OPERATIONAL LAKE MAPPING ON THE SOUTHERN TIBET USING SENTINEL-1 DATA

The Tibet plateau (TP), as the third pole, plays an important role for indicating the global climate change. This study evaluated the potential of SAR data on lake mapping on the southern TP region which is characterized by various complex water conditions. A robust object-oriented lake extraction approach using Sentinel-1 data is proposed. First, an automatic method of selecting regions that obeying bimodal distribution is used to obtain the optimal threshold to identify lakes. Second, Taguchi design experiments are used to obtain the optimal segmentation parameters of multiresolution segmentation and object-oriented classification is applied to extract the lakes. Finally, a postprocessing approach is applied to remove the false alarms due to shadows with auxiliary DEM data and object-based variance of intensity map. Experiments on the Sentinel-1 image demonstrate that the proposed method produces highly reliable lake extents in southern TP, and is potentially applicable to the lake mapping projects on the whole TP.