SPATIOTEMPORAL URBAN THERMAL HETEROGENEITY AND MORPHOLOGY-RELATED SPATIAL ASSOCIATIONS IN A SEMI-ARID URBAN SYSTEM

Abstract

Urban thermal heterogeneity in semi-arid environments is shaped by nonlinear interactions between surface energy balance, three-dimensional morphology, and land cover dynamics, yet existing studies typically treat land surface temperature (LST) as a deterministic, error-free pixel value while ignoring spatial dependence and compounded uncertainties. This study reframes urban thermal heterogeneity as a spatially autocorrelated, uncertainty-perturbed manifestation of the surface energy balance under heterogeneous morphological boundary conditions. We integrate Local Climate Zone (LCZ) taxonomy, spatial econometrics (spatial error and spatial lag models), and nested Monte Carlo block bootstrap uncertainty propagation (B=1,000) to examine housing morphology LST associations in Ain Smara, a semi-arid urban system in northeastern Algeria, using Landsat and Sentinel-2 imagery (2013 - 2023). Results show that collective housing is associated with 2.5°C higher LST than individual housing (95% CI: 1.7 - 3.3) after spatially-aware propensity score matching, with contrasts largest in the urban core. Mediation analysis reveals that NDVI (50%) and sky view factor (38%) dominate the morphological contrast, while albedo plays a minor role (17%). Spatial dependence is substantial (global Moran's I = 0.67), and ignoring it underestimates standard errors by approximately 40%. Uncertainty is dominated by spatial bootstrap (48%) and downscaling residuals (32%), not retrieval errors. Our integrated framework moves beyond deterministic "morphology drives LST" narratives toward a probabilistic, spatially explicit, and uncertainty-aware conceptualization, providing robust inference for urban heat mitigation in data-scarce regions of the Global South.