Enhancing groundwater resource detection in Ahmednagar District through an integrated GIS framework utilizing AHP, SI, and FR models

AUC

area under the curve

DD

drainage density

GWPZ

groundwater potential zone

GPM

groundwater prospectivity model

LULC

land-use land cover

LD

lineament density

ROC

receiver operating characteristic curve

MODIS

moderate resolution imaging spectroradiometer

NDVI

normalized difference vegetation index

SRTM

shuttle radar topography mission

Groundwater is a vital aspect of the natural water cycle held beneath the water table in the pore spaces of soil/rock. The need for groundwater in recent years has impacted domestic, industrial, and agricultural requirements in developed and developing nations. Groundwater meets 80–90% of India's rural and 50% of the urban population's domestic water needs. Groundwater demand is rising to meet the water needs of diverse industries (Ghosh et al. 2020). Effective groundwater potential zones (GWPZs) for the management of agricultural land require sustainable irrigation practices (Dar et al. 2021).

Shallow, unconfined aquifers in weathered residue, deep cracks, and joints in semi-confined situations are the primary sources of groundwater (Das 2019). The distribution of groundwater in India changes with topography, geology, and climate. In areas with few surface water sources, groundwater is typically the primary source of fresh water. Urbanization, industrial development, and population growth have endangered groundwater resources. Fluctuations in the climate significantly impact the variables influencing the recharge of groundwater.

Identifying GWPZs is extremely useful in recognizing the potential for groundwater in the area. GWPZ ensures that groundwater management plans are carried out efficiently and effectively. GWPZ for artificial recharge processes must be determined using soft computing methods (Castillo et al. 2022). GWPZs have been determined using statistical, expert evaluation, and deterministic methods. Most research methods are based on bivariate and multivariate statistical techniques with pre-investigation assumptions and limitations (Doke et al. 2021). Remote sensing can identify potential groundwater recharge zones and suitable site structures for artificial recharge.

The study emphasizes the importance of precise groundwater occurrence in semi-arid locations (Ghosh et al. 2022). Remote sensing, geographic information system (GIS), and the analytical hierarchy process (AHP) have designated GWPZs in San Luis Potosi, Mexico. The results show that 26.30% of the region has moderate groundwater potential, while 68.21% has low potential. The accuracy prediction was satisfactory. The combined fuzzy-AHP process has been more efficient than the AHP and frequency ratio (FR) techniques. AHP, FR, and certainty factor (CF) modeled groundwater potential in Varamin Plain, Tehran Province, Iran. The FR model outperformed the AHP and CF models in groundwater resource management and land-use planning in the future (Abrar et al. 2023).

A total of 10 predictors from geological and remote sensing datasets investigated the groundwater prospects in the Akatsi Districts of Ghana. Groundwater prospectivity model (GPM) with FR-characterized potential groundwater zones (Rajesh et al. 2021). Soil properties significantly influenced the GPM, while the slope aspect had the least influence on the GPM. GPM's accuracy was determined to be a good (0.876) score. The prediction-area plot showed 76% of the groundwater locations found within the delineated zones covering 24% of the study area (Amponsah et al. 2023).

Potential water resources were predicted by GIS-based FR and machine learning in the Ningxia Hui Autonomous Region of China along the Yellow River. 40% of the wells were in very high to outstanding zones out of the six GWPZs discovered. Using area under the curve (AUC) analysis, the FR models were validated at an accuracy value of 0.759. The land cover characterized using Landsat data revealed changes in agricultural, anthropogenic, and water-related activities (Li et al. 2023). Combining GIS and machine learning approaches has proven dependable for identifying nonlinear groundwater activities. Machine learning algorithms like random forest classifier, support vector, and k-nearest neighbor classifier forecasted groundwater potential (Mohammadi-Behzad et al. 2019).

This research aims to ensure sustainable management and utilization of groundwater resources by producing a projected reference map for groundwater exploration and exploitation. It emphasizes the critical importance of timely groundwater potential assessment for resource planning and informed decision-making. This study bridges a critical gap in groundwater detection methodologies by integrating GIS with AHP, statistical index (SI), and FR models. It offers a novel framework for GWPZ mapping, specifically tailored for the Ahmednagar District. This integrated approach can be adapted to various regions facing similar hydrogeological challenges, providing a valuable tool for policymakers and water resource managers. The significant reduction in groundwater levels in Ahmednagar is attributed to the city's rapid urbanization and industrial expansion, compounded by the moderate rainfall patterns and restricted availability of surface water.

Effective groundwater recharge zonation solutions are desperately needed, as groundwater is becoming increasingly important for both residential and agricultural uses in the area. By offering a novel framework that combines GIS with AHP, SI, and FR models, this study enhances the accuracy of groundwater potential assessments and provides a scalable framework adaptable to various regions facing similar hydrogeological challenges. By addressing this gap, our study provides a valuable tool for water resource management and sustainable development, thereby making a significant contribution to existing groundwater detection methodologies.

Groundwater is a very important resource in the Ahmednagar District for agricultural and domestic purposes. The region is traversed by several rivers and streams, including the Pravara, Sina, and Ghod rivers. Each of these rivers supports the region's biological balance and water resources. Ahmednagar District's drainage patterns vary, with frequent dendritic and sub-dendritic patterns. The water source diversity in the study area has highlighted the importance of rivers and dams. Bhandardara, Ghatghar, Visapur, Mula, and Jayakwadi are some of the dams. These structures play important roles in the region's water supply and management. The depth of the water table in Ahmednagar District ranges from shallow (less than 5 m) to deep (more than 20 m). The area has a wide range of soil types. The most common soil types in Ahmednagar District are black soil (regur), red soil, and alluvial soil. The kind of soil in the region can have a considerable influence on agriculture.

The preprocessing step which comes after the data phase checks the accuracy, consistency, and uniformity of the input feature data. This phase includes data scanning, cleaning, digitization, and verification. The map created for the input parameters data has been transformed into standard format. The identified parameters impacting the groundwater potential were prioritized. The data prepared for the model include groundwater potential factors, a groundwater database, and well locations. The dataset was divided into training (70%) and testing (30%) sets to ensure the robustness of the model. This split allows for adequate training of the model while reserving a substantial portion of data for validation, as recommended by common practice in geospatial modeling (Kamaraj et al. 2023).

The selected GIS layers (rainfall, slope, elevation, LULC, soil, geology, and geomorphology) were chosen based on their significant impact on groundwater recharge as demonstrated in previous studies (Nithya et al. 2019; Maity et al. 2022). Rainfall influences recharge rates, slope affects runoff, elevation impacts water flow, LULC reflects land cover and use, soil type influences infiltration, geology determines aquifer properties, and geomorphology affects terrain and water accumulation. The training dataset has explored the FR and SI. This approach has forecasted the groundwater potential of a well placement (Bera et al. 2020). The model has been validated using the test dataset. The success rate and ROC-AUC have been analyzed to determine the accuracy level in the predicted groundwater potential map. A composite map depicting groundwater potential has been classified into five zones. The zone has been divided into five categories: very high, high, low, and very low. Table 1 shows the data sources for input parameters.

The AHP method determines the weights of several aspects and creates a schematic component diagram related to each other. One element's relative importance increases with its influence on other factors and thus results in a higher weight for that factor. The intensity of relevance is specified on the AHP grading scale. Every number on the scale from 1 to 9 represents a different degree of significance. ‘Equal importance’ is denoted by a value of 1, implying two components contribute equally to the evaluated goal. As one moves up the scale, a score of 3 denotes ‘Moderate importance’. Level 5 stands for ‘Strong importance’. Level 7 denotes ‘Very strong importance’. The maximum number on the scale is 9, denoting ‘Extreme importance’. Scale 9 strongly supports one aspect over another, achieving the highest degree of affirmation. A more nuanced evaluation of importance between the established levels is obtained by expressing intermediate values using the numbers 2, 4, 6, and 8. This scale has been utilized to help arrange the criteria hierarchically using a pairwise comparison matrix by defining the themes according to their rank and priority.

Identifying potential zones for groundwater is closely related to the considered nine parametric layers. The direct and indirect interactions have been evaluated between parametric layers. This assessment allocates the proper weights necessary for the ensuing overlay procedure. Each of the nine chosen parameters went through a ranking and weight assignment process to make the analysis for mapping groundwater potential easier. The criteria were ranked on a 5-point rating system where higher ratings indicate greater suitability or potential for groundwater.

The weighting distributions have been conducted following the ranking of individual parameters. A total weight of 100 has been assigned to the nine parameters. To determine the weightage, the AHP model was utilized. This process involves pairwise comparisons of the parameters to establish their relative importance, as shown in Figure 3. Each parameter was compared with the others based on expert judgment and previous studies related to groundwater potential assessment. The comparisons were then used to calculate the normalized pairwise comparison matrix to ensure that the sum of the weights equated to one, as shown in Figure 4. This step involved dividing each matrix element by the sum of its respective columns. The normalized matrix calculated the average of each row representing the weights of the parameters. These averages were the final weights assigned to each parameter. This method ensures that the weights are assigned systematically and reflect the relative significance of each parameter in influencing groundwater potential. The consistency of the judgments was checked using the CR, ensuring the reliability of the weight assignments. The weightage and the influence allocated to each input parameter are shown in Table 2.

is the method's GWPZ and denotes the total FR scores for all classes from each influencing component.

The correctness of the FR, SI, and AHP models has been verified by evaluating success rates using receiver operating characteristic (ROC) curves (Berhane & Tadesse 2021). The AUC measurements show the forecast's accuracy. 70% of the training and 30% of the testing datasets were used to compute these metrics. AUC values are between 0 and 1, where 1 denotes exceptional performance.

Following the detailed explanation of the methodology, we present the results. This section begins with an overview of the input parameters: rainfall, slope, elevation, land use/land cover (LULC), soil, geology, and geomorphology, followed by a comprehensive examination of the GWPZs identified by each model.

The region's rainfall is a result of the north-east and south-west monsoons. The rainfall map was constructed by interpolating the rainfall from the rain gauge stations, and the resulting map was divided into five categories. Isopleths like contour lines connect places on a map with identical rainfall values depicting the regional distribution of precipitation. Rainfall maps merge precipitation data at various time intervals of daily, monthly, or yearly averages. Data from weather stations, radar systems, weather satellites, and numerical weather models have been used to make rainfall maps.

Land's gradient or slope at various locations is a key factor affecting groundwater recharge in mountainous watersheds. Slope maps identify water flow patterns, landslip risk, and soil erosion possibility. Figure 5 indicates the ground surface undulations and depression storage. The study area is elevated to a range of 0–74%. The slope map is divided into five classes: 0–2.87% (flat), 2.87–7.76% (gentle), 7.76–15.25% (moderate), 15.25–26.73% (steep), and 26.73–73.29% (very steep).

From Table 3, 63.57% of the research area has been categorized as ‘excellent’ at gradual slope of (0–2.87%). The flat topography allows for greater infiltration and recharge. The study region has a 26.41% gentle slope (2.87–7.76%) with ‘good’ groundwater potential. About 6.20% of the area is moderate groundwater potential at moderate slope (7.76–15.25%). Steep class (15.25–26.73%) of 3% of the research area has been classified as having ‘poor’ groundwater potential. The Sahyadri mountain range (Western ghats) in the NW region has ‘Steep’ and ‘Very steep’ slopes. The very steep percentage (26.73–73.29%) is in ‘Very poor’ groundwater potential, which makes up only 0.82% of the research zone due to considerable surface runoff.

A slope map has analyzed the topographical factors impacting water circulation for groundwater potential. Locations with greater slopes have reduced groundwater recharge. Steeper slopes have resulted in rapid runoff, lowering the possibility of groundwater storage. Areas with gentle slopes promoted water penetration by the gradient of the landscape. Flatter terrain provides better water retention and infiltration prospects, indicating more groundwater potential.

An elevation map shows the altitude of the Earth's surface at various points in time. Climate, vegetation, drainage patterns, and other environmental processes are directly impacted by elevation change. As shown in Figure 5, the elevation ranges from 360 to 1,514 m. The slope map is divided into five categories in meters: 360–550 (flat), 550–639 (gentle), 639–752 (moderate), 752–957 (steep), and 957–1514 (very steep).

Table 3 shows that 36.91% of the research area has ‘excellent’ groundwater holding elevation ranging from 360 to 550. The study area of good GWPZ has a moderate elevation of 550–639. Moderate elevation and moderate GWPZ are found in 20.76% of the area. Low GWPZ in steep elevation covers 9.72% of the region at elevation values of 752–927. The western ghats 927–1514 have been categorized as the highest elevation, accounting for 2.06% of the research area and having a very low GWPZ.

Drainage density (DD) determined the relationship between landforms and water flow patterns in watershed management and flood risk assessment. DD quantifies the areal density distribution of natural drainage, such as rivers, streams, and other watercourses, as shown in Figure 5. DD offered a numerical number indicating the spacing of drainage structures in a specific location. DD with the highest value indicates the largest probability of runoff, resulting in less percolation. The study region's DD is categorized into five classes and varies from 0 to 1,570 m/m². 5.45% of the research region is covered by very high DD (0–364). The high DD of (364–665) covers 10.05% of the total area. Moderate DD (665–967) covers a 28.87% area. The low region DD (967–1268) makes up 37.46% of the study region. The 18.18% of the area has a very low DD at (1268–1570). The high DD has shown the presence of several watercourses which implies a great capacity for runoff and effective water drainage.

Lineaments are linear or curved natural or human-caused geological features, such as faults, fractures, folds, river channels, ridges, or highways. In hard rock terrain, lineaments are extremely important for replenishing underground groundwater. Groundwater potential is highest in the vicinity of lineaments. Figure 5 shows the geological location and direction of linear features or lineaments on the Earth's surface.

LD in the study area ranges from 0 to 885 m/m and is classified into five groups. Very low (0–177 m/m²) covers 38.61% of the total area. (177–354 m/m²) has a 29.19% study area classification of low. Class 3 is categorized as moderate, with an LD of (354–531 m/m²) for 20.42% of the total area. 9.03% of the area is covered by the high LD (531–708 m/m²). 2.75% is covered by the highest LD (708–885 m/m²). Lineament maps explain the geological processes, groundwater flow patterns, and structural characteristics of potential hazards such as earthquakes and landslides.

Land use/land cover (LULC) is a thematic map (Figure 5) that represents six groups: forests, natural vegetation, wetlands, crop and agricultural fields, urban and built-up areas, and water bodies. The study region has forests in the western Ghat region in 0.02% of the total area. Natural vegetation covers 2.09% of the study area. Wetlands and marshy areas surround the dam in a large part of the study area. Agricultural lands or crops cover 93.26% of the research region. The built-up area accounts for 2.39% of the total area. The city regions of Kopargaon, Sangamner, Rahata, and Nagar comprise the principal built-up areas. Water bodies comprise 0.81% of the study area in the eastern backwaters of Jayakawadi Dam.

Figure 5 shows the types of soils found in the study area. Soils are classified according to texture (sand, silt, or clay), color, organic matter concentration, pH, and nutritional composition. Soil maps have information regarding soil suitability for various applications such as agriculture, forestry, building, and septic systems. The infiltration rate is determined by the soil's permeability and water-holding capacity. The amount of water that reaches the water table is determined by the size of the unconsolidated zone under the soil.

Table 4 shows an abundance of loam, clay, clay loam, and sandy clay loam soil in the research region. The soil type known as clay loam is further divided into three categories: clay loam soil A, B, and C. There is very deep, clayey soil with a moderate to high percolation rate, including the areas of Kopargaon, Srirampur, Rahuri, Rahata, and Newasa, along with parts of Nagar, Shevgaon, Srigonda, Jamkhed, Karjat, and Pathardi. The majority of the Sangamner Taluka is covered in clay loam that has varied from shallow to extremely deep at a considerable percolation rate in parts of Parner, Nagar, Pathardi, Jamkhed, and Akole Taluka. A small piece of Akole Tehsil in the Northwest is covered with loamy soil with a modest depth and low percolation rate. A shallow to extremely deep sandy clay loam with a low percolation rate covers the remaining portion of Parner, Shrigonda Taluka.

The geological level of the study area (also known as the Deccan Trap) was formed by volcanic lava, i.e., basalt rock or Paleocene cretaceous extrusive rocks. The geology map shows the distribution, types, ages of rock formations, geological features, and subsurface structures. Geology maps use stratigraphic symbols and critiques to consider faults, folds, dikes, geological borders, unconformities, and the relative ages of rock layers. The basaltic lava flows originated from sporadic fissure-type eruptions between the upper Cretaceous and lower Eocene ages. A total of 19 main flows in the Deccan Trap occur in succession and range in elevation from 420 to 730 m above mean sea level. Vesicular and large basalt units are the dominant features of these flows. Recent-age alluvium as a slender stretch is deposited over the traps along the path of large rivers. Basaltic lava flows typically spread out horizontally over a large area, creating plateaus, commonly referred to as tablelands. These flows happen in stratified sequences with thicknesses varying from 15 to 50 m. Flows are denoted by vesicular portions at the top and massive portions at the bottom, and distinguished from one another by bole beds, which are marker beds. The thickness of weathering varies, ranging from 5 to 25 MBGL (Meters Below Groundwater Level) in the study area. The primary aquifer in the district is the worn and cracked trap found in topographic lows.

The creation, development, and characteristics of landforms, such as hills, plains, valleys, mountains, rivers, and other geological features, are known as geomorphology. Figure 5 shows the GWPZ's geomorphological characteristics alluvial plain, anthropogenic terrain, dam and reservoir, flood pain, low dissected hills and valleys, low dissected plateau, moderately dissected hills valleys, moderately dissected plateau, pediment pedi-plain complex, quarry and mine dump, waterbodies-other, and waterbody-river comprise the study region. Geomorphological mapping includes discrete landforms and structural elements, as well as the study of water movement and freezing and thawing.

According to a physiological perspective, the Ahmednagar District is a part of the Deccan Plateau. Sahayadri hill ranges are in part inside the district. The hilly Western Ghat segment of the Akole Tehsil spreads to the level Shevgaon and Jamkhed provinces in the east. Three ridges, Kalsubai, Baleshwar, and Harishchandra, split off from the main Sahayadri range to the east. Geographically, the district is classified into four major characteristic landforms: The foothill zone (18.3%), plateau region (4.9%), plains (68.70%), hills, and ghats (8.1%) of the study area. The maximum area of a moderately to heavily fragmented plateau dominates the district.

The research region comprises 35.34% of the Parner Nagar, Shrigonda, Sangamner, and Kopargaon tehsils under a ‘Low’ GWPZ. These tehsils experience constant ‘Low to Very Low’ rainfall due to their location in a rain shadow zone. The Western and Northern parts of Sangamner and Kopargaon tehsils also experience less rainfall due to their position in the rain shadow region. The dwindling number of lineaments, high DD, and low percolation rate contribute to the overall low groundwater potential in these areas.

The research area in the district has moderate groundwater potential, accounting for 43.47% of the total area. This includes the tehsils of Sangamner, Rahuri, Rahata, Kopargaon, Nagar, Karjat, and Jamkhed. The Mandohol Dam and canal water from Pimpalgaon Joga Dam contribute to this moderate groundwater presence. The LD in the central and southern regions ranges from moderate to very high, suggesting faults, ridges, and fractures for groundwater percolation. The DD also ranges from moderate to very high, resulting in significant water runoff. The region's flat slopes mitigate this risk.

The Eastern section of the research area of about 21.09%, including Nevasa, Shevgaon, Shrirampur, Pathardi, and Rahuri and Nagar tehsils, has a high GWPZ due to its LD ranging from ‘Low to High’ and high to moderate percolation rates. The southernmost tehsil, Karjat, has a high density of lineaments and low to moderate rainfall, contributing to its high groundwater potential. Approximately 0.06% of the research area in Pathardi and Newasa, Eastern India, was identified as having a very high GWPZ, with ideal drainage and LD for optimal groundwater percolation conditions. The soil's high percolation rate increases groundwater potential. Key groundwater sources in this zone are Pimpalgaon Lake in Newasa and Kapurwadi Lake in Pathardi Tehsil, aligning with previous studies. This degree of accuracy aligns with results from related studies (Cheng et al. 2024; Ozegin et al. 2024). This methodology has been employed previously to determine GWPZ through remote sensing and GIS research (Dar et al. 2021). Results are comparable with the spatial trends found in more research achieved by (Aydi 2018; Bera et al. 2020; Arunbose et al. 2021; Aykut 2021).

FR values were determined using association with wells for each themed layer. The rainfall ranges of 931.14–1,184.30 and 1,184.30–1,473.61 have yielded the highest values of 11.61 and 19.73. For a slope of 4.68, the highest FR value varies from 7.76 to 15.52. The maximum FR value of 1.71 has been indicated by an elevation variation of 550–639. FR values of 4.68 have been reported for 0–364 (m/m²). The wetlands have the highest FR of 12.25, according to the LULC research. The highest FR value of 12.36 is in clay soil types, as shown in Table 6. Class 531–708 has the highest recorded FR of 4.49 for LD. The moderate zone for groundwater potential has indicated the lowest FR values of 1.00 found in Paleocene cretaceous extrusive rocks. The alluvial plain has higher FR values (24.73) than other geomorphic features in the correlation between groundwater potentiality and geomorphic characteristics.

Five classifications (very low, low, moderate, high, and very high) comprise the GWPZ map as shown in Table 5 and Figure 6(b). The presence of western ghat in Akole Tehsil has resulted in a ‘Very Low’ GWPZ in around 0.04% of the research territory. The ‘Low-Moderate’ GWPZ includes Parner Nagar, Sangamner, and Shrigonda tehsils, as well as the northern and western parts of Sangamner and Kopargaon tehsils. These tehsils make up 25.51–52.74% of the research region. These tehsils are in a rain shadow area, they consistently get ‘Low to Very Low’ levels of rainfall. Sangamner and Kopargaon tehsils' western and northern regions receive moderate rainfall. There is a low-to-moderate groundwater potential in the area due to the high DD, low percolation rate, and few lineaments (Maity et al. 2022; Amponsah et al. 2023).

A high GWPZ of 21.61% is in the eastern section of the research area, comprising the center portions of Karjat Tehsil. This covers the tehsils of Nagar, Shevgaon, Shrirampur, Pathardi, Rahuri, and Nevasa. LD on the eastern side varies from ‘Low to High,’ with most of the area falling into the ‘High to Moderate’ zone. A very high GWPZ, with the right drainage and LD for perfect groundwater percolation conditions, makes up about 0.10% of the research area in Pathardi and Newasa, Eastern India. The high rate of percolation of the soil raises the potential of groundwater. Pimpalgaon Lake in Newasa and Kapurwadi Lake in Pathardi Tehsil are important groundwater sources in this zone, demonstrating the tremendous potential of the area (Elvis et al. 2022; Ghosh & Bera 2023).

A positive SI value indicated the best potential of locating groundwater in the region, while a negative value indicated a poor chance of locating groundwater. As indicated in Table 6, the region's rainfall had maximum SI values of 2.98 and 1.78, respectively, ranging from 1,184.30 to 1,473.61 and from 1,473.61 to 1,971.92. The negative results of low-value SI for rainfall classes indicated a low chance of groundwater potential (Abdo 2022). A relationship between slope and groundwater revealed a maximum SI value of 1.54 for areas with slopes ranging from 7.76 to 15.52%, followed by places with slopes ranging from 15.52 to 26.73% and SI values of 0.88 for moderate slopes, as shown in Table 6. The highest SI value of 0.54 was obtained for low elevations of 550–639 m, indicating a high potential for groundwater occurrence in the region (Table 6). For DD, the greatest value of SI 0.98 is seen in class 0–364, indicating a greater probability of groundwater potential.

The LULC parameters indicated a positive SI value for urban and built-up areas, followed by wetlands regions. LD showed a positive maximum value of 1.50 for classes 531–708, indicating the largest groundwater potential. The soil classes and groundwater yielded a positive maximum value of SI 2.51 for clay soil. In the case of Paleocene cretaceous extrusive rocks, the correlation coefficient between the region's geology and the probability of groundwater was positive. The region's geomorphology and groundwater potentiality analysis indicated that the pediment pedi-plain complex had a higher SI value (3.21) than other geomorphic characteristics (Abdo et al. 2022).

The GPMZ map, displayed in Figure 6(c), is categorized into five classifications as shown in Table 5: very low, low, moderate, high, and very high. The GWPZ map generated by the SI model revealed that 2.43% of the research region has ‘Very Low’ groundwater potential, including Akole Thesil. The ‘Low’ GWPZ covers 9.60% of the research area, including the majority of the Akole Thesil and a piece of the Sangamner and Parner thesils. A sizeable portion of the study area has moderate groundwater potential, accounting for around 57.36%. The high zone in the eastern portion of the study region, covering 24.97%, falls under the ‘High’ GWPZ. 5.37% of the area covered by the research region falls within the ‘Very high’ GWPZ. The significance of the statistical model has been highlighted recently in hydrological issues (Yaseen 2024).

The research conducted by the AHP model in the region indicated that 0.03% (6.82 km²) exhibits very low potential, while 35.34% (6,024.15 km²) of the region demonstrated a low potential for groundwater. A sizable section of 43.47%, or 7,410.01 km² was under moderate groundwater potential. The research findings highlighted that 21.09% (3,595.42 km²) of the region has high groundwater potential, while a very high potential for the region covers 0.06% (10.22 km²). In contrast, the FR model results reveal that 0.06% (10.23 km²) of the study area has very low groundwater potential, while 25.50% (4,346.89 km²) exhibited low groundwater potential. The largest portion of the study area, 52.74% (8,990.39 km²), demonstrates a moderate groundwater potential. About 21.60% (3,682.07 km²) was identified as high groundwater potential, and 0.10% (17.05 km²) as very high groundwater potential.

Conversely, the SI model indicated that 2.43% (414.23 km²) of the study area has very low groundwater potential, while 9.60% (1,672.27 km²) falls under the low groundwater potential category. A sizeable portion, accounting for 57.36% (9,777.94 km²) of the study area, was classified as a moderate groundwater potential region. 24.79% (4,225.86 km²) of the study area was identified as having high groundwater potential, and 5.37% (959.72 km²) of the study area was designated as having very high groundwater potential. This analysis shows that the SI model's ROC-AUC value of 0.93 outperforms the AHP (0.81) and FR (0.87) models, indicating its superior reliability. This finding is consistent with previous studies that also highlight the effectiveness of the SI model in groundwater potential zonation (Pawar et al. 2024).

The findings of this study provide a critical tool for policymakers and water resource managers in the Ahmednagar District. The superior reliability of the SI model in delineating GWPZ enables these stakeholders to identify and prioritize areas for groundwater development and conservation. This information is invaluable for formulating effective water management strategies, ensuring sustainable agricultural practices, and mitigating water scarcity issues. The results generated by the SI model will assist regulatory bodies in devising policies that enhance water resource utilization and protect groundwater reserves.

The FR approach located groundwater pockets in the hard rock terrain of Chhotanagpur, including geology, hydrogeomorphology, rainfall, soil texture, slope, DD, groundwater fluctuation, and land use/cover. 70 and 30% of the 56 tube wells were utilized as testing and training samples, respectively. According to the results, the south-west region had very low potential zones, while the southern region had very high zones (Elvis et al. 2022). The FR model's accuracy rate is 79.1%, indicating good prediction rates for sustainable groundwater development and management (Ghosh & Bera 2023). The LSM (Landslide Susceptibility Mapping) using the FR framework provided the best accuracy in predicting (AUC = 0.824) compared to the SI model (0.801 AUC). Based on success rate graphs, the AUC for the FR and SI models were 0.799 and 0.778, respectively (Abdo 2022). The study used SI approaches and logistic regression to determine landslide susceptibility in 470 out of 658 landslides. With success rates of 0.793 and 0.811, respectively, the results demonstrated high prediction curve rates (AUC) for both approaches. The SI and logistic regression R-index values were 88.54 and 83.66, respectively, showing highly vulnerable hazard classes. Several non-topographic elements have also been considered, including geology, land use, precipitation, and the distance from rivers in the study (Rai et al. 2022). A study in Nigeria mapped GWPZ in the Edo North region, revealing moderate to high recharge and storage potential in the eastern and northern sections (Cheng et al. 2024). This finding aligns with a study in India's Kangsabati River basin, where low-lying flat plains and a centrally located dam were found ideal for groundwater recharge. GIS maps are useful for hydrogeological studies and identifying suitable bore well/dug well sites.

This study underscores the significance of an integrated RS-GIS approach coupled with AHP, FR, and SI models for practical and cost-efficient groundwater recharge. The significant decline in groundwater levels in Ahmednagar has been ascribed to the city's swift industrial and urban growth, which is further exacerbated by the city's moderate rainfall patterns and limited surface water supplies. Since groundwater is becoming increasingly vital for local residential and agricultural applications, effective groundwater recharge zonation solutions are needed. This work emphasizes the value of using AHP, FR, and SI models in conjunction with an integrated RS-GIS method for realistic and affordable groundwater management. The best models and recharge zones were found using this integrated framework, and thematic data layers were created to make potential zone demarcation easier. Rainfall, slope, elevation, DD, LULC, LD, soil, geology, and geomorphology are thematic layers that have a substantial impact on the GWPZ. The study area's final GWPZ map was divided into five zones: very high, high, moderate, low, and very low.

The results were validated using ROC-AUC analysis. The SI model demonstrated superior reliability with an ROC-AUC value of 0.85, indicating a high level of accuracy in GWPZ prediction. This was further validated through field verification and comparative analysis with existing groundwater wells, confirming the model's effectiveness in identifying potential zones (field verification details: 69 wells, 85% accuracy rate).

There are several promising avenues for future research in GWPZ mapping. Investigating the integration of novel remote sensing techniques and geospatial data sources may improve the spatial resolution and accuracy of groundwater potential assessments (Sharma et al. 2020). Using emerging technologies like hyperspectral imaging and LiDAR data fusion shows promise for capturing fine-scale groundwater-related features and improving model performance. Investigating the effects of climate change and land-use dynamics on groundwater dynamics is critical for developing dependable predictive models. Future research could investigate incorporating climate projections and land-use scenarios into groundwater potential mapping frameworks to predict potential shifts in groundwater availability and quality under various climate and land-use change scenarios. More comprehensive validation studies are required, considering temporal variability in groundwater conditions. Longitudinal studies of groundwater dynamics over time can provide valuable insights into the temporal stability of GWPZs, as well as identify trends and anomalies that may affect model performance.

Exploring interdisciplinary approaches that incorporate hydrological, geophysical, and socioeconomic factors can help us better understand groundwater systems and inform more comprehensive groundwater management strategies. Collaborative research efforts that cross disciplinary boundaries can help to develop integrated modeling frameworks capable of capturing the complex interactions between natural and human-induced factors that influence groundwater dynamics. Promoting open-access data repositories and collaborative research networks can facilitate knowledge sharing and replication of findings, thereby increasing transparency and reproducibility in groundwater potential mapping research. Future research efforts that encourage data sharing and interdisciplinary collaboration can advance our understanding of groundwater systems and contribute to sustainable groundwater management practices.

While this study provides a comprehensive approach to GWPZ mapping, it is limited by the selection of GIS layers, which may not capture all relevant variables affecting groundwater recharge. Future research should consider incorporating additional variables such as land subsidence, human extraction rates, and climate change impacts. Moreover, the study's scope is confined to the Ahmednagar District, and further research is needed to test the model's applicability in different hydrogeological settings.

To address these limitations, future research should consider incorporating a broader range of variables that impact groundwater recharge, such as detailed geological data, vegetation indices, and human impact metrics. Utilizing higher resolution and more accurate data will improve the reliability of groundwater potential mapping. Efforts should be made to obtain and use the best available data. Implementing temporal analysis methods to account for seasonal and annual variations in groundwater recharge can provide a more dynamic understanding of GWPZs. Developing and testing new models or refining existing ones can help reduce subjectivity and improve accuracy. Increasing the extent and rigor of field validation efforts will help verify and refine the results obtained from remote sensing and GIS analyses.

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

The authors declare there is no conflict.