Comparison of a Landsat-based logistic regression model and socioeconomic simulation model of deforestation in the Amazon

Modeling deforestation has become increasingly important in analyzing the effects of land-cover change on ecosystems. A simulation model based on socioeconomic variables was developed and applied to a study area in Central Rondonia, Brazil over a 20 year time period. The simulation model integrates three submodels of settlement diffusion, land-use change, and ecological change and simulates the effects of colonization on deforestation by varying a number of parameters (e.g. immigration rates, crop productivity, soil quality). We have previously developed a Landsat-based logistic regression model of deforestation for the same study area in Central Rondonia, Brazil over the 20 year time period. The model incorporates five main parameters in predicting the amount of deforestation on individual farmer settlement lots. These parameters include proximity of a lot to the nearest paved road (PPR); proximity of a lot to the nearest secondary road (PSR); distance to the nearest forest/cleared edge (PFC); distance from a market center (DMC); distance from a major urban center (DUC). This study evaluates and compares the predictive capability of both the image based and integrated model using remote sensing analysis and field measurements. The percent cleared and spatial metrics of landscape pattern including the patch-per-unit area metric (PPU) and square pixel metric (SqP) are used to compare image classifications with model outputs. The logistic regression model performed better than the socioeconomic model in capturing the pattern and location of deforestation over time. Differences in clearing patterns in both models were attributed to local topography, farming obstacles, and stochastic elements of farming. Results of this study help demonstrate some of the advantages in the use of integrated socioeconomic/ecological models for predicting land-use change and will hopefully facilitate the development of integrated models that accurately simulate the processes of land clearing on a regional and global scale