Estimation of particulate matter from aerosol optical depth using collocated low-cost sensors and AERONET station in Lubango, Angola

Air pollution dynamics are still subject to many knowledge gaps, despite being recognized as a serious public health threat around the globe. In Africa, the existence of data is often rare and sparse, forcing the use of as many alternatives as possible, including remotely sensed data, reanalysis, and modeling approaches to fill the gaps. This study developed empirical models, linear, multivariate, and Bayesian regression models for particulate matter (PM) estimation at three sizes (1, 2.5, and 10 µg/m3) from an Aerosol Optical Depth data source at different wavelengths (340, 380, 440, and 500 nm). The observed root mean square error (RMSE) for the linear regression was 7.94, 18.10, and 22.11 µg/m3, while the Bayesian showed errors of 7.86, 17.84, and 21.77 µg/m3 for PM1, PM2.5, and PM10, respectively. The three-parameter Bayesian regression model showed better performance than the four-parameter multivariate regression model. This study has developed empirical models for the estimation of particulate matter using different wavelengths of aerosol optical depth. The result obtained can help local and national authorities to monitor air pollution using AOD measurement.