Economic modeling of tropical deforestation in Antioquia (Colombia), 1980-2000 : an analysis at a semi-fine scale with spatially explicit data
obtained largely from aerial photography was employed to calculate deforestation. In addition, free international databases and national georeferenced biophysical information
are also used to consolidate a set of potential explanatory variables. Econometric models estimated the clearing probability as a function of the variables: distance from roads,
distance from rivers, slope, Gini’s index (a measure of land ownership concentration), soil fertility, protected area, and population density. The robustness of the statistical estimation to the pixel size, the analysis unit used in this dissertation, suggests that a 30 m pixel size model performs better from a statistical viewpoint than models with pixel size from 50 to 300 m. The 30 m pixel size model has the highest values of R2 and area under the ROC curve, an index used to measure discrimination
accuracy in the econometric models. Elasticity magnitudes of all the explanatory variables included in the 30 m pixel size model indicate that the Gini’s index, the slope, and the distance from roads are the main drivers of the observed deforestation in Antioquia. The 30 m pixel size model was used in policy analyses in two different ways:(i)to simulate how new roads may influence deforestation,(ii)to identify the areas under greater threat of deforestation. For the building of new roads a simulated scenario examines the incremental effects on deforestation with respect to a base scenario, after reducing the distance from roads by a fixed percentage of 50%. Simulation results suggest that the building of new roads will induce more deforestation (~1039 hectares) compared to the base scenario. The results of this dissertation were also employed to identify forest areas under greater threat of deforestation. The results indicate that the forests located in the Northwest of Antioquia are under greater threat of conversion most likely because of their accessibility, and that forests located in remote sites may be protected without significant institutional efforts.
Advisor:Adams, Darius M.; Montgomery, Claire A.; Sessions, John; Plantinga, Andrew J.; Adams, Paul W.
School:Oregon State University
School Location:USA - Oregon
Source Type:Master's Thesis
Keywords:tropical deforestation land cover change spatial analysis colombia use conversion economics of columbia antioquia econometric models
Date of Publication:05/12/2009