Semi-Automated Detection of Thaw Lakes in Permafrost Areas in Qinghai-Tibet Plateau from Sentinel-2 Images Using Markov Random Field

Abstract

Over the past few decades, climate warming and increasing human activities severely disrupted permafrost distribution in Qinghai-Tibet plateau. However, existing surveying tools and methods limit our research on permafrost over large areas. The formation of thaw lakes in permafrost areas has been strongly linked to the degradation of permafrost, which can be potentially identified in remote sensing images. For identifying and characterizing the spatial distribution of permafrost degradation, here we present a semi-automated method for thaw lake detection in permafrost area. The method mainly includes two steps: (1) Normalized Difference Water Index (NDWI) combined with multi-threshold for training samples generation and (2) Markov Random Field (MRF) for thaw lake detection. The proposed method was applied to detect thaw lakes from 10 m Sentinel-2 image in permafrost area of Qinghai-Tibet plateau. Results demonstrated the effectiveness of the proposed method and the capability of rapid thaw lake detection over large areas. We consider this method may provide a technical support for future research in permafrost changes.