Machine Learning-Assisted Insights into Sources and Fate of Microplastics in Wastewater Treatment Plants

Wastewater treatment plants (WWTPs) converge multiple sourced microplastics (MPs) and serve as a temporary repository in the case of releasing them into the environment. The process involves two critical scientific problems, including the source composition of MPs and their fate in WWTPs. Therefore, this study conducted a full-scale investigation in each stage of four WWTPs in Hong Kong, with the results showing that the fate of MPs was mainly affected by their physicochemical characteristics and WWTP parameters. Moreover, three conventional machine learning (ML) methods, namely the multilabel decision tree, random forests, and support vector machine, were also applied for figuring out the source compositions of MPs. The results demonstrated that the sources of MPs were mainly composed of domestic (57.3-59.9%), industrial (21.1-21.7%), coastal (11.2-12.7%), domestic/medical (4.6-5.1%), and domestic/agricultural (2.6-3.8%) sources, respectively. The discovery of domestic/medical-sourced MPs should draw the public's attention to the insufficient management of used face masks. This study was a novel attempt to utilize ML to explore the fate and sources of MPs in environmental compartments, which provided new insights into developing the MP source tracing approaches from the source management of plastic contaminants.