Modelling Land Surface Temperature Across Morocco's Complete Climatic Gradient: A National LST Equation Linking Topography, Spectral Indices, and Thermal Dynamics
Year: 2026
Type: conference-paper
Abstract: Accurate prediction of Land Surface Temperature (LST) from topographic and spectral covariates remains a fundamental challenge across heterogeneous landscapes. A multiform regression framework was applied to 125,669 Landsat-8 pixels from 35 Moroccan cities spanning seven zones (Mediterranean, Atlantic, Continental, Mountain, Pre-Saharan, Saharan, Saharan-South), covering an elevation range of 0 to 4,167 m. Eleven algebraic forms were systematically compared using adjusted$\mathbf{R}^{\mathbf{2}}$as the sole selection criterion. The hybrid form F8 (Poly$^{2}+\text{ln}+\sqrt{ }+interactions)$achieved the highest adjusted$\mathbf{R}^{\mathbf{2}}$of 0.7507 (RMSE$=\mathbf{3. 6 8}^{\circ} \mathrm{C})$. XGBoost-SHAP attribution revealed topographic features explain 67.2% of LST variance, with longitude (2.20°C), elevation (1.63°C), and albedo$(\mathbf{1. 3 5}^{\circ} \mathbf{C})$as the three most influential predictors. A LASSOparsimonious simplified equation, the Morocco LST equation (5 variables,$R^{2}{ }_{A d j}=0.710, R M S E=3.97^{\circ} C)$achieves 94.6% of the best model's performance at a fraction of complexity, enabling operational thermal mapping from DEM and Landsat data alone. Calibrated on cloud-free 2023-2025 dry-season composites at Landsat 8's fixed morning overpass, it represents, to the best of our knowledge, the first analytically compact LST equation across Morocco's complete climatic diversity, offering a transferable tool for thermal zoning and drought early-warning across data-sparse North African environments.
Authors: Achraf Rachidi, Sanaa Hayani Mounir
Language: English
Topic: Urban Heat Island Mitigation
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
SDG: Climate action
Open Access status: hybrid
OpenAlex ID: W7204113949