Abstract
Particulate matter (PM2.5) is one of the major threats to public health, particularly Dhaka City in Bangladesh, frequently cited as one of the worst cities in the World in terms of air quality. This study examines the effects of six environmental (land surface temperature (LST), digital elevation model (DEM), water vapor concentration, wind speed, rainfall, and normalized difference vegetation index (NDVI)) and six economic factors (population density, road density, gross domestic product (GDP), poverty rate level, and percentage of low-income groups in rural and urban setting) on PM2.5 concentration in five industrial cities of Bangladesh using geographically weighted regression modelling (GWR) and machine learning (ML) tools. The mean annual rate of PM2.5 concentration increased by > 42% during 2002–2020 in all cities. Dhaka and Narayanganj districts were affected the most. Goodness-of-fit (R2) was 93% (environmental factors) and 73% (economic factors). Environmental factors: LST (100%) and water vapor concentration (100%) were correlated positively with PM2.5, while DEM (100%), rainfall (83%), NDVI (81%), and wind speed (84%) had a negative relationship at 95% confidence level. β-coefficients of DEM (p < 0.02), LST (p < 0.01), water vapor concentration (p < 0.01), NDVI (p < 0.02), and poverty rate (p < 0.01) were correlated negatively. Moreover, machine learning has extracted a good prediction of PM2.5, ranging the R2 between 0.79 and 0.86%. This study can be replicated in other cities by incorporating socio-economical, local geo-environmental, and meteorological with other air pollutants.