PROBING TROPICAL DEFORESTATION - THE USE OF GIS AND STATISTICAL-ANALYSIS OF GEOREFERENCED DATA

The significance and magnitude of the relationship between forest cove r and ten environmental variables were investigated in a case study of Mindoro, Philippines. Landsat Thematic Mapper data and thematic maps were processed for Pearson's chi-square test, Cramer's V calculations and logistic regression analysis. Attribute query and display of compo site layers in GIS were further implemented to analyse the deforested lands. These statistical analyses show that some variables are signifi cantly related to forest cover, although the magnitude of relationship for all variables is weak. In contrast with most studies, the accessi bility and topographic factors in this study, among others, are not si gnificantly associated with forest cover. Associations between forest cover and some topographic and edaphic factors support the hypothesis that the area's many deforested lands are actually 'natural grasslands '. However, external evidence suggests that burning forests for grazin g purposes at least maintains and probably expands the area's deforest ation, Because of the high spatial, temporal and attribute variability of deforestation, extrapolation and generalization from the results s hould be made with caution, especially if they will be used as inputs for planning and management. (C) 1998 Elsevier Science Ltd. All rights reserved.