Examining the impact of land use and land cover changes on land surface temperature in Herat city using machine learning algorithms

Urbanization has made significant changes in Land Use Land Cover (LULC) patterns and these changes have a considerable impact on Land Surface Temperature (LST) in highly populated and rapidly growing cities. The current study aims to analyze the past and future LULC changes and seasonal (summer and winter) LST patterns in Herat City, Afghanistan. Random Forest (RF) algorithm and Landsat thermal bands were applied to retrieve LULC changes and associated LST patterns during 2000 and 2020. In addition, the Cellular Automata-Markov (CA-Markov) and the CA-Logistic Regression (CA-LR) models were used to predict the future LULC and seasonal (summer and winter) LST patterns, respectively for the year 2030. The CA-Markov and CA-LR models' validation for LULC and LST were assured using the predicted and actual data for the year 2020. Results showed a significant increase in the built-up area about 48.06 km2 (42.53%), while the vegetation 27.65 km2 (24.48%) and bare soil 20.45 km2 (18.11%) reduced significantly, however, the water bodies decreased slightly in Herat City. The seasonal (summer and winter) LST patterns revealed higher temperatures in built-up area followed by the bare soil, vegetation and water bodies. Predicted LULC patterns showed a considerable increase in the built-up area of about 16.69 km2 (14.77%) for the year 2030 as compared to 2020. Seasonal analysis predicts that by 2030, around 51.84 km2 (45.88%) of Herat City may experience summer LST > 45 degrees C, while 90.15 km2 (79.78%) could have winter LST ranging from 10-20 degrees C. The results indicated a direct relation between rapid urbanization and rising LST, which negatively impacts both the local population and the environment. To ensure sustainable urban growth and reduce the impact of LST in Herat City, the future city master plans should prioritize urban greenery and the protection of natural resources.