A Meteorology Based Particulate Matter Prediction Model for Megacity Dhaka

ABSTRACT 

Dhaka, the capital of Bangladesh, is one of the megacities in the world with the worst air quality. In this study, we develop statistical models for predicting particulate matter (PM) concentration in ambient air of Dhaka using meteorological and air quality data from 2002 to 2004 of a continuous air quality monitoring station (CAMS). Model for finer fraction of PM (PM_2.5) explains up to 57% variability of daily PM_2.5 concentration, whereas model for coarser fraction (PM_2.5-10) explains up to 35% of its variability, indicating that PM_2.5 is influenced more by meteorology than PM_2.5-10. Temperature, wind speed, and wind direction account for 94% of total PM_2.5 variability explained by the model, while relative humidity contributes to 75% of total PM_2.5-10 variability. Inclusion of PM lag effect increases models’ predictive power by 4–16%. In general, our developed models show promising performance in capturing the seasonal variability of Dhaka’s PM concentration, although overestimate the low concentrations during wet season (April–September). We validate these models using a recent dataset (2013–2017) from the same monitoring site, in which modeled PM show strong positive correlations with observed concentrations (r = 0.81 and 0.76 for PM_2.5 and PM_2.5-10 respectively). Models also exhibit strong predictive power in forecasting PM levels of two other CAMSs in Dhaka. Thus, the developed models have potentials to explain the temporal and spatial variability of daily PM within Dhaka. These models can be helpful to policymakers as they can predict daily PM at any location of Dhaka with reasonable accuracy if daily meteorological data and previous day’s PM concentration are available. The effect of climate change scenarios on air pollution dynamics of Dhaka can also be assessed using these models.