Spatial heterogeneity of remote sensing-based surface heat conditions and environmental associations in Fuzhou, China: Insights from multiscale geographically weighted regression

Abstract Urban thermal conditions exhibit substantial spatial heterogeneity across cities. This study develops a Relative Surface Heat Index (RSHI) to characterize remote sensing-based surface thermal conditions and examines its spatially heterogeneous and multiscale associations with urban environmental factors in the central urban area of Fuzhou, China. Based on Landsat 9 imagery, RSHI integrates normalized land surface temperature (LST) and the Normalized Difference Water Index (NDWI). Spatial autocorrelation analysis and ordinary least squares (OLS), geographically weighted regression (GWR), and multiscale geographically weighted regression (MGWR) were used to examine the relationships between RSHI and built-up land (BU), water bodies (WB), urban green space (UGS), and nighttime light intensity (NTL). RSHI ranged from 0.236 to 0.910, with a mean of 0.648, and showed significant spatial clustering (Moran’s I = 0.452, p < 0.001). NTL had the largest global correlation with RSHI (r = 0.534), whereas BU, WB, and UGS showed weak global correlations. However, MGWR revealed substantial local variation in these associations. Model performance improved from OLS to GWR and MGWR, with adjusted R² increasing from 0.284 to 0.487 and 0.529, respectively. MGWR bandwidths ranged from 52 nearest neighbors for NTL to 2030 for BU, indicating distinct spatial scales of association. These findings highlight the importance of considering spatial heterogeneity and variable-specific spatial scales in urban thermal-environment assessment and planning.