Multi-phase distribution and predictive modeling of polycyclic aromatic hydrocarbons in the East China marginal seas

The East China Marginal Seas (ECMSs), impacted by significant terrestrial inputs, face considerable polycyclic aromatic hydrocarbon (PAH) pollution. However, a comprehensive understanding of PAH multi-phase distribution and robust predictive modeling in this complex region remains limited. This study systematically investigated the concentration of 16 priority PAHs in dissolved, particle, and sediment phases across the ECMSs (Bohai Sea, Yellow Sea, and East China Sea), elucidating their horizontal and vertical distributions. Results revealed distinct patterns, including surface and near-bottom enrichment, strongly influenced by terrestrial emissions, hydrodynamic conditions, and sediment-water interactions. To advance predictive capabilities, we developed a novel main-channel residual dual-channel multilayer perceptron (MCRDC-MLP). This architecture uniquely processes representative PAH congener data and key marine environmental parameters via parallel pathways with subsequent fusion. The MCRDC-MLP demonstrated strong predictive accuracy for total PAH concentrations across dissolved (R² = 0.83), particle (R² = 0.95), and sediment (R² = 0.80) phases. By leveraging only 4-5 selected PAHs and essential oceanographic variables, this model provides a robust and efficient framework for PAH estimation, significantly enhancing marine pollution forecasting, particularly in data-constrained environments.