Locating groundwater artificial recharge sites using random forest: a case study of Shabestar region, Iran

One of the most crucial design steps to guarantee the long life and the sustainability of artificial recharge of groundwater projects is to find the best locations. The present study focuses on identifying potential zones of groundwater artificial recharge in Shabestar region, northwest of Iran. For this purpose, random forest (RF) model, a learning method based on ensemble decision trees, was proposed for locating groundwater artificial recharge. Important factors, including slope and slope aspect, soil texture, erosion, land use, groundwater quality, permeability and geological lithology were integrated in a geographic information science (GIS). According to RF model, permeability and unsaturated zone thickness were identified as the most effective parameters for locating groundwater artificial recharge sites. Based on the proposed model, it was found that 14% of the region is located in suitable site for groundwater artificial recharge projects. The accuracy of the model was evaluated with receiver-operating characteristic (ROC) curve and the mean squared error (MSE). Low MSE and ROC curve of the model with the highest area under curve equal to 0.947, indicated high accuracy of random forest in locating groundwater artificial recharge.