Dataset for: Unraveling scale-dependent and non-linear drivers of land surface temperature in a typical arid oasis city

This dataset contains the processed geospatial data matrix used for the multi-scale XGBoost-SHAP analysis presented in the manuscript. It includes seasonal Land Surface Temperature (LST) and corresponding driving factors (e.g., NDVI, NDBI, DEM, NTL, Wind Speed) for Urumqi, China. The data is aggregated at three spatial granularities: 300 m (block scale), 900 m (neighborhood scale), and 1500 m (district scale), covering four seasons from late 2023 to late 2024.