Machine learning-based assessment of microplastic pollution and human dietary risk in four commercial markets fish species from Yichang, China.

Microplastics (MPs) pose an emerging environmental and public health concern particularly, due to their accumulation in aquaculture species for human consumption. Therefore, the current study considered the abundance, characteristics, and human health risk assessment (HHRA) in four commercially available fish species i.e., Dicentrarchus labrax (Sea bass), Carassius Carassius (Crucian carp), Oreochromis niloticus (Nile tilapia), and Hypophathalmichthys nobilis (Bighead carp) sampled from four various local markets in Yichang, China. A proposed protocol by following (10% KOH, 60 °C, 48 h) used for MPs isolation and characterized using micro-Fourier transform infrared spectrometer (μ-FTIR) and scanning electron microscopy (SEM). The results revealed that the highest average MP abundance (52.5 ± 9.4 items/Kg) was found in sea bass with polymer types dominated by polypropylene (PP) and polyethylene (PE), mainly as black colored fibers within 150-5000 μm size ranges. Moreover, a significant positive correlation (R2 = 0.80; p < 0.001) was observed between fish weight and MPs number. Among the machine learning (ML) models, XGBoost and LightGBM exhibited superior predictive performance for MPs accumulation validated by the Taylor diagram. The HHRA also found that estimated annual MPs intake by fish consumption was significantly higher for adult male than for female across all the sampled fish species. In conclusion, sea bass posed the highest exposure risk, followed by Nile tilapia and bighead fish, whereas crucian carp showed the lowest levels of risk to human, which highlights the urgent need for regulatory measurement for the mitigation of MPs contamination in food chain.