Study of the Atmospheric Pollution Control Framework in the Transboundary Pollution Region

Transboundary atmospheric pollution is a major environmental challenge in northwestern Bangladesh, particularly in Rajshahi, where pollutants transported from the Indo-Gangetic Plain significantly influence local air quality and climatic conditions. This study develops a machine learning-based framework to investigate the relationship between atmospheric pollutants and regional temperature variations. The research aims to identify the key pollution parameters contributing to atmospheric warming. We used publicly available historical environmental datasets comprising PM₂.₅, PM₁₀, O₃, NO₂, and SO₂ concentrations as input features, while average temperature was selected as the target variable. There were four machine learning models, Random Forest (RF), Gradient Boosting (GB), Support Vector Regression (SVR), and Deep Neural Network (DNN), which were implemented and evaluated using an 80/20 training/testing data split. We assessed the model performance through regression-based statistical indicators to determine predictive accuracy and robustness. The results showed that the DNN model outperformed the other approaches, demonstrating superior capability in capturing the complex nonlinear interactions between atmospheric pollutants and temperature. Furthermore, we conducted feature importance analysis to quantify the contribution of individual pollutants to temperature variability. The analysis identified PM₂.₅ as the most influential parameter associated with atmospheric warming in the Rajshahi region, followed by other particulate and gaseous pollutants. The proposed framework provides valuable insights for pollution control strategies and climate-informed environmental management in transboundary pollution regions.