Groundwater Vulnerability Study Of Hengda Mining Area Based On Machine Learning And Numerical Simulation

With the continuous development of our country’s economy,groundwater pollution is becoming more and more serious.Groundwater vulnerability assessment is one of the prerequisites for groundwater protection.Based on the ground water environmental survey in Fuxin,the study carried out a regional water quality forecast and analyzed and summarized the appropriate vulnerability factors in the region,the special vulnerability assessment model of shallow groundwater in Hengda mining area is established by using numerical simulation and machine learning,and the vulnerability assessment is carried out in different zones.The main findings are as follows:(1)Based on the study of water quality during the dry and wet seasons using multiple models,the results indicate that the RF model has a higher overall accuracy.Combining the spatial distribution maps of various factors and the water quality zoning map analysis,the overall quality of shallow groundwater in this area is poor,with 29%,38%,and 33%of groundwater during the dry season classified as Class III,IV,and V water,concentrated in the southwest and northeast regions,the main contributing factors are nitrate and bicarbonate;during the wet season,33%and67%of groundwater quality is classified as Class IV and V water,respectively,concentrated in most areas of the northeast and south,with the main contributing factors being manganese and bicarbonate.The model validation results show that the model has strong applicability and can meet the actual conditions in the constituency.The fluorine concentration in the vicinity of the mining area in the dry and wet periods is high,and its distribution is regular,which can be used as a factor to evaluate groundwater vulnerability.(2)The simulation results show that the water balance is negative,indicating that groundwater recharge is less than discharge,and the water level is declining.The main sources of recharge for the fourth aquifer system are rainfall and surface water,with recharge rates of 4532.1 m~3/d and2153.9 m~3/d,respectively.The discharge through the river boundary is 6712.8 m~3/d.The limited recharge and discharge rates for the second aquifer system indicate a weak hydraulic connection between the groundwater and the second aquifer system.(3)ROC curve,prediction evaluation index and vulnerability density statistics show that the accuracy and applicability of RF and XGBoost prediction models are obviously superior to SVM.RF and XGBoost fragile zoning as a whole showed gradual change,the vulnerability of the northwest mining area is significantly higher than the southeast,no mutation.The extremely high and high vulnerable areas are mainly distributed in areas around Democratic village and hanjiadian.The main contributing factors are net recharge and groundwater depth,strong aquifer permeability,weak groundwater runoff,large net recharge,low vegetation coverage and water-rock interaction.As a key prevention and control area,monitoring should be strengthened to reduce the risk of groundwater pollution.The paper includes 37 figures,19 tables and 95 references.