Introduction

Groundwater is one of the most important freshwater resources in our daily life (Lezier et al. 2017). Under the background of global climate change, untreated rate of urbanization, human interferences and policy variations affect the change of global and local groundwater quality which is very common in developing countries like India (Emenike et al. 2020). Groundwater is the largest reservoir of available freshwater and accounts for nearly one-third of global freshwater withdrawal (Siebert et al. 2010; Famiglietti 2014). Groundwater supplies more than 40 percent in irrigation sector and about 50% of municipal water extractions worldwide (Ross 2016). The estimated worldwide groundwater abstraction is approximately 982 km3/year, corresponding to approximately one quarter of the entire water abstraction. For these groundwater volumes, 70% is for agriculture, 21% is for domestic use, and the remainder is mainly for industrial use (Margat and Van der Gun 2013). Groundwater is the main source of drinking water worldwide, with 2.5 billion people solely depending on this resource to fulfill their daily needs (Zektser and Everett 2004). According to the Central Water commission report in India, groundwater was extensively used in agricultural sectors and 50% of the total irrigated land was dependent on groundwater. (“Central Water Commission, Ministry of jal shakti, Department of Water Resources, River Development and Ganga Rejuvenation, GoI” 2000). In India, 60% of irrigated food production depends on groundwater wells irrigation (Shah et al. 2000; Das and Pal 2019a). India uses 25 percent of all groundwater extracted worldwide at just over 260 cubic km per year, ahead of the USA and China (Barlow 2012). In India, 70% water in agricultural sectors was supplied by groundwater (Foster and van Steenbergen 2011). India’s water supply will remain the lifeline for years to come. Most water issues in India are those relating to groundwater water found beneath the surface of the earth. It is because we are the world’s largest groundwater user and therefore highly reliant upon it. The exploitation and contamination of groundwater have become a global problem. The environmental and hydrogeology researchers recognize that groundwater resources are at risk due to overexploitation and pollutants concentration. Overexploitation of groundwater occurs in a context in which groundwater has shifted from “a reserve resource” used in times of scarcity to “one now systematically abstracted in an uncontrolled or unregulated manner” (Molle et al. 2018).

The consequence is widespread groundwater tension, with approximately 1.7 billion people living in overexploited areas (Wada et al. 2011). At the same time, the quality of the groundwater is under threat from a growing list of anthropogenic pollutants (Lapworth et al. 2012). Groundwater vulnerability assessment has an important method for controlling and securing groundwater (Yu et al. 2012; Kumar et al. 2015; Adimalla and Wu 2019). The unsuitable groundwater management and vulnerability model of this valuable resource in the area have resulted in the depletion of potential natural resource and a continuous deterioration of its quality (Mondal et al. 2019). The term “vulnerability” refers to a changing state where a risk exists or any possibility of harm to a society (Pal et al. 2019; Das and Pal 2019b).

The DRASTIC model was used to classify vulnerable areas for development and management of the groundwater (Huan et al. 2012; Jaseela et al. 2016; Wu et al. 2016; Yang et al. 2017; Kumar et al. 2016). Number of studies indicated that the DRASTIC needs to be modified (Noori et al. 2019; Yang et al. 2017; Karan et al. 2018). Such alternative methods are useful for identifying vulnerable areas of pollution in far more specialized cases (Kumar and Pramod Krishna 2020). The DRASTIC approach adopted by the United States Environmental Protection Agency (EPA) is the commonly used tool for mapping vulnerabilities in pollution (Shirazi et al. 2012; Neshat and Pradhan 2015). The DRASTIC model was widely used in various regions like Colombia, Italy, Greece, Afghanistan (Betancur et al. 2013; Aschonitis et al. 2016; Busico et al. 2017; Gesim and Okazaki 2018). Recent researchers have made changes to rating scales to adjust the model by using ArcGIS data interpolation (kriging method) (Sadat-Noori and Ebrahimi 2016). Furthermore, modification of the model has been done with omission or addition of layers such as land use patterns to estimate contamination risk to pollutants (Alam et al. 2014; Kumar et al. 2017; Kozłowski and Sojka 2019; Noori et al. 2019; Jesiya and Gopinath 2019; Jia et al. 2019) and also many literatures has been applied different weight and rating in respect to agriculture (Saha and Alam 2014; Thapa et al. 2018). The application of the analytical hierarchy process (AHP) to adjust the DRASTIC parameter weights to local settings has become prevalent (Lad et al. 2019; Yang et al. 2017; Chakrabortty et al. 2018; Jhariya et al. 2019). The present study has been carried out in Malda district of West Bengal, India. This district faces tremendous problems related to groundwater, and a major portion of this area is agricultural land that creates a lot of pesticide contamination in groundwater, and various land use like urban sewage, industrial effluents, etc., also affects groundwater. So, this study aims to develop and examine spatial groundwater vulnerability mapping to assess or identify the groundwater vulnerable zones for sustainable management or planning. This work has been done through various groundwater vulnerability assessment methods like DRASTIC, land use DRASTIC, agricultural DRASTIC and AHP land use DRASTIC.

Materials and Methods

Description of the Study Area

The study area consists of the entire Malda districts of West Bengal, India. Malda district is located in between the Rajmahal and the Garo hills. The geographical location of this district is 24°40′20″ N latitude to 25°32′08″ N latitude and 87°45′50″ E longitude to 88°28′10″ E longitude. The geographical area of this district is 3733.66 square kilometers. The district is surrounded by the Murshidabad district in the southern part and North Dinajpur district in the northern part, Bangladesh, and South Dinajpur in east and Jharkhand state in the west (Fig. 1). The region is experiencing subtropical monsoon types of climate with seasonal variation. The annual rainfall of this region is 1450 mm. About 90% of the annual rainfall occurs in the monsoon period (July–August).The Gangetic flat land of the district is very fertile for agricultural crops, and the agricultural productivity is very high in this region.

Fig. 1
figure 1

Location map of study area

Full size image

Geological Setting

The geological location of the area is a result of the tectonic history and subsequent sedimentation of the Garo-Rajmahal Gap, which separates the Bengal Basin into a northern foreground facing the Himalayas and a southern delta. The district is classified into three broad physiographic divisions, namely Tal, Diara and Barind (Mukherjee and Pal 2018; Ghosh and Kar 2018). Eastern part of this area is characterized by successive ridges and depressions sealed with small streams of water in the valleys named as Barind tract, which consists of Gazole, Habibpur and Bamangola administrative blocks. The southern part is defined by well-drained flat land created by the recent alluvium known as Diara, which consists of the Manikchak, Kaliachak-I, Kaliachak-II, Kaliachak-III and English Bazaar administrative blocks. The northern part is a marshy tract known as the Tal region, which includes Harischandrapur-I, Harischandrapur-II, Chanchal-I, Chanchal-II and Ratua administrative blocks.

The geological formation of the study region was divided into three ages, such as the Holocene era, Pleistocene period and the Meghalayan period (Fig. 2). Baikunthapur formation was developed in the early Holocene to Pleistocene period. Shaugaon, Koshi-Ganga, Jalpaiguri, Malda and Bhagirathi formation was formed in the Holocene period. The Barind formation was formed during the Pleistocene period. Present-day deposits were formed in the Meghalayan period. English-Bazar block is covered by Jalpaiguri, Malda, Shaugaon and Present-day deposit formation. Koshi-Ganga formation is found in Kaliachak-I, Kaliachak-II, Kaliachak-III, Harischandrapur-I, Harischandrapur-II, Chanchal-I and Chanchal-II. Baikunthapur and Barind formation is located in the Old-Malda block, and the Present-day formation is located in Manikchak and Ratua-I, Ratua-II block (Fig. 2).

Fig. 2
figure 2

Geological map of the study area

Full size image

Hydrological Setting

The hydrogeological characteristics are closely related to the topography of a region. The hilly areas primarily consist of aquitard, while the central areas consist of unconsolidated water-bearing sediment layers. These aquitard and unconsolidated layers formed a complete hydrogeological unit. On the basis of geological character and hydrogeological setting, aquifer system can be subdivided into confined and unconfined aquifer. The unconfined aquifer mainly consists of sand, clayey silt and sandy silt. Very coarse unconsolidated granular sediments occur in this area and continuous aquifer 950 m to 100 m thick. The depth of groundwater usually 2 m to 6 m below the ground surface. In this district, Tal region is centrally depressed and has a water level between 6–70 m and 5–70 m on an average. The Barind tract shows the lowest level of underground water, and ground water depth of this region is 6 m to 7 m. In Diara, track water level is relatively nearer to the surface. Average water depth of this region is 3.36 m to 4 m. However, this water level is not fixed; it is fluctuated by several factors like, local slope, soil texture, soil structure, rainfall and permeability of this region.

Database and Method

Various data sets have been processed and used to assess the groundwater vulnerability of the Malda district. Applicable secondary data such as elevation data, meteorological data and hydrological characteristics were also used in this analysis (Table 1). The key features of the data used are shown in Fig. 3. A combination of DRASTIC, agricultural DRASTIC, land use DRASTIC and AHP land use DRASTIC models were used to assess groundwater vulnerability associated with pollution.

Table 1 Sources of data used for parameters of DRASTIC and modified DRASTIC models
Full size table
Fig. 3
figure 3

Database and methodological flow chart

Full size image

DRASTIC

DRASTIC is one of the most significant methods, which was used to assess the vulnerability of groundwater; it is a kind of overlay index process by combining various influencing variables. For evaluating groundwater vulnerability to pollution, this model was established by the US Environmental Protection Agency (EPA) (Aller et al. 1985) and National Water Well Association. DRASTIC is the short form of the hydrological factor that manages the groundwater pollution, the hydrologic character is combined with geological and hydrological factors that have an effect on the groundwater movement, and seven influencing factors are depth to water, net recharge, aquifer media, soil media, topography, the impact of vadose zone, and hydraulic conductivity. Each factor determines rating score and a weight categorizing the aquifer by potential pollution. Each DRASTIC factor was assessed with respect to the other factor in order to assess the relative value of each factor. A proportional weight of 1 to 5 has been allocated to each DRASTIC (Table 2). The most important factors are the weights of 5; the least important, the weight of 1. Every range for each DRASTIC factor was assessed with respect to the others in order to evaluate the relative importance of each range with respect to the potential for pollution. The rating of this scale varies 1-10, which has been assigned to the range for each DRASTIC factor.

$${\text{DRASTIC}}\; {\text{Vulnerbility}}\; {\text{Index}}\; = {\text{DrDw}} + {\text{RrRw}} + {\text{ArAw}} + {\text{SrSw}} + {\text{TrTw}} + {\text{IrIw}} + {\text{CrCw }}$$
(1)

where D = depth to groundwater, R = net recharge, A = aquifer media, S = soil media, T = topography, I = impact of the vadose zone, C = hydraulic conductivity, r = rating, w = weight (Secunda et al. 1998).

Table 2 Weight of DRASTIC, land use DRASTIC, agricultural DRASTIC and AHP land use DRASTIC parameters
Full size table

Agricultural DRASTIC

Agricultural DRASTIC is an unusual case with the same set of factors for DRASTIC index. Standard weights are intended to represent the pesticide agricultural practice (Aller et al. 1985). The agricultural DRASTIC has been considered to be the place where the pesticide is applied. The original DRASTIC is used for the industrial and municipality areas, and the agricultural DRASTIC is used in the agricultural land area. Agricultural DRASTIC has been used as most of the region of this study area is agricultural field. Agriculture DRASTIC index was drawn up using the same equation, taking into account the same weight of all parameters but different in Sw, Tw, Iw and Cw.

Land use DRASTIC

Land use DRASTIC method is used to evaluate the vulnerability of groundwater to LULC (Secunda et al. 1998; Al-Adamat et al. 2003; Shirazi et al. 2013). The modification of this model relied on alteration of the rating, weight and addition, subtraction of parameters.

$${\text{Land}} - {\text{Use DRASTIC Vulnerbility Index}} = {\text{DrDw}} + {\text{RrRw}} + {\text{ArAw}} + {\text{SrSw}} + {\text{TrTw}} + {\text{IrIw}} + {\text{CrCw}} + {\text{LrLw}}$$
(2)

where D = depth to groundwater, R = net recharge, A = aquifer media, S = soil media, T = topography, I = impact of the vadose zone, C = hydraulic conductivity, L = land use, r = rating, w = weighting.

AHP—Land Use DRASTIC

The AHP used to consider the effect of each DRASTIC factor on the vulnerability cycle provides a comparison of the value of the variables, the normalization, and the computing consistency ratio (CR). When applying the AHP method, the power of significance (as per Saaty’s scale) between the two variables is filled in a matrix format using ground-based inquiries, field characteristics and subject matter specialists’ view. Such values have simplified to the influence of the subjectivity, involved in the weight assignment process. Standardized weights are tested for accuracy using the CR formula. To calculate the CR, the Consistency Index (CI) was first computed (Eq. 4) from the pair-wise comparison matrix of all Saaty (1980) parameters.

The pair-wise comparisons allow estimating the significance of all factor strength assessment of judgment matrix-based Consistency ratio.

$${\text{C}}.{\text{R}} = {\text{CI}}/{\text{RI}}$$
(3)

Consistency Index (C.I) evaluated as

$$C.I = \frac{{\lambda_{\hbox{max} } - N }}{N - 1}$$
(4)

where λ max is the principal eigenvalue and N is the number of factors λmax = Σ of the products between each element of the priority vector and column totals. This CI can be compared with a random matrix, RI, so the ratio CI/RI is the consistency ratio (CR). In general rule, CR ≤ 0.1 should be maintained in order to be consistent with the matrix.

In this analysis, the weights of all parameters used in the classic DRASTIC and land use DRASTIC were determined using the AHP technique, to enhance the initial weight factors included in the vulnerability equation analysis. The different vulnerability index was determined using new weight values for each of the parameters, and new vulnerabilities maps have been prepared using the DRASTIC, land use DRASTIC and AHP models with the GIS techniques described above (Kumar et al. 2017).

Single Parameter Sensitivity Analysis of the model

Sensitivity analysis demonstrated the differences in outcomes when the input variables are modified. It is vital to identify the factors that should be considered for assessing the intrinsic vulnerability of groundwater (Sener and Davraz 2013). Through single parameter sensitivity analysis test, the mean contribution of every parameter on the final vulnerability index was calculated. The single parameter sensitivity analysis (Napolitano and Fabbri 1996) is used to determine the effect of specific input variables on the vulnerability index. The aim of this operation is to compare the effective weight of the different parameters with the corresponding assigned weight model (Babiker et al. 2005). In SPSA, all parameters have been tested for their interdependence and uncertainty, as parameter independence decreases the risk of decision making, and then the effective weight of each parameter has been calculated (Rosen 1994).

$$W = \left\{ {\left( {Pr \times Pw} \right)/V} \right\} \times 100$$
(5)

where W refers to the effective weight, Pr and Pw are the corresponding rating and weight of each parameter, and v is the overall vulnerability index.

Validation through ROC Curve

Validation of models is an important part of any scientific research. AUC of ROC method has been applied to determine the model accuracy or validation. The plotting of sensitivity (true-positive rate) on the y axis and 1 − specificity (false-positive rate) on the x axis produce these curves (UM and CN 2012; Carter et al. 2016). The groundwater depth has been taken as a reference point for preparing ROC, which determines as an important parameter for groundwater vulnerability analysis. ROC curve is a popular validation method, which is widely used in different fields like floods (Sarkar and Mondal 2020), landslide (Pal and Chowdhuri 2019), gully erosion (Arabameri et al. 2020), livelihood vulnerability (Singha et al. 2020). The AUC was calculated using the ROC and judged on the basis of this acceptability of the models. The area under curve (AUC) is a measure of the likelihood of proper classification of a phenomenon by the model obtained. The AUC nearer to 0.5 indicates predictive efficiency is equal to random sampling, and that nearer to 1 indicates correct prediction (Meyer et al. 2014).

Friedman Test

The Friedman test (Friedman 1935) also applied in this study, which is a nonparametric test for detecting the disparities between the models. This test ranked to every row, taking into account the rank values of every column. The current research’s null hypothesis (H0) shows that there’s no difference between vulnerability model prediction capabilities. If the p value (significance) is less than the significance level (< 0.05), the null hypothesis is rejected. The greatest weakness of this technique is that it only shows whether or not there is a difference in performance between the models and does not have the capability to display pair-wise comparisons between the models (Bui et al. 2018).

Data Layer

Groundwater Depth

Groundwater depth helps to understand groundwater level, the quantification of resources, and groundwater flow, that rely on several factors such as geological and hydrogeological environments, aquifer coverage, groundwater recharge situation, rainfall, discharge levels, etc. (Thapa et al. 2018). Groundwater depth denotes difference between the earth surfaces to the groundwater table. Depth of groundwater is significant to delineate the groundwater vulnerability. Pollution mixed with groundwater is based on its depth, the lower water table depth refers to lower chances, and the higher depth of water table states higher the chances of pollutants entering groundwater. Data of groundwater depth have been taken from Central Ground Water Board (CGWB) Yearbook (2015–2016). A total of 83 well stations were used, and the point is interpolated using Inverse Distance Weighted (IDW) method in ArcGIS environment. The depth of groundwater is alienated into 5 classes, the lowest depth is 2 m, and the highest depth is 21.98 m. The lowest values were identified in the western and southwestern parts of the area, and the highest values were observed in the eastern part of the area (Fig. 4a).

Fig. 4
figure 4figure 4

Data layers a depth of water, b net recharge, c aquifer media, d soil media, e slope, f impact of vadose zone, g hydraulic conductivity, h land use

Full size image

Net Recharge

Rainfall is the principle supply of groundwater, and recharge is the process that infiltrates the rainfall to the ground surface and enters the aquifer. The infiltrated water pollutes vertically the groundwater and horizontally to the aquifer. In this analysis, various factors such as slope, soil permeability and rainfall were used for calculating the recharge portion (Bartzas et al. 2015; Jhariya et al. 2019).

$${\text{Recharge}}\;{\text{Value}} = {\text{Slope }}\;{\text{factor}} + {\text{Rainfall}}\; {\text{factor}}\;\left( {\text{mm}} \right) + {\text{Soil }}\;{\text{permeability}}$$
(6)

The slope (%) in the area was obtained from the digital elevation model of ASTER; the map of soil permeability was derived from a soil texture map. The rainfall of the study area is 1356 mm; the rating of rainfall was 4 as per Table 2 (Piscopo 2001). By using recharge value, the net recharge map was produced (Eq. 5). The net recharge varies 5 to 13; the lowest values were observed in eastern portion, where the highest value was observed in the southwest part and rest of the value varies scatter in the whole area. The highest amount of net recharge carries the maximum rating for groundwater vulnerability (Fig. 4b).

Aquifer Media

It refers to the rock that is consolidated or unconsolidated and serves as an aquifer. An aquifer is characterized as a subterranean rock unit that will yield enough water for human use (Alam et al. 2014). The vulnerability of groundwater depends on the aquifer of the region; the greater porous in the aquifer, the greater the chances on groundwater vulnerability. Particulate migration is controlled by the aquifer media’s both primary and secondary porosity. The time taken to pollute the groundwater is dependent on the aquifer media. The consolidated medium is the low rating and the unconsolidated medium is the high rating of aquifer media. In this area, the sand and gravel are fundamental rock formation, the class of rating is 8, and the weight of this type of media is 4 (Table 3). One type of aquifer media belongs to the total study area. The medium rating of aquifer media depicts medium rating of the vulnerability of the area by aquifer.

Table 3 Recharge ratings and weightings for the study area (Piscopo 2001)
Full size table

Soil Media

Soil is regarded to be the uppermost part of the earth’s surface. Soil is a vital factor for the measurement of groundwater vulnerability. It is the medium that helps to infiltrate the rainfall water; the finer the soil particle, the lesser the chance of infiltration, and the coarser the soil particle, the higher the chance of infiltration. The soil media map has been prepared from NBSS & LUP published soil map (“NBSS & LUP, Indian Council of Agricultural Research, India,” 2001). The soil of this area comprises clay, clay loamy, and silts loamy and coarse loamy (Fig. 4d). Sandy loam is present to a small extent in the northeastern and southwestern parts; this type of soil is promoting heavy infiltration. Clay loam is present to a large extent in the northern, middle and eastern portions. Clay and silt loam also present in the area. The amount of infiltration is comparatively low in clay, clay loam and silt loam soil, sandy loam that signifies the relatively highest penetration and moderate rating of vulnerability (Table 3).

Topography

Topography plays a major role in the movement, magnitude of flow and direction of groundwater. Hence, its topographic features also control pollutant runoff (Thomas et al. 2017). The steep topography represents higher chances of runoff flow, and gentle surfaces represent the least runoff and higher chance of infiltration. The steep slope promotes less vulnerability, and the gentle slope promotes huge vulnerability. The topography of the region is divided into 4 classes: 0–2, 2–6, 6–12, and > 12 (Fig. 4e).

Impact of Vadose Zone

The vadose zone is situated above the aquifer and below the soil layer, and its effect on the aquifer emission potential depends on its permeability and media attenuation characteristics (Shrestha et al. 2016). In groundwater vulnerability analysis, the impact of the vadose zone is a very significant factor, so the rating of this parameter is 5. In the northeastern part of the study area, clay is observed and the remaining part contains sand, silt, and gravel, as a result maximum sections of the area face moderate vulnerability rating because the weight of the sand and silt gravel is 6 (Fig. 4f).

Hydraulic Conductivity

Hydraulic conductivity is a significant aquifer factor that regulates the rate of groundwater flow into the saturated zone and thus regulates contaminant flow direction (Jhariya et al. 2019). The speed at which groundwater travels, also influences the rate at which the contaminant is transported from where it enters the aquifer. Hydraulic conductivity has been determined from soil media, and this is calculated from Table 4 (Smedema and Rycroft 1983). The maximum value 1–3 is very small rating and has been found on the maximum places, and remaining part the hydraulic conductivity value is 0.5–2, 0.02–0.2 and below 0.02 (Fig. 4g).

Table 4 Ranges and ratings of the DRASTIC and modified DRASTIC parameters
Full size table

Land Use

Land use parameter is the important parameter for addressing the groundwater vulnerability. Land use is one of the major variables influencing the development of underground development (Das et al. 2019), and for the modified DRASTIC method land use parameter has been added with DRASIC parameter. In this work, built-up area and agricultural land provided a relatively high rating for their bad effect on groundwater and the vegetation area has been relatively low rating (Fig. 4h).

Results and discussion

Results

Groundwater Vulnerability Based on the DRASTIC Method

The DRASTIC model was used to generate a vulnerability index map of the research area in the GIS setting by taking the total sum of the product of the rate and weights applied to each parameter (Eq. 1). Higher index denotes higher chance of vulnerability. The DRASTIC map is generated by overlaying all seven parameters, and the DRASTIC index ranges from 86 to 167. The groundwater vulnerability area has been classified low (24.95%), moderate (44.30%), and high (30.74%) in three categories (Fig. 5a).

Fig. 5
figure 5

Groundwater vulnerability zone maps (a DRASTIC, b agricultural DRASTIC, c land use DRASTIC, d AHP land use DRASTIC)

Full size image

Groundwater Vulnerability Based on the Agricultural DRASTIC Method

The agricultural DRASTIC method has been differentiated with drastic methods by their different weights. For DRASTIC and pesticide DRASTIC, the four parameters, viz. Soil environment, topography, vadose zone effect and hydraulic conductivity, were regarded with separate weights. The mean of these four parameters showed the value of vadose zone in DRASTIC as maximum, while topography was most prominent in DRASTIC pesticide. After water depth, the soil media has the most prominent weight in agricultural DRASTIC (Table 1). In this model, 43.16% of the area is in low vulnerability, which is between 95 and 143, and 20.51% of the area is in high vulnerability, ranging from 170 to 197 (Fig. 5b).

Groundwater Vulnerability Based on the Land Use DRASTIC Method

A parameter describing land use was added to the assessment to change the original DRASTIC method. Weighing factors for land use have been identified for use in vulnerability mapping (Table 1). The land use DRASTIC vulnerability maps the area prepared by overlay analysis of all eight parameters. The computed Land Use DRASTIC index range is from 94 to 206 (Fig. 5c). In land use DRASTIC, the low-vulnerability index range is from 94 to 146, the area is 39% is higher than DRASTIC, the moderate-vulnerability index range is from 146 to 172 at the area of 42.65%, and the high-vulnerability index range is from 172 to 205 at the area of 18.34% (Fig. 6).

Fig. 6
figure 6

Percentage of vulnerability areas (%) a DRASTIC, b agricultural DRASTIC, c land use DRASTIC, d AHP land use DRASTIC

Full size image

Groundwater Vulnerability Based on the AHP Land Use DRASTIC Method

The range of factors for land use DRASTIC method has been modified to use an AHP to show a more rational technique for categorizing vulnerability in groundwater. For eight factors used in the revised DRASTIC process, the pair-wise correlation matrix was prepared and new rating coefficients were determined for all variables (Table 5). The rating coefficient used by the AHP land use DRASTIC system in the map preparation area is calculated by multiplying the rating values of each variable with the new rating value obtained. In this model, 61% of the area is at moderate vulnerability and 32% of area is at high vulnerability.

Table 5 K-value ranges by soil texture (Ritzema 2006)
Full size table

Single Parameter Sensitivity Analysis of the model

The single parameter sensitivity analysis evaluates the effective weight of the different parameters according to their theoretical weight. In this case, the effective weight corresponds to the function consisting of a single parameter compared to the other as well as the weight assigned to the model (Tomer et al. 2019; Brindha and Elango 2015). The theoretical and effective weights for DRASTIC, agricultural DRASTIC, land use DRASTIC and AHP land use DRASTIC are shown in Table 7. In DRASTIC, the parameter depth to water was the most prominent, as reflected by its effective weight (mean, 31.63%), which exceeded the theoretical weight by (mean, 21.7%). Likewise, impact of the vadose zone has also shown higher effective weight (mean, 21.34%), and also aquifer media has higher effective weight than theoretical weight. The net recharge and soil media, on the other hand, have shown significantly lower effective weight (mean, 15.62% and 4.61%) than the theoretical weight assigned (mean, 17.39% and 8.70%). Higher effective weight of the parameters like depth to water, impact of vadose zone and aquifer media highlighted their importance in the DRASTIC model output. Similarly in pesticide DRASTIC and AHP land use DRASTIC model, the depth of water has also higher effective weight than theoretical weight and the topography parameter is more important; the effective weight more excessive than theoretical weight, in land use DRASTIC depth of ground water and impact of vadose zone has higher effective weight.

Validation Through ROC Curve and Friedman Test

The measured AUC under ROC specifically defined the appropriateness of all models as the AUC of each model is nearer to 1% or above 0.85%. The values are 0.91%, 0.92%, 0.89% and 0.95% (Table 6) for DRASTC, agricultural DRASTIC, land use DRASTIC and AHP land use DRASTIC. Among this AHP land use DRASTIC is more accurate than other models with 0.95% accuracy rate. As a result, vulnerability assessment methods used in this study are considered to be more reliable (Figs. 7, 8).

Table 6 Pair-wise comparison matrix, weights and consistency ratio of the data layers used
Full size table
Fig. 7
figure 7

Model validation trough receiver operating characteristic (ROC) curve

Full size image
Fig. 8
figure 8

Field photography of the study area (a & b. Pond filled by garbage that spoil subsurface water, c & d excess pesticide pollute groundwater, e wastage material of household contaminant wetland, f Wetland encroached by agricultural land.)

Full size image

In the case of the Friedman test, also the same result is found. Mean rank value in the case of AHP land use DRASTIC model is identified as the best model. Table 7 shows that, the p value was < 0.05 (0.00); the null hypothesis was rejected, indicating significant differences between the four models of groundwater vulnerability.

Table 7 Statistics of single parameter sensitivity analysis
Full size table

Discussion

The study of the four different models was used to assess the groundwater vulnerability in Malda district of West Bengal. The vulnerability index values range low, moderate, and high in different models, the area under moderate to high vulnerability is mainly in a built-up area and agricultural land, and the low vulnerability area is seasonal agricultural land. To modify and for a genuine result, the land use layer added, as Noori et al. (2019) modified the DRASTIC and SINTACS methods added with a LULC layer and compared between these methods for southern Tehran aquifer vulnerability assessment, Iran. Related research by Gómez et al. (2017) has also stated the strong impact of human activities on groundwater quality as demonstrated by LULC changes in the vulnerability analysis. The outcome of this study is in harmony with the outcome of Singh et al. (2015) conducting a study in Lucknow City, India, uses a modified DRASTICA model (DRASTICA) in which anthropogenic influence contributed significantly to groundwater pollution. The groundwater resource threatened due to excessive use of groundwater by a human being, the use of pesticide and waste material thrown in water (Al-Rawabdeh et al. 2013). As a result of the increasing population, the dependence on groundwater also increases and the groundwater quality decreases vice versa and that impact on the ground resources (Lathamani et al. 2015). In this study AHP land use DRASTIC is also used to enhance the groundwater vulnerability. In this model weight distributed as per expert review and the groundwater depth, recharge soil media or soil texture and land use are more weight rather than other parameters. Jhariya et al. (2019) investigated the groundwater contamination vulnerability of Tandula watershed by modifying DRASTIC and the analytical hierarchy process. Lad et al. (2019) also use the AHP model in Deccan basaltic province in Pune district. To measure the effective weight of multiple parameters according to their theoretical weight, the single parameter sensitivity analysis has been done, and the two parameters depth to water level and impact of vadose zone were also found to be most effective in the DTASTIC and land use DRASTIC models. In the agricultural drastic model, the depth of groundwater and the topography are most important parameters, and in AHP land use DRASTIC depth of groundwater and net recharge parameter are most efficient. The models are validated by ROC curve and Friedman statistical test; both of this validation have a high accuracy of the model. The ROC curve of AHP land use DRASTIC model has high accuracy (AUC > 85%) (Tables 8, 9). Das and Pal (2020) used the ROC curve for model validation of groundwater vulnerability.

Table 8 AUC value of the ROC curve
Full size table
Table 9 Mean rank value of the computed models under Friedman test
Full size table

Conclusions

Groundwater is important to the source of water for human beings to act as a daily use. Even so, pollutants caused by human behaviors decline groundwater quality that poses various problems related to human health and socio-economic problems. Proper groundwater planning and management efforts require the delineation of vulnerable areas and the identification of contaminant sources. As a result, this study carried out DRASTIC and modified DRASTIC models to define groundwater vulnerability to pollution of the Malda district. The area of high groundwater vulnerability of DRASTIC, agricultural DRASTIC, land use DRASTIC and AHP land use DRASTIC is 30.74%, 20.51%, 18.34% and 32%, respectively. The northeastern and southwestern portion of this area is a high vulnerable area, and middle portion is designated as a moderate vulnerable zone. Based on two different validation tests, it can be defined that AHP land use DRASTIC is the best appropriate method to delineate groundwater vulnerability of this area. Also, the findings of this study recommend that the groundwater situation in the Malda district will become more vulnerable in the future considering land use change. There is also an instant step at vulnerable areas to promote suitable planning of groundwater resources and management, also prohibit the excessive use of pesticides in agricultural areas, and to decrease pollution-related anthropogenic sector in the built-up area.