Estimation of Surface PM_2.5 Concentration from Satellite Aerosol Optical Depth Using a Constrained Observation-Based Model

Studies have established that extreme air pollution is more prevalent and is responsible for more deaths and disability-adjusted life years (DALY) in urban cities, especially in developing economies. However, the paucity of ground-based observation has greatly hindered extensive and long-term monitoring and, as such, a good understanding of the trend and characteristics of air quality where it matters most. Aerosol optical depth (AOD) from satellites retrievals provides good spatial and temporal resolutions of atmospheric aerosols and could be a good proxy for ground-level PM_2.5 concentration. This study used a Bayesian regression model to determine the parameters of a PM_2.5 model at four monitoring stations using AOD and selected atmospheric variables (PBLH and RH) as input. The dry-air reference value (K) and the integrated humidity coefficient (γ) were used to delineate the effects of the aerosol characteristics. The values of K and γ, 0.02<𝐾<0.07 (m²g⁻¹) and 0.54<𝛾<3.14, respectively, are site-specific even within the same country as is the case for Lekki and Benin (both in Nigeria). The PM_2.5 estimates from the developed observation-based model were in good agreement with the ground-based observations (0.55< 𝑟 <0.77
). RH and a combination of PBLH-RH were the best performers in the development of the model. Firstly, this study identifies the unique range of values for K and 𝛾 for site-classes in the sub-Saharan tropical climate. Secondly, PBLH adds more explanatory power to the PM_2.5 estimates in Benin and Douala (both non-coastal cities) while RH improves the performance of the model significantly in Lekki and Owendo (both coastal cities). For West Africa and similar data-sparse regions, the methodology presented here offers a practical pathway to enhance air quality monitoring capabilities.