Abstract
While high resolution spatial variables contribute to a good fit of spatially-explicit deforestation models, socio-economic processes are often beyond the scope of these models. Such a low level of interest in the socio-economic dimension of deforestation limits the relevancy of these models for decision making and may be the cause of their failure to accurately predict observed deforestation trends in the medium term. This study aims to propose a flexible methodology for taking into account multiple drivers of deforestation in tropical forested areas, where the intensity of deforestation is explicitly predicted based on socio-economic variables. By coupling a model of deforestation location based on spatial environmental variables with several sub-models of deforestation intensity based on socio-economic variables, we were able to create a map of predicted deforestation over the period 2001–2014 in French Guiana. This map was compared to a reference map for accuracy assessment, not only at the pixel scale but also over cells ranging from 1 to approx. 600 sq. km. Highly significant relationships were explicitly established between deforestation intensity and several socio-economic variables: population growth, the amount of agricultural subsidies, gold and wood production. Such a precise characterisation of socio-economic processes allows to avoid overestimation biases in high deforestation areas, suggesting a better integration of socio-economic processes in the models. While considering deforestation as a purely geographical process contributes to the creation of conservative models unable to effectively assess changes in the socio-economic and political contexts influencing deforestation trends, this explicit characterisation of the socio-economic dimension of deforestation is critical for the creation of deforestation scenarios in REDD+ projects