Are we underestimating the driving factors and potential risks of freshwater microplastics from in situ and in silico perspective?

The high loads of heterogeneous microplastics (MPs) in water system sparked the exploration of MPs source and impact in the environment. However, the contributions of driving factors to MPs contamination and the potential risks posed by multidimensional characteristics are still poorly understood. By incorporating in situ investigation with machine learning predictions, this study reported widespread MPs contamination in both textile upstream and receiving watershed in the Yangtze River Delta. The dominant MPs categories were fibers (0.1–0.5 mm in size), transparent in color, and composed of polyethylene terephthalate. These morphological characteristics indicated a conditional fragmentation process, suggesting that larger MPs are more prone to fragmentation. Multivariable analysis revealed significant correlations between MPs occurrence and factors of metal concentrations, geographic locations, and water qualities, highlighting the roles of textile production and automotive tire wear in determining MPs abundance. Among five machine learning models, Random Forest outperformed others in predicting MPs abundance. The interpretable analysis indicated that longitude (35.3 %), TN (13.8 %) and Sb (13.4 %) were pivotal nodes in shaping the MPs abundance. Emission point sources from express, autotire and textile yield feature importance from 6.60 % to 7.88 %. A total 12.39 % of the predicted variability can be further explained by interaction effects. Besides, MPERI and MultiMP indices based on abundance, size, color, shape, and polymer distributions suggested that most sampling sites fell within moderate to high-risk categories. Artificial neural network-based assessment results are suitable for explaining the MPs induced risks and polymer type was the most influential variable in determining the risk values. These quantitative insights into the driving factors and potential risks behind MPs occurrence improve our knowledge to manage MPs pollution in large-scale watersheds, providing crucial information for the development of effective mitigation strategies.