Evaluation of groundwater potential using Dempster-Shafer model and sensitivity analysis of effective factors: A case study of north Khuzestan province

Abstract

In order to optimize the management of groundwater resources, accurate measurement of groundwater potential is necessary. Application of learning models and remote sensing is very encouraging in reducing the costs of groundwater potential measurement. In this research, Dempster-Shafer learning model and remote sensing data are used to evaluate the groundwater potential in the north of Khuzestan. The study area has a semi-arid climate and is mostly covered by quaternary sediments except for the north, which is mountainous and different geological formations are seen there. For this purpose, 9 parameter layers including: lithology, soil type, slope percent, land use, altitude, TWI index, lineament density, river density and distance from the river as the most effective parameters in groundwater potential are prepared. Additionally, 474 wells in high potential areas have been selected as suitable real data for training and testing of the model. In the training phase, the weights of the layers were calculated using Dempster-Shafer model based on real data of 70% of wells then the groundwater potential map was obtained by combining the weighted layers. To confirm the efficiency of the model and the accuracy of the groundwater potential map, the ROC curve is plotted on the basis of 30% of wells that were not used during the training phase. The results show that the final groundwater potential map is 86% accurate. After verifying the efficiency of the model, sensitivity analysis of effective parameters in predicting groundwater potential is performed using map removal sensitivity analysis method. The results show that all of the considered parameters have a positive effect on groundwater potential prediction accuracy. However, lithology, soil type, and slope percent are the most effective on the accuracy of the model.