Water Resources and Power

Dissolved oxygen(DO) is a key indicator of water quality health, and its variation is affected by a variety of environmental factors and interactions. Based on the water quality data of river sluice gates in the Zhejiang-Fujian region of Qiantang River from January 2021 to April 2022, the extreme gradient Boost(XGBoost) algorithm was used to build a DO prediction model. The driving effect and interaction effect of water quality indicators on DO were analyzed by considering Shapley value(SHAP). The results show that water temperature(WT) and pH are the main factors affecting DO, contributing 42.7% and 20.9% to the model output. In addition, the nonlinear interaction between WT and pH is particularly significant. Low WT and low pH promote DO, while high WT and high pH may reduce DO concentrations. The interaction between WT and turbidity(TB) significantly reduced the DO concentration, while the appropriate amount of total phosphorus and conductivity had a positive effect on the DO concentration. The results can provide scientific basis for water quality management and ecological protection in the basin.