Exploring the Link Between Ground Based PM2.5 and Remotedly Sensed Aerosols and Gases Data to Map Fine Particulate Matters in Malaysia Using Machine Learning Algorithms

Malaysia experiences serious air pollution and the poor air quality especially during the south west monsoon greatly affects human health. Nevertheless, data on fine particulate matters are not available sufficiently. In order to increase the availability of PM2.5, this study proposes the use of satellite data and robust statistical techniques. Aerosol optical depth (AOD) from Himawari-8 satellite and atmospheric gases data delivered by Sentinel 5p are recommended to be used to map PM2.5 in Malaysia using machine learning algorithms. Random Forest and Support Vector Regression techniques are able to describe the complex and non-linear relationships between PM2.5, AOD and gases pollutants in Malaysia with high accuracy (R2=0.7). The outcome of the study which describes the spatial and temporal distribution of PM2.5 across Malaysia is significant for health care authorities in Malaysia to analyse the impact of air pollution on human health.