Explainable artificial intelligence models for proposing mitigation strategies to combat urbanization impact on land surface temperature dynamics in Saudi Arabia

Abstract

Urbanization in Saudi Arabia has significantly altered land use and land cover (LULC), driving notable changes in land surface temperature (LST) dynamics. This study aims to analyze spatio-temporal LULC changes from 2002 to 2022 and their impact on LST, employing optimized machine learning models like Random Forest, Gradient Boosting, XGBoost, and LightGBM, enhanced by explainable artificial intelligence (XAI). Results show a complete loss of water bodies (from 99.25 km2 to 0 km2), urban expansion (from 1344.38 km2 to 1377.12 km2), and a decline in sparse vegetation (from 231,430.12 km2 to 230,454.50 km2). Concurrently, LST increased, with temperatures rising from 25.08 °C–54.42 °C in 2018 to 26.08 °C–56.31 °C in 2022. The LightGBM model demonstrated the highest predictive accuracy with the lowest mean absolute error (MAE). SHAP analysis revealed that higher aerosol concentrations, air temperatures, and pollutants (CO, NO2, SO2) increase LST, while vegetation (NDVI) and water presence (NDWI) mitigate it. The study emphasizes the significant environmental impact of urbanization on LST and highlights the importance of integrated environmental management strategies, such as enhancing vegetation cover, optimizing water management, and minimizing pollution, to mitigate urban heat island effects and foster sustainable development in Saudi Arabia.