Sulfur dioxide ( \({{\textrm{SO}}_{2}}\) ) is a pollutant primarily emitted through the combustion of fossil fuels and industrial activities, contributing significantly to environmental degradation and public health risks. Monitoring \({{\textrm{SO}}_{2}}\) deposition poses challenges due to spatial variability, technical complexities, and financial constraints. This study develops a machine learning model to predict \({{\textrm{SO}}_{2}}\) deposition in microclimates using reanalysis datasets and land cover data. The model leverages \({{\textrm{SO}}_{2}}\) data from the MERRA-2 reanalysis dataset and urbanization information from the Copernicus Land Cover dataset to account for localized variations. Multiple machine learning algorithms, including Random Forest, ExtraTrees, Support Vector Regression, and Ordinary Least Squares, were evaluated, with Random Forest achieving the best performance. The RF model yielded an R2 score of 0.81 ± 0.09 and an RMSE of 9.39 ± 3.38, demonstrating its ability to explain a substantial portion of the variance in the data while maintaining low prediction errors. This project contributes to the field by developing a methodology for predicting \({{\textrm{SO}}_{2}}\) levels in areas with limited monitoring infrastructure, offering a flexible model applicable across diverse urban and industrial regions. Additionally, \({{\textrm{SO}}_{2}}\) deposition maps generated from this model provide valuable insights for environmental agencies, enabling more effective pollution control strategies and mitigation efforts.
