Microplastics identification framework: Integration of microplastic-derived dissolved organic matter fingerprints and...

Microplastics (MPs), ubiquitous in aquatic environments, pose ecological risks that depend fundamentally on their polymer composition. MPs can release dissolved organic matter (DOM) carrying polymer-specific chemical fingerprints, yet their integration into an accessible and interpretable framework for polymer-related MPs identification remains insufficiently developed. Here, we proposed an MP-DOM based MPs identification framework that integrates an interpretable machine learning model with MP-DOM fingerprints derived from aliphatic (polyethylene, polypropylene), aromatic (polystyrene, polyethylene terephthalate), and biodegradable MPs (polylactic acid) under controlled (Milli-Q) and environmental (river water) conditions. MP-DOM fingerprints were identified from dissolved organic carbon (DOC) levels, UV-Visible absorbance and fluorescence indices. Functional group analyses and molecular-scale simulations further supported that these fingerprints reflect polymer-dependent DOM compositions and release pathways. Specifically, aliphatic MP-DOM exhibited weak and uniform optical signatures linked to backbone-controlled release, aromatic MP-DOM showed enhanced aromaticity and a lower polymerization degree due to structure-selective release from aromatic units with higher local electronic activity, whereas biodegradable MP-DOM was characterized by high DOC release, strong bulk absorbance, and low aromaticity associated with ester-bond hydrolysis. After benchmarking multiple classifiers, an optimized random forest model incorporating seven MP-DOM fingerprints achieved high classification performance (AUC=0.953). When applied to MP-added river water samples, the model retained recognizable classification performance after background correction (AUC=0.903), with near-complete identification of biodegradable MPs and residual overlap mainly between aliphatic and aromatic MPs. This study demonstrates the feasibility of MP-DOM fingerprints for polymer-related MPs identification, providing an accessible approach to support source-related interpretation in aquatic environments.