Estimation of ambient outdoor air quality in data poor locations using open access data, a case study of Kampala.

Abstract

Just like air quality monitoring stations, ancillary data are often lacking in developing countries and even where they exist, access to them is poor and their sparse distribution makes them insufficient for reliable estimation of air pollutants, at least at a local scale. This study explored the possibility of using a combination of remotely sensed data and reanalysis data to estimate a spatially continuous PM2.5 surface. A random forest model was trained and used to estimate a spatially continuous surface of PM2.5 using the data for 26th/06/2020. The R-squared, p-value and standard error of the model training were 0.819, 0.000 and 0.029 respectively while those of the model validation were 0.619, 0.04 and 0.147 respectively.