A Comprehensive Analysis of AOD and its Species from Reanalysis Data over the Middle East and North Africa Regions: Evaluation of Model Performance Using Machine Learning Techniques

The present study examines the spatiotemporal changes in aerosol optical depth (AOD) and five aerosol species over the Middle East and North Africa (MENA) regions and incorporates an advanced Machine Learning model to predict AOD. The study utilizes reanalysis data from the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2), and the Copernicus Atmosphere Monitoring Service Reanalysis (CAMSRA), in conjunction with the MODIS spanning from 2003 to 2020. Seasonal-averaged AOD₅₅₀ showed the dominance of aerosols in the summer and spring seasons, primarily driven by dust and black carbon, with dust being the most significant contributor due to frequent storms and desert conditions. Other contributors such as sea salt, sulfate, and organic carbon also play crucial roles, underscoring the complex interplay between natural and anthropogenic aerosols. The validation results revealed a high coefficient of determination (R²) for AOD ranging from 0.76 to 0.96 across these datasets. This demonstrates strong predictive accuracy with the XGBoost model, which shows a robust correlation between predicted and actual AOD values with minimal error and no significant bias. The AI/ML model analysis further elucidates the contributions of individual aerosol species to AOD predictions, revealing that dust and black carbon consistently enhance AOD. This study reveals that AOD fluctuations in the MENA region are driven by meteorological factors and drought-induced dust emissions. Ultimately, long-term reliable atmospheric composition reanalysis data can supplement ground-based or remote sensing observations in air quality research, emphasizing the need for continuous assessment of aerosols to inform policies aimed at reducing air pollution and mitigating climate change impacts.