Search

Species distribution models were developed for three high economic value timber trees (Calophyllum brasiliensis, Carapa guianensis and Virola surinamensis) that are heavily harvested in the Amazon Basin. A combination of habitat measurements extracted from remote sensing data (MODIS, QSCAT and SRTM) and bioclimatic surfaces was examined to ascertain the most influential factors determining the occurrence of these tree species. The prediction of species’ occurrence rates was tested separately for each species distribution model and the results were examined for their ability to accurately map the spatial distribution of these tree species. By evaluating the omission and commission rates we concluded that species distribution models based on remote sensing data contributed significantly in quantifying environmental properties used to summarize the ecological niche of each tree species. Specific vegetation characteristics (such as percentage of tree cover, vegetation moisture and roughness, annual NDVI and mean LAI during the dry LAI) showed the dependence of these species’ occurrence in more densely vegetated forests. Areas with high leaf area (even during the dry months) and areas with high vegetation moisture were predicted as potential species habitat for C. brasiliensis. The density vegetation during the dry season and vegetation phenology were strongly correlated with climate differences, such as variations in air temperature and precipitation seasonality for V. surinamensis. Lower elevation areas with more exuberant vegetation and a high greenness index were among the most important factors accounting for the geographical distribution of C. guianensis. Species distribution models are increasingly important in many fields of research and conservation. The potential of remotely sensed data to monitor environmental changes in tropical areas, along with the understanding of ecosystem function, are both critical for conservation of biodiversity and the long-term process of sustaining ecosystems. [Copyright &y& Elsevier]