Asian Journal of Geographical Research

Land Surface Temperature is a critical indicator of regional climate dynamics, yet modelling its spatial distribution across heterogeneous landscapes remains methodologically challenging. This study presents a hybrid geostatistical framework integrating Trend Surface Analysis with Regression Kriging to predict decadal Land Surface Temperature patterns (2015–2025) across West Bengal, India. Using Moderate Resolution Imaging Spectroradiometer-derived Land Surface Temperature composites and four environmental covariates, i.e., the Normalized Difference Vegetation Index, the Normalized Difference Built-up Index, the Normalized Difference Water Index, and the Digital Elevation Model. The study hierarchically evaluated polynomial Trend Surface Analysis models (first to sixth order) against the Regression Kriging approach. The sixth-order Trend Surface Analysis captured macro-scale thermal gradients (coefficient of determination = 0.719, root mean square error = 0.966°C), yet retained spatially autocorrelated residuals. In contrast, the hybrid Regression Kriging model substantially outperformed all deterministic surfaces, achieving superior predictive accuracy (coefficient of determination = 0.952, root mean square error = 0.408°C) - a 72.1% reduction in prediction error relative to linear Trend Surface Analysis. Results reveal a pronounced north-south thermal gradient (2.8°C per 100 kilometres), with identified hotspots in the Western Lateritic Plateau (31.27°C) and urban-industrial corridors including Kolkata and the Durgapur-Asansol belt. Elevation, vegetation density, and impervious surface cover emerged as dominant mechanistic controls. This research demonstrates that hybrid geostatistical modelling significantly enhances Land Surface Temperature prediction fidelity in environmentally complex terrains, offering a robust framework for evidence-based climate adaptation and urban planning.