces that, when combined in a chain process, lead to landscape transformation. For example, changes in human activities may lead to changes in the biophysical context. These changes may feed back again on human activities, leading to direct alteration of the landscape. The Land-use and Land-cover Change Cycle also includes the influence of the context variables on each other. For example, a socioeconomic change by itself may not trigger a land manager to change a particular land-use practice. However, the same socioeconomic change may lead to a change in a political process that is perceived as important by the land manager, changing the land use accordingly. As stated above, at least from a conceptual point of view, the land-use and land-cover change cycle is a straightforward process. In practice, however, it is not as easy to identify what is the factor or factors leading to landscape transformations, as there is rarely one single factor influencing the change. It is more common that a set of factors, and some of their interactions and feedbacks, are the key driving forces(Geist and Lambin 2002). To make matters even more difficult, the factors controlling how rapidly
25 humans can modify their land use in response to changing conditions vary according to social, political, and economic characteristics (Ojima, Galvin, and Turner 1994). The characteristics of the households occupying a landscape have enormous influence on how that landscape is used, and therefore on land cover. For example, evidence from the highlands of Peru (Wiegers et al. 1999), and Western Amazonia (Coomes and Burt 1997; Marquette 1998) shows that, in a given region, households with different assets (i.e., ecological base, institutions, infrastructure, household characteristics) will have different land-use practices. Furthermore, small households may have access to more than one production space, and land-use patterns developed for each production zone need to be analyzed taking into account the livelihood strategies in all production spaces (Wiegers et al. 1999). This means that land-use and land-cover change in one production zone may be related to landscape changes in other production zones. Institutional forces seem to have a profound effect in determining landscape transformations. As institutions develop and enforce rules about natural resource management, they also influence the land managerâ€™s options to use the land. As a result, these rules translate into specific landscape configurations. For example, in Honduras (Southworth and Tucker 2001) local institutions have contributed to reforestation through the imposition and enforcement of a logging ban. Commercial interests promoted by government institutions also have major effects in the evolution of a landscape. In Mexico, national coffee policies implemented by the Mexican Coffee Institute (INMECAFE) between 1962-1989 lead to widespread landscape changes (Nestel 1995).
26 Institutional forces also seem to be one of the most important driving forces of tropical deforestation (Geist and Lambin 2002). Demographic forces such as population growth and migration also leave their imprint on the landscape. Although households have land-use practices dependent on each householdâ€™s assets, they also use land management in accordance to the stage of development of the household and the development of markets (Moran and Brondizio 1998; Walker 2003). In areas along the Trans-Amazon highway in Brazil, recently settled households tend to transform more forest into agricultural fields. As the household becomes established, markets arrive at the frontier, and the children are old enough to participate in agricultural production, the forest transformation diminishes dramatically and agricultural intensification and commercialization in the cleared plots takes place. The transmigration programs from the Indonesian government provide examples of landscapes that have totally changed since the colonization of sparsely populated areas began (Imbernon 1999). In Sumatra, the forested landscape of the 1930s has been replaced by an agricultural landscape in the 1990s. Forests have nearly disappeared from this island, and this has profound implications for ecosystem processes. This example also illustrates the complex interaction of driving forces. The transmigration program is essentially a government policy. Therefore, it can also be classified as a political or institutional driving force. However, its main goal is the relocation of people from densely populated areas into sparsely populated islands. As a result, it can be also classified as a demographic driving force. There are two important points that can be highlighted from these examples. In first place, in different socioeconomic, environmental, political, and institutional contexts, the
27 driving factors responsible for triggering landscape transformations may not be the same. In second place, as illustrated by the Indonesian transmigration example, factors interact and combine in complex ways, therefore making it difficult to identify the key driving forces. Key Concepts of Land-Use and Land-Cover Change Spatial patterns are essential Perhaps the most important concept underlying the LUCC research is the need to understand the spatial arrangement of land-cover and land-use change (Turner et al. 1995). Scientists working in natural landscapes have long recognized that patch size and shape are very important in determining the species that can survive (Odum 1997) and that landscape configuration plays an important role in explaining ecosystem processes (Turner, Gardner, and O'Neill 2001). In domesticated landscapes the spatial patterns of land cover do not only depend on changes of biophysical variables but also on the influences of human driving forces. Natural habitats within these domesticated landscapes become extremely important because most of the ecosystem services still come from the ecological processes taking place in the remaining patches of natural ecosystems. It is for this reason that LUCC inspired projects should try to use the principles of landscape ecology in an integrated analysis that includes the biophysical setting, the actual land use, and the effects of human-induced landscape transformation (Etter and Villa 2000). Land-cover change is both a natural and human-induced process Although usually overshadowed by the enormous influence of humans on different landscapes, land-cover change may be the result of natural processes. Unfortunately, it is not always easy to separate natural processes from human influences leading to landscape
28 transformations, as human driving forces may interact with natural drivers and amplify or attenuate the effects of biophysical drivers (Turner et al. 1995). What there is no doubt about is that in the present and recent past, land-cover changes are mostly the result of human actions (McKenzie 1998). There are some major differences between the dynamics of natural and human induced change. Natural change tends to be gradual and slower, punctuated by sudden, rapid changes (McKenzie 1998), while human induced changes are extremely fast, and their rate of change has no precedent in the geologic past (McKenzie 1998; Mannion 2002). While ecosystems are adapted to the natural rates of change, human induced transformations, like land-use change, modify ecosystem properties more rapidly than would naturally occur, putting enormous pressure on the ecosystemâ€™s ability to cope with change (Ojima, Galvin, and Turner 1994). Increased human population and land-use intensification have caused the subdivision of native habitats, the extinction of species, and lower species diversity within managed ecosystems (Flather and Bevers 2002). The fact that human induced landscape change results in more simplified landscapes is critical because the disappearance of the original ecosystem results in a decreased ability for the managed ecosystem to cope with disturbances. This is caused by the fact the managed ecosystem becomes more spatially uniform and less functionally diverse, reducing the resilience of the system (Holling and Gunderson 2002). Natural landscapes may be the result of intentional management Many landscapes throughout the world exhibit characteristics that were previously identified with mature or climax vegetation. More recently, however, it has come to the forefront that many of these, supposedly natural ecosystems, are the result of long-term
29 interaction between human and biophysical drivers. For example, Tropical rangelands are maintained in their current state by these interactions, and grazing is essential to their existence (Lambin et al. 2001). Some of the maple-hickory forests found in areas of Qubec province are the result of land use management and selective logging throughout the 19 th and 20 th century (Bouchard and Domon 1997). In certain areas of New England, the characteristics of present day forests are directly related to earlier land use practices in the area (Motzkin et al. 1996). According to these authors, pitch pine stands are only found in areas that were previously ploughed and used for agriculture. Some authors suggest that the presence of extensive grasslands in the high Andes (pramos) is the result of synergies between culture and nature (Sarmiento and Frolich 2002; Sarmiento 2000). The presence of islands of Andean forests above the current tree line suggests that these forest patches are remnants of a time where the tree line was located higher up in the mountains, and management for grazing of Andean camelids, and later sheep and cattle, eventually converted the formerly forested landscape into grasslands. The existence of similar environments in places where Andean camelids are not present would seem to undermine Sarmiento and Frolichâ€™s arguments, and there is a lot of debate about this explanation of the origins of Andean paramos. Land-cover dynamics One critical question of LUCC research is to understand the temporal and spatial relationships between land use and land cover (Jansen and Di Gregorio 2002). This involves looking at the spatial arrangement of land covers and their evolution through time as different driving factors interact, feedback and change, resulting in landscape transformations. A large measure of understanding land-use dynamics can be gained from the perspective of the land manager, who is the ultimate agent of direct change, and is
30 responding to events and processes that determine future use (Turner et al. 1995). According to the LUCC Science/Research plan, land-use and land-cover dynamics involve the relationships between the land user, the larger socioeconomic setting, the land in question, and the biophysical feedbacks of land-use strategies. Unfortunately, due to its complexity, there has been little advancement in the understanding of how the human land-use decision making process takes place, and there are still many gaps in the scientific understanding of the economic rationale leading to particular land-use strategies (Irwin and Geoghegan 2001). Although both patterns and processes of landscape evolution are important in explaining land-use and land-cover dynamics, most of the LUCC research has focused on patterns (i.e., outcome of land-use decisions) rather than process (i.e., human behavior leading to land-use decisions) due to gaps in the scientific understanding on the land use decision-making process. Market involvement and land tenure: critical factors Although land managers are exposed to a wide variety of factors when making land-use decisions, there are two factors that seem to be of critical importance in determining the selection of land-management strategies. These are market involvement and land tenure. Presently, in most places of the world, land-use and land-cover change are taking place under the influence of the market (Turner et al. 1995). For example, market forces determine the intensification of certain types of agriculture and the selection of certain crops in Nepal (Brown and Shrestha 2000), and the demand for particular products like coffee may influence land managers to focus on these cash crops (Palacios 1980). Land tenure has also a profound effect on how the land is used (Bilsborrow and Okoth Ogendo 1992). In many developing countries, like Colombia, landless peasants can claim a parcel of land in a colonization frontier if they show they
31 have used it for certain time and made some improvements such as building a house or a stable (Brunnschweiler 1972). In most instances, the first step towards showing use of the land is to cut down the rainforest. A similar, but opposite example, comes from Java (Peluso 1992). In this island, forest plantations belong to the State, but countless landless peasants are in charge of reforestation programs after forest plots are harvested. These peasants can use the land where the trees are growing to plant crops while the trees are small. Because these farmers do not own the land, and do not have anywhere else to go, they keep many of these reforestation parcels continuously under cultivation, destroying the tree saplings and replacing them by seedlings so enough space is left for subsistence crops. The net effect is that many reforestation plots never reach a mature state. The LUCC is the result of a few key processes shaping the landscape According to the Extended Keystone hypothesis in Ecology, there is a small set of plant, animal, and abiotic processes structuring natural ecosystems across scales in time and space (Holling 1992). A similar situation has been found in integrated human-environment ecosystems (Holling and Gunderson 2002). In these coupled systems, a few key processes (institutional, demographic, socioeconomic, political, and biophysical), together with disturbances, are responsible for the spatial patterns and evolution of a landscape. Many human-induced processes produce clusters of entities, generated by a small set of self organizing processes that define the â€œnaturalâ€ scales and frequencies that characterize managed landscapes (Gibson, Ostrom, and Ahn 2000). Empirical evidence from natural, disturbed, and managed ecosystems suggests that spatial attributes are not homogeneously distributed at all scales, and they tend to concentrate at certain scales (Holling and Gunderson 2002). This basically means that species and human use of any given ecosystem tends to cluster around these scales where attributes concentrate,
32 because it is at these scales where resources are more readily available. These are also the scales where changes are more noticeable, leading to specific landscape patterns. Any change in the processes shaping the ecosystem results in landscape transformations. Because landscape patterns change across space and time, time is a critical component of the analysis of landscape evolution (Harvey 1967). Landscape structure and composition may change dramatically over time (Baker 1989). The spatial and temporal patterns of the structuring variables become critical elements in understanding land-use and land-cover change dynamics. Unfortunately, in many instances the information available to study the evolution of a particular landscape may not match the temporal and spatial scales of the structuring processes, making more difficult the interpretation of landscape change patterns. Spatial diversity of patterns of landscape change Scientists engaged in LUCC research are overwhelmed by the high spatial heterogeneity of the patterns of land-use change at the local level (Lambin and Geist 2001). Because the local scale is also the level at which most land managers operate, it is the scale at which it is easier to examine the process of land-use and land-cover change. At the local scale, it is possible to see the direct effects of land management, and link them to the factors that lead to land-use change. However, the immense variety of patterns of land-use change at this level makes it difficult to generalize to a regional based conceptualization of landscape dynamics. Another fact that makes it difficult to generalize is that landscape changes are not evenly distributed either in time or space and change tends to concentrate in certain areas (Etter and van Wyngaarden 2000). What makes some areas more susceptible to landscape change than others is a question that has not been successfully answered, although recent developments in multi-level analysis
33 have been somewhat successful in linking regional and local-scale land-cover dynamics (Lambin and Geist 2001; Hoshino 2001). Reconstructing land-use and land-cover change is challenging At first glance, reconstructing land-cover change may seem a relatively easy task, requiring the search, compilation, and analysis of land-cover maps from different moments in time. The availability of aerial photographs and satellite imagery has made it possible to reconstruct land-cover changes very accurately both in time and space. However, in many areas of the world it is difficult to obtain this kind of remotely sensed information for reconstructing landscape transformations. Researchers interested in LUCC have to rely on other methods to analyze landscape evolution. In many instances information about land-cover change can be obtained from other records that may serve as a proxy for land-cover changes. Unfortunately, this method does not necessarily mean that the researcher will be able to compile accurate land-cover maps. However, it can provide a wealth of information about the land-use and land-cover change process. For example, Bouchard and Domon (1997) have been able to create a landscape evolution model for forested areas in Qubec using notary public records of timber sales. In general terms, when historical reconstructions of this type are used, it is only possible to obtain aggregate estimates of landscape transformations, and rarely the spatial configuration of those changes. Historical land-use information is somewhat more difficult to obtain. Although land-cover information provides insights on land use, it does not explicitly incorporate the driving forces of land-use change in a specific area. For this purpose, it is better to understand the land managerâ€™s (e.g., farmers, mining companies, et cetera.) decision-making process, and try to elucidate the factors that influenced land-use decisions and
34 their associated land-cover changes. This approach works fine for reconstructing land-use driving forces in the recent past. However, when trying to elucidate the LUCC driving mechanisms from the distant past (i.e., more than 10 years ago), land managers are not always able to point out the reasons why they decided to change their land-management practices. Therefore, reconstructing land-use history and land-use decisions requires not only talking directly to the land users, but also understanding the economic, political and social processes that were taking place at particular moments in time. Because historical sources are rarely spatially explicit, it becomes more difficult to assess land use change in a spatially explicit framework. Recent Developments in Understanding the LUCC Process As with any other research subject, the scientific understanding on land-use and land-cover change evolves, reflecting the increased knowledge about how this complex process takes place. A recent article summarizes some of the key aspects that have changed in our knowledge about LUCC (Lambin and Geist 2001). In order to see how much scientific knowledge regarding LUCC has changed, these recent developments were compared to a 1994 article containing similar information (Houghton 1994): Land-cover modification is more prevalent than land-cover conversion. Earlier views about LUCC (Houghton 1994) suggested that the majority of land-use and land-cover change was associated with the transformation of natural vegetation into agricultural fields. Because the landscape changes associated with land-cover modification tend to be more subtle, it is more difficult to identify them. It is now acknowledged that land-use intensification and diversification, both forms of land-cover modification, are the most common responses to pressures and opportunities experienced by land managers. As stated before, most LUCC research has concentrated on land-cover conversion. Therefore, this recognition of land-cover modification as a predominant process changes the focus of LUCC research, and makes the study of the landscape transformations associated with agricultural intensification, a type of land-cover modification, of critical importance. Land-cover transformations affect all cover types, not only forests. However, because one of the most dramatic landscape changes is the outcome of
35 deforestation, LUCC research was concentrated on forest conversion to agriculture in the tropics (Houghton 1994; Kummer and Turner 1994; Ojima, Galvin, and Turner 1994). Although this is a very important process, there are many other land-cover transformations associated with other land-management practices. In recent years, there is more awareness and interest in the landscape transformations associated with urbanization, agricultural change, and other activities (Mannion 2002) and LUCC research has started to study the changes caused by these processes. Changes in land cover are not permanent. They are part of complex and reversible trajectories of change. For example, areas of Central America were radically transformed during pre-Columbian times as the Maya kingdoms developed. With the Mayan collapse, the domesticated landscape associated with some areas of the lowlands and highlands of Guatemala reverted to forest (Whitmore and Turner 1992). In earlier LUCC research, changes in most landscapes were thought as permanent (Turner et al. 1995). Although there are examples of irreversible changes in short time scales such as soil degradation, or the acidification of lakes caused by acid precipitation, most forms of landscape evolution are reversible if the key structuring processes change to a previous state. The implications of this point for future research are important because it implies the possibility of ecosystem â€œnaturalizationâ€, an idea used in river systems, where sections of a channelized river are transformed to a more natural state that is compatible with current human land use and improves the supply of ecosystem services in any specific area (Rhoads and Thorn 1996). Extending the concept of naturalization to landscape management has great potential, as it creates opportunities for LUCC research to influence effective land planning, moving from a passive role of recording and explaining landscape changes to a more proactive role in designing and managing landscapes to achieve goals that are compatible with its intended land use and still provide enough ecosystem services for their inhabitants (Hobbs 1997) Changes are spatially heterogeneous, and are concentrated in â€œhot-spotsâ€. Some landscapes have been relatively unchanged for long periods of time, whereas others keep changing dramatically. As stated above, what makes some landscapes to stay relatively unchanged and what makes other landscapes to change rapidly is not well understood. What appears to be the case is that these rapidly changing landscapes are responding to transformations in one or more of the key structuring processes, resulting in a change in specific locations(Holling 1992; Holling and Gunderson 2002; Gibson, Ostrom, and Ahn 2000). Therefore, change exhibits a very spatially heterogeneous pattern. This contrasts sharply with the earlier understanding of the LUCC process, where landscape change was thought to be spatially homogeneous (Lambin and Geist 2001)
36 The LUCC as a Multi-scale Process Land-use and land-cover change, as other global environmental change processes, form a complex and interactive system linking human action to use/cover, to environmental feedbacks, to their impacts, and to human responses. These linkages occur at different spatial and temporal scales. Many human activities have causes and consequences at multiple scales (Gibson, Ostrom, and Ahn 2000). Similarly, ecological processes also have effects at multiple scales (Allen and Hoekstra 1992). As a result, the study of land-use and land-cover change may benefit from a multilevel analysis. The issue of scale, including cross-scale interactions, in land-use and land-cover change studies is extremely important, as not only human activities have causes and consequences operating at different spatial and temporal scales, but each scale has its own units and variables (Turner et al. 1995). This means that variables that may be the driving force of landscape change at one scale may not be a driving force at another scale. According to ecological hierarchy theory, scale is determined by what is being studied (Allen and Hoekstra 1992). As a result, the research question defines the extent and grain of the research protocol. Allen and Hoekstra (1992) also state that the scale of study can be defined as that level showing strong interactions among the variables operating at that level and weaker interactions with higher levels of the hierarchy, creating a boundary between the two levels. Lower levels of the hierarchy tend to have finer grain and smaller extents than higher levels. This means that processes at lower levels (i.e., finer scales) evolve at a faster pace than processes at higher levels. Therefore, a higher level can be regarded as the context in which phenomena of interest change. This higher level of organization (coarser scale) changes more slowly and imposes constraints
37 on lower levels. The choice of scale has important implications for identifying driving forces. For example, a higher-level variable might be considered constant or a changing force depending on the temporal scale selected for a particular study. Despite the fact that the context may determine the key-forces driving land-use and land-cover change, and that these variables change with location and scale, it seems that variations in explanatory variables of land-use change with scale exhibit a consistent pattern (Veldkamp and Lambin 2001). At the farm level, social and accessibility variables seem to control land use. At the landscape level, agroclimatic potential and topographic variables are the most important variables in explaining land-use change. At coarser levels of detail, land-use change is mostly related to climatic, demographic, and macroeconomic variables. Other authors (Hoshino 2001) adopt a hierarchical approach and incorporate variables at different scales in order to reach a good understanding of the landscape evolution process. This author combines variables about the natural conditions (i.e., physiography, soils) with farm level (i.e., size, type of production), local level (i.e., local government policies, cooperatives), national level (i.e., supply and demand of agricultural products), and international level (i.e., food imports and exports, trade) variables for explaining landscape evolution in Japan. Many of the drivers of land-use and land-cover change exhibit scale dependency. Scale dependency simply means that driving forces, or their consequences at one scale, may influence driving forces at other scales. Thus, an analysis of land-use and land-cover change should be carried out at different temporal, spatial, and hierarchical scales (Turner et al. 1995; Veldkamp and Lambin 2001). However, dealing with integrated human-environment systems makes it difficult to undertake this kind of approach, as social and
38 natural systems operate at different spatial and temporal scales, and the relationship between scales is unknown (Turner et al. 1995). This poor understanding of cross-scale dynamics is reflected in current models of landscape change. Most of these models have a very limited ability to capture multi-scale and cross-scale processes (Baker 1989). Although it is clear that the land-use and land-cover change process is a multi-scale process where cross scale dynamics are essential, it is very difficult to identify the right spatial and temporal scales to carry out a given study (Ojima, Galvin, and Turner 1994). In an ideal situation, where perfect knowledge about a specific landscape evolution process is available, it would be easier to identify the critical shaping processes. In reality, data availability dictates the selection of scale and the choice of critical variables. The problem associated with the actual selection of these two parameters is that they may not correspond with the scales at which critical processes of land-use and land-cover change are taking place (Anselin 2001) Complexity of the Land-Use and Land-Cover Change Process In the previous sections I have highlighted how the land-use and land-cover change process varies in different contexts, and how driving forces interact and change with location and scale. Therefore, it is not a surprise to the reader that the process of land-use of land-cover change is inherently complex. Simple answers explaining land-cover change based on population growth or infrastructure development are inadequate to capture both the generalities and nuances of the LUCC process (Lambin and Geist 2001; Geist and Lambin 2002). It is for this inherent complexity that the study of landscape evolution calls for a multidisciplinary analysis (Veldkamp and Lambin 2001). A multidisciplinary approach provides an adequate framework for studying how cultural
39 relations and natural processes interact with each other in the functioning and shaping of ecosystems (Hll and Nilsson 1999). Despite the acknowledgement that the study of landscape evolution requires a multidisciplinary approach that takes into account the complexity of the land-use and land-cover change process, there is little understanding about how the complex relationships between factors emerge and determine land use at the local level (Rao and Pant 2001). In other words, scientists understand that land-use and land-cover change is the result of the interaction of socioeconomic, demographic, institutional, cultural, and biophysical factors, and also recognize the importance of the interactions and feedbacks between driving forces (Reenberg and Lund 1998). However, there is very little understanding on how those interactions take place. Fortunately, this trend seems to be changing. A recent study about driving forces of tropical deforestation highlights how it is possible to analyze these feedbacks and interactions (Geist and Lambin 2002). In these study, the authors use a meta-analysis of 152 transnational deforestation case studies to classify the causes of deforestation into proximate (infrastructure extension, agricultural expansion, wood extraction) and underlying (demographic, economic, technological, policy and institutions, cultural) driving forces of deforestation. Using statistical techniques to analyze these data, Geist and Lambin (2002) reach the conclusion that interactions among driving forces and feedbacks are more important in explaining the process of deforestation than individual causes. Geographers have long recognized the complexity of spatial patterns. The development of spatial patterns is not a haphazard process and principles of spatial evolution can be developed (Harvey 1967). Therefore, it can be argued that it might be
40 possible to unveil the principles of spatial evolution associated with land-use and land-cover change. More recently, geographers and other scientists recognize that no single principle has the key to explanation, and even relying on a few principles does not offer adequate explanation of the complexity of change in the real world (Lambin et al. 2001; Houghton 1994; Bilsborrow and Okoth Ogendo 1992). Due to the extreme variability of patterns of landscape transformation in time and space, and the influence of different socioeconomic and biophysical contexts, it has been difficult to come up with general principles of landscape evolution (Turner et al. 1995). Some scientists have suggested developing an extensive library of local case studies across different historical and geographical contexts, so that it becomes possible to link local and regional LUCC processes, and understand the influence of different factors, or combination of driving forces in different contexts (Lambin and Geist 2001). However, there are some scientists that state that it is impossible to predict the patterns of land-cover change simply because of the complexity of interactions among driving forces (Mannion 2002). It remains to be seen which of these two positions will prevail as our knowledge about landscape evolution increases. An additional dimension that increases the complexity of landscape evolution processes is the fact that current land use and land cover are not simply the result of the equilibrium of driving forces. Land cover at any given location is the result of the myriad of land use decisions made by land managers and the history of the site. The present state of any landscape may be dependent on the initial conditions, and small random events that may lead to different outcomes (Turner et al. 1995). For example, in Chiapas each forest fragment has its own particular history (Ochoa-Gaona 2001). Forest fragmentation
41 does not occur as a continuous process in space and time, and for each forest patch it is defined by the complex interactions between the current state of the land cover, land tenure, environmental conditions, and economic alternatives, which determine where and how much forest will be transformed in each forest patch. Perhaps one of the major difficulties associated with the inherent complexity of the land-use and land-cover change process is the difficulty of linking patterns of landscape evolution with the processes of change. This is caused by the fact that, at local scales, it is possible to identify the direct actors of land-use change, and determine process-based explanations of these changes. As resolution decreases and the extent increases, it becomes more difficult to identify key processes and actors (Veldkamp and Lambin 2001). Some authors have suggested that there are ways to overcome these difficulties to some extent. It appears that the pathways of LUCC are largely the result of cause-connection patterns operating at national or regional scales (Lambin et al. 2001). According to Lambin et al. (2001) rapid land-use changes are often the result of the incorporation of certain areas into the world economy. They also suggest that driving forces under different conditions produce different land-use change pathways, and that globalization may attenuate or amplify these driving forces. Therefore, the key to establishing a link between patterns of landscape transformation and the process of spatial evolution lies on being able to develop the cause-connection patterns among the different driving forces. It was already stated that land-use and land-cover change research requires a spatially explicit framework. So far there has been a lot of effort to develop spatially-explicit datasets to enable scientists to study the landscape evolution process. However,
42 most of these variables are biophysical in nature. Scientists need to put more emphasis on the social context (including driving forces) influencing human impacts on the global environment (Ojima, Galvin, and Turner 1994). However, there are many driving forces of this kind influencing land use, such as attitudes and values, which are difficult to map in a landscape (Riebsame et al. 1994). These socio-cultural elements have been neglected because they are difficult to make spatially explicit, are hard to quantify, therefore making them difficult to include in mathematical simulations, which, as the next section will show, are one of the integral components of the LUCC inspired research. Modeling and LUCC To this date, much of the LUCC inspired research has been based in modeling (Lambin, Rounsevell, and Geist 2000). Models are simplified representations of reality, focusing on the significant relationships for the phenomenon being studied, and only applicable in a limited range of conditions (Haggett and Chorley 1967). One of the main goals of the LUCC initiative is to model regional and global landscape transformations in order to try to predict land-use and land-cover change, and the expected impact of different driving forces on landscape evolution (Turner et al. 1995). Global environmental change research has started to consider historical reconstructions of land use and land cover and to analyze the social forces behind these transformations as important elements that can be incorporated into the existing models of global change (Riebsame et al. 1994). As a result, modeling of landscape transformations can improve the scientific understanding of the land-use and land-cover change process. Land-use models serve as experiments to test the scientific knowledge about key processes, and to improve the understanding of critical variables and their relationships (Veldkamp and
43 Lambin 2001; Liverman 1994). Models of land-use and land-cover change should address at least one of the following questions (Lambin, Rounsevell, and Geist 2000): Which environmental and cultural variables contribute most to an explanation of land-cover change? Why? Which locations are affected by land-cover change? Why? At what rate do land-cover changes progress? When? There are many different approaches to model land-use and land-cover change. Scientists have used a wide variety of techniques including linear programming, statistical modeling (i.e., Montecarlo simulations, Regression analysis), cellular automata, and von Thnnen models among others (Lambin, Rounsevell, and Geist 2000; Liverman 1994; Yapa 1977; Harvey 1967; Turner et al. 1995; Turner 1997). Each model is developed with a particular objective in mind, and to date there is no single model is capable of providing comprehensive, yet geographically detailed assessments of land-use and land-cover change (Turner et al. 1995). This also means that no model can incorporate all the aspects of the complex interactions leading to landscape transformations. A detailed description of all the different modeling approaches utilized in studying landscape evolution is beyond the scope of this dissertation. For interested readers, Briassoulis (2000) provides a thorough review of the different modeling approaches used in LUCC research. Additionally, Agarwal et al. (2002) offer an assessment and review different modeling approaches of land-use change and Dâ€™aquino et al. (2001) give an excellent summary of Agent-based models of land-use and land-cover change. Models of land-use change patterns and rates are still on a case study basis (Turner et al. 1995; Veldkamp and Lambin 2001). Most LUCC studies show that by integrating
44 information about the physical attributes of the landscape changes over time with demographic, legal, and policy transformations, it is possible to formulate cause and effect patterns (Rao and Pant 2001). However, these explanations vary with scale, and the historical and geographical context, making it difficult to generalize the process of landscape change. Some landscape evolution models are interested in the rates of change, while others focus on where those changes occur. Presently, it seems to be easier to determine where changes will occur than the rates of landscape change (Veldkamp and Lambin 2001). Although there has been significant development in modeling the land-use and land-cover change process, there are several limitations associated with them. These limitations are more related to our understanding of the process of landscape change than with lack of adequate methods and technology. It is essential to move away from some of the simplified models of landscape evolution like the idea that tropical deforestation is the result of the migration of small-holder farmers engaged in shifting cultivation. Some authors (Geist and Lambin 2002) have demonstrated that, although an important force in the tropical deforestation process, small-holder farmers are just one of the agents of change, and it is the complex interaction of drivers like the availability of credits for cattle ranchers and the demand for beef in southern Brazil that leads to deforestation for agriculture. There are other landscape transformation myths, like the idea that tropical rangelands are a climax community, or that urbanization has little impact on land-use and land-cover change, that are being slowly debunked as scientific understanding about land-use and land-cover change increases (Lambin et al. 2001). Another issue with models of landscape change is that in many instances the model does not incorporate the
45 land managerâ€™s decision making process explicitly, and models are developed in a way in which they match the land-cover changes, but do not explain the process of change (Irwin and Geoghegan 2001). Therefore, they may fit well a particular situation, but when they are applied to other, similar conditions, the model does not fit the spatial patterns of landscape transformation. Additionally, the modeling units do not necessarily correspond to the land managerâ€™s land-use decision units. For example, in models that rely on cells as their unit of analysis, the minimum modeling unit corresponds to a grid cell. This spatial grain rarely coincides with the land-use decision unit. For example, a farm may be comprised of several cells. Land-use and land-cover change modeling approaches could greatly benefit from some of the methods used in the Political Ecology literature to try to fill the gap in the understanding of how the process of landscape transformation takes place. In its most basic form, political ecology uses an integrated explanatory device that links levels, scales, biophysical and socioeconomic variables called a chain of explanation (Blaikie 1994; Blaikie 1995). The starting point is the local level, where the researcher identifies the key forces influencing landscape evolution. These forces are usually located at different levels. Once these forces are identified, the scientist has to unveil how these driving factors interact with each other. Once the linkages within and across different levels are established, the researcher has a conceptual model explaining the process of landscape transformation. From a modeling perspective, this means some advancement. However, the major problem of how to operationalize these conceptual processes into a spatially explicit model still remains.
46 Agricultural Intensification and Commercialization Farmers are embedded in a context of permanently changing environmental and socioeconomic conditions (Brookfield 2001). Land managers face these changes on a daily basis, and as a result, change is a normal condition in any agricultural activity. New agricultural practices, technology, methods, and crop varieties are developed and adopted to respond to these changes, and old practices and cultivars are adapted to the new conditions. Therefore, agriculture is a dynamic field characterized by change, and agricultural landscapes should reflect agricultureâ€™s dynamic nature. As a result, understanding how agricultural change takes place is critical for understanding agriculture itself, as well as landscape transformation. Agricultural change, and its associated land-cover changes, is the result of two major kinds of transformations (Brush and Turner 1987): 1) Technological: agricultural change is the result of changes in technologies and management, and 2) Structural: changes in social and economic conditions trigger intensification. Agricultural intensification and commercialization are just components of the larger process of agricultural development and evolution (Goldman 1993). However, they are extremely important because they reflect the interaction of the farmer and the larger context in which he/she is embedded. Unlike the agricultural practices, technologies, and crop varieties, that are associated with the management at the plot level, intensification and commercialization of agriculture reflect the land-use decisions that encourage a land manager to adopt new and adapt old agricultural practices under changing environmental, socioeconomic, and institutional context. Agricultural intensification means increasing the utilization or productivity of the land currently under production (Netting 1993). Commercialization of agriculture is
47 defined as the process of increasing the market involvement of agricultural production, gradually shifting from subsistence agriculture to production for market consumption (von Braun 1995). Presently, both of these processes go hand in hand, and it is difficult to find farmers engaged in purely subsistence agriculture. Most farms today produce for the market at some level (Turner and Brush 1987), and the overwhelming majority of smallholder farmers are neither pure subsistence nor pure market farmers (Dorsey 1999). Therefore, it is essential to incorporate the influence of markets when studying the relationship between agriculture and landscape evolution. To exemplify this duality of farmersâ€™ goals, one may look at the conceptual model of agricultural evolution in recently colonized areas of the Amazon basin (Walker 2003). According to this model, in recently established homesteads, subsistence agriculture is prevalent with food security as an immediate concern. As markets arrive to the area, subsistence agriculture gradually gives way to commercial agriculture, and the idea of food security is complemented by the goal of maximizing utility. This view is also supported by other authors (Turner and Shajaat Ali 1996; Pingali and Rosegrant 1995; Hyden, Kates, and Turner 1993). From a theoretical perspective, there are two major explanations of agricultural growth based on changing subsistence needs or market demands (Dorsey 1999): 1) Subsistence needs theory emerges from the works of Boserup and Chayanov, and 2) Market demand theory, mostly the induced intensification theory that states that intensification is induced by the need to produce more for the market and for consumption. These theoretical models will be presented later on in some detail. For much of human history, agricultural growth has been accomplished mostly through expanding the area under cultivation in a process that is called agricultural
48 extensification. However, as human population has increased, much of the land available for agriculture has been already used. As a result, agricultural intensification will probably grow in the coming decades as a growing population demands more food and fiber to support itself (Ojima, Galvin, and Turner 1994; Ruttan 1994). Agricultural commercialization will also expand, as rural areas in low income countries are slowly incorporated into the market, and urbanization and a diversifying rural economy spread (von Braun 1995). Measures of Agricultural Intensity There are many different ways of measuring agricultural intensity. In most instances, it involves an increase of the frequency of cultivation and an improvement of the agricultural practices (Turner and Doolittle 1978). Agricultural intensification usually results in higher yields. Although yield alone does not provide an adequate measure of agricultural intensification, empirical evidence suggests that it correlates well with agricultural intensity (Turner and Shajaat Ali 1996). Therefore, yield can be used as a proxy variable for measuring agricultural intensification. However, one has to be careful because yield increase might be associated with new crop varieties that produce more per unit of area without an associated change in the management strategies or changes in cropping frequency, which reflect better the processes of intensification and commercialization of agriculture (Brookfield 2001; Brush and Turner 1987; Pingali and Rosegrant 1995; von Braun 1995). Since the beginning of the green revolution yields have been growing all throughout the world. This suggests that agricultural intensification is becoming increasingly important. However, in recent years yield growth is slowing down because the best lands have already been used, and the lower yields associated with more
49 marginal lands lower the average yields associated with intensive agriculture. Additionally, the detrimental environmental effects associated with some of the green revolution technologies such as salinization in irrigated areas and soil compaction due to heavy machinery can result in lower yields (Conway 2001). Agricultural Change Requires an Integrated Perspective Analyzing agriculture and agricultural change from a holistic perspective is useful because it integrates the socioeconomic, political, environmental, and technological elements of the production unit (Turner and Brush 1987) Because an agricultural system is the result of material variables (i.e., tools, cultivars, labor, soils, water, nutrients), structural variables (i.e., institutions, markets), and individual and community behavior (i.e., goals, allocation choices, community norms), differences in agricultural systems denote fundamental differences between the biophysical, social, and economic context of the communities involved (Brush and Turner 1987). Furthermore, institutions beyond the community (i.e., market, government) control key production resources and affect the nature of the agricultural system, and therefore land use (Brush and Turner 1987). The previous statements clearly show how the process of agricultural development is a complex, multi-scale process, the study of which benefits from an interdisciplinary framework just as the land-use and land-cover change process described earlier. Conditions for Successful Intensification/Commercialization The processes of intensification and commercialization of agriculture require the right context to be successful. In first place, because intensification implies more work and or energy investment, farmers will not intensify if it is not strictly necessary (Netting 1993). The reason for this is that as agricultural activities intensify, the marginal returns to labor diminish (Boserup 1965). In other words, the less intensive an agricultural
50 system, the more production per unit of labor a farmer obtains. Therefore, if a farmer can obtain enough food and commercial crops with a less intensive system, he/she will maintain the less intensive system as long as productivity is sufficient. Local institutions are very important factors in understanding the patterns of intensification, commercialization, disintensification and degradation in any area (Bebbington 1997). According to Bebbington, these institutions might strengthen the grassroots capacity to negotiate with actors and institutions that regulate technology, information, and knowledge among others. By doing this, these grassroots organizations can influence the processes through which rights and entitlements are defined, affecting the local intensification process and the social distribution of its benefits. The intensification of rural livelihoods, including agricultural intensification and commercialization, also requires secure and expanded rights over natural resources, improved access to markets, and investment in rural areas (Bebbington 1996b). In the South American Andes successful intensification is present in areas where technology-generating institutions exist, the ecology of production allows the intensification with high-value crops, and a market that is open to new actors (Bebbington 1997). Because presently it is difficult to separate the processes of intensification and commercialization, these conditions are similar to the key policy drivers for successful commercialization of agriculture listed by von Braun (1995). According to this author, there are three key forces for successful commercialization of agriculture: 1) macro economic and trade policies, 2) market reform, and 3) rural infrastructure. The potential for intensification and commercialization is also mediated by cultural values. It appears that different peoples and households perceive different opportunities
51 for intensification based on access to different amounts of resources, wealth, and gender, among other differences (Adams and Mortimore 1997). Local heterogeneity determines who intensifies and who does not. In other words, intensification and commercialization are not homogeneous processes, and they are differentially adopted by different groups in any given community. Additionally, households usually have access to different production zones that may be intensified or disintensified as they respond to changes in prices, technology and population (Wiegers et al. 1999). Pathways of Intensification and Commercialization Although there are many ways in which households can adopt agricultural intensification and commercialization, there are three paths that capture much of the intensification underway in agriculture (Lambin et al. 2001): Land scarcity: it triggers intensification in areas not fully integrated to the market. It is usually linked to high population growth. The most common response to land scarcity is the adaptation of an existing agricultural system to increase yield. Commodification: the presence of markets and demand for certain products triggers the commercial intensification of agriculture. Intervention: intensification and commercialization are caused by external interventions (i.e., development projects). These factors are usually linked with each other, creating many different mechanisms through which intensification and commercialization take place. Figure 2 shows these interactions. Agriculture intensifies based on the influence of these factors, which in turn can interact with each other to produce specific intensification and commercialization outcomes.
52 Less-intensiveAgriculture Agriculturalchange IntensiveAgriculture Land Scarcity ExternalIntervention Commodification Less-intensiveAgriculture Agriculturalchange IntensiveAgriculture Land Scarcity ExternalIntervention Commodification Land Scarcity ExternalIntervention Land Scarcity ExternalIntervention Commodification Figure 2. Agricultural intensification and commercialization. Full black arrows depict the agricultural change path. The red arrows represent the factors determining agricultural change, and the broken arrows how these factors may influence each other. Intensification, Commercialization, Agricultural Diversity, and Landscape Change The idea that human use simplifies landscapes was presented before. These simplified landscapes tend to more homogeneous and have lower levels of biodiversity. In agricultural landscapes, agricultural diversity plays a critical role by minimizing the risks associated with market fluctuations and environmental impacts. Agricultural diversification is not new to smallholder farmers. The main strategy for combining high production per unit of area with risk reduction and sustainability is diversification (Netting 1993). When commercial products are grown, diversification is also regarded as an important mechanism to avoid over-reliance on a limited number of agricultural commodities (Dorsey 1999). Therefore, it could be argued that the intensification and commercialization of agriculture could be associated with increased agricultural diversity (agrodiversity).
53 Some of the conceptual models of intensification and commercialization seem to suggest that as agricultural change takes place, agrodiversity diminishes dramatically (Conelly and Chaiken 2000). Some empirical evidence supports this position. In intensive, industrialized agriculture, large areas are planted with the same crop, and even hedgerows and verges, that can be important refugia for biodiversity, are eliminated as they act as obstacles for the heavy machinery necessary to maintain this agricultural system (Pauwels and Gulinck 2000). In other areas of the world like Africa it seems that agricultural intensification is associated with higher levels of diversity, and farmers develop complex systems of intensive agriculture that attempt to maintain soil fertility and maximize production in small areas of land (Conelly and Chaiken 2000). Some authors (Pingali and Rosegrant 1995) argue that at the beginning of the diversification process the farmer adds additional crops to a farm. The same authors suggest that as the commercial orientation of the farm increases, mixed, more diverse farming gives way to specialized, less diverse agriculture. This, however, seems to be against some empirical evidence that shows that farmers consciously maintain certain levels of crop diversity as the commercial intensification of the farm takes place (Zimmerer 1996a; Netting 1993). One of the ways in which the apparent loss of biodiversity associated with more intensive agriculture can be overcome is by adopting complex agroforestry systems. Agroforestry is a good option to improve not only resource management and boost rural income, but also helps in preserving the environment (Smith 1999a). In particular, agroforestry plots in the Amazon basin provide subsistence and cash income while preserving soil, water, and forest resources (Coomes and Burt 1997). Although
54 agroforestry plots never replace the ecosystem services provided by mature forests, they are much better at providing ecosystem services than other agricultural systems. Agroforestry systems are also suitable for areas requiring permanent crops, like the steep slopes of the Colombian coffee lands. Traditional coffee plots are grown under forest canopy. This agroforestry system provides a more structurally diverse environment for different forest dwelling creatures, minimizes soil erosion, and reduces the dependence on chemical fertilizers and pesticides among other benefits when compared to more intensive coffee plots grown without shade (Gobbi 2000; Perfecto et al. 1996). It seems agroforestry provides a more sustainable way of intensifying agriculture, while at the same time preserving the resilience of the system to cope with changes and improving rural livelihoods. There are two characteristics that appear to determine to a great extent the level of landscape and species diversity in an agricultural system. These are the possibility of mechanization and the size of the land holding, and they usually go hand in hand. It seems that in areas characterized by small land holdings, where mechanization of agriculture is more difficult as no tractors or other heavy machinery can be efficiently used, agricultural intensification and commercialization lead to higher diversity of crops. Evidence from Africa seems to support this assertion (Conelly and Chaiken 2000; Dorsey 1999). On the other hand, where the land holding is large enough and suitable for mechanization, the agricultural system evolves usually to a single crop, resulting in a very homogeneous landscape. Evidence from Western Europe (Pauwels and Gulinck 2000) and the American Midwest (Riebsame et al. 1994) supports this position. It has even been suggested that when modern energy intensive technology is used, agriculture
55 becomes more extensive and disintensified as large areas relying on machinery and inputs replace smaller areas of land where intensified agriculture was present (Netting 1993). It can be concluded that in areas where mechanization cannot take place, agricultural diversity is likely to increase as intensification takes place. However, if the land holding is large enough and suitable for heavy machinery, it is expected that agricultural and landscape diversity decrease. Models of Agricultural Change: Intensification and Commercialization The study of agricultural change has been a question addressed by many disciplines, including economics, anthropology, sociology, and geography. Perhaps one of the first explicit frameworks for explaining agricultural change was developed by Thomas Malthus in the late 18th century (Bilsborrow and Carr 2001). According to the Malthusian model of population growth and agricultural intensification, population grows geometrically while food production increases arithmetically (Bilsborrow and Carr 2001; Boserup 1965). As a result, as population grows, more marginal land is brought into production. Because food production cannot keep up with population growth, a Malthusian scenario will eventually lead to environmental degradation, famine and starvation. The Malthusian model has been widely criticized because it is very simplistic and does not consider cultural and institutional factors, (Brush and Turner 1987; Reenberg and Lund 1998)and empirical evidence shows that food production has been able to keep pace with population growth on a global basis without an accompanying environmental collapse, as the model predicts (Conway 2001). However, there is increasing evidence supporting the major environmental effects of human activities on the functioning of the planet (Vitousek et al. 1997).
56 During the 20th century, other frameworks explaining the processes of land-use intensification and commercialization have been developed. Each one of these models tries to explain this process using different variables and processes. The most widely known frameworks for explaining agricultural intensification and commercialization are Boserupâ€™s model of agricultural intensification (1965), Chayanovâ€™s model of peasant farms (Netting 1993), the population cycle based on the ideas of Davis and further developed by Bilsborrow (2001; 1992), the commercialization model underpinned by von Braunâ€™s ideas (1995), and the Induced Intensification theory (Turner and Shajaat Ali 1996). The first two do not consider the influence of the market and because of this they are somewhat unrealistic. The following sections briefly discuss each one of these models, their advantages and disadvantages. Boserupâ€™s model of agricultural intensification Boserup (1965) criticized the way in which classical economics looks at increasing agricultural production. According to the neoclassical economic point of view, under increasing conditions of population pressure, there are two major ways to increase agricultural production: 1) creation of new fields (extensification), and 2) increased production in existing fields (intensification). However, this distinction is not useful for landscapes where cultivated and non-cultivated lands are not sharply distinguished, like agroforestry systems (Boserup 1965). Instead, Boserup proposes to measure agricultural intensity using the frequency of cropping. As land becomes more and more scarce due to increasing population pressure, land-use intensification takes place, and this is reflected in a shortening fallow period (Bilsborrow and Carr 2001; Goldman 1993). According to this model, as intensification takes place, labor productivity declines (Goldman 1993; Netting 1993). In other words, more intensive systems mean more
57 output, but the labor demands grow faster than the output. Therefore, the model also states that as land scarcity decreases, farmers tend to disintensify production. Some empirical evidence supports this point (Netting 1993). Boserupâ€™s ideas also imply the possibility of agricultural intensification without environmental degradation because of improved management of the land, unlike the Malthusian framework (Adams and Mortimore 1997). This framework has been widely revised and amplified by economists, geographers, and anthropologists. Goldman (1993) provides an excellent review of how Boserupâ€™s ideas have been criticized and revised in different disciplines. This model has been criticized in the literature in several ways. Some authors view this model as flawed because it assumes a linear development based on a single variable (population), when in reality agricultural intensification is a process responding to a wide variety of socioeconomic, historical, and biophysical variables (Brookfield 2001; Marquette 1998). Another limitation of the Boserupian framework is that it does not work very well in areas where market forces are important in agricultural production (Netting 1993; Brush and Turner 1987). However, it seems to work well in areas where agricultural activities are not fully integrated into the market (Brush and Turner 1987). Others criticize the model because there is empirical evidence showing that there is not a significant correlation between increased productivity and high population density (Goldman 1993). More recently, Boserupâ€™s model has been criticized because it is difficult to implement in a modeling environment, making it difficult to predict land-use changes associated with intensification (Lambin, Rounsevell, and Geist 2000). Chayanovâ€™s model of peasant farms According to Chayanovâ€™s model, peasant economies are mostly defined by the characteristics of family labor. Peasant farms relying on family labor and limited capital
58 do not show a tendency to maximize profit, and they rather try to maximize income (Netting 1993). In this framework, a peasant farm is regarded as a self-sufficient enterprise without wage labor, with each household dedicated to its own reproduction, and with family labor being the critical variable. According to Chayanovâ€™s ideas, farmers try to satisfy family needs according to a communityâ€™s definition of an acceptable standard of living. This model implies three basic assumptions: 1) Arable land is not in short supply and can be acquired by the household; 2) there is not an opportunity to work outside the farm for wages; 3) each peasant community has a social norm for the minimum acceptable income per person. Under these circumstances, the labor force is the limiting factor of production (Netting 1993). Labor availability changes with the evolution of the household. The composition of the labor pool and the degree of labor activity are determined by the number of able-bodied family members and the number of consuming members for whom they should provide. In reality, finding farms that fit the three conditions specified above for Chayanovâ€™s model is very difficult, and only certain areas of the tropics match them (Netting 1993). Although this model has been very appealing to anthropologists studying pre-capitalist societies producing for self-consumption rather than for exchange, it breaks apart when dealing with societies engaged in market production. In farms engaged in commercial production, labor is no longer restricted to reproduction activities of the household. It also fails in areas where increasing land scarcity is an issue. However, the cyclical nature of agricultural change based on the evolution of the family has been a powerful tool in explaining land-use patterns. For example, Moran and Brondizio (1998)
59 explain farm land-use patterns in the Amazon basin as a function of the development stage of the household. More recently, Walker (2003) presents a model of landscape evolution in recently colonized areas of the Amazon basin where the amount of forest cleared and land planted depends on the number of able-bodied members of each household. In his model, he builds upon some of Chayanovâ€™s ideas, and introduces the influence of markets in determining farmersâ€™ goals and land-use decisions, switching gradually from a subsistence based agriculture towards a more commercially oriented agriculture as markets arrive to recently colonized areas. Population cycle According to Bilsborrow and his collaborators (2001; 1992) Kingsley Davis suggested that demographic responses to population pressure occur before agricultural intensification. Critics of this view state that this demographic perspective does not take into account the economic behavior of people. Bilsborrow and Carr (2001) tried to reconcile these two positions by combining economic behavior and demographic responses into a single model of agricultural change that I have called the population cycle. According to this model, migration has a critical role to play when dealing with increasing population pressure. There are four phases that can occur simultaneously in any given area: Tenurial Extensification Technological Demographic According to the model, the first signs of population pressure result in a clear delineation of access to land and resources under increasing scarcity. At this stage, outmigration is not a desirable outcome. The second phase involves extensification, and
60 may involve migration to new, less populated areas. The technological stage implies new techniques of land use (i.e., cash crops) that is accompanied by an intensification of production. The final stage, the demographic phase, involves the postponement of marriage, and or reduction of fertility, and outmigration to other areas. These responses are mediated by factors at different scales in the human and physical environments that include the availability of arable land, quality of natural resources, government policies, land tenure, accessibility to labor and market, and institutional factors among others. There are several authors presenting empirical evidence of this sequence of events (Bilsborrow and Carr 2001; Bilsborrow and Okoth Ogendo 1992; Murphy, Bilsborrow, and Pichn 1997). Bilsborrow and Carr (2001) present evidence of the extensification stage in Latin America, as population pressure increases. Murphy et al. (1997) provide evidence from Ecuador of the first and second stages. The recent commercial intensification of coffee production in Colombia could be placed in the third and fourth stages. Bilsborrow recognizes that this model has limitations, and that it can only be applied in certain instances. He closes his discussion by stating that no single theory can adequately explain the complex interactions between population growth and land-use change (Bilsborrow and Okoth Ogendo 1992). Commercialization of agriculture Commercialization of agriculture not only means an increased marketing of agricultural outputs. It also implies the selection of products and inputs based on profit maximization (Pingali and Rosegrant 1995). Commercialization occurs for both staple foods and cash crops. The commercialization of agricultural systems implies greater market orientation of farm production, a progressive substitution of non-traded inputs in
61 favor of purchased inputs, and a decline of integrated farming systems and their replacement by specialized enterprises (Pingali and Rosegrant 1995). In order for commercialization to be successful, adequate infrastructure, market, macroeconomic and trade policies should be present. Von Braun (1995) cites four major determinants of commercialization: 1) population change, 2) new technologies and crops, 3) infrastructure and market creation, and 4) macroeconomic and trade policy. According to Pingali and Rosengrat (1995), in the early stages of commercialization farms diversify out of the staple food production. As rural economies grow and markets develop, there is a gradual movement out of subsistence agriculture to a more diversified market oriented system (Walker 2003). This change is triggered by rapid technological change, improved rural infrastructure, and diversification in food patterns in urban and rural markets. This diversification implies the initial addition of crops, but as the level of commercial orientation increases, mixed, complex farming systems give way to specialized production. At the national level, commercialization leads to higher diversity of marketed products, while at the regional and farm level it leads to specialization (Pingali and Rosegrant 1995) With increasing commercialization, mechanical and chemical technologies will substitute for human labor in the more control-intensive operations (i.e., weeding) and the power-intensive activities (i.e., tilling) (Pingali and Rosegrant 1995). These authors present a nice summary of the differences in input use and nutrition between subsistence, semi-commercial and commercial farms. There is a transition from human/animal power to motor sources of power. Traditional systems are usually based on family labor, while commercial systems rely increasingly on hired labor. Traditional systems derive most of
62 their income from the farm, while off-farm income becomes increasingly important on the commercial side. In terms of soil fertility maintenance, subsistence farms rely mostly on animal manure and household refuse, while commercial farms use commercial fertilizers. Commercial farms rely more on chemical pesticides, herbicides and insecticides for pest and disease control. Finally, subsistence farms produce most of their food needs, while commercial farms buy most of their food. The commercialization of agriculture also involves a change in the decision-making process of small farmers. A subsistence farmer just requires producing enough to feed his/her family. A commercial-crop producer, on the other hand, faces two decisions: 1) how much output to produce, and 2) what to produce and how to produce it based on market signals (Binswanger 1986). There are many impacts to peasant families resulting from commercialization. According to Von Braun (1995), smaller farmers allocate a larger share of their land to commercial agriculture than larger farms. This is also the case in the Colombian coffee lands, where smaller farms have a larger proportion devoted to coffee than larger farms (Palacios 1980; FNC 1997). In all the case studies reviewed by von Braun (1995), farms maintain a considerable portion of the farm in staple foods. This is because farmers decide to maintain certain food security. In general, women work less in the commercial plots than men. This may potentially lead to conflict, as women are marginalized from commercial production and may lose access to certain resources that they customarily controlled or shared with men. Evidence from coffee producing farmers in Ethiopia (Hamer and Hamer 1994) and from wetlands converted to agricultural fields in Gambia (Carney 1996) support this statement. The transition to commercial agriculture also
63 brings with it changes in the allocation of labor. As stated above, hired labor becomes increasingly important as commercialization takes place. Additionally, the types of labor may change, as the new crops associated with the commercialization of agriculture may imply different field operations and processing requirements. In terms of general household welfare, the commercialization of agriculture increases household income (von Braun 1995). As a result, households will be able to buy more food, work less in the fields (improving child care), improve household sanitation, water availability (quality and quantity), and better health care access. Commercialization may also lead to improved education for children. Although commercialization has the potential to increase overall nutrition, it also involves health risks associated with the increased exposure to chemical fertilizers and pesticides (Pingali and Rosegrant 1995). Induced intensification Induced intensification emerged as an alternative explanation of agricultural development that tries to incorporate the exceptions and nuances that Boserupâ€™s and Chayanovâ€™s general models failed to address, in particular, the influence of market forces in agricultural intensification and commercialization, and variations among households (Turner and Shajaat Ali 1996). This framework was first introduced by Hyden, et al. (1993) and basically explains changes in agricultural intensity, technology and management. In this model, farmers have different production goals and rules for manipulating labor and capital towards these goals. This variation follows from the proportion of cultivation for subsistence and for the market at the farm level (Turner and Shajaat Ali 1996). In any small farm there are two basic rules of production: 1) minimize risk in supplying basic needs, and 2) minimize labor for production. As farmers become
64 involved in market production, their aspirations change, and allocate their inputs according to how well they are integrated into that market. However, smallholder farmers may not respond to market signals because(Turner and Shajaat Ali 1996): 1) they imply high levels of risk, 2) their production goals and logic are not fully market oriented, and 3) farms differ from each other in terms of their socioeconomic status and market orientation, so their responses to market signals vary accordingly. As market demands increase or decrease, management strategies change. A particular level of demand (i.e., a mix of market and subsistence) leads to a specific land-management strategy. As the demand increases, it is first met by increasing labor and capital inputs within the management strategy, until marginal returns are so low that they prompt a switch to another strategy (Turner and Shajaat Ali 1996). Therefore, the trajectory of agricultural growth and intensity is a stair-stepped one, marked by thresholds where strategies change. The trajectory of change may be impeded or facilitated by environmental, socioeconomic, political, and institutional factors. According to the same authors, at any of these thresholds there may be intensification, involution, or stagnation, and commercialization is an integral part of agricultural change. Involution implies that production increases, but with significant declines in the marginal utility of inputs, and very few other options (i.e., agricultural strategies) are available for farmers. In stagnation, production does not increase, and may even decline. Empirical evidence from Bangladesh (Turner and Shajaat Ali 1996) and different places in Africa (Hyden, Kates, and Turner 1993) suggests that in many areas agricultural growth follows a stair-stepped path very similar to the one described above. However, its
65 ability to explain the process of agricultural growth varies according to the specific context in which the studies were carried out. No Existing Single Model Can Adequately Explain Agricultural Development The availability of numerous models of agricultural development suggests that no single model is sufficient to explain the process of agricultural change. Furthermore, many of these models are based on a limited number of factors. It has become increasingly clear that the process of agricultural change requires a multi-causal approach that incorporates not only population pressure, but also factors such as the presence of markets, policy environment, social norms and customs, and biophysical variables among others (Lambin, Rounsevell, and Geist 2000). The Malthusian, Boserupian or Chayanovâ€™s models of agricultural change, relying on population as its main explanatory variable, are simplistic. In reality, there are many other factors determining agriculture and other land-use practices (Reenberg and Lund 1998). Bilsborrow and Carr (2001) argue that neither a Malthusian nor a Boserupian model offer a satisfactory explanation of the population-environment linkages caused by agriculture. They present case studies where contextual factors affect the agricultural development path in different ways. Goldman (1993) presents a similar situation for Kenya. According to Goldman, agricultural innovation leading to intensification can follow different paths in different social, economic, and biophysical contexts. An important point is that the process of agricultural intensification and commercialization is not spatially homogeneous. Certain production strategies may be intensified while others are disintensified at the same time (Wiegers et al. 1999). As farmers allocate their resources among different production spaces, they choose which land-use practices to intensify and disintensify based on the context, availability of labor, capital, and other income-generating activities.
66 It is becoming increasingly difficult to separate the intensification effects caused by population pressure, market integration (i.e., commercialization), and changes in the biophysical environment. Rural population pressure is likely to lead to one or more of the following (Hyden, Kates, and Turner 1993): Intensify inputs to agriculture Intensification of outputs Expansion of land under cultivation Change in the proportion of crops marketed Greater economic diversification Changes in well-being in terms of diet quality and environmental degradation Changes in social relations such as increased inequality, and changes in gender relations. Models of intensification need to take into account the differences between households. There are many examples showing how households with different assets follow different land-use paths, including agricultural intensification and expansion. Zimmerer (1996a) shows how households with different levels of wealth follow different agricultural development paths in the Peruvian Andes, with wealthier households maintaining more of the traditional potato varieties. Murphy and her collaborators (1997) also present convincing evidence of how several factors differentiate households in terms of their agricultural strategies in the Ecuadorian Amazon. In order to move beyond explanations at the household level and try to explain the landscape transformations associated with agricultural change, it will be necessary to understand how different kinds of households respond to population and market pressures. However, given current models of agricultural intensification and commercialization, translating household models into landscape models seems to be extremely difficult to accomplish (Lambin, Rounsevell, and Geist 2000). However, there are some examples of successful scaling up from household to landscape like Walkerâ€™s (2003) model of household evolution in the
67 Brazilian Amazon. Models of intensification are also difficult to translate to a landscape when considering agroforestry systems, as well as in areas where there is continued use of abandoned plots purposefully enriched with valuable species is common, like in the Amazon basin (Coomes and Barham 1997; Coomes and Burt 1997; Smith 1999b) Putting It All Together: Agriculture and Landscape Change Contemporary consumption demands can only be met with major land-cover conversion and modification (Turner et al. 1995), and these rates of cover change depend on the demand of land-based commodities (Veldkamp and Lambin 2001). When considering agricultural products, land-cover modification in the form of agricultural intensification and commercialization will become more important in supplying the growing demand for agricultural goods (Ruttan 1994). Because commercial intensification of agricultural production usually implies significant changes in how the land is used and managed (Binswanger 1986; Pingali and Rosegrant 1995; von Braun 1995), as farms intensify and commercialize, they change their land-use practices. Although agricultural land-use decisions are usually made at the farm level, the transformations associated with agricultural change are also manifest at the landscape level. The land-use and land-cover changes of individual farms aggregate to produce specific landscape-evolution patterns. Therefore, it is very important to understand how the effects of agricultural changes in individual farms aggregate and translate into specific landscape patterns. Landscape Manifestations of Agricultural Intensification and Commercialization Conceptual Frameworks Despite the limitations of many of the intensification and commercialization models presented before, it is an interesting conceptual exercise to discuss some of their
68 manifestations at the landscape level. Because most of these models were developed at the household level, there has been relatively little work in analyzing their landscape effects. One of the few examples found in the literature involves the successful coupling of Boserupâ€™s model of agricultural change and Von Thnenâ€™s model of land-use development around a central market (Brush and Turner 1987; Netting 1993). A secondary objective of this dissertation is to analyze if the factors that, according to these models, are responsible for agricultural intensification at the household level, can also be used to explain agricultural intensification at the landscape level. According to the Malthusian view, the best lands will be used first and, as population growth takes place, more marginal lands are brought into production, leading to environmental degradation. (Bilsborrow and Carr 2001; Bilsborrow and Okoth Ogendo 1992). The landscape patterns in a Malthusian scenario are characterized by areas of widespread degradation and exhaustion of lands, forcing farmers to open new land for cultivation. Some evidence seems to undermine this argument, as there are examples or areas where agricultural intensification is taking place, like certain regions of the United States, where agricultural change is associated with enhanced environmental quality (Lee, Ferraro, and Barrett 2001). Some case studies from Kenya also suggest that in cases of extreme population pressure, intensification can take place without environmental degradation, something opposite to what is expected under a Malthusian scenario (Conelly and Chaiken 2000). In her model of agricultural intensification, Boserup (1965) recognizes that in a region undergoing intensification, the best lands will be intensified first. This means that a landscape going through agricultural intensification is a mosaic of more intensive and
69 less intensive production systems where agroecological conditions determine which lands will be intensified first. Eventually, as population grows, these less suitable lands will be intensified as well. There appears to be some evidence against this situation. Southworth and Tucker (2001) present results from a mountainous area of Honduras where intensification of agricultural production has meant the abandonment of some areas that are either too inaccessible or not well suited to the technology used in the intensive agricultural system. A similar situation is presented by Di Pietro (2001) in the French Pyrenees, where agricultural lands in flat areas are intensified while agriculture in sloping terrain is abandoned. In this second case, intensification occurred under declining population, something that Boserupâ€™s model does not predict. A similar situation where agricultural intensification is taking place without the necessary population density pressure comes from Latin America (Bilsborrow and Carr 2001). In many countries, land-use intensification is taking place as a result of export operations rather than population pressure. These cases suggest a spatial pattern characterized by intensified agricultural fields interspersed with abandoned agricultural land and or less disturbed natural habitats, rather than a landscape characterized by more intensive agriculture surrounded by less intensive agriculture, as suggested by Boserup. The basic premise of Chayanovâ€™s model of peasant farms is that the amount of land farmed in each household depends on the number of able-bodied family members that can work in the fields. As a result, one can expect that as a family grows, the area farmed also increases, and as a family shrinks by children establishing their own homestead or death of the parents, the size of the farm decreases. At the landscape level, in an area not experiencing land scarcity, the area farmed should reflect the stage of development of the
70 families inhabiting that landscape. Larger families with more able-bodied workers will have more land under cultivation. Therefore, the average age and the number of dependents per family in a given region should provide a good indicator of how much land is used for agriculture. Walker (2003) uses this relationship between number of able-bodied family members and land under agricultural production for his model of peasant agriculture along the trans-Amazon highway. A similar pattern of landscape evolution associated with the stage of development of rural families has been documented for other parts of the trans-Amazon highway by Moran and Brondizio (1998). These authors have shown that as a wave of colonizers arrives, only a small portion of the rainforest is cleared to provide an area for a subsistence plot. As families become established and grow, children become an available labor source for clearing the land and cultivating new fields. Finally, as children leave the homestead, the forest starts to regenerate in what were previously agricultural fields. This example shows clearly how changes at the farm level can influence the landscape spatial configuration, and how the evolution of the landscape is tied to the evolution of individual families. When talking about models of commercialization of agriculture at the farm level there are several landscape implications (von Braun 1995; Pingali and Rosegrant 1995). In first place, Pingali and Rosengrat (1995) suggest that when commercialization begins, the number of crops in a farm usually increases as a region becomes integrated to the market. As this incipient commercial agriculture is intensified, and technology for specific products is adopted, the initial surge in agrodiversity gives way to farms specialized in a few products. These authors suggest that commercialization of agriculture leads to regional specialization and national/global diversification, as different
71 regions with specific agroecological conditions specialize in different products. This would suggest very homogeneous landscapes at the regional level, with a few dominating crops. However, most small farmers do not behave as profit maximizing decision makers from an economic perspective (Walker 2003). Many of them may maintain a diverse array of crops, both subsistence and commercial, in order to minimize risks associated with market and environmental fluctuations, but also as a mechanism to earn sufficient income (Zimmerer 1996a; von Braun 1995). Bebbington (1996a) indicates that rural livelihoods have diversified as rural areas are incorporated to the market economy in the South American Andes. This diversification involves the adoption of new crops and their integration to the traditional rural livelihood, as well as other strategies such as off-farm income. Therefore, it could be argued that commercialization leads to higher diversity of crops, which translates into more heterogeneous landscapes. However, this is not always the case. According to some authors, commercialization leads to the consolidation of farm landholdings as urbanization takes place, which usually results in less homogeneous landscapes (Pingali and Rosegrant 1995). This seems to be the case in many developed countries like Finland, where agricultural production is being concentrated in larger production patches (Hietala-Koivu 2002). Because the concentration of land is accompanied by mechanization, the net effect is that the landscape becomes simplified (Di Pietro 2001; Pauwels and Gulinck 2000). In other areas of the world, commercialization has not resulted in the consolidation of farms. For example, in the Colombian coffee lands, the number of coffee farms has increased from around 300,000 in 1970 to more than 600,000 in 1997, a period also characterized by a very strong commercial-intensification process (FNC 1970, 1997). At the same time, the coffee
72 growing landscapes appear to have diversified, as it will be shown later on. Therefore, there is no clear evidence on what are the effects of agricultural commercialization on land concentration. The landscape transformations of the induced intensification conceptual framework are not as straight forward as the effects of other models discussed so far. This is because, as a hybrid framework combining Boserupâ€™s and Chayanovâ€™s ideas with agricultural commercialization, the induced intensification framework is more complex. Differences among households (i.e., socioeconomic, goals, available labor, market involvement) make it difficult to identify a single landscape pattern. Each household is engaged in its own trajectory of change (Turner and Shajaat Ali 1996). According to this model these trajectories of change are stair-stepped, therefore, landscapes should reflect this characteristic. This means that landscape changes are associated with households moving to the next, higher intensity management strategy, and areas without change associated with farms that have not yet reached that threshold. As a result, landscapes should be patchy in terms of management strategies, with many different levels of agricultural management coexisting in the same landscape. Empirical evidence from Bangladesh suggests that areas with better conditions (i.e., environmental, accessibility, market) show a higher correlation with more intense agricultural systems than areas with worse conditions, and these more intense systems are related to market demand (Turner and Shajaat Ali 1996). The same authors show that population density accounted for more of the explained variance in intensity than market participation. In other words, although the induced intensification framework results from the interaction of demographic variables
73 and market involvement, in this case the demographic component is more important in explaining agricultural intensification. Landscapes and Agricultural Change Most of the models of agricultural intensification and commercialization assume that farmers are engaged in the production of a single or a few crops in a homogeneous production space. However, farmers have access to different production spaces determined by environmental, social, institutional and economic factors. In the Peruvian Andes, people have access to different production spaces determined by the environmental variability associated with the mountainous terrain. These spaces of production are associated with cropping and animal husbandry systems(Wiegers et al. 1999; Zimmerer 1999). Certain land uses in these production spaces have intensified while others have disintensified based on farmerâ€™s land-use decisions under this ever-changing context. Access to some resources, such as communal grazing lands or forests, is mediated through institutional factors. These institutions determine who has access to what resources, therefore delineating how farmers may use these resources (Southworth and Tucker 2001; Ostrom et al. 1999). Based on the opportunities and constraints offered by social, economic, political, environmental, and accessibility factors, farmers allocate their resources (labor, capital, technology) among these production spaces to determine a specific land-use pattern in their farm. As these conditions change, farmers change their land use accordingly, meaning that landscapes are constantly evolving as the factors affecting the land user change. Innovation is an essential part of agricultural change. Agricultural innovations have been classified as yield increasing, labor saving, and quality enhancing (Binswanger 1986). Yield-increasing innovations are those that raise the productivity of land. Labor
74 saving innovations reduce the amount of labor required for agricultural production. Quality-enhancing innovations improve the quality of agricultural products. Each one of these has different implications in terms of how landscapes change. According to Binswanger (1986), yield increasing innovations tend to reduce the area required to produce a unit of output. Labor saving innovations do not usually reduce the area required for production. Finally, quality-enhancing innovations leave the area under cultivation unaffected. Furthermore, not all farmers adopt the innovations at the same time. Yapa (1977) suggests that farmers that rely more on labor (i.e., small farmers) tend to favor labor-saving techniques, while farmers with more access to material inputs tend to favor yield-increasing and quality-increasing innovations. Therefore, landscapes reflect these differences, and patches of more input-intensive innovation agriculture are interspersed with labor-saving innovation agriculture. The level of resources of a household also has an influence on what activities farmers choose. Coomes and Burt (1997) show how farmers with more land tend to have a more diverse array of productive strategies that are also more sustainable. Murphy and her collaborators (1997) present evidence from the Ecuadorian Amazon colonization frontier where the initial assets of the homesteaders, as well as their area of origin, have a tremendous influence on how land is used. In this area, land use also reflects that labor is in short supply, favoring productive strategies that can be intensive in their use of land but with relatively little need for labor such as cattle raising. In areas engaged in commercial production, in certain instances farmers with smaller farms tend to leave a larger share of their farms in commercial crops than larger landholders (Palacios 1980).
75 The selection of productive strategies also depends on changes at higher levels. For example, market changes also determine how farm resources will be allocated. Pinto-Correia (1999) provides a case study in Portugal in which certain land-use practices became uncompetitive when this country joined the European Union, and some productive strategies were abandoned. Bilsborrow and Carr (2001) also present evidence on how market conditions induce intensive use of lands even at low population densities in Latin America, mostly for planting cash crops in large plantations. Southworth and Tucker (2001) show convincing evidence of how land tenure and accessibility determine the spatial patterns of change in Honduras. These examples clearly highlight how landscape transformations associated with agricultural intensification and commercialization are inseparable from important sociodemographic changes (Bouchard and Domon 1997), market changes (Pinto-Correia 1999), institutional transformations (Bebbington 1996a), and environmental factors. The inherent spatial variability of these variables also creates production spaces with different opportunities and constraints. Farmers adopt land-use practices and crops according to these limitations and advantages, resulting in a very diverse landscape, where intensification of some land uses is taking place because of the opportunities presented by the landscape and its context, and other land uses are being disintensified and abandoned based on the limitations. Additionally, these examples and others from the scientific literature suggest that different factors seem to be more important in determining agricultural intensification and commercialization in different contexts. Research Questions and Hypotheses The main research question was introduced in Chapter 1. By now, it should not be a surprise to the reader that agricultural intensification and commercialization lead to
76 landscape transformations. The purpose of this dissertation is to study the landscape changes associated with the commercial intensification of coffee production in the Colombian coffee lands. The following paragraphs will introduce each one of the hypotheses that I will try to test with this case study. Landscapes diversify as commercial intensification of agriculture takes place Farmers have access to different production spaces determined by environmental, social, economic, political, and institutional factors. This translates into a variety of land-use practices at the farm level that, as discussed earlier in this chapter, are also manifest at the landscape level. As commercial intensification of agriculture takes place, some crops might be introduced, traditional subsistence agriculture is also maintained, and some areas not suited for the intensive techniques might be abandoned or used in different ways. The net result is an agricultural landscape characterized by more diversity of land covers. Also, as markets develop, and accessibility improves, different crops appear on the landscape reflecting the differential accessibility to markets. Therefore, I hypothesize that a landscape becomes more diverse and heterogeneous as commercial intensification of agriculture takes place. H 0 : Landscapes do not change as commercial intensification takes place H 1 : Landscapes become more diverse and heterogeneous as commercial intensification takes place In the Colombian coffee lands, if the area planted in coffee and pasture decreases and the area in other crops increases as coffee production intensification takes place, the landscape is expected to become more heterogeneous and the null hypothesis can be rejected.
77 In a region undergoing intensification, some land-use practices intensify while others disintensify The selection of land-management practices depends on socioeconomic, political, institutional, and environmental conditions that offer opportunities and determine limitations. Farmers chose the better options based on these factors, and change their land use accordingly. However, what is the overall effect when multiple households are combined in a landscape? In a hypothetical region where cultural and biophysical differences are minimal and families are very similar in terms of their size, assets and cultural background, one would expect that in the same way in which certain land uses become more predominant and intensification of production takes place at the farm level, the same should happen at the landscape level. Therefore, families allocate their limited resources to a set of productive activities where certain land uses intensify, while others disintensify. When all the factors that differentiate households from each other are considered, a similar landscape pattern, although not as strong as in the hypothetical situation, should emerge. Although not all households are equal, biophysical conditions and access infrastructure make a specific product prevalent throughout a region. Despite the fact that the amount of land devoted to this crop in each farm depends on family labor, family assets, and other factors, it is expected that the landscape will exhibit a predominance of the product or products that have a comparative advantage given certain market conditions. At the same time, farmers devote less time and resources to other crops that do not have a strong market demand, or less effort invested in subsistence agriculture, as it is more desirable for the farmer to devote more land to the intensive crop than to produce food locally. Therefore, it can be hypothesized that in any given
78 landscape, the intensification of certain production systems is accompanied by the disintensification of other production systems. H 0 : In any given landscape, all different land uses will intensify at the same time H 1 : In any given landscape, not all land uses will intensify at the same time The conditions necessary for rejecting the null hypothesis are that yields do not increase for all agricultural products, and there are no crops that become less important in each landscape. Factors accompanying intensification vary with the context. Agricultural intensification and commercialization of agriculture are caused by a multitude of factors rather than a single variable, as suggested by some of the conceptual models presented above. These forces interact in different ways in various geographical and historical contexts. Therefore, it is expected that the relative importance of these variables changes from region to region. Using demographic variables as suggested by Boserupâ€™s model (Population density) and Chayanovâ€™s conceptual framework (Family size, average age of the family, number of dependants per family), market access variables as suggested by Von Braun, and Pingali and Rosengrat (accessibility measured as the average distance to the nearest major town in each unit of analysis, number of state institutions, number of banks), and environmental factors determining the agroecological offer (rainfall in the two driest months, rainfall in the two wettest months, latitude, longitude) I will try to determine, among these broad categories, which is the one that has a stronger relationship with the process of intensification in the Colombian case. I would argue, that in a landscape that has been integrated to the market since the early 20th century, like the Colombian case, market forces will have a stronger influence than the demographic or environmental factors. Also, because the importance of these factors may
79 change from one region to another, I will try to test their relative contributions in different regions. H 0 : Market forces (accessibility, number of banks, number of state institutions) have the same influence as demographic forces (rural population density, family size, family average age, dependents per family) and environmental factors (rainfall in driest two months, rainfall in wettest two months, location) in their relationship with the commercial intensification of coffee (area planted in sun grown coffee) in all areas of the country H 1 : Market forces, demographic factors, and environmental factors have different relative importance in their relationship with commercial intensification of coffee in all areas of the country The conditions necessary to reject the null hypothesis require that the significant variables from the regression analysis change from region to region and that the relative contribution of each of the variables is different.
CHAPTER 3 COFFEE AS A CROP IN THE INTERNATIONAL MARKET Coffee is a crop that has transformed societies and landscapes in a dramatic way. When coffee was introduced to Europe in the 17 th century, coffeehouses soon became the center of social, political, commercial, and artistic life (Smith 1985). As the demand for coffee increased, and the Arab monopoly on coffee production was broken in the early 17 th century, European powers quickly planted coffee in colonial outposts. The main coffee-production centers shifted to new locations as the balance of power in Europe shifted, and in response to some diseases. However, one thing remains constant through time: coffee is produced in poor countries and consumed in rich and metropolitan areas of the world (Dicum and Luttinger 1999). Today, coffee is the second most valuable legal export commodity (after oil) traded in the world market, representing more than 10 billion dollars in revenues (Dicum and Luttinger 1999; Rappole, King, and Vega-Rivera 2003). Presently close to 11 million hectares in more than 70 countries are planted with this crop (FAO 2001). Therefore, understanding how coffee production intensification influences landscape evolution has relevance for many areas of the world. The purpose of this chapter is to describe the characteristics and history of coffee as a crop. Coffee as a Crop Taxonomically speaking, coffee belongs to the genus Coffea, in the Rubieaceae (Willson 1985b; Wrigley 1988; Cherrier and Berthaud 1985). Because it grows naturally in the understory of African rainforests, it is an evergreen bush adapted to shade (Willson 1985b; Wrigley 1988; Willson 1999). As a result, it has not developed mechanisms to 80
81 reduce water loss in times of stress, making it very sensitive to extended periods of dry weather (Willson 1985a). There are many wild varieties adapted to a wide range of environmental conditions, including some resistance to drought and disease. This variety of wild coffees is the result of the interaction between genetic and ecological variability in natural habitats (Cherrier and Berthaud 1985). However, there are only two coffee species that are of commercial interest (Willson 1999; Dicum and Luttinger 1999; Cherrier and Berthaud 1985): 1) Coffea arabica Linnaeus, usually known as Arabica coffee, and 2) Coffea canephora Pierre ex Froehner, also known as Robusta coffee. They are native to the highlands of Ethiopia and the Congo River rainforests in Uganda respectively. Historically, Coffea arabica has played a more important role. Robusta coffee only became an important crop after the Second World War (Smith 1985). Arabica coffee produces a better tasting bean than Robusta, and it is one of the reasons why it is preferred. The increasing importance of Robusta coffee is the result of two processes. In first place, it became the coffee of choice for the instant coffee market. Although this process was developed in the 1930s by Nestl, it only became a popular drink during WWII, when instant coffee was part of the Allied forces rations (Dicum and Luttinger 1999). Because this process does not result in a particularly good tasting cup of coffee, Arabica coffee beans would be wasted if used to produce instant coffee. In second place, Robusta coffee was promoted during the late 1940s and 1950s by the European powers as a mechanism of producing coffee in their own colonies. Soon Ivory Coast and Angola, colonies of France and Portugal respectively, became important producers of Robusta coffee (Dicum and Luttinger 1999). This variety is more resistant to disease and grows in
82 a different environment, making it a good crop for areas where Arabica coffee cannot be grown for environmental reasons (van der Vossen 1985). Perhaps one of the most appealing characteristics of coffee is that it is a mild stimulant. Coffee contains three purines that are responsible for this situation. These substances are caffeine, theobromine, and theophylline. Of these, caffeine is the most widely known and the least desirable from a health point of view (Willson 1999). Besides producing a bean that is not as good tasting as Arabica beans, Robusta has more caffeine than Arabica, 2-3% vs. 1-1.3% respectively (Willson 1999; Smith 1985; van der Vossen 1985). Coffee roasters have developed methods to remove caffeine from coffee (Wrigley 1988) and in recent years plant breeders developed caffeine-free coffee varieties (Willson 1999). However, millions of people wake up every day with a desire for a full-bodied cup of coffee and its associated stimulant properties. Because of its tropical rainforest origins, coffee is a crop that cannot tolerate frosts. Arabica coffee is a mountain-loving plant that enjoys a subtropical, almost temperate climate while Robusta coffee is better adapted to higher temperatures and is more sensitive to frost (Wrigley 1988). The ideal temperature range for growing Arabica coffee lies between 17-25 C, with the mean temperature as close as possible to 20 C (Willson 1985a; Wrigley 1988). Lower or higher temperatures restrict the development of the coffee cherries (Wrigley 1988). Robusta coffee grows better at higher temperatures. The ideal temperature range is between 24-30 C (Wrigley 1988; Willson 1985a). In both cases, temperatures should have small daily and yearly variations. Coffee is a crop that is very sensitive to very-wet or very-dry conditions and rainfall should be well distributed throughout the year with a definite dry period (Wrigley
83 1988). A dry season from 6 to 12 weeks (14 at most) is needed to build the water stress necessary to break the dormancy of flower buds (Wrigley 1988). After the first rains, the flowers open. If the dry season is too long, the dormant flower buds never develop, leading to a poor crop. On the other hand, very-wet conditions are not suitable for the coffee plot, as soils may become waterlogged. Furthermore, intense rainfall may force the coffee cherries to fall to the ground before they are ripe. This may also occur during the flowering period, leading to a poor crop if flowers fall before pollination takes place. Arabica needs at least 1200-1500 mm yr -1 , but less than 2500 mm yr -1 (Wrigley 1988). Robusta coffee requires precipitation above 1550 mm yr -1 (Willson 1985a). The rainfall distribution controls the cropping pattern (Willson 1985a). If the rainfall distribution is unimodal, there is one flowering season followed by a period of crop development. This is the situation in the subtropical areas. When the rainfall distribution is bimodal, the coffee plants experience two flowering and two cropping seasons. This condition becomes prevalent as one moves towards the Equator. Coffee prefers deep, well drained, loamy soils that are slightly acid, and rich in organic matter and exchangeable bases (Willson 1985a; Wrigley 1988). Arabicas grow better in soils with pH 5.2-6.5. Robustas need soils with pH higher than 4.5, and are more resistant to more basic soils (Willson 1985a). Arabica tends to have a deeper root system than Robusta, making it less susceptible to extended periods of dry weather, as the root system can reach deeper into the soil to obtain water. Additionally, because of its shallower root system, Robusta coffee has stronger competition from weeds and other plants (Wrigley 1988). Because of its origins as a forest plant, coffee is well adapted to the physical and chemical characteristics of forest soils. Therefore, it is not a surprise that
84 historically cleared-forest soils have been favored for planting coffee(Willson 1999; Wrigley 1988). The best Arabicas are grown on recent volcanic deposits all throughout the world. Examples include a large area of the Colombian coffee lands, regions in Costa Rica, Kenya, Hawaii, and Indonesia (Wrigley 1988). As a result of different pollination characteristics, Arabica and Robusta coffees are propagated in different ways. Because Arabica coffee is self pollinated, it is propagated by seed, as self pollination assures very homogeneous characteristics of the progeny (Willson 1985b). The seed loses its viability quickly, and in a commercial operation coffee cherries need to be pulped, fermented, and washed before they can be used as seed. Robusta coffee requires cross-pollination. Because with this kind of pollination there is no guarantee on the characteristics of the progeny, this coffee type is usually planted through vegetative propagation (Willson 1985b). The most common Arabica varieties are Bourbon (Coffea arabica v. bourbon) and Typica (Coffea arabica v. typica). All other varieties are believed to have derived from these, including large bean varieties like Maragogipe, and dwarf varieties like Caturra, San Ramn, and Mokka (Willson 1985b). In the case of Robusta coffees, the main types are Coffea canephora v. nganda, which is present mostly in Uganda, and Coffea canephora v. Kouilou, which is more widely distributed (Willson 1985b). Despite the wide variety of wild coffees, breeding is mostly restricted to Coffea arabica and Coffea canephora (van der Vossen 1985). The breeding efforts have concentrated in developing resistance to certain diseases. Most of these efforts produce a lower quality bean (van der Vossen 1985). As it will be discussed later on, as a result of its historical spread, Coffea arabica species have a very narrow genetic base. Coffee
85 bushes, mostly Arabica but also Robusta, have biennial bearing, a distinctive cropping pattern in which a year of high productivity is alternated with a year of low productivity (Dicum and Luttinger 1999; Wrigley 1988). This characteristic can be managed by pruning the coffee bush, which modifies the ageing process, and apparently forces the plant to maintain high productivity in secondary branches (Wrigley 1988). The establishment of a coffee plot begins by setting up a nursery. The beds should be kept moist and the coffee seedlings need to be shaded. As they grow up, shade is gradually reduced, until eventually the young coffee bushes are exposed to direct sunlight for a few weeks before they are transplanted to the field (Willson 1985b). Coffee plants are transplanted to their permanent location when they are 6-18 months old, preferably at the beginning of the rainy season (Wrigley 1988). After planting, young coffee plants need some shade from temporary trees or nurse crops (Willson 1985b). As any other crop, coffee is susceptible to pests and diseases. Some of these have played a very important role in the geographical spread and shift of the major centers of production. Perhaps the most influential disease is the Coffee Leaf Rust (Hemileia vastatrix Berk. and Br.). This was the disease responsible for wiping out the plantations in the island of Ceylon in the mid to late 19 th century. Wrigley (1988) provides a good overview of this disease. According to Wrigley, the disease is characterized by orange-yellow circular spots on the coffee leaves. As these lesions become older they become necrotic, slowly killing the leaves, defoliating the coffee bush. If the leaves are unable to supply the needs of the developing coffee cherries, the carbohydrate reserves of the roots and stems are used to continue cherry development, weakening the coffee bush. If repeated defoliations occur, the coffee bush is killed. It should be emphasized that Coffee
86 Leaf Rust disease does not normally kill healthy coffee bushes. These become susceptible to die when they are old, exposed to extreme environmental conditions, or in areas with poor soils, and poorly managed. The most important pest of coffee is the Coffee Cherry Borer (Hypothenemus hampei Ferrari). Again, Wrigley (1988) describes the impacts of this pest. This is a miniature black beetle (2 mm) that causes serious damage to the coffee cherries. The female beetle bores a circular hole at the tip of the cherry and lays its eggs inside it. In periods of heavy infestation, it is usual for more than one female to lay its eggs in the same coffee cherry. The eggs hatch in five to nine days. The larvae feed of the bean during their larval stage (10-26 days). As the larvae grow, they destroy one or both of the seeds, and the cherry appears unharmed to the naked eye. The damage can only be assessed once the coffee cherry is harvested and processed, when it becomes evident that the coffee beans are ruined. Brief History of Coffee Production: Shifting Centers of Production Early Beginnings in Africa and Arabia Because the wild coffee plant is indigenous to Africa, one should start its history there. The reader should keep in mind that this story refers mostly to Arabica coffee, as Robusta coffee only become important after the Second World War. As stated before, this species is indigenous to Ethiopia (Chalarc 1998; Dicum and Luttinger 1999; Smith 1985; Wrigley 1988), and this is where the story begins in prehistoric times. Coffee has long been used by the mountain peoples of Ethiopia, but not as a beverage. Coffee beans are crushed and mixed with animal fat. This mixture is shaped into small balls that are used as energy food for long trips and warfare (Dicum and Luttinger 1999). In the western culture, there are many legends surrounding the discovery of coffee (Dicum and Luttinger 1999; Smith 1985). Perhaps the most widely known is the story of a goat herder
87 that noticed how his goats became excited after eating the cherries of a bush. He tried the cherries himself and felt his tiredness dissipate, and his energy soar. Other stories connect the goat herder legend with a monk from a nearby monastery that, when hearing the herder story, collected the cherries, pulped them, roasted and brewed the beans and produced a beverage that allowed monks to stay awake for prayers during long periods of time. Some legends even tie the discovery of coffee to Islamâ€™s prophet Mohammed. In a dream, the Angel Gabriel came to Mohammed showing him the cherry and telling him about the coffee cherryâ€™s potential to stimulate the prayers of his followers (Dicum and Luttinger 1999).The first tangible historical evidence about coffee production and consumption as we now it today is dated 850 AD, where Arabica coffee was cultivated in the Arabian colony of Harar (present day Ethiopia), and eventually spread to Mecca and Yemen (Smith 1985). During the first few centuries among the Arabs, coffee was regarded more as a medicine than just a regular beverage. The first written account of coffee comes from an Arab philosopher and astronomer called Rhazes, who lived in the 10 th century (Smith 1985; Dicum and Luttinger 1999). By the early 16 th century, European travelers had discovered the joys of coffee in the Ottoman Empire. However, production was monopolized by the Arabs in areas of the Arabian Peninsula, particularly in Yemen (Dicum and Luttinger 1999). At this time, Venetian merchants commercialized the beans in Europe, but Arabs maintained a strong control on production. Laws prevented foreigners to visit coffee plantations and coffee beans were heated or boiled before being commercialized to prevent the export of viable seeds (Dicum and Luttinger 1999; Smith 1985). However, this monopoly came to a close in the early 16 th century, when an Indian pilgrim in Mecca was able to smuggle some
88 seeds out of Arab control, and take them back to the Mysore region of India, a region which still produces a good-quality coffee (Dicum and Luttinger 1999; Smith 1985) To the East Indies and Europe Coffee as a beverage started to become popular in Europe in the early 16th century. The Turks introduced it in a massive way to Europe in the 17 th century as they retreated from the siege of Vienna. From here, coffee spread rapidly, and coffee houses opened in all major European cities soon after. To exemplify this rapid spread in coffee consumption one can look at the evolution of English coffee houses. The first coffee house in England was opened in 1650 in Oxford, and by 1715 there were more than 2000 coffee houses in London alone (Dicum and Luttinger 1999). All over Europe coffee houses became the centers for social, commercial, political and literary life (Smith 1985). However, major coffee production was still in the hands of non-European countries. By 1616, a Dutch trader successfully stole a coffee plant that was taken to the East Indies. After the Dutch took control of Ceylon and Java from the Portuguese a few decades later, a major production shift took place. Soon after the Dutch gained control of these islands they established large coffee plantations. The first beans from the East Indies plantations were brought to Amsterdam in 1706 along with a coffee plant to be part of the Amsterdam botanical garden collection (Smith 1985; Dicum and Luttinger 1999). The Dutch were so successful with coffee cultivation in the Far East that they controlled coffee prices for a long time. In 1714, Louis XIV received a coffee plant, descendant from the plant that arrived to Amsterdam in 1706, as a gift from Amsterdamâ€™s major. This bush was taken to the botanical garden (Jardin des Plantes) in Paris. A few years later, in 1723, a naval officer took several plants from the Jardin des Plantes to Martinique (Smith 1985; Dicum and
89 Luttinger 1999). These plants became the â€œmothersâ€ of almost all the coffee in the New World. Soon, other European powers followed the example of planting coffee in their New World colonies. Arrival to the New World There are some authors that suggest, based on evidence on old Chinese maps and toponyms in the island of Puerto Rico, that coffee was brought by the Chinese to the New world during the travels sponsored by admiral Zheng He in the early 15 th century (Menzies 2002). However, there is little evidence that, even if coffee was brought to the New World by the Chinese, it was used by Amerindians in this Caribbean island. The more traditional and believable story tells us that after coffee was introduced successfully to Martinique in 1723, it spread rapidly to the Spanish and Portuguese colonies (Smith 1985; Dicum and Luttinger 1999). The French soon became the major producers of coffee in Haiti, replacing the Dutch as the largest producer of the bean (Dicum and Luttinger 1999). Coffee had also been introduced to the Dutch colonies in South America by 1718. However, it did not develop as strongly as in other places because the Dutch had the control of the world coffee market with their massive plantations in Java. There are also some accounts of coffee being introduced to Haiti in 1715 (Dicum and Luttinger 1999). However, the major milestone to establish its production in the New World was the arrival of coffee plants to Martinique, as mentioned before. Coffee was introduced to Jamaica in 1730, and since then, Jamaican coffee is known as one of the best in the world (Dicum and Luttinger 1999). The arrival of coffee to Brazil and the Spanish colonies in the New World is also a very interesting and colorful story. In 1727, some coffee plants were smuggled out of French Guyana by a Portuguese army lieutenant that was serving as an intermediary in a
90 boundary dispute between the French and the Dutch in the Guyanas, and took them to the state of Par, in the Amazon basin (Dicum and Luttinger 1999; Smith 1985). Soon after, coffee plantations started to emerge in Brazil, but it was not until the early 1800s when this country became an important player in the coffee market. Coffee was also successfully planted by Jesuit missions in the 18 th century in the Eastern lowlands of present day Colombia and Venezuela in the Orinoco River basin (Chalarc 1998). Today is still possible to find feral coffee bushes growing in the riparian forests of these savannas. Coffee was already an important product by the late 18 th century in Brazil, Costa Rica, and Venezuela (Parsons 1968). In the late 18 th century there was another shift in the centers of coffee production, as the repercussions of the French Revolution reached Haiti (Dicum and Luttinger 1999). Because the production of coffee in this island was based on large numbers of slaves, and the French Revolution promulgated the idea that all men were equal and free, a major slave revolt took place in the island. Production never recovered, and the major center of production soon moved to the island of Ceylon, which was then a British colony. In the mid 1860s another center of production shift took place. At this time the Coffee Leaf Rust disease wiped out the coffee plantations of Ceylon, and the major center of production moved to Latin America, particularly Brazil (Dicum and Luttinger 1999; Palacios 1980). In Ceylon, coffee plantations were replaced by tea plantations, which have been very successful since then. Brazil emerged as the major producer in the world, joining Costa Rica and Venezuela, who, by the 1850s, were important coffee producers in the world market (Palacios 1980). Coffee arrived relatively late to Colombia when
91 compared to other Latin American countries (Parsons 1968). However, it soon became a very important export crop. In the 19 th century, the growing demand for coffee, mostly from the expanding U.S. market, transformed the agriculture of many Latin American countries. For example, in Mexico, the switch from traditional cacao and tobacco production to coffee took place mostly as a result of the increasing U.S. demand (Nestel 1995). During the first half of the 20 th century coffee production concentrated mostly in Latin America, with Brazil and Colombia becoming the most important coffee producers in the world, and with Venezuela rapidly transforming from the second coffee producer in the world in the 19 th century, to one of the leading oil producers in the world (Price 1994). Coffee production became a more important activity in Africa and South East Asia as the former European colonies reached their independence and turned to coffee as a means to earn foreign currency. Today, African countries like Tanzania and Uganda, and Asian countries like Vietnam and Indonesia are major coffee producers. Coffee and Its Transforming Power It was mentioned earlier how coffee transformed the social fabric of coffee consuming countries starting in the 17 th and 18 th centuries. The transformations were not restricted to these countries alone. Coffee agriculture had also significant impacts on the societies and landscapes of the coffee-producing countries. For example, the coffee boom in the 19 th century fueled by the increased demand for the beverage in the rich countries of the northern hemisphere transformed the human geography and landscapes of tropical America (Price 1994). Perhaps the most important change that occurred in the coffee-producing countries is that peasant families that were previously producing mostly for subsistence became more market oriented when coffee was adopted (Parsons 1968). Even
92 in areas engaged in large-scale agricultural production, coffee became a crop linking the local level with the international market. Agricultural production became the result of land-use decisions based on local conditions and international markets. At the landscape level, the historic evolution of coffee production was also responsible for major transformations. For example, when production shifted from Haiti to Ceylon in the early 19 th century, it was accompanied by major land conversion. By the time Coffee Leaf Rust disease wiped out coffee production in this island in the late 1860s, about 71,000 hectares of forest had been transformed into coffee plantations (Dicum and Luttinger 1999). When coffee arrived to Martinique in the early 18 th century, it rapidly became a very popular crop. Within the first three years of arrival, there were 3,000,000 coffee bushes planted in the island (Dicum and Luttinger 1999). Coffee has been promoted as a good crop for homesteaders in the long term. For example, in areas of Vietnam (Caspersen 1999), Robusta coffee and rubber have been promoted as a desirable agroforestry system in agricultural frontiers with so much success that this Southeast Asian country is now the second largest coffee producer in the world (FAO 2001). A similar situation where only coffee is promoted, although at a much smaller scale, is present in certain areas of Colombia, where coffee production has expanded in regions of active colonization (Chalarc 2000). Presently, coffee covers more than 5.9 million hectares in Latin America and the Caribbean, 2.9 million hectares in Africa, 2.0 million hectares in Asia, and 66,000 in Oceania (FAO 2001). In Northern Latin America, where roughly 2.7 million hectares are planted with coffee, the establishment of the crop brought major transformations to the native ecosystems of the mid-elevation belts of mountain ranges (Rice 1996). The same author discusses that in the last few decades,
93 coffee-producing landscapes in Latin America have again changed dramatically as a result of the intensification of production. In Brazil, the establishment of coffee plantations was also accompanied by the disappearance of native forests (Wrigley 1988). Institutions in charge of promoting coffee production have also played a very significant role in changing rural landscapes all throughout the world. A good example is the National Coffee Growers Federation of Colombia (FNC), a semi-private organization in charge of improving coffee production in this South American country. Starting in the early 1970s, the FNC began a large-scale campaign to modernize coffee production throughout the country. This involved the intensification of production and the replacement of the more traditional and diverse agroforestry system by coffee monocultures. In Mexico, the Mexican National Coffee Institute (INMECAFE) also promoted a similar transformation, where coffee monocultures were preferred to the traditional system (Rice 1996). This was a response to an increasing demand for coffee. The landscape change was accomplished mostly through technical assistance. The area transformed represents 30% of the coffee growing area in Mexico, demonstrating the major landscape transformations that can result from government organization interventions and policies. Despite of its power to change landscapes, there are relatively few detailed studies describing and analyzing the landscape transformations associated with coffee production. Some of these studies are detailed descriptions of landscape changes in Chiapas, Mxico (Rice 1996), Puerto Rico (Rudel, Perez-Lugo, and Zichal 2000), and Vietnam (Caspersen 1999). Although Colombia has good information sources regarding landscape transformations for three moments in the last thirty years (FNC 1970, 1976,
94 1983, 1997), there has been little effort to analyze the spatio-temporal evolution of the Colombian coffee lands. This dissertation seeks to contribute to this understanding of spatial evolution associated with coffee production. Coffee Production Systems All over the world, coffee is planted under several production systems that differ in terms of their vegetational and architectural structure as well as their management. These systems are part of a continuum that ranges from the more traditional systems, where coffee is planted under the canopy of rainforest trees, to the most intensive, which is essentially a coffee monoculture with no shade at all (Moguel and Toledo 1999; Gobbi 2000). Moguel and Toledo (1999) have broken up this continuum into a limited number of production systems based on their management level and vegetational and structural complexity. Although the original classification was developed for Mexico, it can be easily extended to most coffee-producing areas in the world. In this classification system, there are 5 distinct production systems characterized as follows (Moguel and Toledo 1999): Rustic: understory of tropical forests is cleared and coffee is planted under the original rainforest canopy. Very low use of inputs Traditional polyculture: coffee is planted under the cover of the original forest alongside useful plants. A "coffee garden" is created. Highest vegetational and structural complexity. Commercial polyculture: complete removal of the original forest trees and their replacement with shade trees suitable for coffee cultivation (usually useful species). Shaded monoculture: leguminous trees are used almost exclusively for shade. Monospecific coffee plantation under a specialized shade. Unshaded monoculture: no shade at all. Requires the highest level of inputs.
95 It is important to make some remarks regarding this classification system. As the production system moves away from the traditional polyculture there is a major change. In the first two systems the original rainforest canopy is maintained, while in the last three a human-made forest replaces the original forest. There is also a gradient in terms of biological diversity associated with these systems. While the first two are the most diverse, diversity decreases sharply as a few species of planted useful trees replace the diverse canopy of the rainforest. It appears that the overall biological diversity of the coffee agroecosystem is related to the number of tree and herb species (Moguel and Toledo 1999), as more trees and herbs mean more habitat types for different organisms. The commercial polyculture, shaded monoculture, and unshaded monoculture systems are related to farmers that are more commercially oriented. The other two systems can be found in areas not yet fully integrated into the market economy, or regions where indigenous populations are present. Furthermore, it appears that more traditional production systems (rustic, traditional polyculture, commercial polyculture) are more sustainable in the long run, and at the same time provide a commercial strategy profitable for farmers (Richter 2000). According to Richter, these systems remained nearly unchanged in areas of Chiapas for more than 100 years, and maintained soil fertility through the use of leguminous-shade trees that replenished not only nitrogen, but also contributed large amounts of leaf litter that transformed into soil organic matter. Additionally, the tree canopy and litter layer minimized erosion. Therefore, these systems provided an agricultural production strategy that combined natural-resource conservation with market production.
96 Another aspect that differentiates these coffee agroecosystems is the presence of shade. More traditional agroforestry systems tend to have a more diverse canopy, while more intensive agricultural systems tend to have a simpler canopy or no shade at all. Although using shade by itself as a distinguishing characteristic does not capture the full variability of coffee production systems as defined above, it has been widely used in the scientific literature to distinguish between traditional and intensive coffee production (Niehaus 1992; Ortiz 1989; Palacios 1980; Perfecto et al. 1996; Rice 1997; Rappole, King, and Vega-Rivera 2003; Brown 1996b, 1996a; Dicum and Luttinger 1999; Errzuriz 1986; FNC 2001; Muschler 2001; Baggio et al. 1997; Parsons 1968; Wrigley 1988; Willson 1985b). For the Colombian case this distinction is useful because nearly all coffee production is under the shaded monoculture or the unshaded monoculture as described above. The following sections describe the characteristics of shaded and unshaded coffee production systems, as well as their advantages and disadvantages. The most important characteristics of these systems are summarized in Table 2. Table 2. Coffee production systems characteristics Traditional System Characteristics Intensive System Characteristics Coffee bushes planted under shade trees Low planting densities (<2500 bushes/ha) Traditional (tall) varieties Little use of inputs Requires less strict agroecological conditions Long coffee plot life cycle (>10 years) Coffee bushes planted with little or no shade High planting densities (up to 10000 bushes/ha) Short varieties Very high input use Requires more strict agroecological conditions Short coffee plot life cycle (< 7 years) Another interesting characteristic of coffee production is that, regardless of the coffee production system chosen, there is an increasing trend towards producing this crop in smaller farms (Wrigley 1988). According to UNCTAD (1984) only 1% or 2% of the
97 coffee in the world market is produced in large plantations. According to OXFAM (2002), 70% of coffee is produced in farms smaller than 10 hectares. Wrigley (1988) states that in Brazil there is a significant number of large coffee estates (>10 ha). However, more than 75% of the farms growing coffee have less than 10 ha. In Brazil, Colombia, and Mexico, less than 3% of the growers are considered large-scale operators (> 10 ha in coffee) (Wrigley 1988). This, accompanied by a clearly intensifying trend in coffee production at the global level, suggests that the commercial intensification of coffee production has not been accompanied by land consolidation, as some authors suggest (Pingali and Rosegrant 1995; Hietala-Koivu 2002). Traditional Production System The traditional coffee production system, also known as shade-grown coffee, has recently come to the forefront as an alternative to provide farmers with a profitable economic activity while contributing to biodiversity conservation (Rappole, King, and Vega-Rivera 2003; Gobbi 2000). Traditional coffee plots are usually planted under the canopy of trees with low planting densities. In most instances, traditional (tall) varieties are planted. The management of these plots usually requires little or no use of chemical inputs. The moderating and regulating effects of the shade make it possible to plant coffee in a very wide range of environmental conditions. This production system is characterized by a long coffee-plot cycle, usually more than 10 years. The most important effects of shade on the coffee plot from an agroecological perspective are (Perfecto et al. 1996; Rice 1996; Baggio et al. 1997; Nestel 1995; Willson 1985b; Wrigley 1988): Moderates the temperature range. Reduces insolation rates and winds.
98 Minimizes soil erosion. Reduce frost damage. Shade trees pump nutrients from deep soil horizons. Regulates water availability. Creates habitats. Better nutrient recycling. Trees provide fuel wood and other valuable products. The beneficial effects of shade trees do not include only agroecological aspects of the coffee production system at the plot level. Gobbi (2000) provides a list of scientific studies showing how the traditional coffee production system also supplies some ecosystem services like carbon sequestration, pollination of crops and other plants, and sources of clean air and water. Additionally, this production system implies improved health conditions for the farmers and coffee workers through reduced use of agrochemicals (Gobbi 2000), and farmers can derive significant income from harvesting the shade trees (Rice 2000), so the benefits of shade trees outweigh the detrimental effects caused by the competition with the coffee bushes (Baggio et al. 1997). The presence of shade trees in sub-optimal coffee producing areas in Costa Rica has also been linked with better quality of the produced coffee when compared to unshaded coffee plantations (Muschler 2001). Table 3 presents a summary of the most important advantages and disadvantages of the traditional coffee production system. Shaded-coffee plantations are potential areas for biodiversity conservation. More traditional coffee plots benefit from the structural and floristic complexity of shade trees, resembling in part the original forest (Perfecto et al. 1996; Gobbi 2000). This is also very important in Latin America, where most natural habitats in the coffee producing areas have disappeared, and shaded-coffee plantations can provide suitable habitat to forest-dwelling species (Perfecto et al. 1996). However as the production system changes from
99 traditional polyculture, to commercial polyculture, to shaded monoculture, the shade becomes simplified (i.e., less diversity of tree species) (Moguel and Toledo 1999). The potential for conservation drops dramatically as the shade becomes less structurally complex and floristically diverse (Gobbi 2000). Table 3. Traditional coffee production system advantages and disadvantages Advantages Disadvantages Provides a more structurally diverse ecosystem Reduces the need for chemical inputs and pumps nutrients from the soil Moderates temperatures in the coffee plot Trees usually provide other useful products Higher levels of biodiversity when compared to sun-grown coffee Hydrologic regulation Appropriate for small, medium, and large farmers Replacement of native forest by useful trees. This destroys natural habitats Shaded environment suitable for certain pests and diseases, and some human diseases If shade species are very few, this system is not that different from intensive coffee production The shaded coffee production system also has some disadvantages. In first place, coffee yields tend to be low in the more traditional systems, which are also those with most of the benefits mentioned so far. Additionally, the moderating effects of shade trees provide a more suitable environment for pests and diseases such as the coffee leaf rust (Willson 1985b) and some human diseases (Parsons 1968). It is important to bear in mind that the traditional coffee production system never replaces the native forest because it does not provide the same ecosystem services as the natural habitats (Rappole, King, and Vega-Rivera 2003). Furthermore, the commercialization of coffee production has meant the replacement of the rainforest canopy by anthropogenic forests, still destroying the natural habitats. In recent years, the emergence of organic markets has meant that the traditional production system is becoming a profitable option for small farmers. Consumers are
100 willing to pay a premium price for coffee produced in an environmentally-friendly way, assuring the conservation of certain habitats and an improvement in the standard of living of small coffee farmers that receive more for their coffee (Brown 1996b; Gobbi 2000; Pulschen and Lutzeyer 1993). Countries like Mexico (Moguel and Toledo 1999) and El Salvador (Gobbi 2000) have moved to this specialized market very successfully. The Intensive Coffee Production System The intensive coffee production system, also known as sun-grown coffee, is a system that emerged in the late 1960s in several Latin American countries, particularly Colombia and Costa Rica (Gabriel Cadena personal communication). The development of this new production system was based on the threat of the arrival of Coffee Leaf Rust to certain areas of the continent (Brazil, Central America), because, as it was already mentioned, this disease was responsible for wiping out the coffee plantations in Ceylon in the late 1860s (Dicum and Luttinger 1999). The basic characteristics of the new production system are the reduction or elimination of shade, the replacement of the traditional, tall coffee varieties, for dwarf varieties that can be planted at higher densities, the heavy use of agrochemicals, which guarantee very high levels of production, and the ease of management for certain pests and diseases (Gobbi 2000; Moguel and Toledo 1999; Nestel 1995; Perfecto et al. 1996; Palacios 1980; Niehaus 1992; Rice 1996). Other important characteristics include a shorter coffeeplot cycle, and it can only be adopted in a limited range of environmental conditions when compared to the traditional System. Table 2 summarizes the main characteristics of this production system in comparison to shade-grown coffee. Although the idea of planting coffee under direct exposure to solar radiation is not new, the intensive coffee production system has been labeled as sun-grown coffee.
101 Historically, coffee has been grown under direct sunlight in many areas of the world and different time periods. The plantations established by the Dutch in Java did not have shade trees (Wrigley 1988). A similar situation was present in Ceylon before the collapse of its coffee plantations caused by the Coffee Leaf Rust disease (Wrigley 1988; Willson 1985b; Dicum and Luttinger 1999). In Latin America, Brazilian production is mostly sun grown, with shade only present in a certain areas as a mechanism to protect the crop from frosts (Dicum and Luttinger 1999; Baggio et al. 1997). The production of coffee in Venezuela in the 19 th century was under direct sunlight (Price 1994). According to Price (1994), the coffee production pattern in this South American country was characterized by a slash-and-burn agriculture, where coffee was planted after a plot of Andean rainforest was cleared, and when the soil was exhausted, the farmer moved to a new location. Only as land became more scarce, permanent coffee plots became prevalent. At the end of the 19 th century these permanent sun-grown plots became less important as Venezuela shifted to oil as its most important product (Topik and Wells 1998; Parsons 1968). Also in 19 th century Guatemala, Colombian entrepreneurs promoted sun-grown coffee as the most desirable production system (Parsons 1968). In Colombia, some of the earliest plantations in the provinces of Cauca and Santander did not have shade trees (Palacios 1980). Therefore, what is new in the intensive-production system is the use of dwarf coffee varieties, the extensive use of agrochemicals, and higher planting densities. One of the most important implications of the intensive coffee production system is that plot renovation to maintain a high productivity is done at higher frequencies than in the traditional coffee production system. This means that in the traditional production system the landscape changes more slowly, because the plot cycle is much longer.
102 Therefore, landscapes where traditional coffee production takes place tend to be more stable. Another important implication of coffee production intensification is that there is more output per unit of land. As a result, the same, or even more coffee can be produced in less land, making land available for other land uses. Table 4. Intensive coffee production system advantages and disadvantages Advantages Disadvantages Easier to manage certain pests and diseases Higher yields As coffee production intensifies, land becomes available for other land uses Soil is more exposed. Higher erosion Higher runoff rates It does not yield well without heavy fertilization Better suited for medium and large producers The intensive coffee production system also has some major disadvantages. Table 4 presents the advantages and disadvantages of this system. Among the most important disadvantages is the increased erosion potential. The lack of tree canopy prevents the formation of a thick layer of leaf litter that buffers the ground surface from the elements, leading to increased soil erosion and higher runoff rates when compared to the traditional production system (Richter 2000). In a study comparing an advanced secondary succession plot, a coffee plot shaded with Inga sp., and a coffee plot under direct sun light, Jaramillo and Chves (1999) found that the amount of effective rain (i.e., percentage of the rainfall event that reaches the ground) was highest in the sun-grown coffee, followed by the secondary succession plot, and the lowest value was for the shade-grown coffee. That the plot in the intensive production system has the largest effective rainfall is expected, because the amount of effective rainfall is inversely proportional to the amount of foliage (Jaramillo and Chves 1999). Therefore, because more rain is reaching the ground, there is increased erosion potential in the intensive system. Additionally, this production system has lower yields than shaded coffee when
103 no fertilizer is used (Willson 1985b). This makes coffee farmers highly susceptible to price swings in agricultural inputs and coffee, making it difficult for them to buy all the fertilizer needed if coffee prices are low. This obviously decreases the productivity of the intensive coffee production system. The dependence on industrial inputs makes this production system very dependent on capital, therefore more suitable for medium and large producers (Nestel 1995). Some authors suggest that this fact has lead to land concentration in coffee-producing landscapes where the intensive system becomes prevalent (Nestel 1995; Ortiz 1989). The evidence from Colombia suggests that land fragmentation has taken place alongside the intensification of coffee production (FNC 1970, 1983, 1997). Although initially the switch to the intensive production system was based on an improved protection and easier management of Coffee Leaf Rust, the anticipated impact of this disease was overemphasized (Perfecto et al. 1996). Very early on, the most important reason for the switch to the intensive system became its higher productivity rather than its ease to control pests and diseases, and it continues to be so (Perfecto et al. 1996; Gobbi 2000). In northern Latin America, there is an increasing trend to replace shaded-coffee plantations with sun-grown coffee plots. By 1990 it was estimated that half of the coffee area in Latin America had been converted to the intensive system (Perfecto et al. 1996). However, the degree of this transformation changes dramatically from country to country. While, by 1996 it was estimated that 70% of the area planted in coffee in Colombia had been transformed to the intensive system, in countries like El Salvador 90% of coffee production was still carried out with the traditional system (Rice 1996).
104 Preparing the Coffee Cherry for the Market Regardless of the production system chosen, the end product is the same: coffee cherries. The harvesting of the coffee cherries can be in two ways: 1) collecting just the ripe cherries, and 2) collecting all the cherries in a coffee bush (Dicum and Luttinger 1999; Wrigley 1988). In the first method, ripe cherries are hand picked in an extremely labor-intensive process. The resulting coffee has higher quality than the other collection method. However, labor costs are very high as workers have to carefully screen each coffee bush branch for ripe cherries. Furthermore, as not all the cherries ripen at the same time, harvesting coffee usually involves workers collecting cherries several times in the same coffee plot during the harvest period. In the second method, the worker collects all the cherries in any branch. This results in a mixture of ripe and green cherries that produces a poor tasting coffee if too many green cherries are collected. In this harvesting method, the worker only collects cherries once in each bush. Additionally, machinery has been developed to harvest coffee this way, reducing enormously the labor costs associated with the collection of the ripe cherries by hand. Once the coffee cherries are collected they need to be further processed before they can be taken to the market. At this stage, farmers have two options: 1) the wet-processing method, and 2) the dry-processing method (Dicum and Luttinger 1999; Wrigley 1988). Both methods result in different tasting coffees. The flavor of the coffee beans produced through the wet-processing technique tends to be the preferred coffee taste. In this method, coffee cherries are depulped with a mechanical depulper, and then they are put in large tanks with water for up to 36 hours so fermentation can take place. The fermentation of the coffee beans adds special qualities to their flavor, and removes the mucilage that is left on the coffee seed after depulping. Once the fermentation process is
105 finished, the coffee beans are washed, and set in large flat areas to dry. This method produces enormous amounts of organic matter that are usually dumped into water courses, causing serious water pollution (Dicum and Luttinger 1999). In recent years, the discarded pulp is composted and used as fertilizer, and new methods that use less water, like the BECOLSUB system in Colombia, have been developed (FNC 2001). In the dry-processing method, the cherries are set to dry under the sun, and once they are dry, the coffee beans are extracted and the dried pulp discarded. The wet system is the most common around the world, and the dry system is used in Brazil, Indonesia, and Angola among others (Dicum and Luttinger 1999). After the coffee beans are dried, there is still one more step required before they are ready for the market. The coffee seed has a hull that has to be removed. At this stage coffee is called parchment coffee. The coffee seeds are hulled and the resulting coffee beans, now called green coffee, are ready to be marketed. Coffee in the International Market There are certain aspects that are unique to the coffee market. Coffee is a crop that is very sensitive to environmental conditions, and as a result, events such as frosts in Brazil, or drought or excessive rains in other areas of the world have profound effects on the supply, making the market very unpredictable (Wrigley 1988; Dicum and Luttinger 1999). In addition to this situation, coffee production has been historically controlled by one or a few countries which tend to monopolize production, and to a certain extent control the market price of this commodity (Wrigley 1988; Smith 1985). Since the 19 th century, Brazil has been the major coffee producer in the world. Its share of the market was more than 75% in the late 1800s (Wrigley 1988), and has dropped to around 25% in 2000 (FAO 2001). The combination of a few major producers with the susceptibility to
106 environmental factors makes the forecast of coffee supply very difficult. For example, a frost or drought in Brazil has major implications on the supply of coffee in the following few years. Because of lower than expected supply, coffee prices rise and farmers all around the world are encouraged to plant more coffee. However, by the time their coffee plots are producing their first crop two to three years after the hike in prices, the market has usually stabilized. The extra coffee produced by the new coffee plots means a higher supply, which lowers prices. Therefore, farmers who expected higher prices when they planted their new coffee plots end up receiving less than before due to low coffee prices, even though they are producing more coffee. This cycle of a few years of high prices caused by a lower supply followed by a longer period of low prices and higher supply is characteristic of the coffee market. This agricultural product, as many other export crops, is characterized by these boom-and-bust cycles (Palacios 1980). Since Brazil emerged as the single most important producer in the late 1800s, it has tried to minimize these market swings and to maintain a stable price (Dicum and Luttinger 1999; Wrigley 1988). The first attempts to increase the price started in the early 1900s in a scheme called valorisao (valorization), in which excess coffee was stockpiled in warehouses with the idea to maintain the supply close to the demand, resulting in higher and stable prices. This system worked for a while. However, the higher prices for coffee encouraged other countries, particularly Colombia, to plant more coffee, therefore increasing the supply of the crop. Also, the amount of stored coffee increased dramatically, and millions of coffee sacks had to be destroyed. These unilateral measures by Brazil were repeated several times until the late 1930s, while countries like Colombia continued to push coffee as a major export crop to take advantage of Brazilâ€™s efforts. By
107 the late 1920s both Brazil and Colombia had established national organizations to manage and market coffee. These institutions, the Instituto Brasileiro du Caf and the Federacin Nacional de Cafeteros de Colombia, have played a key role in the management of the coffee market, the spread of the intensive production system in Colombia and Brazil, and the creation of the International Coffee Organization in the early 1960s. (Dicum and Luttinger 1999). During the economic depression of the 1930s coffee demand dropped dramatically. By that time, the U.S. accounted for about 80% of the world demand for coffee (Wrigley 1988). As conditions deteriorated in the U.S., people stopped drinking coffee, and Brazil, the major supplier of the U.S. at the time, was forced to destroy 78 million sacks of coffee, equivalent to the world supply of 2.5 years at that time (Wrigley 1988). After the depression, coffee prices recovered somewhat, and another major blow to demand set on the coffee market: the Second World War. During this period, the European markets were closed, reducing the demand for coffee. As a mechanism to prevent Latin American countries to sympathize with the Axis powers, the U.S. signed an agreement with the Latin American producing countries ensuring a constant demand for coffee in the U.S. This guaranteed a relatively stable period of coffee prices. After the war, coffee prices increased rapidly because the European demand recovered. However, there were major problems to reactivate production, leading to higher prices. The main reasons for this were that in some of the countries affected by the Japanese occupation, particularly the East Indies, plantations had been heavily damaged, and Latin American producing countries had not increased their area in coffee as a result of the limited international demand during the war. Frosts in Brazil in 1953 exacerbated this increasing-price trend
108 (Dicum and Luttinger 1999), but prices soon decreased as new producers, mostly former colonies in Africa and in the far East, entered the market. During the mid 1950s, many of the Latin American countries tried to form an economic cartel to withhold coffee from the market, and maintain prices artificially high (Wrigley 1988). These early strategies in which only the producing countries were engaged in controlling the market price for coffee did not work, but set the stage for the next major change in the coffee market: The International Coffee Agreement. In the early 1960s, falling coffee prices started to become a significant issue for the U.S. government. Because lower prices meant lower incomes for coffee growers, this situation, according to the Kennedy administration, made it more feasible for Marxist ideologies to leak to the Latin American republics (Dicum and Luttinger 1999). As a result, the U.S. government decided, as the major consumer in the world, to sign an agreement with the coffee producing countries to try to balance supply and demand, and keep prices close to an agreed upon level by both consumers and producers. By 1962, the International Coffee Agreement (ICA) was signed and the International Coffee Organization (ICO) was founded to administer it. This situation prevailed until 1989. According to this agreement, each producing country had a quota that could be traded with the consumer country members of the international coffee cartel. Any excess coffee in the producing countries should be stored in warehouses, or it could be sold to non-ICA countries at lower prices. The quotas for each one of the producing countries were renegotiated every 5 years. The agreement was also designed so as to give more power to the largest producers, Brazil and Colombia. Therefore, smaller producers and newcomers did not have a voice and the quota system reflected largely the supply of the largest
109 producers, putting smaller producers at a disadvantage. The agreement defined the prices for a period of 27 years with minor interruptions caused by extremely damaging environmental phenomena, like the major frosts in Brazil in the mid 1970s, which changed the supply conditions. During these interruptions, coffee supply dropped dramatically, leading to higher coffee prices. As these events unfolded, the quota agreement was suspended, and coffee producing countries were able to take advantage of free market trade (i.e., no quotas), higher prices, and reduce their stored coffee stocks. The maintenance of artificially high prices did not deter producing countries to expand or intensify their coffee production (Wrigley 1988). Additionally, newcomers like Laos and Vietnam planted large areas in coffee, contributing to the already over-supplied world market. By 1989 the parties of the ICA could not reach a new agreement, and free-market conditions started to prevail. As producing countries struggled to sell as much as possible of the coffee stocks stored in warehouses, coffee prices plummeted to the lowest price in decades, and made a major impact in countries whose major foreign exchange source was coffee. By 1993, the conditions had changed, and the U.S. was no longer interested in preventing Marxist regimes in Latin America. The economic policies promoted by the Reagan administration dictated that the free market should set the price for commodities. The situation has remained more or less similar since then, and excepting a few short-lived price hikes associated with frosts in Brazil and other environmental events, prices continued to be extremely low. Current coffee prices (2003) are at their lower levels in the last hundred years. During the last meeting of the ICO there were some talks to try to revive the quota system in order to increase prices, and prevent further degradation of the coffee farmersâ€™ standard of living. A recent letter from
110 the U.S. Congress International Relations Committee asks the Secretary of State to reactivate the quota system, so it would provide an acceptable income from coffee production, and prevent coffee farmers in Colombia and elsewhere from planting illegal crops (ICO 2003b). The low price of coffee in the international market has prompted coffee farmers to look for new alternatives. The emergence of the organic and specialty markets in the United States and the European Union provides good opportunities for coffee producers to improve their income. (Brown 1996a, 1996b; Gobbi 2000; Pulschen and Lutzeyer 1993). In the developed world, there is an increasing number of consumers that are willing to pay a premium price if their coffee is produced in certain way, or if it satisfies certain quality requirements. The certification of coffee producers is one of the alternatives to reach a higher price in the international market. However, specialty and certified coffees are an option not suitable for the 25 million coffee farmers in the world (OXFAM 2002). There are many types of certified coffees, each one with different certification criteria. Examples include Bird friendly (SMBC 2003), Conservation coffee (CI 2003) and ECO-OK (RA N.D.) among others. These certification programs guarantee that coffee is produced in certain conditions. Similar programs that also raise the income of coffee farmers are the Fair-Trade coffees. As it will be shown in the next section, coffee farmers receive a very small portion of the retail price in the developed world markets. The idea behind Fair-Trade coffee is to pay the coffee farmer a fair price for his efforts, and reduce the earnings of the middlemen and roasters, who traditionally have monopolized the largest share of the profits. Although well intentioned, these options do
111 not really address the main problem of the coffee marketing chain: the concentration of the profits on the roasters and distributors in the developed world (OXFAM 2002). Farmers who become part of these programs do obtain higher prices for their coffee. While a coffee farmer receives on average about 6.5% of the coffee retail price in the developed world (OXFAM 2002), certified coffee farmers can receive up to 25% of the retail price in the U.S. and Canada (LGT 2003). However, they are still receiving a very small fraction of the retail price. The Marketing Chain From the time a ripe coffee cherry is picked from the coffee bush to the time it is drank in a cup, the coffee bean has switched hands many times. It has also increased in price several times. In addition to lower market prices in recent years, there has been a shift of much of this value to the distributors and roasters in the developing world. While in 1985 $0.38 of each dollar spent for retail roasted coffee in the U.S. reached a developing country, by 1995 only $0.23 made the trip back to the producing country (Dicum and Luttinger 1999). Although relatively little money gets back to the coffee producer, this crop is better at bringing capital through export earnings to the developing world than other primary commodities like tea, sugar, cocoa, bananas, oranges, cotton, and tobacco among others (Dicum and Luttinger 1999). For example, Costa Rican black pepper growers in the San Carlos area are paid $1.25 per pound, while the retail price for this product in the U.S. is about $1.00 per ounce (Jones N.D.). This means the producer is getting only about 8% of the retail price of a crop that requires little processing after being harvested. Because of different arrangements in the coffee producing countries, it is difficult to generalize how coffee marketing takes place. However, there are some general links
112 that all coffee goes through as it travels from the farm to the cup of the consumer (Wrigley 1988). Figure 3 presents a simplified Coffee marketing chain as developed by the United Nations Conference on Trade and Development (UNCTAD 1984). It clearly shows how there are many intermediaries between the grower and the consumer. At each one of these steps one or more intermediaries may be present. For example, in the case of Mexico, the marketing chain contains several middlemen between the grower and the exporter (Waridel 2002). In countries like Colombia, the marketing chain is much simpler. The grower usually sells directly to the National Coffee Growers Federation, a semi-private organization in charge of coffee marketing (Broker and Importer/Exporter in Figure 3 ), although the farmer is not prevented from selling it to other private exporters (Palacios 1980). Dicum and Luttinger(1999) present a marketing chain that tries to incorporate the regional differences just described. However, for the sake of simplicity, it is better just to concentrate on the major steps that apply to all coffee marketing chains. The most important effect of this marketing chain is that coffee prices increase dramatically from the grower to the consumer. Figure 3 presents the case of Ugandan Robusta coffee that is processed for instant coffee in the United Kingdom (OXFAM 2002). The coffee grower only receives 0.5% of the retail price. The broker pays $0.07 for bagging and transporting the coffee to Kampala, and takes a profit of $0.05. Transportation costs, insurance, and selecting the best coffee beans cost the Exporter/Importer from Figure 3 $0.37. When the coffee gets to the instant coffee factory, its value skyrockets, and the roaster and retailer pocket more than $24.00, which cover more than enough processing, packing, and distribution costs. The price of a pound of coffee inflates between 4000% and 7000% from the moment it leaves the coffee farm
113 until it reaches the consumer, with most of the profits staying in the hands of the roasters and retailers (OXFAM 2002). ConsumptionDistributionTransformationIntermediationProduction Coffee importing countryCoffee producing country Grower Broker* Roaster Retailer Consumer Exporter/Importer* Roaster Retailer Consumer * Private or governmental agencies Price paid ateach level$0.14/kg$0.26/kg$1.64/kgN.A.$26.40/kgConsumptionDistributionTransformationIntermediationProductionConsumptionDistributionTransformationIntermediationProduction Coffee importing countryCoffee producing country Grower Broker* Roaster Retailer Consumer Exporter/Importer* Roaster Retailer Consumer * Private or governmental agencies Price paid ateach level$0.14/kg$0.26/kg$1.64/kgN.A.$26.40/kg Figure 3. Coffee marketing chain. Price information for average Ugandan Robusta coffee in the U.K (OXFAM 2002). Dicum and Lutinger (1999) present an interesting graphic breaking up how much of each US dollar spent on a cup of coffee in the United States went to each part of the marketing chain in 1998, when coffee prices were much higher. According to this graph, only 5 of every dollar are paid to the coffee grower, and 67 are the value added (i.e., roasting, grinding, packaging, trucking and profits) for the roasters and wholesalers in the U.S. Even if half of the value added in the producing country is profits, it still represents more than 6 times the amount paid to the grower, who tended and cared for the crop. It should be fair that the farmer, who is the one who spends more time and invests more effort in producing the crop, got a higher share of the market value. Some producing countries have tried to add value locally to coffee by exporting instant or roasted coffee.
114 Brazil and Ivory coast have large industrial plants for producing instant coffee, and they are successful exporting this product (Dicum and Luttinger 1999; Wrigley 1988). Colombia has been also experimenting with instant coffee production that is exported to many countries in the world, and canned coffee drinks for the Japanese market (Gabriel Cadena personal communication). This unfairness to the producer is very striking, but still, coffee is the export crop that does a better job in getting money to the producer in the developing world (Wrigley 1988). Even if the coffee prices in the international market oscillate dramatically, the consumer in the developed world will not notice anything, because the retail price changes very little. Large multinational companies like Nestl, Sara Lee, Kraft, and Proctor and Gamble pocket the most of the difference between the low prices paid to growers and the high prices charged to consumers in the developed world. Breaking this unequal distribution of profits and improving the standard of living of coffee farmers is the major challenge facing the coffee market right now, but unfortunately it is not likely to change in the near future.
CHAPTER 4 THE STAGE: COFFEE IN COLOMBIA Since the mid 19 th century, coffee has been one of the most important products of the Colombian economy. In 1870, coffee represented 17% of the legal exports (Palacios 1980). Since 1890 coffee became the most important Colombian export crop (Escobar and Ferro 1991) By 1897, coffee was responsible for 47% of the legal export earnings of the country (Palacios 1980). In 1970 coffee was the most important export, accounting for 63% of the total legal export earnings (BANREP 2002a, 2002b). By 1990 the coffee were still important, but their relative contribution had decreased to 21.1% of the total legal export earnings of Colombia (BANREP 2002a, 2002b). By 2000, coffee only represented 8.1% of the Colombian legal export earnings. Oil, gold and other primary commodities became important export products in the mid 1990s and since the early 1980s, exports such as flowers, bananas, and some manufactured products have diversified the Colombian exports. Because of its economic importance for the country, coffee has been widely studied from the economic and politic perspectives (Palacios 1980). However, there is relatively little research on the social and environmental impacts of coffee production (Ortiz 1989). This is also the case for landscape evolution of the coffee-growing regions. In general terms, landscape change in Colombia has not been a popular research subject, limited to regional studies describing the evolution of landscapes in a generalized way (Etter and van Wyngaarden 2000). One of the reasons for this is undoubtedly the lack of historical 115
116 information regarding land-use and land-cover change, including the spatial characteristics of landscape evolution. The ideal conditions for coffee production in Colombia are between 1000 and 2000 meters above sea level, with the most productive coffee areas concentrated in the 1200-1800 meter altitude belt (FNC 2001). Because of its location near the Equator, the Colombian rainfall regime exhibits a bimodal distribution that results in two coffee harvesting seasons. The ideal rainfall should be between 1500 mm yr -1 in the higher (colder) locations to 2500 mm yr -1 in the lower (warmer) elevations (FNC 2001). $$$$$$$$BOGOTAMEDELLINMANIZALESIBAGUENEIVAPOPAYANCALIBUCARAMANGA . 0200400600100Kilometers Ideal conditions for coffee production-Elevations between 1000 and 2000 mabove sea levelIn reality, coffee is grown in a muchwider range of environmental conditions-Rainfall between 1500 and 2500mm per year Legend Coffee producing area$Major cities Figure 4. Location and general characteristics of the Colombian coffee lands. Figure 4 presents the location of these ideal rainfall and elevation conditions. Coffee is planted on the slopes of the Andean ranges, usually in very steep slopes. It is worth noting that the coffee belt is located in the temperate elevation range of the three
117 Colombian mountain ranges. Coffee is rarely the only economic activity in which farmers engage in their land. This results in a very patchy landscape, where coffee plots are interspersed with pasture, vegetable production, and riparian vegetation among others (Figure 5). Figure 5. Coffee growing landscape (La Siria, Manizales-Chinchin Road, May 2000) For 2002, Julin Garca (personal communication) estimated that there were roughly 775,000 hectares planted with coffee, of which about 613,000 were under production. Although this area is not very large, the effects of coffee on rural landscapes in Colombia are far reaching. For the period 1993-1997, when the last large-scale coffee survey was carried out by the National Coffee Growers Federation (FNC), the coffee producing landscapes occupied more than 3.6 million hectares, with coffee occupying about 870,000 hectares (FNC 1997). This clearly emphasizes the fact that coffee is almost always accompanied by other land-use practices. At the same time, roughly 4 million people, about 35% of the Colombian agricultural workforce at the time, derived their livelihoods from coffee (FNC 1997).
118 Despite the fact that the majority of the profits in the primary commodity chains from the developing to the developed world stay in the latter, coffee has brought prosperity to many farmers in Colombia (Chalarc 2000; Arango et al. 1998). The rural living conditions of the Colombian coffee lands tend to be higher than those of the rest of the countryside. The Quality of Life index (ICV) developed by the Colombian Statistical Agency (DANE) summarizes the standard of living of people based on access to basic infrastructure, living conditions, and access to health and education among others. The ICV ranges from 0 (none of the basic needs are met) to 100 (all basic needs are met) 1 . The coffee growing municipalities have a rural ICV value of 56.3 whereas the rest of the rural areas of the country have an ICV value of 53.4 (Sarmiento et al. N.D.). Although the difference does not appear to be that high, it is statistically significant (Mann-Whitney U test, p<0.001). The reason for these better living conditions in the coffee growing areas is that the National Coffee Growers Federation has invested a large portion of the coffee-export earnings to improve the standard of living of coffee farmers. This has been accomplished through the construction of roads and water distribution systems, improved education, and health services among others (Arango et al. 1998). For most of the 19 th century, coffee production in Colombia was associated with relatively large haciendas (Palacios 1980). However, in the late 19 th century and much of the 20 th century, coffee production became characterized by smallholder operations (Palacios 1980; Parsons 1968; Escobar and Ferro 1991; Chalarc 1998). As it will be shown in a later chapter, there are regional variations in the size of the coffee farm and the area of the farm devoted to coffee. What is evident is that coffee production has 1 For a more detailed description of the ICV index refer to the Methods and Data Sources Chapter.
119 become more and more a commercial strategy for farmers, and as a result intensification has been taking place. Nestel presents evidence from Colombia showing that if the price of coffee does not fluctuate drastically, households will slowly commercialize by adopting innovations, in this case the intensive coffee production system. Furthermore, the same author suggests that if coffee production does not satisfy the minimal consumption and reproduction requirements of the household, farmers may switch to other crops and reduce the amount of energy and assets devoted to coffee. The trend towards smaller, intensive coffee production farms suggested by Wrigley (1988) at the global level is also present in Colombia, where the number of farms since 1970 has more than doubled, and the area under intensive production system represents about 70% of the total area planted in coffee (FNC 1970, 1976, 1997). Arrival and Spread of Coffee in Colombia Although coffee was first planted in the Jesuit missions of the Eastern lowlands of Colombia in the 18 th century, it did not spread to other parts of the country (Chalarc 1998). This was probably because the Jesuit order was expelled from the Spanish empire by Charles III during the second half of the 18 th century, bringing the Jesuit missions to a sudden end. Coffee arrived to other areas of the country and became an important crop only in the 19 th century, which is relatively late when compared to other Latin American countries (Parsons 1968). The spread of coffee was relatively slow at the beginning, but accelerated during the 1870s. A timeline of arrival to different areas of the country has been developed piecing information from different bibliographic sources and is presented in Figure 6 (Carriker 2001; Escobar and Ferro 1991; Chalarc 1998, 2000; Palacios 1980; Parsons 1968).
120 0180360540720Km. 18th Century 1860s1820s1840-18601870s1880-1910 Early 1900s Antioquia Boyac Caldas Cauca Cesar Cundinamarca Guajira Huila Magdalena Nario N. Santander Quindo Risaralda Santander Tolima Valle del Cauca Figure 6. Spread of coffee in Colombia Coffee arrived to Colombia from Venezuela to what are today the northeastern departments of Santander and N. de Santander in the first decades of the 19 th century (orange arrow in Figure 6 ). From there, it slowly spread southward along the Eastern cordillera. By 1840-1860 coffee had reached Cundinamarca, Tolima, parts of Huila and Cauca and was mostly a large land holder undertaking (blue arrow). The first coffee plantations in Huila were established in 1862. At the same time, coffee plantations were set up in the Sierra Nevada de Santa Marta in the Caribbean coast of Colombia (black arrow).By 1870, coffee had reached the department of Antioquia, and soon became a very popular crop (red arrow). From 1880-1910 coffee spread southwards towards what
121 is today the most important coffee producing area in the country, the departments of Caldas, Risaralda, Quindo and northern Valle (dark green arrow). Unlike the rest of the country, which was characterized by relatively large coffee-producing estates, this area was characterized by smallholder farmers settling in an agricultural frontier and adopting coffee after establishing their homestead. During the 20 th century, coffee concentrated in this area, and new important centers of production have emerged in Tolima (Errzuriz 1986) and Huila (Chalarc 2000). Coffee is still an important product in most of these provinces, and although associated with the opening of new agricultural frontiers, it has been maintained as an integral product in the farmerâ€™s crops. It is important to emphasize that historically coffee is very rarely the only product in a farm. It is usually planted alongside traditional crops such as corn, sweet cassava, and beans among others (Palacios 1980; Parsons 1968; Rice 1996). This is a more marked pattern in areas where smallholder production prevails, but it is also a strategy used by large holders in order to minimize the risks associated with relying only on a crop like coffee that is very susceptible to major price swings. Another major implication of coffee production is that, when it became an important crop, it dramatically increased the value of hillside land (Palacios 1980). Because the colonial hacienda production system implanted by the Spanish gave preference to flat areas along valleys, and indigenous populations decreased dramatically after the first few decades of European contact, many hillsides were abandoned since early colonial times. In these abandoned lands, secondary succession took place. The establishment of a large number of the Colombian coffee plantations in the 19 th and early 20 th century involved the clearing of these secondary forests (Parsons 1968).
122 Coffee production in Colombia has experienced two major periods of different characteristics. The first period, starting in 1850 and ending around 1970, was characterized by a traditional coffee production system where the expansion of production was the result of the increase in the planted area. This period had two phases of expansion, mostly associated with increased demand in the international market (Palacios 1980). The first stage took place between 1870 and 1910. During this time coffee exports increased five-fold and the major limitation to further development was the poor transportation infrastructure. The second phase of expansion started around 1915 and ended with the beginning of the intensification trend in 1970. This increase in the area planted in coffee was mostly caused by the increasing demand for coffee in North America, and production in Colombia was characterized by small and medium land holders colonizing new lands in what is today the main coffee growing area of the country. The coffee production system during this time period remained nearly unchanged, with the exception of the introduction of the mechanical depulper, which allowed farmers to process their coffee in their farms (Palacios 1980). The second period began in 1970 and continues to this date. This phase is characterized by a switch to the intensive coffee production system, and a decline in the area planted in coffee while maintaining coffee production. In this second period, farmers still have a diverse array of agricultural activities. Despite the fact that farms have many different livelihood strategies, coffee tends to monopolize production (Ortiz 1989). Some authors also suggest that coffee farmers are abandoning this diversity of products and focusing only in coffee (Escobar and Ferro 1991). The evidence from the last coffee survey seems to
123 undermine this argument, as the diversity of crops has increased rather than decreased in the coffee-growing regions (FNC 1997). Models of Landscape Change 1850-1970: The Coffee Economy This distinction between smallholder and large-holder farmers adopting coffee as a crop also resulted in two different models of landscape evolution (Figure 7). While smallholders were establishing their own farm, large land holders were using landless peasant labor to enlarge their coffee plantations (Palacios 1980; Parsons 1968; Escobar and Ferro 1991). However, in both cases the hard labor of opening up the land for production was accomplished through capitalizing the labor of peasant families. The reason for this is that coffee production is ideally suited for the family enterprise, which provides a very cheap source of labor for an extremely labor intensive crop (Parsons 1968; Palacios 1980) Forests Shadecoffee SubsistenceAg. Shadecoffee CoffeeLands Smallholders Largeholders Forests Shadecoffee SubsistenceAg. Shadecoffee CoffeeLands Smallholders Largeholders Figure 7. Landscape evolution paths 1850-1970 Large land holders expanded their plantations using the labor of landless peasant families. Rich land owners provided the landless peasant families with the necessary cash for each family to open up a new coffee plot (Palacios 1980; Escobar and Ferro 1991).
124 The peasant family used this money to buy the supplies necessary to establish a homestead, plant a home garden and subsistence crops, and later on to establish a coffee and shade-tree nursery, and finally begin clearing the forest and transplanting the coffee bushes and shade trees. The first step involved taking care of the farmerâ€™s family subsistence needs: clearing a small plot of land, building a house, and planting the garden and subsistence crops. Once these were satisfied, the attention switched to the establishment of a coffee plot. A nursery for coffee bushes and shade trees was established, and the landless peasant family tended for these plants. Once the coffee bushes and shade trees were ready to be transplanted, a plot of forest was cleared, and the young coffee bushes and shade trees were transplanted. The family provided the labor necessary for all these activities, and continued to do so until the coffee bushes produced the first crop. At this point, the large land holder took control of the land, and paid the landless peasant for the improvements (i.e., the home garden, other crops) present in the plot, and the landless family moved to a new area of the hacienda to establish another coffee plot. Since the time the landless family began establishing a coffee plot until this first crop, the process took approximately 5 years (Palacios 1980). A constant stream of landless peasant families coming from the overcrowded highlands provided a steady supply of cheap labor for the large land-holder coffee producers. In other areas, like the Sierra Nevada de Santa Marta, which was a colonization frontier, no labor from landless peasants was available, and hacendados were forced to import laborers from other areas of the country, and build houses and infrastructure in the hacienda before starting the coffee plantation (Carriker 2001). In this situation, workers had their home gardens and subsistence crops close to their houses, and the establishment of coffee plots was not
125 associated with a landless family clearing land, but rather by paid, permanent hacienda employees planting new coffee plots. However, this situation was less prevalent in Colombia. Although coffee provided good economic opportunities for these large land holders, it forced them to be indebted for long periods of time (Palacios 1980). Establishing a coffee estate was a very risky business for large holders because the returns of the investment only began after 5 years of the initial investment, the coffee prices in the international market were volatile, and large sums of money were needed to establish the coffee plantations and associated processing infrastructure. Despite the fact that the smallholder-farmer landscape evolution path shares some of the characteristics of the establishment of large coffee estate, it has an intermediate step that makes it very different. The best source of information on this landscape change model comes from Parsonsâ€™ (1968) study of the colonization of the temperate altitude belts of Antioquia, Caldas, Risaralda, Quindo and Valle departments in Colombia in the late 19 th and early 20 th centuries. This large area of colonization shares cultural and historical ties to Antioquia, and currently keeps strong links to this area. In these regions, a colonization frontier advanced southward into areas that had been uninhabited since early colonial times, and that by the 19 th century were covered with old secondary succession. Only a few and small pockets of indigenous population were displaced in this process. As the colonization front moved south, families established their homesteads. Because the first goal of these peasants was to obtain a title to their land and not to establish a coffee plot, the first activities involved clearing a plot of land, building a house, and planting subsistence crops. It is important to mention at this point that coffee is not a suitable crop for homesteaders, as it takes up to 5 years to start producing, and it
126 does not address the basic subsistence needs of the family in areas that are probably isolated from markets (Parsons 1968; Palacios 1980). A similar situation has been reported by Rice (1997) for landless smallholders in Chiapas, who plant subsistence crops after colonizing the rainforest of this Mexican region. They bagan planting coffee when the Mexican Coffee Institute arrived to the area, and provided a suitable environment for the production of this commercial crop through incentives and technical assistance. Going back to the Colombian case, obtaining the title of the lands where the colonization was taking place was somewhat difficult because of conflicts with large land owners also claiming the ownership of these public lands. However, in the long run the smallholder farmers prevailed, and farmers were usually given 5 hectares per family member in the recently founded towns (Parsons 1968). There were strict rules in place to prevent land concentration, and land was not given to families with more than 30 hectares. Once subsistence was taken care of, other plots of land were open for cultivation of maize for raising pigs, sugarcane for aguardiente and brown sugar, and pasture for the mules and horses necessary to take these products to the mining regions where they were required. Because most of these farms were located on very steep slopes, crops grew well only for a few years, as soils were rapidly exhausted and eroded. At this point, farmers cleared a new plot of land, converted the exhausted plot into pasture, and it eventually was let to fallow. After about 5-7 years, the fallow plot was cleared, and crops planted again. This slash-and-plant cycle was prevalent throughout the landscape, and soon there were very few forests left. Additionally, soils had been dramatically eroded due to their exposure to the elements with crops that provide little
127 ground protection like maize and cassava. Soon, there were major shortages of fuel wood and construction materials. At this point, coffee appeared as an alternative and dramatically transformed these landscapes. Coffee became a very attractive for smallholder farmers because it provided a high-valued crop that once processed could be safely stored until it reached a market, offered permanent ground cover protecting the soil from erosion, and was associated with shade trees that solved the problem of fuel wood and construction material shortages. Shade trees also provide other products that can represent up to 20%-30% of the small farmâ€™s income (Rice 2000). The traditional coffee production system offered an environmentally-friendly alternative very well suited to the sloping terrain of the Colombian Andes (Parsons 1968). This production system transformed the land from a mostly deforested agricultural landscape to a landscape that looked like a forest, even though it was a man-made forest with very few shade-tree species (Parsons 1968). The labor in these small coffee farms was supplied mostly by family members. Even though shaded coffee appeared as a more environmentally-friendly option than shifting cultivation on the Andean slopes, coffee production exhausted soils in certain areas after a few harvests. These regions were later transformed to pastures, and the arrival of African grasses helped to reduce erosion and restore damaged lands (Parsons 1968). This point highlights that, although more friendly to the environment, traditional coffee production system is not suited for certain agroecological conditions. It is interesting to note that despite the fact that most coffee producers are smallholder farmers, their contribution to production has decreased as time goes by. While in 1932 smallholder farmers represented 98% of the farms and accounted for 60%
128 70% of the total coffee production, by 1970, smallholder farmers represented 71.3% of farms and were responsible for only 29.5% of the total production (Palacios 1980). The reason for this is probably that smallholder farmers did not respond in the same way as medium and large coffee producers to the effects of the International Coffee Agreement. As discussed in the previous chapter, this marketing treaty helped to maintain stable and high prices by trying to balance supply and demand. Higher prices acted as incentives for farmers with more capital and land to concentrate their resources in coffee production, and disintensify other products. Although smallholder farmers were also attracted to these price incentives, they did not expand production at the same scale, mostly because they lacked the necessary capital and labor was constrained family labor. As a result, middle and large coffee farmers, although fewer in number, put more resources per unit of land, and were producing more coffee than small producers, accounting for the drop in the latterâ€™s contribution to total production. This trend towards a higher reliance on coffee was also evident at the national level. While world coffee demand grew by 2.2% per year between 1915 and 1970, coffee production grew 4.2% annually in Colombia for the same time period (Palacios 1980) Contraction and Expansion in the Coffee Economy Export-oriented economies are usually cyclic, very dependent on the rise and fall in the income earned through exports (Palacios 1980). Colombian entrepreneurs were well aware of this fact, as they participated in major export cycles in the 19 th century for products like tobacco and indigo. However, coffee was the first product that was successful in withstanding the negative effects of the low-price period of the cycle. Coffee entrepreneurs had learned from previous experiences the need to diversify in order to minimize the risks associated with the market troughs.
129 For most of its coffee-production history, Colombia has depended very heavily on coffee exports and a single major buyer, the United States, as the main source of export revenues (Palacios 1980). This made the country very susceptible to changes in consumption patterns in this North American country. There are many examples in Colombian history where changes in demand in the U.S. due to recession or other difficult economic situations have a serious impact on the coffee growing landscapes. For example, between 1880 and 1888, a period where the American economy went through a phase of recession, a large number of coffee plots was converted into pasture because of this market crisis (Palacios 1980). After 1890, when the U.S. market recovered, the Colombian coffee-producing area expanded again as a response to the increased demand. Because the area planted in coffee had contracted, supply was limited and this resulted in higher prices. During these market troughs, small coffee farmers were not as affected as large and medium producers. The reason for this is that in small farms, production was barely above subsistence levels producing both for the market and local consumption. When coffee prices were low, smallholder farmers devoted less effort to the coffee plots and invested more energy on the other livelihood strategies of the coffee farm. As stated above, the emergence of the International Coffee Agreement increased coffee prices by leveling supply and demand. Sudden changes in production caused by environmental factors, such as frosts in Brazil, drastically changed supply, and prices soared. Higher prices promoted better management and new investments in coffee plantations in Colombia. For example, the 1975 frosts in Brazil destroyed most of the coffee plantations of this South American country, affecting supply very drastically. This sent prices to their highest levels in history and encouraged coffee farmers in Colombia
130 and elsewhere to establish new plots and improve management of existing ones (Palacios 1980). This increase of prices, as well as the emergence of African and Asian coffee producers in the 1950s and 1960s encouraged Colombian coffee farmers to seek a more efficient and productive coffee agricultural system. The increased competition contributed greatly to the modernization of Latin American coffee farms that started in the 1970s, replacing traditional, more diverse coffee production systems into more homogeneous coffee producing units (Nestel 1995). Model of Landscape Change 1970-2002: The Intensification of Production By the late 1960s, the coffee leaf rust disease had arrived to Brazil and Central America. As it was presented in the previous chapter, this disease was responsible for wiping out coffee production in Ceylon in the 19 th century. As a result, the National Coffee Growers Federation was very worried about the consequences that a similar outbreak could have in a country that, at the time, was heavily dependent on coffee. The National Coffee Research Center (Cenicaf) was put in charge of designing a coffee production system that made the crop less susceptible to this disease, and at the same time increased the productivity of coffee production. By the late 1960s, the first experiments were taking place in the department of Caldas, in the main coffee-growing area (Parsons 1968). These involved reducing or removing the shade associated with the traditional production system, and switching to shorter coffee varieties (Coffea arabica v. Caturra) in order to be able to fit more bushes per hectare. For most coffee growers it is a known fact that unshaded plantations have higher yields and a shorter life span (Parsons 1968). Therefore the new system increased productivity not only by increasing the planting density, but also by increasing sun exposure. During the 1970s, Cenicaf designed a coffee production system and its associated management practices including
131 soil conservation measures, fertilizer and pesticide application calendars, and frequency and methods of plot renovation, among others. This system was strongly promoted by the National Coffee Growers Federation extension service. The intensification process in Colombia was slow in the beginning. Smallholder farmers were reluctant to accept the new coffee production system, and the earliest adoptions of the new system were mostly among medium and large-holder coffee producers. By 1974, only 10% of the coffee plots had been transformed into the intensive production system. However, these producers accounted for 30% of the total coffee production, showing clearly the increased productivity of the new production system (Palacios 1980). By 1981, 34% of the area planted in coffee used the intensive coffee production system (FNC 1983). The overall effect of the adoption of the new production system was that it differentiated between small, medium and largeholder coffee producers, something that was not present in the traditional production system. Because the intensive system requires high levels of capital and labor, these were mostly associated with the middle and large producers (Rice 1997). Therefore, while the smallholder producers tended to maintain the less-intensive production system, middle and large-holder farmers adopted the intensive system. Another major effect was that, because its higher productivity, it required less land to produce the same amount of coffee. The area planted in coffee dropped from more than a million hectares in 1970 to about 870,000 hectares in 1997 (FNC 1970, 1997), making the nearly 200,000 hectares available for other crops. The increased productivity also meant that whereas with the traditional systems farmers relied on local labor and family members as the major source of labor for the harvest, now
132 middle and large-holder coffee producers require more labor than the local market can provide. This created a large force of migrant workers that travels throughout the country during the harvest season (Duque, Restrepo, and Velsquez 2000). Duque et al. also present evidence that highlights the large labor requirements of coffee production. Presently, harvest costs are around 40% of the total production costs, and are mostly associated with the migrant workers required to harvest the coffee cherries. In Colombia and Central America, the harvest accounts for 60% of the labor input required for coffee (Palacios 1980). The permanent need for workers has also changed the livelihood strategies of smallholder farmers. Presently, their main source of income is wage labor in other farms rather than production in their own land (Palacios 1980). These two facts, the presence of a large pool of migrant labor and the increased reliance of small-holder farmers in wage labor, coincide with the ideas presented by Von Braun (1995) about commercialization of agriculture. According to von Braun, in areas undergoing commercialization of agriculture there is a significant increase in hired labor. The major landscape transformations are summarized in Figure 8. For much of the 1970s and until the mid 1980s the intensification of coffee production was a relatively slow process. The area in coffee decreased and the percentage planted in the intensive coffee production system increased. Some of the land formerly planted in coffee was planted with other crops like vegetables, sweet cassava, citrus, tomatoes, macadamia nuts, and different fruits among others. The area in pasture decreased at the same time, making more land available for other crops.
133 70s-Mid 80sMid 80s-Mid 90sPests, Diseases, Markets, Eradication Abandonment Coffee Trad% Int% Other Pasture Mid 90s-PresentMarkets Traditional coffee Coffee Trad% Int% Other Pasture Coffee Trad% Int% Other Pasture Trad: Traditional Coffee Production SystemInt: Intensive Coffee Production SystemOther: Other crops like maize, beans, fruit orchards, yuca, vegetables, among others: Increase in area: Decrease in area %: Increasing proportion %:Decreasing proportionArrows represent the possible changes for each land cover70s-Mid 80sMid 80s-Mid 90sPests, Diseases, Markets, Eradication Abandonment Coffee Trad% Int% Other Pasture Mid 90s-PresentMarkets Traditional coffee Coffee Trad% Int% Other Pasture Coffee Trad% Int% Other Pasture 70s-Mid 80sMid 80s-Mid 90sPests, Diseases, Markets, Eradication Abandonment Coffee Trad% Int% Other Pasture Mid 90s-PresentMarkets Traditional coffee Coffee Trad% Int% Other Pasture Coffee Trad% Int% Other Pasture Trad: Traditional Coffee Production SystemInt: Intensive Coffee Production SystemOther: Other crops like maize, beans, fruit orchards, yuca, vegetables, among others: Increase in area: Decrease in area %: Increasing proportion %:Decreasing proportionArrows represent the possible changes for each land cover Figure 8. Landscape evolution path 1970-2002 In the early 1980s the dreaded coffee rust disease arrived to Colombia, and this accelerated the transition to the intensive system. By this time Cenicaf had developed a coffee variety resistant to this disease (Coffea arabica v. Colombia), and it was widely promoted. The same trends of decreasing coffee and pasture, and increasing the area under other crops were present until the mid 1990s. During this period and as a major consequence of the arrival of the coffee leaf rust disease, the proportion of the area on coffee planted with the intensive system kept increasing. In the mid 1980s the coffee-cherry borer also arrived to the country, and this was a major blow to coffee producers. Heavy use of pesticides was the first mechanism to control this pest, but later on methods of bio-control were developed in Cenicaf using an African wasp who is a natural enemy of the coffee-cherry borer. Although the International Coffee Agreement ended in 1989,
134 its effects were not felt until the mid 1990s. After 1989 prices dropped for a few years, but Colombia, as will be explained later with in more detail, had an institutional framework that buffered coffee farmers from the vagaries of the international market. Prices went up again in 1994 as a result of frosts in Brazil (Dicum and Luttinger 1999). It was only after supply increased again that prices slumped dramatically. Since the mid 1990s coffee prices have been at their lowest level in decades (ICO 2003a). The institutions that buffered coffee producers ran out of money and farmers started to feel the brunt of the coffee crisis. Coffee prices reached prices so low that production costs were higher than the price of coffee in the international market. Many coffee farmers eradicated or abandoned their plots. Many farmers hung to coffee, as there is a very strong cultural attachment to coffee. In many areas of the country farmers maintain coffee because they feel that changing their emphasis to other crops betrays their own culture (Escobar and Ferro 1991). The area in coffee kept decreasing, and the area planted with other crops increased. In the same way as it has happened in the past with other coffee market crisis, the area of pasture started to increase. Many of the eradicated coffee plots were converted to pasture. The adoption of the intensive coffee production system also had some major ecological implications. In first place, shade trees have mostly disappeared. As it was discussed before, this has serious implications for areas where there are little natural habitats left, like the Colombian coffee lands. However, because the trees in these shaded coffee plantations were mostly planted, the benefits of Colombian traditional coffee plantations were not as marked as the rustic system or traditional polyculture system (Moguel and Toledo 1999). Another major implication of the widespread adoption of the
135 intensive production system is that soil erosion has increased (Siavosh Sadeghian personal communication). Figure 9. Erosion in a poorly managed intensive coffee plot. The reason for this is that, even if the technical recommendations for the intensive production system include soil-conservation measures, most farmers adopt only the components of the intensive system that increase productivity (higher planting density, short coffee varieties, increased use of agrochemicals). As a result, the soil is exposed and severe erosion takes place (Figure 9). In many areas, the top soil has been totally eroded, and productivity is maintained at high levels by fertilizer applications (Siavosh Sadeghian personal communication). The widespread use of chemicals has also had some implications for human health and for biodiversity. Some pesticides are used despite the fact that they are forbidden in the country (i.e., endosulfan, an organochloride). When this pesticide is applied to a coffee plot to control de coffee-cherry borer, it also kills almost all other animals and insects (Jorge Botero personal communication). However, it is
136 important to mention that, when compared to coffee (either intensive or traditional), it is the other crops that have appeared in the landscape the ones that use more chemicals. Shifting Centers of Production In the same way that the major centers of coffee production have historically shifted to new locations, Colombiaâ€™s main producing regions have changed through time. A timeline of these changes has been compiled from the available literature on the history of coffee in Colombia (Chalarc 1998, 2000; Errzuriz 1986; Ortiz 1989; Palacios 1980; Parsons 1968; FNC 1970, 1976, 1983, 1997)Jorge Botero personal communication). The reader should refer to Figure 6 , as specific department names are mentioned, and this figure presents the coffee producing departments in the country. In the early 19 th century, most coffee was produced in N. de Santander, near the Venezuelan border. In the late 19 th and early 20 th century, production shifted to Cundinamarca and Santander. By 1900 Santander was producing 60% of Colombian coffee. Between 1910 and 1930 the area of Caldas, Risaralda and Quindo became the leading producers. By the late 1960s, 75% of the Colombian coffee was produced in the departments of Antioquia, Caldas, Risaralda, Quindo, and Northern Valle. In the 1980s, a municipality in northern Tolima (El Lbano) was the largest coffee producer in the country. In recent years, production is becoming increasingly important in the Huila department. The other departments not mentioned in this timeline have never played a major role in coffee production in the country. National Coffee Growers Federation: The Supporting Role of Juan Valdez The National Coffee Growers Federation (FNC) has played a key role in shaping the landscapes and improving the livelihood conditions of farmers in the Colombian coffee lands (Chalarc 1998, 2000; Arango et al. 1998). Its important role began in 1927,
137 when a group of coffee growers decided to create a private organization to defend the interests of coffee growers, and promote coffee consumption both in Colombia and abroad. Since its beginnings the FNC has been in charge of coordinating the official Colombian government coffee policies, controlling the coffee market in the country, and collecting and managing the taxes from coffee exports (Chalarc 1998). It has also been in charge of promoting Colombian coffee abroad through a character called Juan Valdez who represents a Colombian coffee farmer. The FNC marketing strategy has been very successful, and the Juan Valdezâ€™s symbol has become one of the most recognized trademarks worldwide. Any coffee farmer with more than 0.5 hectares on coffee can be a part of the FNC. As a result, it is regarded as a very democratic organization. FNC members elect the representatives that attend the yearly national meeting to discuss the official coffee policies, and the members of the departmental committees, who are in charge of implementing these policies at the regional and municipal level through municipal committees. The departmental committees also coordinate the agricultural extension services that the FNC provides in most coffee growing municipalities for coffee and other agricultural activities, as well as the health, nutrition, and education programs sponsored by the FNC. They are also in charge of the programs to improve rural infrastructure (roads, electricity, drinking water, telephone) by investing money coming from the revenues of coffee exports in the producing areas. Additionally, they help with the establishment of coffee cooperatives for the marketing of coffee and other products (Chalarc 1998). Presently, there are more than 40 cooperatives with 386 purchasing outlets where farmers can sell coffee and other crops and buy farm supplies and other
138 products at favorable prices (Jos Jaramillo personal communication). Additionally, many municipal committees also sell farm and household equipment and supplies at attractive prices. The FNC also sponsors scientific research through Cenicaf, the National Coffee Research Center. Research in this center includes not only how to improve coffee production, but also better soil and water management practices, and research on other agricultural products associated with coffee production. Recently, biodiversity conservation has become an important research subject at this center. Besides the taxes collected from coffee exports for the Colombian government, the FNC also collects a voluntary contribution from the coffee farmers for the institutionâ€™s operating costs. The amount of the contribution is agreed upon at the FNCâ€™s annual meeting. Since the first coffee quota agreement signed with the U.S. during the Second World War, there has been a trust fund called the National Coffee Fund in charge of managing the purchase of coffee in Colombia and the export of the allowed quota, and setting the internal price of coffee for Colombia. This fund was established in the early 1940s and is in charge of storing the excess coffee, marketing of this excess coffee in non-ICO countries, and providing other services such as insurance companies and banks to coffee farmers (Chalarc 2000). This fund also acts as a buffer between the coffee farmers and the international market. When coffee prices were high, coffee farmers received more money, but at the same time, a larger share of the export price was saved in this fund. The idea behind this savings was to allow the fund to pay a higher price to the farmers when coffee prices were low. This worked very well to cope with the periods of high and low prices while the ICO Agreement was in place. The National Coffee Fund became a very rich and powerful fund, and started providing important services for the
139 FNC. For example, a shipping company was created with other Andean countries to make it easier to market coffee internationally. It also created banks for coffee farmers, insurance companies, storage facilities, and even owned a large portion of a commercial airline (Chalarc 2000). However, after the collapse of the ICA in 1989, prices dropped dramatically. In order to help the coffee growers, the National Coffee Fund started to sell its assets to keep paying slightly higher prices to coffee farmers. By the late 1990s, there were few assets left, and the fund stopped paying higher prices to farmers, as there were not â€œsavingsâ€ left to pay for the price premium. Because the National Coffee Fund ran out of money, in 2002 and part of 2003 the FNC requested funds to the Colombian government in order to pay a subsidy to coffee farmers. However, the future of this subsidy is not clear at the moment, as the Colombian government is in the midst of an International Monetary Fund Structural Adjustment program, and one of the critical components is to reduce government spending and subsidies are an integral part of this. However, paying this subsidy helps the government to avoid a socioeconomic crisis of large proportions, as it helps farmers to survive rather than having to help them by providing other livelihood strategies which would require significant investments. The FNC has had a major impact on landscapes mostly through agricultural extension services. Since the 1970s, the FNC has aggressively promoted the switch to the intensive coffee production system, with all the landscape transformations that came with it and were described earlier in this chapter. Additionally, it has promoted diversification campaigns to try to reduce the reliance of farmers on coffee, and to provide farmers with additional earnings during the time of the year when coffee is not providing any income, while at the same time minimizing the risks for the Colombian economy of coffee
140 overproduction (Uribe 1996). However, many of these efforts have failed because weak marketing links that did not provide an outlet for the new products, and the excellent agroecological potential for coffee production, that despite the efforts of the FNC to diversify production, meant that farmers preferred coffee to other products. The landscape impacts of the FNC also include coffee-plot renovation campaigns, where an economic incentive is given to the coffee farmer that renovates plot. Plots can be renovated in two ways, and when the plot is in this state, coffee bushes are not producing (FNC 2001). The first one is to replace old coffee bushes by new coffee plants. The second one is a procedure called â€œzocaâ€, where the coffee bush is cut about 15 cm above the ground, and then it grows again from the stump. The zoca tricks the coffee bush to behave as a young coffee plant, increasing the future productivity of that coffee bush. After 2 to 3 years, the plot starts producing coffee again. Between 1998 and 2000, the FNC financed a massive plot-renovation program. The idea behind this policy was to pay farmers to renovate their plots, while at the same time reducing coffee production (i.e., helping to stabilize market prices) and improving productivity in the near future (3-5 years), when it was expected that coffee prices would have rebounded somewhat. About 196,000 hectares were renovated between 1998 and 2000 as a result of these incentives (Jos Jaramillo personal communication), which represents roughly 22% of the total area planted in coffee according to the 1993/97 coffee census (FNC 1997). These examples show how the FNC has played a major role not only in coffee production but also in the economics and politics of the Colombian Government. Coffee changed, and as the renovation example attests, keeps changing dramatically the Colombian coffee lands. Through the FNC efforts, coffee farmers improved their
141 standards of living. Furthermore, the benefits of coffee exports, the only major Colombian export for a very long period of time, also reached the rest of the country. Coffee farmers now feel that the Colombian government should give them a hand to help them during the coffee crisis in the same way that the earnings from coffee during much of the 20 th century fueled the Colombian economy. Additionally, the FNC is starting to look for different strategies to market value-added coffee products. The export of instant coffee to more than 40 countries is an example of such strategies. In 2003, several Juan Valdez stores that sell gourmet coffee have opened in Colombia and in the U.S. in an effort to increase the proportion of the price paid by the consumer reaching the coffee farmers, therefore improving the viability of the coffee business in Colombia. There are many alternatives that need more development, like the specialty and organic coffee markets, where there has been a lot of growth in the last decade (Pulschen and Lutzeyer 1993; Brown 1996b, 1996a). The future of the coffee industry in Colombia depends on how successful is the FNC in developing strategies to switch from primary commodity production to strategies that add value locally to coffee and exploit new markets, while at the same time continuing with its efforts to improve the livelihood of coffee farmers in the country.
CHAPTER 5 METHODS AND DATA SOURCES Data Sources In the Colombian coffee lands there is relatively little spatial information on land-use and land-cover change at a fine spatial scale. The reason for this is that in many areas of the coffee lands there are no recent aerial photographs, with the most recent photos available being more than 20 years old. In terms of satellite imagery, the mid-elevation altitude range of the Colombian mountain ranges also coincides with areas usually covered in clouds, making it difficult to find relatively cloud-free imagery to generate detailed land-cover maps. As a result, the landscape evolution patterns are mostly derived from other sources that allow this analysis to take place only at the municipal level. Primary Data Sources Agricultural extension agent questionnaires Because the FNC has been trying to improve the conditions for coffee production since 1927, large sums of money have been funneled towards agricultural extension services. In most instances, farmers and extension agents work in close cooperation, and it is more or less common that extension agents stay for long periods of time in the same area 2 . As a result, extension agents are witnesses of landscape transformation, and can provide valuable information on the dynamics and evolution of particular areas. Although in most instances these key informants can provide only qualitative information about 2 This was more prevalent before the beginning of the â€œcoffee crisisâ€ caused by the collapse of the international coffee agreement in 1989. After this date, extension programs were reduced, and many of the people with long trajectories in one location benefited from early retirement. 142
143 landscape transformations, it still made it possible to document land cover processes associated with coffee intensification with enough detail. A questionnaire with open-ended questions was developed to collect some of this information (See Appendix A). The questionnaire was divided into three parts. The first part contained questions about the evolution of the area planted in coffee, its yield, and if these changes are experienced by all coffee producers in the same way (small, medium, and large coffee producers). The second part collected information about the evolution of other land-use practices in the region (e.g., new crops, crops disappearing, other livelihood strategies), the presence or absence of home gardens, and the use of inputs (fertilizers, pesticides) for coffee and other crops. The last part of this questionnaire inquires about the possible reasons for the changes in land use and land cover as perceived by the extension agents. This research protocol was approved by the University of Florida Institutional Review Board (Protocol #2001-326). A preliminary version of the questionnaire was tested with some extension agents during a preliminary field season in June-July 2001. The questions that were not clear for the respondent were reworded, and this revised protocol was approved by the IRB office in early 2002. This questionnaire was distributed in two ways. Because of the impossibility of visiting the more than 500 coffee-growing municipalities in the country, most of the surveys were mailed to each departmental extension office, where they were distributed to each municipality. This is the standard procedure for most studies relying on questionnaires to gather information in a wide geographical area and a limited budget (Berdie, Anderson, and Niebuhr 1986; Neuman 2000). Additionally, questionnaires are a very common research protocol in Geographic research (Dixon and Leach 1978).A
144 review of the recent scientific literature suggests this kind of research method is mostly used as the most important data collection tool in psychology and medical studies (ISI 2003). I also visited some of the agricultural-extension offices where I interviewed the extension agent who had been in the area for the longest time. A total of 117 extension agents (117 municipalities) returned the answered questionnaire. In the present study, the response rate was only 20.93%. Although some authors argue that high response rates can be obtained if the research design follows certain rules (Berdie, Anderson, and Niebuhr 1986), others point out that mail questionnaires usually have low response rates (Neuman 2000). The scientific literature reports a very wide range of response rates. For example, while a study on demographics, attitudes, and reef management among sport divers reports a response rate of 56% (Ditton et al. 2002), a study about the willingness of Italian farmers to create or improve habitat for wildlife on their farms only report a response rate of 3.2% for the mailed questionnaires (Genghini, Spalatro, and Gellini 2002). Although the response rate in this study is low, it still provides information for all the municipalities in 5 of the 16 coffee producing provinces. Additionally, extension agents from two provinces, which were not able to provide information at the municipal level, filled the questionnaire via e-mail for the entire province. In these cases, the provincial extension chief officer answered the questions. Although the answers from these provinces are at a different spatial scale than those from the other questionnaires, they still provide valuable information about landscape evolution at a coarser spatial scale. It is critical to make some important remarks about the design of this questionnaire. In first place, this questionnaire uses open-ended questions even though it is not advisable
145 to use this type of questions as part of mailed or self-administered questionnaire (Neuman 2000). The reason for this is that the information gathered using this type of question may differ significantly from respondent to respondent, as the questions may be interpreted in different ways. Furthermore, the quantity of information provided can vary significantly, with some people providing very detailed answers while others not giving enough information (Neuman 2000; Berdie, Anderson, and Niebuhr 1986). Furthermore, open-ended questions are more difficult to analyze for the researcher, as there are not set answers as in a closed-ended questionnaire. On the other hand, open-ended questions are very valuable for gathering information about complex issues, such as land-use and land-cover change, because they allow richness of detail, and reveal the respondentâ€™s logic and his/her frame of reference (Neuman 2000). Additionally, open-ended questions are well suited for instances where all the possible answers for a closed-ended question are known, and therefore, open-ended questions are used in preliminary surveys to determine the possible answers for a closed-ended questionnaire (Berdie, Anderson, and Niebuhr 1986). Although it is clear that open-ended questions have some major disadvantages when used in mailed questionnaires, this kind of question was chosen because 1) all the possible answers were not known, and 2) to allow the respondents to elaborate on the causes of land-use and land-cover change. Another major disadvantage from this questionnaire is the use of recall questions to try to reconstruct land-cover changes. Respondents should not be given too much credit for good memory, and recalled information may differ significantly from the actual facts (Berdie, Anderson, and Niebuhr 1986). Presently, it is acknowledged that memory is less trustworthy than previously thought, and recalled information may be influenced by the
146 topic (e.g., socially undesirable facts), the significance of the event for the respondent, and many other factors (Neuman 2000). However, in this case it was hard to obtain other sources of information documenting land-cover change, and because of the nature of the work of the respondents (i.e., agricultural extension agents) and some cross-validation with recent agricultural statistics at the municipal level (not ground truthed) suggest that the information provided by the extension agents indicates land-cover change trends accurately. Farmer interviews Coffee farmers are the ultimate land-use decision makers in the Colombian coffee lands, and as a result, they can provide information on the reasons behind different land-use changes. I met with farmers whenever it was possible. I asked questions regarding different land-use practices, livelihood strategies, and if they had increased the area planted in coffee and other crops. Additionally, in order to try to get an idea of the degree of intensification, I asked them about the adoption of the intensive coffee production system, and the fertilizer and pesticide use. Unfortunately, because my main interviewing period coincided with the main harvest, most farmers did not have time to answer my questions, and I was only able to meet with a few of them. Although this is a very important weakness for a study concentrating at the local scale, it is less important for this particular case, as the scale of analysis is coarser and other factors different from farmerâ€™s land-use decisions (e.g., presence of guerrillas or paramilitary groups) have a stronger relationship with land-cover changes in a specific area. Interviews with key informants I also had the opportunity to meet with some people working in different offices of the FNC who were able to provide information about coffee production in the country. In
147 particular, I met with the technical manager of the FNC, the director of Cenicaf, FNCâ€™s coffee research center, and an agricultural-extension advisor to the FNC. The long-term experience in many areas of the coffee lands of the Director of Cenicaf and the agricultural-extension advisor helped me in getting a better understanding of the spatial and temporal dynamics of coffee production in Colombia. Because of the willingness of Cenicafâ€™s director, this institution provided office space and logistical support while doing my fieldwork in Colombia. While at this institution I interacted with experts in many topics associated with coffee production. In particular, my conversations with researchers from the conservation biology, agrometeorology, soil science, biometrics, and agricultural economics programs were very helpful in increasing my understanding of the coffee-production process in Colombia, and its associated consequences. The meetings with these researchers were open interviews, but I always asked how much land cover had changed in recent years. Based on these conversations and a careful literature review of Cenicafâ€™s library, it is possible to point out that most research carried out at this research institute does not incorporate a spatial dimension. Secondary Data Sources The FNC coffee censuses These are perhaps one of the most important sources of information. They provide spatial and temporal data on the area planted in coffee and other crops for each one of the coffee growing municipalities. Since 1970, the time around which intensification of coffee production began in Colombia, there have been 3 large-scale surveys on this product. The 1970 coffee census (FNC 1970, 1976) contains information about the number of coffee farms, the total area planted in coffee, the area in unproductive coffee
148 fields, and in intensive coffee, pasture, sugarcane, temporary crops, and other land uses in each coffee growing municipality. Temporary crops refer to products like cotton (Gossypium spp.), chili peppers (Capsicum spp.), arracacha (Arracacia xanthorrhiza), rice (Oriza sativa), green peas (Pisum sativum), beans (Phaseolus spp.) and others that are planted only during certain part of the year. Other land uses refer to other permanent or semi-permanent crops like fruit orchards, cacao (Theobroma cacao), plantains (Musa X paradisiaca), yuca (Manihot esculenta), pineapple (Ananas comosus), home gardens, and to built areas. This census was carried out using photo interpretation and field surveys. Preliminary land-cover maps were derived from the aerial photos. Later on, surveys were used to collect information about coffee production at the farm level and to check and correct the maps. Maps for some areas of the country (1:25.000) present the spatial arrangement of these land covers (FNC 1976). Although the aerial photo interpretation did include all the land in the temperate altitude belt, these maps only depict the land covers of areas where coffee is produced, leaving large gaps in areas where this crop and its associated land covers are not planted. The maps also include farm boundaries, and areas of forest and bamboo groves. This publication (FNC 1976) also includes estimates of the area in forest and secondary succession at the departmental level calculated from the aerial photographs. Although I searched for this information at the municipal level, and I was assured by FNC and Cenicaf personnel that it existed, it was impossible to locate it. The information present in this census is accurate, reliable and covers 97.18% of the area of the coffee lands (FNC 1975). According to the FNC (1975), the figures of coffee production only differ in about 1% with the historical data compiled by FAO-ECLAC for the period 1956-1970.
149 In 1980-1981 a similar coffee census was carried out (FNC 1983). Unlike the 1970 coffee census, the 1980-81 census did not compile information about any land covers different from coffee. Information about the area planted in this crop is presented at the municipal level, discriminating between the traditional and intensive coffee production systems. At the provincial level, this census also includes information on planting densities and yields discriminated by coffee production system. It also includes data on the distribution of the area by elevation ranges in each coffee producing province and coffee production system. The 1980-81 census used a similar method as the 1970 survey, that is, using recent aerial photos and field work to check the photo interpretation. The statistical and spatial information compiled for this census has a confidence level of more than 98%, according to its accuracy assessment (FNC 1983). The last coffee census was carried out during the period 1993/97. Although one of its main goals was to record the area planted in coffee, it also involved a comprehensive assessment of the socioeconomic conditions of the coffee producing households (FNC 1997). As a result, it required an extensive period of fieldwork to gather information of all coffee producing households in the country. Additionally, it included many of the characteristics about coffee production in each province as the 1980-81 census (e.g., planting densities, elevation distribution). The socioeconomic variables include the characteristics of the house (i.e., building materials, access to clean water, electricity, sewage, and telephone, level of schooling, number of people per household, and other characteristics). In terms of land covers, this census is more detailed than the 1970 coffee census. It includes information about the area in coffee (discriminated by traditional or intensive), sugarcane, cacao, plantains, other permanent crops like fruit orchards,
150 temporary crops such as corn, beans, tomatoes and vegetables, associated crops (i.e., plots in which crops different from are planted together. For example, corn and beans), pasture, forest and secondary succession, planted forests, and other uses (i.e., built up area). In order to compare land cover between 1970 and 1993/97, this detailed land-cover information was reclassified to match the 5 classes of the 1970 census. Unfortunately, I was not able to find any quantitative information on the reliability and accuracy of the information gathered during this census. However, the foreword to the final report of the 1993/97 census (FNC 1997) suggests that the results are of very high quality, and the associated Geographic Information System was at the time one of the most complete and advanced agricultural information systems in Colombia and Latin America (FNC 1997). Additionally, the fact that this information is still used and updated by many FNC provincial and municipal committees suggests that the information contained in this census is useful, reliable and accurate. The average farm size for each municipality was derived from the 1993/97 coffee census. It was calculated as the sum of the area in each land cover divided by the number of farms. This derived variable was used to compare the temporal evolution of farm size since 1970. Population census data The last Colombian population census was carried in 1993 (DANE 1993). The coverage of this population census was 88.3% (Rueda and DANE 1999). At the national level, the coverage for rural areas (82.5%) was lower than for the urban population (91.2%). From the 16 coffee growing departments, 4 had a lower coverage than the national average. The average and the median coverage in the Colombian coffee lands were 88.2% and 88.7% respectively (Rueda and DANE 1999). Demographic and
151 socioeconomic information is available at the provincial level for some variables and at the municipal levels for most variables. The following information was selected for the rural areas of each one of the coffee growing municipalities: Average age Average family size Number of people younger than 15 years of age per municipality Number of people older than 65 years of age per municipality Total number of dependents per municipality Total number of households Number of young people per household Number of old people per household Number of dependents per household Population density Total Urban and Rural population Municipios y regiones de Colombia This database, commissioned by Fundacin Social, a Jesuit organization in Colombia, provides a characterization of all the municipalities in the country (Sarmiento et al. N.D.). It compiles socioeconomic, demographic, environmental, and institutional information, data about government spending, major economic activities, financial and banking data, the agricultural sector, quality of life, recreational and cultural events, and armed conflict at the municipal level. When applicable, information is presented for the urban and rural portions of each municipality. The research team of the Municipios y Regiones de Colombia project used this information to create a typology of municipalities based on an index calculated as the sum of scores derived from the conditions in each one of the categories just mentioned. Unfortunately, the documentation of this data source does not describe the methods used to carry out the typology with enough detail, and in many instances the typology is calculated based on other indices (e.g., Index of environmental degradation) that are not described.
152 Furthermore, for many variables there is only partial coverage. As a result, I was only able to use variables that had complete coverage for the coffee growing municipalities and whose meaning was clearly stated. The following variables were complete and selected from this database: Number of financial institutions Number of state institutions Quality of life index for the rural areas (ICV) The quality of life index provides a measure of the standard of living and human well-being in any given region. It integrates information about 12 variables in 4 groups. These four groups of variables are (DNP 2000): 1. Quality of the house: materials used for the walls and floors of each house (e.g., brick vs. wood walls, dirt vs. tile floor) 2. Access and quality to utilities: access to clean water, sewage, electricity and telephone, and cooking energy source 3. Human capital and education: educational level of the head of household, highest level attained by children younger than 18, number of children attending school. 4. Size and composition of the household: number of children younger than 6 as a proportion of the household, number of people per room. For each variable, there is a table that translates its possible values into a numerical score. The Quality of life index is the sum of these scores. The minimum value is 0, and the maximum is 100, when all basic needs are satisfied. Ministerio de agricultura. Estadsticas agrcolas por consenso The Ministry of Agriculture has compiled yearly information about the area planted in different crops, total production, their yields, and in most instances consumer and producer prices since 1988 (MinAgricultura N.D.). These figures are the result of the compilation of statistics from different organizations dealing with agriculture and crop production and marketing (e.g., cattle raising cooperatives, citrus growers associations).
153 The information is collected by each municipality, and submitted to the provincial office of rural planning, where data for all the municipalities in a given province is assembled. Finally, the Ministry of Agriculture puts together the information of all municipalities, where it is aggregated and validated at the provincial level. The information for the period 1988-2000 is available to the public and can be downloaded at the Ministry of Agriculture website (http://www.minagricultura.gov.co) (MinAgricultura N.D., ). For each of the 16 coffee growing provinces I obtained the information about area, production, and yield for each crop listed that is also planted in the coffee-growing region for the period 1988-2000. Accessibility The accessibility to markets is one of the major determinants of commercialization of agriculture (Southworth and Tucker 2001; von Braun 1995; Pingali and Rosegrant 1995). Originally I intended to measure accessibility for each municipality as the road density (total length of roads/area of the municipality). Unfortunately, it was not possible to obtain the length of roads at the municipal level or a Geographic Information Systems (GIS) road coverage with enough detail suitable to perform a map overlay with the municipal boundaries, and derive this information. As a result, I used a GIS derived measure of accessibility based on the distance to the closest municipality. The municipal boundary data set was compiled by the National Statistical Office of Colombia (DANE) for the 1993 population census. In a GIS environment I selected all the coffee growing municipalities (polygons). Using the GIS tools, all these polygons were merged into a single one. Then, a new polygon was created from this file, using a 10 km buffer outside it (buffered polygon). The municipal data set also contains the coordinates of each municipal seat of power as attributes. This location information was transformed to a
154 point data set. Then, I selected all the municipal seats within the buffered polygon. Using the analytical capabilities of the GIS software, a surface (raster) was created measuring the distance from any point to the closest municipal seat. I used a cell size of 100 m for this process. This calculation of distance assumes a straight line between any two points and does not include the effects of topography. Therefore, it does not represent the actual distance coffee farmers have to travel to sell their coffee and other agricultural products. However, it can be used as a proxy of the actual distance to the market. Finally, using the GIS function of zonal statistics, I calculated the average distance to the closest town of each of the polygons representing municipalities. These values were stored as attributes of the municipality data layer (polygons). Rainfall Weather station information for the coffee growing area was extracted from the Atlas Hidrometeorolgico de Colombia (IDEAM-PROSIS 1995). I used mean multi-annual monthly records in order to get a measure how wet and dry can get in any given municipality. Each station is represented as a point in a GIS file with the weather data stored as attributes for each month. For each station, I compiled two new attributes from the mean monthly precipitation called Minimum2 and Maximum2. These variables are defined as the sum of the rainfall of the two consecutive driest months, and the two consecutive wettest months respectively. In order to characterize these variables in each municipality I used a GIS. I first selected all the coffee growing municipalities (polygons). Using the GIS tools, all these polygons were merged into a single one. Then, a new polygon was created from this file, using a 4 km buffer outside it (buffered polygon). The next step involved selecting all the weather stations that fell inside the buffered polygon. A surface (raster) was interpolated for each one of the variables of
155 interest using an Inverse Distance Weighting algorithm (2 nd power) with a cell size of 100 m. Finally, in order to characterize each municipality (polygons) I used the zonal statistics procedure to calculate the average Maximum2 or Minimum2 values for each municipality, and this values were stored as attributes of the municipality data layer (polygon). Ecotopes Ecotopes are defined as the physical environment of a biotic community (Thomas and Goudie 2000). Therefore, each ecotope represents the set of biophysical conditions where specific organisms live, and may be found in different parts of the world. The FNC adopted this concept for classifying coffee production environments. According to the FNC, there are 86 ecotopes in the four mountain ranges of Colombia that are homogeneous in terms of their agroecological potential and production characteristics. The main variables distinguishing ecotopes are soils and climate. Using a GIS data layer of ecotopes provided by FNC, I reclassified them into 4 categories that correspond to each of the mountain ranges. Then, municipalities were assigned to their corresponding mountain range. This classification was used to analyze regional differences in the intensification of coffee production. Location Geographical coordinates were used as an additional variable to characterize municipalities. The latitude and longitude of the centroid of each polygon were selected as the position that best represents the location of each municipality. These coordinates were stored as attributes for all municipalities.
156 Software Packages Used for Research The software packages used in this study include Geographic Information Systems, Statistical Packages, Spreadsheet and Word Processing software. The Geographic Information System packages used were ArcView 3.2a and 3.3, and ArcGIS 8.2 and 8.3 from ESRI. For carrying out the statistical analyses I used NCSS 2001 and SPSS 11.0. Finally, Microsoft Office XP provided the spreadsheet and Word processing capabilities for this project. Methods As Intensification Takes Place, Landscapes Diversify This part of the analysis relied mostly on the coffee census information and the extension agent questionnaires. I also used the few farmer interviews and the interviews with key informants as anecdotal information. Land-cover evolution To analyze land-cover evolution I first calculated the total area for each land cover for the entire coffee growing region for the 1970 and 1993/97 coffee censuses. Because the total area covered in each census is different, the data was converted to percentages of the total area surveyed in each coffee census, making data comparable. Then, I selected the municipalities present in both the 1970 and 1993/97 coffee censuses in order to compare land-cover evolution. Data on each land cover for each municipality was converted to a percentage of the total area. Using the spreadsheet with the information from the previous step, the percentages for each land cover in 1970 and in 1993/97 were compared, and I obtained the covers with the largest increase and the largest decrease for each municipality. This information was summarized in a contingency table showing the percentage of municipalities where certain land covers have increased (i.e., the land cover
157 with the largest increase) at the expense of others (i.e., the land cover with the largest decrease). If the landscape has diversified, the land covers that were more prevalent in 1970 (coffee and pasture) should have decreased in a higher percentage of the municipalities while the remaining covers should have increased. In order to compare the two coffee censuses information for the three land covers that have shown more change (pasture, other, coffee), I calculated the ratio of pasture to coffee and other to coffee for 1970 and 1993/97 for each municipality. These ratios show the hectares of pasture or other crops per hectare of coffee in each municipality. If the landscape diversifies, it is expected that the ratio of pasture to coffee should decrease and the ratio of other to coffee should increase. In order to test if the changes in these ratios actually take place I used the a non-parametric statistical test (Mann-Whitney) to compare the distributions of these values in the 1970 and 1993/97, as well as the number of municipalities where both conditions are true. Land-use system evolution Land use can be inferred from land cover. A constant assemblage of land covers through time suggests land-use practices that do not change and land managers still engaged in the same management practices. In order to assess if land use had changed in the Colombian coffee lands, I combined the information about the percentage of each municipality covered with coffee, pasture, sugarcane, temporary, and other crops into land-use systems. Then I analyzed if the land-use system in each municipality changed for the study for the period determined by the 1970 and 1993/97 coffee censuses. Using a k-means cluster analysis the land-cover information for each municipality in the 1970 census was classified into 4 land-use systems that have all the five covers being studied. Three of these land-use systems have one or two land covers that dominate (mainly
158 pasture, coffee and pasture, other and pasture). The last productions system (mixed) does not have a any land cover dominating. The 1993/97 land-cover information was classified into the same land-use systems. This was accomplished by assigning each 1993/97 municipality to the 1970 land-use system whose cluster means were closer to the 1993/97 land-cover values in each municipality. The next step involved comparing the land-use systems in 1970 and 1993/97 for each municipality. This allowed to see which municipalities had changed their land-use system (and to what land-use system they have changed to) and the percentage of municipalities that have maintained their land-use system. Finally, I used a GIS to create maps of the land-use system type for each municipality in1970 and 1993/97, and then compare where the changes had taken place. Increasing agricultural diversity The coffee censuses data provide information about land cover only until 1997. To get an idea of the recent patterns of land-cover change I used the information from the extension agent questionnaires. First, I compiled a list of all the crops that have become more important (i.e., increasing area or farmers engaged more time in its production) and less important (i.e., decreasing area or farmers engaged less time in its production) in the period 1997-2002 for each municipality. Then I calculated the percentage of municipalities where a specific crop is becoming more or less important. If the landscape is becoming more diverse, the products becoming important should be reported in a higher number of municipalities than the products becoming less important. Furthermore, there should be more crops becoming more important than less important. Following this step, I used a GIS to portray this information spatially. Municipalities were classified into classes with the following rule: If a municipality has the same or a higher number of crops becoming more important than of crops becoming less important, the number of
159 crop has increased. If this condition is not met, the number of crops has decreased. A municipality where the number of crops has increased suggests that there might be more crops planted in the same area, leading to higher agricultural diversity. Some Land Uses Intensify while Others Disintensify For testing and supporting this hypothesis I used the coffee census information, anecdotal information supplied by key informants from the FNC, the Ministry of Agriculture statistics, and data from the extension agent questionnaires. The analysis of this hypothesis was divided in two different parts. First, I looked at the spatial and temporal evolution of coffee production intensification in the 1970 to1993/97 time period. Secondly, I used the Ministry of Agriculture statistics time series to determine whether if certain crops had intensified while others disintensified based on the temporal evolution of their yields. I complemented this information with the answers about crops becoming more/less important from the extension agents at the municipal level. Evidence of coffee production intensification In the case of coffee production intensification, the intensive coffee production system involves a different management strategy than the traditional system. In order to analyze the spatial evolution of intensification, I selected all the municipalities that were listed in the three coffee censuses (1970, 1980/81, 1993/97). This information was joined to a GIS data layer of the municipalities, and I prepared a map showing the adoption of the intensive system through time. The availability of information at three different times defines two time periods that can show how dynamic is coffee as a land cover in a specific municipality. For each time period (1970-1980, 1980-1993/97) I compared the area planted in coffee in each municipality at the beginning of the period with that at the end. This comparison for two time periods lead to four different classes: increased in both
160 time periods, decreased in both time periods, increased in the first time period and decreased in the second, and, decreased in the first time period and increased in the second. By comparing the overall trend (1970-1993/97) with the trends from the two time periods defined by the three coffee censuses, it is possible to obtain a picture of the evolution of coffee production. For example, if in any given municipality the area in coffee increased between 1970 and 1993/97, but decreased between 1970-1981 and increased between 1980 and 1993/97, this means that coffee became a predominant crop relatively rapidly (after 1981). An increasing trend for the entire time period may be masking a period of growth followed by a period reduction in the coffee area. By having two (or more) time periods, it is possible to analyze these dynamics. Finally, in order to see if intensive coffee production was concentrating in certain areas of the country, an analysis of spatial autocorrelation was carried out for 1970, 1980, and 1993/97 using the percentage of the total area planted in coffee using the intensive production system for each municipality. For each year, the Moranâ€™s I value was calculated, and it was tested if this value could be the result of a random process or not. Evidence of simultaneous crop intensification and disintensification In order to see if intensification of certain crops or land-use practices is simultaneous with the disintensification of other land-management practices I used the Ministry of Agriculture data on yields for the period 1988-2000, and I complemented these results with an analysis of the extension agent questionnaire answers. The agricultural statistics are available at the provincial level. Because the 16 coffee growing departments include elevation ranges where a variety of crops is planted, I only selected the crops for which complete records existed that grew in the same elevation range as coffee. In order to minimize the variability in annual yields caused by weather patterns
161 (e.g., droughts), pest outbreaks, or other phenomena, yields were smoothed using a 3 year moving average. This decreased the temporal range of the time series from 1988-2000 to 1989-1999. Then, Spearman-rank correlation coefficients were calculated for every pair of crops in each province both for the original data (not-smoothed out) and the 3-yr moving average data. This non-parametric test was chosen because the time series data on yields for most crops in the 16 provinces does not follow a normal distribution. To determine whether the correlations calculated with the original data and the 3-yr moving average data on yields were the result of a random process, I simulated yield time-series data for fifty crops for a 50 year period. For each crop, the yield in any given year was a function of the previous year plus a random quantity. Then, these 50 time series were used to calculate Spearman-rank correlation coefficients among crops using the original data, and 3, 5, and 7 year moving averages. These coefficients were used to derive a probability distribution of the correlation coefficients using different moving averages. These semi-random probability distributions were also used to determine if the correlation coefficients obtained from the crop yield data for 1988-2000 were outside the 95% confidence interval of the semi-random probability distributions mean. If the actual yield data correlation coefficients were outside this confidence interval, they are not likely caused by a random process. If crops are intensifying together, it is expected that they exhibit a positive correlation coefficient. If a crop is intensifying while another is disintensifying, they should have negative correlation coefficients. In order to include the influence of climate variability on agricultural yields, I used annual rainfall time series (1988-2000) from representative weather stations in each of the coffee growing departments. The influence
162 of climate was minimized by using partial correlations (i.e., excluding the influence of rainfall patterns). The Factors Accompanying Intensification Change from Region to Region. The literature review chapter presented evidence of how intensification of agricultural production depends on a wide variety of factors that changes from household to household and region to region. Furthermore, most models of agricultural intensification and commercialization were developed at the household level. As a result, it is important to determine if the variables deemed responsible for intensification at the household level are also important at the regional level (i.e., if the intensification process can be scaled up). For this purpose, I used the variables that lead agricultural intensification according to some of the household level intensification models and tested if these were also the most important variables at the regional level using multivariate regression. Exploratory analysis The first task involved an exploratory analysis to select the appropriate variables that lead to agricultural intensification. From the socioeconomic, demographic, and environmental information for each municipality I selected variables that, according to the models of Chayanov, Boserup, von Braun should be critical in determining agricultural intensification. I also tried to include some environmental factors, as it appears that at landscape and regional scales, climatic and topographic variables are important in determining land-use change (Veldkamp and Lambin 2001). Table 5 presents the variables that were pre-selected for this analysis. In order to minimize redundancy between variables and eliminate unnecessary factors, I used a correlation matrix. I kept the variables that were highly correlated with the area in intensive coffee
163 (assuming that higher correlation coefficients imply some kind of direct or indirect relationship). I also wanted to have at least one variable from each one of the four classes in Table 5. With the results from the correlation matrix (Spearman rank, as most variables do not follow a normal distribution), I was able to rank the importance of the variables and eliminate variables with high correlation coefficients. Table 5. Variables pre-selected for the regional analysis Description Variable Boserup Rural population density Farm size Coffee plot size Chayanov Average family age Dependents per household Index of Quality of Life (ICV) Von Braun Distance No. financial institutions No. State institutions Urban population (%) Environmental factors Total Rainfall Precipitation Concentration Index (PCI) Minimum 2 Maximum 2 Location (Long, Lat) Mountain range Using the variables in Table 5 as independent variables, and the area in intensive coffee as the dependent variable, I ran a multivariate regression model. The basic purpose behind this model was to identify the variables which exhibited a great degree of multicollinearity. Multicollinearity is the existence of near-linear relationships between independent variables, and it can lead to misleading results in a multivariate linear regression such inaccurate regression coefficients and degrade the predictability of the regression model (Hintze 2001). To minimize the effects of multicollinearity, I ranked the independent variables according to their variance inflation factor (VIF), a measure of
164 multicollinearity. Two criteria were used to eliminate variables from the regression model. Variables whose VIF was extremely high (>10) were eliminated while trying to maintain at least one of the variables of each type in Table 5. Three variables were selected to carry out regional breakdowns. These variables divide the coffee growing municipalities into spatial datasets. The first variable was the mountain range. This divides the municipalities into 4 contiguous areas. The second variable was farm size. This variable allows to see if there were any significant differences among small (<3 ha), medium (3-10 ha), and large (> 10 ha) coffee producers. The last variable selected for a regional breakdown was the Index of Quality of Life. The municipalities were divided in two regions based on this index (above average, below average), to see if intensification of coffee production is associated with better living conditions, as von Braun (1995) hypothesizes. The selection criteria described lead to a subset of 12 variables. Although originally the location variables were included up to the third power, I decided to use only the original values for longitude and latitude. The reason for this is that in the analyses that will be presented below, these higher-order location terms overwhelmed the effects of all the other variables. The variables selected are shown in Table 6. Principal components analysis Principal Components Analysis (PCA) is a technique used to reduce the dimensionality of a multivariate dataset while preserving the majority its information (Hintze 2001). The PCA produces a group of new variables that are linear combinations of the original variables, and, unlike the original data, are uncorrelated. The interpretation of the new variables provides some insights about how the original variables combine and capture much of the variability of the dataset.
165 Table 6. Variables selected for multivariate regression analysis Description Variable Boserup Rural Population Density Chayanov Average family age Dependents per household Von Braun Distance No. financial institutions No. State institutions Environmental factors Minimum 2 Maximum 2 Location (Long, Lat) Variables for regional breakdown Mountain range Index of quality of life (ICV) Farm size For the present study, I ran a PCA using the 10 variables selected in the previous analysis (rural population density, average family age, dependents per household, distance, number of financial institutions, number of state institutions, minimum2, maximum2, latitude, longitude). The four principal components selected captured more than 10% of the dataset variability each. The PC scores for each municipality were then used as independent variables in a multivariate regression model. The dependent variable (Area in intensive coffee) was transformed (Intensive Area 1/3 ) to ensure the regression residuals follow a normal distribution, therefore complying with the assumptions for the multivariate regression coefficient significance tests. A power transformation was chosen because it eliminates most of the problems associated with a logarithmic transformation (i.e., need to add a constant to zeros before transforming the variable). This transformation was also selected because it was the first transformation I tried that yielded residuals with a normal distribution. With this set of independent variables (PC scores) and the transformed dependent variable (Area in intensive coffee) I ran two different regression models: 1) with first-order interactions among independent variables,
166 and 2) without interactions among independent variables. For each regression model, the regression coefficients were interpreted based on their significance and their relative contribution to the model as standardized coefficients. Multivariate regression In order to determine the factors that have a stronger relationship with intensification a multivariate linear regression model was used. As in the PCA regression analysis, the dependent variable (Intensive Area) was transformed (Intensive Area 1/3 ) to assure that the regression residuals follow a normal distribution. Using this transformed variable and 10 independent variables (rural population density, average family age, dependents per household, distance, number of financial institutions, number of state institutions, minimum2, maximum2, latitude, longitude), I ran a stepwise forward multivariate regression model considering interactions among variables and no interactions among variables. The purpose of this was to see if the explanatory power of the regression model improved significantly by incorporating the interactions among variables, and by using a stepwise forward variable selection method it is possible to see which variables explain the most (i.e., single variables, interacting variables) by looking at the order in which they were entered in the regression. Regional regression analysis In order to see if the factors most closely related to intensification change from region to region, I used a similar multivariate regression analysis but incorporating regional breakdowns. I used three of the variables selected (farm size, ICV, mountain range) to create 3 regional breakdowns. Each regional breakdown was analyzed independently. The breakdown by mountain range divided the country into 4 regions. The
167 subdivision by farm size divided the country into 3 regions. Finally, the ICV criteria divided the country in two regions. For each one of the regional breakdowns, the regions were incorporated into the regression models using dummy variables (Chattaerjee and Price 1991). The following general model exemplifies the approach used assuming 2 regions (only 1 dummy variable D 1 ): nnnnXDXXDXXDXDY1,1212,122111,11110.... By multiplying D 1 by each one of the variables, it is possible to know if the coefficients and intercepts for the regression equation are different in the two regions under study. If any of the coefficients of the variables multiplied by the dummy variables ( 1,n ) is significantly different from zero, this means that the variable within that region behaves differently than outside the region. This approach can also be extended to include interactions among variables. However, in this case that was not possible because when dummy variables are included for all interactions the number of variables results larger than the number of observations. Regression models of this type were run for each one of the regional breakdowns.
CHAPTER 6 RESULTS Evidence of the Intensification of Coffee Production The area planted in coffee in Colombia has declined steadily since 1970 ( Figure 10 ). At the same time, total production and productivity have increased (Figure 11). This situation clearly suggests that intensification of production has taken place because, in a smaller area, total production has increased through the adoption of new management practices. 5006007008009001,0001,10019701972197419761978198019821984198619881990199219941996199820002002YearArea (Thousand ha)70.0%75.0%80.0%85.0%90.0%95.0%100.0%Area in Production/Total Area (%) Total Area Area in Production Area in production (%) Figure 10. Evolution of the coffee growing area in Colombia. Total area refers to the area planted in coffee (hectares). Area in production refers to the coffee plots that are actually in production (hectares). Area in production (%) is the ratio of the area in coffee in production to the Area planted in coffee. 168
169 6,0008,00010,00012,00014,00016,00018,00019701972197419761978198019821984198619881990199219941996199820002002Production (1000 60-kg sacks)68101214161820Yield (60-kg sacks/ha) Total Production Yield Figure 11. Evolution of total coffee production and yield 1970-2002. $300,000$400,000$500,000$600,000$700,000$800,000$900,000$1,000,000Jan-70Jan-72Jan-74Jan-76Jan-78Jan-80Jan-82Jan-84Jan-86Jan-88Jan-90Jan-92Jan-94Jan-96Jan-98Jan-00Constant Colombian $ per 125 kg of parchment coffee Figure 12. Price paid to the farmer for 125 kg of parchment coffee. Parchment coffee is green coffee that still has the husk attached to it (Constant price 2002)
170 Despite the fact that the year-to-year production shows large oscillations (Figure 11) coffee production has two distinct phases, one before 1977, and one after 1977. In the first stage, the total coffee production of Colombia was around 8,000 thousand 60kg-sacks per year, and after 1977 it jumped quite fast to about 11,000 thousand 60kg-sacks per year. Total production reached a peak in 1992, decreased to its lowest level since 1978 in 2000, and started to rebound in 2001. It is interesting to note that the largest increases in productivity took place around the time when the coffee farmer received more money for his coffee (Figure 12), showing the link between intensification of coffee production and favorable conditions in the international market. What is surprising is that yield and production were maintained during much of the 1980s as coffee farmers received less and less for their coffee. The decreasing trend on the price paid to the farmer also highlights how, as discussed in an earlier chapter, farmers get less money for his coffee as time goes by. For most of the time period under study (1970-2002), the ratio between the area of coffee in production and the total area in coffee has remained above 90% (Figure 10). After 1999 this proportion decreased to values around 80%. This sharp drop corresponds to a major coffee-renovation program launched by the FNC in 1998. By 2000, more than 22% of the coffee growing area had been renovated (Jos Jaramillo personal communication), therefore taking these coffee plots out of production for a few years and reducing the producing area. By 2002 this trend continued, possibly as a result of low coffee prices in the international market that encourage coffee farmers to invest their resources in other land-use practices that provide higher returns.
171 As the total area planted in coffee decreased by more than 18.5% since 1970, the number of municipalities engaged in coffee production has also declined (Table 7). The average area surveyed per municipality decreased from 9822 ha in 1970 to 8539 ha in 1993/97. The average area planted in coffee per municipality also decreased from 1818 ha in 1970 to 1555 ha in 1993/97. While the proportion of the total area planted in coffee under the intensive production has increased to 70%, the number of coffee farms has more than doubled (Table 7). The average and the median coffee farm sizes have decreased more than two times, and the standard deviation has increased between 1970 and 1993/97. Table 7. Summary of FNC coffee censuses 1970 1980 1993/97 Number of municipalities 587 577 559 Total Area surveyed (ha) 5765587 N.A. 4773557 Area in Forest and Secondary Succession (ha) 1026968 N.A. 1151474 Area planted in coffee (ha) 1067113 1,003,940 869158 Intensive coffee/Total area in coffee (%) 0.22% 34.14% 70.04% Total Area in coffee/Total Area Surveyed (%) 18.51% N.A. 18.21% Number of coffee farms 301708 N.A. 609432 Average coffee farm size (ha) 24.2 N.A. 11.7 Median coffee farm size (ha) 15.5 N.A. 6.1 Std. Dev coffee farm size (ha) 54.7 N.A. 64.1 Smallest farm (ha) 1.3 N.A. 0.8 Largest farm (ha) 1064.2 N.A. 1509.5 These facts suggest that although farms are generally getting smaller, the range on size has increased. This is confirmed by the decrease in the minimum farm size has decreased and the maximum farm size has increased. This means that as commercial intensification takes place, the majority of land holdings are becoming smaller, but also some land holdings are becoming larger. This is somewhat contrary to the land-holding
172 consolidation associated with commercialization postulated by Pingali and Rosegrant (1995). These authors indicate that as commercialization of agriculture takes place, farms tend to become larger to accommodate more efficiently agricultural machinery. Although the total area in coffee in Colombia has declined since 1970, there are many areas of the country where coffee cultivation is expanding (Figure 13). Coffee is decreasing in the majority of municipalities (59.7%), and increasing in 28.2% of the coffee growing municipalities for which records exist in the three FNC coffee censuses (FNC 1970, 1983, 1997). The areas where coffee is increasing are concentrated in certain regions of the country. In the northern part of Colombia, the increase in the area planted in this crop is associated with an agricultural frontier that is still expanding. In the southern section of the coffee growing region (department of Huila, Figure 6), the increase is the result of a shift in the center of production. The department of Huila is one of the areas that has more land suitable for intensive coffee production in the country, and as a result, the crop has expanded in this area (Chalarc 2000) The expansion and contraction of coffee in Colombia has been more complex than what Figure 13 suggests (Figure 14). While the area in coffee increased in 145 municipalities between 1970 and 1993/97, it only increased steadily in 61 of those municipalities between both 1970 and 1980, and 1980-1993/97 (Table 8). The area in coffee decreased in 174 of the municipalities for the same time periods. This suggests that the area associated with coffee production in Colombia does not necessarily follow the linear trend of expansion or contraction presented in Figure 13. In reality, about 46% of the municipalities exhibit a constant trend for the two time periods (Table 8), and more than half of the coffee producing municipalities have an
173 Figure 13. Evolution of the area planted in coffee 1970-1997 in the Colombian coffee growing municipalities. Coffee is grown between 1000 and 2000 masl on the slopes of the Colombian Andes
174 Figure 14. Evolution of the area planted in coffee by time period
175 Table 8. Evolution of the area planted in coffee by time period Total Change 1970-1993/97 Change by Time period (1970-1980, 1980-1993/97) Data Decrease > 10% Increase > 10% No change Grand Total Decrease, Decrease Municip. 174 1 175 Percentage 33.9% 0.0% 0.2% 34.0% Decrease, Increase Municip. 62 42 20 124 Percentage 12.1% 8.2% 3.9% 24.1% Increase, Decrease Municip. 71 46 36 153 Percentage 13.8% 8.9% 7.0% 29.8% Increase, Increase Municip. 57 5 62 Percentage 0.0% 11.1% 1.0% 12.1% Total Municipalities 307 145 62 514 Total Percentage 59.7% 28.2% 12.1% 100.0% increasing-decreasing trend (29.8%) or an increasing-decreasing behavior (24.1%). Although there are some municipalities (6) with consistent trends (decrease-decrease, increase-increase) in the â€œNo Changeâ€ column, this does not mean the area in coffee has not changed. The reason why they appear in this column is that the area planted in coffee has increased or decreased less than 10% in between 1970-1993/97. When each time period is analyzed independently, it is striking to see that between 1970 and 1980 the area planted in coffee decreased in 298 municipalities, and then from 1980 to 1993/97 the number of municipalities where coffee decreased had grown to 328 (Table 8). These facts suggest that coffee production is concentrating in certain areas of the country, as the number of municipalities where coffee is increasing is getting smaller. It is interesting to see that there are 88 (17.1%) municipalities that have increased the area planted in coffee between 1970 and 1993/97, but that show opposite behaviors for the 1970-1980 and the 1980-1993/97 time periods. There are 42 municipalities whose coffee area decreased in
176 the first time period, and increased in the second (i.e., very rapid increase in the second time period), and 46 whose area increased in the first time period and decreased in the second (i.e., large increase in the first time period). A similar situation is present for 133 municipalities (25.9%) whose coffee planted area decreased from 1970 to 1993/97 and have opposite behaviors for the two time-periods. 62 of these municipalities have a decrease in the coffee area between 1970-1980 and an increase for 1980-1993/97 (i.e., large decrease in the first time period). The remaining 71 municipalities show the opposite behavior (i.e., growth followed by decrease in area). The variety of possible outcomes in Table 8 clearly suggests that the evolution of the area planted in coffee in Colombia is very dynamic, and more than half of the municipalities exhibit a non-steady trend in the time period studied where periods of rapid growth/ contraction of coffee production are followed by periods of fast changes to less/more area planted in coffee. The increasing values of the Moranâ€™s I statistic suggest that coffee production is concentrating in certain areas of the country (Table 9). When this measure of spatial autocorrelation is calculated with the total area in coffee per municipality, the values have a slight increase from 1970 to 1993/97 (Table 9). This suggests that the municipalities with similar areas planted in coffee are clustered together (i.e., municipalities with large coffee areas next to each other). When I tested if these Moranâ€™s I values could be produced by a random process, all of the calculated Z values were statistically different from the critical value (= -0.00208, =0.0299). Therefore, the municipalities with the largest areas planted in coffee are close to each other, and this clustering is not the result of a random process. As a result, there should be some biophysical, demographic, socioeconomic, political, and institutional
177 factors, as well as their interactions at different temporal and spatial scales that are responsible for these spatial patterns. Table 9. Evidence of coffee production concentration in certain areas Area planted in coffee (ha) Area in Intensive Coffee (%) Year Moranâ€™s I Z value Random Moranâ€™s I Z value Random 1970 0.36 12.50 No 0.23 7.84 No 1980 0.38 13.38 No 0.64 21.42 No 1997 0.41 14.37 No 0.66 21.96 No The evidence of concentration of production is even more striking when the percentage of the area planted in coffee per municipality under the intensive production system is used in a similar Moranâ€™s I analysis (Table 9). The Moranâ€™s I value nearly triples from 1970 to 1980, and then it continues to grow at a slower rate until 1993/97. The increase suggests that the municipalities with the highest proportion of coffee planted with the intensive production system are clustered together, and this clustering is not caused by a random process (= -0.00208, =0.0299). In the same way that the overall decrease in the area planted in coffee in Colombia has not been homogeneous throughout the country, the adoption of the intensive coffee production system is also a process that changes from location to location (Figure 15). The 1980 map shows a large cluster of intensified coffee production in the northwest part of the country, while the rest of the municipalities in the coffee lands had transformed less than 50% of their coffee area into the intensive production system. The pattern reverses for 1997, when there are islands of more traditional coffee production surrounded by municipalities where more than 50% of the area in coffee was transformed to the intensive coffee production system. Some authors suggest that the adoption of the intensive system in Colombia was very slow at the beginning because farmers were reluctant to change their production system (Palacios 1980). Although Figure 15 would
178 Figure 15. Evolution of coffee intensification in Colombia. Intensification is equivalent to the conversion of the traditional coffee production system to the Intensive (sun-grown) coffee production system coffee.
179 seem to support this assertion, when one calculates the area transformed to the intensive production system from 1970 to 1980, and then from 1980 to 1993/97 with data from Table 7, there was a larger area transformed between 1970 and 1980 (340397 ha) than between 1980 and 1993/97 (266013 ha). However, before 1980 the transformation from the traditional to the intensive production system was concentrated in one area of the country. After 1980, the rate of transformation slowed down, but spread throughout the coffee lands as the coffee leaf rust disease, which arrived to Colombia in the early 1980s, spread through the country. From 1970 to 1980, the percentage of the area planted in coffee under the intensive production system increased in 489 municipalities (99.2%). From 1980 to 1993/97, the proportion of the under the intensive system continued to grow in 460 municipalities (93.3%), but shows a decreasing trend for 33 municipalities (6.7%). Although the four classes depicted in Figure 15 are arbitrary, it is interesting to see how the number of municipalities in each class has evolved (Table 10). The major change is that in1980 more than 40% of the municipalities had less than 25% of their coffee area under the intensive coffee system, while in 1993/97, only 12.3% of the municipalities had less than 25% of their area under intensive coffee production. Only 120 municipalities (24.3%) have remained in the same class, and more than 70% of the municipalities have more than 50% of the area planted with coffee under the intensive production system in 1993/97, compared to 24.3% of the municipalities in 1980. Although the general trend for a municipality is to move to the next higher-intensity class (e.g., from 25%-50% to 50%-75%), there are 9 municipalities that have disintensified coffee production. The presence of a small proportion of municipalities
180 where the proportion in intensive coffee has decreased since 1980 also suggests that the intensification of agricultural production for one product in one area can be associated with the disintensification of production for the same product in another area, as coffee has both intensified and disintensified in different municipalities. Table 10. Evolution of the proportion of the total area planted in coffee under the intensive production system Area in intensive coffee 1993/97 Area in intensive coffee 1980 Data Less than 25% 25%-50% 50%-75% More than 75% Grand Total Municipalities 53 69 82 25 229 Less than 25% Percentage 10.7% 14.0% 16.6% 5.1% 46.4% Municipalities 8 15 48 74 145 25%-50% Percentage 1.6% 3.0% 9.7% 15.0% 29.4% Municipalities 0 0 12 67 79 50%-75% Percentage 0.0% 0.0% 2.4% 13.6% 16.0% Municipalities 0 0 1 40 41 More than 75% Percentage 0.0% 0.0% 0.2% 8.1% 8.3% Total Municipalities 61 84 143 206 494 Total Percentage 12.3% 17.0% 28.9% 41.7% 100.0% Evidence of Landscape Diversification Because the area planted in coffee has decreased since 1970, it is also expected that the area in other crops and land uses that are usually associated with coffee production have also decreased (Table 7). In terms of total area, pasture has decreased drastically, while other crops, forest, and secondary succession have increased slightly (Figure 16). Because the area of the coffee lands is different for 1970 and 1993/97, I converted the areas of each land cover to percentages of the total area in order to be able to compare and analyze land-cover evolution between 1970 and 1993/97 (Figure 17). In relative terms, there has been a major decrease in the area in pasture, while other and forest and secondary succession have increased (Figure 17). The proportion of the total area planted in coffee has remained nearly constant. The remaining land covers also
181 2,147,691141,169 201,3031,026,9681,181,3441,067,113 1,313,60390,097119,709 1,229,516 869,158 1,151,4740500,0001,000,0001,500,0002,000,0002,500,000CoffeePastureSugarcaneTemporaryOtherForest SSArea (ha ) 1970 1993/97 Figure 16. Evolution of land covers 1970-1993/97 2.4%24.1% 37.3%18.5%20.5%17.8%3.5% 25.8%18.2%2.5% 1.9%27.5%0.0%10.0%20.0%30.0%40.0%CoffeePastureSugarcaneTemporaryOtherForest SSPercentage of total area 1970 1993-97 Figure 17. Evolution of land covers as a percentage 1970-1993/97
182 remained nearly unchanged in their relative contributions. Unfortunately, there is no information available about forest and secondary succession for 1970 at the municipal level, making it impossible to see if these patterns of land-cover change that are observed in the coffee lands as a whole are also represented in the majority of the municipalities. Because the major changes in absolute terms have been in coffee, pasture, and other crops (Figure 16), if the landscape has become more diverse one can expect that the number of hectares of pasture per hectare of coffee has decreased, while the hectares in other crops per hectare of coffee should have increased for most of the coffee growing municipalities. From 1970 to 1993/97, the number of hectares of pasture per hectare of coffee has decreased in all the coffee growing municipalities (511), while there are more hectares of other crops per hectare of coffee in 318 (62.2%) of the coffee growing municipalities. The 1970 average number of hectares of pasture per hectare of coffee for all the coffee growing municipalities is statistically higher than the 1993/97 value (Table 11, Mann-Whitney test, p=< 0.001). Additionally, the 1970 average number of hectares of other per hectare of coffee is statistically lower than the 1993/97 (Table 11, Mann-Whitney test, p=<0.001). This would suggest that agricultural diversification is taking place based on the conditions specified above. The comparison of the coefficients of variation of the two land-cover ratios for 1970 and 1993/97 (Table 11) suggests that extreme values are becoming less common, and there are more municipalities where the ratios between land covers are closer to the mean values shown on the table. Although the coefficients of variation for both ratios
183 (Pasture/Coffee, Other/Coffee) are much smaller in 1993/97 than in 1970, they are still very large, indicating the presence of a very broad range of values. Table 11. Descriptive statistics for the 1970 and 1993/97 coffee censuses ratios. Distributions of hectares of Pasture/hectares of Coffee and hectares of Other/hectares of Coffee 1970 1993/97 Average Pasture (ha) /Coffee (ha) 3.46 2.98 Std. Dev. 4.96 2.24 Coefficient of Variation 143.30% 75.20% Average Other (ha)/Coffee (ha) 1.66 2.51 Std. Dev. 5.52 3.11 Coefficient of Variation 332.30% 123.80% So far, the results presented treat the area in coffee and other land covers in absolute terms. However, it is also interesting to analyze how the proportions of each land cover have changed at the municipal level (Table 12). Table 12. Evolution of the proportion of the area in coffee,pasture, and other land covers 1970-1993/97 Coffee Pasture Other Municipalities Percentage Increase Increase Increase 11 2.10% Increase Increase Decrease 31 5.90% Increase Decrease Increase 189 36.10% Decrease Increase Increase 52 9.90% Decrease Decrease Increase 143 27.30% Decrease Increase Decrease 37 7.10% Increase Decrease Decrease 59 11.30% Decrease Decrease Decrease 1 0.20% The proportion of coffee increased in 290 municipalities (55.4%) (Table 12). This contrasts with the results from Figure 13, where the actual area planted in coffee grew in only 145 municipalities of the country. This means that although the total area planted in coffee is decreasing in most of the country, the proportion of the area planted in coffee for the municipality was higher in 1993/97 than in 1970 in 55.4% of the coffee growing municipalities. When the total area in other crops is analyzed, it increased only in 263
184 municipalities (51.4%). Interestingly, the proportion of the area in other crops had increased in 75.5% of the municipalities from 1970 to 1993/97. What these analyses about coffee and other crops suggest is that coffee farmers are relying more (measured as the proportion in each land cover) in coffee and other crops and less in pasture. This is the situation for more than 36% of the municipalities (Table 12), while the proportion of the municipality on coffee and pasture decreased and the percentage of other crops increased in 27.3% of the municipalities (Table 12). The proportion of devoted to pasture has decreased in 392 municipalities (75.0%). If one looks at the land-cover changes from 1970 to 1993/97 (Table 13), a similar pattern emerges. The land cover that proportionally has increased the most is other in 53.3% of the municipalities. Pasture is the land cover whose percentage has decreased the most in 59.1% of the municipalities. These results suggest that the landscape is becoming more agriculturally diverse, and potentially more spatially heterogeneous as there are more land covers per unit of area. Table 13. Major land-cover increase and decrease as a proportion 1970-1993/97 Major Land Cover Increase 1970-1993/97 Number of municipalities % Major Land Cover Decrease 1970-1993/97 Number of municipalities % Coffee 132 25.2 Coffee 96 18.4 Other 279 53.3 Other 56 10.7 Pasture 77 14.7 Pasture 309 59.1 Sugarcane 16 3.1 Sugarcane 38 7.2 Temporary 19 3.6 Temporary 24 4.6 Using a cluster analysis (k-means), I used the proportion of each land cover in each municipality in 1970 to classify it into one of four land-use systems. The average percentages on each land cover in any given municipality for each land-use type are presented in Figure 18. The relatively low standard deviation from these mean values
185 (Table 14) shows that in general terms, most municipalities are close to the average values. Then, the land cover values for 1993/97 were assigned to the four classes depicted in by calculating the Euclidean distance between the municipality percentages of each land cover and each land use system in Figure 18 as if these percentages represented coordinates in a 5-dimensional space, and selecting the land use type closest to each municipality. 21.4%24.5%18.3%18.0%17.8% 14.6 % 32.8%2.9%3.8%45.9% 40.8%34.9%4.7%3.2%16.6% Mainl y Pasture Coffee and Pasture Other and Pasture Mixe d 15.1%59.2%4.3%3.1%18.4% Coffee Pasture Sugar Cane Temporary Others Figure 18. Land-use systems (n=503). The percentages represent the average proportion of each land cover in a municipality belonging to any of the land-use systems. The spatial distribution of these land-use systems in 1970 and 1993/97 is shown in Figure 19. Overall, 51.9% of the municipalities maintained their land-use system since 1970 (Table 15). The most important changes were from â€œMainly pastureâ€ to â€œOther and pastureâ€ in 20.9% of the municipalities, from â€œMainly Pastureâ€ to â€œCoffee and Pastureâ€
186 in 7.6% of the municipalities, and from â€œCoffee and Pastureâ€ to â€œOther and Pastureâ€ in 6.2% of the municipalities (Figure 19 and Table 15). Table 14. Standard deviations for each land cover around the cluster means of the K-Means Cluster Analysis for 1970 Land Covers Mainly pasture Mainly coffee and pasture Mainly other and pasture Mixed system Coffee 7.5% 10.5% 8.8% 11.1% Pasture 9.7% 10.5% 11.9% 12.7% Sugar Cane 4.2% 4.2% 4.0% 13.3% Temporary 2.9% 4.1% 4.7% 15.9% Others 8.1% 7.3% 13.2% 8.3% The cluster of municipalities in the west-central section of the country under the â€œCoffee and pastureâ€ production system remained unchanged, which is something to be expected from the major coffee producing area of the country (Figure 19). The major changes, when comparing the breakdown of municipalities in the four land-use systems for 1970 and 1993/97 are the decrease of the proportion of municipalities in the â€œMainly pastureâ€ system from 47.3% in 1970 to 23.7% in 1993/97, and the increase of the percentage of municipalities in the â€œOther and pastureâ€ category from 15.1% in 1970 to 41.4% in 1993/97 (Table 15). Table 15. Evolution of land-use system types, 1970-1993/97. Percentage of municipalities. Land use system 1970 Land use system 1997 Mainly pasture Coffee and pasture Other and pasture Mixed Grand Total Mainly pasture 17.5% 2.6% 2.0% 1.6% 23.7% Coffee and pasture 7.6% 18.9% 1.0% 2.0% 29.4% Other and pasture 20.9% 6.2% 11.9% 2.4% 41.4% Mixed 1.4% 0.4% 0.2% 3.6% 5.6% Grand Total 47.3% 28.0% 15.1% 9.5% 100.0%
187 Figure 19. Evolution of land-use systems in Colombia 1970-1993/97 (n=503)
188 Because the most significant changes are from a land-use system where a homogeneous land cover predominates (Pasture) to a system where a very diverse land-cover group (Other) dominates, this suggests again the diversification of the Colombian coffee lands. The information from the FNC coffee censuses regarding land-cover change ends in 1997. To see if the trends towards diversification continued after 1997 I used the information provided by the extension agent. I obtained answers from 117 out of 559 municipalities, a response rate of 20.9%. I also received two questionnaires answered at the provincial level. On average, extension agents from these 117 municipalities have been in their municipality for 9 years (Figure 20). There are 47 municipalities which have extension agents who have served in that area from 1 to 7 years, 49 whose extension agents have served from 8 to 14 years, 16 municipalities have extension agents who have been practicing in their area from 15 to 21 years, and 5 who have served in the same area for more than 22 years. Therefore, the answers from the extension agents provide reliable information on coffee production and land-cover changes from the last 5 years. It is interesting to note that the spatial distribution of these municipalities shows 2 clusters (one in the west, one in the east) that correspond to the Departmental Coffee Growing committees that were willing to provide this information. The questionnaire asked extension agents which crops or land covers were becoming more important, where important was defined as crops that were increasing in area or that had appeared in the last few years. In the same way, they were asked which land covers were becoming less important, therefore providing information on the crops decreasing in area or that had disappeared from the landscape.
189 Figure 20. Time of service of the agricultural extension agents by municipality
190 The answers from these questions for the 117 municipalities report that there are 28 crops or land covers becoming less important and 33 crops or land covers becoming more important. Therefore, there are more crops appearing or expanding (more important) in the landscape than land-management practices contracting or disappearing (less important), suggesting landscape diversification. These answers also show that crop are becoming more important in a larger percentage of municipalities (Figure 21 and Figure 22). 0%20%40%60%PastureSugarcaneFruitsPlantainsCitrusCoffeeTomatoIllicitVegetablesCocoaCassava% municipalitie s Figure 21. Municipalities where crops are increasing in importance 1997-2002 Coffee is the land cover that is becoming less important in most municipalities (Figure 22), while pasture, sugarcane, and fruit orchards are becoming more important in more municipalities (Figure 21). It is also interesting to see that crops that are becoming important in some municipalities are becoming less important in others. Coffee, sugarcane, cassava, and cocoa are good examples of this situation. Overall, there are 23 crops or land covers that are becoming both more important and less important in different municipalities. There are 11 crops that are only increasing in importance,
191 0%20%40%60%CoffeePlantainsFruitsMaizeCassavaForestBeansCocoaTobaccoCitrusSugarcane% municipalitie s Figure 22. Municipalities where crops are decreasing in importance 1997-2002 and 6 crops only decreasing in importance. Pasture is the land cover that is becoming important in the largest number of municipalities (54.7%). This trend is consistent with periods when similar market crises took place and coffee prices were extremely low. In Colombia, coffee plots were replaced with pasture during these crises (Palacios 1980). Although it is not reported by a large number of municipalities, illicit crops, mostly coca and poppy, are becoming important in certain areas (11.1% of the municipalities). These crops are usually intercropped with coffee, using the presence of coffee bushes as camouflage so the illicit crops can be used to produce illegal drugs. This practice just started to become appear in recent years(Umaa 2003; Pealoza 2002), possibly sparked from the low prices associated with the coffee crisis, and this is one of the reasons why some members of the U.S. congress are interested in reinstating the International Coffee Agreement to improve coffee prices and prevent coffee farmers from switching to illegal crops (ICO 2003b).
192 If the number of crops increasing in importance is larger than the number of crops becoming less important in any given municipality, there are more crops being planted in the same area, suggesting a more diverse landscape. Agricultural diversity, as defined above, has increased or maintained in 70.1% of the municipalities (Figure 23). Using the net change in crops (increasing importance â€“ decreasing importance), I calculated the Moranâ€™s I (I=0.06) for the municipalities depicted in Figure 23. This value is not significantly different from the value that could be produced by a random pattern (=-0.009, =0.068, Z=1.015), making it difficult to identify the driving forces of diversification. Simultaneous Intensification and Disintensification of Crops It was not possible to obtain yield information for the same crops for all coffee growing departments (Table 16). Yield data for arracacha , cocoa, sugarcane, and technified corn from 9, 15, 14, and 9 departments respectively. The rest of the crops had productivity information in all coffee growing departments. The departments with more crops have more potential crop associations. (Table 16). Figure 24 presents yield evolution of these crops for 1988-2000 in the 16 Colombian coffee-growing departments. In general terms, there is a large variability in yield from year to year. However, many crops that exhibit smoother patterns. Some of the crops that are highly variable in one department are the crops that are less variable in another department. This is the case of cocoa, which exhibits a relatively smooth decline in Antioquia and large oscillations in yield values for departments such as Boyac and Cauca (Figure 24). Other crops like traditional corn tend to vary less from year to year in most departments (Figure 24).
193 Figure 23. Recent trends in agricultural diversity. Agricultural diversity refers to the number of agricultural crops planted in any given municipality.
194 Table 16. Crops with annual yield information (1988-2000) in the coffee growing departments. A= Arracacha; B=Beans, C=Cacao, M= Traditional Corn, P= Plantains, S= Sugarcane, TC= Technified Corn, Y=Yuca Department Crops Number of possible associations per crop Antioquia A, B, C, M, P, S, Y 6 Boyac A, B, C, M, P, S, Y 6 Caldas B, C, M, P, S, TC, Y 6 Cauca B, C, M, P, S, TC, Y 6 Cesar B, C, M, P, S, TC, Y 6 Cundinamarca A, B, C, M, P, S, Y 6 Huila A, B, C, M, P, S, TC, Y 7 La Guajira B, M, P, Y 3 Magdalena B, C, M, P, Y 4 Nario B, C, M, P, S, Y 5 N. Santander A, B, C, M, P, S, Y 6 Quindo B, C, M, P, S, TC, Y 6 Risaralda A, B, C, M, P, S, TC, Y 7 Santander A, B, C, M, P, S, TC, Y 7 Tolima A, B, C, M, P, S, TC, Y 7 Valle del Cauca A, B, C, M, P, S, TC, Y 7 The cumulative variation in yield for each crop and department is presented in Table 17. The lower the values in the table, the less year-to-year variability in yield for that particular crop. The provinces where yields are more variable are La Guajira (average cumulative year-to-year variation 161.7%) and Caldas (average cumulative year-to-year variation 145.1%), and the departments where crop yields are less variable are Huila (average cumulative year-to-year variation 77.6%) and Antioquia (average cumulative year-to-year variation 79.8%) (Table 17). The most variable crop in terms of yield is plantains (average cumulative year-to-year variation 143.5%) and the least variable crop is technified corn (average cumulative year-to-year variation 86.1%). The crop that exhibits the largest cumulative variation in yield for the period of study is cacao in the department of Boyac, while technified corn in the province of Huila is the crop that shows the smallest variability (Table 17).
195 Antioquia 40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Boyac30%40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Figure 24. Evolution of crop yields from 1988 to 2000 in the coffee growing departments. For each crop, values have been converted to percentages by dividing them by the maximum yield in the period of study.
196 Caldas 30%40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Cauca40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Figure 24. Continued.
197 Cesar 50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Cundinamarca 20%30%40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Figure 24. Continued.
198 Huila 50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca La Guajira40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Figure 24. Continued.
199 Magdalena 20%30%40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Nario 30%40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Figure 24. Continued.
200 N. Santander40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Quindo40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Figure 24. Continued.
201 Risaralda40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Santander10%20%30%40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Figure 24. Continued.
202 Tolima30%40%50%60%70%80%90%100%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Valle del Cauca40.0%50.0%60.0%70.0%80.0%90.0%100.0%1988199019921994199619982000Year% of maximum recorded yield Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Figure 24. Continued.
203 Table 17. Cumulative year-to-year variations (regardless of sign) presented as a percentage for the coffee growing departments (1988-2000). The lower the values in the table, the less year-to-year variability in yield. Arracacha Cocoa Sugarcane Beans Tech. Corn Trad. Corn Plantains Yuca Antioquia 94.7% 60.2% 41.5% 56.5% N.A. 75.7% 130.0% 99.7% Boyac 160.3% 289.8% 82.3% 104.1% N.A. 51.3% 137.6% 137.9% Caldas N.A. 131.5% 144.4% 130.9% 119.9% 134.8% 205.0% 149.3% Cauca N.A. 144.5% 125.2% 79.8% 87.9% 109.8% 150.1% 222.8% Cesar N.A. 154.4% 84.9% 91.3% 117.2% 58.9% 104.6% 150.3% Cundina-marca 201.2% 186.6% 90.9% 285.8% N.A. 55.7% 115.7% 62.2% Huila 42.9% 47.6% 125.0% 96.2% 23.9% 60.7% 128.4% 95.8% La Guajira N.A. N.A. N.A. 91.6% N.A. 173.0% 238.4% 143.8% Magdalena N.A. N.A. N.A. 115.3% N.A. 113.7% 172.2% 79.1% Nario N.A. 208.6% 88.8% 113.1% N.A. 151.2% 100.6% 199.5% N. Santander 162.8% 110.7% 112.2% 202.7% N.A. 98.2% 108.1% 129.1% Quindo N.A. 126.6% 34.4% 107.9% 110.1% 178.2% 204.6% 128.7% Risaralda 113.4% 133.5% 119.5% 96.1% 110.6% 171.8% 136.2% 87.1% Santander 170.2% 106.7% 78.5% 66.9% 75.0% 57.9% 76.4% 109.4% Tolima 50.3% 72.7% 85.5% 124.7% 86.0% 76.1% 177.5% 174.9% Valle del Cauca 75.3% 84.2% 80.6% 85.6% 44.0% 74.3% 110.7% 118.4% It is worth noting that a significant part of the variability in yearly crop yields can be explained by changing weather patterns. For example, during cold El Nio Southern Oscillation (ENSO) events, rainfall patterns in Colombia change, and certain areas receive more rain while others receive less. This obviously has a direct effect on crop yields. Between 1988-2000 there were 3 cold ENSO phases in 1988, 1991-92, and 1997-98 (Caviedes 2001). However, when the annual rainfall time series from representative weather stations in the coffee growing provinces is included in the analysis (1 station per province), only 11% of the crop-yield time series at the departmental level show a statistically significant correlation with rainfall patterns. Although this suggests little influence of climate on crop yields, it is more likely that yield information aggregated at the departmental level smoothes out the influence of climate, as some locations in the
204 same province would have higher yields and others lower yields. This indicates that the analysis of the influence of climate in annual yields would benefit from either more weather station records for each province or a finer spatial scale in the annual yield data. At first glance it is hard to discern any patterns (i.e., crops with similar or opposite behavior) in the yield evolution graphs presented in Figure 24. However, it is evident that in some departments (e.g., Boyac, Caldas, Cauca, Huila, La Guajira) there were major changes in yields between 1992-1994. This corresponds with the moment when there was a major neoliberal economic reform in the country, and import taxes and agricultural subsidies disappeared (BANREP N.D). This probably had an impact in certain departments, where some crops could not compete (i.e., corn) with the cheap imports from other countries. When the yield time-series information is analyzed using Spearman-Rank correlations (Table 18), it is evident that for each coffee growing province there are not only crops that exhibit similar behavior (e.g., crops whose yields increase or decrease together) but also crops that show opposite behavior (e.g., crops whose yields increase while the yields of other crops decrease). Although there are instances of both cases for most provinces (10 departments), it is more common to find more crops with oppositeyield behavior (7 departments) than more crops with similar crop-yield behavior (6 departments). There is one department where the number of crops with similar-yield behavior is the same as the number of crops with opposite-yield behavior, and two provinces where there are no statistically significant correlations among yields for different crops.
205 Table 18. Correlation coefficients among different crop yields through time (1988-2000). * Significant at =0.05, ~ Significant at =0.10 Department (Number of crops with full records) Crops with similar behavior Spearman correlation Crops with opposite behavior Spearman correlation Antioquia (7) Arracacha-Cacao Arracacha-Plantains 0.58 * 0.80 * Arracacha-Beans Beans-Cacao -0.49 ~ -0.76 * Boyac (7) Arracacha-Plantains Beans-Plantains Beans-Sugarcane Plantains-Sugarcane Plantains-Yuca 0.49 ~ 0.89 * 0.72 * 0.70 * 0.49 ~ Caldas (7) Cacao-Yuca 0.73 * Beans-Sugarcane Plantains-Tech. Corn -0.65 * -0.56 * Cauca (7) Cacao-Trad. Corn Sugarcane-Yuca -0.59 * -0.53 ~ Cesar (7) Beans-Plantains Beans-Tech. Corn 0.70 * 0.61 * Beans-Cacao -0.61 * Cundina-marca (7) Beans-Yuca Plantains-Yuca Trad. Corn-Yuca 0.60 * 0.66 * 0.62 * Sugarcane-Yuca -0.53 ~ Huila (8) Arracacha-Plantains Arracacha-Yuca Beans-Sugarcane Beans-Tech. Corn Sugarcane-Tech. Corn 0.65 * 0.65 * 0.77 * 0.75 * 0.71 * Arracacha-Sugarcane Arracacha-Beans Beans-Yuca -0.50 ~ -0.52 ~ -0.55 ~ La Guajira (4) Magdalena (4) Nario (6) Trad. Corn-Yuca 0.60 * Beans-Cacao Beans-Plantains -0.56 * -0.57 * N. Santander (7) Cacao-Plantains -0.68 * Quindo (7) Beans-Cacao Plantains-Sugarcane Sugarcane-Tech. Corn 0.71 * 0.62 * 0.56 * Beans-Plantains Beans-Sugarcane Cacao-Plantains Cacao-Sugarcane Trad. Corn-Yuca -0.78 * -0.69 * -0.58 * -0.82 * -0.49 ~
206 Table 18. Continued Department (Number of crops with full records) Crops with similar behavior Spearman correlation Crops with opposite behavior Spearman correlation Risaralda (7) Beans-Sugarcane Trad. Corn-Yuca Plantains-Yuca -0.60 * -0.57 * -0.58 * Santander (8) Beans-Plantains Beans-Sugarcane Beans-Tech. Corn Beans-Trad. Corn Beans-Yuca Plantains-Tech. Corn Plantains-Yuca Sugarcane-Trad. Corn Sugarcane-Tech. Corn Sugarcane-Yuca Tech. Corn-Trad. Corn Tech. Corn-Yuca Trad. Corn-Yuca 0.67 * 0.81 * 0.82 * 0.75 * 0.80 * 0.70 * 0.68 * 0.67 * 0.51 ~ 0.74 * 0.70 * 0.74 * 0.76 * Cacao-Trad. Corn -0.55 ~ Tolima (8) Arracacha-Beans Arracacha-Tech. Corn Arracacha-Trad. Corn Beans-Tech. Corn Beans-Trad. Corn Cacao-Plantains Trad. Corn-Tech. Corn 0.71 * 0.80 * 0.90 * 0.84 * 0.80 * 0.49 ~ 0.85 * Arracacha-Sugarcane Beans-Sugarcane Sugarcane-Tech. Corn Sugarcane-Trad. Corn -0.68 * -0.64 * -0.65 * -0.49 ~ Valle del Cauca (8) Cacao-Sugarcane Trad. Corn-Yuca 0.62 * 0.49 ~ Cacao-Trad. Corn Cacao-Yuca Sugarcane-Yuca -0.60 * -0.86 * -0.51 ~ The fact that there are more departments where there are more crops with opposite behavior shows that the intensification of some crops is accompanied by the disintensification of other crops. To see if these correlations could be generated by a random process, I simulated random yield time series, and calculated the 95% confidence interval for the mean. The mean correlation for a random process is 0, the lower confidence limit is -0.345, and the upper confidence limit is 0.372. Therefore, it is clear
207 that the correlation coefficients presented in Table 18 are not likely the result of a random process. There are certain crops that exhibit mostly positive correlations (e.g., similar yield behavior) like arracacha (8 correlations), beans (16 correlations), and plantains (13 correlations) (Table 19). Others, like cacao (10 correlations) and sugarcane (12 correlations) show mostly negative correlations (e.g., opposite yield behavior). Table 19. Number of statistically significant correlations among crop yields (yearly data). +,-; Positive or negative correlation. (n) Number of provinces where the relationship is significant (<=0.10) Arracacha Beans Cacao Plantains Sugarcane Tech. Corn Trad. Corn Yuca Arracacha + (1) (2) + (1) + (3) (2) + (1) + (1) + (1) Beans + (1) (3) + (3) (2) + (3) (4) + (4) + (2) + (2) (1) Cacao + (1) (2) + (1) (1) (3) + (1) (1) Plantains + (2) + (1) (1) + (3) (1) Sugarcane + (3) (1) + (1) (1) + (1) (3) Tech. Corn + (1) + (1) Trad. Corn + (4) (2) Overall, there are 43 crop associations that exhibit statistically significant positive correlations and 30 crop associations that show statistically significant negative correlations (Table 19). When one looks if there are certain crops more associated with positive (i.e., similar yield behavior) than negative (i.e., opposite yield behavior) correlations, with the exception of sugarcane and cacao, all the crops show more positive than negative
208 statistically significant correlations (Table 20). Overall, significant correlations account for more than 15% of the total possible crop associations for each crop (Table 20). Table 20. Number and percentage of statistically significant correlations for each crop (annual yields) * Calculated from the potential number of associations per crop in Table 16 Number of possible crop associations * Positive correlations % Negative correlations % Total correlations % Arracacha 59 8 13.5 4 6.8 12 20.3 Beans 95 16 16.8 12 12.6 28 29.4 Cacao 92 5 5.4 10 10.9 15 16.3 Plantains 95 13 13.7 6 6.3 19 20.0 Sugarcane 88 11 12.5 12 13.6 23 26.1 Tech. Corn 59 11 18.6 2 3.4 13 22.0 Trad. Corn 95 9 9.5 6 6.3 15 15.8 Yuca 95 13 13.7 8 8.4 21 22.1 When a 3-year moving average is considered for yields, the number of statistically significant correlations as well as the magnitude of the correlation coefficients increases (Table 21). This rise in statistically significant correlations is caused by the fact that some of the year-to-year variability caused by phenomena like pests and diseases is smoothed out. There is some evidence that some crops might be changing at coarser time scales. Crop associations like Arracacha-Yuca in N. Santander and Beans and Traditional Corn in Valle del Cauca do not show statistically significant correlations at the yearly time scale, while they are statistically significant at the 3-yr moving average time scale. The opposite situation may also occur. There are some crops that might be changing at a shorter time scale than the 3-yr moving average, and crop associations that were statistically significantly correlated with the annual yield are not statistically significantly correlated using the 3-yr moving average. Examples of this type of crop associations includes Beans-Plantains in Nario and Traditional Corn-Yuca in Quindo.
209 Table 21. Correlation coefficients among different crop yields through time (3-yr moving average), 1988-2000. * Significant at =0.05, ~ Significant at =0.10 Department (Number of crops with full records) Crops with similar behavior Spearman correlation Crops with opposite behavior Spearman correlation Antioquia (7) Arracacha-Cacao Arracacha-Plantains Cacao-Plantains 0.82 * 0.86 * 0.76 * Arracacha-Beans Beans-Cacao Beans-Plantains Plantains-Sugarcane -0.76 * -0.89 * -0.88 * -0.57 ~ Boyac (7) Beans-Plantains Beans-Sugarcane Plantains-Sugarcane Cacao-Sugarcane Cacao-Yuca Sugarcane-Yuca 0.97 * 0.70 * 0.60 ~ 0.62 * 0.58 ~ 0.74 * Caldas (7) Beans-Tech. Corn Cacao-Sugarcane Cacao-Yuca Plantains-Yuca 0.64 * 0.57 ~ 0.80 * 0.61 * Beans-Cacao Beans-Sugarcane Sugarcane-Tech. Corn -0.66 * -0.77 * -0.79 * Cauca (7) Sugarcane-Tech. Corn 0.76 * Cacao-Trad. Corn Plantains-Tech. Corn Sugarcane-Yuca Tech. Corn-Yuca -0.74 * -0.55 ~ -0.77 * -0.57 ~ Cesar (7) Beans-Plantains Beans-Sugarcane Beans-Tech. Corn Beans-Trad. Corn Plantains-Sugarcane Plantains-Tech. Corn Plantains-Trad. Corn Sugarcane-Tech. Corn Sugarcane-Trad. Corn 0.85 * 0.81 * 0.73 * 0.63 * 0.79 * 0.55 ~ 0.71 * 0.71 * 0.58 ~ Beans-Cacao Cacao-Plantains Cacao-Sugarcane Cacao-Tech. Corn Trad. Corn-Yuca -0.75 * -0.67 * -0.88 * -0.64 * -0.65 * Cundina-marca (7) Arracacha-Plantains Arracacha-Yuca Beans-Trad. Corn Cacao-Sugarcane Plantains-Yuca Trad. Corn-Yuca 0.63 * 0.57 ~ 0.65 * 0.76 * 0.96 * 0.54 ~ Arracacha-Beans Arracacha-Cacao Arracacha-Sugarcane Cacao-Yuca Plantains-Sugarcane Sugarcane-Yuca -0.54 ~ -0.53 ~ -0.71 * -0.60 * -0.74 * -0.82 *
210 Table 21. Continued Department (Number of crops with full records) Crops with similar behavior Spearman correlation Crops with opposite behavior Spearman correlation Huila (8) Arracacha-Plantains Arracacha-Yuca Beans-Sugarcane Beans-Tech. Corn Plantains-Yuca Sugarcane-Tech. Corn 0.85 * 0.72 * 0.90 * 0.98 * 0.53 ~ 0.89 * Beans-Yuca Arracacha-Beans Arracacha-Sugarcane Arracacha-Tech. Corn Sugarcane-Yuca Tech. Corn-Yuca -0.91 * -0.67 * -0.71 * -0.64 * -0.90 * -0.87 * La Guajira (4) Beans-Plantains -0.59 ~ Magdalena (5) Plantains-Yuca 0.93 * Cacao-Plantains Cacao-Yuca -0.70 * -0.73 * Nario (6) Beans-Sugarcane Cacao-Yuca Plantains-Trad. Corn Trad. Corn-Yuca 0.73 * 0.69 * 0.75 * 0.70 * Beans-Cacao Cacao-Sugarcane -0.55 ~ -0.75 * N. Santander (7) Arracacha-Yuca Plantains-Sugarcane 0.57 ~ 0.63 * Arracacha-Trad. Corn Cacao-Sugarcane Cacao-Yuca -0.56 ~ -0.64 * -0.88 * Quindo (7) Beans-Cacao Beans-Trad. Corn Plantains-Sugarcane Sugarcane-Tech. Corn 0.69 * 0.66 * 0.82 * 0.62 * Beans-Plantains Beans-Sugarcane Cacao-Plantains Cacao-Sugarcane Cacao-Tech. Corn Tech. Corn-Yuca -0.83 * -0.62 * -0.84 * -0.94 * -0.57 ~ -0.55 ~ Risaralda (8) Arracacha-Cacao Plantains-Trad. Corn Sugarcane-Tech. Corn Tech. Corn-Yuca 0.62 * 0.77 * 0.60 ~ 0.55 ~ Arracacha-Yuca Beans-Plantains Beans-Sugarcane Beans-Trad. Corn Cacao-Tech. Corn Cacao-Yuca Plantains-Yuca Trad. Corn-Yuca -0.63 * -0.56 ~ -0.72 * -0.63 * -0.70 * -0.64 * -0.84 * -0.68 *
211 Table 21. Continued. Department (Number of crops with full records) Crops with similar behavior Spearman correlation Crops with opposite behavior Spearman correlation Santander (8) Beans-Plantains Beans-Sugarcane Beans-Tech. Corn Beans-Trad. Corn Beans-Yuca Plantains-Sugarcane Plantains-Tech. Corn Plantains-Yuca Sugarcane-Tech. Corn Sugarcane-Trad. Corn Sugarcane-Yuca Tech. Corn-Trad. Corn Tech. Corn-Yuca Trad. Corn-Yuca 0.85 * 0.89 * 0.91 * 0.79 * 0.95 * 0.61 * 0.86 * 0.84 * 0.80 * 0.87 * 0.83 * 0.74 * 0.94 * 0.78 * Beans-Cacao Cacao-Sugarcane Cacao-Tech. Corn Cacao-Trad. Corn Cacao-Yuca -0.55 ~ -0.66 * -0.61 * -0.77 * -0.59 ~ Tolima (8) Arracacha-Beans Arracacha-Tech. Corn Arracacha-Trad. Corn Beans-Tech. Corn Beans-Trad. Corn Plantains-Sugarcane Trad. Corn-Tech. Corn 0.87 * 0.84 * 0.91 * 0.97 * 0.98 * 0.91 * 0.96 * Arracacha-Plantains Arracacha-Sugarcane Beans-Plantains Beans-Sugarcane Plantains-Tech. Corn Plantains-Trad. Corn Sugarcane-Tech. Corn Sugarcane-Trad. Corn -0.97 * -0.90 * -0.79 * -0.75 * -0.77 * -0.82 * -0.77 * -0.79 * Valle del Cauca (8) Beans-Cacao Beans-Sugarcane Cacao-Sugarcane Tech. Corn-Yuca 0.75 * 0.69 * 0.85 * 0.72 * Beans-Trad. Corn Cacao-Trad. Corn Cacao-Yuca Plantains-Tech. Corn Sugarcane-Trad. Corn Sugarcane-Yuca -0.62 * -0.69 * -0.78 * -0.53 ~ -0.53 ~ -0.62 * As with the annual yield data, these correlation coefficients were compared to those generated by a random process. The mean of the simulated random correlation coefficients is 0.00, the lower 95% confidence limit is -0.500, and the upper 95% confidence limit is 0.513. Because all the coefficients presented in Table 21 are higher or lower than these confidence limits, it can be concluded that the observed correlations among crops are not likely the result of a random process.
212 The trends reported for the annual yield data are maintained in the most part. There are 9 departments where there are more crops with opposite behavior (i.e., negative correlation coefficients) than with similar behavior (i.e., positive correlation coefficients), 5 where there are more crops with similar behavior than with opposite behavior, and 2 provinces with an equal number of crops with opposite and similar behavior. With the annual yield data, there were 7 departments where there were more crop associations with opposite-yield behavior and 6 provinces with more crop associations with similar-yield behavior (Table 18). Therefore, when the 3-yr moving average for crop yields is used the evidence of intensification of certain crops with the simultaneous disintensification of other crops is present in a larger area (i.e., more departments) of the Colombian coffee lands than when the annual-yield data are used. Overall, there are 74 crop associations with statistically significant positive correlations and 67 crops crop associations with statistically significant negative correlation coefficients (Table 22). This contrasts with the results of the annual-yield data, where only 43 and 30 crop associations had a statistically significant positive or negative correlation coefficient respectively (Table 20). The number of crops associations with opposite-yield behavior more than doubles when the coarser time scale is considered. The percentage of significant crop associations increases when the 3-yr moving average, with the total number of correlations representing more than 30% of the total number of crop associations for all crops excepting traditional corn (Table 23). Although both positive and negative correlations increase when using the 3-yr moving average, the number of negative correlations (i.e., crops with opposite-yield behavior) increases much more than the number of positive correlations (Table 20 and Table 23)
213 Table 22. Number of statistically significant correlations among crop yields (3-yr moving average). +,-; Positive or negative correlation. (n) Number of provinces where the relationship is significant (<=0.10) Arracacha Beans Cacao Plantains Sugarcane Tech. Corn Trad. Corn Yuca Arracacha + (1) (3) + (2) (1) + (3) (1) (3) + (1) (1) + (1) (1) + (3) (1) Beans + (2) (5) + (3) (4) + (6) (4) + (5) + (5) (2) + (1) (1) Cacao + (1) (3) + (4) (5) (4) (3) + (3) (5) Plantains + (6) (2) + (2) (3) + (3) (1) + (5) (1) Sugarcane + (5) (3) + (2) (2) + (2) (4) Tech. Corn + (2) + (3) (3) Trad. Corn + (3) (2) Table 23. Number and percentage of statistically significant correlations for each crop (3-yr moving average) Possible crop associations * Positive correlations % Negative correlations % Total correlations % Arracacha 59 12 20.3 11 18.6 23 38.9 Beans 95 23 24.2 19 20.0 42 44.2 Cacao 92 12 13.0 17 18.5 29 31.5 Plantains 95 23 24.2 15 15.8 38 40.0 Sugarcane 88 25 28.4 23 26.1 48 54.5 Tech. Corn 59 18 30.5 14 23.7 32 54.2 Trad. Corn 95 16 16.8 11 11.6 27 28.4 Yuca 95 20 21.0 17 17.9 21 38.9 * Calculated from the potential number of crop associations in Table 16 The number of statistically significant correlations can also be looked from the perspective of each coffee-producing department (Table 24). The overall patterns (i.e., comparison of positive and negative correlations for each department) are maintained in 10 departments when the annual yield and the 3-yr moving average are used. Patterns change in Cundinamarca and Huila, where there are more positive correlations than negative correlations using the annual yield data, and an equal number of positive and
Table 24. Comparison of number of statistically significant crop relations for the annual yield time series and the 3-yr moving average time series Positive Correlation Negative Correlation Original data 3-yr average Original data 3-yr average Department (possible crop-relation pairs) Relations with similar behavior % Relations with Similar behavior % Relations with opposite behavior % Relations with opposite behavior % Antioquia (21) 2 9.5 3 14.3 2 9.5 4 19.0 Boyac (21) 5 23.8 6 28.6 0 0.0 0 0.0 Caldas (21) 1 4.8 4 19.0 2 9.5 3 14.3 Cauca (21) 0 0.0 1 4.8 2 9.5 4 19.0 Cesar (21) 2 9.5 9 42.9 1 4.8 5 23.8 Cundinamarca (21) 3 14.3 6 28.6 1 4.8 6 28.6 Huila (28) 5 17.9 6 21.4 3 10.7 6 21.4 La Guajira (6) 0 0.0 0 0.0 0 0.0 1 16.7 Magdalena (10) 0 0.0 1 10.0 0 0.0 2 20.0 Nario (15) 1 6.6 4 26.6 2 13.3 2 13.3 N. Santander (21) 0 0.0 2 9.5 1 4.8 3 14.3 Quindo (21) 3 14.3 4 19.0 5 23.8 6 28.6 Risaralda (28) 0 0.0 4 14.3 3 10.7 8 28.6 Santander (28) 13 46.4 14 50.0 1 3.6 5 17.9 Tolima (28) 7 25.0 7 25.0 4 14.3 8 28.6 Valle del Cauca (28) 2 7.1 4 14.3 3 10.7 6 21.4 214
215 negative correlations using the 3-yr moving average, in Caldas and Nario, where there are more negative than positive correlations when the annual yield is used, and more negative correlations when the 3-yr moving average yield data is used, in La Guajira, where there are the same number of positive and negative correlations when the annual yields are used, and a larger number of negative correlations when the 3-yr moving average is used, and Tolima, where when using the annual yield data the number of positive correlations is larger than the number of negative correlations, and the pattern is exactly the opposite behavior when the 3-yr yield data are used. As the time scale gets coarser, the results from the correlation analyses show that in ten departments the difference between the number of positive significant correlations and the number of negative significant correlations get smaller, suggesting that the pattern of intensification of some crops with the simultaneous disintensification of other crops occurs at larger time scales, and possibly indicating agricultural change trends. In summary, the correlation analyses indicate the following: Simultaneous intensification and disintensification of agricultural production is taking place in the Colombian coffee lands, as the production of certain crops intensifies as others disintensify. The significant crop relations change from department to department. Crop associations significant in certain departments are not significant in other areas of the country. The same crop associations can have positive or negative correlations in different areas of the country, indicating how different factors (i.e., economic, demographic, environmental, institutional) influence land-use decisions. As the time scale gets coarser, the evidence of simultaneous intensification and disintensification is stronger and indicates how agricultural change may be taking place The first three facts indicate that, although intensification and disintensification are taking place, the way farmers choose to intensify the production of certain products and
216 disintensify the production of other products changes from region to region, as different socioeconomic, environmental, demographic, and institutional factors interact. The last point provides a relatively easy way of looking at possible longer-term agricultural changes taking place in a region. Factors Accompanying Coffee Production Intensification in Different Areas of Colombia Principal Components Analysis The results of the principal components analysis using the 10 variables 3 postulated as key variables driving the commercial intensification of agriculture by Boserup, Chayanov, Pingali and Rosegrant, and Veldkamp and Lambin show that the first four principal components (PC) on these 10 variables account for 82.98% of the variability in this dataset (Table 25). I selected these four principal components based on their individual contribution, selecting those PCs that accounted for 10% or more of the variability in the original 10 variables. Table 25. Principal component analysis results on the factors accompanying agricultural intensification (Eigenvalues) Principal Component Eigenvalue Individual Percent Cumulative Percent 1 3.341 33.41 33.41 2 2.199 21.99 55.40 3 1.696 16.96 72.36 4 1.061 10.61 82.98 5 0.436 4.36 87.33 6 0.394 3.94 91.28 7 0.382 3.82 95.09 8 0.217 2.17 97.26 9 0.185 1.85 99.11 10 0.089 0.89 100 Using the standardized Eigenvectors in order to compare the relative contribution of each variable to each principal component, it can be seen that PC1 is mostly the result 3 Appendix B presents maps of these variables and a brief description of their spatial patterns.
217 of the two rainfall variables, the location variables, and to a lesser extent the institutional variables (Table 26). In PC2, the most important variables are socioeconomic and demographic (average family age, dependents per household, rural population density) and accessibility as measured by the distance to the closest town (Table 26). The most important variables in PC3 are the institutional factors, with location and accessibility contributing to a lesser extent (Table 26). Finally, in PC4, the variables that contribute the most are rural population density, the number of dependents per household, the average family age, and accessibility (Table 26). These four PCs can be interpreted broadly as summarizing environmental (PC1), socioeconomic (PC2), institutional (PC3), and socioeconomic/demographic (PC4) factors. PC2 and PC4 are very similar in terms of their composition, but in PC4 rural population density makes a larger contribution. Table 26. Principal component analysis standardized Eigenvectors Variables PC1 PC2 PC3 PC4 Longitude -0.445 -0.147 0.137 0.202 Latitude -0.435 -0.026 0.294 -0.072 Average family Age 0.025 -0.564 -0.038 0.362 Financial Institutions 0.227 -0.002 0.634 -0.062 State Institutions 0.251 0.036 0.628 -0.057 Dependents per household -0.114 0.504 -0.182 -0.381 Rural Population Density 0.146 -0.336 -0.003 -0.717 Minimum rainfall in 2 driest months 0.498 -0.058 -0.179 0.157 Maximum rainfall in 2 wettest months 0.463 0.103 -0.102 0.108 Distance to closest town 0.032 0.526 0.145 0.347 The Principal component factor loadings (i.e., correlation between original variables and PCs) simply stress the points made earlier about the interpretation of each PC (Table 27). These loadings make more clear the contribution of some of the variables and the interpretation of the four principal components. For example, an analysis of Table 26 and Table 27 emphasize the fact that the environmental factors are the most important in PC1 and the institutional variables are the critical factors in PC3.
218 Table 27. Degree of correlation between variables and principal components Variables PC1 PC2 PC3 PC4 Longitude -0.814 -0.218 0.178 0.208 Latitude -0.795 -0.039 0.384 -0.074 Average family Age 0.046 -0.837 -0.050 0.373 Financial Institutions 0.416 -0.003 0.826 -0.063 State Institutions 0.458 0.053 0.817 -0.059 Dependents per household -0.208 0.747 -0.237 -0.392 Rural Population Density 0.267 -0.499 -0.003 -0.739 Minimum rainfall in 2 driest months 0.910 -0.086 -0.233 0.162 Maximum rainfall in 2 wettest months 0.845 0.152 -0.133 0.112 Distance to closest town 0.059 0.781 0.189 0.357 When the first two principal components for each one of the coffee growing municipalities (accounting for more than 50% of the variability of the original 10 variables) are plotted in a scatter plot it is clear that there are no clusters of municipalities (Figure 25). There are a few municipalities with extremely high values of PC1. These tend to be departmental capitals or very large and important municipalities. The fact that this kind of municipalities exhibits a large value for PC1 is the result that they tend to have a high number of state and financial institutions, and therefore â€œinflateâ€ the PC1 values that should correspond to mostly environmental factors. There are also some high and low values along the PC2 axis. Although it is harder to come up with a reasonable explanation for some municipalities having extreme PC2 values, it probably has to do with a large number of dependents per household, and a long distance to the closest municipalities (positive PC2 values) and municipalities with older families with extremely high population densities (extreme negative PC2 values). The PC values derived for each one of the coffee growing municipalities (PC1, PC2, PC3 and PC4) were used as independent variables in a multivariate linear regression model with the area in intensive coffee per municipality as the dependent variable. The
219 spatial distribution of these independent variables is analyzed in more detail in Appendix B. The R 2 from this regression was 0.161 (Adj. R 2 = 0.155). The variables that explain the most variability are PC1 and PC3 (entered first in the stepwise forward regression). The regression coefficients are statistically different from zero for PC1, PC2 and PC3. -6.0-4.0-2.00.02.04.06.08.010.0-3.0-2.0-1.00.01.02.03.04.05.06.0PC1PC2 Figure 25. Scatter plot of PC1 and PC2 values for the coffee growing municipalities The standardized coefficients suggest that PC1 is the variable that contributes the most in explaining the area in intensive coffee (Std. Coeff. 1.26), followed by PC3 (Std. Coeff. -1.21), and with a much smaller contribution from PC2 (Std. Coeff. 0.126). This means that environmental and institutional factors have the strongest relationship with the area in intensive coffee at the municipal scale. Although this regression model yields normally distributed (i.e., random) regression residuals, when these are plotted on a map, it is clear that the residuals are spatially autocorrelated (Figure 26). The Moranâ€™s I value for this spatial pattern is 0.210, and the Z-score 7.513 (= -0.00186, =0.0282). In
220 Figure 26. Principal component multivariate linear regression residuals (no interactions among independent variables)
221 Figure 27. Principal component multivariate linear regression residuals (interactions among independent variables)
222 municipalities with positive residuals, the estimated area in intensive coffee is lower than the actual value, while in municipalities with negative residuals the regression model is overestimating the area in intensive coffee. On average, the area in intensive coffee is underestimated by 446.8 hectares. The standard deviation of the residuals is 1530.9 hectares, with maximum overestimation of the area in intensive coffee of 4475.3 ha, and a maximum underestimation of the area in intensive coffee of 10051.4 ha. When the regression model is run allowing the variables to interact (e.g., PC1*PC2), the R 2 goes up to 0.250 (Adj. R 2 = 0.236), which is better than in the case where the independent variables do not interact, but still low. The most important variables have not changed from the regression model without interactions (PC1 and PC3), followed by the PC1*PC3 interaction. The regression coefficients for PC1, PC2, PC3, PC1*PC2, PC1*PC4, PC2*PC3 and PC2*PC4 are statistically significant different from zero. The standardized coefficients indicate that PC2*PC3 is the variable that contributes the most (Std. Coeff. 1.094), followed by PC3 (Std. Coeff. -0.745), PC1 (Std. Coeff. -0.471), PC2 (Std. Coeff. -0.249), PC1*PC2 (Std. Coeff. 0.223), PC2*PC4 (Std. Coeff. -0.171), and finally PC1*PC4 (Std. Coeff. 0.141). As in the regression without considering the interaction of the independent variables, the regression residuals are random. However, when they are plotted on a map (Figure 27 above) it is clear that there are clusters of low and high residuals, and their patterns are very similar to those of the residuals from the regression not including the interactions. The Moranâ€™s I for this spatial pattern is 0.2152 and the Z-score is 7.702 (= -0.00186, =0.0282). Therefore, it is unlikely that this spatial pattern is caused by a random process. The area in intensive coffee is underestimated in the main coffee
223 growing area in the center of the country (e.g., positive residuals) while it is underestimated in the southernmost coffee growing municipalities as well as in the north-eastern part of the country. The average residual for the model considering interacting independent variables is 403.5 ha, which is slightly less than in the model not considering the interactions. The standard deviation of the residuals from the regression with interactions decreases slightly (=1403.9 ha) when compared to that of the regression model without interactions, the maximum overestimation increases to 4737.0 ha., and the maximum underestimation decreases to 9359.6 ha. Although in both cases the regression coefficients are low, these results highlight several important issues. First, at the municipal scale, environmental and institutional forces seem to be the most important factors related to the area in intensive coffee. This contrasts with the models of agricultural intensification at the household level, where socioeconomic forces are the key factors driving the process of intensification. In second place, although introducing the interactions between the independent variables increases the adjusted R 2 values from 0.155 to 0.236 (a 52% increase), thus increasing the explanatory power of the regression model, still the most important variables are the same (PC1, PC3), and interacting variables are entered last (with the exception of PC1*PC3 which is the third variable entered in the stepwise multiple linear regression) and contribute less to the regressionâ€™s correlation coefficient. Finally, it is very important to mention that despite the fact that the regression residuals are randomly distributed, once they are plotted on a map it is clear that there are clusters of high and low residuals. Therefore this means that there are variables and processes that cause these spatial patterns that are not captured in the multiple regression models, and that a linear model
224 does not necessarily is the best way to describe the relationship between the PCs and the area in intensive coffee at the municipal level. Multiple Linear Regression Analysis The results of using a stepwise regression using the 10 variables selected based on models of intensification and commercialization of agriculture at the household level as independent variables and the area in intensive coffee per municipality as the dependent variable for the entire coffee growing area indicate that about 37% of the variability in the area in intensive coffee per municipality is related to the variability of these 10 variables (Table 28). The top three variables selected in every model are the same ones that are more closely related to the area in intensive coffee based on their standardized coefficients (Table 28). Furthermore, these three variables refer to the environmental conditions associated with coffee production as defined in Table 6 (Methods and Data Sources Chapter). Therefore, it would appear that at the municipal scale, agroecological variables relate better to intensive agriculture than the socioeconomic, demographic, accessibility, and market and credit characteristics. Although the regression model residuals were random and normally distributed, when they are plotted spatially they clearly indicate a non-random pattern (Figure 28). The Moranâ€™s I value is 0.223, and the Z-score is 8.183 (= -0.00186, =0.0282). This means the regression model, although including the latitude and longitude as dependent variables, is not capturing the spatial characteristics of the processes that lead to a specific area in intensive coffee at any given municipality throughout the country. Furthermore, municipalities with positive residuals (i.e., model underestimating the area planted in intensive coffee) and areas with negative residuals (i.e., model overestimating
Table 28. Multiple linear regression models characteristics No Regional Breakdown Regions by Mountain Range Regions by Farm Size Regions by Quality of Life Index Number ofVariables 10 41 32 21 Dummy variables None D1=1 Western Cordillera D2=1 Central Cordillera D3=1 Eastern Cordillera D1=D2=D3=0 Sierra Nevada D1=1 Farms smaller than 3 ha. D2=1 Farms between 3-10 ha. D1=D2=0 Farms larger than 10 ha D1=1 Quality of life index above average R 2 (Adj. R 2 ) 0.37 (0.36) 0.48 (0.44) 0.44 (0.41) 0.41 (0.38) First 5 variables selected (Stepwise Regression) X, Y, Rainfall in 2 driest months, Average Age, Rural Population Density Average Rural Age_D3, Rainfall in 2 driest months, Y, X_D3, D3 X, Average Rural Age_D1, Y, Rainfall in the 2 driest months, Average Rural Age X, Rainfall in the 2 driest months_D1, Average Rural Age, Rainfall in the 2 driest months, Y Statistically Significant variables (Std. Coeff.) Rainfall in 2 driest months (0.72), Y (0.61), X (-0.34), Average Age (-0.23), Rural Pop. Density (0.14), Distance (0.12) State Institutions_D2 (0.41) Rainfall in the 2 wettest months (-2.10), Rainfall in the 2 driest months_D2 (1.18), Rainfall in 2 driest months (0.53), Y (0.42), Rural Population Density (0.39), Dependents per Household_D1 (0.37), Rural Population Density_D2 (-0.27) Financial Institutions (7.49), Financial Institutions_D1 (-7.69), Rainfall in the 2 driest months_D1 (-0.75), Rainfall in the two driest months (0.40), Y (0.34), X (-0.34), Average Rural Age (-0.21), Distance (0.13), Rainfall in the 2 wettest months (-0.12) 225
226 Figure 28. Multiple linear regression residuals. No regional breakdown
227 the area planted in coffee) are clustered together. The area in intensive coffee is underestimated in the south of the coffee-growing area, excepting the southernmost municipalities where the area in intensive coffee is overestimated. The area in intensive coffee in each municipality is also overestimated along many of the municipalities of the western edge of the Colombian coffee lands. The residuals are positive in the main coffee-producing area of the center of the country. On average, this model underestimates by 343.6 ha the area in coffee in each municipality. The maximum underestimation is 9232.6 ha and the maximum overestimation is 4736.9 ha. The distribution of these residuals also has a standard deviation of 1360.6 hectares. These results suggest that, although the R 2 value is not extremely low, the linear regression model is not doing a good job in predicting the area in intensive coffee, indicating that there are variables and processes that are not included in the regression model that probably can relate better, not only in their correlation, but also spatially, to the area in intensive coffee in each municipality. As with the PC regression models, a multivariate linear regression is not necessarily the best to describe the relationship between the 10 variables selected and the area in intensive coffee. Regression Analysis at the Regional Level I used three regional breakdowns to assess if the effect of the 10 independent variables changes from region to region. The coffee growing municipalities were divided into groups based on their geographical location (4 Mountain Ranges, Figure 29), average farm size (3 regions, Figure 30), and Quality of Life Index (2 regions, Figure 31). Figure 32 shows that the distribution of the area in intensive coffee per municipality is very similar for the Central and Western Cordilleras, and has a wider distribution in the Sierra Nevada and a very narrow distribution of smaller areas in the Eastern Cordillera. The
228 Figure 29. Regional breakdown by mountain range for regression analysis
229 Figure 30. Regional breakdown by farm size for regression analysis
230 Figure 31. Regional breakdown by Quality of Life Index (ICV) for regression analysis
231 0.01000.02000.03000.04000.0 CentralEasternSierra NevadaWesternMountain RangesArea in Intensive Coffee per Municipality (ha) Figure 32. Box-Plot of the area in intensivecoffee per municipality by mountain range. Boxes represent the 25% quantile, median, and 75% quantile (interquantile range). Whiskers represent 2 Interquantile Ranges distribution of area in intensive coffee is clearly different in the Central and Eastern Cordilleras (Mann-Whitney test, p < 0.001) and the Eastern and Western Cordilleras (Mann-Whitney test, p < 0.001). On the other hand, the distributions of the area in intensive coffee per municipality in the Sierra Nevada and the Eastern Cordillera do not show a statistically significant difference (Mann-Whitney test, p= 0.09). When the distributions of the Western and the Central cordilleras are compared, they are not different from each other (Mann-Whitney test, p = 0.14). The same occurs among the distributions of the area in intensive coffee per municipality in the Western Cordillera and the Sierra Nevada (Mann-Whitney test, p= 0.11). Finally, the area in intensive coffee per municipality is not statistically different in the Central cordillera and the Sierra Nevada (Mann-Whitney test, p= 0.28).
232 The distribution of the area in intensive coffee per municipality indicates that municipalities where middle-sized farms (3 ha â€“ 10 ha) predominate tend to have a larger area devoted to intensive coffee (Figure 33). The distribution of the area in intensive coffee in the municipalities with smaller average farms (less than 3 ha) is different from municipalities with intermediate-sized farms (3 ha â€“ 10 ha) (Mann-Whitney test, p<0.001) and the distribution of the area in intensive coffee the regions characterized by smaller farms and large farms (more than 10 ha) (Mann-Whitney test, p<0.001). On the other hand, the distributions of the area in intensive coffee per municipality in the areas with small and large average farms are not different from each other (Mann-Whitney test, p = 0.53). Figure 34 indicates that the municipalities with higher standards of living as measured by the Index of Quality of life (ICV) tend to have larger areas planted in intensive coffee (Mann-Whitney test, p<0.001). These regional breakdowns suggest that coffee production intensification is not a homogeneous process, and the area planted in intensive coffee per municipality differs not only by geographical region, but also by socioeconomic characteristics such as farm size and the standard of living of the population. In the same way that the area in intensive coffee varies regionally, I expected the 10 variables used in the multiple linear regression model to be more or less related to coffee production in different areas of the country. To explore this situation I used a multiple linear regression incorporating dummy variables to include the regional effects of these socioeconomic, demographic, environmental, and accessibility variables. The regional breakdown by mountain range increases the adjusted correlation coefficient when compared to the Multiple Linear Regression without a regional breakdown (Table 28).
233 0.01000.02000.03000.0 3 ha 10 haLess than 3 haMore than 10haAverage Farm SizeArea in Intensive Coffee per Municipality (ha) Figure 33. Box-Plot of the area in intensive coffee per municipality by average farm size. Whiskers and boxes as explained in Figure 32. 0.01000.02000.03000.0 < Mean ICV> Mean ICVAbove/Below Mean Quality of Life Index (ICV)Area in Intensive Coffee per Municipality (ha) Figure 34. Box-Plot of the area intensive coffee per municipality by Quality of Life Index (ICV). Whiskers and boxes as explained in Figure 32.
234 However, only one variable has a statistically significant coefficient, suggesting that, despite a relatively high R 2 value, this model is not capturing the factors that are more closely related to the area in intensive coffee production at the municipal level. The number of state institutions is a closely related variable to the area planted in intensive coffee in the municipalities of the Central Cordillera, but not anywhere else in the country. Although not statistically significant at =5%, the first 5 variables selected in the stepwise regression process suggest that latitude and rainfall in the two driest consecutive months are variables that might be related to the area in intensive coffee, whereas average rural age and longitude are only important in the Eastern Cordillera. The residual map for this regression model has a very clear spatial pattern, with residuals decreasing from west to east and municipalities with similar values of residuals clustered together (Figure 35). The Moranâ€™s I value for this spatial pattern is 0.788 and the Z-score is 28.009 (= -0.00186, =0.0282). This model overestimates on average 822.6 ha the area in intensive coffee (=866.9). The maximum overestimation is 4394.7 ha, while the minimum overestimation is just 11.6 ha. The average residual of this model is larger than the residual from the regression model without regional breakdowns considered (average residual 343.6). Additionally, the clear spatial distribution of the residuals indicates that this model does a poor job in capturing the spatial character of the area in intensive coffee when 4 mountain ranges are considered as separate regions. The regional breakdown by farm size has a higher adjusted R 2 value (0.41) compared to 0.36 for the regression model without a regional breakdown. However, the increase in the number of independent variables from 10 to 32 only increases the adjusted R 2 value by 0.05, or 13.9%.
235 Figure 35. Multiple linear regression residuals. Regions by mountain range
236 Figure 36. Multiple linear regression residuals. Regions by farm size
237 Figure 37. Multiple linear regression residuals. Regions by Quality of Life Index (ICV)
238 This model (i.e., regional breakdown by farm size) also includes longitude, latitude, and the rainfall in the two driest months as some of the first variables entered into the regression model. In terms of demographic characteristics, average rural age is also one of the few variables entered into this regression model, as in the case without a regional breakdown and a breakdown by mountain range. Statistically speaking, this model clearly shows differences in two areas. While Rural population density has a significant coefficient for all the coffee growing municipalities; it also has a different behavior in municipalities characterized by average farm sizes between 3 ha â€“ 10 ha. For all the coffee growing municipalities, the standardized coefficient for rural population density is 0.39, while for the intermediate farm sizes (3 ha -10 ha) the coefficient is smaller (i.e., the sum of the coefficient for the entire country minus the coefficient for this variable). This means that rural population density does not relate as strongly to the area in intensive coffee in municipalities where the average farm size is between 3 and 10 hectares than in the rest of the country. For these municipalities, the rainfall in the two driest months seem to be very important (standardized coefficient 1.18) in its relation to the area in intensive coffee. Another important variable that only has statistically significant relations with the area in intensive coffee at the regional level is the number of dependents per households in municipalities with an average farm size of less than 3 hectares. In these municipalities, the area in intensive coffee increases as the number of dependants increases, mostly because in small farms most of the labor is provided by the family members. By far, the most important variable is the rainfall in the two wettest consecutive months, with the area in intensive coffee decreasing as the rainfall increases. Finally, latitude and longitude are also
239 variables with statistically significant coefficients, but their contribution is not as marked as the rainfall related variables, but more than the rural population density related variables. The regression residuals of this model (Figure 36) do not have such a marked spatial pattern as the residuals of the model with a breakdown by mountain range (Figure 35). However, it is apparent that there are clusters of municipalities with positive residuals (area in intensive coffee is underestimated) and negative residuals (area in intensive coffee is overestimated). On average, the area in intensive coffee per municipality is underestimated by 314.9 ha (=1309.2), with the maximum underestimation being 9287.8 ha, and a maximum overestimation of 3841.3 ha. The Moranâ€™s I value of this spatial pattern is 0.155 and the calculated Z-score is 5.573 (= -0.00186, =0.0282,). As a result, the pattern is not likely caused by a random process in space. Once again, the regression model is not capturing the spatial dimension of the area of intensive coffee production at the municipal level. The last regional breakdown also has a higher adjusted R 2 value than the model without regional breakdowns (Table 28). However, the addition of 11 extra independent variables only increase the predictive capability of the model by 5.5% when compared to the model without regional breakdowns. This regression is one of the two where the institutional variables appear (the other model that includes institutional variables is the regression by mountain range), and the access to credit seems to be an extremely important variable for municipalities below the average ICV. In municipalities where the ICV value is greater than the average, the standardized coefficient is negative, and it is greater in magnitude that the value for financial institutions. This means that while in
240 areas with lower ICV values financial institutions might be extremely important for explaining the area in intensive coffee production, in areas with higher ICV this importance seems to have faded. A possible explanation for this situation is that, as households improve their quality of life, they may have assets and savings that allow them to invest in intensive coffee fields without the need for a loan from a financial institution. It is only in areas where the quality of life is low that farmers do not have assets different from their land, therefore using it as collateral to obtain money to plant intensive coffee. Once again, and with the exception of the financial institutions just mentioned, the variables associated with the agroecological conditions (i.e., rainfall, location) seem to be more important in their relation to the area in intensive coffee than socioeconomic, demographic and accessibility factors. Rainfall in the driest two months is again an important variable. As this variable increases so does the area in intensive coffee. In municipalities above the average ICV, as rainfall in the two consecutive driest months increases, the area in intensive coffee decreases. A possible explanation for this is that farmers in these municipalities are engaged in other productive activities (e.g., wage labor, other agricultural products better suited to the area) different than coffee. For example, vegetable production, a crop that usually requires significant amounts of inputs, is very profitable. Therefore, in areas with more assets, farmers might be looking for better alternatives different from coffee. The residual map (Figure 37) also indicates that the regression model does a poor job in capturing the spatial dimension of the area in intensive coffee. There are clusters of municipalities where the area in intensive coffee is over and underestimated. On average,
241 the area in intensive coffee per municipality is underestimated by 325.2 hectares (=1312.5), with a maximum underestimation of 9088 hectares, and a maximum underestimation of 4147 hectares. As in the other regression models with regional breakdowns, the areas of underestimation correspond to the regions of the country that have been traditionally more oriented towards coffee production. The Moranâ€™s I for this pattern is 0.223 and the associated Z-score is 7.973 (= -0.00186, =0.0282). Therefore, this pattern is very unlikely to be caused by a random process. There are several remarks that can be made with regards to the regression models. In first place, it appears that the variables that define the agroecological potential (e.g., location and rainfall related variables) tend to be more strongly correlated with the area in intensive coffee at the municipal level than the socioeconomic, demographic, and market access variables that are important according to the theoretical models of commercialization and intensification of agriculture at the household level. Therefore, it would seem that these variables are not scaleable, and variables such as population density are not good indicators of agricultural intensification at coarser spatial scales. Another interesting finding is that different variables are important in different regions of the country, and these change with different regional breakdowns. For example, while distance to the closest town has a statistically significant regression coefficient for the models without a regional breakdown, and a breakdown by quality of life index, it does not have significant coefficients in the other regression models. The variables that seem to be more common across models are the location variables and the rainfall in the two driest months, indicating the very strong influence of the agroecological potential in predicting the area in intensive coffee. The area in intensive
242 coffee is growing in areas that have higher rainfall in the two driest months, and higher rainfall in the two wettest months. When considering the total area in intensive coffee per municipality for the 1980 and 1993/97 FNC coffee censuses, the average rainfall in the two driest months is for the municipalities where the area is intensive coffee is increasing (400) is 111.6 mm while it is only 96.7 mm for the municipalities where the area in intensive coffee is decreasing (96). This difference is statistically significant (Mann-Whitney test, p<0.01), therefore, the area in intensive coffee is concentrating in municipalities with higher minimum rainfall. A similar analysis was carried out for the average rainfall in the two wettest consecutive months. The results show that the average rainfall in the two wettest months for the municipalities where the area in intensive coffee is increasing (400) is 671.5 mm, while it is only 626.5 mm for the municipalities where the area in coffee is decreasing. Again, the difference in these distributions is statistically significant (Mann-Whitney test, p<0.01), indicating that municipalities that have an increase in the area planted in coffee with the intensive coffee production system tend to have higher rainfall in the two wettest consecutive months. The comparison of the regression model using the principal components as independent variables and the model using the original 10 variables without a regional breakdown is interesting and puzzling. While using the PC scores, PC3 (institutional forces) has a very significant contribution (measured as the standardized coefficient) to predicting the area in intensive coffee, the institutional variables do not have statistically significant coefficients when the original variables are used. Finally, the residual maps for all regression models show very clear clusters of municipalities where the area in intensive coffee is over or underestimated. The logical
243 conclusion is that these models are not capturing the spatial dimension of the dependent variable. A possible reason for this is that the variation in the area in intensive coffee with space does not follow a linear relationship, as in the multivariate regression models. In order to capture this, other methods that incorporate the location variables more explicitly (e.g., co-kriging, other spatial interpolation methods) or higher order polynomials could be tried in the future. The results from this chapter show that diversification of agricultural production is a process that occurs as agricultural intensification takes place. Furthermore, the intensification of production of certain crops is accompanied by the disintensification of production of others. As a result, rather than looking at coffee as an isolated product, it has to be looked at as part of a production system where certain crops intensify and others disintensify as conditions change. By considering the area in intensive coffee as the response variable of socioeconomic, demographic, environmental and institutional factors without including the effects of other crops on the area planted in coffee does not capture the complex dynamics of coffee production in the Colombian coffee lands. Although the regression analyses just presented highlight some important points regarding the relationship of intensive coffee and other variables at the municipal level, the analysis would greatly benefit from incorporating data on the area planted with other crops as independent variables, so the relationship between coffee and other crops can be better captured in the regression models. Unfortunately, I was not able to obtain information on the area planted in other crops at the municipal level. Furthermore, these regression analyses are only considering one point in time. As a result, there is no temporal dimension in the coffee production dynamics. Another important way to improve the
244 analyses would be by obtaining time-series data on the area planted and the yield of different products at the municipal level and using them with time-series information of the other socioeconomic, demographic, institutional, and environmental independent variables used for this dissertation to run regression models where the area in intensive coffee is the dependent variable. With this information it would be possible to better capture the relationship between coffee and other products, as well as the simultaneous intensification and disintensification of coffee production alongside other crops.
CHAPTER 7 DISCUSSION The previous chapters presented evidence of the landscape transformations accompanying the commercial intensification of coffee production in the Colombian coffee lands. These findings point out to several important issues about the methodological challenges associated with carrying out LUCC research and the understanding of the process of Land-use and Land-cover change. Additionally, the results presented also suggest that models of commercial intensification of agriculture can contribute greatly to analyzing land-use and land-cover change, as different elements of the theoretical models of intensification and commercialization of agriculture are clearly related to landscape changes. This chapter will discuss the implications of these results. The LUCC and Coffee Production By now, the reader should be aware of the importance that history plays in landscape evolution. A careful review of historical events sheds light on the driving forces of landscape change, as well as the major landscape transformations, therefore contributing to the explanation of current land-cover patterns. Although the recognition of this fact is not new in the LUCC research agenda (Turner et al. 1995), its role is becoming more and more important. History can tell how landscapes have changed, and when no other information is available, it can provide qualitative knowledge on landscape evolution. Additionally, current landscape patterns are not the result of the equilibrium of driving forces (Turner et al. 1995). There is path dependency on the initial 245
246 conditions and time lags between the moment when a driving force changes and the time where its effects can be observed on the landscape. For example, in the Colombian case there is a very strong attachment to coffee as a product. Despite the fact that the market conditions for coffee production are not favorable at the moment, many farmers maintain their coffee fields because they cannot perceive themselves (yet) as something other than coffee farmers. Perhaps the best example of this situation is the novel â€œDel Caf a la Cocaâ€ (From Coffee to Coca) (Trujillo Restrepo 1996), where one of the characters is an older small-holding coffee farmer badly hit by the coffee crisis of recent years. Despite the poor economic prospects of coffee as a crop, this farmer refuses to switch to other products, and continues to plant coffee. In the end of the book, this attachment to coffee meant his death, as the coffee crisis deepens and it slowly takes its toll on this old man and his land. This situation, although might have been exaggerated because of the literary nature of this work, summarizes the feeling of most farmers in the main coffee-growing area of the center of the country, where strong cultural attachment to coffee has created a very distinct culture that has left its mark in the Colombian countryside. The effects of this driving force do not coincide with the trends in the international market that would suggest abandoning coffee for other products, clearly demonstrating how landscape evolution is a complex process that depends on the initial conditions and, to some extent on the â€œinertiaâ€ of the landscape, where socioeconomic, demographic, and cultural variables might act as negative feedbacks in the process of landscape change. However, combining a historical perspective with an analysis of landscape dynamics requires an interdisciplinary approach where a common language between historians and spatial scientists needs to be developed (Brgi and Russell 2001). Building this common
247 language between different disciplines is not an easy task. Fortunately, there have been some important advances in linking landscape ecology and history. Case studies from Switzerland (Brgi and Russell 2001) and Massachusets (Foster, Motzkin, and Slater 1998) attempt to bridge these two disciplines, and provide some methodological tools to analyze landscape patterns from an integrated perspective. A major issue regarding LUCC research is that, although ideally spatially-explicit datasets are necessary, there are many sources of information that can provide insights on landscape evolution patterns that, although not spatially explicit, still allow researchers to infer what are the major driving forces of landscape change, and how that region has been transformed. A good example of this situation is the case study presented by Bouchard and Gomon (1997), where notary records on wood sales were used to reconstruct land-cover changes since the early 19 th century in Quebec. Scientists interested in LUCC research should try to capitalize on these sources of non-spatial data that provide information on the land-use and land-cover change process. Even if it is not possible to analyze the spatial patterns of landscape evolution, non-spatial information can be also used to answer the questions of what changed, when it changed, and how and why it changed, which are the 4 of the 5 key questions of LUCC research. In some instances, this kind of information even allows the scientist to infer, in a general way, where landscapes have changed (the fifth key LUCC research question). For example, there is a wealth of information in traditional and indigenous knowledge that can be used to determine land-cover patterns and recent land-cover evolution in areas where there is no possibility to obtain aerial photographs or satellite imagery. Other kinds of records such as traveler diaries and land titles can provide historical information about land-use and
248 land-cover in specific locations. The problem that still exists is how to link this type of information, of a more qualitative nature, to a research agenda that prefers quantitative and spatially-explicit analyses. For the case of the Colombian coffee lands, I used qualitative information (e.g., extension agent surveys, description of establishment of coffee farms from history books) to try to analyze, in a very broad and general sense, patterns of recent land-cover change, and landscape-evolution paths. In the Colombian case, the influence of institutions in shaping the coffee-growing landscape has been very strong. Since its creation in 1927, the FNC has promoted coffee cultivation throughout the country. For the period of this study (1970-2002), the FNC played (and still plays) a very important role in landscape transformations. Since 1970, it has been an official institutional policy to promote the intensive production system all over the country in order to achieve a more efficient coffee production system that enables Colombia to be more competitive in the international market (FNC 1997). As the results of this research show, this process was restricted to the municipalities of western Colombia before 1981, but later on it became a process that spread throughout the country. Similar results for Mexico (Rice 1997) suggest that National Coffee Institute for that country (INMECAFE) also played a significant role in transforming and shaping coffee growing areas in the state of Chiapas. Some other examples show clearly this link between institutions and landscape evolution (Zimmerer 1996a; Southworth and Tucker 2001; Moran and Brondizio 1998). Although LUCC research is starting to concentrate more on the questions of â€œhowâ€ and â€œwhyâ€ landscapes are changing (to complement â€œwhatâ€ and â€œwhereâ€ is changing), there is still a lot of room for improvement in our
249 understanding of the linkage between institutional arrangements and landscape transformations. In theory, LUCC based research should try to address land-use and land-cover changes at different spatial and temporal scales (Turner et al. 1995). The reason for this is that different driving forces lead to landscape changes at different spatial and temporal scales. Furthermore, some of the variables responsible for leading to those transformations change through time. Unfortunately, this is a very strong requirement in terms of information requirements. Additionally, data availability usually determines the spatial and temporal scales, and not the nature of the research question, as in an ideal situation. In practice, most LUCC research projects end up considering one single scale for space and time, and this project is not an exception to this rule. In the present case, the spatial scale and the temporal extent were chosen based on data availability. Although this might seem as a major limitation for a LUCC study because the available data do not necessarily reflect the critical driving forces leading to landscape evolution, in this case it contributed to the body of knowledge about LUCC in a very specific way. Until now, most LUCC research has concentrated at the local or global scales. For this study, the finest possible spatial scale was regional, therefore contributing to the understanding of land-cover dynamics at the meso-scale. For this particular level of spatial detail, the evidence from the scientific literature suggests that the driving forces that are responsible for land-use and land-cover change are climatic and agroecological (Veldkamp and Lambin 2001) which is consistent with the results from the Colombian coffee lands, where rainfall characteristics and location are the variables that are more strongly associated with the area in intensive coffee. If coffee production intensification has
250 resulted in dramatic landscape transformations, then rainfall patterns and location are also closely related to spatial changes in the coffee lands. It was quite unexpected to see how relatively little influence variables like rural population density or accessibility had in estimating the area in intensive coffee per municipality. As a commercial crop, I expected the area planted in intensive coffee to be very sensitive to institutional, accessibility, and market forces. There is evidence of the importance of these driving forces in other areas of the world as markets develop and intensification of agricultural production takes place (Walker 2003; Southworth and Tucker 2001). However, environmental variables were more strongly associated with the area in intensive coffee than socioeconomic, demographic, and institutional variables. Perhaps one of the reasons for this is that presently in the Colombian coffee lands coffee is only one among a wide variety of commercial products. As a result, when the institutional and marketing conditions are adequate for a product or products, farmers tend to emphasize and intensify the production of those crops, while at the same time preserving some of the products that are less important at the moment. Therefore, although institutions and accessibility have a profound effect on the farmerâ€™s land-use decisions, the fact that farmers maintain a diverse array of crops makes the landscape manifestations of those decisions more difficult to identify. Related to the concept of scale in LUCC research, as the spatial resolution decreases, it becomes more difficult to identify the key processes and actors leading to landscape transformation (Veldkamp and Lambin 2001). This is probably caused, at least in part, because as the scale gets coarser it is more difficult to identify clearly the impacts of human activities (Vitousek et al. 1997). Therefore, if it is not possible to pinpoint the
251 impacts of specific human activities at coarser spatial levels, it is also difficult to identify the main driving forces leading to landscape transformations. The results from the research in the Colombian coffee lands support this point. Although intuitively variables such as population pressure (measured as population density and number of dependents per household) and access to markets (measured as the distance to the closest municipality, number financial and state institutions) have a direct link to agricultural intensification, and therefore to landscape transformations, the regression coefficients for most of these variables are not statistically significant, suggesting no link to agricultural intensification and its associated landscape transformations at the meso-scale. There are two different and clear trends regarding landscape evolution and commercial intensification of agriculture. On the one hand, where farmers can afford machinery and land holdings are large enough to benefit from it, the trend is towards landscape homogenization (Burel and Baudry 2003; Di Pietro 2001; Hietala-Koivu 2002; Pauwels and Gulinck 2000; Schuller et al. 2000; Turner, Gardner, and O'Neill 2001). This is usually the case in more developed countries, or in developing countries where plantation schemes are feasible like the Bolivian and Brazilian lowlands, where soybean production has skyrocketed (FAO 2001). In other instances, land holdings are small and it does not make economic sense to use heavy machinery like tractors or combines. Farmers intensify agricultural production mostly by maximizing the efforts of their labor and the efficiency of input use, and this is usually accomplished by diversification. This is the situation in many landscapes where small land holdings predominate like Central Kenya (Conelly and Chaiken 2000) and
252 certain areas of the Andes (Bebbington 1996b; Bebbington 1997; Zimmerer 1996b, 1996a). However, these two trends are not the only possible outcomes of the combination of farm size and access to technology. In many areas of the world, there is machinery adequate for the land-holding size and financial resources of small farmers like in rice producing areas in China and Southeast Asia. I would argue that in these cases, the availability of machinery also leads to landscape homogenization because land that was previously set aside for pasture to feed draft animals can now be converted to rice paddy. However, this land might also be used to plant other crops that might take advantage of the small-sized agricultural machinery. The final possible situation would be farmers in large holdings without access to machinery and industrial inputs. Although this possible combination is not likely to be found in the presently, it characterized the tea and coffee plantations of the Island of Ceylon in the 19 th century, where most of the labor came from agricultural workers (Willson 1999). In these cases the potential to intensify is limited by the availability of human labor (either hired or provided by the family members), animal power, and access to non-industrial inputs. If animal power is used, the agricultural area will probably expand, but at the same time, a large chunk of the farm has to be set aside for pasture to feed the draft animals. Therefore, this leads to a patchwork of intensive (agriculture) and extensive (pasture) land-management practices. Because there is no land scarcity, the crops planted in the intensive area are likely the result of market demands, and to take advantage of animal power, relatively large areas of a single or a few commercial crops are planted. Therefore, the number of crops might increase, but probably less than in the
253 case where smallholder farmers do not have access to machinery and chemical inputs. If human labor is used, the entire farm could be used for agricultural products, as no land is set aside for pasture to feed draft animals. However, it requires a well-developed labor market in rural areas. Thus, a larger area will be planted in one or a few commercial crops, leading to a relatively more diverse farm. These four possible combinations of farm size and access to inputs are summarized in (Table 29). The landscape transformations caused by agricultural intensification for small farms with little or no access to machinery and industrial inputs and large farms with access to machinery and industrial inputs are well documented in the scientific literature. The other two situations have not been addressed with the same depth, and although there are some case studies dealing with these situations, they rarely consider the landscape implications of intensification. These are the reasons of the uncertainty on the possible outcomes under these circumstances. Therefore, the proposed landscape outcomes of commercial intensification of agriculture can be used as hypotheses in future research. Furthermore, for simplicity, industrial inputs and machinery were considered as a single category. There are many instances in which farmers have access to one of these factors of production and not the other. As a first approach, the possible outcomes presented in Table 29 are adequate, but as the literature review of case studies builds up, I would add a third dimension (i.e., land tenure, access to machinery, access to external inputs) to carry out the analysis of landscape transformations. Does Coffee Follow the Models of Commercial Intensification of Agriculture? Although no single conceptual model of intensification and commercialization of agriculture can explain the process of coffee production intensification, most of the
254 Table 29. Possible landscape transformations associated with the commercial intensification of agricultural production. Relation to farm size and access to external inputs. Question marks indicate the likelihood of these outcomes is uncertain Small Farms Large Farms Access to Machinery and industrial inputs Homogenization or Diversification (?) Homogenization No access to machinery and limited industrial inputs Diversification Diversification, but less marked than in small farms (?) models considered (Boserup , Chayanov, Commercialization of Agriculture, Population Cycle, Induced Intensification) contribute to the explanation of this process either by supporting or rejecting some of the basic tenets of the models. According to Boserupâ€™s ideas, more intensive use of agricultural lands should be associated with higher population density (i.e., land scarcity triggers intensification). For the Colombian coffee lands, population density does not seem to be closely related to the area in intensive coffee, as it is the variable with one of the smallest standardized regression coefficients when compared to other variables such as location and rainfall. Additionally, there is evidence of land scarcity as the number of farms has more than doubled since 1970, with the farm size decreasing by a factor of 2. For 1997, more than 70% of coffee farmers had farms smaller than 3 hectares (FNC 1997). In the case of Chayanovâ€™s model of peasant farms, the area in intensive coffee production is not strongly correlated with the number of dependents per household, or the average age of the families. The number of dependents per household is not related to the area in intensive coffee as measured by the multivariate linear regression models, and the average age only shows up making a marginal contribution when compared to agroecological and climatic factors. The reasons for the failure of these models in explaining the area in intensive coffee is probably because they do not consider the market explicitly. However, the fact that population
255 density and average age have weak but statistically significant relationships with the area in intensive coffee suggests they are variables that contribute to agricultural intensification in areas integrated to the market. The conceptual models of agricultural commercialization (Pingali and Rosegrant 1995; von Braun 1995) suggest that commercial intensification of coffee production should take place in areas with better accessibility and access to credit and state institutions. The results from the regression analysis indicate that the distance to the closest municipality (i.e., measure of accessibility) is not related to the area in intensive coffee production. This was a surprising result because accessibility to markets has been documented in the literature as a critical variable determining the commercial intensification potential of a parcel of land (Southworth and Tucker 2001). This situation might be the result of the approximate way in which distance was measured (see Methods and Data sources Chapter) due to the lack of a detailed road map to conduct a spatial analysis to provide an adequate measure or accessibility. It is also be possible that at this particular scale (regional), accessibility does not play such a significant role in explaining the commercial intensification of agriculture, but it is clear from the field visits I performed to several areas of the coffee lands that areas with better road networks are usually associated with more area planted with the intensive coffee production system as well as for other market crops such as vegetables and fruit orchards. One of the aspects clearly associated with intensive coffee production is a higher standard of living of the rural populations (von Braun 1995).The rural inhabitants of the coffee growing municipalities have a higher standard of living than the rest of the Colombian countryside (Mann-Whitney U, p<0.001). However, other expected outcomes of commercialization
256 are not present in the Colombian coffee lands. In first place, the land consolidation postulated by Pingali and Rosegrant (1995) has not occurred, and the number of farms has more than doubled in the last 30 years. The trend towards diversification of production as commercialization of agriculture takes place has been present in the study area. At the municipality level, there is more diversity of agricultural crops than in 1970. Additionally, as postulated by Pingali and Rosegrant (1995), farms have specialized, and today they devote a larger share of the land to coffee production and other market products than 30 years ago (Palacios 1980; FNC 1970, 1997). One of the aspects that differ from this conceptual model of commercialization of agriculture is that the transition towards chemical inputs and engine-powered machinery has only partially taken place. Although the intensive coffee production system requires chemical fertilizers, pesticides and herbicides to maintain its high productivity, most of the tasks associated with the production of this crop are still based on human labor. Of course there has been an increase in the use of powered backpack sprayers, gas-powered trimmers, and other engine-powered tools, but the tasks that require the largest share of labor (e.g., harvesting, fertilizing) are still carried out by hand. Harvesting is done by hand to guarantee that only ripe coffee cherries are collected, and in this way help to maintain the quality that has made Colombian coffee famous all around the world. Fertilization is usually carried out by hand using granulated fertilizers or humus derived from the waste of the coffee-depulping process that are carefully applied to the base of each coffee bush. The historical evolution of coffee production in Colombia also correlates well with the Population cycle model proposed by Bilsborrow based on the ideas of K. Davis (Bilsborrow and Carr 2001; Bilsborrow and Okoth Ogendo 1992). As described in the
257 literature review Chapter, there are 4 stages that link population growth, land tenure, and the intensity of land management. In the last 30 years most of the Colombian coffee lands have been subject to agricultural intensification and outmigration. The results presented in the previous chapters suggest that commercial intensification of agriculture is definitely taking place. This puts the Colombian coffee lands between stages 3 and 4 of the Population Cycle, which basically refer to agricultural intensification (stage 3) and outmigration (stage 4) as responses to population growth. The extension agent surveys clearly indicate that young people are migrating from the coffee-growing areas to the cities and towns of the coffee lands. People with access to better education usually migrate to cities and prefer urban-based employment opportunities than working the land (Duque, Restrepo, and Velsquez 2000). Finally, it appears that the evolution of the coffee-growing landscapes closely resembles the Induced Intensification model (Turner and Shajaat Ali 1996). According to this model, as farmers perceive different market opportunities, they change their land-management strategies accordingly. This process at the farm level translates to the landscape levels as a patchwork of more intensive and less intensive agricultural practices as some farmers change their land use and others do not. In the Colombian case, the coffee crisis has lead to the conversion of coffee plots into pasture, or the abandonment of coffee plots. The farmers that have converted more coffee into pasture are usually large landholders, whereas production in intermediate sized coffee farms tends to maintain the area in coffee. Finally, evidence from field observations, the extension agent surveys, and interviews with some of the researchers at Cenicaf indicate that small-holder farmers are those who usually abandon coffee plots because it has become more attractive to devote
258 themselves full time to wage labor than to work in their own coffee fields. The net effect is that the area in pasture and abandoned coffee plots has increased while the area in coffee has decreased. This clearly responds to how different farmers perceive and react to the coffee crisis. Furthermore, the abandonment and eradication of coffee plots and the expansion of pasture has been a common response of coffee farmers during previous crises in the coffee market (Palacios 1980). Another aspect that supports the induced intensification model is that some land uses intensify while others disintensify. This means that farmers are putting more emphasis on those land-management strategies that are more attractive to them at this particular moment in time. In the Colombian coffee lands this simultaneous intensification/disintensification of agricultural production is taking place in most of the coffee growing departments. A similar situation of intensification of land use practices at the expense of others has been also documented in coffee growing areas of Honduras(Southworth and Tucker 2001), the Upper Canete valley in Peru (Wiegers et al. 1999)and in rural areas of India (Yapa 1977). The Future of Coffee in Colombia: Future Landscapes and Prospects In the same way that coffee production has been intimately linked to landscape evolution since the 19 th century in Colombia, it will continue to exert a significant influence on how coffee-growing landscapes evolve. The current coffee crisis is playing, and will continue to play, a critical role on landscape transformations. Coffee farmers all around the world are starting to change their management practices to take advantage of the growing market for specialty, fair-trade and organic certified coffees in the developed world (Brown 1996a, 1996b; Dicum and Luttinger 1999; OXFAM 2002). It is expected that this market niche will continue to grow faster than the demand for regular coffee because the number of consumers concerned with the environment or the working and
259 living conditions of coffee-producing communities is growing, and people in the developed world are willing to pay a premium price to assure that farmers that produce under certain conditions get higher returns for their efforts (Pulschen and Lutzeyer 1993). Most of the specialty coffees are associated with more traditional coffee production system (i.e., shaded-coffee plantations). This kind of production system shows potential for biodiversity conservation in areas with little native forest left like the coffee growing areas of Latin America (Perfecto et al. 2003; Perfecto et al. 1996) as well as the possibility of higher profits for the farmer per pound of coffee produced (Gobbi 2000). Unfortunately, it is only in the most traditional coffee production systems (i.e., shaded coffee plantations under the canopy of rainforest trees) that have this potential, and they only represent a minimal fraction of the area planted in coffee and even a smaller margin in terms of the amount of coffee produced for the market. For example, in most of Latin America shaded coffee plots are planted under a few selected species, therefore leading to an artificial forest with more structural diversity than an intensive production system coffee plot, but with very little extra biological diversity (i.e. a few more tree species). Although the growth in the specialty coffee market has been extremely fast, the future growth in the different kinds of certified coffee products is somewhat at risk. The reason for this is that consumers do not understand all the differences between the different conservation-oriented, environmentally oriented, or socially oriented certified coffees. Some scientists believe that there needs to be a growing campaign to convince the coffee-drinking public to purchase shade-grown coffee instead of other kinds of coffee as a mechanism to preserve biodiversity (Rappole, King, and Vega-Rivera 2003).
260 Presently, the number of organic or fair-trade coffee producers in Colombia is relatively small because until recently the premium price paid by the consumers for Colombian coffee maintained the price of this commodity higher than that of other types of coffee. However, in the last few years there has been an effort to promote the production of these special coffees in Colombia. For example, Conservation International has successfully established a certified environmentally friendly coffee production program in certain areas of the country (Andrs Dicker personal communication). In this program, coffee consumers pay a premium price that is passed on to farmers as a reward for maintaining their traditional shaded coffee plots. Also, the FNC has started to open Juan Valdez stores in Colombia, several cities in the U.S. and Europe where they sell specialty coffees from different parts of the country (Londoo 2003). At these stores, the FNC sells very high quality coffee at higher prices than the regular blends, and more money trickles down to the farmer. Many of these certified coffees represent win-win situations, where the consumer is willing to pay a higher price for a better product and to improve the life of the farmer, with the added bonus of preserving the shade trees, and their associated potential for biodiversity conservation. However, the certification programs have failed to certain extent because farmers need to pay to be certified, and most small farmers are not being able to pay the fees. Here is where the role of institutions like Conservation International or the FNC has been crucial, providing the know-how and money necessary to organize small farmers and develop the local institutions to maintain the links with the international market. The intervention of these institutions to organize farmers is important because they supply the two most important elements missing for successful commercialization of specialty coffees: 1) the link with
261 the international market, and 2) the management skills needed to maintain a more or less constant supply of coffee. However, switching to specialty-coffee production is not the best solution to face the coffee crisis because it does not address the structural imbalances of the coffee market, where coffee farmers only receive a small portion of the retail price (OXFAM 2002). If the majority of small-holder coffee farmers in the world switched to organic, fair trade or environmentally friendly production, the specialty market would be swamped, and prices would drop as supply would outstrip demand. In order to continue to be successful, the growth in supply should keep pace with the growing demand. As consumer preferences are changing towards specialty coffees, it is also possible that cheaper, generic coffees become the desirable product for most consumers once again. If this occurs, producers of specialty coffee all throughout the world would be in dire straits, because very few consumers will be available to buy their product. Although ideally the coffee market should be reorganized to benefit more the producers, this is not likely going to happen in the near future. Therefore, coffee farmers should try to diversify their farms as a mechanism to minimize the dependence on a single product. Diversification will definitely continue impacting the evolution of coffee-growing landscapes in Colombia. Although this process has been taking place in the study region, it has failed in certain respects. For many farmers, diversification is not an attractive strategy because coffee production offers a secure alternative because it is a crop that has a guaranteed buyer (if it meets the quality standards of the FNC) and a known price, taking care of the marketing of the crop. With other agricultural products, farmers need to find a buyer and rarely know the price before hand, as with coffee. Therefore, many
262 farmers choose to forego a possible higher income and eliminate the economic risks of other products, and keep planting coffee. This is one of the major factors why diversification (including those campaigns promoted by the FNC) has not been more successful, and it is caused in part by the institutional policies of the FNC that indirectly encourage farmers to keep their coffee instead of other products. Another aspect that is critical to consider when thinking about the future of the Colombian coffee lands is land tenure. As stated before, the number of farms has doubled since 1970, and land has become fragmented mostly due to inheritance. If this process continues, farms will become so small that they might not be able to support a family, leading to more intensive use of land, and the disappearance of less intensive land-management strategies like pasture. Dorsey (1999) presents evidence of a similar situation in certain areas of Kenya, where the average farm size and the area in pasture have decreased, while at the same time land is used more intensively to try to support a family. Furthermore, in some areas of the coffee lands farms are getting so small that they are becoming economically unfeasible. Despite the fact that the process of agricultural intensification has led to more diversity of agricultural practices, which in turn would suggest a more sustainable (e.g., diverse) landscape (Di Pietro 2001), it is a landscape that might become characterized by economically unsustainable farms. At the same time that there is a clear trend of land fragmentation, there is a trend towards land consolidation. In recent years many people involved in drug trafficking have started buying farms all over Colombia as a means to launder some of the money coming from this illegal activity. Drug lords buy the land but they not use its full agricultural potential. It is vox populi that the agricultural lands in Colombia are in the hands of the drug
263 traffickers, and the coffee-growing areas are not an exception. A recent study by the Colombian Geographical Institute (IGAC) and the Colombian Agriculture Research institute (Corpoica) show that 0.4% of land owners in the rural areas of the country own more than 47 million hectares, or about 61.2% of the titled land for a total of (Anonymous 2004b). According to the same study, these lands are usually in farms larger than 500 hectares. Furthermore, in all the titled lands of Colombia, only 3.6% of the total area is devoted to agriculture, and nearly 18 million hectares (about 27%) are underutilized (Anonymous 2004b). Although not all of these large land-holders are involved in illegal activities, it clearly demonstrates how land is concentrated in a few hands, and underused lands are usually associated with larger land holdings. As a result, there is one trend where small farms are getting smaller and have a more intensive land management, while the number of large farms is increasing, and because agriculture is not the main goal of these large properties, only extensive land management takes place The answers from the extension agent surveys, the interviews with personnel from Cenicaf, and the literature review indicate that wage labor is becoming a predominant activity for small coffee farmers. This suggests that, from the small-holder farmer perspective, it is more attractive economically to earn a living thorough off-farm employment than working in his/her own coffee fields. Therefore, many farmers would simply abandon agriculture in their farms, leaving the land essentially fallow, or even migrating to towns and cities where opportunities for an unqualified workforce can be found. The latter trend is suggested by the answers to the extension agent surveys. This may result in land abandonment, and a similar situation is presented for rural areas of
264 Japan, where farmers abandon agriculture and work as construction and industrial workers in nearby cities and towns (Hoshino 2001). From a demographic perspective, there is another factor that will definitely contribute to the evolution of the area planted in coffee and coffee production in general. The answers from the extension agent surveys indicate that the younger generation is not interested in working in coffee production. Young people are usually more educated, and tend to prefer jobs in urban areas, where they do not have to work the land. This situation has also been acknowledged by Arango et al. (1998)in Risaralda, and Duque et al. (2000) indicate that in coffee-growing areas it is usual to find lower population growth rates, outmigration of educated people, and a larger proportion of people over 50 years of age. What this shows is that the typical coffee farmer is getting older. As people age, they slowly stop being able to carry out labor-intensive tasks, therefore, making it more difficult for older people to take care of coffee plots as they require significant labor demands. Therefore, coffee plots will eventually be abandoned or replaced by other products that do not require as much labor, and are still suitable for older farmers. Two additional factors that have a very significant effect on the future of the area planted in coffee in Colombia are the relatively recent appearance of illegal crops in coffee-growing areas, and the armed conflict that has characterized many areas of Colombia in the last 40 years (but not in the coffee-growing areas). Illegal crops are a great threat to the ecosystems in the Colombian Andes (Etter and Villa 2000). Most farmers choose to plant these products because there are no other economically feasible opportunities, and the marketing of other crops is extremely difficult due to poor infrastructure or lack of markets. Illegal crops represent an option in which farmers do
265 not get rich, but at least can earn a decent living. In many coffee-growing areas, as the answers to the extension agent survey show, farmers are starting to plant illegal crops (i.e., coca) among coffee bushes as a means to try to earn a living that coffee does not provide anymore (Umaa 2003; Pealoza 2002). Because of this illegal activity, the presence of armed groups (either paramilitary or guerrilla groups) has also become prevalent, as these groups finance a large portion of their activities with money coming from drug trafficking. Thus, coffee is becoming a secondary crop that is used as a disguise for illegal crops, and farmers are not planting new coffee plots. As a result, the intercropping of coca bushes among coffee plants results in a slight decrease in the area planted in coffee, and the aging of coffee plots. Because coffee is no longer the most important crop, there will be significant reduction in coffee production in the areas where both coffee and illegal crops are present, as coffee is only used as a decoy. The presence of illegal groups also leads to some land abandonment. Farmers who prefer not to engage in illegal activities might be forced out of their lands because they represent a security risk to the nearby farmers who are engaged in illegal-crop production. This means that these two elements tend to decrease the area planted in coffee. The impacts of all these factors on the area planted in coffee are summarized in Figure 38. The impact of pests and diseases, particularly the coffee leaf rust and the coffee cherry borer, will continue to influence the productivity of coffee plots, but it is unlikely to affect the area planted in this crop. In 2003, Cenicaf released a new coffee variety resistant to the coffee leaf rust that is well suited for shade-coffee production (Moreno 2002). This provides a solution for areas where the intensive production system was not suitable and that did not have a coffee leaf rust resistant variety. As a result, it is expected
266 that many shaded coffee plots will be replanted with this new variety. In terms of the coffee cherry borer, the low economic returns of coffee mean that farmers cannot buy all the necessary products necessary to maintain the populations of this insect at bay. However, the FNC continues with its aggressive campaign of biological control and integrated pest management, and farmers have to pay a very little amount to buy the tiny wasps that control the coffee cherry borer. Therefore, the levels of infestation of this insect will probably maintain at current levels or even decrease a little bit as coffee prices have had a very slight increase in the last few months. Area Planted in Coffee FNC Market SpecialtyMarket Wage Labor Diversification No GenerationalReplacement Land Fragmentation InsecurityIllegal crops (+)(=)(-)(-)(-)(-)(-)(-) Area Planted in Coffee FNC Market FNC Market SpecialtyMarket Wage Labor SpecialtyMarket Wage Labor Diversification No GenerationalReplacement Diversification No GenerationalReplacement Land Fragmentation InsecurityIllegal crops Land Fragmentation InsecurityIllegal crops (+)(=)(-)(-)(-)(-)(-)(-) Figure 38. Factors affecting the area planted in coffee in Colombia and their effects. (+) Increase in area in coffee; (-) Decrease in area in coffee; (=) No change. Based on the previous discussion, I would expect the area in coffee to keep decreasing, while at the same time there is a slowdown in the conversion from the traditional coffee production system to the intensive production system, or even an increase of the area in traditional coffee, as farmers try to take advantage of the opportunities of the specialty coffee market. At the landscape level, I expect the trends
267 that were described in the chapter about coffee production in Colombia and the results chapters to be maintained. The landscape has evolved towards a more agriculturally diverse countryside. As the area in coffee has decreased, the area in other crops has increased, and the area in pasture has decreased (but in recent year it has started to increase again). I would argue that agricultural diversification will be the factor leading landscape transformations, as the coffee farmers that were waiting for the coffee crisis to end realize that they need to plant different products to be able to earn a decent living once again. In the early stages of this agricultural change, the area in pasture will expand, and cattle raising will grow in the coffee-growing areas. At the same time farmers will start to plant other products that provide good market opportunities and area compatible with the assets and preferences of each farmer. Due to the increased urbanization, the production of vegetables, fruits, and other products with high demand in urban areas will increase. As population continues to grow in the coffee growing areas, both land fragmentation and urban migration will continue, therefore the population still engaged in farming will keep intensifying agricultural production. Additionally, areas in natural or semi-natural state (secondary forests, secondary succession plots) will probably increase as many small-holder farmers, which represent the majority of coffee farmers, will devote more time to wage-labor, abandoning to a large extent agricultural production on their own farms. According to Lee et al. (2001), the role of commercial intensification of agricultural production in the future should increase food security, provide a higher income, and at the same time help in environmental conservation. In the Colombian coffee lands only one of these goals (i.e., provide a higher income) was accomplished during the process of
268 commercial intensification of coffee production. It is clear that coffee farmers have achieved a higher income and standard of living when compared to other rural areas of the country. As it was already mentioned, the quality of life index (ICV) for the coffee-growing municipalities is significantly higher than in the rest of the country. However, this characteristic of the coffee growing areas having a higher standard of living is disappearing as the coffee crisis persists, and widespread unemployment rates are common, leading to poor living conditions (Anonymous 2004a). The disappearance of home gardens indicated by the extension agents and Escobar and Ferro (1991) suggests that food security has decreased because farmers are depending more on food that is bought rather than produced on the farm. While coffee allowed farmers to earn a sufficiently high standard of living, they were able to buy more than enough food. However, as the coffee crisis hit them, their income decreased drastically, and many families were not able to feed themselves adequately. In terms of economic security at the farm level, although at the landscape level (i.e., municipality) there are more different crops planted now than in 1970, the proportion of the average coffee farm planted in coffee and other few commercial products has increased, suggesting a higher reliance on a few commercial products, making farmers more susceptible to market swings (and the present coffee crisis is a perfect example). Additionally, there have been some campaigns sponsored by the FNC (Jos Jaramillo personal communication) to try to encourage small-holder farmers to re-establish their home gardens and improve their food security. If these campaigns are successful, I would expect the home gardens to reappear in the coffee growing areas, mostly among the small-holder coffee producers.
269 In terms of environmental conservation, the prospects are also rather grim. The adoption of only certain aspects of the intensive production system (e.g., not adopting the soil conservation practices) has lead to widespread soil erosion, and the increased coffee production has meant the dumping of large amounts of coffee pulp to watercourses, higher use of pesticides, fertilizers and other chemical inputs. The increase in the area planted in other crops has also contributed to increased pollution. The extension agent survey results indicate that most of the other products require more chemical inputs than coffee (even intensive coffee). Therefore, as the diversification trend continues, the use of chemical inputs will increase, possibly leading to higher pollution levels. Additionally, the paradigm of the FNC until recent years had very little emphasis on environmental conservation, focusing mostly in increasing productivity, leading to very little institutional support for more sustainable production systems (Arango et al. 1998). Many of the new crops are not suitable for the sloping terrain, and will lead to widespread environmental degradation in a few years. For example, the production of cassava is presently a very attractive economic alternative for farmers, but is a disaster in terms of soil conservation, as the harvesting of this crop requires digging out tubers. This loosens and leaves exposed the soil, and being in an sloping terrain is easily washed away (Siavosh Sadeghian personal communication). Additionally, there are other clear signs of environmental degradation that are not being paid attention to like water pollution and biodiversity loss (Jorge Botero personal communication). It remains to be seen how long can the coffee growing landscape take this abuse, and hopefully this dissertation can contribute to show land managers and decision makers the need for a more environmentally friendly agriculture in the Colombian coffee lands.
CHAPTER 8 CONCLUSIONS, LIMITATIONS, AND FUTURE RESEARCH In the introductory chapter of this dissertation I clearly stated that land-cover modification was an important, often neglected component of land-use and land-cover change (Lambin and Geist 2001; Mannion 2002; Turner et al. 1995). Unlike land-cover conversion, that has direct effects on landscapes, land-cover modification is more subtle and indirect, and the landscape transformations it produces are not immediate as in land-cover conversion. Despite this subtlety, land-cover modification is a very powerful agent in landscape transformation, as the results from this study show. In the Colombian coffee lands, land-cover modification in the form of commercial intensification of agriculture had profound impacts on the landscape evolution of the coffee growing area. What is very interesting to note is that the area planted in coffee is not the only land-management strategy that has been affected by this process of landscape change. In the Colombian case, the commercial intensification of coffee production had the effect of changing land-covers in the non-coffee areas of coffee farms. Therefore, the process of commercial intensification of one product had the effect of not only changing the area in that specific product, but also in other land-management strategies associated with a coffee farm. In this sense, commercial intensification of agriculture for one product can be regarded as having â€œmultiplierâ€ effects that cascade to other land uses in the landscape. The intensification of coffee production in Colombia not only meant the decrease in the area planted in coffee, but also a decrease in the area in pasture, an increase of the area in other crops, and the disappearance of the shade trees that were previously associated with 270
271 the traditional coffee production systems. The intensification of coffee production has concentrated in regions with higher average rainfall in the two driest consecutive months and in the two wettest consecutive months. Although the growth in the area in intensive coffee in regions with a less marked dry season is expected, as coffee is very sensitive to a very prolonged dry season, it was somewhat unexpected that the area in intensive coffee increased in areas where the rainfall in the two wettest months is higher because conditions that are too wet are not ideal for intensive coffee production. It is likely that in regions where the area of coffee planted with the intensive production system is increasing these two wettest consecutive months correspond to the early development stages of the coffee cherries, where the coffee bush needs all the resources available to produce seed, therefore, abundant rainfall is important. What these two facts point out is the critical importance of agroecological potential in the process of agricultural intensification. This fact, although important at a local level, becomes more important at a regional level because it determines to certain extent the areas of the landscape where intensification of agricultural production is more likely to take place, creating a patchy landscape of intensive and less intensive agricultural fields. This dissertation started by identifying three basic hypotheses: 1) Landscapes undergoing commercial intensification of agriculture diversify, 2) As this process takes place, some crops will intensify and others will disintensify, and 3) In different areas, there are different factors associated with the intensification process. The results suggest that the first two hypotheses are clearly supported, and in this case the Colombian coffee lands show an increase in the number of different crops planted in the coffee-growing municipalities since 1970. This is probably caused by the combination of sloping terrain
272 and small land holdings, and the difficulties associated with the use heavy agricultural machinery to replace human labor in coffee production, leading to intensification of agricultural production through diversification. The analysis of yield time-series data indicates that as commercial intensification of coffee production takes place, there are crops that intensify along with coffee, while others disintensify as farmers react to different market opportunities. The last hypothesis is more difficult to accept because the results of the regression analysis in different regions of the country suggest that climatic and agroecological variables are almost always important regardless of the regional breakdown used. Additionally, the fact that different factors have different impacts on the area of intensive coffee per municipality when different regional breakdowns are used also highlights a problem that has been long recognized by geographers but that has rarely being addressed in LUCC research: The Modifiable Areal Unit Problem (MAUP) (Barber 1988). The basic issue regarding the MAUP is that when data are aggregated or analyzed with different spatial extents containing the same spatial units of analysis (i.e., a circle containing all the coffee-growing municipalities versus a square containing all the coffee growing municipalities), the variability is lost (when data is aggregated), and the results from different spatial extents at the same spatial scale can be different. This makes it difficult to identify the key driving forces leading to a specific landscape pattern, because the significant variables may change as the spatial configuration of the units of analysis varies within the spatial extent. In land-use and land-cover change research, the fact that the results from different spatial extents may change is rarely acknowledged, and the spatial extent at a given scale is considered as a constant and the driving forces identified
273 using that spatial extent through spatial analysis the actual variables leading to landscape evolution. In this dissertation I show how the analysis of the same data in different regional breakdowns leads to different driving forces within regions within the same regional breakdown, but also for the different regional breakdowns as a whole. It remains to be seen how changing the spatial boundary or the level of aggregation of the data changes the results about the LUCC process. However, I would expect that, in the same way that at the local level households with different assets have different land management practices (Murphy, Bilsborrow, and Pichn 1997; Coomes and Burt 1997; Zimmerer 1996a), driving forces of landscape change also change from region to region, as each region also has a different set of assets determined by its socioeconomic, demographic, institutional, and environmental conditions. The results from the regressions also show that some of the variables hypothesized as critical for agricultural intensification at the household level such as population density, the number of dependents per household, or the average age are not necessarily significant for the municipalities as units of analysis. A similar situation occurs with variables theoretically associated with commercialization of agriculture such as access to markets and credit. On the other hand, the results also suggest that other models of intensification of agricultural land use might be better suited for an analysis at a coarser scale. The fact that there are crops that are intensifying while others are disintensifying is consistent with the Induced Intensification model, and the outmigration to urban areas from younger and more educated people is consistent with the ideas of Bilsborrowâ€™s Population Cycle. In the case of the Induced Intensification model, it appears that is a conceptual model of commercial intensification of agriculture that allows taking into
274 account household differences, therefore translating in a landscape characterized by intensive patches surrounded by less intensive parcels that change through time as farmers react to different market opportunities. This pattern can be inferred from the simultaneous intensification and disintensification of crops through time. The time series analysis shows how there are certain crops that intensify for a few years, and later on other crops may become important. Therefore, the Induced Intensification framework appears to be ideally suited to analyze landscape transformations associated with the commercial intensification of agriculture at this particular scale. When the results from the Colombian coffee lands are compared to the conditions associated with the different models of intensification and commercialization, several similarities and differences emerge. In the Boserupian model, land scarcity (i.e., high population density and no land available for planting new fields) triggers agricultural intensification to feed the increasing population. In this model, markets are not considered, and agricultural production is at the subsistence level. In the Colombian coffee lands there is evidence of land scarcity, but the fact that coffee is a cash crop makes conditions for intensification very different. Rather than producing for subsistence only, Colombian coffee producers are mostly commercial farmers. Therefore, the intensification of production of coffee and its associated products is heavily dependent on the conditions of the market. When prices are favorable and the farmer resources sufficient, it is likely that certain products will intensify. This does mean that population density is not important, but in areas fully engaged in market production it takes a secondary role in explaining intensification of agricultural production. Despite the fact that Boserupâ€™s model does not consider the influence of markets, it postulates a patchy
275 landscape with more intensive and less intensive agriculture plots interspersed with each other. Interestingly enough, this is one of the similarities of this model with the actual pattern of landscape change in the Colombian coffee lands, where patches of different crops planted with different levels of intensity coexist in the same area, and the intensity of management of those plots changes as conditions for certain products improve or deteriorate. A clear difference with the Boserupian model is the fact that although the number of farms has more than doubled in the Colombian coffee lands, there has been a simultaneous trend of land consolidation. Therefore, the landscape is characterized by a large number of small farms that become smaller, and a few large farms that become larger. In Chayanovâ€™s model of a peasant farm, the intensification and expansion of production depends on the number able-bodied family members, and the ratio of household dependents to the number of people who can work in the field. This model assumes a different kind of economic rationality (i.e., satisfiers instead of optimizers), abundance of land to plant new fields, and no production to the market. As the number of able-bodied family members increases, the ratio of dependents to able-bodied family members decreases, the amount of agricultural labor per person diminishes. However, if the number of dependents increases faster than that of able-bodied family members, the latter have to work harder and intensify production or open new land for cultivation in order to care for the dependents. The intensification of coffee production has been the result of market forces and the diffusion of certain production technologies among coffee farmers. The introduction of this new system has changed dramatically labor demands, and the family is no longer able to provide enough labor to care for the crop, as it would
276 be expected in Chayanovâ€™s peasant farms. In the coffee-growing area of Colombia some farmers have abandoned land because it is either not suitable for the intensive production system, but also because it becomes easier and economically more attractive for small farmers to engage in wage labor in the farms of the intermediate and large land holders than to work in their own farm. Chayanovâ€™s model also assumes that each community has a standard of what constitutes decent living conditions, and this defines the amount of agricultural work for each family. In the coffee-growing regions of Colombia influences beyond the community determine how much people are willing to work in the land. Young people tend to prefer urban jobs, and there is a continuous outmigration from the countryside to the towns and cities, usually people looking for better opportunities than in the rural areas or escaping from the violence that plagues many areas of Colombia. In summary, the Chayanov model differs from the situation of the Colombian coffee lands in the fact that in the latter land is limited, agricultural production is mostly commercial, and there is a desire of the younger generations to move out of the country side. In Bilsborrowâ€™s population cycle, there are 4 different possible responses in the relationship between population pressure and natural resource use in any given region increasing population (Tenurial, Extensification, Technological, Demographic). The coffee growing landscapes of Colombia are experiencing technological (i.e., intensification of production) and demographic (i.e., outmigration) responses to population pressure. The development of these model comes from a wide variety of case studies from Asia and Latin America (Bilsborrow and Carr 2001; Bilsborrow and Okoth Ogendo 1992), and they present similar responses to those observed in the Colombian coffee lands. Once again, the impacts of the market are not explicitly acknowledged,
277 and in the Colombian case, intensification has been a response to market conditions rather than population pressure. The results of this study share many elements with the ideas about commercialization of agricultural production of von Braun (1995) and Pingali and Rosegrant (1995). According to these authors, commercialization takes place as governments develop markets and adequate macroeconomic policies for certain products. Additionally, there is a shift from subsistence to commercial production and human or animal labor and organic inputs are slowly replaced by machinery and industrial inputs. Furthermore, regions tend to specialize in certain products, leading to lower agricultural diversity at the regional level, but a higher diversity at the national level. The results from this study show that farmers are more commercially oriented now than 30 years ago, as a higher proportion of the farm is planted with coffee and other cash crops, but there has not been a strong transition to agricultural machinery or chemical inputs. The reasons for this were explained before, and have to do with the difficulty of using machinery in steep slopes and the expenses associated with chemical fertilizer. As a result, farmers intensify by diversifying production. The net effect is that the coffee-growing areas have more agricultural diversity than 30 years ago. One final similarity with the ideas about commercialization is that the increasing commercial orientation of production leads to a higher standard of living. The coffee-growing areas of Colombia definitely have a higher standard of living as a result of investment of coffee profits in rural infrastructure and health and education services. The induced intensification model states that households with different assets and access to markets will intensify production only when it is strictly necessary. As a result,
278 the evolution of land management resembles a stair-stepped line. Farmers maintain a particular production system for a specific product until it is not economically feasible to continue producing in the same way, and switch to other management practices and or products. Each household has its own intensification path dependent on family assets. Therefore, the landscape is composed of farms with more intensive management interspersed with farms with less intensive management, and there may be many levels of intensive management present in the same landscape. Turner and Ali (1996) developed this model with empirical information about farms engaged in commercial rice production in villages with different access to markets in Bangladesh. The Colombian coffee-growing region shares many elements of the Induced Intensification model, but in a totally different context. Although both in Bangladesh and Colombia farmers are engaged in commercial production, in Bangladesh it is a food crop that is consumed locally and in Colombia is a sumptuary crop that is consumed abroad. Thus, the market linkages in Colombia act at much longer distances than in Bangladesh, and the nature of the crop means that if Bangladeshi farmers are not able to sell their crop, they do not starve, while Colombian farmers rely heavily on the earnings from coffee to buy their own food. Therefore, I would argue that Colombian coffee farmers are more dependent on the market than Bangladeshi rice farmers. Another of the contributions of this dissertation is answering the questions of what land covers have changed, and when and where those changes have occurred. This research addressed more these three questions than the â€œhowâ€ and â€œwhyâ€ land-covers change, but it provides the starting point to improve the scientific understanding of land-use change in coffee growing-areas, now that the patterns of change have been
279 documented. Additionally, this dissertation tries to link the spatial and temporal characteristics of landscape change with the evolution of the societies that induced those changes. Therefore, understanding the history of coffee in Colombia is as important as the recent patterns of landscape change in explaining land-use and land-cover change. Focusing at the regional scale has both advantages and disadvantages. Perhaps the most obvious disadvantage is that for this particular case study, the spatial grain is not directly linked with the agent of land-use change (i.e., coffee farmers). Therefore, it is difficult to link patterns of landscape evolution with the processes leading to those changes. However, this can be considered as an advantage as well if the analysis at the regional scale is accompanied by analyses at finer and coarser scales. By concentrating on coarser scales, it is possible to identify other actors in the landscape change process that operate at coarser scales, and have more indirect impacts on landscape evolution. Furthermore, an analysis of the landscape changes associated with agricultural intensification also allows the linkage between agricultural change and socioeconomic processes at regional and national scales that are not necessarily evident when processes at the local scale are considered, and certain processes can be conceptualized as constant for a group of households in a specific region (i.e., access to markets). Therefore, by combining an analysis at different scales, it is potentially possible to address the emerging properties of the land-use and land-cover change system. In other words, the change in a landscape is not only the sum of the changes in individual farms, but also of other forces that act at the landscape level. This research also contributes significantly to the understanding of coffee production in Colombia, as it is one of the handful of studies in the country that
280 incorporates the spatial dimension of production, and it is definitely the first one that does it for the entire coffee growing area of the country. Coffee (both traditional and intensive production systems) has a great potential for coffee to transform landscapes. For example, the transforming power of coffee has been acknowledged in the scientific literature for Latin America (Rice 1996; Perfecto et al. 1996). However, only a few studies documenting the landscape evolution of coffee-growing landscapes in Chiapas, Mexico (Rice 1997), Vietnam (Caspersen 1999) and Puerto Rico (Rudel, Perez-Lugo, and Zichal 2000) have been published in the scientific literature, mostly due to lack of spatial and historical land-use information. Although the quantity, resolution and scale of spatial information related to coffee production in Colombia is not ideal, it certainly provides enough detail to analyze landscape evolution at a relatively coarse scale. In this regard, this dissertation contributes not only to the understanding of the landscape evolution of coffee growing landscapes in Colombia, but also as another case study documenting this process at the global level. In terms of methodological contributions, there are several important contributions from this research. In many instances there is relatively little information on landscape evolution both spatially and historically. This research shows how it is possible to combine these spatial sources with other non-spatial sources such as descriptions in history books, and government statistics among others to try to reconstruct landscape change. Although this combination of spatial and non-spatial sources usually only allows recreating landscape patterns at coarse scales, this is a great improvement for many regions where little spatially-explicit information is available. Furthermore, this combination of non-spatial information with a series of snapshots at different moments in
281 time (e.g., land-cover maps from different years) allows the scientist to fill the gaps of these unconnected land-cover patterns, and infer landscape change processes. The use of correlation coefficients of different crop yield time-series data to look at agricultural change is another contribution of this research. For many developing countries, annual yield information is one of the few variables that is readily available (e.g., FAOSTAT database) and can be used to analyze the agricultural sector. The use of correlation coefficients provides an easy, yet powerful tool to look at year-to-year changes, but also to detect transformations at coarser time scales. Concentrating at the intermediate scale and using the methods outlined in this dissertation has some disadvantages and disadvantages. Perhaps the most important advantage is that the methods used in this research rely on information that can be obtained relatively easy, and show how landscape evolution can be analyzed with relatively sparse datasets. By combining quantitative and qualitative information in the analysis of landscape change I was able to develop a story line of how landscapes evolve that could not be developed by looking at quantitative data alone. Another important advantage of the methods used in this dissertation is the pivotal role played by history. Without understanding how coffee became an important crop in Colombia or in the world, it is not possible to understand how the current landscape patterns evolved. Many historical processes shape how the landscape is or were used, and determine how it will be used in the future. Without a careful consideration of these processes it is hard to understand the dynamics of the landscape and move beyond the description of static snapshots. The choice of methods also contributes to LUCC research by helping to fill the gap between global and local studies of land-use and land-cover evolution.
282 The methods point out several disadvantages of working at the intermediate scale. Perhaps the most important is that the models of agricultural intensification developed at the household level do not work well at the regional level. At this scale, other properties and variables of the land-use and land-cover system become more important, and the variables hypothesized by the household-level agricultural intensification models can not be scaled up. Additionally, by not considering other scales it is impossible to analyze the dynamics of cross-scale interactions, and how certain variables at one level influence other variables at other levels. Another issue related to the lack of analysis at multiple scales is that focusing at the regional level masks lower scale spatial variability. Despite these methodological disadvantages, the results and the analysis presented show landscape evolution paths that are adequate to explain current landscape patterns at the regional scale. Thus, it would seem that in this case, the pros are more than the cons in using these methods. Limitations of this Research As in any scientific study, this dissertation has several limitations. Perhaps the most important is that it is relaying heavily on secondary data. There are two major problems when dealing with secondary data: 1) quality, and 2) it is not necessarily compatible with the objectives of other projects. The quality of the information is of critical importance, and in some instances the secondary data sources do not include an explicit statement of their quality, or how certain variables were measured or calculated. Therefore, this reduces the usefulness of the data because it is impossible to know how accurate it is. The researcher faces the dilemma of not using the sources for which there is no clear statement of the quality of the data, or to incorporate them and assume they are adequate. It is probably better to avoid using these sources, but in many instances they are the only
283 information available related to a particular aspect of the research project. In this latter case, the dilemma for the researcher becomes even greater. For this research, some of the sources that captured critical variables (e.g., environmental degradation, water pollution) ended up not having any statement of their accuracy, and were not used in the final analyses. In some instances, it is possible to use other sources of information to cross-reference and validate the source without a statement of data quality, or assess the internal consistency of the data source (i.e., the information makes sense and does not provide contradictory data). The second issue related to secondary data is that the parameters recorded do not necessarily fit the research questions. All data are generated to answer specific questions, and this information may not be suited to answer other questions. In this case, I tried to use information about environmental degradation at the municipal level (Sarmiento et al. N.D.), but I was not able to use it because it was not clear how the indices of environmental degradation were calculated. However, the information was adequate to answer the research questions of the original project. However, it also has to be stated that using secondary information is also extremely important for LUCC research because it is necessary to fill the gaps of knowledge about the spatial and temporal evolution of landscapes. However, because of the reasons described, researchers have to be careful when incorporating secondary data. Another weakness of this study is that some of the methods were not the best suited to collect the information. Specifically, the extension agent surveys had open-ended questions and that required the person answering it to recall information. As stated in the Methods and Data Sources Chapter, open-ended questions are not ideally suited for
284 mailed questionnaires, and in questions requiring recalling past information, what is remembered is not necessarily fact. However, in this case the choice of open-ended questions was based in the fact that there was no prior information to provide a limited but comprehensive number of choices to the extension agents answering the questions. Furthermore, the answers were consistent with each other in terms of level of detail and information content. Regarding recalling information, I was only able to assume that extension agents would be able to recall information regarding agricultural activities, given the fact that working with farmers and traveling through the countryside is an integral part of their job. Although these results might be treated with some skepticism, they provide a suitable baseline and suggest the general trends of landscape evolution to carry out more detailed studies that use more adequate research protocols. From the point of view of the LUCC Science/Research plan, this study has two limitations: 1) little spatial detail, and 2) little quantitative information. These limitations are significant, but the fact that the temporal and spatial scales of this research were clearly defined from the beginning makes them less important because the spatial detail was adequate for the spatial scale (i.e., municipalities as units of analysis and a regional scale), and the temporal information was adequate for the temporal scale (i.e., statistics, coffee censuses, and a 32-year time period). The lack of spatial data is something that cannot be solved in many regions of the world, and instead of choosing not to conduct a research study of a particular area because there is not enough information with adequate spatial detail, I believe that is better to work with the available information to, at least, get a general, qualitative understanding of landscape evolution. The lack of quantitative information is also related to the lack of data. This kind of information is critical to try to
285 confirm the trends that are suggested by the qualitative data. However, if records of the variables of interest do not exist, it is difficult to solve this deficiency. One of the possibilities to address this limitation is to use other variables with quantitative records that are not directly related to the landscape evolution process, and develop relationships with landscape-evolution related variables for which there are no records available. Additionally, it would be possible to collect data to answer specific questions, but this is dependent on the availability of time and resources to do so. Further Questions and Future Research Most research projects lead to more questions than the ones that are actually answered. This study on the Colombian coffee lands is not an exception. The results from this research suggest that it is very important to explore further the linkages between commercial intensification of agriculture, mechanization, land tenure, and landscape evolution. Perhaps a good approach to better understand the relationships between these variables would be to carry out a meta-analysis of case studies similar to that carried out by Geist and Lambin (2002) for the causes of tropical deforestation. If a typology of landscape evolution patterns could be developed based on different conditions of agricultural intensification, access to industrial inputs, mechanization, and land tenure, and how they interact with each other, much could be gained in understanding the process of landscape transformations associated with agricultural change. Such a study would contribute by creating a typology of landscape changes in coffee growing areas. As in the case of tropical deforestation, I would suspect that the interactions among variables are more important than the individual variables in explaining land cover transformations. Although the results of this research indicate that the effects of the interactions of driving forces were apparently less important than the impacts of the
286 individual forces, this topic requires further research to try to elucidate how variables are interacting with each other. The most logical research question coming from the results of this dissertation is to try to develop a detailed LUCC model for the Colombian coffee lands linking biophysical, demographic, and socioeconomic variables, as well as the different actors involved in landscape change. Completing this model would complement the what, when, and where questions on LUCC answered in this dissertation with the how and why landscapes are changing. Based on the success of the Induced Intensification model in conceptually explaining patterns of intensive and less intensive agriculture in any given landscape, I would like to explore the possibility of implementing this model in a GIS environment in order to simulate landscape patterns in the future as conditions change. This tool then could be used by the FNC and Government planning agencies in order to explore future landscapes, and perform landscape planning to try to manage the slopes of the Colombian Andes in a more sustainable way, and reduce the impacts to the remaining patches of forest. However, in order to create this model, I would use the information gathered during this research project as a baseline, and I would need to obtain more information on landscape change from land users and land managers, and explore the possibility of successfully incorporating remotely sensed data, something that was not possible to accomplish during this study. The results suggest that the variables associated with intensification and commercialization of agriculture at the household level do not scale up to the municipal level. This does not mean that there are not variables that provide information at the local level that can be scaled up to the landscape, municipal, regional, or national level and still
287 maintain significant relationships with the process of agricultural intensification. Finding these factors would be a great contribution to LUCC and agricultural change research, because they would provide variables that could be used in multi-scale models of land-use and land cover-change, and in this way , fulfill one of the major goals of the LUCC Science/Research Plan (Turner et al. 1995). Additionally, much can be learned from management strategies and intensification of agricultural production in other areas of the world where the products being intensified is subject to phytosanitary constraints. For example, coffee production in Colombia might benefit from the lessons learned by the cacao producers in the Brazilian state of Bahia, and vice-versa (Nigel J. H. Smith personal communication). These case study comparisons will contribute to the development of the agricultural change and landscape evolution typology mentioned above. Therefore, another future direction for research is to look for scientific research on the intensification of other agroforestry systems to try to learn effective management strategies to deal with pests and diseases. Another topic that can be tied with the results of this dissertation is how the demand in urban areas and foreign countries triggers changes in rural locations. Understanding these urban-rural and transnational linkages is critical for understanding commercial intensification of agricultural production and its associated landscape transformations. It would be very interesting to establish a typology of interactions, looking at the size and distance to the urban agglomerations and or countries that are more related to certain areas of the landscape, how these cities influence the patterns of agricultural intensification or disintensification for different products, and how the relationships change as time goes by. With this information, it would be possible to
288 develop maps of areas of the landscape influenced by certain cities, regions, or countries. This kind of map would help in determining how changes in one location may trigger landscape transformations in other regions of the world. Humanityâ€™s ecological footprint can be defined as the resources needed to support the demands of people given the present levels of consumption (Wackernagel and Rees 1996). Our footprint has exceeded the capacity of the planet to provide resources at a sustainable level, and this means that humankind is slowly depleting the Earthâ€™s resources, jeopardizing our future and that of our planet. Here is where the role of LUCC and other Global Environmental change research are critical. Understanding the patterns and processes leading to landscape evolution are essential for determining sustainable land-management practices. Although changing current patterns of consumption will not happen overnight, identifying and developing land-management strategies that are more sustainable and have the potential to adapt to changing conditions is a necessary condition for reducing our footprint. Understanding the patterns of landscape change is a first step towards a more sustainable land-management strategy, and hopefully the results of this dissertation will lead to more steps in the direction of a sustainable landscape in the coffee growing areas of Colombia and other parts of the world.
APPENDIX A AGRICULTURAL EXTENSION AGENT QUESTIONNAIRE Comit de: __________________________Fecha: ________________ Nombre del entrevistado: _____________________________________ Telfono:__________________________________________________ CAF 1. Hace cunto trabaja en el programa de extensin para esta rea?________________ 2. Desde que usted empez a trabajar qu ha pasado con el rea sembrada en caf (aumenta, disminuye)? 3. Y qu ha pasado con la productividad del caf (aumentado, disminuido, igual.)? Cul cree que sea la razn principal de estos cambios? 4. Se observan los mismos cambios en la productividad y rea para pequeos, medianos, y grandes caficultores? Qu diferencias hay? 289
290 OTROS CULTIVOS 5. Qu cultivos u otros usos del suelo han disminuido? Cules han desaparecido? (por favor sea especfico en cules han desaparecido y cules han perdido importancia) dnde? (de una idea aproximada de dnde han ocurrido los cambios, en el sur del departamento/municipio, o en las veredas/municipios XXX) 6. Qu cultivos u otros usos del suelo han aumentado? Cules han aparecido? (por favor sea especfico en cules han aparecido y cules han aumentado su importancia) dnde? (de una idea aproximada de dnde han ocurrido los cambios, en el sur del departamento/municipio, o en las veredas/municipios XXX) 7. Qu ha pasado con las huertas caseras? Siguen existiendo? Han desaparecido? Qu caficultores (pequeos, medianos, grandes) las mantienen y cules no?
291 8. Qu ha pasado con el consumo (aumentado, disminuido, sin cambio) de insumos qumicos (abonos, pesticidas, herbicidas) desde que usted trabaja en el rea? En su opinin, se usan ms en una parcela de caf o en una parcela del mismo tamao de otros cultivos? Hoy en da quines usan ms insumos qumicos por hectrea (pequeos, medianos, o grandes productores)? y antes quines usaban ms insumos qumicos? RAZONES PARA LOS CAMBIOS 9. Fuera de la agricultura, hay otras actividades econmicas (jornaleo, trabajo en zonas urbanas, etc) a las que se estn dedicando los caficultores? Comparando con el momento en que empez a trabajar en la zona, hoy en da se dedican ms o menos que antes a estas actividades? 10. Qu factores han influenciado estos cambios (por favor trate de dar por lo menos dos o tres factores)? Cul es el factor ms importante? Qu otros factores tenan importancia antes y ya no son importantes?
APPENDIX B SPATIAL DISTRIBUTION OF THE VARIABLES USED IN THE REGRESSION ANALYSIS There were ten variables selected that according to different theoretical models are responsible for the intensification and commercialization of agriculture at the household level (Table 30). These variables were used to run some multiple linear regressions and principal component linear regressions. Table 30. Variables selected for regression analyses and regional breakdowns Variable Description Rural Population Density Boserup Average family age Dependents per household Chayanov Distance No. financial institutions No. State institutions Von Braun Minimum 2 Maximum 2 Location (Long, Lat) Environmental factors Mountain range Index of quality of life (ICV) Farm size Variables for regional breakdown This section presents the spatial distribution of these variables, and a brief discussion of their associated patterns. Original Variables The location variables (Table 30) do not require a map to describe their spatial patterns. The reason for this is that all the coffee growing municipalities are in the western hemisphere and north of the Equator. As a result, latitude increases from south to north, and longitude increases from east to west. The other two environmental factors
293 (rainfall in the wettest consecutive two months, rainfall in the driest consecutive two months) reveal some interesting spatial patterns. While in the two consecutive driest months there is a clear trend from lower rainfall in the northeast part of the country and more rainfall in the southwest section of the coffee growing area (Figure 39), for the wettest months the pattern is different. The highest cumulative rainfall in the two wettest months is largest in the north, south and western parts of the country (Figure 40). This clearly shows the marked seasonality of rainfall in the northern part of the coffee growing region, with a stronger dry season as one moves north. A visual examination of these figures indicates that the spatial distribution of the cumulative rainfall of the two wettest and the two driest consecutive months is not randomly distributed. The Moranâ€™s I value for the cumulative rainfall in the two consecutive driest months is 0.924 and the Z-score is 36.862 (= -0.00186, =0.0282,). The pattern for the cumulative rainfall in the two consecutive wettest months is slightly less clustered (Moranâ€™s I= 0.899, Z-score= 31.961), and it is not likely the result of a random process (= -0.00186, =0.0282,). The importance of the amount of rainfall in the driest months for coffee production is highlighted by the fact that the cumulative rainfall in the two driest consecutive months has a higher correlation with the area in coffee (Spearman-Rank correlation = 0.41, p < 0.0001) than with the cumulative rainfall in the two wettest consecutive months (Spearman-Rank correlation = 0.38, p < 0.0001). The highest values of rural population density are concentrated in areas of the western part of the country (traditionally associated with coffee production), and in the areas of the Colombian Andes in the East-Central portion and the southernmost regions, which are characterized by very small holdings (Figure 41).
294 Figure 39. Average cumulative rainfall in the two driest consecutive months
295 Figure 40. Average cumulative rainfall in the two wettest consecutive months
296 Figure 41. Rural population density
297 The population density is not highly correlated with the area in coffee at the municipal level (Spearman-Rank correlation = 0.18, p < 0.001), but exhibits a slightly larger correlation coefficient with the area in intensive coffee for each municipality (Spearman-Rank correlation = 0.22, p < 0.001). This would provide some support to Boserupâ€™s idea that intensification of agriculture is the result of land scarcity (i.e., increased population density). The Moranâ€™s I value for this spatial pattern is 0.421 and the Z-score is 15.01 (= -0.00186, =0.0282,), suggesting a non-random spatial pattern, where municipalities of high population density are located close to each other. The majority of municipalities in the country tend to be relatively young with the exception of a large group of municipalities in the Eastern Cordillera in the center of the country (Figure 42). The average family age in the rural areas of coffee growing municipalities is 25.6 years. These areas of relatively high average age also correspond to municipalities with fewer dependents per household (Figure 43). The average number of dependents per household is 2.1. According to Chayanovâ€™s model of peasant farms, as the number of able bodied members (i.e., fewer dependents) increases as families age, the drudgery associated with agricultural work decreases, and more hands means more land could be planted and larger farms (in a situation without land scarcity). The Spearman-Rank correlation among average age and the number of dependents is -0.65 (p < 0.001, indicating that as families age, there are fewer dependents as Chayanovâ€™s model suggest. The correlation between the number of dependents per household and farm size (Spearman-Rank correlation = 0.12, p =0.007) and the average age and farm size (Spearman-Rank correlation=-0.35, p <0.001) do not support the idea of larger farms as
298 Figure 42. Average age in rural areas
299 Figure 43. Number of dependents per rural household
300 Figure 44. Average size of coffee farms
301 as families have more able-bodied members because the correlation coefficient indicates that as families age, farms tend to become smaller. The reason for this is that there is land scarcity in the Colombian coffee lands, with the number of coffee farms doubling in the period from 1970 to 1997 (FNC 1970, 1997). Farms are smaller in locations where population density tends to be high, suggesting land scarcity (Figure 41 and Figure 44). The correlation among population density and farm size indicates that as farm size decreases, population density increases (Spearman-rank correlation= -0.55, p < 0.001). The spatial pattern of average age has a Moranâ€™s I value of 0.624 and a Z-score of 22.201 (= -0.00186, =0.0282), therefore, it is not the result of a random process. The number of dependents per household is also a non-randomly generated spatial pattern, however, the clustering of similar values is less marked than in the average age case (Moranâ€™s I = 0.315, Z-score= 11.234, = -0.00186, =0.0282). The spatial pattern of the average farm size has a Moranâ€™s I value of 0.585 and a Z-score of 20.824 (= -0.00186, =0.0282), therefore having a very strong clustering of similar values. The theoretical models of commercialization of agriculture (von Braun 1995) indicate that for successful commercialization it is essential to improve accessibility, access to markets and credit, and an adequate institutional framework. Figures Figure 45, Figure 46, Figure 47 present the spatial distribution of three variables that try to capture these aspects of the theoretical model (number of financial and state institutions, and the average distance to the closest town). The number of financial institutions in most municipalities is very low, and only a handful of municipal seats of power have large numbers of banks and similar institutions. These municipalities correspond to departmental capitals.
302 Figure 45. Number of financial institutions per municipality.
303 Figure 46. Number of state institutions per municipality
304 Figure 47. Average distance to the closest town
305 Figure 48. Municipal rural Quality of Life Index (ICV)
306 The spatial pattern of the number of financial institutions per municipality yields a Moranâ€™s I value of -0.010 and a Z-score of -0.275, making it one of the few variables whose spatial distribution could be the result of a random process (= -0.00186, =0.0282). The correlation coefficient between the area in intensive coffee and the number of financial institutions per municipality is 0.42 (Spearman-Rank, p < 0.001), indicating that it tends to be more area in intensive coffee in municipalities with more financial institutions (i.e., more access to credit). The spatial distribution of the number of state institutions per municipality is the other variable that could be the result of a random process (Figure 46). The Moranâ€™s I value for this pattern is 0.007 and its Z-score is 0.318 (= -0.00186, =0.0282). The number of state institutions shows a statistically significant correlation with the area in intensive coffee, but weaker than the number of financial institutions (Spearman-rank correlation= 0.36, p<0.001). This would seem to suggest that the access to credit is more important than the institutional framework in its relation to the intensification of coffee production. The distance to the closest municipality shows a pattern very similar to the Rural Population density map (Figure 41) and the Average Farm Size Map (Figure 44). In Figure 47, there are two clusters of municipalities, one in northwestern part of the country, and the other in the Eastern Cordillera in the middle of the coffee growing area, that are, on average, closer to towns than the rest of the country. There is a very weak correlation between the area in intensive coffee and the distance to the closest town (Spearman-rank correlation= 0.12, p= 0.004), suggesting either that the way in which the average distance was calculated is not a good measure of accessibility, or that in this
307 particular case, accessibility is not as important as the institutional forces described above. The spatial pattern yields a Moranâ€™s I of 0.462 and a Z-score of 16.459 (= -0.00186, =0.0282). This confirms the fact that low distance values are clustered in certain areas of the country. The Index of Quality of Life (ICV) is closely related to coffee production. The highest values of ICV are associated with the location where most of the coffee in Colombia is produced (Figure 48). However, the correlation coefficient between the ICV and the Area in intensive coffee is only 0.27 (Spearman-rank, p< 0.001), suggesting that the improved standards of living in these municipalities with higher ICV are only marginally related to coffee production. This is surprising because the revenues from coffee have been used to improve the infrastructure, education, health, and living standards of coffee farmers in many areas in Colombia (Arango et al. 1998; Chalarc 1998, 2000). The spatial pattern of this variable is clearly clustered, and yields a Moranâ€™s I of 0.611 and a Z-score of 21.730 (= -0.00186, =0.0282). Principal Components The spatial analysis of the principal component scores for each municipality reveals some interesting patterns. For the first principal component (PC1) (Figure 49) there is an astonishingly clear spatial pattern increasing from NE to SW. A visual analysis of this figure clearly suggests that the values of PC1 are not randomly distributed. The Moranâ€™s I value for this pattern is 0.909, and the Z value 32.415, clearly not the result of a random process (= -0.00186, =0.0282). This pattern is related to the X and Y coordinates of each municipality and the influence of the cumulative rainfall in the two driest and wettest months, which shows a very clear spatial pattern (Figure 39 and Figure 40) similar to that one of PC1 (Figure 49). In areas where the value of PC1 is negative the
308 influence of location is stronger than the influence of rainfall. In regions where the value is positive, the values of the wettest and or driest months contribute more than the location variables because the factor loadings of this principal component for the X and Y coordinates are negative, while those of the rainfall variables are positive. The other three principal components do not have such marked spatial patterns. The second principal component (PC2) (Figure 50) also shows some clusters of high and low values. As the family average age and the rural population density increase, the value of PC2 decreases. On the other hand, as the number of dependents per household and the distance to the closest town increases, so does the value of PC2. Therefore, in areas of positive PC2 values the number of dependents per household and the distance to the closest town tends contribute more to the value of PC2. Areas of negative PC2 values are associated with increasing family age and higher population densities. The areas of high PC2 values are associated with areas of very small land holdings, smaller municipalities (i.e., towns closer to each other), high population densities, and large families (Figure 50). This is somewhat contrary to the interpretation of PC2 based on the standardized contribution of each variable, and highlights the difficulties of interpreting correctly the principal components. The Moranâ€™s I value for this spatial pattern is 0.4384 and its associated Z value is 15.616 (= -0.00186, =0.0282). Therefore, the spatial pattern of PC2 is not likely caused by a random process. The third principal component (PC3) shows a very similar pattern to that of PC2 (Figure 51 and Figure 50 respectively). The clusters of low and high values are in similar locations. PC3 summarizes mostly institutional factors (number of financial institutions and state institutions in each municipality). As the number of institutions increases, so
309 Figure 49. Spatial distribution of the first principal component in the coffee growing municipalities
310 Figure 50. Spatial distribution of the second principal component in the coffee growing municipalities
311 Figure 51. Spatial distribution of the third principal component in the coffee growing municipalities.
312 Figure 52. Spatial distribution of the fourth principal component in the coffee growing municipalities.
313 does this principal component. In areas with positive PC3 values, the influence of institutions is very strong, while in areas with negative PC3 values, other variables that also contribute to this principal component (i.e., rainfall in the two driest consecutive months, Y coordinate) contribute more to the principal component score than the institutional forces. The Moranâ€™s I value for this spatial pattern is 0.2631 and its associated Z value is 9.399 (= -0.00186, =0.0282). Therefore, the spatial pattern of PC3 is not likely caused by a random process, but is closer to the critical value for Z, making this pattern more random than those of PC1 and PC2. The fourth principal component (PC4) appears visually as a more random pattern (Figure 52). Its composition is similar to that of PC2, summarizing mostly socioeconomic forces, however some of the relationship between PC4 and the original variables are the opposite to those of PC2 (i.e., relationships dependents and PC4, and average family age and PC4). As rural population density and the number of dependents increases, PC4 values decrease. On the other hand, as the average family age and the distance to the closest town increases, so does PC4. Therefore, this principal component is a good indicator of overcrowded areas, large and young families. The areas of negative values are associated with regions of very small land holdings with large families, so as the population density and the number of dependents per household increase they contribute more to the PC4 score. In areas of positive values, the distance to the closest town and the average family age seem to contribute more to the PC4 score. The Moranâ€™s I value for this spatial pattern is 0.3865 and its associated Z value is 13.778 (= -0.00186, =0.0282). As a result, it can be concluded that the spatial pattern of PC4 is not likely caused by a random process.
LIST OF REFERENCES Anonymous. 2004a. El desempleo es 'la otra broca' del eje cafetero. El Tiempo, 02/01/2004. Anonymous. 2004b. En Colombia hay mucha tierra concentrada en pocas manos. Portafolio, 03/18/2004, N.A. Adams, W. M., and M. J. Mortimore. 1997. Agricultural intensification and flexibility in the Nigerian Sahel. The Geographical Journal 163 (2):150-160. Agarwal, C., G. M. Green, J. M. Grove, T. P. Evans, and C. M. Schweik. 2002. A review and assessment of land-use change models: Dynamics of space, time, and human choice. Newton Square: USDA Forest Service, CIPEC Indiana University. Allen, J. C., and D. F. Barnes. 1985. The causes of deforestation in developing countries. Annals of the Association of American Geographers 75 (2):163-184. Allen, T. F. H., and T. W. Hoekstra. 1992. Toward a unified ecology. Edited by T. F. H. Allen and R. V. O'Neill, Complexity in ecological systems series. New York: Columbia University Press. Anselin, L. 2001. Spatial effects in econometric practice in environmental and natural resources. American Journal of Agricultural Economics 83 (3):705-710. Arango, O., Federacin Nacional de Cafeteros de Colombia (FNC), Corporacin Autnoma Regional de Risaralda (CARDER), and Fundacin Espiral. 1998. Caf, medio ambiente y desarrollo social. Bogot: FNC, CARDER, Fundacin Espiral. Baggio, A. J., P. H. Caramori, A. Androcioli Filho, and L. Montoya. 1997. Productivity of southern Brazilian coffee plantations shaded by different stockings of Grevillea robusta. Agroforestry Systems 37 (2):111-120. Baker, L. 1989. A review of models of landscape change. Landscape Ecology 2 (2):111-133. Banco de la Repblica (BANREP). 2002. Exportaciones no tradicionales 1970-2001 [MS Excel Spreadsheet]. Banco de la Repblica, Colombia 2002a [cited August 2002]. Available from www.banrep.gov.co . 314
315 â€”â€”â€”. 2002. Exportaciones tradicionales 1970-2001 [MS Excel Spreadsheet]. Banco de la Repblica, Colombia 2002b [cited August 2002]. Available from www.banrep.gov.co . â€”â€”â€”. 2004. Apertura econmica [WWW]. Banco de la Repblica N.D [cited January 2004]. Available from http://www.lablaa.org/ayudadetareas/economia/econo89.htm . Barber, G. M. 1988. Elementary statistics for geographers. New York: Guilford Press. Bebbington, A. 1996a. Movements, modernization, and markets. Indigenous organizations and agrarian strategies in Ecuador. In Liberation ecologies. Environment, development, social movements, edited by R. Peet and M. Watts. New York: Routledge. â€”â€”â€”. 1996b. Organizations and intensifications: Campesino federations, rural livelihoods and agricultural technology in the Andes and Amazonia. World Development 24 (7):1161-1177. â€”â€”â€”. 1997. Social capital and rural intensification: Local organizations and islands of sustainability in the rural Andes. The Geographical Journal 163 (2):189-197. Berdie, D. R., J. F. Anderson, and M. A. Niebuhr. 1986. Questionnaires: Design and use. 2nd ed. Metuchen: The Scarecrow Press, Inc. Bilsborrow, R. E., and D. L. Carr. 2001. Population, agricultural land use and environment in developing countries. In Tradeoffs and synergies: Agricultural intensification, economic development and the environment, edited by D. R. Lee and C. B. Barrett. Wallingford: CABI Publishing. Bilsborrow, R. E., and H. W. O. Okoth Ogendo. 1992. Population-driven changes in land use in developing countries. Ambio 21 (1):37-45. Binswanger, H. P. 1986. Evaluating research system performance and targeting research in land-abundant Sub-Saharan Africa. World Development 14 (4):469-475. Blaikie, P. 1995. Changing environments or changing views? A political ecology for developing countries. Geography 80 (348):203-214. Blaikie, P. M. 1994. Political ecology in the 1990s: An evolving view of nature and society. East Lansing: CASID. Boserup, E. 1965. The conditions of agricultural growth: The economics of agrarian change under population pressure. Chicago: Aldine Publishing Company. Bouchard, A., and G. Domon. 1997. The transformations of the natural landscapes of the Haut-Saint-Laurent (Qubec) and their implications on future resource management. Landscape and Urban Planning 37:99-107.
316 Briassoulis, H. 2003. Analysis of land use change: Theoretical and modeling approaches [Online book]. Regional Research Institute, West Virginia University 2000 [cited August 2003]. Available from http://www.rri.wvu.edu/WebBook/Briassoulis/contents.htm . Brookfield, H. 2001. Intensification, and alternative approaches to agricultural change. Asia Pacific Viewpoint 42 (2/3):181-192. Brown, L. R. 2001. Eco-economy: Building an economy for the earth. First ed. New York: Earth Policy Institute, W.W. Norton & Company. Brown, S., and B. Shrestha. 2000. Market-driven land-use dynamics in the middle mountains of Nepal. Journal of Environmental Management 59:217-225. Brown, S. J. 1996a. Low impact, high interest. Organic coffees: An overview, part II. Tea and Coffee Trade Journal 168 (10):32-36. â€”â€”â€”. 1996b. Organic coffees: An overview, part I. Tea and Coffee Trade Journal 168 (9):120-123. Brunnschweiler, D. 1972. The Llanos frontier of Colombia; environment and changing land use in Meta. East Lansing: Michigan State University Latin American Studies Center. Brush, S. B., and B. L. Turner. 1987. The nature of farming systems and views of their change. In Comparative farming systems, edited by B. L. Turner and S. B. Brush. New York: The Guilford Press. Burel, F., and J. Baudry. 2003. Landscape ecology: Concepts, methods and applications. Enfield: Science Publishers Inc. Brgi, M., and E. W. B. Russell. 2001. Integrative methods to study landscape changes. Land Use Policy 18:9-16. Campbell, J. B. 1996. Introduction to remote sensing. 2nd ed. New York: The Guilford Press. Carney, J. 1996. Converting the wetlands, engendering the environment. The intersection of gender with agrarian change in Gambia. In Liberation ecologies. Environment, development, social movements, edited by R. Peet and M. Watts. New York: Routledge. Carriker, M. R. 2001. Vista nieve. Rio Hondo: Blue Mantle Press. Caspersen, O. H. 1999. Population and land use changes in the central highland province of Dak Lak, Vietnam. Danish Journal of Geography Special Issue (1):47-54.
317 Caviedes, C. N. 2001. El Nio in history. Storming through the ages. Gainesville: University Press of Florida. Chalarc, J. 1998. Vida y hechos del caf en Colombia. 1. ed. Santaf de Bogot: J. Chalarca. â€”â€”â€”. 2000. La caficultura en el Huila. Historia y desarrollo. Bogot: Comit Departamental de Cafeteros de Colombia-Comn Presencia Editores. Chattaerjee, S., and B. Price. 1991. Regression analysis by example. 2nd ed. New York: John Wiley & Sons. Cherrier, A., and J. Berthaud. 1985. Botanical classification of coffee. In Coffee: Botany, biochemistry and production, edited by M. N. Clifford and K. C. Willson. Westport: Croom Helm Ltd AVI Publishing. Conelly, W. T., and M. S. Chaiken. 2000. Intensive farming, agro-diversity, and food security under conditions of extreme population pressure in western Kenya. Human Ecology 28 (1):19-51. Conservation International (CI). 2003. Conservation newsroom [WWW]. Conservation International 2003 [cited November 2003]. Available from http://www.conservation.org/xp/news/campaigns/coffee.xml . Conway, G. 2001. The doubly green revolution: Balancing food, poverty and environmental needs in the 21st century. In Tradeoffs and synergies: Agricultural intensification, economic development and the environment, edited by D. R. Lee and C. B. Barrett. Wallingford: CABI Publishing. Coomes, O. T., and B. L. Barham. 1997. Rain forest extraction and conservation in Amazonia. The Geographical Journal 163 (2):180-188. Coomes, O. T., and G. J. Burt. 1997. Indigenous market-oriented agroforestry: Dissecting local diversity in western Amazonia. Agroforestry Systems 37 (1):27-44. XVI censo nacional de poblacin y vivienda 1.0. Departamento Administrativo Nacional de Estadstica (DANE), Bogot. D'aquino, P., P. August, A. Balmann, T. Berger, F. Bousquet, B. E., D. G. Brown, H. Couclelis, P. Deadman, M. F. Goodchild, N. M. Gotts, G. J. Gumerman, M. J. Hoffmann, M. G. A. Huigen, E. Irwin, M. A. Janssen, R. Johnston, T. Kholer, N. L. R. Law, V. Lee, C. Le Page, K. Lim, S. M. Manson, W. J. McConnell, S. McCracken, E. F. Moran, R. Najlis, J. I. Nassauer, J. J. Opaluch, S. E. Page, D. C. Parker, G. J. Polhill, D. Robinson, R. Thompson, P. Torrens, and K. Warren. 2001. Agent-based models of land-use and land-cover change. Bloomington: LUCC Focus 1 Office.
318 Denevan, W. M. 1963. Pre-spanish earthworks in the Llanos de Mojos of northeastern Bolivia. Annals of the Association of American Geographers 53 (4):586-586. Departamento Administrativo Nacional de Planeacin (DNP). 2003. Indicadores sociodemogrficos para el sector rural [MS Word file]. Direccin de Desarrollo Agrario, Departamento Administrativo Nacional de Planeacin 2000 [cited November 2003]. Available from http://www.dnp.gov.co/ArchivosWeb/Direccion_Desarrollo_Agrario/boletines/Boletin1/ind_sect_rural.doc . Dicum, G., and N. Luttinger. 1999. The coffee book: Anatomy of an industry from crop to the last drop, The bazaar books series. New York: The New York Press. Di Pietro, F. 2001. Assessing ecologically sustainable agricultural land-use in the central Pyrnes at the field and landscape level. Agriculture, Ecosystems and Environment 86:93-103. Ditton, R. B., H. R. Osburn, T. L. Baker, and C. E. Thailing. 2002. Demographics, attitudes, and reef management preferences of sport divers in offshore Texas waters. Ices Journal of Marine Science 59:S186-S191. Dixon, C. J., and B. Leach. 1978. Questionnaires and interviews in geographical research, Concepts and techniques in modern geography ; 18. Norwich Eng.: Geo Abstracts. Dorsey, B. 1999. Agricultural intensification, diversification, and commercial production among smallholder coffee growers in central Kenya. Economic Geography 75 (2):178-195. Duque, H., M. Restrepo, and R. Velsquez. 2000. Estudio sobre cosecha de caf y mano de obra en Palestina, Caldas. Chinchin: CENICAF. Errzuriz, M. 1986. Cafeteros y cafetales del Lbano: Cambio tecnolgico y diferenciacin social en una zona cafetera. Bogot: Universidad Nacional de Colombia. Escobar, O. L., and G. Ferro. 1991. Cultura del hombre cafetero. Pereira: Banco de la Repblica-Area cultural. Etter, A., and W. van Wyngaarden. 2000. Patterns of landscape transformation in Colombia, with emphasis in the Andean region. Ambio 29 (7):432-439. Etter, A., and L. A. Villa. 2000. Andean forests and farming systems in part of the eastern cordillera (Colombia). Mountain Research and Development 20 (3):236-245. Federacin Nacional de Cafeteros de Colombia (FNC). 1970. Censo cafetero. Anexo. Bogot: Federacin Nacional de Cafeteros de Colombia.
319 â€”â€”â€”. 1975. Calidad de los datos estadsticos del censo. Revista Cafetera de Colombia 24 (161):10-11. â€”â€”â€”. 1976. Atlas cafetero de Colombia. Bogot: Divisin de Investigaciones Econmicas, Federacin Nacional de Cafeteros de Colombia. â€”â€”â€”. 1983. Censo cafetero 1980-81. Bogot: Federacin Nacional de Cafeteros de Colombia. â€”â€”â€”. 1997. Sistema de informacin cafetera. Encuesta nacional de cafeteros SICA. Informe final. Santaf de Bogot: Federacin Nacional de Cafeteros de Colombia. â€”â€”â€”. 2001. Cartilla cafetera caf: Comit de Cafeteros del Valle del Cauca CENICAFE. Flather, C. H., and M. Bevers. 2002. Patchy reaction-diffusion and population abundance: The relative importance of habitat amount and arrangement. The American Naturalist 159 (1):40-56. Foster, D., G. Motzkin, and B. Slater. 1998. Land-use history as long term broad-scale disturbance: Regional forest dynamics in central New England. Ecosystems 1:96-119. Geist, H. J., and E. F. Lambin. 2002. Proximate causes and underlying driving forces of tropical deforestation. BioScience 52 (2):143-150. Genghini, M., F. Spalatro, and S. Gellini. 2002. Farmers' attitudes toward the carrying out of wildlife habitat improvement actions (whia) in intensive agricultural areas of northern Italy. Zeitschrift Fur Jagdwissenschaft 48:309-319. Gibson, C. C., E. Ostrom, and T. K. Ahn. 2000. The concept of scale and the human dimensions of global change: A survey. Ecological Economics 32:217-239. Gobbi, J. A. 2000. Is biodiversity-friendly coffee financially viable? An analysis of five different coffee production systems in western El Salvador. Ecological Economics 33 (2000):267-281. Goldman, A. 1993. Agricultural innovation in three areas of Kenya: Neo-boserupian theories and regional characterization. Economic Geography 69:44-71. Gregory, P. J., and J. S. I. Ingram. 2000. Global change and future food and forest production: Future scientific challenges. Agriculture, Ecosystems and Environment 82:3-14. Haggett, R., and R. J. Chorley. 1967. Models, paradigms and the new geography. In Integrated models in geography, edited by R. J. Chorley and P. Haggett. London: Methuen & Co. Ltd.
320 Hamer, J., and I. Hamer. 1994. Impact of a cash economy on complementary gender relations among the Sadama of Ethiopia. Anthropological Quarterly 67 (4):187-202. Harvey, D. 1967. Models of the evolution of spatial patterns in human geography. In Integrated models in geography, edited by R. J. Chorley and P. Haggett. London: Methuen & Co. Ltd. Hietala-Koivu, R. 2002. Landscape and modernizing agriculture: A case study of three areas in Finland in 1954-1998. Agriculture, Ecosystems and Environment 91:273-281. Hobbs, R. 1997. Future landscapes and the future of landscape ecology. Landscape and Urban Planning 37:1-9. Hll, A., and K. Nilsson. 1999. Cultural landscape as subject to national research programmes in Denmark. Landscape and Urban Planning 46:15-27. Holling, C. S. 1992. Cross-scale morphology, geometry, and dynamics of ecosystems. Ecological Monographs 62 (4):447-502. Holling, C. S., and L. H. Gunderson. 2002. Resilience and adaptive cycles. In Panarchy: Understanding transformations in human and natural systems, edited by L. H. Gunderson and C. S. Holling. Washington: Island Press. Hoshino, S. 2001. Multilevel modeling on farmland distribution in Japan. Land Use Policy 18:75-90. Houghton, R. A. 1994. The worldwide extent of land-use change. BioScience 44 (5):305-313. Hyden, G., R. W. Kates, and B. L. Turner. 1993. Beyond intensification. In Population growth and agricultural change in africa, edited by B. L. Turner, G. Hyden and R. W. Kates. Gainesville: University Press of Florida. Atlas hidrometeorolgico de colombia 1.0. IDEAM, Bogot. Imbernon, J. 1999. Changes in agricultural practice and landscape over a 60-year period in north Lampung, Sumatra. Agriculture, Ecosystems and Environment 76:61-66. International Coffee Organization (ICO). 2003. Historical statistical data [Database]. International Coffee Organization 2003a [cited September 2003]. Available from www.ico.org . â€”â€”â€”. 2003. Letter from the u.S. Congress international relations committee to the u.S. Secretary of the interior [Adobe Acrobat File]. International Coffee Organization 2003b [cited September 2003]. Available from http://www.ico.org/ed/colombia/congress.pdf .
321 International Geosphere Biosphere Program (IGBP). 2003. International geosphere biosphere program [WWW] 2003 [cited October 2003]. Available from http://www.igbp.kva.se/ . Irwin, E. G., and J. Geoghegan. 2001. Theory, data, methods: Developing spatially explicit economic models of land use change. Agriculture, Ecosystems and Environment 85 (1-3):7-23. Jansen, L. J. M., and A. Di Gregorio. 2002. Parametric land cover and land-use classification as tools for environmental change detection. Agriculture, Ecosystems and Environment 91:89-100. Jaramillo, A., and B. Chves. 1999. Aspectos hidrolgicos en un bosque y en plantaciones de caf (coffea arabica l.) al sol y bajo sombra. Cenicaf 50 (2):97-105. Jones, B. 2003. Citrus and black pepper in san carlos, northern costa rica [WWW]. The Vegetable Liberation Front N.D. [cited November 2003]. Available from http://www.veggielib.org/cragro/ . Knapp, G., K. Mathewson, and W. M. Denevan. 1985. Pre-hispanic agricultural fields in the Andean region. Paper read at 45 International Congress of Americanists, at Bogot. Kummer, D., and B. L. Turner. 1994. The human causes of deforestation in southeast Asia. Bioscience 44 (5):323-328. Lambin, E. F., X. Baulies, N. Bockstael, G. Fisher, T. Krug, R. Leemans, E. F. Moran, R. R. Rindfuss, Y. Sato, D. Skole, B. L. Turner, and C. Vogel. 1999. Land-use and land-cover change: Implementation strategy. Stockholm and Geneva: International Congress of Scientific Unions and International Science Council. Lambin, E. F., and H. J. Geist. 2001. Global land-use and land-cover change: What have we learned so far? Global Change Newsletter, June 2001, 27-30. Lambin, E. F., M. D. A. Rounsevell, and H. J. Geist. 2000. Are agricultural land-use models able to predict changes in land-use intensity? Agriculture, Ecosystems and Environment 82:321-331. Lambin, E. F., B. L. Turner, H. J. Geist, S. B. Agbola, A. Angelsen, J. W. Bruce, O. T. Coomes, R. Dirzo, G. Fisher, C. Folke, P. S. George, K. Homewood, J. Imbernon, R. Leemans, L. Xiubin, E. F. Moran, M. Mortimore, P. S. Ramakrishnan, J. F. Richards, H. Sknes, W. Steffen, G. D. Stone, U. Svedin, T. A. Veldkamp, C. Vogel, and J. Xu. 2001. The causes of land-use and land-cover change: Moving beyond the myths. Global Environmental Change 11 (4):261-269.
322 Lee, D. R., P. J. Ferraro, and C. B. Barrett. 2001. Introduction: Changing perspectives on agricultural intensification, economic, development and the environment. In Tradeoffs and synergies: Agricultural intensification, economic development and the environment, edited by D. R. Lee and C. B. Barrett. Wallingford: CABI Publishing. Level Ground Trading (LGT). 2003. How the profits flow? [WWW]. Level Ground Trading 2003 [cited November 2003]. Available from www.levelground.com . Liverman, D. 1994. Modeling social systems and their interaction with the environment: A view from geography. In Integrated regional models: Interactions between humans and their environments, edited by P. M. Groffman and G. E. Likens. New York: Champan & Hall. Londoo, J. G. 2003. Casi lista la reestructuracin de los estatutos de la Federacin Nacional de Cafeteros. El Tiempo, 10 Dec 2003, N.A. Mannion, A. M. 2002. Dynamic world. Land-cover and land-use change. New York: Arnold Press. Marquette, C. 1998. Land use patterns among small farmer settlers in the northeastern Ecuadorian Amazon. Human Ecology 26 (4):573-598. McKenzie, F. T. 1998. Our changing planet: An introduction to earth system science and global environmental change. Upper Saddle River: Prentice Hall. Menzies, G. 2002. 1421: The year China discovered America. New York: Harper Collings Publishers. Ministerio de Agricultura y Desarrollo Rural (MinAgricultura). 2002. Estadsticas agrcolas por consenso [Database]. Ministerio de Agricultura y Desarrollo Rural N.D. [cited Nov 2002]. Available from http://www.minagricultura.gov.co/sistemas/interfaces/cnsevaluaciones.asp . Moguel, P., and V. M. Toledo. 1999. Biodiversity conservation in traditional coffee systems of Mexico. Conservation Biology 13 (1):11-21. Moran, E. F., and E. Brondizio. 1998. Land-use change after deforestation in Amazonia. In People and pixels. Linking remote sensing and social science, edited by D. Liverman, E. F. Moran, R. R. Rindfuss and P. C. Stern. Washington D.C.: National Academy Press. Moreno, L. G. 2002. Tabi: Variedad de caf de porte alto con resistencia a la roya. Chinchin: Cenicaf. Motzkin, G., D. Foster, A. Allen, J. Harrod, and R. Boone. 1996. Controlling site to evaluate history: Vegetation patterns of a New England sand plain. Ecological Monographs 66 (3):345-365.
323 Murphy, L., R. Bilsborrow, and F. Pichn. 1997. Poverty and prosperity among migrant settlers in the Amazon rainforest frontier of Ecuador. Journal of Development Studies 34 (2):35-65. Muschler, R. G. 2001. Shade improves coffee quality in a sub-optimal coffee-zone in Costa Rica. Agroforestry Systems 85:131-139. NCSS and PASS. Number cruncher statistical systems 2001. NCSS, Kaysville. Nestel, D. 1995. Coffee in Mexico: International market, agricultural landscape and ecology. Ecological Economics 15 (2):165-178. Netting, R. 1993. Smallholders, householders: Farm families and the ecology of intensive, sustainable agriculture. Stanford, CA: Stanford University Press. Neuman, W. L. 2000. Social research methods: Qualitative and quantitative approaches. Edited by S. L. Kelbaugh. 4th ed. Needham Heights: Allyn and Bacon. Niehaus, D. J. 1992. Slope instability hazard assessment for natural disaster reduction, an approach using remote sensing analysis and geographical information systems: A case study in the central cordillera of Colombia. M.Sc. Thesis, International Institute for Aerospace Survey and Earth Sciences (ITC), Eschende, The Netherlands. Ochoa-Gaona, S. 2001. Traditional land-use systems and patterns of forest fragmentation in the highlands of Chiapas, Mexico. Environmental Management 27 (4):571-576. Odum, E. P. 1997. Ecology: A bridge between science and society. Sunderland: Sinauer Associates, Inc. publishers. Ojima, D. S., K. A. Galvin, and B. L. Turner. 1994. The global impact of land-use change. BioScience 44 (5):300-304. Ortiz, A. P. 1989. Sombros y caturrales del Lbano-Tolima. Transformacin y crisis ecolgica de un paisaje cafetero. Anlisis metodolgico y cartografa integrada. Vol. 13, Anlisis geogrficos. Bogot: IGAC. Ostrom, E., J. Burger, C. B. Field, R. B. Norgaard, and D. Policansky. 1999. Revisiting the commons: Local lessons, global challenges. Science 284:278-282. OXFAM International (OXFAM). 2002. Mugged: Poverty in your coffee cup. London: OXFAM International. Palacios, M. 1980. Coffee in Colombia 1850-1970. Cambridge: Cambridge University Press. Parsons, J. 1968. Antioqueo colonization in western Colombia. 2nd ed, Iberoamericana. Berkeley: University of California Press.
324 Pauwels, F., and H. Gulinck. 2000. Changing minor rural road networks in relation to landscape sustainability and farming practices in west Europe. Agriculture, Ecosystems and Environment 77:95-99. Peluso, N. 1992. Rich forests, poor people. Resource control and resistance in Java. 1st ed. Berkeley: University of California Press. Pealoza, M. T. 2002. Ya hay ms de 700 hectreas de coca en el eje cafetero. La Patria, 19 Sept 2003, N.A. Perfecto, I., A. Mas, T. Dietsch, and J. Vandermeer. 2003. Conservation of biodiversity in coffee agroecosystems: A tri-taxa comparison in southern Mexico. Biodiversity and Conservation 12:1239-1252. Perfecto, I., R. A. Rice, R. Greenberg, and M. E. Van Der Voort. 1996. Shade coffee: A disappearing refuge for biodiversity. BioScience 46 (8):598-608. Perrings, C. 2001. The economics of biodiversity loss and agricultural development in low income countries. In Tradeoffs and synergies: Agricultural intensification, economic development and the environment, edited by D. R. Lee and C. B. Barrett. Wallingford: CABI Publishing. Pingali, P. L., and M. W. Rosegrant. 1995. Agricultural commercialization and diversification: Processes and policies. Food Policy 20 (3):171-185. Pinto-Correia, J. M. 1999. Contribution to the extensification/intensification debate: New trends in the Portuguese Montado. Landscape and Urban Planning 46:125-131. Price, M. 1994. Hands for the coffee: Migrants in western Venezuelaâ€™s coffee production, 1870-1930. Journal of Historical Geography 20 (1):62-80. Pulschen, L., and H. J. Lutzeyer. 1993. Ecological and economic-conditions of organic coffee production in Latin America and Papua-New-Guinea. Angewandte Botanik 67 (5-6):204-208. Rainforest Alliance (RA). 2003. Sustainable agriculture [WWW]. The Rainforest Alliance N.D. [cited November 2003]. Available from http://www.rainforest-alliance.org/programs/cap/program-description7.html . Rao, K. S., and R. Pant. 2001. Land use dynamics and landscape change pattern in a typical micro watershed in the mid elevation zone of central Himalaya, India. Agriculture, Ecosystems and Environment 86:113-123. Rappole, J. H., D. I. King, and J. H. Vega-Rivera. 2003. Coffee and conservation. Conservation Biology 17 (1):334-336. Reenberg, A., and C. Lund. 1998. Land use and land right dynamics -determinants of resource management in eastern Burkina Faso. Human Ecology 26 (4):599-620.
325 Rhoads, B. L., and C. E. Thorn. 1996. The scientific nature of geomorphology : Proceedings of the 27th binghamton symposium in geomorphology, held 27-29 september, 1996. Chichester ; New York: Wiley. Rice, R. A. 1996. Coffee modernization and ecological changes in northern Latin America. Tea and Coffee Trade Journal 168 (9):104-113. â€”â€”â€”. 1997. The land use patterns and the history of coffee in eastern Chiapas, Mexico. Agriculture and Human Values 14:127-143. â€”â€”â€”. 2000. Managed biodiversity in shade coffee systems: The non-coffee harvest. Paper read at Primer foro internacional en caf y biodiversidad, at Manizales, Colombia. Richter, M. 2000. The ecological crisis in Chiapas: A case study from Central America. Mountain Research and Development 20 (4):332-339. Riebsame, W. E., W. J. Parton, K. A. Galvin, I. C. Burke, L. Bohren, R. Young, and E. Knop. 1994. Integrated modeling of land use and cover change. BioScience 44 (5):350-356. Rudel, T. K., M. Perez-Lugo, and H. Zichal. 2000. When fields revert to forest: Development and spontaneous reforestation in post-war Puerto Rico. Professional Geographer 52 (3):386-397. Rueda, J. O., and DANE. 1999. Memoria tcnica del censo 93, Estudios censales, 14. Santaf de Bogot, D.C., Colombia: Departamento Administrativo Nacional de Estadstica DANE. Ruttan, V. W. 1994. Sustainable agricultural growth. In Agriculture, environment, and health, edited by V. W. Ruttan. Minneapolis: University of Minnesota Press. Sarmiento, F. O. 2000. Breaking mountain paradigms. Ecological effects on human impacts in man-aged tropoandean landscapes. Ambio 29 (7):423-431. Sarmiento, F. O., and L. M. Frolich. 2002. Andean cloud forest tree lines. Naturalness, agriculture and the human dimension. Mountain Research and Development 22 (3):278-187. Sarmiento, L., D. C. Castillo, J. Ramos, A. E. Castaeda, M. E. Alvarez, R. Villar, L. Y. Sandoval, W. C. Tellez, L. M. Cuervo, J. Gonzlez, P. M. Ferreira, M. J. Prez, A. Sarmiento, C. L. Ramrez, J. M. Barbero, A. Reyes, M. E. Naranjo, O. Fresneda, P. I. Moreno, and O. A. Alfonso. N.D. Municipios y regiones de Colombia. Una mirada desde la sociedad civil. Bogot, Colombia: Fundacin Social.
326 Schuller, D., H. Brunken-Wincler, P. Busch, M. Frster, P. Janiesh, R. von Lemm, R. Niedringhaus, and H. Strasser. 2000. Sustainable land use in an agriculturally misused landscape in northwest Germany through ecotechnical restoration by a 'patch-network-concept'. Ecological Engineering 16:99-117. Smith, N. J. H. 1999a. Biodiversity and agroforestry along the Amazon floodplain. In Traditional and modern natural resource management in Latin America, edited by F. J. Pichn, J. E. Uquillas and J. Frechiore. Pittsburgh: University of Pittsburgh Press. â€”â€”â€”. 1999b. The natural river forest: A natural history of plants, animals, and people. New York: Oxford University Press. Smith, R. 1985. A history of coffee. In Coffee: Botany, biochemistry and production, edited by M. N. Clifford and K. C. Willson. Westport: Croom Helm Ltd AVI Publishing. Smithsonian Migratory Bird Center (SMBC). 2003. Shade management criteria for "bird friendly" coffee [WWW]. Smithsonian Migratory Bird Center 2003 [cited November 2003]. Available from http://nationalzoo.si.edu/ConservationAndScience/MigratoryBirds/Coffee/Certification/criteria.cfm . Southworth, J., and C. Tucker. 2001. The influence of accessibility, local institutions, and socioeconomic factors on forest cover change in the mountains of western Honduras. Mountain Research and Development 21 (3):276-283. Thomas, D. S. G., and A. Goudie, eds. 2000. The dictionary of physical geography. 3rd ed. Malden: Blackwell Publishers. Thomson-ISI (ISI). 2003. Web of knowledge [Database]. ISI 2003 [cited August 2003]. Available from www.isiknowledge.com . Topik, S. C., and A. Wells. 1998. Introduction: Latin Americaâ€™s response to international markets during the export boom. In The second conquest of Latin America: Coffee, henequen, and oil during the export boom, 1850-1930, edited by S. C. Topik and A. Wells. Austin: University of Texas Press. Trujillo Restrepo, C. A. 1996. Del caf a la coca: Vivencias alrededor de una metamorfosis. Cali, Colombia: Car Tres. Turner, B. L. 1997. The sustainability principle in global agendas: Implications for understanding land-use/cover change. The Geographical Journal 163 (2):133-140.
327 Turner, B. L., and S. B. Brush. 1987. Purpose, classification, and organization. In Comparative farming systems, edited by B. L. Turner and S. B. Brush. New York: The Guilford Press. Turner, B. L., and W. E. Doolittle. 1978. The concept and measure of agricultural intensity. Professional Geographer 30 (3):297-301. Turner, B. L., and A. M. Shajaat Ali. 1996. Induced intensification: Agricultural change in Bangladesh with implications for Malthus and Boserup. Proceedings of the National Academy of Sciences 93:14984-14991. Turner, B. L., D. Skole, G. Sanderson, G. Fisher, L. Fresco, and R. Leemans. 1995. Land-use and land-cover change: Science/research plan. Stockholm and Geneva: International Congress of Scientific Unions and International Science Council. Turner, M. G., R. H. Gardner, and R. V. O'Neill. 2001. Landscape ecology in theory and practice. Pattern and process. New York: Springer-Verlag. Umaa, F. 2003. Ya hay ms de 700 hectreas de coca en el eje cafetero. El Tiempo, 03 March 2003, N.A. United Nations Conference on Trade and Development (UNCTAD). 1984. Studies in the processing, marketing and distribution of commodities. The processing and marketing of coffee. Areas for international co-operation. Geneva: United Nations Conference on Trade and Development. United Nations Food and Agriculture Organization (FAO). 2003. FAOSTAT [Database]. United Nations Food and Agriculture Organization 2001 [cited July 2003]. Available from http://apps.fao.org/default.htm . Uribe, H. 1996. La diversificacin cafetera en Colombia. In Memorias del seminario internacional de experiencias de reestructuracin, edited by J. Vallecilla. Manizales: Centro de estudios cafeteros y empresariales (CRECE). van der Vossen, H. A. M. 1985. Coffee selection and breeding. In Coffee: Botany, biochemistry and production, edited by M. N. Clifford and K. C. Willson. Westport: Croom Helm Ltd AVI Publishing. Veldkamp, A., and E. F. Lambin. 2001. Predicting land-use change. Agriculture, Ecosystems and Environment 85 (1-3):1-6. Vitousek, P. M., H. A. Mooney, J. Lubchenco, and J. M. Melillo. 1997. Human domination of earth's ecosystems. Science 277 (25 July 1997):494-499. von Braun, J. 1995. Agricultural commercialization: Impacts on income and nutrition and implications for policy. Food Policy 20 (3):187-202.
328 Wackernagel, M., and W. E. Rees. 1996. Our ecological footprint : Reducing human impact on the earth, New catalyst bioregional series ; no. 9. Gabriola Island; Philadelphia: New Society Publishers. Walker, R. 2003. Mapping process to pattern in the landscape change of the Amazonian frontier. Annals of the Association of American Geographers 93 (2):376-398. Waridel, L. 2002. Coffee with pleasure. Just java and world trade. Montral: Black Rose Books. Whitmore, T., and B. Turner. 1992. Landscapes of cultivation in Mesoamerica on the eve of the conquest. Annals of the Association of American Geographers 82 (3):402-425. Wiegers, E. S., R. J. Hijmans, D. Herv, and L. D. Fresco. 1999. Land use intensification and disintensification in the upper Caete valley, Peru. Human Ecology 27 (2):319-339. Willson, K. C. 1985a. Climate and soil. In Coffee: Botany, biochemistry and production, edited by M. N. Clifford and K. C. Willson. Westport: Croom Helm Ltd AVI Publishing. â€”â€”â€”. 1985b. Cultural methods. In Coffee: Botany, biochemistry and production, edited by M. N. Clifford and K. C. Willson. Westport: Croom Helm Ltd AVI Publishing. â€”â€”â€”. 1999. Coffee, cocoa and tea. New York: CAB International. Wrigley, G. 1988. Coffee. Edited by G. Wrigley, Tropical agriculture series. Singapore: Longman Scientific & Technical. Yapa, L. S. 1977. The green revolution: A diffusion model. Annals of the Association of American Geographers 67 (3):350-359. Zimmerer, K. 1996a. Changing fortunes. Biodiversity and peasant livelihood in the Peruvian Andes, California studies in critical human geography. Berkeley: University of California Press. â€”â€”â€”. 1996b. Discourses of soil loss in Bolivia. Sustainability and the search for socioenvironmental "middle ground". In Liberation ecologies. Environment, development, social movements, edited by R. Peet and M. Watts. New York: Routledge. â€”â€”â€”. 1999. Overlapping patchworks of mountain agriculture in Peru and Bolivia: Toward a regional-global landscape model. Human Ecology 27 (1):135-165.
BIOGRAPHICAL SKETCH Andrs Guhl was born in Colombia in 1971. After finishing High school in 1989 he went to Universidad de Los Andes in Bogot, where he majored in civil engineering and anthropology. During his time in college he spent a year as an exchange student at the University of Illinois at Urbana-Champaign (UIUC). In 1996 he got a B.Sc. in civil engineering, and started working for a consulting firm in Bogot, Colombia. In the Fall of 1997, he started his graduate studies in geography at UIUC. He completed his M.Sc. in physical and environmental geography in 2000, and then moved to the University of Florida for his Ph.D. in geography. Here at UF he worked on land-use and land-cover change issues, focusing on tropical environments in Latin America. 329