3.1 Metadata
In order to estimate the spatial distribution of groundwater in the central region, airborne magnetic and radiometric datasets and geo-referenced borehole logs of existing successful and unsuccessful boreholes were correlated and analyzed. Airborne magnetic and radiometric datasets were obtained from the Ghana Geological Survey Authority; a topographic map was obtained from the Survey Department. A-30 m resolution digital elevation model for the study area was deduced from Shuttle Radar Topographic Mission (SRTM) data sourced from the U.S. Geological Survey Earth Explorer (www.earthexplorer.usgs.gov). Other evidential layers generated from SRTM include slope, stream order, stream accumulation, stream density, land use/land cover, and soil types. The potential zones for groundwater were proposed using a knowledge-based approach: the Weighted Overlay Model (WOM). Classes of evidential layers generated from relevant geophysical and remote sensing datasets were scored 1 to 9 based on experts’ opinions on how each class of a thematic layer influences the distribution of groundwater in the study area. The most favourable and significant class was assigned a class score of 9, and the least favourable class was assigned a score of 1(e.g., proximity close to a fault or fault intersection was regarded significant as compared to farther proximities). Hence, the class of closest proximity was assigned the highest score, the class of farthest was assigned the lowest score and the intermediate class was assigned scores between 1 and 9.
3.2 Space Radar Topographic Mission (SRTM)
Digital Elevation Model (DEM) and shaded relief maps were generated from the SRTM data on an ArcGIS platform. The SRTM was subjected to various illuminations (0°, 45°, 90°, 135°, 180°, 225°, 270° and 315°) with a constant inclination of 45° in order to generate different relief maps. During this process, structural lineaments that were orthogonal to the light incident whenever the illumination was varied were delineated and analyzed (e.g., fault and fracture zones). The DEM was subjected to hydrologic modelling methods using the ArcGIS extension toolbox to determine the stream flow direction, stream flow accumulation, watersheds and stream order/network. A slope map was also generated.
3.3 Magnetic data
Various lineament enhancement and edge detection filters were applied to the magnetic data to delineate lithological contacts and structural lineaments (e.g., faults, fractures, dykes) that have the potential to influence or control the groundwater system in the study area. Some of the filters include Reduction to the Pole (RTP), Analytic Signal Amplitude (ASA), First Vertical Derivative (1VD) and Tilt Angle Derivative (TDR).
In order to interpret the geology and structures from the magnetic data, the total magnetic intensity map was reduced to the pole using a magnetic inclination of -15, magnetic declination of -5.8, and total intensity of 31756 nT in order to shift the anomalies directly over the spatial location of the sources (Hansen & Pawlowski, 1989; Blakely, 1995; Rajagopalan, 2003). Since Central Region, and for that matter Ghana, lies within low magnetic latitudes (< 20°), interpreting RTP images at low magnetic latitude (inclinations within 20° of the magnetic equator) presents some difficulties
due to the induced magnetization producing a north-south noise in the data (Rajagopalan, 2003). For example, the transform is not able to completely reconstruct NS-trending anomalies of the dykes and regional faults in the area. The Analytical Signal Amplitude (ASA) filter, which is independent of magnetization direction (Nabighian, 1972; Rajagopalan, 1997), was applied to the Total Magnetic Intensity (TMI, Fig. 2a) to give the ASA image (Fig. 2c). Compared to the RTP (Fig. 2b), this gives a better fit of the anomaly-to-source location. The maxima of the anomalies approximately coincided with the Birimian metavolcanic, with the low coinciding with the Birimian metasediment. The edges of the dykes, faults, and lithological contacts (e.g., Tarkwaian Group and Sekondian Group at the NW part of the study) were enhanced by the ASA filter.
The 1VD and TDR filters were applied to the magnetic image to enhance local and subtle fault and dyke anomalies in the study area, as well as isolate diffused anomalies from closely spaced magnetic sources (Blakely, 1995; Cooper and Cowan, 2004, 2006; Ma et al., 2014). The zero values of both filters are placed on the edges of the magnetic sources (Nasuti et al., 2019). These led to an enhanced number of regional structures in the area, 1VD (Fig. 2d) and TDR (Fig. 2e). The TMI image (Fig. 2a) was subjected to the Centre for Exploration Targeting (CET) grid analysis tool in Oasis Montaj platform for magnetic texture analysis and lineament extraction (Holden et al., 2012). The standard deviation, phase symmetry, amplitude threshold, skeletonization, and vectorisation processes guided the extraction of the magnetic lineaments (Fig. 2f). From the Rose diagram in Fig. 2f, it can be seen that most of the structures trend in the NE-SW direction, the main structure trend for Birimian structures (Kesse, 1985; Griffis et al., 2002).
3.4 Radiometric data
Airborne radiometric data was used to ascertain the variation in the concentration of naturally occurring radioelements such as potassium (K) (Fig. 3a), Uranium (eU) (Fig. 3b), and thorium (eTh) (Fig. 3c) in the various lithological units in the area (Anderson & Nash, 1997; Wilford et al., 1997). K (%) is a dominant constituent of most lithology, with mobile eU (ppm) and immobile eTh (ppm) present in trace amounts as mobile and immobile elements, respectively (Elawadi et al., 2004). Derived ratio and ternary maps from the radioelements were used to verify the mapped lithological units, as well as detect weathered zones (Telford et al., 1994; Lo and Pitcher, 1996; Wilford et al., 1997; Elawadi et al., 2004) which have major control on groundwater occurrence.
To delineate K-enrichment zones, high K values related to hydrothermal alteration had to be separated from that of lithology and weathering through the following procedure (de Quadros et al., 2003): First, the deviations or anomalies of measured K were calculated as: \(\:{\text{K}}_{\text{d}}\) = (K map -\(\:{\:\text{K}}_{\text{i}}\)) /\(\:\:{\text{K}}_{\text{i}}\) with \(\:{\text{K}}_{\text{i}}\) = (Kaverage / eThaverage) × eTh map, where the \(\:{\text{K}}_{\text{i}}\) is the ‘ideal thorium-defined potassium’ value for each station having measured thorium values and \(\:{\text{K}}_{\text{d}}\) is the deviations (anomalies) of measured K values from ideal calculated K values (i.e.\(\:\:{\text{K}}_{\text{i}}\)). Secondly, the abundance of measured K related to the ratio of eTh and eU was estimated as F parameter = K map × (eTh map/eU map), where the F parameter defines the abundance of measured K related to eTh/eU ratio. Lastly, a ratio map of K/eTh was generated. The radiometric ternary RGB image (Fig. 3d) was generated to highlight anomalous hydrothermal alteration zones by assigning the following parameters to the following channels: (i) Red channel: \(\:{\text{K}}_{\text{d}}\); (ii) Green channel: F parameter; and (iii) Blue channel: K map/eTh map ratio.
The resulting Kd, F parameter and K/eTh ratio images were then combined in a Principal Component (PC) Analysis, and the high PC1 image values were identified as hydrothermal alteration zones (K-enrichment) (Fig. 4b). The accumulations of K-alteration in the rivers were masked using river channels in the study area. The k-rich alteration zones extracted from the ternary map in Fig. 3d are shown in Fig. 4. These alteration zones overlaying heavily faulted regions were interpreted as prospective zones (IAEA, 2003; Sharma, 2000; Telford, 1994) that are favorable for groundwater accumulation.
3.5 Evidential layers and class scores
In this study, thirteen evidential layers, namely, geology, soil, lineament, lineament intersection, lineament density, lineament intersection density, dyke, alteration zones, land use/cover, digital elevation model, slope, stream flow accumulation and drainage density, were used. The evidential layers were each assigned a weight based on the relevant importance of the layer to groundwater occurrence, accumulation and storage (Table 1). The assigned scores were done such that the total weight scores add up to 100%. Mapped lineaments, which are possible fracture zones and weathered zones, have major control on groundwater occurrence (Joel et al., 2016). Therefore, areas with lineament (Fig. 5a) were given a score of 10. Also, lineament density, which influences the availability of trap zones for infiltration of rainwater for groundwater storage, was given a score of 8 (Fig. 5b). Lineament intersection (Fig. 5c) was given the highest score of 15 because groundwater occurred within developed secondary porosities in crystalline terrain. Furthermore, high groundwater potential areas are associated with areas of high lineament intersections (Reddy and Raju, 1998). Lineament intersection density (Fig. 5d), which is the total length of lineaments per unit area, was given a score of 10. Proximity to mafic dyke (Fig. 5e) was assigned a lower score of 5 since dykes primarily act as barriers to groundwater flow and proximity to alteration zones (Fig. 5f), derived from the airborne radiometric dataset) was given a score of 7. Geology (Fig. 1a) was also given a score of 10 because it plays a significant role in the weathering and fracturing of rock types, similar to lineaments. This makes geology and lineament each to be assigned a score of 10 since both have a significant influence on groundwater availability. Soil types (Fig. 1b) in the study were assigned a score of 8 since the soil types will influence the rate of infiltration. Land use/cover and lineament density were each assigned a score of 8. This is because the soil type will influence the rate of infiltration, and the land use/cover determines to what extent the land has been put to use. Thus, land use/cover (Fig. 6a) is dependent on the soil type underlying the area and was given a score of 8, just like soil types. The digital elevation model (Fig. 6b) was assigned a score of 5. Slope (Fig. 6c) was given a score of 7. Stream flow accumulation (Fig. 6d) and drainage density (Fig. 6e) were assigned scores of 4 and 6, respectively (Table 1). The stream flow accumulation (Fig. 6d) was given the lowest score of 4 because the important of stream does little in the weathering or fracture for groundwater accumulation except for bank storage or when the stream flow tends to recharge a weathered zone or some extensive fracture extending across some section of the stream. The drainage density was given score of 6 because drainage reflects the inability of surface flow to infiltrate into the subsurface environment. Areas of high drainage density are low groundwater potential zones since they have high surface flow and vice versa (Al-Djazouli et al., 2020).
Table 1
Weights assigned to evidential layers used in the GIS-based Weighted Overlay Model (WOM)
No.
|
Evidential layer
|
Weight
|
1
|
Geology
|
10
|
2
|
Soil type
|
8
|
3
|
Proximity to lineament
|
7
|
4
|
Density of lineaments
|
8
|
5
|
Lineament Intersection
|
15
|
6
|
Density of lineament intersection
|
10
|
7
|
Proximity to mafic dyke
|
5
|
8
|
Proximity to alteration zone
|
7
|
9
|
Land use/cover
|
8
|
10
|
Digital elevation model
|
5
|
11
|
Slope
|
7
|
12
|
Flow accumulation
|
4
|
13
|
Drainage Density
|
6
|
The classes of each evidential layer deduced from the geology, magnetic, radiometric and remote sensing datasets (e.g., DEM) were given scores ranging from 1 to 9. Nine was chosen because, for each evidential layer, the maximum number of classes is 9. For each class, the subcategory that has the least contribution to groundwater occurrence was given a score of 1, and the most favourable a score of 9. The intermediaries that lie between the two extremes were given scores between 1 and 9 according to their relative importance to groundwater occurrence. The geology layer was reclassified into 8 classes due to the fact that there are 8 groups/structural units within the study area. The regions of the Sekondian Group, the Tamnean Plutonic Suite and the Eburnean Plutonic Suite were regarded as favourable classes. They were assigned high scores as compared to the overlying Tarkwaian Group and Birimian Supergroup. This is based on the composition and ability of the rocks contained in the group, formation or structural unit to have varied porosities and can undergo weathering. Groups/structural units containing rocks with high porosities and can undergo rapid weathering are given higher scores. On the other hand, groups containing rocks with low porosities and can undergo less weathering state are given lower scores (Shama, 2002; Telford et al., 1994; Blyth & de Fraites, 1984). The intrusive suites serve as an aquitard to the overlying aquifers. High scores were assigned to the land cover layer classes for grass without dispersed trees, moderately closed trees, and moderately dense herbs with scattered trees. These locations are low-elevation and easily accessible.
The soil cover layer was used to identify regions with high water retention capability. The favourable classes were the leptosols, lixisols and acrisols, with high scores of 7, 8, and 9, respectively. Lower values were assigned to other soils with lower porosities. The favourable classes for the DEM and slope layers were the 35–63 m class (lowest elevated region) and the 0-10.45o class (low lying-gentle slope). The stream flow accumulation layer was classified into eight classes, and the favourable class, 128, was assigned the highest score of 9. This is because the 128 class represents the area with the concentrated flow and stream channels. Lower scores were assigned to flow accumulations for 64, 32, 16, 8, 4,2, and 1. The stream order/drainage density layer is defined as a ratio of the total length of all streams to the total area of the basin. The favourable drainage density classes are 0.59–0.69, 0.69–0.82, and 0.82–1.18, with class scores of 9, 8, and 7, respectively. Lower scores were accordingly assigned to less favourable drainage densities. The proximity layers generated were proximity to the faults layer, proximity to the fault intersection layer, proximity to dykes, and proximity to the hydrothermal alteration zones layer. The favourable proximity classes were 100, 200, and 300 m, with 9, 8, and 7 class scores, respectively. Less favourable proximity classes: 400, 500, 1000, 2000, 3000 and > 3000 m was assigned scores of 6, 5, 4, 3, 2 and 1, respectively. The lineament density layers generated were the fault density layer and fault intersection density layer. The proximity and lineament density layers were used to identify regions with favourable permeability. Higher scores were assigned to areas with close proximity, and lower scores were assigned to areas with less proximity. Since dykes serve as barriers to groundwater occurrence (Comte et al., 2017), areas closest to dyke were given lower scores and areas farthest were given the highest score (Table 2.0). The closest proximity to dyke was given a score of 8 instead of 9 since Manso dyke in the study may have some substantial thickness (Antonio et al., 2019).
3.6 Weighted Overlay Model
In order to integrate thirteen (13) evidential layers (proximity to faults, fault density, fault intersection, faults intersection density, dykes, radiometric alteration zones, stream order, stream flow density, stream flow accumulation, geology, land use/ land cover, slope, and digital elevation model) derived from processed airborne magnetic and radiometric datasets and relevant remote sensing data, Weighted Overlay Model in a GIS environment was employed. In such a suitability study, the weighted overlay modelling is the intersection of standardised and differently weighted layers to solve multicriteria problems such as site selection for potential groundwater (Steinel et al., 2016). The assigned weights quantify the relative importance of the evidential layers. The weighted evidential layers were combined through the equation below, which calculates an average weighted output score (\(\:\overline{\text{S}}\)) for each pixel (Bonham-Carter, 1994; Yousefi and Carranza, 2016):
where \(\:{\text{S}}_{\text{i}\text{j}}\) is the class score assigned to each of the jth classes of the ith evidential layer according to their relative importance based on experts’ knowledge of the study area, and \(\:{\text{W}}_{\text{i}}\) is the weight assigned to the evidential layers based on relative importance as compared to other layers. The output score (\(\:\overline{\text{S}}\)) is written to new cells in an output layer. The symbology in the output layer is based on these output scores.
These suitability models of the thirteen evidential layers are superimposed on georeferenced borehole yields to obtain the area's groundwater potential. To validate the groundwater potential map obtained, a set of new geo-referenced boreholes are used to determine the dispersion patterns of borehole yields and the distribution of such yields of boreholes based on categorized yields of boreholes defined within the study.
3.8 Validation of Results
The groundwater potential map was validated using borehole yields in the area to determine the distribution of borehole yields or categories that constitute the study area. Two validations were done using Modified Index of Agreement (Mod., IOA) and Modified Nash-Sutcliffe Error (Mod., NSE) to determine the accuracy of groundwater potential map. According to Krause et al., (2005), the Modified Index of Agreement(dj) and Modified Nash-Sutcliffe Error (Ej) are given as:
where M is the mean, \(\:{O}_{i}\), the observed value, \(\:{P}_{i}\) is the predicted or modeled value and jЄN, ∀=1. The statistical determination of the errors associated with the groundwater potential map obtained is then classified as unsatisfactory, satisfactory, good, and very good for error or modified values of x: x ≤ 0.5, 0.5 < x ≤ 0.65, 0.65 < x ≤ 7.5 and 0.75 ≤ x < 1.0 respectively (Bouslihim et al., 2016). A conclusion is then drawn as to the error associated with using the model and the map as a basis for future groundwater exploration in the study area.