Assessment of groundwater potential zone using GIS-based multi-influencing factor (MIF), multi-criteria decision analysis (MCDA) and electrical resistivity survey techniques in Raipur city, Chhattisgarh, India

Urban areas are dynamic networks that experience rapid population growth, surface water shortage, and high groundwater demand. The groundwater potential of a region depends on different facts and it varies from place to place according to its change. Variation of the groundwater potential within a short distance and the same geological formation has also been observed (Dar et al. 2010; Nasir et al. 2018; Choudhari et al. 2018; Pradhan et al. 2018). Compared to soft rock aquifer with high yield capacity, hard rock terrain possesses a limited quantity and is mostly concentrated in the weathered zone and fractured zone. In such a situation, the proper identification of the potential zones is necessary to prevent financial loss and wasting time and effort. This type of proper identification is possible with knowledge of the geological, hydrogeological, and geophysical aquifer characteristics. In groundwater hydrology, evaluation of potentiality plays an essential domain in groundwater resources through effective planning and management, in terms of its occurrences and accumulation (Yadav et al. 2014, 2016; Pradhan et al. 2018).

There are different methods and tools available for the discovery of groundwater probable zones in a particular area (Sadeghfam et al. 2016; Mogaji & Lim 2018; Termeh et al. 2019), among which, tools like remote sensing (RS) and geographic informatic system (GIS) are categorized as the most useful and inexpensive tools which require very little mechanical and physical work. The present study involved the classification of groundwater probable zones in and nearby Raipur city area by considering tools like GIS, RS, multi-criteria decision analysis (MCDA), and resistivity survey. Different studies have been conducted all around the world, including in Chattisharg, to find out potential zones of groundwater with the help of GIS and mathematical models (Saraf & Choudhury 1998; Mustak et al. 2016a, 2016b; Thakur et al. 2016; Maity & Mandal 2017; Choudhari et al. 2018; Kumar et al. 2018; Murmu et al. 2019; Pande et al. 2019). These studies mainly emphasize the successful application of geology, geomorphology, rainfall, land use land cover (LULC), drainage density, slope, groundwater level depth, soil texture, lineament, etc., amalgamated by the weighted index analytical hierarchy process (AHP) method to produce the groundwater potential model (Singh et al. 2010, 2015, 2018; Mustak et al. 2016a, 2016b; Varga et al. 2019).

MCDA is a technique that has wide applications in different fields. It is mainly used to solve complex problems by dividing them into different sections and solving and integrating each of them to get the ultimate result. It is used in fields where decision-making is a little tough and complex. Since, compared to others, MCDA is considered as one of the approachable techniques, within that, the AHP is marked as an important one. The method was developed and introduced by Thomas L. Saaty in 1977 (Saaty 1977). The AHP (Saaty 1980) is accepted worldwide for quantitative analysis. It is a dependable decision-making tool for problems with different criteria and with different natures, which can also be used to evaluate the probable zones of groundwater occurrence chosen in this study. AHP has been accepted as a very useful tool by the international scientific community due to its ability to deal with complex problems and making suitable decisions. This method itself introduced the concept of pairwise comparison. In the absence of a quantitative rating, one can still manipulate each controlling factor's rank by proper assignment of the rank of each parameter gained from the literature study and field observation, according to its importance. In this case, the pairwise comparison is converted into a set of numbers with the help of AHP, to categorize it into different ranks according to its relative priority (Saaty 1980; Agarwal & Garg 2016). The pairwise comparison technique is considered as a theoretical-based approach that applies to the computation of weights representing their relative importance. Comparing all possible pairs from the eigenvector of the square reciprocal matrix (normalized matrix) derives a set of weights from the best fit, used for the assignment of weight for thematic layers.

Finding and locating the water-bearing/fractures zones, with the provision of electrical resistivity method is considered as the best and most accepted and generalized technique (Zohdy 1989; Taylor et al. 1992; Ayolabi 2005; Majumdar & Pal 2005; Venkateswaran et al. 2014; Pradhan et al. 2018). Vertical change in resistivity has been conducted by performing vertical electrical soundings (VES) utilizing the Schlumberger electrode setup. The simplicity of the technique, easy interpretation, and rugged nature of the associated instrumentation makes vertical electrical sounding one of the effective methods used for groundwater studies. The technique is widely used in soft and hard rock areas (Urish & Frohlich 1990; Ebraheem et al. 1997; Bredewout et al. 1998).

Therefore, the integrated use of AHP, GIS, RS, and electrical resistivity techniques by taking into account hydrogeological, geomorphologic, and meteorological data is much more reliable than applying the techniques individually. Integration of all these factors helps to produce a more appropriate result, which can refer to any other area, especially in a highly populated, developing area like Raipur. The current study employed the above-integrated method for the evaluation of potential zones of groundwater in Raipur city.

Raipur city is the capital of Chhattisgarh province, which is located in India and is positioned between latitudes 21°16′54.523″N and 21°5′30.553″N and longitudes 81°32′31.883″E and 81°53′12.433″E (Figure 1), covering around 490.43 km² geographical area. The average elevation of the area varies from 223 m to 345 m above mean sea level.

The maximum temperature is recorded around 47 °C during May. The monsoon starts in June and extends up to the end of September. The average annual precipitation is 1,240 mm, and the climate is humid. The Kharun River bounds the study area on the western side of Raipur city.

In the present study, numerous spatial data sets have been utilized to analyze probable potential zones of groundwater within the study area (Table 1).

The base map of Raipur city was prepared according to the Survey of India (SOI) toposheets (1:50,000 scale). To assess groundwater potential zones, multi-parametric data set, namely, geology, geomorphology, rainfall, lineament, LULC, drainage density, soil type, slope, and soil texture were prepared using topographic maps, existing map, data collected from the field study, and satellite image using integrated techniques such as RS and GIS (Thakur et al. 2016). The satellite data, Sentinel-2 geocoded imagery, were accessed from Earth to explore the site (http://glovis.usgs.gov/). The LULC, drainage, lineament, and soil texture map were prepared based on the Sentinel-2 (spatial resolution 10 meter) false color composite (FCC) images in ERDAS IMAGINE software using visual interpretation techniques with field check.

A geology and geomorphology map of the area has been produced from the existing maps from the reports of Central Groundwater Board (CGWB 2009), subsequently updated with Geological Survey of India (GSI) toposheets and satellite image and field check. Topography (slope map) has been prepared by using Shuttle Radar Topography Mission-Digital Elevation Model (SRTM-DEM) with a 30 m spatial resolution. Groundwater level data collected from the bore wells and the inventory wells of the CGWB located in the study area were employed for the generation of groundwater depth maps, with the help of interpolation techniques in the ArcGIS environment. Meteorological data collected from the Water Resource Department, Raipur Chhattisgarh (WRD) were used to generate rainfall maps with the help of ArcGIS software. The whole technical flowchart of the methodology adopted in this study is given in Figure 2.

In the present study, nine criteria, geomorphology, geology, lineament, slope, soil, groundwater depth, rainfall, drainage density, and LULC were considered to assess potential groundwater zones in the area. The RS and GIS techniques were employed to prepare different thematic layers which are discussed as follows.

Geomorphology deals with the formation of surface landforms that, up to a limit, controls groundwater movement and occurrence. Satellite imagery is a better tool that enables the identification of surface structures that can indirectly be applied to identify groundwater potential areas. This surface identification from the satellite is aided by the application of visual image-interpretation methods like size, shape, tone, texture, relief, location, association, physiography, landforms, and the existence of the different type of rocks or sedimentary formations and the geological structures present on it (NRSC 2010).

Identification of geomorphological features within the basin can be classified into buried pediplain, pediment zones, alluvial plain, and valley fill shallow, which were considered for finding out potential zones as per their response on groundwater occurrence (Figure 3). The geomorphic units/landforms prepared from the CGWB annual reports (CGWB 2009) and validated and rechecked by the image interpretation methods from the aerial photograph were verified and validated in the field.

The study area's lithology is considered as one of the controlling factors influencing the groundwater flow and its existence. The serial arrangement of different rocks or lithological units and their interaction determines the area's total infiltration capacity. Porous and permeability of the litho units refer to the storage and transmitting capacity, which supports the groundwater occurrence and occurrence of an area. An extensive field check with literature review and satellite data analysis by visual interpretation found that the study area comprises calcareous and argillaceous sedimentary formations, specifically limestone and shale within the Raipur group of rocks (GSI 2005; CGWB 2009). Delineated rock units in the study area were further classified into five classes, namely, alluvium, stromatolite dolomitic limestone, laterite, stromatolitic dolomitic limestone with sandstone, and shale as shown in Figure 4.

Like primary porosity, secondary porosity is also essential for the determination of hydrogeological conditions. Lineaments represent secondary porosity and are linear features of tectonic origin. Due to their linear, direct, curvilinear form, they can easily be demarcated in satellite imagery. Some other indications like tone, texture, relief, drainage, and vegetation soil tone's linearity also give valuable information for lineament differentiation. Correlation of structural features like faults, fractures, joints, and bedding planes with these lineaments is an excellent practice to determine the potential areas of groundwater occurrence (Pandian et al. 2013). Four multi-buffer zones with dimensions of 50 m, 100 m, 150 m, and 200 m (Figure 5) have been defined according to their influence on areas with good potential of groundwater. Some of them are in a crisscrossed pattern, which indicates the intersection of lineaments, and a good indicator of groundwater potential zones.

The slope is an important criterion that helps to delineate the groundwater potential zone. It directly affects infiltration and surface runoff. Low/nearly level slope has high infiltration and low runoff, resulting in good groundwater recharge, while moderate to steep slope enhances surface runoff. A slope map was prepared from the SRTM elevation data with the help of ArcGIS software (Szabó et al. 2015; Rawat et al. 2018). The slope map is categorized into three classes, nearly sloping (0–1%), very gently sloping (1–3%), and gently sloping (3–5%) (Figure 6).

Soil is a significant parameter for the identity zone of potential groundwater occurrence. The study area consists of four types of soil, i.e., sandy loam soil, sandy clay loam, clay loam, and clay soil (Figure 7). The soil texture of the area is one of the major factors that control the surface runoff and infiltration of rainwater. Sandy group soil has a low runoff rate and high groundwater potential, whereas the clay soil group has a high runoff rate and very low groundwater. Sandy soil shows a high infiltration rate, and clayey soil offers the least infiltration capacity. The majority of the study area (52.52%) was covered by clay loam soil with a low to medium infiltration rate.

The depth to the groundwater level depends upon the recharge and discharge of the groundwater. Groundwater level data collected from CGWB Raipur, used for water level maps, has been generated by the inverse distance weighting (IDW) interpolation method. Normal depth to water level within the study area ranges between 3 and 12 mbgl. The depth to water level map is categorized into four classes, 3–6 mbgl, 6 and 9 mbgl, 9 and 12 mbgl, and >12 mbgl. This parameter is important for the groundwater potential zone. The thematic representation of the depth to water level shows that the eastern part of the study area is found as an area of low groundwater level (Figure 8).

Rainwater is considered the primary source of the groundwater resource. The rainfall distribution associated with surface gradient influences the runoff and infiltration rate, hence indicating the possible groundwater potential zones. The southwest monsoon is prominent within the study area, and is active from June and extends up to September. With the aid of five rain gauge stations, the recorded daily rainfall of the year 2016 indicates that rainfall was above normal with significantly high intensity than that compared to the past 15 years. The maximum rainfall is recorded at approximately 1,267 mm. According to the available data, four classes have been derived to characterize the potential zone (Figure 9).

Drainage density has been divided into four, ranging from very low to high, >1.00 to 1.43 km/km² (very low), 1.34 to 2.24 km/km² (low), 2.24 to 3.23 km/km² (medium), and 3.23 to 4.89 km/km² (high) (Figure 10). According to the recharge rate, the weight of the subclasses of drainage density is assigned in a manner that high weight is given to the areas of low drainage density, as a representative of low runoff rate and high infiltration rates, and similarly, low weight has been given to the regions having high drainage density with a high rate of surface runoff and low rate of infiltration. This indirectly infers that the area with high drainage density represents the land of impermeable rock and low drainage density represents a permeable basement. In conclusion, the possible groundwater occurrence zones can be identified by the presence of low drainage density areas and of high infiltration rates.

Consumption of land for different processes also affects the pattern of infiltration within that area. LULC of the study area was captured and monitored by satellite image using visual interpolation techniques in ERDAS IMAGINE software, followed by field verification. The LULC was categorized into eight extensive classes, namely, cultivation, settlement, vegetation, open land, industry, drainage, lake, and road, that has been identified and demarcated, as shown in Figure 11. The majority of the study area is covered with cultivation (43.78%), settlement (24.04%), open land (22.18%), road (2.21%), and lake (5.37%). Each subclass in the land uses land cover class assigned with different weights according to their participation in groundwater infiltration. Within the subclasses, the lake and drainage possess high weightage due to less runoff. In contrast, the settlement and road possess low weight due to high runoff, and moderate to good groundwater potential can be found in the areas that cover open land and cultivation land, which consist of medium weight (Figure 11).

The AHP is a subunit of the multi-criteria decision-making method, which involves the analysis and associated decision-making of multiple objectives (Wang et al. 2009; Cay & Uyan 2013; Uyan & Cay 2013; Murmu et al. 2019). The GIS-based multi-criteria assessment applies to define the rates of various classes in each layer. Weights of each thematic layer are allocated according to its influence on groundwater potential and by considering Saaty's AHP method (Saaty 1980). In MCDA, weightage assignment for each influencing factor is applied by considering its practical role in a particular area (Chow & Sadler 2010; Agarwal & Garg 2016; Jhariya et al. 2017; Murmu et al. 2019). The AHP is a subjective approach in which subunit selection and its weightage allocation are based on the comparison between various criteria derived from the appropriate strategy of decision-making proposed by Saaty (1980). The AHP method is applied according to the calculation of weightage from a preference matrix representing map layers. The weightage is generated by the comparison of relevant criteria based on preference factors. The ability to manage a vast number of heterogeneous data for the required weightage, even for vast data in a straight forward manner, has made the method a popular one within various GIS methods (Chen et al. 2010; Feizizadeh & Blaschke 2012; Jhariya et al. 2016; Khan & Jhariya 2019).

In AHP, decision-making functions through different processes such as division of different parameters, hierarchical arrangement, judgments based on the relative significance of a set of elements, and derivation of its results (Saaty 1999; Agarwal et al. 2013). The AHP was first introduced by Saaty for solving a complex problem by splitting it into different categories and integrating each subsection to find out the big picture that has to be solved. In this study, nine thematic layers have been developed and their relationships are defined with the aid of the AHP. The steps involved in the derivation of weights of major thematic layers and their subclasses are shown below (Saaty 2004; Agarwal et al. 2013; Kumar et al. 2014; Jhariya et al. 2016; Khan & Jhariya 2019; Murmu et al. 2019).

The first step involves the determination of the relative importance values. Saaty's 1–9 scale is the standard reference used for the finalization of relative important values (Table 2, in which the score of 1 represents two themes of equal importance, whereas 9 is an indicator of a highly important theme (Saaty 1980).

The RI values representing different numbers of n are shown in Table 3 (Saaty 1980). The value of CR should be less than or equal to 0.1 (Maity & Mandal 2017). If it deviates from the stated condition, re-evaluation of corresponding weight should practice avoiding inconsistency; otherwise, the AHP may provide faulty results (Chakraborty & Banik 2006).

The depth-wise variation of the formation has been studied with the help of the electrical survey by the successive increase in the current electrodes’ separation. The method is simply defined as the vertical electrical survey carried out by keeping the central point of the electrode configuration fixed, by the gradual increase of the separation between the current electrodes. According to the increasing current electrode separation, the graphical representation of the obtained apparent resistivity record is known as the apparent resistivity curve or field curve at the point of observation. An increase in the distance between the potential electrode and current electrode increases the depth of penetration of the current, results in the generation of the curves, which reveals the depth-wise variation in resistivity. The resulting curves subject to the curve matching method with the theoretical curve of the known vertical distribution of the resistivity and thickness gives the information of sequences representing the geo-electrical layer at the point and depth of investigation.

In sounding with Schlumberger configuration, the movement of electrodes is carried out to always follow a straight line by keeping the potential electrodes closely spaced. In practice, the potential electrode spacing is kept at not more than 1/5th of the current electrode spacing. The successive length of current electrode spacing is usually increased in geometric progression. The apparent resistivity curve should be plotted on double log graph paper with the help of at least four to eight points of equal distribution. The break in the curve can be determined by keeping the potential electrode position as a fixed one and outward symmetrical movement of the current electrodes at an increment of 2–5 m. Conventionally, electrodes and spacing have been accepted as, A, B, M, and N. A and B represent the current electrodes, while AB represents its spacing. Likewise, M and N are the representatives of potential electrodes with the spacing of MN:

• The apparent resistivity ρa = π{(AB/2)₂− (MN/2)²}/MN * R = K * R.

• The geometric factor K = π{(AB/2)²− (MN/2)²}/MN.

Deviation from conventional methods and the adoption of potential mapping techniques/tools such as RS, GIS, and electrical resistivity can be applied for the delineation of groundwater probable zones. Different thematic layers were developed according to each of them and their importance on groundwater occurrence of the study area (Figures 3–11). The rate of each assigned factor decides the contribution of each factor on groundwater storage and potentiality. In this process, the GIS layers of different factors such as geomorphology, geology, LULC, soil, lineament buffer, slope, rainfall, and drainage density were analyzed carefully, and weights were assigned to corresponding thematic maps. Allocation of rates from 1 to 5 indicate very low, low, medium, high, and very high in ascending order, associated with each class, were selected based on the control of each factor on the groundwater potential.

The representative weight of each thematic layer and associated classes derived from applying the AHP method are given in Tables 4 and 5. The linear combinations of these weights that are adopted for the evaluation of groundwater potential are shown in Table 6.

The calculation based on the AHP method in this study has been conducted by considering different factors, such as n = number of factors involved (i.e., 9) and λ = average value of the consistency vector λ = 9.25 + 9.46 + 9.26 + 9.45 + 9.50 + 9.47 + 9.48 + 9.51 + 9.42/9 = 9.42 and CI = (9.42-9)/(9-1) = 0.064.

Where RI is the Ratio Index, the value of RI for selected ‘n’ values are given in Table 3. For n = 9, CR is 0.044, as 0.044 (CR) <0.10, is under the acceptable limit which indicates reliability in the pairwise comparison (Saaty 1999, 2004; Dalalah et al. 2010; Agarwal et al. 2013; Jhariya et al. 2016; Jhariya et al. 2017) (Table 4). The weights of the different criteria and corresponding CR are shown in Table 4. Finalized weights for slope, geology, geomorphology, LULC, drainage density, slope, groundwater depth, soil texture, and rainfall are derived as 0.023, 0.202, 0.066, 0.201, 0.080, 0.152, 0.087, 0.117, and 0.072, respectively.

Application of AHP techniques by considering the weighted parameters for demarcation of potential zones involves calculations in the raster format module and development of potential maps in the GIS environment. The adopted weight results from the normalization of individual parameters by considering the ratio of weight assigned to the specific parameter and the corresponding layer's geometric mean. Normalized weight derived from the thematic layers’ individual features considered for producing a potential groundwater index map was created (Figure 12). According to the spatial variation of groundwater potential, the study area was split into five zones, namely, low, medium, medium-high, high, and very high, whose spatial distribution and extents 58.14 km² (11.86%), 141.34 km² (28.82%), 166.33 km² (33.92%), 103.27 km² (21.06%), and 21.35 km² (4.35%) are given in Table 7.

Delineation of the aquifer geometry is possible with the help of a vertical electric survey (VES). Around 14 locations have been selected differentially for the survey using the Schlumberger configuration (Zhdanov & Keller 1994) with a maximum current electrode spacing (AB) of 300 m. This electrode spacing was sufficient to provide information about the resistivity variation from near the subsurface aquifer condition and delineate the shallow and deeper aquifer systems. Data were collected for different depths (AB/2 = 2, 3, 4.5, 6, 8, 10, 12, 15, 20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, 100, 120, and 150 m). The VES data were interpreted manually using two-layer master curves and auxiliary curves, and the results were also rechecked by using IPI2WIN software.

The interpreted results of resistivity data indicate there are four different layers (the results are given in Table 8 and Figure 13) as given below:

• The first layer is topsoil cover, except at Kurru (VES-64), Banjari (VES-65), and IIIT Naya Raipur (VES-68), where the top layer is laterite. The resistivity ranges of this layer vary from 3 to 360 Ohm-m. The depth ranges vary from 1.1 m to 3.6 m.

• The second layer is weather formation (shale/limestone), and its resistivity varies from 2 to 115 Ohm-m. The thickness varies from 1.3 to 19.4 m.

• The third layer is the fractured formation (fracture shale/limestone), and its resistivity varies from 15 to 345 Ohm-m. The thickness of this layer is between 12 and 39.3 m.

• The fourth layer, i.e., the last layer, is the massive formation (massive shale/limestone) below the fracture formation. The resistivity of this layer varies from 150 to 950 Ohm-m.

The maximum resistivity value was observed in Dharampura (VES No.43) and the resistivity value is 950 Ohm-m below a depth of 16.7 m. The high resistivity indicates that the formation is compact at this depth and the rock type is limestone.

To identify fractures at depth in hard rock area by conducting VES, a factor analysis method is used. In this method, first of all, the value of apparent resistivity should be taken for the same potential dipole (MN/2) value. The factor for any AB/2 value is the ratio of apparent resistivity value of that AB/2 and the sum of all the apparent resistivity values of all the earlier AB/2. If the total number of apparent resistivity values of a sounding is ‘n’ then the total factor will be n-1, as there will be a factor for the first AB/2. We can identify the same factor value for two consecutive readings of AB/2 from the obtained factor values to indicate the fracture zone at the respective depth. The provable fracture zones are given in Table 8.

Assessment of potential zones within the study area reveals that the high potential zone has been detected in the west and north-eastern portion. In contrast, low to medium groundwater potential is located in the central and eastern parts. Derived groundwater potential results from the integrated operation of various factors such as slope, rainfall, lineament, drainage density, and soil patterns including the geomorphic and geological control.

The result shows that LULC and geology are the main factors that control the groundwater occurrence in the area. The geomorphology and slope do not have much influence on the groundwater potential because of the low slope and moderately horizontal and plain nature of the terrain.

Groundwater occurrence of the area is mainly controlled by the geological succession of the area. Even though the LULC may play a significant role in the infiltration of the water and recharge of the groundwater, the geology itself provides the space for its free movement. Comparison of the resulting groundwater potential and the geology of the area shows that the zone of high to a very high potential is located in the area with formations of alluvium, stromatolitic dolomitic limestone, and stromatolitic dolomitic limestone with sandstone.

The result obtained from the resistivity survey also supports the result obtained from the demarcation of probable zones. The resistivity survey helped to differentiate the vertical variation of the layers that present beneath the study area. The soil cover and the laterite formation extend up to a 20 m depth, and along with the LULC distribution provides a positive signal concerning the infiltration of water to the groundwater resources. Further, the occurrence is controlled by the fractured formation that is situated beneath it. The fracture zones located in the dolomitic limestone and similarly with the association of sandstone, reveal the adequate density of fracture zones that are present in the formation.

The obtained result is much more applicable to the categorization of the land into different grades. Moreover, the prediction of the groundwater probable zones is important in the future, in the sense that studies reveal the impact of climate change on water resources is considerable (variability of rainfall patterns and surface water flow) and may cause a positive and negative impact on the replenishable water resources (especially to surface water) (Havril et al. 2018; Kaini et al. 2019, 2020a, 2020b). The prediction of groundwater availability can help consumers to control the unwanted usage/excess usage of the groundwater, especially in a region that is running out of groundwater resources.

Since the majority of the area consists of cultivated land and settlement area, the major findings are helpful for the community that resides there, and also for the farmer community to utilize the groundwater and the area in a proper manner. Further, the results may help policymakers and development planners, who will apply their projects in various fields, which have a direct or indirect connection with the groundwater consumption in the area. Even though the result can help them, it is important to assess and consider the socio-economic impact on the areas that may arise by the application of the projects (Kaini et al. 2020a, 2020b; Suhardiman et al. 2020).

In the present study, the output result has been validated using well yield data and the ROC method.

An accuracy check of the prediction model is highly essential to prevent errors and improve environmental studies’ decision-making. The obtained potential zones derived by integrating different techniques like RS, GIS, and MCDA were validated with correlation studies of data collected from ten wells in the study area (Table 9 and Figure 14). Locations of the selected wells, yield attained from the prediction map, actual yield data collected from the pumping test, and the acceptance/rejection of values that denote borehole deviation yield data between expected/real in the form of the agreement are shown in Table 9.

Accuracy check of predicted values is estimated as follows:

• Total boreholes = 10.

• Number of borehole value which attains agreement between actual yield and expected = 8.

• Number of borehole value which shows disagreement between actual yield and expected = 2.

• Accuracy of prediction = (8/10) * 100 = 80%.

The accuracy prediction proved that the selected methodology implemented in this study is notably reliable and accurate.

In the present study, the sample size is 10 and the middle and high yield is 1, 2 and the low yield is 0. The area under the ROC curve (AUC) is utilized to measure how well the proposed model predicts low and middle/high yield areas. It is a plot between false-positive percent and true positive percent. The true positive percent indicates the sample for which the expected yield description from the prediction map matches the actual yield description.

On the other hand, false-positive percent indicates the percentage of wrongly detected positive cases (middle/high yield). The ROC plots are obtained using Medcalc software at a 95% confidence interval. To plot ROC, the 10 samples shown in Table 9 (sample size = 10) are used. The classification model output corresponding to middle and high yield is considered positive, while that corresponding to low yield is negative. The ROC plot obtained is shown in Figure 15. The area under the ROC curve (AUC) is found to be 0.857 (85.7%) at a significance value of less than 0.001. Thus, the high value of AUC justifies the efficacy of the proposed approach in predicting low and middle/high yield areas.

In the present study, the successful application of techniques such as RS, GIS, MCDM, and electrical resistivity helped evaluate and identify probable groundwater potential zones. Different steps chosen for the study include the development of the thematic layer followed by the appointment of weight for each influencing factor with the help of the AHP method and, last, overlay analysis for the demarcation of groundwater potential zone. Electrical resistivity is also carried out to locate the fracture zone. The interpreted results of resistivity data indicate there are four different layers, topsoil, weathered formation (shale/limestone), fractured formation (fracture shale/limestone), and massive formation (massive shale/limestone). The highest resistivity value was observed in Dharampura (VES No. 43) and the resistivity value is 950 Ohm-m below the depth of 16.7 m. The elevated resistivity indicates that the formation is compact at this depth and the rock type is limestone.

The groundwater potential zone of the study area was split into five zones, namely, low, medium, medium-high, high, and very high, whose spatial distribution and extents are 58.14 km² (11.86%), 141.34 km² (28.82%), 166.33 km² (33.92%), 103.27 km² (21.06%), and 21.35 km² (4.35%). The result reveals that the study's potential zones are high in the west and north-eastern portion, while low to medium groundwater potential is located in the central and eastern portion. Derived groundwater potential results from the integrated operation of various factors such as slope, rainfall, lineament, drainage density, and soil patterns, including the geomorphic and geological control. The result revealed that the LULC and geology play a major role in the groundwater condition of the area. This study validated using well yield from which accuracy of 80% was attained and validated using the ROC method. The area under the ROC curve (AUC) is found to be 0.857 (85.7%) at a significance value of less than 0.001. Thus, the high value of AUC justifies the efficacy of the proposed approach in predicting low and middle/high yield areas. Overall, the integrated application of RS and GIS method supported by the resistivity analysis increases the values of the result, therefore, it can be considered as authentic data for future planning, especially for policymakers and development planners that interact with or affect the groundwater resources of the area. However, a slight variation in the future possibly considers the variables rainfall and LULC as the major factors for the computation of groundwater potential zones.

The authors are highly grateful to the Chhattisgarh Council of Science & Technology (CGCOST) Raipur, Chhattisgarh, for providing research funds to carry out the present study. Enormous thanks go to the Central Groundwater Board (CGWB) NCCR, Raipur, Chhattisgarh, Government of India for providing the data and help extended to us for the present study. Heartfelt thanks go to those who knowingly and unknowingly have helped to make this paper successful.

All relevant data are included in the paper or its Supplementary Information.