Drastic model application to groundwater vulnerability elucidation for decision making: the case of south western coastal basin, Ghana

Abstract

The health of the population is intertwined with the availability and supply of potable water. With many people around the world resorting to groundwater resources as a source potable water supply, it becomes imperative that the state of groundwater susceptibility to contamination is known. The DRASTIC model has been applied to determine the state of aquifer vulnerability to contaminants. The seven hydrogeological parameters were considered in the standard model and the resulting vulnerability, DI were classified into three vulnerability zones; low, moderate and high. The model was then modified to include land use/land cover (LU) and lineament density (LIN) parameters and net recharge (rm). The modified vulnerability indices were designated as DI-LU, DI-LIN, DIrm and MDI (combined modification). The results revealed the land area under high vulnerability under DI, DI-LU, DI-LIN, DIrm and MDI were 32.2, 26.2, 33.7, 26.2 and 40.4% respectively. There were no significant difference between the standard moderate vulnerability class and the modified derivatives (DI + LU, DI + Lin). The combined modification (MDI) however led to 17% decrease in this class of vulnerability. The model was validated using a nitrate concentration of 36 samples collected for the purpose. The validation assessment revealed that the performance of the model was improved with modification.

Introduction

Groundwater is chief source of freshwater reserve in the world. Increasingly it’s becoming a source of choice in several portions of the world. It has found use in domestic, agriculture and industry for purposes such as drinking, irrigation, and manufacturing. It is estimated that 50% of potable water used in France, 70% in China and 45% in Ghana come from groundwater, (Buamah et al. 2008; Tai et al. 2012). This is partly due to the contamination of surface water. Furthermore, the need for freshwater supply sources is likely to rise due to population pressures. This phenomenon will be profound in the third world where physical development outpaces water supply infrastructure. People would turn to groundwater because the saturated zone houses twenty-one percent (21%) of the world total freshwater and about ninety-seven (97%) of the world unfrozen freshwater (Dunne and Leopold 1978). Thus, groundwater resource is of enormous importance and need to be protected for human kind. The first step towards groundwater protection is to map areas of vulnerability where the resource could be easily contaminated.

Vulnerability assessment methods are developed to estimate aquifer vulnerability. These include the overlay and index method, the process-based method and the statistical methods (Tesoriero et al. 1998). The index and overlay method include factors that control the migration of contaminant from the ground surface to the zone of saturation. This results in different locations having different vulnerability. The overlay and index method have the advantage that large areas could be covered because of readily available data (Thapinta and Hudak 2003). A major setback of this method is the subjectivity inherent in assigning weights and ratings to the input parameters. The statistical methods involve the use of statistics to evaluate the association between the actual occurrence of a contaminant in groundwater and the spatial variables. This method has a limitation in that there is often not enough available and accurate water quality data (Babiker et al. 2005). On the other hand, the process-based method employs simulation models to determine the migration of contaminant. The method is however constrained by data availability (Barbash and Resek 1996).

Vulnerability assessment has been carried in many places by many researchers employing different models. So far the DRASTIC model appears to be the most widely used model. This model has been applied to delineate vulnerability of aquifer systems in Africa (Saidi et al. 2011; Ojuri and Bankole 2013; Ouedraogo et al. 2016; Hamza et al. 2017; Lahjouj et al. 2020; Omotola et al. 2020), Australasia (Hasiniaina et al. 2010; Wang et al. 2012; Shirazi et al. 2013; Neshat et al. 2014a; Hussain et al. 2017; Pokhriyal et al. 2020), Europe (Gogu et al. 2003; Albuquerque et al. 2013; Sener and Davraz 2013; Zouhri and Armand 2019), Persia (Chitsazan and Akhtari 2009; Awawdeh and Jaradat 2010; Kattaa et al. 2010; Hassan et al. 2012; Nazzal et al. 2019), Northern America (Rundquist et al. 1991; Denny et al. 2007) and South America (Tovar and Rodríguez 2004; Herlinger et al. 2007). In Ghana the DRASTIC model has been applied to assess aquifer vulnerability of some major basins (Anornu et al. 2012; Agyare et al. 2017) and aquifer systems of densely populated urban areas (Ewusi et al. 2016). Other studies have involved an extensive review of the DRASTIC model (Shirazi et al. 2012) and a combination of models (Gogu et al. 2003; Bordbar et al. 2019; Lad et al. 2019) for aquifer vulnerability.

The study and measurement of vulnerability indices are important in groundwater resource management. Vulnerability assessment is the first step taken in the management and protection of groundwater resources. It is usually undertaken to determine which areas of an aquifer system have greater possibility of contamination because of human activity at the surface. Once ascertained, the delineated areas then become a subject of effective monitoring and targeted for proper land use to safeguard groundwater resource.

The thrust of this study, therefore, is to identify the locations within the study area where groundwater could be vulnerable to contamination by human activities using the DRASTIC model and modified versions and also to elicit the impact of lineament density on vulnerability. The outcome of this work is expected to providing information that leads to the protection of groundwater resource and thus ensures good health.

Study area

The area of study for this research is Sekondi-Takoradi Metropolis, which forms part of the coastal basin of Ghana. It lies within the coordinates latitudes 4° 52.6′–4° 59.5′ N and longitudes 1° 40.2′–1° 49.4′ W. The area has a population of 559,548 with a growth rate of 2.95% (GSS 2010). The climate is described as tropical monsoon (Obuobie et al. 2018) with a mean temperature of 22 °C spanning January and March. Rainfall is bimodal with the main period happening between March and July and a minor period happening within August and November. The monthly climatic conditions are presented in Fig. 1.

Fig. 1

Monthly weather of study area (courtesy Ghana Meteorological Agency)

The area has a diverse landscape with the coast line having bays and capes, which is mainly eroded. The middle part of the Metropolis has a low altitude lies approximately 6 m beneath the sea level. Three main types of vegetation; the tropical forest, savannah woodland and the mangrove exist in the area. The area is characterized by faulted shale and sandstone resting on a solid subterranean vault of schist, gneiss and granite. The early cretaceous to late Ordovician Sekondian Group overlies some igneous rocks, which serve as a basement rock (Fig. 2). Provenance studies suggest that the Sekondian Group mainly originated from the Birimian granitoids (Asiedu et al. 2005). Palaeontological evidence, sedimentary structures and textures the surroundings of deposition of the group is deduced as non-marine to coastal marine (Asiedu et al. 2000; Crow 1952).

Fig. 2

Simplified geological map of Sekondian Group of study area (after Atta-Peters 2000)

Materials and methods

The DRASTIC model

The DRASTIC method is applied in the present study, for estimating potential groundwater vulnerability to contamination. The acronym DRASTIC relates to the initials of the seven parameters that defines vulnerability according to Aller et al. (1987) and presented in Table 1.

Table 1 Ratings and weighting factors used in the DRASTIC model (Aller et al. 1987; Awawdeh and Jaradat 2010)

The DRASTIC vulnerability index was computed using the linear summation of the discrete products (rating × weight of corresponding parameter), employing the equation below:

$${\text{DI}} = \mathop \sum \limits_{i = 1}^{7} \left( {Wi \times Ri} \right),$$

(1)

where DI represents the DRASTIC index; D, R, A, S, T, I, and C denotes the seven parameters, according to Table 1; and R, i, and W are the resultant rating for grid cell i and associated weights.

The weights denote the comparative significance of each DRASTIC parameter in connection with the other parameters. In addition, each DRASTIC parameter is assigned a rating varying between 1 and 10. These ratings were determined according to the intrinsic hydrogeological conditions in the study area. An in-depth understanding of hydrogeology and geology of the study area was considered a requirement to establish the rating of the parameters. As much as possible, the ratings designated in the current study were similar to the characteristic ratings suggested in the original DRASTIC study (Aller et al. 1987).

DRASTIC model modification

In this study, four modifications were done to the original model. The first involves the use of the “effective weight” instead of the assigned weights and further reclassification done. The second involved adding land-use/land cover (LULC) parameter to elicit influence of anthropogenic impact on the groundwater susceptibility to the potential contaminant. LULC directly influence the amount of water that penetrates the soil zone and also as runoff. When the land cover is impervious such as roads, pavement, roof tops, et cetera, there is more runoff to natural reservoirs. On the contrary, surfaces such as vegetation covers retain water for a longer time and allow percolation to occur. The third modification involved the introduction of another parameter; lineament density. The inclusion of the lineament parameter is to enable a further precise valuation of the groundwater potential to contamination. This is because lineament is associated with the flow of groundwater and, therefore, contaminant migration.

The introduced parameters were applied to modify the DRASTIC vulnerability index according to Eq. (2) below;

$${\text{MDI}} = {\text{DI }} + {\text{ L}}_{w} {\text{L}}_{ri} + {\text{ Lin}}_{w} {\text{Lin}}_{ri} ,$$

(2)

where “MDI is the modified DRASTIC vulnerability index, L and Lin are the land use and lineament parameters and subscripts r, i, and w were the subsequent rating for grid cell i and weights”. After modification reclassification was then done. The fourth modification involved the use of land use/land cover and lineament density in computing the net recharge.

Data acquisition and database compilation

The data used for the study came in various sources and form and are summarized in Table 2. GIS database for the geology, hydrogeology, groundwater recharge and soil was created for the area. The data were processed with ArcGIS 10.3.

Table 2 Data source for the model

Development of the DRASTIC parameters

The methodology employed to develop the intrinsic/standard groundwater vulnerability map was based on the original work put forward by Aller et al. (1987). Each parameter processed and used to develop the model is defined beneath.

Depth of groundwater (D)

The 'Depth to water table' (D), is the vertical distance from the surface of the land to the topmost of the zone of saturation. It gives indication of distance a probable pollutant has to traverse before touching the aquifer. Therefore, D has an effect on the extent of interactions between the infiltrating pollutant and surface and sub-surface materials such as water, minerals and air and thus, on the quantity and magnitude of the chemical and physical changes as well as the process of degradation (Rahman 2008). Generally, the susceptibility to contamination reduces as D increases. The D was determined from field measurement of drill wells. Applying the inverse distance weighting interpolation technique to the data, the depth to water layer was created. Ratings of 9 were assigned for depth (1.5–4.6 m), 7 to depth (4.6–9.1 m), et cetera according to Table 1 and five categories were delineated. The D index (rating factor × weighting factor) varied from 10 (the least vulnerable) in the northeastern parts to 45 (the most vulnerable) towards the south-west.

Net recharge (R)

The 'Net recharge', R, indicates the magnitude of water infiltrating the ground surface per unit area of land to access the water table. The amount of water reaching the saturated zone is influenced by factors such as soil permeability, the quantity of surface cover, the surface land sloping and the quantity of water reaching the aquifer. Diffusion and attenuation of pollutant depend to a large extent on the quantity of water available in both unsaturated and saturated zone and, therefore, on the net recharge. High recharge areas are more susceptible compared to areas of low recharge. The net recharge was determined using Piscopo method shown in Table 3 applying Eq. (3) as described in (Awawdeh and Jaradat 2010).

$${\text{Recharge }} = {\text{ rainfall factor}} + {\text{slope factor}} + {\text{soil permeability factor}}{.}$$

(3)

Table 3 Net recharge generated from Eq. (3) (Piscopo 2001; Awawdeh and Jaradat 2010)

Data for this parameter came from the Meteorological Agency, for rainfall, Soil Map of the World by FAO-UNESCO for permeability and the digital elevation model from SRTM 90m for topography.

In assigning the ratings, factors were awarded according to the topographical map, soil map. A rating of 4 was assigned based on the mean annual rainfall of 1077 mm/year. A factor of 5 was assigned to the soil permeability. The layer was generated and the net recharge was calculated using Eq. 3. The R index varied from 32 (the least vulnerable) to 40 (the most vulnerable) and reclassified into three: 37–40, 34–36 and 32–34. Ratings of 10, 9 and 8 were assigned, respectively.

Aquifer media (A)

The 'Aquifer media', A, denotes the kind of unconsolidated or consolidated material that host the aquifer. The vulnerability of aquifer increases as the size of grain and the fractures or openings within the aquifer increases. Aquifer media was inferred from well logs data from the Community Water and Sanitation Agency (CWSA). Based on this information two types of media were classified; sandy, assigned a rating of 6 and moderately weathered sandstone, rated 4. The A index varied from 12 (the least vulnerable) to 18 (the most vulnerable). The aquifer media layer was generated and classified accordingly.

Soil media (S)

Soil is the foremost media the contaminant penetrates as it permeates into the ground. Soil has a major effect on the quantity of recharge that can penetrate the ground, and therefore on the capability of contaminant to travel vertically into the unsaturated zone. Soil media was inferred from well logs data and Soil Map of the World by FAO-UNESCO. The soils were identified as sandy. Two soil types were delineated: the Orthic Acrrisols (Ao1) and Ferric Acrisols (Af1) making up 30.7% and 69.3%, respectively. The Af1 is composed of 81 and 75% sandy topsoil and subsoil, respectively, whilst the Ao1 is composed of 82 and 68% sandy topsoil and subsoil, respectively. Ratings of 9 and 8 were thus assigned, respectively, and the soil media layer map generated.

Topography (T)

The 'Topography', T, regulates the quantity of percolation and runoff capability of surface water into the soil, and thus the capability to initiate contaminants into the soil. Steep slopes engender more runoff and therefore low groundwater contamination risk. Flat surfaces on the other hand is inclined to hold water over a longer period of time and consequently increases the possibility for contaminants movement. The T was inferred from the 90 m Shuttle Radar Topography Mission (SRTM90) database. The values for the slope were generated with the SRTM90 by employing the Spatial Analyst Software within the ArcGIS10.3. The T index ranged from 1 (the least vulnerable) to 10 for the most vulnerable. Five categories were classified and ratings assigned according to Table 1.

Impact of the vadose zone (I)

The function of the vadose zone is incorporated in the ‘I’ parameter. This zone is crucial in estimating the vulnerability of aquifers because of its ability to affect residence time of contaminants and thus the attenuation potential. The impact of the vadose zone was inferred from well log data. The I index ranged from 20 to 30 with two categories identified and classified with assigned ratings of 4 and 6 for moderately weathered argillite and sandstone respectively.

Hydraulic conductivity (C)

The 'Hydraulic conductivity' is the extent of the capability of the aquifer to transfer water under the influence of hydraulic gradient. It measures the speed of movement of contaminants, their residence time and therefore dilution ability. High conductivity values indicate greater risks to contamination exist (Rahman 2008). The thematic layer for the hydraulic conductivity map was “inferred from the global hydrogeological map of permeability and porosity (Gleeson et al. 2011). The global permeability map is given in log permeability (log k)”. The values were converted to hydraulic conductivity, K, as follows:

$$K = k \times \rho \times \frac{{\text{g}}}{\mu },$$

(4)

where K (m/s) is hydraulic conductivity and is dependents on fluid density, $\rho$ (kg/m³) is the density of the fluid, usually for water = 999.97 kg/m³, g (m/s²) is the acceleration due to gravity = 9.8 m/s², and $\mu$ (kg/m s or Pa s) is the viscosity of the fluid, usually for water = 0.001 kg/ms. Hydraulic conductivity ranged between 12.28 and 60.82 m/day. Ratings of 4, 6 and 8 were assigned. The C index ranged from the value of 12–24 and three categories were classified according to Table 1.

Land use/land cover (LULC)

Land cover connotes the physical materials such as crops, grass, land, forest and water that covers the land surface. Land use in contrast connotes the use of land for human activities.

The land-use/land cover (LULC) parameter was included to elicit the influence of human activity on groundwater susceptibility to the potential contaminant. The LULC map was inferred from a geo-referenced Google Earth Pro image and classified accordingly into five categories according to Table 1.

Lineament density (Lin)

The lineament parameter was included to enable a more precise assessment of the groundwater potential to contamination due to its association with the flow of groundwater and thus contaminant migration. Lineament was inferred from a Landsat Satellite image downloaded from USGS Earth Explorer. Using the PCI Geomatica Banff software, lineament lines were extracted and line density done to calculate line magnitude per unit area to produce a raster map. Lineament classification was done according to Table 1.

Sensitivity analysis

A fundamental benefit of the DRASTIC model has been its ability to incorporate the quantity of input data layers (Evans and Myers 1990). It is established that altering the quantity of data layers decreases possible uncertainties or error effects of individual parameters on the final output (Rosen 1994; Panagopoulos et al. 2006). A couple of researchers opined that the vulnerability estimation of groundwater can be evaluated without involving wholly the DRASTIC model factors (Merchant 1994; Li and Merchant 2013). Other researchers hold a contrary view (Napolitano and Fabbri 1996; Ghosh and Kanchan 2016; Hasiniaina et al. 2010). To this end, an assessment of sensitivity analysis to provide evidence on the impact of the assigned rating and weights to each parameter was performed on the model. This included the single parameter and map removal sensitivity analyses.

Single parameter sensitivity analysis

The single parameter sensitivity assessment put forward by Napolitano and Fabbri (1996) was conducted in this current study. This single-parameter sensitivity assessment was established to measure the effect of each DRASTIC parameter upon the vulnerability index. This assessment permits comparison of the “effective” weight to the “theoretical” weight (Babiker et al. 2005). The “effective” weight of each parameter is calculated applying the equation that following:

$$Wi = \frac{{P_{ri} \times P_{w} }}{{{\text{DI}}}} \times 100,$$

(5)

where “Wi refers to the “effective” weight of each parameter, P_r,i and P_w are the ratings and the weights of each parameter, respectively, DI is the overall vulnerability index”.

Map removal sensitivity analysis

The map removal sensitivity assessment put forward by Lodwick et al. (1990), was conducted in this current study. This was performed to corroborate the significance of each parameter employed in the DRASTIC model. This analysis identifies the sensitivity of the overall vulnerability map regarding the removal of one or more thematic map from the initial analysis. It is determined by using the equation;

$${\text{Si}} = \left( {\frac{{|\left( {{\text{Vi}}/N} \right) - (V^{^{\prime}} /n)|}}{{{\text{Vi}}}}} \right)*100$$

(6)

where Si is the sensitivity index, Vi is the vulnerability index computed using Eq. 1 (the unperturbed vulnerability), V′ is the vulnerability obtained by removing one map (perturbed vulnerability), N is the number of parameters used in computing Vi and n is the number of parameters used in computing V′. The variation index is also used to give a measure of the sensitivity when one parameter is removed at a time and the resulting effect compared to the initial vulnerability index. The variation index has been widely applied (Rahman 2008; Li and Merchant 2013; Ghosh and Kanchan 2016; Ouedraogo et al. 2016; Agyare et al. 2017). The variation index is computed using the following equation;

$${\text{Vai}} = \left( {\frac{{\left( {{\text{Vi}}} \right) - (V^{^{\prime}} )}}{{{\text{Vi}}}}} \right)*100,$$

(7)

where Vai is the variation index, Vi and V′ are unperturbed and perturbed DRASTIC vulnerability index, respectively.

Results and discussion

Spatial groundwater vulnerability distribution based on the DRASTIC model

The classification matrix for DRASTIC index model using Table 1 identifies three classes of intrinsic vulnerability towards pollutants as the DI varied between 116 and 168. In terms of spatial distribution, only three percent (3%) of the area was assessed to have low vulnerability towards pollutants. A larger proportion of the area (65%) was assessed to have moderate vulnerability towards pollutants whilst 32% has high vulnerability; Table 4. The rated maps used for the computation of the DRASTIC vulnerability index are exhibited in Fig. 3 and the resulting DI map in Fig. 4.

Table 4 Classification for DRASTIC index model

Fig. 3

Thematic maps used for the overlay and development of DI

Fig. 4

Standard vulnerability map based on the DRASTIC model

Sensitivity analysis

Single parameter sensitivity analysis

The measurement of single parameter sensitivity analysis has been established to measure the effect of each of the DRASTIC model parameters on the overall vulnerability index. This is needed to contrast the ‘theoretical’ weight designated to each of the input parameters with their ‘effective’ weight that is designated by the analytical model. The results of this analysis are presented in Table 5

Table 5 Statistical summary of single parameter sensitivity analysis

The deviations of the percentage ‘effective’ weight from their theoretical could be observed on the able. It is observed that soil media and depth to water had over thirty percentage point changes in their mean effective weights. These two parameters (mean percentage ‘effective’ weight 12.11 and 28.28 against a theoretical weight of 8.70 and 21.74 respectively) are thus considered the most effective parameters influencing groundwater vulnerability index. The necessity to acquire detailed, correct and representative information on these parameters for vulnerability index measurement is thus imperative.

Net recharge exhibited near conservation of its influence on the vulnerability index with only 3.43 percentage point higher ‘effective’ weight than its theoretical weight of 17.39. The near conservation of the recharge parameter, R, underscores the dynamic equilibrium that exists between the phreatic aquifers and that groundwater is replenished in parts by some other means other than percolation. This assertion agrees with previous studies (Hasiniaina et al. 2010; Hassan et al. 2012). The four remaining parameters T, I, C and A exhibited smaller  “effective weights” relative to their “theoretical weights” and are, therefore, considered less influential as input parameters for model evaluation.

Map removal sensitivity analysis

The variations in the sensitivity of the DRASTIC vulnerability index by removing one parameter layer at a time is presented in Tables 6 and 7. The mean sensitivity index varied from the least value of 0.37 (S layer) to the highest value of 2.38% (D layer).

Table 6 Statistical summary of map removal sensitivity analysis

Table 7 Statistical summary of multiple map removal sensitivity analysis

The sensitivity of the final DRASTIC vulnerability index when a parameter layer or more is removed is presented in Table 7. The parameter causing the least sensitivity on the overall index is removed first. The removal of the S layer led to the least variation in the sensitivity index with a mean value of 0.37%. The largest sensitivity index variation was due to the removal of the depth to water layer (D) with a mean value of 5.61. The differences in percentage change in the average sensitivity that results from the removal of a layer indicated that the seven hydrogeological parameters are all essential in computing the DRASTIC vulnerability index.

The removal of a parameter results in the variation of the DRASTIC vulnerability index. This variation is captured in the map removal sensitivity analysis presented in Table 8. The results indicated that depth to water has a profound effect on the final DRASTIC vulnerability index. The removal of the D parameter resulted in a 28.03% average change in the variation index. The impact of the vadose zone layer (I) and the net recharge layer (R) also impacted the DRASTIC vulnerability index by 17.88 and 17.86% respectively. These three-parameter layers (D, I and R) accounted for 63.75% of the total variation. The greater contribution of these three parameter layers could be attributed to the high weightage (D = I = 5, R = 4) associated with the parameters.

The parameters soil media (S), aquifer media (A) and hydraulic conductivity (C) had a moderate impact on the DRASTIC vulnerability with a variation index of 12.07, 11.08 and 10.96% respectively. Topography impacted the least in terms of variability in the model and could be attributable to the low weight (T = 1) assign to it by the model. An average cumulative variation index of 99.40% was observed due to randomness.

Table 8 Statistical summary of map removal sensitivity analysis

Summary of the DRASTIC parameters

This analysis has been undertaken to determine which of the seven parameters used in the model has the greatest influence on the intrinsic vulnerability index. The statistical summary is presented in Table 9.

Table 9 Summary statistics of the DRASTIC parameters

An examination of the means (and medians) of the analysed parameters indicate that “soil media” and “depth to water” has the topmost impact on the intrinsic vulnerability of groundwater as they have the highest mean rate of 8.93 (9.00) and 8.43 (9.00), respectively. Thus these two parameters: S and D has the highest impact on potential groundwater contamination risk. The parameters net recharge, aquifer media, hydraulic conductivity and impact of the vadose zone with means values 6.78, 5.48, 5.43 and 5.31 respectively have a moderate impact on intrinsic aquifer vulnerability whilst topography has the lowest contribution to groundwater pollution potential.

With respect to contribution to variations of the vulnerability index, the coefficient of variation is an important indicator. Accordingly, (Table 9) the topography (83.04%) indicated that a small change in its value could alter significantly the value of the vulnerability index. Moderate contributions are expected from net recharge (39.01), hydraulic conductivity (33.80) and impact of vadose zone (26.38). The least impact to variation of the vulnerability index is caused by changes in the soil media (2.97).

The statistical analysis outcome of the sensitivity analyses indicated that the depth to water and soil parameter were the important parameters significantly controlling and impacting on the susceptibility of the groundwater to pollution. This assertion is consistent with groundwater originating from unconfined or semiconfined aquifers (Napolitano and Fabbri 1996).

Outcome of DRASTIC modification

Sensitivity analysis on the DRASTIC model

When computing the vulnerability of the aquifer system using the effective weights, a new pattern emerged (Fig. 5a). The DI values changed from 116–168 to 89.2–109.9. The entire system is thus classified as having low vulnerability according to the original classification put forward by Aller et al. (1987). The new DI was reclassified at the equal interval (natural breaks) to obtain Fig. 5b

Fig. 5

Maps of DI based on sensitivity analysis (a) and reclassified at an equal interval (b)

There was an effort to determine whether a relation existed between the percentage change in weight (theoretical versus effective) and the percentage coefficient of variation. This is significant because whilst percentage changes in weight results in the reduction of subjectivity in the DRASTIC model, the percentage coefficient of variation indicates the parameter that would tremendously alter the DI. The outcome indicated a strong negative correlation between the two indicators (r² = − 0.74) and is presented in Fig. 6. The S parameter caused the least variation in DI whereas the T parameter greatest variation in DI. On the other hand, whilst the S is considered the most effective impacting DI, the T is the least effective parameter. It is thus concluded that the most effective parameters cause the least variation in the DI.

Fig. 6

Relationship between the two sensitivity factors

Effect of land use on the DRASTIC model vulnerability index

The combination of conventional DRASTIC and land use/land cover is designated DI-LULC. The land use/land cover classification results are presented in Table 10. The area is essentially a built-up class with some natural forest predominantly prevalent at the northern frontiers making the second large class.

Table 10 Types and areas of land use

The inclusion of LULC in mapping out the aquifer susceptibility to potential contamination is important because human activities such as mining, farming, industrial, urbanization could alter the hydrogeological conditions. The resulting modified and classified map is presented below together with the LULC map (Fig. 7).

Fig. 7

Maps of LULC (a) and modified DI (b) based on LULC

The resulting modification apportioned 13.4% of land area to low vulnerability compared to 3% of DI. This is very significant and important for future land use planning. The land area classified as a moderate and high vulnerability on the other hand witnessed a reduction in their magnitude (Fig. 8). The increase in the low vulnerability class and the decrease in the moderate class were consistent with other results (Shirazi et al. 2012). Forest and vegetation cover appears to be the main drivers of shifts in the vulnerability classes. This could be observed in the western portions and areas around Esikado which host forest and vegetation covers tend to have a moderate vulnerability.

Fig. 8

Graphical representation of DI modification results

Effect of lineament on DRASTIC model

The combination of conventional DRASTIC and lineament is designated DI-LIN. Lineament is associated with groundwater flow and, therefore, migration of contaminants. The resulting modified and classified DI map is presented below together with the lineament map (Fig. 9).

Fig. 9

Maps of lineament (a) and modified DI (b) based on lineament

The inclusion of lineament did not bring about marked changes in the statistical apportionment to the standard DI in terms of vulnerability zones (Fig. 8). However, on the ground, there were variations in the vulnerability zones (Fig. 9). Areas that were classified as high vulnerability zones (including Asakae, Takoradi, Kwesimintsim, Apremdo, Esikado, New Takoradi) on the standard DI but having low lineament density became classified as moderate zones. The marginal increase in the high vulnerability zone (32–33.7%) were observed around Anaji area which also was classified as one of the areas with high lineament density. Though other areas of study were classified as having high lineament density, the impact did not outweigh the seven hydrogeological parameters considered in the standard vulnerability assessment. In those areas, the inclusion of lineament density did not contribute significantly. This indicated the usefulness and contribution of the lineament parameter on the overall computation of the vulnerability index.

The combine effects of LULC and Lin

The combination of the conventional DRASTIC index with LULC and lineament is designated MDI. The range and vulnerability classification used in the modification are presented in Table 11. The resulting changes in DI of modified vulnerability zones are illustrated in Fig. 10.

Table 11 Modified DRASTIC index classification

Fig. 10

Map of modified DI (MDI) based on LULC and LIN

For the area classified as low vulnerability, the resulting MDI allocated a land area that was 12.2% higher than the mean areas of the standard DI and its two modifications (ie DI + LU + LIN). Between the DI and MDI, the percentages change were more than one hundred and seventy (> 170%). This implies the inclusion of land use and lineament resulted in a greater area being classified as low vulnerability region. These areas are located between Nkroful and Kojokrom.

For the area classified as moderate vulnerability, there was no significant difference between the DI and its derivatives (ie DI + LU, DI + LIN). However, the resulting MDI allocated a landmass that was 17.5% lower than the mean areas of the standard DI and its two modifications (ie DI + LU + LIN) and 21% less than the standard DI. Thus the inclusion of land use and lineament resulted in less areas being demarcated as moderately vulnerable.

The zones of high vulnerability were of greater concern because they represent the sources through which groundwater could be contaminated much easily. The MDI allocated 40.4% of the total land area to be of high vulnerability. This is higher than the standard DI or any of the derivatives. The inclusion of land use was rather promising and of good news. The high vulnerability area was reduced to 26.2%. This buttress the importance of considering human activities on the surface that has the potential to impact the hydrogeological conditions and therefore vulnerability of groundwater.

The effects of modified net Recharge

The development of DRASTIC index based on modified recharge is designated DIrm. This took into account land use/land cover and lineament as factors of recharge. The original Piscopo equation was thus modified as:

$${\text{Recharge}}\, = \,{\text{rainfall factor}}\, + \,{\text{slope factor}}\, + \,{\text{soil permeability factor}}\, + \,{\text{LU factor}}\, + \,{\text{Lin factor}}$$

(8)

This yielded a recharge rate of 15–24 and was classified and rates assigned as 6 (15–17), 8 (18–20) and 10 (21–24). Based on this a net recharge index (rate × weight) of 24–40 was obtained and was subsequently reclassified into four as shown in Fig. 11a. The net recharge ratings were 4 (24–28), 6 (28–32), 8 (32–36) and 10 (36–40). The resulting vulnerability index, DIrm, ranged from 123 to 176. This was reclassified as low (123–140), moderate (141–158) and high (159–176). The range and vulnerability classification used in the modification are presented in Table 8. The resulting changes in net recharge classification and the resulting DI of modified vulnerability zones are illustrated in Fig. 11.

Fig. 11

Map of modified recharge (a) and DI based on modified recharge (b)

For the area classed as low vulnerability, the resulting DIrm allocated a land mass that was 193.3% higher than those of the standard DI. This implies the inclusion of LULC and lineament in the net recharge resulted in a greater area being classified as low vulnerability region. These areas are located between Adiembra and Fijai.

For the areas classified as moderate vulnerability, there was no difference between the DI and its variant DIrm. The zones of high vulnerability were of great concern because they represent the sources through which groundwater could be contaminated more easily. The DIrm allocated 26.2% of the total land mass to be of high vulnerability class. This is however lower than the standard DI by 5.8 percentage points and the MDI by 14.2%. The inclusion of land use/land cover and lineament density in net recharge therefore led to a reduction in high vulnerability zones. Areas around Asakai and Kwesimintsim were declassified as high vulnerability area.

Validation of the DRASTIC model

The DRASTIC model was validated using nitrate concentrations of 35 sampling points collected for the purpose (Fig. 11). The results indicated a moderate association between the DI and the nitrate concentration with a correlation coefficient, R² value of 0.55 (Fig. 12).

Fig. 12

Map of DI showing nitrate concentration

However, the modified DI based on sensitivity analysis involving the use of effective weights yielded an improved R² of 0.85. Thus the use of sensitivity analysis does actually improve the efficiency of the model as observed in this model validation. With respect to the monotonous increase in the correlation between nitrate and DI, the results being consistent with other reported research (Neshat et al. 2014a, b) (Fig. 13).

Fig. 13

Correlation between nitrate and a DI and b MDI

However, the marginal increase in correlation (between DI and MDI with nitrate) in this study (54.3%) compared to 200.8% (Neshat et al. 2014a, b) underscores the need to include the LULC layer in accurate determination of groundwater susceptibility to pollution. The area of study in this case is highly urbanized (cf Table 9 and Fig. 7) with much less vegetation/agricultural activities and therefore less use of fertilizer with accompanying low nitrate concentration. In the future parameters other than nitrate may have to be considered in vulnerability studies involving densely populated and urbanized regions. Nonetheless, other researchers have obtained a DI-nitrate correlation in excess of 0.80 (Ouedraogo et al. 2016; Agyare et al. 2017). Other validation procedure employed in similar studies involved a correlation between the nitrate concentration and the seven thematic layers in the model (Chitsazan and Akhtari 2009), whilst chloride has been used to compliment nitrate in the validation of the model (Shirazi et al. 2013).

Conclusions

The current study has endeavored to investigate the vulnerability of aquifers in the southwestern coastal basin using the overlay approach with the aim to preemptively protect and manage the groundwater resources. Standard and modified DRASTIC vulnerability index were employed to ascertain the workability of the method in the study area and application of nitrate to validate the model. Areas demarcated for groundwater protection (high vulnerability class) were marked. About thirty-two percent (32%) of the land area were delineated based on the standard model to need protection. These were mainly in the southwestern portions and pockets around the central parts. With modification that includes land use and lineament, the area size increased to 40.4%. Areas around Apremdo, Kwesimintsim, Asakai and Sekondi were demarcated to require protection as is the most prone areas to contamination. With modified net recharge, however, the western portions (Asakai and its surroundings) were de-classified as high vulnerability area. The summary statistics and the sensitivity analysis revealed that depth to water and soil media parameters were the most effective parameters impacting on the vulnerability of groundwater in the area. This is consistent and typical with groundwater originating from unconfined and semiconfined aquifers. Modification led to an improvement in the performance of the model with a correlation coefficient between model and nitrate concentration increasing from 0.54 to 0.85. The use of the model and its modification is acknowledged to be an effective method for the assessment of groundwater susceptibility to contamination. It is recommended for resource managers and engineers seeking to provide for the future needs of its people in a sustainable manner.