Satellite-based land surface temperature (LST) products are frequently affected by cloud cover and atmospheric conditions, resulting in missing data that significantly limits the continuous monitoring of the thermal environment in complex terrains, such as the Tibetan Plateau. Existing spatiotemporal interpolation methods face clear accuracy limitations when addressing extensive data gaps, while physical models often struggle due to insufficient meteorological inputs in complex landscapes. Moreover, conventional data-driven approaches usually overlook local spatial variations, resulting in smoothed thermal patterns and systematic errors. To overcome these issues, we propose a Physically Constrained Spatial Residual Learning framework. In this framework, we use the Enhanced Annual Temperature Cycle (EATC) model to capture the temporal baseline of LST first. Then, we integrate multi-source auxiliary data into the Geographical-XGBoost (G-XGBoost) algorithm to model spatial nonlinear residuals. Using simulated cloud masks on the 2017 MODIS LST dataset from the Qinghai–Tibet Engineering Corridor, we show that the hybrid model outperforms both individual physical models and global machine learning models in accuracy and spatial detail recovery. Validation results yield an R² of 0.88, an RMSE of 1.92 K, and a mean bias of 0.07 K. Seasonal evaluations indicate best performance in winter (RMSE = 1.19 K) with robust performance in summer. Furthermore, the framework reduces boundary artifacts and accurately reproduces thermal spatial patterns in complex terrain through adaptive local bandwidth and weight adjustments. This approach provides a reliable method for high-precision LST reconstruction over heterogeneous alpine surfaces.

The Source Region of the Yangtze River is a high-altitude area with extensive permafrost on the Tibetan Plateau. While temperature, precipitation, and radiation significantly affect vegetation phenology, the influence of permafrost changes remains unclear. Using the daily Long-term Seamless NOAA AVHRR NDVI Dataset of China (2003–2022), we extracted the start (SOS) and end (EOS) of the growing season in the Source Region of the Yangtze River (SRYR). Soil thawing date (SOT) was obtained from freeze–thaw state products, while active layer thickness (ALT) was estimated using the Stefan model based on MODIS land surface temperature (LST). Partial least squares regression and mediation analysis quantified the direct and indirect effects of permafrost degradation. Results show: (1) The end of the growing season (EOS) became significantly earlier in 64.33% of the region, while the start of the growing season (SOS) showed little change. (2) The effect of SOT on SOS depends on moisture conditions. Earlier SOT leads to earlier SOS in wetter areas by supplying meltwater, but delays SOS in cold–dry areas by increasing soil water loss. (3) Thicker ALT strongly promotes earlier EOS, accounting for up to 42.61% of EOS variation in cold–dry zones, because a deeper active layer potentially promotes downward movement of water, which may further lead to the potential leaching of nutrients from the shallow root zone, limiting resources for shallow-rooted plants. (4) Alpine meadows respond more strongly to permafrost changes than alpine grasslands. Overall, water loss caused by permafrost degradation may reduce the potential lengthening of the growing season under climate warming, highlighting the key role of soil water in linking permafrost and vegetation dynamics.