The added value of integrated correction models for the detection of forest transitions in mountain areas

Mapping of vegetation in mountain areas on the basis of remote sensing is considerably hampered by atmospheric distortions and differential illumination effects due to topography. A variety of atmospheric and topographic normalization techniques have been proposed to minimize the atmospheric and topographic effects and to detect the often subtle land cover changes in sparsely populated mountain areas. The purpose of this study was to evaluate the effect of these normalizations on the land cover classification accuracy. Therefore, 14 combinations of atmospheric and topographic correction models were applied in a case study in the Romanian Carpathian mountains. The impact of the IRC models on the accuracy of land cover classifications was evaluated visually and statistically. Statistical analysis showed that all 14 preprocessing models resulted in a better overall classification accuracy. The separability of the land cover classes coniferous and mixed forest was significantly enhanced after correction. The preprocessing models are the most efficient on weakly illuminated slope segments. In general, the topographic component of the correction models contributed more to the improvement of classification accuracy than the atmospheric component.status: publishe