Assessment and enhancement of Landsat 8 land surface temperature retrieval using mono window algorithm and machine learning approaches

Urban regions such as Klang Valley in Malaysia are increasingly affected by rising Land Surface Temperatures (LST) driven by rapid urbanization and climate change. Accurate LST retrieval is essential for environmental monitoring, climate analysis, and urban heat island studies. However, the challenges remain in validating satellite-derived LST against ground-based measurements, particularly in tropical regions with frequent cloud cover. This study aims to retrieve LST using the Mono Window Algorithm (MWA) applied to the thermal infrared data from Landsat 8 and 9 Operational Land Imager/Thermal Infrared Sensor (OLI/TIRS) imagery from 2015 to 2022. Selected images with less than 40 % cloud cover were used to ensure data quality. The retrieved LST values were validated against the air temperature dataset obtained from the Malaysian Meteorological Department (METMalaysia) at several ground stations. To enhance prediction accuracy, machine learning regression models including Fine Tree of Regression Trees, Fine Gaussian Support Vector Machine (SVM), and Wide Neural Network (NN) were tested. Model performance was evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R²). The Fine Tree of Regression Trees model achieved the highest accuracy, with RMSE of 0.8876 °C, MAE of 0.7878 °C, and R² of 0.7011. These findings demonstrate the potential of combining MWA with machine learning for reliable LST estimation and highlight its applicability in environmental and urban climate analysis research.