Mapping permafrost thaw stages in interior Alaska

Abstract

Permafrost degradation has been recognized for decades due to climate warming, wildfire, and infrastructure development. However, a large-scale characterization of permafrost thaw status has not been attempted before due to difficulties in ground data collection, inherent complications and heterogeneity of thaw in ecosystem-protected permafrost, and constraints of remote sensor observations and process-based modeling techniques. Here we made a first effort to map the status of permafrost thaw across a large ice-rich lowland fire-influenced landscape (2500 km²) in interior Alaska by developing a new protocol and combining decades of field measurements, repeat airborne lidar, spaceborne WorldView-2, Sentinel-2, Landsat time series products, and a terrain elevation dataset. The repeat lidar and fine-resolution imagery offered a key to solving the bottleneck issue of thaw reference data collection, which further provided an opportunity to track post-fire thaw caused by six large fires in the past 25 years in four stages over time: old thaw, lateral thaw, vertical shallow thaw and vertical deep thaw. The developed protocol achieved an overall accuracy of 79 % in classifying these thaw stages and generated a reasonable thaw pattern mainly controlled by fires and locally modified by other drivers. Identifying degradation patterns can help understand the permafrost-fire-climate system. The protocol is a valuable alternative to current thermokarst mapping techniques.