Spatiotemporal analysis of airborne pollutants and health risks in Mashhad metropolis: enhanced insights through sensitivity analysis and machine learning.
Journal:
Environmental geochemistry and health
PMID:
39724450
Abstract
The study delved into an extensive assessment of outdoor air pollutant levels, focusing specifically on PM, SO, NO, and CO, across the Mashhad metropolis from 2017 to 2021. In tandem, it explored their intricate correlations with meteorological conditions and the consequent health risks posed. Employing EPA health risk assessment methods, the research delved into the implications of pollutant exposure on human health. Results unveiled average annual concentrations of PM, SO, NO, and CO, standing at 27.22 µg/m, 72.48 µg/m, 26.8 µg/m, and 2.06 mg/m, respectively. Intriguingly, PM displayed positive correlations with temperature and wind speed, while exhibiting negative associations with relative humidity and precipitation. Conversely, both SO and NO concentrations showcased negative correlations with temperature, relative humidity, wind speed, and precipitation. Furthermore, CO demonstrated negative relationships with both wind speed and precipitation. The analysis of mean hazard quotients (HQ) for PM and NO indicated values exceeding 1 under 8- and 12-h exposure scenarios, pointing towards concerning health risks. Spatial distribution revealed elevated CO levels in the northwest, north, and east areas, while NO concentrations were predominant in the north and south regions. Through Sobol sensitivity analysis, PM, EF, and NO emerged as pivotal influencers, offering valuable insights for refining environmental models and formulating effective pollution mitigation strategies. Air pollution index (AQI) forecasting was modeled using advanced machine learning comprising Random Forest (RF), Decision Tree (DT), K-Nearest Neighbors (KKN), and Naive Bayesian (NB). Results showed that the RF model with the highest accuracy (R = 0.99) was the best prediction model.