Several TerraSAR-X satellite images acquired in high resolution spotlight mode with different polarisations for two study sites in southern Germany were used to distinguish forest from other land cover classes (agriculture, built-up, water bodies) using logistic regression models. In general, we observed that the mean and particularly the standard deviation of the backscatter were viable measures to discriminate land cover classes. Both measures were lowest for water bodies and highest for built-up areas, with agricultural areas and forest in intermediate positions. Trees outside forest were not differentiable from forest with the applied method. The HHpolarised image was better suited for a classification of built-up areas, whereas the VV-polarised image was more appropriate when classifying agricultural areas. Consequently, the combination of the two differently polarised images leads to a significantly better model. Since forests in one study area were generally found on steeper slopes in comparison to other land cover classes, the inclusion of terrain slope further improved the classification, which resulted in an overall accuracy of 92-95%. Systematic differences in the parameter values of the explanatory variables for one class between the study areas may be caused by differing class probabilities. Thus, transferring the model of one study area to the image of another resulted in a 7-9% loss of accuracy.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.