Regional Groundwater Potential Analysis Using Classification and Regression Trees

Population growth increases the need for food and water, resulting in an increase in water demand around the world. Since groundwater is the main source of consumption in arid and semiarid regions, it is important to understand the groundwater processes in a given watershed. The objective of this study is to use the classification and regression trees (CARTs) algorithm to predict groundwater potential in a semiarid region, Firoozeh watershed, Iran. A total of 11 condition factors, including topographic wetness index, distance to river, slope percent, drainage density, aspect, elevation, land use, lithology, distance from fault, relative slope position (RSP), and topographic position index were employed. Model calibration and validation were conducted based on the random partition in the R software environment. To assess the accuracy of a diagnostic test, relative operating characteristic curve analysis was considered. Sensitivity analysis (SA) was also performed to assess the importance of groundwater conditioning factors. The validation results indicated that the area under the curve value for CART model was 88%. SA showed that the most sensitive factors are RSP, lithology, distance from fault, and distance to river. The findings of the current research can be helpful for decision-makers and managers for sustainable planning, irrigation, and town water supply purposes to achieve water demand goals.