The TRIMS LST product was obtained from the National Tibetan Plateau Data Center (https://doi.org/10.11888/Meteoro.tpdc.271252 .). This product was generated using an enhanced satellite thermal infrared remote sensing–reanalysis data fusion approach. The primary input datasets include the Terra/Aqua MODIS LST products and GLDAS reanalysis data, while auxiliary inputs consist of satellite-derived vegetation indices, surface albedo, and other surface parameters. By fully exploiting the high-frequency and low-frequency components of satellite thermal infrared observations as well as the spatial autocorrelation of land surface temperature, this method reconstructs a high-quality, all-weather LST dataset with improved spatiotemporal continuity and consistency. The ESA land cover data were obtained from the European Space Agency (ESA) Climate Change Initiative Medium Resolution Land Cover project (CCI MRLC) (https://data.ceda.ac.uk/neodc/esacci/land_cover/data/pft/v2.0.8/ ). By integrating multiple high-resolution auxiliary datasets, this product provides annual fractional cover estimates of 14 global plant functional types (PFTs) from 1992 to 2020, thereby effectively reducing the uncertainty associated with traditional land cover–to–PFT conversion approaches. The dataset is provided at an annual temporal resolution with a spatial resolution of 300 m. The Copernicus DEM was released by the European Space Agency and is based on global radar satellite observations acquired during the TanDEM-X mission (2011–2015). After interferometric processing and subsequent semi-automatic manual editing, the WorldDEM product was generated and further resampled to produce the Copernicus DEM at a spatial resolution of 30 m. The dataset is publicly available from the Copernicus Data Space (https://dataspace.copernicus.eu/explore-data/data-collections/copernicus-contributing-missions/collections-description/COP-DEM ).

  （1）Using ArcGIS 10.8 spatial analysis tools, the TRIMS LST data were mosaicked, clipped, and projected, with the projection coordinate system set to WGS84.（2）In the Python programming environment, a statistical fitting model between GST and LST was established using ridge regression. The model was optimized via 10-fold cross-validation and subsequently applied to calculate regional GSTs.（3）DEM data were resampled to a 1 km resolution using bilinear interpolation, and ESA land cover data were resampled to 1 km using the mode method to match the spatial resolution of TRIMS LST.（4）The distribution of permafrost was derived using the TTOP model. The resulting permafrost map was binarized (permafrost/seasonally frozen ground) to generate a 2005–2020 permafrost distribution dataset for the Tarim Basin.

  The simulated permafrost map of the Tarim Basin was comprehensively validated using in situ subregional permafrost survey maps (Aerjin and West Kunlun), field borehole data, ground-penetrating radar measurements, and existing Northern Hemisphere permafrost maps. The accuracy of the subregional permafrost simulations was quantitatively assessed using overall accuracy (OA) and the Kappa coefficient (k) derived from a binary confusion matrix.The validation results indicate that the simulated permafrost spatial distribution is highly consistent with both wetland survey maps and existing large-scale permafrost maps. The area of permafrost was estimated to be approximately 21.75×104 km2 (21.02% of the basin), while seasonal frozen ground covered about 79.53 × 104 km2 (76.86%), excluding glaciers (1.79 × 104 km2, 1.73%) and lakes (0.41 × 104 km2, 0.39%). Permafrost is mainly distributed in the Karakoram, Kunlun, and Aerjin Mountains, as well as the Pamir Plateau and the eastern section of the Tianshan Mountains. Comparisons with survey maps in the West Kunlun and Aerjin subregion yielded k values of 0.84 and 0.52, with corresponding OAs of 0.97 and 0.87 (average k ≈ 0.82), significantly outperforming the Obu map (k = 0.49 and 0.43) and the Ran map (k = 0.42 and 0.02). Borehole data indicate the coexistence of stable and unstable permafrost in the Aerjin region; both types were correctly identified as permafrost in the simulation, and seasonally frozen ground was also reasonably well captured. Temporal analysis from 2005 to 2020 shows that permafrost degradation primarily occurred at the margins of continuous permafrost and in discontinuous permafrost zones, consistent with regional warming trends.Therefore, this dataset demonstrates high spatial and temporal reliability in permafrost regions of the Tarim Basin and provides a robust foundation for studies on regional permafrost dynamics, hydrological processes, and ecological responses