Predictions of heavy metal concentrations by physiochemical water quality parameters in coastal areas of Yangtze river estuary

Abstract

Due to the degradation-resistant and strong toxicity, heavy metals pose a serious threat to the safety of water environment and aquatic ecology. Rapid acquisition and prediction of heavy metal concentrations are of paramount importance for water resource management and environmental preservation. In this study, heavy metal concentrations (Cr, Ni, Cu, Pb, Zn, Cd) and physicochemical parameters of water quality including Temperature (Temp), pH, Oxygen redox potential (ORP), Dissolved oxygen (DO), Electrical conductivity (EC), Electrical resistivity (RES), Total dissolved solids (TDS), Salinity (SAL), Cyanobacteria (BGA-PE), and turbidity (NTU) were measured at seven stations in the Yangtze river estuary. Principal Component Analysis (PCA) and Spearman correlation analysis were employed to analyze the main factors and sources of heavy metals. Results of PCA revealed that the main sources of Cr, Ni, Zn, and Cd were steel industry wastewater, domestic and industrial sewage, whereas shipping and vessel emissions were typically considered sources of Pb and Cu. Spearman correlation analysis identified Temp, pH, ORP, EC, RES, TDS, and SAL as the key physicochemical parameters of water quality, exhibiting the strongest correlation with heavy metal concentrations in sediment and water samples. Based on these results, multiple linear regression as well as non-linear models (SVM and RF) were constructed for predicting heavy metal concentrations. The results showed that the results of the nonlinear model were more suitable for predicting the concentrations of most heavy metals than the linear model, with average R values of the SVM test set and RF test set being 0.83 and 0.90. The RF model showed better applicability for simulating the concentration of heavy metals along the Yangtze river estuary. It was demonstrated that non-linear research methods provided efficient and accurate predictions of heavy metal concentrations in a simple and rapid manner, thereby offering decision-making support for watershed managers.

Keywords: Heavy metal prediction; Random Forest model; Support vector machine; The coastal areas of Yangtze river estuary.