Using an improved XGBoost model to predict and analyze nitrate nitrogen content in groundwater of wetland hyporheic zones
The hyporheic zone in wetlands is an important area for nitrogen cycling in groundwater.The hyporheic zone of Dongting Lake wetlands is taking as the research object,this study explores the influencing factors and mechanisms of nitrogen migration and transformation in groundwater.4 profiles and a total of 16 monitoring wells were set up in the wetland at the entrance of the Xiangjiang River,and groundwater samples were tested and analyzed for one hydrological year.The selected characteristic parameters for the study include redox potential(Eh),dissolved oxygen(DO),water temperature(T),groundwater level(H)and burial depth,pH,and dissolved organic carbon(DOC).An XGBoost machine learning model is established to predict the relative concentration of nitrate nitrogen.The optimal XGBoost prediction model(BO XGBoost)is obtained by using Bayesian Optimization(BO),Sparrow Search Algorithm(SSA),and Particle Swarm Optimization(PSO)algorithms to optimize the hyperparameters of the XGBoost prediction model.Based on this,the SHAP(Shapley Additive exPlans)method is used to analyze the interpretability of the BO-XGBoost model.The research results indicate that the BO-XGBoost model has the best performance,with determination coefficients exceeding 0.90 in both the training and testing sets.The interpretability analysis results and correlation analysis reveal that the impact of factors such as Eh,DO,T,H,pH,and DOC on the nitrate nitrogen content in groundwater in wetland hyporheic zone gradually decreases.
国家科技期刊平台
NSTL
万方数据
