Sources, pattern, and possible health impacts of PM2.5 in the central region of Bangladesh using PMF, SOM, and machine learning techniques

Abstract

Particulate Matter 2.5 (PM2.5) is a major environmental and public health threat in Bangladesh. It is important to explore the relationship between PM2.5 and other variables to mitigate its adverse health impacts. This study aims to understand the sources, patterns, and health impacts of PM2.5 in five central districts of Bangladesh using fourteen variables. These variables have been analyzed by PMF, SOM, Machine Learning, and Multi-regression analysis. This paper has found that PM2.5 is correlated positively with NO2 (0.55), BC (0.45), CH4 (0.38), and NOx (0.22), while correlated negatively with Rainfall (−0.10), CO (−0.33), and SO2 (−0.24). In PMF modeling, the R2 values of settlement density (1.00), SO2 (0.99), DEM (0.94), Rainfall (0.77), NO2 (0.74), and Brickfield (0.66) have found as the most correlated variables. In this study, the dominant variables NO2, CO, Rainfall, O3, AOT, CH4, and BC are found in Factor 1; SO2, settlement density, and DEM are found in Factor 2; and population density and brickfield are found in Factor 3. In SOM mapping, most of the variables are concentrated in the north-eastern, central, and south-eastern parts of the study area. The prediction of PM2.5 using machine learning is significant, showing reasonable R2 for Random Forest (0.85), Extreme gradient boosting (0.81), and Stepwise Linear (0.76). The impact of PM2.5 on child Acute Respiratory Infection (ARI) is significant (p = 0.002, R2 = 0.75); while child mortality is not significant (p = 0.268; R2 = 0.55). These results will be useful for creating and implementing local and regional PM2.5 mitigation plans. Concerned institutions and academia may also use these outputs for reducing health impacts, particularly child mortality and acute respiratory infections.

Keywords

• PM2.5
• ARI
• PMF
• SOM
• Machine learning
• Bangladesh