Application of decision support system/remote sensing/GIS techniques in groundwater recharge assessment

One of the most important worldwide problems is the lack of fresh water, particularly in arid and semi-arid regions. Globally, groundwater has a significant impact on personal, economic, and environmental well-being (Mallick 2021). However, excessive consumption of groundwater has a direct impact on sea level rise, making it the most prominent human signal in the hydrological cycle (Koutsoyiannis 2020). Water resource management is greatly affected by the Hurst Phenomenon for the River Nile (Hurst 1951). According to (Di Baldassarre et al. 2011; Dimitriadis et al. 2021), better prognoses of severe droughts and floods required for agricultural planning, water resource allocation, and infrastructural development are possible only through an improved comprehension as well as Long-Term Plan (LTP) modeling in the River Nile's hydrological flow system. Besides, this has consequences on the designs of dams and reservoirs because such traditional models lack LTP which may make them underestimate likelihoods that are extreme either low or high. The broader hydrological literature suggests that the research on the Nile and its demonstration of LTP has given rise to more advanced stochastic models which take account of this kind of memory enhancement, leading to improved forecasts and utilization of water resources in the Nile basin and the world over. This has greatly improved our knowledge of how water moves throughout the years and how climate-related events differ in time.

After the River Nile, groundwater serves as Egypt's second source of fresh water. The Egyptian government has been attempting to increase the area of its fertile region for many years by reclaiming desert territory using groundwater or River Nile channels. The largest reclamation project in Egypt is Tushka, which, is a part of a larger effort to relieve congestion in the Nile River Basin and aims to create an extensive agricultural zone throughout Upper Egypt's arid region. The Tushka project, where groundwater is the primary supply of water, includes the region west of Lake Nasser. Groundwater exploitation in this area has already started through drilled wells for a considerable amount of time without effective management and evaluation of groundwater recharge, leading to a reduction in the groundwater yield of some wells over time. In this context, the identification and evaluation of groundwater prospective regions is vital for enhancing projects' long-term profitability, lowering the danger of a water shortage, and decreasing drilling costs.

Traditional groundwater research methods, such as geophysical techniques, and data from boreholes, are expensive and time-consuming. Alternative methods for mapping recharge groundwater zones include the integration of geographic information systems (GISs) with remote sensing data and decision-support system (DSS) methodologies (Al-Shabeeb et al. 2018). Weighted overlay (influencing factor) is one of them that is frequently used by researchers (Riad et al. 2011). For quick evaluation, accurate forecast, and long-term planning of groundwater, Analytical Hierarchy Process (AHP), one of the DSS methodologies analyzing several aspects, was initially developed by (Saaty 1990). The AHP model has been used in several research all over the world to outline groundwater recharge maps (Rahmati et al. 2015). Other approaches were also used such as Fuzzy Analytical Hierarchy Process (FAHP) (Maity et al. 2022), Shannon Entropy (SE) (Al-Abadi et al. 2016), Multi-Influencing Factors (MIFs) (Anbarasu et al. 2020), and Frequency Ratio (FR) (Abu El-Magd & Embaby 2021).

Groundwater recharge has recently been addressed by a number of methodologies and approaches, including the logistic model tree (Rahmati et al. 2018), and the certainty factor (Razandi et al. 2015). Applying GIS indicators, geostatistical analysis, and the stochastic model are some other methods that have been utilized to evaluate water stress (Witkowski & Hejmanowski 2020).

Selecting the best parameters for regulating groundwater recharge is crucial in locations with moderate rainfall, such as the region northwest of Lake Nasser, where hydrogeological and structural influences become more powerful. In this research, the properties of the aquifer as well as the hydraulic interaction between groundwater and Lake Nasser water have been studied (Kim & Sultan 2002; Abdelmohsen et al. 2020).

The main objective of the current research is to define the groundwater recharge zones west of Lake Nasser, Egypt, using AHP, FAHP, FR, SE, and MIF models.

The north of the study area is dominated by Limestone formations (Figure 1(b)). The Precambrian basement through Cenozoic sedimentary succession are the exposed rock units in the area. The Cretaceous sandstones of the Abu Simbel, Sabaya, and Lake Nasser Formations are overlain by Mesozoic successions of limestones in the area covered by the Precambrian basement rocks (Arabian Nubian Shield), which include granites, granodiorite, gneiss, and schist (Stern & Kroner 1993). The area is covered by Tertiary Dakhla, Kiseiba shale, Kurkur, Garra, and Thebes formations of Cenozoic rocks. The main part of Playa deposits surrounds Toshka lakes in the center of the study area, and the other part is overlain by Kiseba formations in the northeast of the main part. The Timsah formations lie in the northeast of the study area covered by the Sabaya and Kiseba formations, while the Travertine formations are located in the west of the Timsah formation and east of the Limestone formations. The entire succession is then covered by Quaternary deposits made up of piedmont gravels, sand sheets, Nile deposits, and tuff. Oligocene and Quaternary volcanic intrusions are also present (Darwish 2013). The research area is characterized by arid conditions with sporadic rainfall and is typically part of the North African dry belt.

The region has both a lengthy, hot summer and a mild winter, with mean air temperatures ranging from 14.2 to 38.4 °C. The primary groundwater aquifer in the region is represented by Nubian sandstone (Moneim et al. 2014). It directly covers the basement rocks and is exposed to the study area's surface. It is made up of claystone inter-beds and sandstone with fine to extremely coarse grain. At some locations in the region, it has a hydraulic connection to the water of Lake Nasser. The aquifer has an average effective porosity of 26% and a hydraulic conductivity of 4.5 m/d (Ghoubachi 2012).

Because geology and lithology regulate the groundwater's subsurface flow, surface lithology is a critical component of groundwater recharge. The Egypt geological map was used to digitize the lithology layer (Egy_Geo_Conoco 1984; RIGW_ Hydogeological map of Tushka area 2016).

Based on how it affected the recharge of groundwater in the research region, the lithology was divided into five groups as shown in Figure 3(a). Due to the igneous and metamorphic rocks' limited permeability, it is believed that groundwater activities are lower toward the north and west (subclass 1). Impermeable units include gneiss, granodiorite, and granite. The sand sheets and clay deposits were thought to be inappropriate locations for drilling wells (subclass 2) despite the geological successions having moderate to high groundwater recharge. Travertine and the limestone aquifer are regarded as modest groundwater recharge (subclass 3). The groundwater recharge of the wadi deposits is above moderate (subclass 4). The areas where Nubian sandstone rocks were present were thought to have a significant recharge for groundwater (subclass 5). The previous five subclasses are shown in Figure 3(a).

Data on study area LULC were taken from ESRI [Link1] Sentinel-2 Land Cover Explorer (pixel size = 10 × 10 m) (Karra et al. 2021). As shown in Figure 3(b), there were five main land use/cover classifications identified in the region. Among these are water bodies and flooded vegetation (2.07%), bare ground (85.21%), crops and trees (1.77%), shrubs (10.92%), and built-up regions (0.03%).

Soil data for the study area were obtained from the ‘Digital Soil Map of the World – ESRI Shapefile format’ from the FAO website [Link2]. Additionally, Diva-GIS data were used to derive soil classifications as shown in Figure 3(c). Water bodies have the highest rate of infiltration; loam, sandy loam, and clay loam follow in order of importance.

Water from Lake Nasser replenishes the Nubian aquifer in the research region, raising the aquifer's groundwater level. The Nubian aquifer and Lake Nasser are hydraulically connected, as evidenced by the fact that the hypothetical salts that were discovered in Lake Nasser were identical to those found there (Ghoubachi 2012). According to Kim & Sultan (2002), Nasser Lake affects the groundwater level up to 30 km away from the lake. ArcGIS's distance extension and buffer tools were used to extract the distance from the lake layer and reclassify it into five categories as shown in Figure 3(d), where the likelihood of groundwater recharge increases with decreasing distance from the lake.

Using the spatial analysis function in ArcMap, topography and slope layers with a 30 m resolution were produced from NASA's Shuttle Radar Topography Mission (SRTM). A ‘fill’ function was applied before processing the digital elevation model (DEM) to get rid of pixel value issues. Groundwater is less likely to exist in high-altitude areas due to gravity action (Alikhanov et al. 2021). Elevations vary dramatically from 81 to 500. While low-land regions are dispersed along the study area's periphery, particularly close to the lake, the highly elevated portions are found in the northern part of the study area (see Figure 3(e)).

Figure 3(f) shows that the slope of the majority of the study area ranges between 1.5° and 4.0°. However, the slope gets steeper near the lake.

Based on prior research and the interpolation of data from pumping test analysis, the hydraulic conductivity layer was developed (Sallam 2006; Aggour et al. 2012). The area's hydraulic conductivity values range from 0.42 to 21.25 m/day, with the lowest values corresponding to wadi deposits, playa deposits, and Limestone aquifer, and the greatest values corresponding to regions where the sandstone aquifer, Timsah formations and most of Nubian aquifer are predominated. The resulting hydraulic conductivity map was divided into five groups and resampled as shown in Figure 3(g).

A high density of lineaments is a characteristic of zones with considerable groundwater recharge. Lineaments were manually digitized from the Egyptian geological map (Egy_Geo_Conoco 1984) and from hill shades from SRTM DEM, and automatically recovered from the Landsat 8 OLI picture (Band 8) using the LINE module of PCI Geomatica program (Link3). The ArcGIS environment was used to import and analyze the linear structures to produce a lineament thematic map, which was then utilized to produce the lineament density map in ArcGIS. As shown in Figure 3(h), the research region was divided into five groups and had lineament density that varied from 0 to 0.54 km/km².

High drainage density indicates less infiltration and is correlated with the infiltration rate. By statistically dividing the sum of the lengths of all the streams in a drainage basin by its entire area, the drainage density is determined. Using the ArcGIS platform's line density tool, the drainage density map was created from the drainage map as shown in Figure 3(i). The research area's densities varied from 0 to 2.02 km/km². For groundwater recharge, high weight was allocated to low drainage density and low weight to high drainage density.

Storms recharge groundwater levels by charging aquifers, especially the shallow ones, after flooding streams to increase groundwater supplies.

To depict the value of the precipitation in the study area:

• 1. The monthly precipitation data from the CRU TS version 4.07 [Link4] (Harris et al. 2020) were gathered from 1 January 1901 to 31 December 2022 and were downloaded as one file in Nc format (type of data is multi-dimensional raster).

• 2. From the Arc-toolbox we selected to make the Net-CDF Raster Layer, while the Band dimension was chosen by time. The number of bands in this file equals the number of months from January 1901 to December 2022 (each band consists of 1 month).

• 3. Every 12 consecutive bands give 1 year. So, by composite bands tool, we gathered every 12 bands to give 1 year (each year now is in a single raster file consisting of 12 bands).

• 4. By applying cell statistics tool on each annual raster file, we get an annual raster file with one value for each cell in this raster which represents the annual precipitation data. Then, by applying a raster calculator and taking the average for all annual precipitation years raster files to represent the best prediction of the final map on this study area.

• 5. Convert the final raster file to a point shapefile (each point value equals to precipitation value).

• 6. Applying the inverse distance weighted (IDW) method to get the final map of rainfall (by interpolation) as shown in Figure 3(k).

Groundwater was given a high weight and high recharge for recharging due to the high value of precipitation. According to the geographical distribution, the north region has a greater average rainfall intensity than the rest of the region as shown in Figure 3(k).

Additionally, to obtain detailed information on rainfall patterns, we also used seasonal rainfall maps for the period 1901–2022 for the four seasons (Winter–Spring–Summer–Autumn) as shown in Figure 4(b)–4(i). These maps provide a detailed insight into rainfall of seasonal changes. Understanding seasonal rainfall distribution is essential for estimating groundwater recharge rates and water resources management. Seasonal maps reveal the seasonal distribution, providing important information for making informed decisions on groundwater management and mitigation of groundwater scarcity.

The curvature of the land surface has a significant effect on infiltration and accumulation rates. The DEM is used to create the land surface curvature layer, which is divided into five classes: very concave, concave, convex, and flat and very flat as shown in Figure 3(l). The regions that would catch the water resources produced from the precipitation are defined by the curvature of the land surface (radians/100 m). Convex regions are less able to produce infiltration capacity than curved and flat surfaces. Since flat and concave land surfaces are conducive to water accumulation, locations with high curvature values were given higher weight values, and vice versa (Mukherjee & Singh 2020a).

The existing depressions in the study area act as natural recharge bonds. Depths and locations of Depressions were determined using a spatial analysis by utilizing the SRTM DEM to subtract the original DEM from the filled-DEM. The subclasses are shown in Figure 3(m).

The TWI describes the amount of flow accumulation at a certain location in the study area and the water's propensity to flow downward against gravity (Ghorbani Nejad et al. 2017), which speeds up the accumulation of flow and may also be used to define the moisture conditions of an area (Mukherjee & Singh 2020). The TWI has been used in several research to locate possible groundwater regions (Al-Abadi et al. 2017; Mukherjee & Singh 2020).

The depth-to-water map of the area illustrated in Figure 3(o) was prepared using the annual measured head in observation wells during 1998–2015 which their number was 70 observation wells, was collected from the Ministry of Water Resources and Irrigation (MWRI) (Zakaria et al. 2023) which this reference used some of the observation wells to study a portion of the study area. The gaps in the null data areas were filled using the IDW interpolation technique of the spatial analysis tool in ArcGIS. The research area's depth to water varied significantly; it was between 8 and 190 m.

Following the AHP model by (Saaty & Katz 1990), Using the AHP model, each factor's weight (w) was calculated in three basic phases:

• 1. The creation of the matrix that compares each of the (15) influential parameters pair-wise. Saaty's 1–9 scale was used to rank the relative relevance of each attribute on groundwater recharge. Criteria with a rank of 9 indicate that they have a very strong impact over the other, while a rank of 1 indicates that the two compared characteristics are equally significant. The evaluation of prior research, expert views, and field experience were used to establish the significance of each element. As a result, all relevant parameters were graded in relation to one another in the pair-wise comparison matrix.

• 2. After generating the normalized pair-wise comparison matrix, the normalized weights are determined. Equations (4) and (5) were used to calculate the consistency ratio (CR), which was used to verify the consistency of the generated matrix. According to (Saaty & Katz 1990), if the CR is less than or equal to 0.1%, the model prediction can be considered valid and trustworthy. On the other hand, the prediction model must be reevaluated if the CR exceeds (0.1).

• 3. Every factor was divided into smaller subclasses and given a ranking according to how much of an impact it had on groundwater recharge. The class with the least amount of impact got the lowest grade, while the class with the greatest influence got the highest. Lastly, each rank value was divided by the sum of all ranks for each factor to produce the normalized rate I of each subclass.

  

  (4)

  where λ_max is the principal eigenvalue of the matrix, which is equal to 16.03, this number is described in the following equation:

  

  (5)

CR is the consistency ratio, n is the number of parameters applied in the matrix, CI is the Consistency index, and RI is the Random consistency index.

FAHP is widely used in physical and environmental fields to analyze environmental vulnerability, flood susceptibility, and groundwater recharge mapping (Şener et al. 2018). There are several steps to delineate the groundwater recharge map using FAHP as follows:

• (1) Creating the pair-wise comparison matrix like the AHP approach, then replacing the numbers with the fuzzy number as shown in Equation (6):

  

  (6)
• (2) A geometric mean is used to calculate the weights (Buckley 1985), as shown in Equation (7):

  

  (7)
• (3) Calculate the fuzzy weights w_i for each factor:

  

  (8)
• (4) De-fuzzification to get crisp numerical values for the weights as shown in Equation (9):

  

  (9)
• (5) Calculating normalized weights as shown in Equation (10):

  

  (10)
• (6) Calculate groundwater recharge index =

  

  (11)

The thematic layers of conditioning factors were ranked and assigned classes using Saaty's scale as shown in Table 1 where only the odd numbers of Saaty were selected.

The subclasses were selected based on natural break classifications by editing the properties of each thematic layer raster to classify them as natural breaks method.

A, B, and W_i stand for the major, minor, and individual parameter weights, respectively.

The suggested score for each layer was used to calculate the weights of its subclasses. The weight of the concerned layer (W_i) was equal to that of the first subclass (C_i1).

After dividing the W_i value by the total of the layer's subclasses, P_i was calculated, subtracted from C_i1, and allocated to the second subclass.

GWRZ indicates the groundwater recharge zone, W_i is the weight of each theme layer, and R_i is the rank of each subclass inside the thematic layer.

The AHP, FAHP, FR, SE, and MIF integrated techniques were evaluated using a weighted linear combination in the ArcGIS environment for analyzing groundwater recharge maps of the study area. Figures 5(a)–5(e) show four primary classifications representing: low, moderate, high, and extreme using the five aforementioned techniques respectively. Displays the computed results for each technique, together with the proportion of area covered by these classifications.

Regarding Figure 5 and Table 2, the following can be noticed:

• (1) For the AHP method:

• • • 47.53% of the entire area has intermediate groundwater recharge.

  • • Low, high, and extreme susceptible zones represent 17.3, 28.3 and 6.87% of the total area, respectively.

  • • Areas dominated with Nubian aquifer and sand sheets, slope areas more than 10°, drainage density of <1.3 km/km², elevations <75 m, NDVI > 0.3, hydraulic conductivity >7.8 m/day, distance from lake <10 m, waterbodies, trees, and crops are all considered to be in the high to extreme zone for groundwater recharge.

  • • The findings verified that lithology, LULC, topography, slope, hydraulic conductivity, NDVI, drainage density, and lineament density are the main factors influencing groundwater recharge in the AHP method.

• (2) For the FAHP method:

• • • 32.5% of the total area has low groundwater recharge.

  • • Medium and high susceptible zones, 25.47 and 25.04%, are relatively close to each other.

  • • The extreme part consumes 16.98% of the total area.

  • • The groundwater recharge findings from the FAHP method are controlled mostly by lithology, LULC, soil, topography, slope, hydraulic conductivity, NDVI, drainage density, and lineament density.

• (3) For FR method:

• • • 41.68% of the total area has low groundwater recharge.

  • • 17.86, 27.77, and 12.67% of the total area are in the medium, high, and extreme susceptible zones, respectively.

  • • Areas dominated by clay loam, loam, and Limestone aquifers are considered to be in the low recharge zones.

  • • The high to extreme zones for groundwater recharge are near the same in relation to the AHP method, in addition to Lake Nasser formations.

  • • The findings verified that lithology, LULC, soil, topography, hydraulic conductivity, NDVI, drainage density, and lineament density are the main factors influencing groundwater recharge for the FR method.

• (4) For SE method:

• • • 42.98% of the total area has low groundwater recharge.

  • • The medium, high, and extreme vulnerable zones cover 14.13, 27.01, and 15.95% of the total area, respectively.

  • • The main influencing factors are near to the same in relation to the FR method.

• (5) For MIF method:

• • • 53.89% of the total area has intermediate groundwater recharge.

  • • 3.36, 34.05, and 8.69% of the entire region are comprised of low, high, and extreme sensitive zones, respectively.

  • • Areas with sand sheets and the Nubian aquifer predominate, slope areas greater than 10%, drainage density of less than 1.3 km/km², elevations less than 75 m, NDVI > 0.3, hydraulic conductivity greater than 7.8 m/day, and distances from lakes less than 10 m are all classified as being in the high to extreme zone for groundwater recharge.

  • • The results confirmed that the primary elements impacting groundwater recharge for the MIF method include lithology, LULC, topography, slope, hydraulic conductivity, NDVI, drainage density, and lineament density.

The results led to the following summarized remarks:

• • The majority of high-yield well locations were found in classes with high and extreme groundwater recharge which were gathered from RIGW, according to a reliability analysis of the maps as shown in Table 2 indicating that the models performed well in classifying the study region in terms of well placements.

• • A detailed examination of these maps reveals regions with high to extreme groundwater recharge in the study area's central, northeastern, southeastern and some the southern parts of the study area.

• • The zones of low and moderate groundwater were found in the west, the western north and the north area of the center of the study area.

• • Since Lake Nasser is nearby, the permeable surface lithology, high hydraulic conductivity, and the spread of sandy loam in these places contribute to the high-recharge zones’ location, according to the resulting maps created by utilizing the AHP, FAHP, MIF, SE and FR methodologies.

• • Even though there is high-intensity rainfall in some of the northern portions of the research region compared to the rest of the study area, it is characterized by low recharge zones according to the results of five methods because limestone aquifers occupy this area, and the recharge from the lake is much stronger than rainfall intensity in this area.

• • In addition, the north and the west portions of the research site were situated in low groundwater recharge zones because of the low permeability, high lands, and steep slopes.

• • There was some variation in the spatial layout of certain zones, even with the great matching between the maps produced by all the five methods. This discrepancy might be attributed to the pumping wells’ placement within the noted low and moderate recharge zones, which raised these regions’ computed recharging using the FR and SE methods.

• • Compared to another study that was in the right part of the current study area (Masoud et al. 2022), the following can be pointed out:

• • o This study did not suffer the lack of good data compared to Masoud's study as the total number of production wells in this study area was 112 wells, while in Masoud's study was only 44 wells.

  • o In this study, fifteen factors were studied, while in Masoud's study, only eight factors were considered. It was found that the seven other factors significantly influenced the results of the current study.

  • o The DSS approaches of Masoud's study were AHP and FR only, but in this study, AHP, FAHP, FR, SE and MIF were used.

  • o This increases the credibility of the decision-makers when using the resulting maps.

• • When compared with other studies of other regions (Benjmel et al. 2022; Senapati & Das 2022; Shekar & Mathew 2023; Bora 2024), the below table shows the comparison between the studies:

• • o The discrepancy in the final results of the best accuracy to determine GWRZs is due to many reasons:

  • • ▪ The AHP and FAHP approaches create a pair-wise comparison matrix and evaluate the CR after taking into account the variables' preferences and their relationships with each other.

    • ▪ The MIF approach is more individualized and heavily reliant on the expertise of the expert in the topic of study.

    • ▪ The FR and SE approaches require the distribution of the wells to determine the weights of each class.

    • ▪ The lack of production and observation wells -if lack exists- mainly affects the validation of models. Also, not use all factors that may affect GWRZs in the study area. This study did not suffer from the lack of data.

    • ▪ In this study, we compared all the approaches in Table 3, and used all the factors that may affect GWRZs. In other words, this study covers most of the approaches used in all papers related to DSS/GIS/RS for GWRZ maps.

    • ▪ By considering many different aspects of the hydrogeological system, data uncertainty, and decision-making approaches between influencing factors and groundwater occurrence, each approach makes a distinct contribution to the delineation of groundwater recharge zones. When these approaches are used, groundwater evaluations for a range of uses, including resource management, land use planning, and environmental preservation, can become more thorough and accurate.

    • ▪ In determining groundwater recharge zones, it is necessary to use various parameters such as range, resolution, and data sources to improve the accuracy and reliability of the analysis. Understanding extremes, seasonality, and mean values, alongside statistical measures like standard deviation, skewness, and kurtosis, helps assess data variability and distribution. Spatial variability is critical for identifying areas with high recharge potential. These parameters facilitate model calibration, and effective spatial and temporal analysis, leading to informed and sustainable groundwater management decisions. Please refer to ‘statistics.xlsx’ in the supplementary material for calculations of some applicable statistics for the thematic layers.

In order to establish confidence in the spatial techniques proposed in this study and to determine their physical significance, the results were validated.

Since the majority of these wells were pumped with a high yield, the groundwater recharge zone maps in the current study were confirmed by the overlaying number of productive wells on the maps (Figure 5(a)–5(e)).

The ArcGIS overlaying and extraction procedures were carried out in order to estimate each well's recharge zone. In the research region, a total of 112 pumping wells are used.

A high degree of validation was demonstrated by the AHP approach, as no wells were discovered to be in the low groundwater recharge zones, 19 wells were discovered to exist inside the medium GWRZ, and 93 wells were located within the high and extreme GWRZ (Table 2).

Similarly, of these 112 wells, only 1 was found in the FAHP method's low GWRZ, whereas 98 wells were found in the high and extreme GWRZ. No wells were found in low GWRZ for FR, SE, or MIF approaches, and 111, 111, and 108 wells were found in high and extreme GWRZ for FR, SE, and MIF in the same order.

The model showed that 83, 87.5, 96.4, 99.1, and 99.1% of wells were located in the high and extreme GWRZ for AHP, FAHP, MIF, FR, and SE approaches, respectively. It was thus valid.

According to these graphs, the areas under the ROC curve (AUC) are equal to 84.4, 81.9, 91, 91.1, and 89.9% for AHP, FAHP, FR, SE, and MIF models, respectively.

Based on 3, the prediction accuracy's AUC values were categorized as follows: poor (0.5–0.6), average (0.6–0.7), good (0.7–0.8), very good (0.8, 0.9), and exceptional (0.9–1).

The FR and SE techniques proved to be more efficient than the other models in this study, although all models, when it came to spatial groundwater recharge prediction, showed very good accuracy.

If we compare FR and SE approaches in this study to get the most validated approach, SE will be more accurate than FR for determining GWRZs, because the higher AUC percentage and the number of pumping wells in extreme GWRZs in SE are more than FR.

Even though the fifteen-parameter GWRZ maps have experienced extensive validation and are thought to accurately depict the actual conditions.

Plenty of research has examined the sensitivity analysis of groundwater vulnerability, but not many have examined groundwater recharge (Barbulescu 2020; Pathak 2021).

In order to check the importance of inputs for each approach to see which of the factors (parameters) have a strong or weak impact on groundwater recharge zones, a Sensitivity analysis of factors that affect GWRZs has been prepared using various combinations of parameters:

• 1. Case 1 (Three parameters: geology, LULC, and soil).

• 2. Case 2 (Six parameters: geology, LULC, soil, distance from lake, rainfall, and depth to water).

• 3. Case 3 (nine parameters: geology, LULC, soil, distance from the lake, rainfall, depth to water, hydraulic conductivity, lineament density, and drainage density).

• 4. Case 4 (twelve parameters: geology, LULC, soil, distance from the lake, rainfall, depth to water, hydraulic conductivity, lineament density, drainage density, hydraulic conductivity, topography, slope, and NDVI).

The combination of parameters has mainly been selected based on four categories seeking to better understand their influences on GWRZs:

• 1. Category 1 (geospatial classifications).

• 2. Category 2 (external natural and artificial recharge).

• 3. Category 3 (groundwater ability to move and the potential for surface water penetration).

• 4. Category 4 (ground feature and its specifications).

The GWRZs for all cases were prepared through the same five approaches used in the main topic.

Due to the urgent need to increase reclamation activities in order to meet the growing population. As a result, arid conditions and a noticeable increase in water use are present in the southern region of Egypt. The goal of the study was to enhance the management of the water resources in the region by examining and analyzing the groundwater recharge zones in the region west of Lake Nasser. To evaluate the groundwater recharge zones, AHP, FAHP, FR, SE, and MIF techniques were used. Fifteen groundwater regulating elements were used to generate the maps of groundwater recharge zones. The following may be concluded:

• • According to the study, for AHP, FAHP, FR, SE, and MIF, the sum of the percentages of the total area of moderate, high, and extreme-recharge zones accounted for 82.7, 67.5, 58.31, 57.1, and 96.6%, respectively.

• • 83, 87.5, 99.1, 99.1, and 87.5% of wells existed in high and extreme GWRZ for AHP, FAHP, FR, SE, and MIF, respectively.

• • The ROC curve was used to assess the effectiveness and validate the models that were in use.

• • The AUC are equal to 84.4, 81.9, 91, 91.1, and 89.9% for AHP, FAHP, FR, SE, and MIF models, respectively.

• • The five methods showed very good accuracy in determining different GWRZs.

• • The FR and SE methods had higher predictive performance rates than the AHP, FAHP, and MIF approaches, as shown by the ROC curve accuracy. In particular, SE is the most validated approach in this study.

• • In the five approaches, the ROC curve accuracy increases more between Cases 1 and 2 than between Cases 2 and 3 and between Cases 3 and 4. This demonstrates the noteworthy impact of incorporating the parameters of (distance from lake), (rainfall), and (depth to water) with (geology, LULC, and soil).

• • The ROC curve accuracy is further increased by adding the following factors: Hydraulic conductivity, lineament density, drainage density, NDVI, topography, and slope, which together equal 12 parameters (Case 4). However, the 15 factors which we have used in the main topic give the best accuracy for all five approaches as shown in Figure 6, which means the influence of all factors that we have used in the main topic.

• • The statistical significance of the results and comparison between the five approaches can be summarized in three points:

• • 1. Consistency and sensitivity: The statistical significance considers the consistency and sensitivity of the results to changes in input parameters (factors). All five approaches can demonstrate sensitivity to changes in influencing factors or thresholds.

  • 2. Accuracy and validation: The precision of the results obtained through the five approaches is an important issue. Sensitivity analysis is an important statistical validation method that could be used to evaluate the accuracy as well as the reliability of each approach' groundwater recharge zones.

  • 3. Model performance metrics: statistical metrics can be used to evaluate model performance and compare the effectiveness of the used approaches in the study like Area under the receiver operating characteristic curve (AUC-ROC). Higher values of these metrics generally indicate better model performance in delineating groundwater recharge maps in this study.

• • More information and data, that have direct or indirect effects on groundwater retention and occurrence, can be manipulated to increase the effectiveness of the remote sensing approach in groundwater recharge.

• • The research findings can be utilized by decision-makers to formulate intelligent development plans and efficient groundwater management techniques, therefore ensuring sustainable groundwater usage in the region. Making well-informed decisions to guarantee the best possible use of groundwater resources in the study area is possible by adopting the integrated strategy and methodology used in this study.

• • In search of a deeper understanding of the intricate dynamics of groundwater recharge, future research endeavors should give precedence to the exploration and enhancement of soil water models. This entails looking into innovative technology that can enhance model projections as well as adding more environmental factors for a thorough analysis. We can improve DSSs and remote sensing technologies and develop more sustainable groundwater resource management techniques by deepening our understanding of soil water dynamics.

I want to sincerely thank ‘Research Institute for Groundwater’ for granting me access to the well data that were utilized in this study. Their assistance has been crucial in streamlining the data-gathering procedure and making the analysis carried out in this study possible.

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

The authors declare there is no conflict.