Abstract
In mountain ecosystems, forest damages are of high importance and a challenge for forest management and conservation. However, a quick assessment of these changes is difficult in complex terrain. In response to this constraint, rapid mapping using earth observation data has become an attractive solution. In particular, Synthetic Aperture Radar (SAR) offers a unique opportunity for forest management applications due to its independence of weather conditions. In this study, the potential of X-band COSMO-SkyMed Stripmap SAR images for forest change detection in mountainous environments is evaluated. A forested area in the Southern Alps (South Tyrol) is chosen which was affected by a storm event in June 2011. We mapped the forest changes using a statistical approach (Principal Component Analysis (PCA)). To evaluate the potential of radar relatively to optical data, two RapidEye images were obtained and forest changes by thresholding image differences between both years were mapped. The results show that optical data showed higher forest change accuracies than the SAR images. Using RapidEye imagery we accurately captured changes with a minimum area of 0.1 ha while the minimum detectable area was 0.5 ha with COSMO-SkyMed. The difference in results is explained by the high level of speckle noise introduced by the SAR resolution and by the high Local Incidence Angles (LIAs) within the study area (55° to 95°). The combination of these two effects introduced an increase in false alarm changes. The advantage of using SAR data is the capability of a quick assessment of changes especially in difficult weather conditions. However, the use of radar for rapid mapping of forest damages in mountains can be limited depending on the acquisition geometry. Thus, the acquisition geometry should be carefully chosen with respect to the topography to reduce layover and shadowing effects.