Distribution, transport and ecological risk prediction of organophosphate esters in China seas based on machine...

Organophosphate esters (OPEs), widely used globally, have been detected in significant amounts in various environmental media, raising concerns about their persistence, bioaccumulation, and associated risks. Traditional sampling and detection methods are time-consuming and labor-intensive, limiting a comprehensive understanding. This study employs Extreme Gradient Boosting (XGBoost) and Light Gradient Boosting Machine (LGBM) models, using 12 feature variables and 463 OPEs concentration data points, to investigate the distribution and ecological risk of total OPEs (T-OPEs), chlorinated alkyl OPEs (Cl-OPEs), and aryl-OPEs in seawater of China Seas. The LGBM model proved optimal for predicting T-OPEs and Cl-OPEs concentrations, with RMSE of 0.48 and 0.46 and R² values of 0.79 and 0.76. XGBoost was superior for aryl-OPEs, yielding an RMSE value of 0.82, and an R² value of 0.87. Analysis revealed complex nonlinear relationships between features and OPEs concentrations. Maps showed higher OPEs pollution in urban agglomerations and estuaries, particularly in summer. The XGBoost model was the best predictor for ecological risks, with most sites categorized as low-risk, and a few as moderate-risk. This study offers valuable data and insights for managing OPEs pollution and ecological risks in the China Seas.