Identifying the spatial patterns and drivers of deforestation is a critical task in geographic research. In addition to mapping deforestation, it’s important to determine the statistical effects of the spatial configuration of tropical landscapes on current deforestation trends. In order to model drivers of deforestation is important to take into account the spatial structure of data (i.e., whether observed deforestation is spatially clustered or not). We calculated deforestation rates at the village level in the Rio Grande basin, Colombia, using land cover information derived from Landsat TM/ETM satellite imagery (1986-2012). We used econometric models to understand the deforestation patterns using a set of socioeconomic, biophysical, and accessibility variables. Exploratory Spatial Data Analysis shows the existence of globally and locally positive spatial autocorrelation. The Spatial Lag Model, which considers spatial data autocorrelation, explained most of the variability in deforestation patterns. The main drivers of deforestation for the region over a 26-year period were annual average temperature, population density, road density, and distance to main rivers. Results show that observed deforestation is closely related to dairy farming; this is due to the long history of human intervention in the watershed. We found that forest recovery in recent years occurred, however forest loss continues to be the dominant land transition of Andean landscapes, with positive spatial interdependencies. The identification of drivers of deforestation using methods that account for spatial autocorrelation can inform national conservation policy and programs aimed at the provision of ecosystem services.
Sustainable Development Goals