Modeling air pollution in Dhaka with Bayesian generalized additive mixed model and quantile regression

Abstract Air pollution in Dhaka, Bangladesh, presents a serious challenge to public health and environmental sustainability, largely driven by rapid urbanization and industrial growth. In this study, we investigate the nonlinear and heterogeneous effects of meteorological variables, including temperature, humidity, wind speed, and precipitation, on PM2.5 concentrations across different seasons. To capture these complex dynamics, we employ a Bayesian generalized additive mixed model (GAMM) and quantile regression. The Bayesian GAMM framework captures nonlinear relationships between PM2.5 and meteorological variables while accounting for seasonal random effects. To further explore the distributional impacts of these variables, we apply Bayesian quantile regression, which reveals that their effects are more pronounced during extreme pollution episodes in the dry season and more stable during the rainy season. By integrating these two modeling approaches, we provide a robust, data-driven understanding of air quality dynamics in Dhaka. Our findings offer valuable insights for developing targeted mitigation strategies aimed at improving urban air quality and safeguarding public health.