Land Surface Temperature (LST) is a critical metric for understanding surface-atmosphere interactions and serves as a key variable in analyzing the Urban Heat Island (UHI) effect, a phenomenon where urban areas exhibit elevated temperatures compared to surrounding rural regions. UHI effects, driven by urbanization and surface thermal properties, significantly impact urban climates, particularly under a changing climate. This study evaluates the utility of MODIS daily and 8-day composite data in modeling summer LST and UHI effects across 137 cities in continental France over a 10-year period. Using Random Forest machine learning model and SHAP analysis, we assessed the role of structural, temporal, and meteorological variables in predicting LST and UHI dynamics for both daytime and nighttime. The results reveal that MODIS daily data achieves higher model accuracy (R 2 = 0.85 for daytime, R 2 = 0.85 for nighttime) compared to 8-day composites (R 2 = 0.75 for daytime, R 2 = 0.70 for nighttime) when meteorological inputs are included. However, in scenarios without weather data, the MODIS 8-day data outperformed daily data for daytime LST prediction (R 2 = 0.57 vs. R 2 = 0.48), emphasizing its potential for regions with limited meteorological coverage. Month emerged as a key predictor across all models, serving as a proxy for seasonal temperature variability. Diurnal analyses revealed stronger daytime UHI prediction accuracy whereas nighttime models showed lower accuracy, particularly for 8-day composites. Overall, this study clarifies when coarser-temporal satellite products can reliably substitute daily observations and identifies the dominant drivers of urban heat, providing practical guidance for large-scale UHI monitoring and heat-resilient urban planning in regions with limited ground-based meteorological data. • Utilized Random Forest model to analyze the relative contributions of meteorological, structural, and temporal variables to LST and UHI dynamics. • Average daytime UHI was significantly higher (6°C) compared to nighttime UHI effect (1.8°C). • 8-day composites perform comparably to daily data for daytime but less effectively for nighttime UHI analysis. • Highlighted the role of Month as a key predictor in UHI models capturing seasonal variations in urban-rural temperature dynamics.
Kporha et al. (Thu,) studied this question.