A new inventory and future projections of thermokarst lakes in the permafrost regions of the Qilian Mountains, northeastern Qinghai-Tibet Plateau, China

Thermokarst lakes-lacustrine basins formed by thermokarst processes-are key components of permafrost ecosystems that significantly influence the geomorphology, hydrology, ecology, and the stability of infrastructure, while also contributing to greenhouse gas emissions. High-resolution mapping of thermokarst lakes' current and future distribution presents a considerable challenge. In this study, we leverage a deep learning approach to delineate the boundaries of thermokarst lakes in the permafrost regions of the Qilian Mountains at the northeastern margin of the Qinghai-Tibet Plateau (QTP) using Sentinel-2 imagery, followed by an estimation of future susceptibility to thermokarst lake formation by employing a machine learning method. We identified a total of 2795 thermokarst lakes covering an area of approximately 51.6 km2. Based on future climate scenarios, there will be a significant reduction in the proportion of high-susceptibility areas. Specifically, under the most extreme SSP5-8.5 scenario, the proportion of areas highly susceptible to thermokarst lake formation decreases from the current 6.7 % to 0.01 % in the period 2081-2100. Our comprehensive inventory provides essential data for research on permafrost hydrology and the evaluation and management of water resources, and enhances our understanding of thermokarst lake formation in relation to environmental factors.