This study evaluates land use and land cover (LULC) changes in Fars Province, Iran, using machine learning algorithms, specifically comparing the performance of two non-parametric support vector machine (SVM) and random forest models. With rapid urbanization and agricultural expansion, accurate LULC classification is critical for environmental monitoring and land management.  Applying the Google Earth Engine platform, multi-temporal Landsat 8 imagery was assessed. The findings demonstrated all classification methods presented high accuracy metrics and kappa coefficient values. The SVM algorithm attaining a mean overall accuracy of 91.42% for Landsat 8 imagery to show best performance among all evaluated methods. According to LULC change detection performed by the most accurate classification algorithm, the results indicated an increase in urban parks, gardens, and mountainous rangelands, while barren lands experienced a decline. The evaluation of LULC changes impacts on land surface temperature (LST) shows that enhanced vegetation cover played a key role in reducing LST. A remarkable decrease in both maximum and minimum LST values was observed, declining 37.31°C and 22.47°C in 2019 to 34.45°C and 19.98°C in 2023, respectively. Furthermore, the findings highlight integrating high-resolution satellite imagery with the SVM algorithm leads to achieve a highly accurate and efficient approach for LULC mapping. Consequently, this method proves to be a valuable tool for decision-making in natural resource management and urban planning in similar regions.