Groundwater potential modeling over the eastern part of Ghana's Northern Region using evidence belief functions and weight of evidence

In arid and semi-arid areas such as Ghana's northern regions, water scarcity is prevalent particularly during the dry seasons (between the middle of November to April) and thus, the yearly demand for groundwater during these periods is very high in these areas. Hence, the delineation of prospective zones of groundwater resource occurrence offers an invaluable response towards the mitigation of the water scarcity issues that arise particularly during the dry seasons. In view of this, the employability capacity of the evidence belief functions (EBF) and weight of evidence (WOE) approaches towards the preparation of groundwater prospectivity models (GPM) have been assessed and compared in the eastern part of Ghana's Northern Region. In carrying out the aforesaid task, multiple groundwater-related geospatial layers comprising drainage density (DD), digital elevation model (DEM), geology, lineament density (LD), slope, soil type, stream power index (SPI), topographic position index (TPI), topographic roughness index (TRI) and topographic wetness index (TWI) sourced from geophysical, geological, geomorphological and remote sensing datasets were used as inputs for both the EBF and WOE models. Inventory data comprising 230 existing locations of productive groundwater boreholes and wells, obtained from historical data and field surveys were applied. 161 of the groundwater inventory data prepared were randomly selected to train and subsequently generate groundwater prospectivity models based on the EBF and WOE approaches. The remaining 69 groundwater inventory data points were used as testing data to determine the efficacy of the GPM produced based on the two aforementioned data integration approaches using the receiver operating characteristics (ROC) curve. From the area under the ROC (AUC) scores obtained, the predictive performance of the GPM based on the EBF and WOE approaches were respectively 0.96 and 0.98. The validation results of the models show the ability of the models to correctly predict the potential of groundwater sources in relation to the geology, with the Bunya sandstone member and the Afram formation towards the eastern part of the study area verified as the most producible lithologies for groundwater occurrence. It is envisaged that the outputs developed in this study would be useful for planners to plan and manage groundwater resources effectively, especially in difficult geologic terranes with limited success of boreholes.