Evolution of land surface temperature in Beijing and its multi-source driving mechanisms

Rapid urbanization and climate change have markedly shifted Beijing’s climate in recent decades from cold–dry toward warm–humid conditions, raising urgent questions about the dominant controls of its surface thermal environment. Using MODIS land surface temperature (LST) observations from 2000–2024, combined with NDVI, nighttime lights, the Human Footprint index, ERA5-Land meteorology, and surface albedo, we investigate the spatiotemporal evolution of LST and quantitatively attribute its drivers across spatial scales.

Long-term LST trends were robustly identified using Theil–Sen slopes and the Mann–Kendall test, while the relative contributions of natural and anthropogenic factors were quantified through ensemble machine-learning models (Random Forest and XGBoost) coupled with SHAP-based interpretability. This integrated framework enables a scale-aware attribution of LST dynamics rather than simple correlation analysis.

Pronounced urban heat island patterns are observed in Beijing’s core districts (Dongcheng, Xicheng, Haidian, Chaoyang, Fengtai, and Shijingshan), gradually weakening toward suburban areas. Between 2000 and 2024, LST increased significantly or highly significantly across 34.67% of the city—mainly in the urban core and southeastern districts (Daxing and Tongzhou)—while 63.52% experienced cooling, particularly around the Miyun Reservoir and along the Guishui River in Yanqing. Attribution results reveal that Human Footprint intensity and nighttime light activity exert the strongest warming effects, whereas vegetation greenness (NDVI), relative humidity, and soil moisture consistently mitigate LST. The maximum cooling rate is associated with NDVI values between 0.25 and 0.55. SHAP rankings identify Human Footprint, air temperature, NDVI, and nighttime lights as the dominant drivers at the metropolitan scale, while surface albedo plays a more prominent role within the urban core.

These findings provide a quantitative and interpretable assessment of the scale-dependent drivers shaping Beijing’s surface thermal environment and offer actionable insights for urban climate adaptation, including optimized green-space allocation, high-albedo surface renovation, and land-use planning.