关键词: 地下水污染风险, 随机森林, 权重优化, 污染风险防控, 浅层地下水, 河北保定

Abstract:  Aiming at the deficiency that it is over subjective to determine index weights in the regional groundwater pollution risk assessment system, this paper takes Shallow groundwater in Baoding plain, Hebei Province as the study area, using random forest algorithm to construct the mapping relationship between groundwater pollution risk assessment indicators and nitrate concentration in order to optimize the weights of each index，and on this basis, the risk of groundwater nitrate pollution in shallow groundwater in Baoding plain is evaluated. The results show that when the number of decision trees is 130, the minimum sample number required for decision tree splitting is 82, and the maximum depth of decision trees is 2, the optimal weight is obtained. The results showed that the depth of groundwater level, types of vadose zone media, the net recharge rate of groundwater, and the ratio of cultivated land area account for larger weights, with weights of 5.00, 4.20, 4.14, and 3.40 respectively. The Spearman correlation coefficient between groundwater nitrate pollution risk assessment results and measured groundwater nitrate concentration has significantly increased based on random forest method, and the prediction accuracy was improved by 25.6%. 

Key words:  , groundwater pollution risk, random forest, weight optimization, pollution risk prevention and control, shallow groundwater, Baoding City, Hebei Province