Satellite remote sensing classification of thaw lakes and drained thaw lake basins on the North Slope of Alaska

The purpose of this research was to map thaw lakes and drained thaw lake basins (DTLBs) on the North Slope of Alaska using satellite remote sensing. This research is the first to map DTLBs on a large scale for Alaska's Arctic Coastal Plain. Thaw lakes and DTLBs were classified from seven Landsat-7 scenes using texture analysis, spectral transformations, and image segmentation. The overall classification accuracy was 97.7% with, a Kappa coefficient of 0.96. Thaw lakes had a producer accuracy of 99.1% and a user accuracy of 98.6%, while DTLBs had a producer accuracy of 93.8% and user accuracy of 98.1%. A total of 7054 km(2) of thaw lakes were mapped, accounting for 20.4% of the Arctic Coastal Plain; there were 8917 km(2) of DTLBs covering 25.7% of the study area. Continued research in the analysis of thaw lakes and DTLBs is crucial to our understanding of the global carbon cycle, atmospheric methane concentrations, heat flow and climate change. (C) 2005 Elsevier Inc. All rights reserved.