Table 3 Multi-source data, preprocessing and factor selection rationale.

Landsat5/8 Collection 2Level-2L2)

usgs.gov²⁶

30 m resolution LST raster

LST is the core indicator for quantifying SUHI intensity and its spatial–temporal variability. Satellite-derived LST is widely adopted in urban climate studies due to its spatial continuity and comparability across periods²⁷.

Administrative Boundaries

National Geographic Information Public Service Platform²⁸

/

Administrative boundaries are used to delineate the study area and ensure spatial consistency in data integration and factor analysis.

Land Cover

Yang & Huang (2024)²⁵

30 m resolution land cover raster

Land cover types strongly influence surface thermal behaviour, as impervious surfaces and vegetated areas exhibit contrasting heat storage and radiation properties, making land cover a primary driver of SUHI patterns²⁹.

DEM Data

Copernicus Global DEM³⁰

100 m resolution DEM raster

Topographic attributes, such as elevation, slope and aspect, are included because they influence the receipt of solar radiation, airflow patterns and microclimatic conditions, which in turn affect spatial variations in LST. Studies using spatial econometric and machine-learning approaches have demonstrated that slope and other terrain features contribute significantly to urban heat island intensity, especially in landscapes with noticeable surface undulation^[31].

100 m resolution slope raster

100 m resolution aspect raster

Vector Map Data

OPENSTREETMAP³²

100 m resolution road density raster

Urban form metrics (density, height, road density) strongly influence surface thermal conditions through shading, heat storage and reduced ventilation^33,34.

100 m resolution building height raster

100 m resolution building density raster

Population Density

Chen, Xu, Ge, Zhang, and Zhou (2024)³⁵

100 m resolution population density raster

Population density is commonly used as a proxy for anthropogenic activity intensity and associated heat emissions, which contribute to urban thermal anomalies³⁴.

Normalised Difference Vegetation Index (NDVI)

National Ecosystem Science Data Center³⁶

100 m resolution NDVI raster

The NDVI quantifies vegetation cover, which mitigates surface temperature through shading and evapotranspiration and typically shows a negative correlation with LST³⁴.