Spatial analysis of interannual changes in thermokarst lakes across the Qinghai-Tibet Plateau with Time-Series SAR imagery

• Obtained the changes at the freezing-thawing transition point of thermokarst lake. • Constructed thermokarst lake features from Sentinel-1A SAR images. • Introduced CBAM to improve U-Net and enhanced thermokarst lake extraction accuracy. • Obtain the interannual variation dataset of thermokarst lake on the QTP (2015–2022) • Thermokarst lake area changes were highly correlated with temperature from 2015 to 2022. The climate changes on the Qinghai-Tibet Plateau (QTP) have led to the widespread development of thermokarst lakes in permafrost regions. The seasonal variations in temperature and precipitation in the permafrost regions of the QTP affect the intra-annual state of thermokarst lakes. The intra-annual seasonal variations of thermokarst lakes pose challenges for accurately capturing inter-annual changes. To accurately obtain inter-annual change information of thermokarst lakes, this study proposes a method for capturing inter-annual changes of thermokarst lakes based on the “thaw-freeze” transition points and time-series SAR imagery. This method identifies the “thaw-freeze” transition points for thermokarst lakes across the entire Qinghai-Tibet Plateau, ensuring that each lake is in a similar thermal state, and proposes an improved U-net-based extraction algorithm for thermokarst lakes. This algorithm leverages the transition points and SAR imagery’s polarization, statistical, and temporal features to accurately capture the changes in thermokarst lakes over the years. The study obtained the extent and inter-annual changes of thermokarst lakes from 2015 to 2022, achieving an accuracy of 91.01% and an F1 score of 0.92. The change extraction results identified spatial heterogeneity in the changes of thermokarst lakes on the QTP. Further investigation explored the impacts of surface temperature, precipitation, and evapotranspiration on the changes in thermokarst lakes. The analysis revealed a higher correlation between thermokarst lake area changes and land surface temperature, while the correlation with precipitation and evapotranspiration changes was relatively lower.