Application of machine learning and multivariate approaches for assessing microplastic pollution and its associated risks in the urban outdoor environment of Bangladesh

Abstract

Microplastics (MPs) are an emerging global concern due to severe toxicological risks for ecosystems and public health. Therefore, this is the first study in Bangladesh to assess MP pollution and its associated risks for ecosystems and human health in the outdoor urban environment using machine learning and multivariate approaches. The occurrences of MPs in the urban road dust were 52.76 ± 20.24 particles/g with high diversity, where fiber shape (77%), 0.1-0.5 mm size MPs (75%), blue color (26%), and low-density polyethylene (24%) polymer was the dominating MPs category. Pollution load index value (1.28-4.42), showed severe pollution by MPs. Additionally, the contamination factor (1.00-5.02), and Nemerow pollution index (1.38-5.02), indicate moderate to severe MP pollution. The identified polymers based on calculated potential ecological risk (2248.52 ± 1792.79) and polymer hazard index (814.04 ± 346.15) showed very high and high risks, respectively. The occurrences of MPs could effectively be predicted by random forest, and support random vector machine, where EC, salinity, pH, OC, and texture classes were the influencing parameters. Considering the human health aspect, children and adults could be acutely exposed to 19259.68 and 5777.90 MP particles/ year via oral ingestion. Monte-Carlo-based polymers associated cancer risk assessment results indicate moderate risk and high risk for adults and children, respectively, where children were more vulnerable than adults for MP pollution risks. Overall assessment mentioned that Dhaka was the most polluted division among the other divisions.

Keywords: Ecological risks; Machine learning; Microplastics; Pollution indices; Public health; Road dust; Urban environment.