Industrial Air Quality Assessment through Multi-Source Data and Random Forest Modeling: A Case Study of Assaluyeh, Iran

Air pollution is one of the most challenging environmental problems in industrial areas such as Assaluyeh, Iran. In this study, three sources of terrestrial, meteorological, and Sentinel-۵P satellite data are integrated from ۲۰۲۱ to ۲۰۲۳ to predict and analyze the concentrations of nitrogen dioxide (NO₂) and sulfur dioxide (SO₂) to create a single dataset. A random forest regression model was designed and trained to estimate surface concentrations, and the SHAP library was used to analyze the importance of meteorological data features and analyze the impact of these features. The prediction results show the appropriate performance of the model with R² values of ۰.۷۶ and ۰.۷۵ and RMSE values of ۲.۸۳ μg/m۳ and ۹.۴۷ μg/m۳ for SO₂ and NO₂, respectively. Using SHAP analysis, we found that meteorological parameters such as sunshine hours and relative humidity have the greatest impact on changes in nitrogen dioxide pollutants. Also, regarding sulfur dioxide pollutants, we concluded that the parameters of maximum temperature and the average total ۲۴-hour solar radiation have the greatest impact. By designing and training the aforementioned model, we can have a practical view of how to manage and control pollutant concentrations.

کلیدواژه ها:

Air pollution ، Random Forest ، SHAP ، NO₂ ، SO₂ ، Assaluyeh