Merging Landsat Image Information with Georeferenced Biophysical and Socio-economical Datasets to Describe Forest Cover Change in a Philippine Province

This paper describes a combined remote sensing-GIS-logistic regression approach of merging extracted information from Landsat images with georeferenced biophysical and socio-economic datasets in the detection and analysis of the driving forces of forest cover change in Agusan del Norte (ADN) in Mindanao Island, Philippines, a province where forest resource use have been historically extensive. Year 1976 Landsat 2 MSS and year 2001 Landsat ETM+ images were independently classified using Support Vector Machines (SVM) to produce land cover maps with overall classification accuracies of 95% and 98%, respectively. Changes in forest cover and other types of land-cover change in the 25-year period were then detected from these maps through post-classification comparison in a GIS. To investigate what has driven these conversions, the associations between these changes and a selection of biophysical and socio-economical variables were explored through logistic regression analysis. The results show that while both the biophysical and socio-economical variables were significantly associated with the occurrences of forest cover change, the models containing only the socio-economical variables predict better the occurrences of change than those containing only the biophysical variables. This implies that most of the forest cover change detected in the year 2001 in ADN is much more a socio-economical matter, and is less forced by biophysical limitations. With these results, this study demonstrated the usefulness of RS, GIS and statistical analysis as exploratory tools in understanding the underlying processes and identification of driving forces of forest cover change, especially in areas of extensive forest resource use.