Assessing the effects of urbanization and vegetation on land surface temperature using machine learning and GIS-based approach in Abu Dhabi, UAE

Introduction Abu Dhabi has experienced rapid urban growth in recent decades, raising concerns about changes in land surface temperature (LST) within its hyper-arid environment. This study assess the influence of urban expansion and vegetation growth on LST over time and examines the role of land-cover transitions in shaping local thermal conditions. Methods Landsat 7 satellite imagery from 2001 to 2019 was analyzed using machine learning and GIS-based methods. Supervised classification was conducted using QGIS and Google Colab to map urban, vegetation, sand/desert, and water classes. Classification performance was evaluated using accuracy metrics, while linear regression was used to quantify the relationships between land-cover composition and mean LST. Results Mean LST decreased from 45.3 °C in 2001 to 39.4 °C in 2019. Over the same period, sand/desert cover declined by 9%, while urban and vegetation cover increased by 93% and 68%, respectively. Regression analysis identified vegetation as the strongest predictor of lower LST, with each one-percentage-point increase in vegetation share corresponding to an estimated 6.92 °C decrease in mean LST. Urban cover showed a weaker and statistically non-significant relationship after accounting for vegetation and the simultaneous reduction in sand/desert cover. Discussion The findings indicate that the observed temperature reduction is strongly associated with local adaptation strategies, particularly vegetation expansion and the reduction of exposed desert surfaces. This land-cover transition provides valuable evidence to guide future decision‐making for climate resilience in hyper‐arid environments.