Fine particulate air pollution estimation in Ouagadougou using satellite aerosol optical depth and meteorological parameters

Abstract

This study estimates PM_2.5 concentrations in Ouagadougou using satellite-based aerosol optical depth (AOD) and meteorological parameters such as temperature, precipitation, relative humidity, wind speed, and wind direction. First, Simple Linear Regression (SLR), Multiple Linear Regression (MLR), Decision Tree (DT), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost) models were developed using the available labeled data (AOD and meteorological parameters with corresponding PM_2.5 values) in the city. The XGBoost model outperformed all other models that were used, with a coefficient of determination (R²) of 0.87 and a root-mean-square error (RMSE) of 15.8 μg m⁻³ after a five-fold cross-validation. The performance of the supervised XGBoost model was upgraded by incorporating a semi-supervised algorithm to use large amounts of unlabeled data in the city and allow for a more accurate and extensive estimation of PM_2.5 for the period 2000–2022. This semi-supervised XGBoost model had an R² of 0.97 and an RMSE of 8.3 μg m⁻³ after a five-fold cross-validation. The results indicate that the estimated 24 hour mean PM_2.5 concentrations in the city are 2 to 4 times higher than the World Health Organization (WHO) 24 hour guidelines of 15 μg m⁻³ in the rainy season and 2 to 22 times higher than the WHO 24 hour guideline in the dry season. The results also reveal that the average annual estimated PM_2.5 concentrations are 11 to 14 times higher than the WHO average annual guideline of 5 μg m⁻³. Finally, we find higher PM_2.5 concentrations in the city's center and industrial areas than in the other areas. The results indicate a need for future air pollution policy and mitigation in Burkina Faso to achieve desired health benefits such as reduced respiratory and cardiovascular problems, which will, in turn, lead to decreased PM_2.5 mortality rates.