Tracing Microplastic\nAging Processes Using Multimodal\nDeep Learning: A Predictive Model for Enhanced Traceability

The aging process of microplastics (MPs) affects their\nsurface\nphysicochemical properties, thereby influencing their behaviors in\nreleasing harmful chemicals, adsorption of organic contaminants, sinking,\nand more. Understanding the aging process is crucial for evaluating\nMPs’ environmental behaviors and risks, but tracing the aging\nprocess remains challenging. Here, we propose a multimodal deep learning\nmodel to trace typical aging factors of aged MPs based on MPs’\nphysicochemical characteristics. A total of 1353 surface morphology\nimages and 1353 Fourier transform infrared spectroscopy spectra were\nachieved from 130 aged MPs undergoing different aging processes, demonstrating\nthat physicochemical properties of aged MPs vary from aging processes.\nThe multimodal deep learning model achieved an accuracy of 93% in\npredicting the major aging factors of aged MPs. The multimodal deep\nlearning model improves the model’s accuracy by approximately\n5–20% and reduces prediction bias compared to the single-modal\nmodel. In practice, the established model was performed to predict\nthe major aging factors of naturally aged MPs collected from typical\nenvironment matrices. The prediction results aligned with the aging\nconditions of specific environments, as reported in previous studies.\nOur findings provide new insights into tracing and understanding the\nplastic aging process, contributing more accurately to the environmental\nrisk assessment of aged MPs.