Surface Urban Heat Island Dynamics in a High-Elevation Valley City on the Tibetan Plateau: Diurnal, Seasonal, and Spatial Evidence from Nyingchi, 2000–2025

The thermal environment of high-elevation valley cities is jointly shaped by urbanization and complex geographic settings. Using Landsat and MODIS data from 2000 to 2025, annually varying built-up-area boundaries, and elevation-constrained rural references, we characterized the diurnal and seasonal patterns of the surface urban heat island (SUHI) in Nyingchi, Tibet Autonomous Region, China, and used an XGBoost model with out-of-fold SHAP interpretation to examine predictive associations with summer daytime surface thermal anomalies in 2025. The results showed that, since 2000, the urban built-up area of Bayi District expanded from 2.15 to 12.34 km2. Daytime SUHI was strongest in summer (5.24 °C) and was clearly distinct from that in other seasons, whereas nighttime SUHI remained a weak warming signal in all four seasons. A conventional buffer reference that included cooler mountain slopes overestimated summer daytime SUHI by 4.00 °C relative to the elevation-constrained reference, indicating that rural-reference definition is decisive in this terrain. Spatial cross-validation of the XGBoost model yielded an R2 of 0.764; moisture and vegetation variables were associated with lower predicted anomalies, whereas built-up status was associated with higher predicted anomalies. These findings identify a pronounced diurnal-seasonal asymmetry in Nyingchi’s SUHI and provide a basis for terrain-aware thermal-environment assessment; the machine-learning associations are not interpreted as causal intervention effects.