Delineation of groundwater potentials of Sfax region, Tunisia, using fuzzy analytical hierarchy process, frequency ratio, and weights of evidence models

Abstract

Groundwater in semiarid regions is of extreme importance due to limited water resources and increasing population demand. Hence, a better knowledge of aquifer potentialities is required for better management of this precious resource. This study aims to assess the groundwater potential map (GPM) by both statistical methods and geographical information system (GIS) in Sfax region, Tunisia. A number of 11,868 wells in the region were mapped in GIS and divided into two data sets: 8308 wells (70%) were selected in a random way and defined as training, wells while the remaining ones (3560 as 30%) were considered as testing wells for model validation. First, the groundwater conditioning factors, namely altitude, slope, lithology, drainage density, lineament density and land use maps, were evaluated and then processed by the fuzzy analytical hierarchy process (FAHP), frequency ratio (FR), and weights of evidence (WOE) statistical models. Then, the groundwater potential index (GWPI) was generated from an overlap of the weighted and rated of all the conditioning factors for the three methods. The groundwater potential maps were produced in ArcGIS 10.3 for the three models and classified by means of the quantile classification method. As a final step, the receiver operating characteristics (ROC) curves validated these maps. The validation results show from the areas under the curves (AUC) that the WOE model (AUC = 71.4%) performs similarly to the FR model (AUC = 71.1%), and both are slightly better than the FAHP (AUC = 65.1%) model. Thus, the delineation of groundwater potential zones supports the decision makers for the management of the aquifer exploitation. The appropriate groundwater potential map is established for a suitable management and a better planning of the water resources. Preventive works in low potential coastal areas should be conducted in the present and considered for a long-term management. Given the close results of the three adopted methods, it is possible to apply them in other regions with a consideration of their specific characteristics.