The broad aim of this study is, to contribute to a broader and more comprehensive understanding of the patterns, processes and drivers of unplanned land cover change in the tropics, using Colombia as a study case. Land cover change is an important global issue because of the expanding ecological footprint of a rapidly increasing human population and per capita level of resource consumption. This has a major impact on natural ecosystems and their function at the local (hundreds of square kilometers) and global scales. The understanding of extent and rate of land cover change is an important issue confronting biodiversity conservation, land use planning, protected area management, and global climate change analysis. Tropical deforestation is the major source of global land cover change, with the highest absolute rates occurring in South America, especially in the Brazilian Amazon, where government planning is an important driver of deforestation. However, unplanned deforestation for cropping and ranching is also occurring in Colombia. This is of international concern because Colombia’s diverse ecosystems support high levels of species richness and endemism. Improving the understanding of the spatial and temporal patterns and drivers ofland cover change (both deforestation and regeneration) is an important step in developing planning and conservation strategies to address this problem. I applied a spatial and temporal statistical modelling approach to predict changes in land cover in Colombia at the local (100 km2), regional (104 km2) and national (106 km2) levels, with a timeframe spanning from decades to centuries. As dependant variable data, binary forest/non-forest data are used. Explanatory variables comprise biophysical and socioeconomic data sourced from a broad range of information sources, including remotely sensed data from aerial photographs and satellite images, secondary sources of biophysical and socioeconomic data, and historical data. At the local-level, I addressed the deforestation process over the last 60 years using six case studies of 100 km2 of humid lowland forests, by applying logistic regression and spatial analysis. At the regional-level, I studied the deforestation in the Caquetá colonization front of the Colombian Amazon region from 1988-2004 by applying a forest-cover zoning method and logistic regression models to predict deforestation and forest regeneration from biophysical and socio-economic explanatory variables. At the national level, I quantified and analysed patterns and drivers of land cover change over the past 500 years for key periods of Colombian history, and identified the extent and duration of impacts on broad ecosystem types. At the national and regional-levels, I also modelled current landscape transformation patterns and predicted areas at a high risk of future deforestation using a joint logistic regression and regression tree approach. I discovered that the rate of deforestation across several lowland regions of Colombia follows a simple sigmoid pattern composed of four phases of transformation: an initial phase of gradual forest loss; an intermediate phase of rapid loss; a second intermediate phase where the rate of decline slows; and a final phase where the forest loss stabilises and is balanced by forest regeneration creating a dynamic equilibrium. At the end of this final phase, the landscape is in a highly transformed state with forest cover stabilizing at 2 to 10% of the original extent and an average forest patch size of 15.4 (± 9.2) ha. As a general rule, the transformed landscape will have two forest components: a stable component of remnant mature tropical forests, and a dynamic component of secondary forests of different ages that is repeatedly cleared. A second important discovery was that unplanned deforestation in the Colombian Amazon moves as a colonisation wave, extending from population centres. The rate of movement was 0.84 km.yr-1 between 1989 and 2002. The regional average annual deforestation rate was 2.6%, but varied locally between –1.8% (regeneration) and 5.3%. The parallel deforestation and regeneration processes operating within the colonization front showed consistent patterns and rates directly related to the proportion of forest in the neighbourhood, with the highest rates of deforestation occurring in the areas with intermediate (40-60%) forest cover, following an overall quadratic function, and therefore confirming the sigmoid pattern across an entire colonization front. Landscapes with intermediate forest cover also have the highest density of edge habitat, with the deforestation process mimicking the spread of disease. At the national-level, the study reveals two important outcomes. First, there are significant regional differences in the spatial and temporal patterns and drivers of land cover change. The importance of such regional differences in factors explainingland cover change is highlighted by the greatest discrimination ability shown by a regional-level classification tree model. The coefficients and significance of variables in a regional logistic regression model confirmed these differences. Overall, factors related to accessibility (distance to roads and towns) had the strongest influence on the probability of deforestation. The second national-level outcome highlights the need for a longer-term historical perspective spanning centuries to understand present-day landscapes and their level of human impact. The historical analysis reveals that the main drivers of landscape change varied in the early colonial period, with cattle grazing becoming increasingly important, and finally the high impact of economic globalization in the 1990s. The historic land cover maps show the transformed areas increasing from approximately 15 million ha in 1500 to 42 million ha in 2000. Also, during this period, the transformed areas changed from predominantly cropping land uses in 1500 to predominantly (< 75%) grazing in 2000. The research outcomes collectively provide an understanding of the spatial dynamics of unplannedland cover change in tropical forest landscapes, by showing how deforestation and regeneration processes vary along a transformation gradient, and linking the amount of remnant forests is related to the rate of change. The work has implications for policy and management. For example, the method of calculating the movement speed of a colonization front provides a spatially explicit prediction of threat that can be used in conservation planning. The improved understanding of the deforestation and regeneration dynamics over an entire colonization fronts permits more accurate calculation of carbon budgeting for climate change applications. I recommend that future work test of the generalities presented here in both, countries with unplanned deforestation and countries with planned land clearing, to evaluate the effects government controlled planning has on the end result of human transformed landscapes in tropical and subtropical forest regions.

The broad aim of this study is, to contribute to a broader and more comprehensive understanding of the patterns, processes and drivers of unplanned land cover change in the tropics, using Colombia as a study case. Land cover change is an important global issue because of the expanding ecological footprint of a rapidly increasing human population and per capita level of resource consumption. This has a major impact on natural ecosystems and their function at the local (hundreds of square kilometers) and global scales. The understanding of extent and rate of land cover change is an important issue confronting biodiversity conservation, land use planning, protected area management, and global climate change analysis. Tropical deforestation is the major source of global land cover change, with the highest absolute rates occurring in South America, especially in the Brazilian Amazon, where government planning is an important driver of deforestation. However, unplanned deforestation for cropping and ranching is also occurring in Colombia. This is of international concern because Colombia’s diverse ecosystems support high levels of species richness and endemism. Improving the understanding of the spatial and temporal patterns and drivers ofland cover change (both deforestation and regeneration) is an important step in developing planning and conservation strategies to address this problem. I applied a spatial and temporal statistical modelling approach to predict changes in land cover in Colombia at the local (100 km2), regional (104 km2) and national (106 km2) levels, with a timeframe spanning from decades to centuries. As dependant variable data, binary forest/non-forest data are used. Explanatory variables comprise biophysical and socioeconomic data sourced from a broad range of information sources, including remotely sensed data from aerial photographs and satellite images, secondary sources of biophysical and socioeconomic data, and historical data. At the local-level, I addressed the deforestation process over the last 60 years using six case studies of 100 km2 of humid lowland forests, by applying logistic regression and spatial analysis. At the regional-level, I studied the deforestation in the Caquetá colonization front of the Colombian Amazon region from 1988-2004 by applying a forest-cover zoning method and logistic regression models to predict deforestation and forest regeneration from biophysical and socio-economic explanatory variables. At the national level, I quantified and analysed patterns and drivers of land cover change over the past 500 years for key periods of Colombian history, and identified the extent and duration of impacts on broad ecosystem types. At the national and regional-levels, I also modelled current landscape transformation patterns and predicted areas at a high risk of future deforestation using a joint logistic regression and regression tree approach. I discovered that the rate of deforestation across several lowland regions of Colombia follows a simple sigmoid pattern composed of four phases of transformation: an initial phase of gradual forest loss; an intermediate phase of rapid loss; a second intermediate phase where the rate of decline slows; and a final phase where the forest loss stabilises and is balanced by forest regeneration creating a dynamic equilibrium. At the end of this final phase, the landscape is in a highly transformed state with forest cover stabilizing at 2 to 10% of the original extent and an average forest patch size of 15.4 (± 9.2) ha. As a general rule, the transformed landscape will have two forest components: a stable component of remnant mature tropical forests, and a dynamic component of secondary forests of different ages that is repeatedly cleared. A second important discovery was that unplanned deforestation in the Colombian Amazon moves as a colonisation wave, extending from population centres. The rate of movement was 0.84 km.yr-1 between 1989 and 2002. The regional average annual deforestation rate was 2.6%, but varied locally between –1.8% (regeneration) and 5.3%. The parallel deforestation and regeneration processes operating within the colonization front showed consistent patterns and rates directly related to the proportion of forest in the neighbourhood, with the highest rates of deforestation occurring in the areas with intermediate (40-60%) forest cover, following an overall quadratic function, and therefore confirming the sigmoid pattern across an entire colonization front. Landscapes with intermediate forest cover also have the highest density of edge habitat, with the deforestation process mimicking the spread of disease. At the national-level, the study reveals two important outcomes. First, there are significant regional differences in the spatial and temporal patterns and drivers of land cover change. The importance of such regional differences in factors explainingland cover change is highlighted by the greatest discrimination ability shown by a regional-level classification tree model. The coefficients and significance of variables in a regional logistic regression model confirmed these differences. Overall, factors related to accessibility (distance to roads and towns) had the strongest influence on the probability of deforestation. The second national-level outcome highlights the need for a longer-term historical perspective spanning centuries to understand present-day landscapes and their level of human impact. The historical analysis reveals that the main drivers of landscape change varied in the early colonial period, with cattle grazing becoming increasingly important, and finally the high impact of economic globalization in the 1990s. The historic land cover maps show the transformed areas increasing from approximately 15 million ha in 1500 to 42 million ha in 2000. Also, during this period, the transformed areas changed from predominantly cropping land uses in 1500 to predominantly (< 75%) grazing in 2000. The research outcomes collectively provide an understanding of the spatial dynamics of unplannedland cover change in tropical forest landscapes, by showing how deforestation and regeneration processes vary along a transformation gradient, and linking the amount of remnant forests is related to the rate of change. The work has implications for policy and management. For example, the method of calculating the movement speed of a colonization front provides a spatially explicit prediction of threat that can be used in conservation planning. The improved understanding of the deforestation and regeneration dynamics over an entire colonization fronts permits more accurate calculation of carbon budgeting for climate change applications. I recommend that future work test of the generalities presented here in both, countries with unplanned deforestation and countries with planned land clearing, to evaluate the effects government controlled planning has on the end result of human transformed landscapes in tropical and subtropical forest regions.

The broad aim of this study is, to contribute to a broader and more comprehensive understanding of the patterns, processes and drivers of unplanned land cover change in the tropics, using Colombia as a study case. Land cover change is an important global issue because of the expanding ecological footprint of a rapidly increasing human population and per capita level of resource consumption. This has a major impact on natural ecosystems and their function at the local (hundreds of square kilometers) and global scales. The understanding of extent and rate of land cover change is an important issue confronting biodiversity conservation, land use planning, protected area management, and global climate change analysis. Tropical deforestation is the major source of global land cover change, with the highest absolute rates occurring in South America, especially in the Brazilian Amazon, where government planning is an important driver of deforestation. However, unplanned deforestation for cropping and ranching is also occurring in Colombia. This is of international concern because Colombia’s diverse ecosystems support high levels of species richness and endemism. Improving the understanding of the spatial and temporal patterns and drivers ofland cover change (both deforestation and regeneration) is an important step in developing planning and conservation strategies to address this problem. I applied a spatial and temporal statistical modelling approach to predict changes in land cover in Colombia at the local (100 km2), regional (104 km2) and national (106 km2) levels, with a timeframe spanning from decades to centuries. As dependant variable data, binary forest/non-forest data are used. Explanatory variables comprise biophysical and socioeconomic data sourced from a broad range of information sources, including remotely sensed data from aerial photographs and satellite images, secondary sources of biophysical and socioeconomic data, and historical data. At the local-level, I addressed the deforestation process over the last 60 years using six case studies of 100 km2 of humid lowland forests, by applying logistic regression and spatial analysis. At the regional-level, I studied the deforestation in the Caquetá colonization front of the Colombian Amazon region from 1988-2004 by applying a forest-cover zoning method and logistic regression models to predict deforestation and forest regeneration from biophysical and socio-economic explanatory variables. At the national level, I quantified and analysed patterns and drivers of land cover change over the past 500 years for key periods of Colombian history, and identified the extent and duration of impacts on broad ecosystem types. At the national and regional-levels, I also modelled current landscape transformation patterns and predicted areas at a high risk of future deforestation using a joint logistic regression and regression tree approach. I discovered that the rate of deforestation across several lowland regions of Colombia follows a simple sigmoid pattern composed of four phases of transformation: an initial phase of gradual forest loss; an intermediate phase of rapid loss; a second intermediate phase where the rate of decline slows; and a final phase where the forest loss stabilises and is balanced by forest regeneration creating a dynamic equilibrium. At the end of this final phase, the landscape is in a highly transformed state with forest cover stabilizing at 2 to 10% of the original extent and an average forest patch size of 15.4 (± 9.2) ha. As a general rule, the transformed landscape will have two forest components: a stable component of remnant mature tropical forests, and a dynamic component of secondary forests of different ages that is repeatedly cleared. A second important discovery was that unplanned deforestation in the Colombian Amazon moves as a colonisation wave, extending from population centres. The rate of movement was 0.84 km.yr-1 between 1989 and 2002. The regional average annual deforestation rate was 2.6%, but varied locally between –1.8% (regeneration) and 5.3%. The parallel deforestation and regeneration processes operating within the colonization front showed consistent patterns and rates directly related to the proportion of forest in the neighbourhood, with the highest rates of deforestation occurring in the areas with intermediate (40-60%) forest cover, following an overall quadratic function, and therefore confirming the sigmoid pattern across an entire colonization front. Landscapes with intermediate forest cover also have the highest density of edge habitat, with the deforestation process mimicking the spread of disease. At the national-level, the study reveals two important outcomes. First, there are significant regional differences in the spatial and temporal patterns and drivers of land cover change. The importance of such regional differences in factors explainingland cover change is highlighted by the greatest discrimination ability shown by a regional-level classification tree model. The coefficients and significance of variables in a regional logistic regression model confirmed these differences. Overall, factors related to accessibility (distance to roads and towns) had the strongest influence on the probability of deforestation. The second national-level outcome highlights the need for a longer-term historical perspective spanning centuries to understand present-day landscapes and their level of human impact. The historical analysis reveals that the main drivers of landscape change varied in the early colonial period, with cattle grazing becoming increasingly important, and finally the high impact of economic globalization in the 1990s. The historic land cover maps show the transformed areas increasing from approximately 15 million ha in 1500 to 42 million ha in 2000. Also, during this period, the transformed areas changed from predominantly cropping land uses in 1500 to predominantly (< 75%) grazing in 2000. The research outcomes collectively provide an understanding of the spatial dynamics of unplannedland cover change in tropical forest landscapes, by showing how deforestation and regeneration processes vary along a transformation gradient, and linking the amount of remnant forests is related to the rate of change. The work has implications for policy and management. For example, the method of calculating the movement speed of a colonization front provides a spatially explicit prediction of threat that can be used in conservation planning. The improved understanding of the deforestation and regeneration dynamics over an entire colonization fronts permits more accurate calculation of carbon budgeting for climate change applications. I recommend that future work test of the generalities presented here in both, countries with unplanned deforestation and countries with planned land clearing, to evaluate the effects government controlled planning has on the end result of human transformed landscapes in tropical and subtropical forest regions.

The broad aim of this study is, to contribute to a broader and more comprehensive understanding of the patterns, processes and drivers of unplanned land cover change in the tropics, using Colombia as a study case. Land cover change is an important global issue because of the expanding ecological footprint of a rapidly increasing human population and per capita level of resource consumption. This has a major impact on natural ecosystems and their function at the local (hundreds of square kilometers) and global scales. The understanding of extent and rate of land cover change is an important issue confronting biodiversity conservation, land use planning, protected area management, and global climate change analysis. Tropical deforestation is the major source of global land cover change, with the highest absolute rates occurring in South America, especially in the Brazilian Amazon, where government planning is an important driver of deforestation. However, unplanned deforestation for cropping and ranching is also occurring in Colombia. This is of international concern because Colombia’s diverse ecosystems support high levels of species richness and endemism. Improving the understanding of the spatial and temporal patterns and drivers ofland cover change (both deforestation and regeneration) is an important step in developing planning and conservation strategies to address this problem. I applied a spatial and temporal statistical modelling approach to predict changes in land cover in Colombia at the local (100 km2), regional (104 km2) and national (106 km2) levels, with a timeframe spanning from decades to centuries. As dependant variable data, binary forest/non-forest data are used. Explanatory variables comprise biophysical and socioeconomic data sourced from a broad range of information sources, including remotely sensed data from aerial photographs and satellite images, secondary sources of biophysical and socioeconomic data, and historical data. At the local-level, I addressed the deforestation process over the last 60 years using six case studies of 100 km2 of humid lowland forests, by applying logistic regression and spatial analysis. At the regional-level, I studied the deforestation in the Caquetá colonization front of the Colombian Amazon region from 1988-2004 by applying a forest-cover zoning method and logistic regression models to predict deforestation and forest regeneration from biophysical and socio-economic explanatory variables. At the national level, I quantified and analysed patterns and drivers of land cover change over the past 500 years for key periods of Colombian history, and identified the extent and duration of impacts on broad ecosystem types. At the national and regional-levels, I also modelled current landscape transformation patterns and predicted areas at a high risk of future deforestation using a joint logistic regression and regression tree approach. I discovered that the rate of deforestation across several lowland regions of Colombia follows a simple sigmoid pattern composed of four phases of transformation: an initial phase of gradual forest loss; an intermediate phase of rapid loss; a second intermediate phase where the rate of decline slows; and a final phase where the forest loss stabilises and is balanced by forest regeneration creating a dynamic equilibrium. At the end of this final phase, the landscape is in a highly transformed state with forest cover stabilizing at 2 to 10% of the original extent and an average forest patch size of 15.4 (± 9.2) ha. As a general rule, the transformed landscape will have two forest components: a stable component of remnant mature tropical forests, and a dynamic component of secondary forests of different ages that is repeatedly cleared. A second important discovery was that unplanned deforestation in the Colombian Amazon moves as a colonisation wave, extending from population centres. The rate of movement was 0.84 km.yr-1 between 1989 and 2002. The regional average annual deforestation rate was 2.6%, but varied locally between –1.8% (regeneration) and 5.3%. The parallel deforestation and regeneration processes operating within the colonization front showed consistent patterns and rates directly related to the proportion of forest in the neighbourhood, with the highest rates of deforestation occurring in the areas with intermediate (40-60%) forest cover, following an overall quadratic function, and therefore confirming the sigmoid pattern across an entire colonization front. Landscapes with intermediate forest cover also have the highest density of edge habitat, with the deforestation process mimicking the spread of disease. At the national-level, the study reveals two important outcomes. First, there are significant regional differences in the spatial and temporal patterns and drivers of land cover change. The importance of such regional differences in factors explainingland cover change is highlighted by the greatest discrimination ability shown by a regional-level classification tree model. The coefficients and significance of variables in a regional logistic regression model confirmed these differences. Overall, factors related to accessibility (distance to roads and towns) had the strongest influence on the probability of deforestation. The second national-level outcome highlights the need for a longer-term historical perspective spanning centuries to understand present-day landscapes and their level of human impact. The historical analysis reveals that the main drivers of landscape change varied in the early colonial period, with cattle grazing becoming increasingly important, and finally the high impact of economic globalization in the 1990s. The historic land cover maps show the transformed areas increasing from approximately 15 million ha in 1500 to 42 million ha in 2000. Also, during this period, the transformed areas changed from predominantly cropping land uses in 1500 to predominantly (< 75%) grazing in 2000. The research outcomes collectively provide an understanding of the spatial dynamics of unplannedland cover change in tropical forest landscapes, by showing how deforestation and regeneration processes vary along a transformation gradient, and linking the amount of remnant forests is related to the rate of change. The work has implications for policy and management. For example, the method of calculating the movement speed of a colonization front provides a spatially explicit prediction of threat that can be used in conservation planning. The improved understanding of the deforestation and regeneration dynamics over an entire colonization fronts permits more accurate calculation of carbon budgeting for climate change applications. I recommend that future work test of the generalities presented here in both, countries with unplanned deforestation and countries with planned land clearing, to evaluate the effects government controlled planning has on the end result of human transformed landscapes in tropical and subtropical forest regions.

The broad aim of this study is, to contribute to a broader and more comprehensive understanding of the patterns, processes and drivers of unplanned land cover change in the tropics, using Colombia as a study case. Land cover change is an important global issue because of the expanding ecological footprint of a rapidly increasing human population and per capita level of resource consumption. This has a major impact on natural ecosystems and their function at the local (hundreds of square kilometers) and global scales. The understanding of extent and rate of land cover change is an important issue confronting biodiversity conservation, land use planning, protected area management, and global climate change analysis. Tropical deforestation is the major source of global land cover change, with the highest absolute rates occurring in South America, especially in the Brazilian Amazon, where government planning is an important driver of deforestation. However, unplanned deforestation for cropping and ranching is also occurring in Colombia. This is of international concern because Colombia’s diverse ecosystems support high levels of species richness and endemism. Improving the understanding of the spatial and temporal patterns and drivers ofland cover change (both deforestation and regeneration) is an important step in developing planning and conservation strategies to address this problem. I applied a spatial and temporal statistical modelling approach to predict changes in land cover in Colombia at the local (100 km2), regional (104 km2) and national (106 km2) levels, with a timeframe spanning from decades to centuries. As dependant variable data, binary forest/non-forest data are used. Explanatory variables comprise biophysical and socioeconomic data sourced from a broad range of information sources, including remotely sensed data from aerial photographs and satellite images, secondary sources of biophysical and socioeconomic data, and historical data. At the local-level, I addressed the deforestation process over the last 60 years using six case studies of 100 km2 of humid lowland forests, by applying logistic regression and spatial analysis. At the regional-level, I studied the deforestation in the Caquetá colonization front of the Colombian Amazon region from 1988-2004 by applying a forest-cover zoning method and logistic regression models to predict deforestation and forest regeneration from biophysical and socio-economic explanatory variables. At the national level, I quantified and analysed patterns and drivers of land cover change over the past 500 years for key periods of Colombian history, and identified the extent and duration of impacts on broad ecosystem types. At the national and regional-levels, I also modelled current landscape transformation patterns and predicted areas at a high risk of future deforestation using a joint logistic regression and regression tree approach. I discovered that the rate of deforestation across several lowland regions of Colombia follows a simple sigmoid pattern composed of four phases of transformation: an initial phase of gradual forest loss; an intermediate phase of rapid loss; a second intermediate phase where the rate of decline slows; and a final phase where the forest loss stabilises and is balanced by forest regeneration creating a dynamic equilibrium. At the end of this final phase, the landscape is in a highly transformed state with forest cover stabilizing at 2 to 10% of the original extent and an average forest patch size of 15.4 (± 9.2) ha. As a general rule, the transformed landscape will have two forest components: a stable component of remnant mature tropical forests, and a dynamic component of secondary forests of different ages that is repeatedly cleared. A second important discovery was that unplanned deforestation in the Colombian Amazon moves as a colonisation wave, extending from population centres. The rate of movement was 0.84 km.yr-1 between 1989 and 2002. The regional average annual deforestation rate was 2.6%, but varied locally between –1.8% (regeneration) and 5.3%. The parallel deforestation and regeneration processes operating within the colonization front showed consistent patterns and rates directly related to the proportion of forest in the neighbourhood, with the highest rates of deforestation occurring in the areas with intermediate (40-60%) forest cover, following an overall quadratic function, and therefore confirming the sigmoid pattern across an entire colonization front. Landscapes with intermediate forest cover also have the highest density of edge habitat, with the deforestation process mimicking the spread of disease. At the national-level, the study reveals two important outcomes. First, there are significant regional differences in the spatial and temporal patterns and drivers of land cover change. The importance of such regional differences in factors explainingland cover change is highlighted by the greatest discrimination ability shown by a regional-level classification tree model. The coefficients and significance of variables in a regional logistic regression model confirmed these differences. Overall, factors related to accessibility (distance to roads and towns) had the strongest influence on the probability of deforestation. The second national-level outcome highlights the need for a longer-term historical perspective spanning centuries to understand present-day landscapes and their level of human impact. The historical analysis reveals that the main drivers of landscape change varied in the early colonial period, with cattle grazing becoming increasingly important, and finally the high impact of economic globalization in the 1990s. The historic land cover maps show the transformed areas increasing from approximately 15 million ha in 1500 to 42 million ha in 2000. Also, during this period, the transformed areas changed from predominantly cropping land uses in 1500 to predominantly (< 75%) grazing in 2000. The research outcomes collectively provide an understanding of the spatial dynamics of unplannedland cover change in tropical forest landscapes, by showing how deforestation and regeneration processes vary along a transformation gradient, and linking the amount of remnant forests is related to the rate of change. The work has implications for policy and management. For example, the method of calculating the movement speed of a colonization front provides a spatially explicit prediction of threat that can be used in conservation planning. The improved understanding of the deforestation and regeneration dynamics over an entire colonization fronts permits more accurate calculation of carbon budgeting for climate change applications. I recommend that future work test of the generalities presented here in both, countries with unplanned deforestation and countries with planned land clearing, to evaluate the effects government controlled planning has on the end result of human transformed landscapes in tropical and subtropical forest regions.

The broad aim of this study is, to contribute to a broader and more comprehensive understanding of the patterns, processes and drivers of unplanned land cover change in the tropics, using Colombia as a study case. Land cover change is an important global issue because of the expanding ecological footprint of a rapidly increasing human population and per capita level of resource consumption. This has a major impact on natural ecosystems and their function at the local (hundreds of square kilometers) and global scales. The understanding of extent and rate of land cover change is an important issue confronting biodiversity conservation, land use planning, protected area management, and global climate change analysis. Tropical deforestation is the major source of global land cover change, with the highest absolute rates occurring in South America, especially in the Brazilian Amazon, where government planning is an important driver of deforestation. However, unplanned deforestation for cropping and ranching is also occurring in Colombia. This is of international concern because Colombia’s diverse ecosystems support high levels of species richness and endemism. Improving the understanding of the spatial and temporal patterns and drivers ofland cover change (both deforestation and regeneration) is an important step in developing planning and conservation strategies to address this problem. I applied a spatial and temporal statistical modelling approach to predict changes in land cover in Colombia at the local (100 km2), regional (104 km2) and national (106 km2) levels, with a timeframe spanning from decades to centuries. As dependant variable data, binary forest/non-forest data are used. Explanatory variables comprise biophysical and socioeconomic data sourced from a broad range of information sources, including remotely sensed data from aerial photographs and satellite images, secondary sources of biophysical and socioeconomic data, and historical data. At the local-level, I addressed the deforestation process over the last 60 years using six case studies of 100 km2 of humid lowland forests, by applying logistic regression and spatial analysis. At the regional-level, I studied the deforestation in the Caquetá colonization front of the Colombian Amazon region from 1988-2004 by applying a forest-cover zoning method and logistic regression models to predict deforestation and forest regeneration from biophysical and socio-economic explanatory variables. At the national level, I quantified and analysed patterns and drivers of land cover change over the past 500 years for key periods of Colombian history, and identified the extent and duration of impacts on broad ecosystem types. At the national and regional-levels, I also modelled current landscape transformation patterns and predicted areas at a high risk of future deforestation using a joint logistic regression and regression tree approach. I discovered that the rate of deforestation across several lowland regions of Colombia follows a simple sigmoid pattern composed of four phases of transformation: an initial phase of gradual forest loss; an intermediate phase of rapid loss; a second intermediate phase where the rate of decline slows; and a final phase where the forest loss stabilises and is balanced by forest regeneration creating a dynamic equilibrium. At the end of this final phase, the landscape is in a highly transformed state with forest cover stabilizing at 2 to 10% of the original extent and an average forest patch size of 15.4 (± 9.2) ha. As a general rule, the transformed landscape will have two forest components: a stable component of remnant mature tropical forests, and a dynamic component of secondary forests of different ages that is repeatedly cleared. A second important discovery was that unplanned deforestation in the Colombian Amazon moves as a colonisation wave, extending from population centres. The rate of movement was 0.84 km.yr-1 between 1989 and 2002. The regional average annual deforestation rate was 2.6%, but varied locally between –1.8% (regeneration) and 5.3%. The parallel deforestation and regeneration processes operating within the colonization front showed consistent patterns and rates directly related to the proportion of forest in the neighbourhood, with the highest rates of deforestation occurring in the areas with intermediate (40-60%) forest cover, following an overall quadratic function, and therefore confirming the sigmoid pattern across an entire colonization front. Landscapes with intermediate forest cover also have the highest density of edge habitat, with the deforestation process mimicking the spread of disease. At the national-level, the study reveals two important outcomes. First, there are significant regional differences in the spatial and temporal patterns and drivers of land cover change. The importance of such regional differences in factors explainingland cover change is highlighted by the greatest discrimination ability shown by a regional-level classification tree model. The coefficients and significance of variables in a regional logistic regression model confirmed these differences. Overall, factors related to accessibility (distance to roads and towns) had the strongest influence on the probability of deforestation. The second national-level outcome highlights the need for a longer-term historical perspective spanning centuries to understand present-day landscapes and their level of human impact. The historical analysis reveals that the main drivers of landscape change varied in the early colonial period, with cattle grazing becoming increasingly important, and finally the high impact of economic globalization in the 1990s. The historic land cover maps show the transformed areas increasing from approximately 15 million ha in 1500 to 42 million ha in 2000. Also, during this period, the transformed areas changed from predominantly cropping land uses in 1500 to predominantly (< 75%) grazing in 2000. The research outcomes collectively provide an understanding of the spatial dynamics of unplannedland cover change in tropical forest landscapes, by showing how deforestation and regeneration processes vary along a transformation gradient, and linking the amount of remnant forests is related to the rate of change. The work has implications for policy and management. For example, the method of calculating the movement speed of a colonization front provides a spatially explicit prediction of threat that can be used in conservation planning. The improved understanding of the deforestation and regeneration dynamics over an entire colonization fronts permits more accurate calculation of carbon budgeting for climate change applications. I recommend that future work test of the generalities presented here in both, countries with unplanned deforestation and countries with planned land clearing, to evaluate the effects government controlled planning has on the end result of human transformed landscapes in tropical and subtropical forest regions.

In order to understand the ecology of species, it is important to know how animals use their environment. This information can be determined at a range of spatial and temporal scales, and results may vary accordingly. The habitats that animals use determine resources available to them for different purposes (e.g. feeding and resting), and risks of predation to which they are exposed. Consequently, patterns of behaviour in relation to the environment are likely to influence survival and fitness. In Tasmania, Australia, three common and widely distributed native marsupial herbivores are the red-necked or Bennett's wallaby (Macropus rufogriseus rufogriseus), the red-bellied pademelon (Thylogale billardierii) and the common brushtail possum (Trichosurus vulpecula). Information on the behaviour of these species in relation to their environment is largely unavailable. This thesis describes the abundance, spatial organisation and habitat selection patterns of these three species, within a patchy forestry landscape. The five major habitat types within this environment were: (1) a prepared site that was planted with commercial Eucalyptus nitens seedlings during the study (referred to as 'young plantation'); (2) 5-7 year old E. nitens plantation; (3) grassland; (4) native forest; and (5) harvested uncleared land. Patterns of habitat use and selection were examined at three sequential spatio-temporal scales, within a hierarchy of decisions. These were: (1) location of home-range within the landscape, (2) feeding area within the home-range, and (3) vegetation consumed within one habitat, the young plantation. A radio-telemetry study of Bennett's wallabies, pademelons and possums was used to examine Scales 1 and 2 at the individual animal level. Animal surveys were carried out to examine Scale 2 for the entire herbivore community at the population level. These data were also used to estimate herbivore densities for the overall area and individual habitats. Fenced and unfenced vegetation plots, located within the young plantation, a highly used habitat, were monitored over time to examine Scale 3. As part of this research, modifications to common line-transect sampling methods were made. These enabled methods that are usually applied to daytime surveys in open habitat, to be used in nocturnal surveys in densely vegetated habitats. Accuracy testing of the radiotelemetry system is also described, as the patchiness of the landscape required careful interpretation of results. Results showed that, at night, wallabies and pademelons used all habitats, but consistently selected for open habitats (young plantation and grassland) across spatio-temporal scales. The use of these open habitats for feeding was confirmed by the large biomass of grass and forbs consumed by herbivores in a detailed study of vegetation on the young plantation. These patterns are consistent with their feeding strategies of grazer or mixed-feeder. During the day, the two macropod species avoided open habitats and showed strong selection for closed habitats. Wallabies selected for older plantation, while pademelons selected for native forest. This difference reflects their respective predator avoidance strategy (crypsis for pademelons) or escape response (flight for wallabies). Although shelter habitat was important to the two macropod species, their lack of selection at the home-range scale was suggested to reflect the fact that resting animals require little space. Patterns of habitat use and selection were difficult to interpret for possums, because results varied between the spatio-temporal scales. Spotlighting data showed that at night, possums selected for native forest, young plantation and particularly grassland at the population level. Radio-collared animals selected only for native forest. Older E. nitens plantations were avoided by possums at every level, and appeared to represent a biological desert to this species. High overall densities of wallabies and pademelons (0.3 and 1.5 animals.ha-1, respectively), and small, round, home-ranges (61.6 ha and 22.3 ha, respectively) suggested that these species benefited from the patchiness of this environment. This is attributed to the highly heterogeneous habitats, providing complimentary resources in the absence of ecotones or transitional flora zones, existing side by side, over a small spatial scale. In contrast, extremely low possum population density (0.04 animals.ha-1) and very large home-ranges (39.1 ha) suggested that resources, presumably den sites and/or food, were limited within this forestry environment. Results on the ecological aspects of the three herbivore species, described above, are put in the context of the Tasmanian forestry industry, particularly in relation to management of herbivore browsing damage to planted seedlings. Based on this work, I suggest that future management strategies could involve: (1) reducing fragmentation of the natural environment, which supports small home-ranges and high macropod densities, by designing larger, rounder plantations; (2) considering the placement of plantations in relation to the proximity of open (feeding grounds) and closed (shelter) habitats; (3) reducing or removing windrows from newly established plantations to restrict pademelons to the plantation edge; (4) deliberately retaining groundcover or using cover crops to provision herbivores with an alternative food source, as grasses and herbaceous dicots are eaten in preference to Eucalyptus nirens seedlings; (5) recognising that wallabies and pademelons remove a large biomass of groundcover and therefore, could play a positive role in weed control, reducing the need to herbicide plantations; (6) monitoring newly planted plantations at short and regular time intervals so that damage caused by insects versus mammals can be differentiated; and (7) avoiding planting in winter when macropods may have little alternative food to eat on newly established plantations.

No description available.

No description available.

Fragmentation of forest habitat by urban and rural development has had profound effects on the distribution and abundance of many native species; however, little is known about the ecological processes driving patterns in community structure (species richness and composition) of host-specialised herbivores in modified habitats. I examined patterns in community structure of 9 specialist and 19 generalist invertebrate herbivores of cabbage trees (Cordyline australis Laxmanniaceae) across a highly-modified landscape. I found that, although species richness of specialists was highest in forest sites, the majority of host-specialised species were not restricted to forest habitats and were as widespread as many generalists. In terms of site occupancy, only two specialist and four generalist species were rare. I show that patterns of species occupancy and abundance reflect differing susceptibility to habitat modification, with landscape-level variation an important predictor of abundance for nearly all species. When species occurrences and life history traits were considered I did not find strong evidence for the importance of dispersal ability, which suggests that habitat variability had a stronger organising effect on the community. In a replicated common garden experiment, I found distinct regional patterns in the community structure of the specialist invertebrates occurring on different phylogenetic groups of C. australis. In contrast, community structure of generalist herbivores did not differ significantly among host genotypes. I speculate these patterns are due to historical changes in the distribution of cabbage trees in the Southern phylogenetic region that caused specialised herbivores to become locally adapted on populations of low genetic diversity following expansion after the last glacial maximum. However, this consistent selection pressure did not occur in the Northern region where C. australis habitat has been more consistently available over the past tens of thousands of years, reflected in higher host genetic diversity. This study has advanced our understanding of the patterns in community structure of an indigenous, host-specialised fauna in a highly modified and fragmented urban and rural landscapes.