Machine learning-based assessment of land use change effects on land surface temperature fluctuations in Ho Chi Minh...

Sustainable urban development requires actionable insights into the thermal consequences of land transformation. This study examines the impact of land use and land cover (LULC) changes on land surface temperature (LST) in Ho Chi Minh city, Vietnam, between 1998 and 2024. Using Google Earth Engine (GEE), three machine learning algorithms-random forest (RF), support vector machine (SVM), and classification and regression tree (CART)-were applied for LULC classification. RF delivered the best performance, achieving an average overall accuracy of 93.09% and a kappa coefficient of 0.916. The findings show that built-up areas expanded by over 12%, primarily at the expense of agricultural lands, while mean LST increased by approximately 1.75 °C during the study period. Urban zones exhibited the highest LST values due to the urban heat island (UHI) effect, whereas vegetated and water-covered areas showed strong cooling functions. A positive correlation was identified between LST and built-up areas, while vegetation and water indices were negatively correlated with LST. The study integrates LULC classification, LST retrieval, and biophysical indices to capture urban heat dynamics in a tropical megacity context. The results highlight the importance of green-blue infrastructure in mitigating urban warming and provide valuable guidance for policymakers seeking to enhance urban climate resilience through sustainable land management.