Machine learning aided meta-analysis of microplastic polymer composition in global marine environment

Conventional meta-analysis of the literature on marine microplastic monitoring is not effective at producing patterns and trends in microplastic polymer distribution on a global level, mainly due to the limited number of studies considered for the re-analysis. As a solution, for the first time, we adopted a machine learning-based approach to demonstrate the distribution of microplastic polymers in four different compartments (viz: beach sediment, bottom sediment, water column, and biota) of global oceans. Polyethylene (79.9%), Polypropylene (77.2%), and Polyamide (52.3%) were identified as the most abundant polymers in the marine environment. The abundance of microfibres reported by previous studies was found to be underestimated; moreover, biota contained a disproportionately high amount of microfibres. The morphological characteristics of microplastics, the practice of picking up large particles for spectroscopic identification, and the lower resolution of FTIR found to be inducting bias in polymer characterization. Importantly, this work also illustrates the prominent role of ocean currents in the transport of microplastics. In essence, our study proposes machine learning-aided meta-analysis as an effective technique to facilitate the large-scale analysis of microplastic data to help formulate data-driven policies for combating microplastic pollution.