A long-term (1950–2024) machine learning-corrected land surface temperature dataset over the Tibetan Plateau

Description: The Tibetan Plateau (TP) is a critical amplifier of global climate change and a regulator of Asia's water resources, demanding accurate land surface temperature (LST) records. However, the ERA5-Land reanalysis data exhibit substantial biases across this complex terrain. This dataset provides a long-term, high-resolution (0.1°), monthly LST dataset over the Tibetan Plateau for the period 1950–2024. It is generated by fusing high-density meteorological observations with ERA5-Land using an ensemble machine learning approach. Key improvements over the original ERA5-Land product: Coefficient of determination (R²): Improved from 0.79 to 0.96 Root-mean-square error (RMSE): Reduced from 9.26°C to 2.33°C Warming trend uncovered (Corrected vs. ERA5-Land): 0.32°C per decade (p < 0.05) vs. 0.24°C per decade (p < 0.05) – revealing a significantly stronger warming trend Seasonal pattern: Warming is most pronounced in winter, locally exceeding 0.80°C per decade Spatial heterogeneity: Marked spatial variability with maxima in western Ngari and minima in the southeastern TP Dataset characteristics: Spatial resolution: 0.1° × 0.1° (approximately 11 km at the equator) Temporal resolution: Monthly Data format: NetCDF (.nc) Time range: 1950–2024 Region: Tibetan Plateau Method: Ensemble machine learning approach Significance:This dataset reveals a pervasive ERA5-Land cold bias, particularly acute in cryosphere-sensitive high-altitude zones, necessitating urgent recalibration of climate models and impact projections dependent on these data. It provides an indispensable benchmark for understanding accelerated TP warming and its profound consequences for Asian water security, cryosphere loss, and global climate feedbacks.