Groundwater pollution risk assessment is an important measure to protect groundwater safety. At present, the risk management of contaminated sites in China mainly focuses on soil, while the emphasis on groundwater pollution risk management is insufficient and needs to be strengthened. The risk assessment of contaminated sites mainly adopts the human health risk assessment method. Groundwater is considered as the pollution exposure pathway in the health risk assessment, and the risk caused by the migration and diffusion of pollutants in the vadose zone and aquifer is not concerned.

In this paper, based on the requirements of environmental management of industrial enterprises, industrial sites in China, and adhering to the idea that groundwater protection is as important as human health, and the risks and analyzing the influencing factors of site pollution through vertical infiltration and horizontal diffusion, A multi-level risk assessment method for site groundwater pollution was established. The parameters of the evaluation were researched by the numerical model. Based on environmental management requirements, the concept of time was introduced as a risk early warning classification standard. The groundwater pollution risk assessment and early warning method was applied to a coking site, which provided a reference for practical application.

The main conclusions are as follows :

(1) Based on the analysis of the current demand for groundwater pollution risk management in industrial enterprises and industrial parks in China, a multi-level groundwater pollution risk assessment and early warning method system with groundwater as the first receptor and downstream control points as the second receptor was established for the infiltration process of unsaturated zone pollution and the migration process of saturated zone pollution.

(2) The pollution infiltration model of vadose zone was established by Hydrus-1d, and the influencing factors of the first level vadose zone vertical infiltration pollution risk assessment process were analyzed. The key parameters affecting the assessment results were determined by sensitivity analysis method : groundwater depth (D), organic matter content (f_oc), degradation coefficient (k) and soil bulk density (ρ).

(3) The pollution migration model of saturated zone was established by Visual MODFLOW Flex, and the influencing factors of pollution risk assessment process of horizontal migration of saturated zone at the second level were analyzed.The key parameters affecting the assessment results were determined by sensitivity analysis method : control point distance (L), degradation coefficient (k), permeability coefficient (K) and hydraulic gradient (I).

(4) Taking the site of a coking plant as an example, Hydrus-1d and Visual MODFLOW Flex were respectively used to construct the infiltration model of vadose zone pollution and the migration model of saturated zone pollution. The application of risk assessment and early warning methods for groundwater pollution in the site was carried out, and the suggestions for risk control of site pollution were put forward. The early warning response plan of site risk was preliminarily established, which provided guidance for the risk management of groundwater pollution in the site. In the application process of groundwater pollution risk assessment, two assessment models were adopted, namely the conservative model without degradation and the non-conservative model with degradation. Conservative assessment model has shorter warning time, higher peak concentration, higher risk and higher warning response level.

 (5) Considering the key parameters of vertical or horizontal pollution risk assessment, the stochastic theory was used for uncertainty analysis. A vertical assessment scenario and a horizontal assessment scenario were selected to simulate the uncertainty of key parameters by means of alternative models. Compared with the support vector machine model, the random forest model has the advantages of high precision, high reliability, high stability and low error as an alternative model for vertical or horizontal pollution risk assessment. Uncertainty analysis of key parameters made the results of groundwater risk assessment more reliable.