Using Random Forest to Estimate Low-Cost Sensor PM₂.₅ from MODIS MAIAC AOD in Kampala

Air pollution, particularly, fine particulate matter (PM<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>.<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">5</inf>), poses significant health risks in urban areas of developing countries. However, many such regions lack comprehensive ground-based air quality monitoring networks. This study aims to address this gap by developing a method to estimate daily PM<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2.5</inf> concentrations in Kampala, Uganda, using satellite-derived Aerosol Optical Depth (AOD) data and machine learning techniques. We combined ground-based PM<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2.5</inf> measurements from low-cost sensors at Makerere University with 1 km resolution MODIS MAIAC AOD data for the period January 2022 to December 2023. A Random Forest model was developed to predict PM2.5 concentrations using AOD and auxiliary variables such as columnar water vapour, month, and day of the week. The model demonstrated good performance, achieving an R2 of 0.62 and a correlation of 0.79 on the test dataset. These results are comparable to similar studies in other urban areas, highlighting the potential of this approach for air quality monitoring in data scarce regions. This research contributes to the development of cost-effective air quality monitoring methods and provides valuable information for urban planning and air quality management in Kampala. The approach presented here can be adapted for other urban areas in Uganda and similar contexts, supporting efforts to address the public health challenges associated with urban air pollution in developing countries.