Ground-based data analysis and combined approaches for particulate matter 2.5 prediction

Abstract

Effective air quality management requires accurate prediction of particulate matter 2.5 levels, which are influenced by various factors, including weather and human activities. This study explores integrating ground-based meteorological and traffic data with satellite-derived datasets to improve particulate matter 2.5 prediction accuracy in Tehran. The results demonstrate that while ground-based data can offer valuable insights, combining these datasets with satellite information significantly enhances predictive performance. Accurate prediction of particulate matter 2.5, a harmful air pollutant linked to respiratory and cardiovascular diseases, is critical for managing air quality in densely populated cities. This study compares remote sensing data with four configurations of ground data, meteorological and traffic data, and a combination of remote sensing and meteorological data in predicting particulate matter 2.5 concentrations across Tehran’s 22 districts. Ground data included meteorological factors, traffic data, and direct air quality measurements, supplemented by satellite-based aerosol optical depth estimates from NASA’s HD4 archives via Google Earth Engine. Using machine learning, deep learning, and statistical models, study evaluated the predictive accuracy of each dataset. The findings show that remote sensing data consistently outperforms all ground data configurations, offering superior performance and flexibility. This indicates that satellite-based remote sensing is an effective, independent tool for particulate matter 2.5 prediction, particularly in regions lacking ground monitoring infrastructure. These results underscore the potential of satellite-derived particulate matter 2.5 estimates for public health research and air quality management. The study emphasizes the importance of remote sensing in air pollution monitoring and proposes its integration into future air quality forecasting systems.