Enhancing Groundwater Recharge Assessment in Mediterranean Regions: A Comparative Study Using Analytical Hierarchy Process and Fuzzy Analytical Hierarchy Process Integrated with Geographic Information Systems for the Algiers Watershed

Abstract

Groundwater recharge is critical for sustainable water management in water-scarce regions like North Algeria, where climate change and urbanization exacerbate resource challenges, particularly in the populous Algiers watershed. This study evaluates groundwater recharge potential using the Analytical Hierarchy Process (AHP) and its fuzzy extension (FAHP), integrated with Geographic Information Systems (GIS), to map recharge zones. Employing open-source data, AHP and FAHP assessed factors such as lithology, slope, and rainfall, classifying the watershed into high, moderate, and low recharge potential zones. Results show AHP identifying 44.01% (528.95 km²) as high, 52.82% (634.93 km²) as moderate, and 3.18% (38.14 km²) as low potential, with FAHP yielding similar outcomes (44.35%, 52.47%, and 3.17%, respectively). Validation using borehole drawdown data confirmed a 73.3% accuracy and an AUC of 0.72, indicating moderate to good reliability. High recharge zones were concentrated in the central and northern areas with permeable soils and gentle slopes, moderate zones dominated the region, and low zones were minimal. This study concludes that both AHP and FAHP are effective for preliminary recharge assessments, with AHP favored for its simplicity, though FAHP excels with uncertain data. Limited high-resolution hydrogeological data highlight the need for further refinement, yet the approach offers a replicable framework for managing groundwater in arid, urbanized regions facing similar environmental pressures.

1. Introduction

Groundwater recharge is essential for sustaining water resources, particularly in water-scarce regions like North Algeria. This process involves water from precipitation or surface sources infiltrating the ground to replenish aquifers, maintaining groundwater balance and ensuring a sustainable water supply [1]. In arid and semi-arid areas, such as North Algeria, where rainfall is limited and unreliable, groundwater often becomes the primary and most dependable water source, crucial for agriculture, industry, and domestic needs [2]. This reliable supply is vital for the survival and development of communities in these regions.

North Algiers, the northern region of Algeria’s capital city, is a key area for groundwater recharge studies due to its unique geographical, climatic, and socio-economic characteristics. Situated along the Mediterranean coast, it features diverse landscapes, from coastal plains and gentle slopes to the rugged northern slopes of the Tell Atlas mountains. It experiences a Mediterranean climate with wet, mild winters and hot, dry summers, with annual precipitation ranging from 600 to 800 mm, primarily between October and March. As a highly urbanized capital with extensive residential, commercial, and industrial development, Algiers faces urban sprawl, land use, and water management challenges [3]. The hydrological network includes several small rivers and streams that contribute to local water supplies and influence groundwater recharge. The growing population and urbanization increase water demand, making efficient groundwater management crucial for sustainability [4]. Urbanization often increases impervious surfaces, reducing natural infiltration and affecting groundwater recharge. Additionally, surrounding rural areas rely on groundwater for irrigation, underscoring the need to identify recharge zones to support sustainable agriculture and enhance food security.

The north of Algiers, including the Mitidja Plain, has a complex geological and hydrogeological structure influencing its groundwater resources. The region is part of the Tellian domain within the Alpine orogenic belt and features folded sedimentary rocks. The Mitidja Plain, situated between the Tellian Atlas and the Sahel, is filled with sediments ranging from the Pliocene to the Quaternary, forming two principal aquifers: the Asian aquifer (sandstones and conglomerates) and the Quaternary aquifer (alluvial deposits from surrounding rivers) [5]. Groundwater primarily comes from coastal aquifers recharged by rainfall and surface water. Over-extraction has led to seawater intrusion, causing salinization and nitrate contamination from agricultural activities, which is also a concern. Both aquifers are vulnerable to overexploitation, which affects water quality and availability, especially in coastal areas. In recent years, the water table has experienced a significant decline [6].

Understanding groundwater recharge in North Algiers is essential for developing strategies to mitigate the impacts of climate change, such as altered precipitation patterns and increased drought frequency. This study aims to enhance the understanding of groundwater recharge dynamics in an urbanized coastal setting in North Algeria by assessing potential groundwater recharge zones using the Analytical Hierarchy Process (AHP), a decision-making method that helps prioritize criteria by weighting them, and its fuzzy extension, fuzzy AHP, which incorporates uncertainty by assessing degrees of membership to criteria [7]. These methods are integrated with Geographic Information Systems (GIS). By integrating these advanced methods with GIS, this study offers a comprehensive understanding and practical tools for effective groundwater recharge zone management, supporting sustainable water resource management and contributing to the region’s resilience against environmental and climatic challenges.

Global research on groundwater recharge zone assessment has been crucial for sustainable water management, employing various methods to identify and evaluate these zones [8]. Hydrological modeling, which simulates water movement through the hydrological cycle, helps estimate recharge rates and identify recharge zones by considering factors such as soil properties, land use, and climatic conditions [2,9,10]. Machine learning approaches such as Random Forest, Support Vector Machines (SVMs), and artificial neural networks (ANNs) have been successfully used to improve the accuracy of groundwater potential mapping. Studies have shown that hybrid models, particularly the ensemble machine learning (EML) model based on logistic regression, outperform individual techniques in terms of predictive accuracy [11,12,13]. However, none of these studies compared AHP or FAHP methods; the comparisons were solely within the family of machine learning approaches. Isotopic and tracer studies utilize stable isotopes of hydrogen and oxygen to understand groundwater recharge processes and sources, providing insights into the age and origin of groundwater and aiding in delineating recharge zones [14]. Field measurements, such as direct assessments of soil moisture and water table levels, provide empirical data for groundwater recharge assessment and are often combined with modeling and remote sensing techniques for a comprehensive analysis [15,16].

On the other hand, the integration of remote sensing, GIS, and Multi-Criteria Decision Analysis (MCDA) has proven effective and accurate in delineating groundwater potential zones, while field verification further ensures reliability [17]. The Analytical Hierarchy Process (AHP) is widely used to assign weightings to thematic layers such as lithology, geomorphology, lineament density, soil type, slope, land use/land cover, topographic wetness index, and precipitation [18,19]. Several variations of AHP refine the evaluation process: standard AHP involves a pairwise comparison matrix validated against well data; fuzzy AHP incorporates fuzzy logic to handle uncertainties in weight distribution [20]; AHP-ANP (Analytical Network Process) accounts for interdependencies among factors to improve groundwater mapping precision [21]; and AHP-MIF (Multi-Influence Factor) integrates statistical relationships for a more comprehensive evaluation [22]. Coupling AHP with machine learning, particularly artificial neural networks (ANNs), further enhances groundwater potential mapping. ANN models trained with AHP-derived weightings refine groundwater evaluation, achieving higher precision than traditional GIS methods. Additionally, ANN-AHP models optimize the identification of artificial recharge locations, demonstrating the potential of hybrid approaches in groundwater resource management [23]. In Algeria, GIS and AHP have been integrated with hydro geophysical data to assess groundwater resources in the Telidjene Basin, showcasing the effectiveness of hybrid approaches in data-scarce environments [17]. However, implementing AHP-ANN integration is costly and requires large datasets, contrasting with the approach used in this study, which advocates for the judicious use of easily accessible and open-source data.

Several studies in Mediterranean regions with similar climates and drought challenges due to climate change have applied comparable techniques to assess groundwater recharge potential. For instance, in regions such as Spain, Tunisia, Turkey, and Morocco, researchers have employed GIS-based multi-criteria analysis, machine learning techniques, and isotopic methods to map and evaluate recharge potential zones, demonstrating the adaptability of these approaches across different environmental settings. Many studies have been conducted in North Algeria, highlighting the methodologies used to identify groundwater recharge potential zones and their implications for sustainable water resource management. For instance, integrating AHP, remote sensing, and GIS in the Mostaganem Plateau, Northwest Algeria, has revealed significant changes in groundwater potential zones due to climate change from 2000 to 2014, with projections extending to 2050. These studies have shown a decrease in moderate potential zones and an increase in low potential zones, influenced by changing rainfall patterns and agricultural land use [24]. Similarly, in Algeria’s Upper Cheliff alluvial aquifer, AHP within a GIS environment was utilized to map natural groundwater recharge areas. Validation against well data confirmed its accuracy, highlighting the importance of rainfall and watercourse infiltration in determining recharge areas [19]. Similarly, in Algeria’s Tabelbala region, AHP was combined with remote sensing and GIS to delineate groundwater potential zones, with cross-validation using borehole and well data [25]. A study in the Djelfa region employed a hybrid model incorporating annual rainfall, average annual temperature, and geological characteristics, validated by a chemical tracer method, to quantify groundwater recharge in semi-arid zones. This approach has provided valuable insights for groundwater management [26]. Additionally, in the Mitidja plain, a multi-criteria approach identified three groundwater recharge classes: low, moderate, and high. The most favorable zones were located near the piedmont of the Blidean Atlas and along the western aquifer limit, characterized by high hydraulic conductivity and good water quality [27].

The significance and importance of this study lies in its application to the Algiers watershed. This highly populated area has experienced significant stress on its groundwater resources in recent years. The overexploitation of groundwater has led to a drastic and concerning decline in water levels, exacerbated by prolonged drought conditions [28]. This has triggered a rush toward groundwater extraction, mainly by farmers, where drilling a single well can cost millions of dinars. A failed or “negative” well represents a colossal financial loss, forcing many farmers to abandon agriculture, with severe socio-economic consequences [29]. The absence of reliable groundwater recharge maps in local administrations and the lack of field data have motivated this study to develop a preliminary approach for assessing recharge potential [5]. By adopting both deterministic AHP and fuzzy AHP methods, which account for uncertainties, this study aims to produce maps that enhance resilience to climate change, which is expected to worsen in the future [6]. Another key objective of this study is to rely exclusively on globally available and open-access data from platforms such as USGS EarthExplorer, SRTM, FAO-UNESCO, CHIRPS, and others [30,31,32]. This approach circumvents the challenges posed by local bureaucratic hurdles and the lack of reliable or high-quality data from local administrations. By leveraging these accessible datasets, this study provides a practical and replicable framework for groundwater recharge assessment, even in regions with limited local data infrastructure.

This study utilizes the AHP and fuzzy AHP methods to assess groundwater recharge potential zones by integrating various criteria such as rainfall, soil type, land use, and slope. By applying these methodologies, this study aims to comprehensively evaluate potential recharge zones, offering valuable insights for effective groundwater management and policy making in North Algeria. It also aims to guide and optimize groundwater prospecting campaigns, which incur significant budget costs.

2. Materials and Methods

Recent advancements in decision making for environmental studies employ sophisticated methodologies to manage complexity and uncertainty. Key tools include the Analytical Hierarchy Process (AHP), FAHP, and Geographic Information Systems (GIS), [33]. Developed by Thomas L. Saaty in the 1970s, AHP is a structured decision-making technique that prioritizes and makes decisions based on multiple criteria by breaking down complex problems into a hierarchy of subproblems. Steps include defining the problem, structuring the hierarchy, making pairwise comparisons, calculating weights, and synthesizing results. AHP is used in site selection, resource allocation, and environmental impact assessment [34]. An extension of AHP, FAHP incorporates fuzzy logic to address uncertainty. It uses fuzzy numbers for pairwise comparisons, allowing the use of linguistic terms converted into fuzzy numbers. Steps are similar to AHP but include fuzzy comparisons, constructing fuzzy matrices, calculating fuzzy weights, and defuzzification. FAHP is helpful in climate modeling, ecological risk assessment, and sustainable development planning [35,36]. Geographic Information Systems (GIS) are essential for spatial data collection, analysis, and visualization in environmental studies. GIS supports mapping, environmental monitoring, resource management, and risk assessment, with applications in land use planning, water resource management, and conservation [37]. This study utilized open-source QGIS for data processing and analysis.

Although AHP is a powerful tool in the decision-making process and, by extension, in water resource management, it has several limitations. One of the main drawbacks is its reliance on expert judgment for pairwise comparisons, which can introduce biases, inconsistencies, and variability in results due to differing expert opinions [34,38]. Additionally, AHP faces scalability issues, as the number of comparisons increases exponentially with the number of criteria, making its application complex and resource-intensive for large-scale projects [39]. Another limitation is its static decision-making framework, which struggles to adapt to dynamic environmental changes, such as climate variations and new regulations [40]. AHP alone is not well suited for spatial decision making, particularly in water resource management, but when combined with Geographic Information Systems (GIS), it becomes a powerful tool for geospatial analysis. However, this integration requires technical expertise, computational resources, and additional data processing, which adds to its complexity [41]. Furthermore, in water resource management, AHP provides qualitative decision-support insights, which are valuable but may be insufficient when quantitative accuracy is required. In such cases, decision makers must rely on physical laws and hydrological models to complement AHP-based evaluations.

To overcome these limitations, integrating AHP with GIS-based decision support systems, conducting sensitivity analyses, and adopting hybrid methods such as FAHP or multi-objective optimization can enhance its effectiveness in dynamic water resource management [7]. However, a significant drawback of AHP is that it relies on crisp numerical values of pairwise comparisons, which assumes that decision makers make judgments exactly. To address this drawback, FAHP was developed as an advancement of AHP, involving fuzzy logic to allow uncertainty and imprecision in expert judgments [42]. The significant difference between FAHP and AHP is in treating subjective judgments: AHP uses precise numerical values, whereas FAHP allows decision makers to express their preferences using linguistic terms (e.g., “slightly more important”, “much more important”). These terms are then converted into fuzzy numbers, in the form of triangular or trapezoidal fuzzy values, to better represent human reasoning [43].

2.1. Study Area

The Algiers watershed, situated in northern Algeria (Figure 1), spans the capital city and extends from the coastal plains to the northern slopes of the Tell Atlas mountains, covering an area of approximately 1206.372 square kilometers. This large basin has a perimeter of 226.248 km. It features a diverse topography, including coastal plains, hilly areas, and elevations ranging from −9.08 m to 1620.9 m above sea level, with a mean elevation of 805.9 m a.s.l. The basin’s topography is steep, with a mean slope of 12.177 degrees (21.578%), and the area exhibits a high relief and ruggedness number of 1.484, indicative of mountainous terrain. The watershed has a form factor of 0.463, an elongation ratio of 0.768, and a circularity ratio of 0.296, indicating a moderately elongated shape with a high susceptibility to flash floods.

Hydrologically, the Algiers watershed includes significant rivers such as Oued El Harrach and Oued Hamiz, which play crucial roles in surface water resources and groundwater recharge [44]. As shown in

Table 1. Hydrological and topographical characteristics of the Algiers watershed.

Attribute	Details	Attribute	Details
Area-Perimeter	1206.372 km2—226.248 km	Stream Frequency	0.715 streams/km2
Topography	Coastal plains, hilly areas, with elevations from −9.08 m to 1620.9 m a.s.l., and mean elevation of 805.9 m a.s.l.	Bifurcation Ratio	1.998
Mean Slope	12.177° (21.578%)	Main Channel Length	41.616 km
Relief and Ruggedness Number	1629.97 m-1.484	Total Channel Length	1098.4 km
Shape Characteristics	- Form Factor: 0.463	Time of Concentration	Kirpich: 159.863 min
	- Elongation Ratio: 0.768		Kerby: 187.565 min
	- Circularity Ratio: 0.296		Giandotti: 374.054 min
Major Rivers	Oued El Harrach, Oued Hamiz		Témez: 228.755 min
Drainage Density	0.911 km/km2		Mean Time: 237.559 min

, the main characteristics of the watershed are summarized. The drainage density of 0.911 km/km² and stream frequency of 0.715 streams/km² indicate a coarse drainage texture and low stream frequency, respectively [45]. The bifurcation ratio of 1.998 suggests uniform lithology and gentle slopes. The main channel length is 41.616 km, with a total channel length of 1098.4 km. The mean time of concentration for the Algiers watershed is 237.559 min. Individual methods show a wide range of values, from 159.863 min (Kirpich) to 374.054 min (Giandotti), indicating the varying responses based on different calculation approaches [45]. These watershed characteristics were calculated using the QGIS ArcGeek Calculator plugin and SRTM DEM.

Land use in the watershed is dominated by urban areas, particularly Algiers city, which features extensive residential, commercial, and industrial development. Surrounding rural zones are primarily used for agriculture, contributing to the region’s socio-economic dynamics [3]. The watershed faces numerous challenges, including urban sprawl, water pollution, and over-extraction of groundwater, necessitating integrated management strategies to ensure sustainable water resource utilization [4].

2.2. Data Collection and Processing

This study collected data from various sources, including field surveys and remote sensing datasets (

Table 2. Datasets and sources utilized in this study.

Dataset [Unit]	Description/Properties	Source(s)	Resolution
Slope [degrees]	Derived from DEM	SRTM 1-arc-second DEM (USGS EarthExplorer) [46]	30 m × 30 m spatial grids
Land use/cover [-]	Land use/cover classification	Landsat 9 (USGS) [30]	30 m × 30 m spatial grids
Lithology [-]	Detailed rock type classification	Global Lithological Map (GLiM) [31]	Vector
Soil [-]	Soil classification based on FAO-UNESCO	FAO-UNESCO Soil Map of the World	Vector
Drainage density	Total length of the hydrographic network by the basin area.	Derived from DEM	30 m × 30 m spatial grids
Lineament density [m/m2]	Length of lineament per unit area	Derived from DEM	30 m × 30 m spatial grids
Rainfall [mm/year]	20 years of precipitation data	CHIRPS, processed with Google Earth Engine [32]	0.05°
Evapotranspiration [mm/year]	Actual evapotranspiration (AET) data	GLDAS, processed with Google Earth Engine [32]	0.25°

). The Digital Elevation Model (DEM) was derived from the Shuttle Radar Topography Mission (SRTM) with a 30 m × 30 m spatial resolution, accessed through the USGS EarthExplorer platform [46]. The slope data for the analysis was also derived from the SRTM 1-arc-second DEM, downloaded via USGS EarthExplorer. The land use and land cover (LULC) data, also at a 30 m × 30 m resolution, was produced using Landsat 9 OLI/TIRS satellite imagery provided by the U.S. Geological Survey (USGS) [30]. Lithological data was sourced from the Global Lithological Map (GLiM), offering a detailed classification of the Earth’s surface rock types, while soil data was derived from the FAO-UNESCO Soil Map of the World, ensuring comprehensive soil classification across the study area [47]. Lineament density was calculated based on terrain data derived from the DEM. Rainfall and evapotranspiration data were sourced from CHIRPS and GLDAS, respectively [32], with adequate rainfall calculated using the Google Earth Engine. All GIS operations and analyses were performed using QGIS 3.34.12-Prizren software.

2.3. GIS Data Integration

Geographic Information Systems (GIS) are critical in managing and analyzing spatial data for groundwater recharge zone assessment and other environmental studies.

As a decision support tool, GIS provides spatially explicit information that aids in resource management. The integration of these data layers in GIS involves these steps [48]:

• Data Preprocessing: Convert all data layers to a standard coordinate system and resolution. The coordinate system is WGS UTM 31N, with a final resolution of 30 m · 30 m for all used rasters.
• Reclassification: Convert all layers into raster format and reclassify into three classes.
• Layer Overlay and Weighting: The GIS tool assigns each layer a weight based on its importance for groundwater recharge. The weights are derived using AHP and FAHP.
• Map Generation: Combine the weighted layers to generate maps of groundwater recharge potential zones. Use spatial analysis tools in GIS to create a final map highlighting areas with high, moderate, and low recharge potential.
• Validation: Validate the generated maps using groundwater-level measurements, also called piezometric maps.

The groundwater recharge potential index (GWRPI) [49] is computed as the weighted sum of the criteria scores, calculated pixel by pixel using the Formula (1):

$$GWRPIt=\sum _{i=1}^{n=6}{W}_i{C}_i$$

(1)

where $W_i$ is the associated weight for the i-th criterion, determined using the AHP, and $C_i$ is the score of the i-th criterion.

2.4. Criteria for Assessing Groundwater Recharge Potential: Selection and Influencing Factors

The assessment of potential groundwater recharge involves a variety of factors and criteria, often analyzed using Geographic Information Systems (GIS) and Multi-Criteria Decision Analysis (MCDA). Key factors identified in the literature include:

-

Aquifer Lithology and Soil Texture: The type of soil and underlying rock formations significantly influence the permeability and infiltration rates, which are crucial for groundwater recharge [50]. Soil texture refers to the relative proportions of sand, silt, and clay in the soil, which affect its physical properties such as permeability, water retention capacity, and aeration. These properties are essential for determining groundwater recharge potential. It is important to differentiate soil texture from geomorphology: while geomorphology analyzes geographical features and reliefs, soil texture refers to the characteristics at the particle scale within a soil sample. Lithology, conversely, pertains to the composition, texture, and structure of rocks, which also play a vital role in groundwater recharge.

-

Topography and Slope: The slope of the land affects the runoff and infiltration rates, where flatter areas tend to have higher potential for recharge due to slower runoff and more significant infiltration opportunities [51].

-

Land Use/Land Cover (LULC): Vegetation cover, urbanization, and land management practices directly impact the recharge rates, with forested and less urbanized areas typically allowing more recharge [52]. Land use refers to how land is utilized, whether through human activities or natural processes, encompassing urbanization, agriculture, forestry, and wetlands. Human activities like irrigation or the development of residential, commercial, and transportation spaces influence the area’s economic potential and recharge capacity. Using satellite imagery, land use and land cover (LULC) can be effectively analyzed [53].

-

Climate Factors: Precipitation and evapotranspiration rates are major climatic factors that govern groundwater recharge, with higher rainfall contributing to increased recharge potential [54].

-

Hydrogeological Characteristics: Features such as drainage density, fault density, and the presence of lineaments also play critical roles in determining how and where groundwater recharge occurs [55]. Drainage density (Dd) is a key indicator of how easily water flows through soil, influenced by factors such as retention capacity and infiltration rate [56]. It quantifies the linear extent of landform features, illustrating the spacing between channels, which impacts water flow and infiltration. A lineament, typically a straight feature, can describe alignments of mountain peaks, boundaries of elevated areas, river courses, shorelines, and the linear boundaries of geological formations [57].

The selection of appropriate criteria is a fundamental step in evaluating groundwater recharge potential, as each factor directly influences infiltration and subsurface water movement. Commonly used criteria include soil type, slope, rainfall, land use/land cover, and proximity to water bodies. While these factors are widely cited in the literature [58], this study adopts a refined selection based on their direct impact and practical applicability: land use/land cover, lithology, slope, adequate rainfall, drainage density, and lineament density. This targeted approach ensures a comprehensive yet streamlined assessment, optimizing the accuracy of recharge zone identification while minimizing data redundancy.

To enhance methodological efficiency, specific criteria were adjusted or combined. The “slope” criterion inherently accounts for geomorphology and topography, eliminating the need for separate parameters [59]. “Effective rainfall” integrates both precipitation and evapotranspiration [60], replacing land surface temperature as an indicator of recharge potential. Additionally, lithology and soil type collectively substitute for hydraulic conductivity, aquifer transmissivity [61], and geology, providing a practical alternative for assessing subsurface characteristics. Parameters such as water table depth, the vadose zone, and groundwater quality are considered a posteriori, as they require field measurements after drilling [62]. Given the study’s predictive focus, the selected criteria prioritize data accessibility and efficiency, enabling the development of a reliable recharge potential map to support informed groundwater management decisions.

This study selected key criteria to evaluate groundwater recharge potential based on their critical role in influencing infiltration and recharge dynamics. The selection process focused on factors with the most significant impact on groundwater movement and availability, ensuring a comprehensive yet efficient assessment [63]. Utilizing Geographic Information Systems (GIS) and Multi-Criteria Decision Analysis (MCDA), the following criteria were prioritized for analysis: land use/land cover, lithology, slope, adequate rainfall, drainage density, and lineament density. These factors were carefully chosen to enhance the accuracy of recharge zone identification while maintaining methodological efficiency.

2.4.1. Soil Type

The soil classification for this study is based on the FAO-UNESCO Soil Map of the World [64]. Sandy soils, with their high permeability, allow greater water infiltration, while clayey soils, with lower permeability, tend to retain more water.

Several studies have emphasized the importance of soil texture in recharge modeling, although the weight assigned to it varies. For instance, many studies attribute a moderate importance to soil texture, [65], while in [66,67], it is assigned a lower weight. However, Ref. [66] highlights the higher significance of soil texture in specific contexts. In addition to soil texture, rock formations also influence recharge potential. Reference [68] points out that rock blocks (boulders) enhance drainage and facilitate recharge, while massive rocks (mass rocks) generally hinder it. Calcareous loamy soils, such as loamy-skeletal and mixed (calcareous), are known for their high infiltration rates and are prioritized for recharge [66].

Alluvial sediments, with their significant permeability, are another key factor in recharge assessments [65]. Coarse-textured soils, like sandy soils, are better suited for groundwater recharge due to their higher infiltration capacity, whereas finer-textured soils, such as clay, have poor permeability and a reduced capacity to support recharge [67].

2.4.2. Aquifer Lithology

Lithology plays a crucial role in determining the potential for groundwater recharge, as it influences the underlying rocks’ composition, texture, and structure. Rocks’ ability to store and transmit water largely depends on their type and formation. Reference [69] emphasized the importance of lithology in groundwater recharge assessments, assigning it a significant weight in evaluating recharge potential. The lithological classification used in this analysis is based on the Global Lithological Map (GLiM), which provides a high-resolution representation of the Earth’s surface rock types through 1,235,400 polygons derived from existing regional geological maps. The classification system follows an extension of the approach introduced by [31,70], with additional lithological classes and two new levels of detail.

2.4.3. Slope

The slope of the land influences the runoff and infiltration rates. Gentle slopes facilitate water infiltration into the ground, while steep slopes increase surface runoff, reducing the potential for groundwater recharge [71]. The slope data used for this analysis were derived from the Shuttle Radar Topography Mission (SRTM) 1-arc-second Digital Elevation Model (DEM). The data were accessed and downloaded through the USGS EarthExplorer platform, which provides global elevation datasets with high-resolution coverage suitable for hydrological and terrain analyses.

2.4.4. Rainfall

Rainfall is a primary source of recharge. Areas with higher rainfall generally have a higher potential for groundwater recharge, provided the precipitation infiltrates the ground rather than contributing to surface runoff [72]. More abundant rainfall increases groundwater recharge potential [69]. In Ref. [65], the authors assigned a very high weight to precipitation and pointed out that the number and distribution of rain gauges can sometimes be inadequate.

In assessing the potential recharge of the Algiers watershed, the effective rainfall, a key factor, was calculated by analyzing 20 years of precipitation data (2004–2024) sourced from the CHIRPS dataset, complemented by actual evapotranspiration (AET) data derived from the GLDAS dataset.

Missing values were replaced by calculating the mean of neighboring points within a 5000 m radius. This process was implemented in Python (Version 3.13.2), where a function identifies neighboring points, computes the mean for a specific field, and updates the None values accordingly.

Annual precipitation was calculated using Google Earth Engine by summing daily precipitation values into annual totals. The actual evapotranspiration (AET) was derived by converting the GLDAS “Evap_tavg” band from kg/m²/s to millimeters. To determine effective rainfall, the mean annual AET was subtracted from the mean annual precipitation. The results were initially obtained as point grids across the region, which were subsequently converted to raster format using a cubic spline interpolation algorithm implemented in SAGA, with the entire process facilitated by QGIS software.

2.4.5. Land Use/Land Cover (LULC)

Land use patterns affect the amount of water that infiltrates the ground. Urban areas with impervious surfaces reduce infiltration, while agricultural or forested lands can enhance groundwater recharge through increased infiltration and reduced runoff [73]. Vegetation, especially in agricultural and forested areas, is instrumental in retaining moisture in the soil, thereby aiding groundwater replenishment. The high porosity of soils in agricultural lands enhances water percolation, while urban or barren areas, with low permeability, limit water infiltration [74].

Land use and land cover (LULC) analysis via satellite imagery provides insights into the suitability of areas for groundwater recharge. Forested regions and aquatic vegetation are particularly effective in this regard, owing to their ability to retain water. Agricultural lands and shrubs are considered moderately favorable for recharge, while urbanized and barren areas are less effective due to their limited capacity to retain water [74]. Additionally, residential zones, rocky terrains, and salt marshes are ranked lower for recharge potential due to challenges like high costs and implementation difficulties. The semi-automatic classification tool in QGIS software was used for analysis, with classification performed using a maximum likelihood algorithm after determining the training surfaces.

2.4.6. Drainage Density

Drainage density is calculated by dividing the total length of the drainage network by the basin area. Low Dd values indicate high subsoil permeability and dense vegetation, promoting better infiltration and groundwater recharge. In contrast, high Dd values suggest reduced infiltration and increased runoff due to denser drainage networks [75].

The authors of [48] assigned a moderate weight to drainage density (Dd) compared to other factors, while [67,68] gave it a very low weight. A low drainage density, generally less than 1 km/km², indicates permeable subsoil and dense vegetation, suggesting a coarser drainage texture. These areas are conducive to groundwater recharge rather than dense runoff [48]. Dd is influenced by topography, geomorphology, geology, land use, vegetation cover, infiltration capacity, resistance to erosion, surface roughness, the runoff intensity index, and climatic conditions [53,67].

Using GIS, drainage density maps help identify potential areas for groundwater recharge. The calculation is similar to lineament density but incorporates the Strahler stream order as a weighting factor, assigning more importance to streams with higher tributaries and larger flow capacities.

2.4.7. Lineament Density (Ld)

Lineament density, which measures the length of geological features like faults, joints, and fractures, is crucial for assessing groundwater recharge potential. These features increase porosity and permeability, facilitating water infiltration [76]. High lineament density indicates areas with more significant recharge potential [53,69].

A two-phase method was used to enhance lineament detection. First, the i.zc zero-crossing edge detection algorithm identified significant edges by detecting changes in slope direction in cell values, marking initial lineament boundaries. Next, hillshade analysis was applied by adjusting azimuth and altitude values to enhance the visualization of geological features, improving lineament detection.

The Line Density Interpolation algorithm in QGIS was then applied to calculate the spatial density of lineaments. A 500 m search radius was used, evaluating all lineament segments within this range for each raster cell. This process produced a continuous lineament density map, highlighting areas of higher or lower geological feature concentration, crucial for groundwater recharge assessments.

2.5. AHP Calculation

2.5.1. Pairwise Comparison of Criteria Using AHP

A pairwise comparison matrix (PCM) was constructed using the Analytical Hierarchy Process (AHP) to evaluate the relative importance of each criterion for groundwater recharge. The matrix follows Saaty’s 1–9 scale, where 1 indicates equal importance between two criteria, 3 suggests slight importance, 5 reflects substantial importance, and 9 represents absolute importance. Values below 1 signify lesser importance in reverse order [7,34].

The following factors, lithology, land use/land cover (LULC), slope, rainfall, lineament density, and drainage density, were considered when constructing the matrix. These factors have been widely studied in various regions. Studies always note the significance of lithology and LULC in groundwater recharge. Lithology, in particular, determines the permeability of the soil, influencing the flow of water into the ground, while land use influences infiltration rates and recharge potential [77]. Similarly, [78] lithology and LULC were two of the most substantial factors in groundwater potential mapping in the Genale-Dawa Sub-Basin. Therefore, factors 8 and 7 are ranked based on the soil, as seen in

Table 3. Pairwise comparison matrix.

Criteria	Lithology	LULC	Slope	Lineament Density	Rainfall	Drainage Density	Soil Type
Lithology	1	2	3	5	6	7	8
LULC	1/2	1	2	4	5	6	7
Slope	1/3	1/2	1	3	4	5	6
Lineament density	1/5	1/4	1/3	1	2	3	4
Rainfall	1/6	1/5	1/4	1/2	1	2	3
Drainage density	1/7	1/6	1/5	1/3	1/2	1	2
Soil type	1/8	1/7	1/6	1/4	1/3	1/2	1

.

The factors (slope, lineament density, rainfall) are rated as moderate factors. These factors are often considered secondary but still significant. It was found [67,79] that slope, lineament density, and rainfall significantly influence groundwater recharge, albeit with lower weight than lithology and land use. Regarding the soil type, they are rated 6, 4, and 3 (Table 3).

Drainage density and soil type are generally considered less influential in groundwater recharge. This is supported by the findings of [80], who assigned these factors lower weight in their assessment of recharge zones in arid areas of Pakistan [81]. The rates for drainage density and soil type are 2 and 1, respectively.

In light of the calculated weights, lithology and LULC emerge as the dominant criteria for groundwater recharge. Lithology holds the highest weight, underscoring its crucial role in subsurface water movement, while LULC significantly influences recharge potential. Though significant, slope, rainfall, and lineament density are assigned moderate weights as they contribute to water flow and recharge processes. Drainage density and soil type are considered the least influential, as reflected by their lower scores.

Findings from multiple studies support this logical ranking. The importance of lithology and land use/land cover (LULC) are emphasized by [67,82] in determining water percolation and retention capacity, making them crucial for groundwater recharge. Slope, rainfall, and lineament density are highlighted in [83] to play moderate but significant roles in controlling surface water runoff and subsurface pathways, contributing to recharge processes. Other studies, such as [84], reinforce that while slope and rainfall are important, their influence is more moderate than lithology and LULC, which directly impact subsurface structure and land cover, ultimately shaping groundwater recharge potential. All these considerations are summarized in Table 3 using the pairwise comparison matrix.

2.5.2. Normalization and Calculation

Each element in a column is divided by the sum of the respective column. This results in normalized values for each criterion. The normalized matrix is averaged across each row to calculate the priority weights for each criterion. These weights represent the relative importance of each factor in determining groundwater recharge potential (

Table 4. Priority weights of criteria for groundwater recharge potential assessment.

Criteria	Lithology	LULC	Slope	Lineament Density	Rainfall	Drainage Density	Soil Type
Weight	0.3605	0.2493	0.1742	0.0881	0.0596	0.0402	0.0280

).

2.5.3. Check for Consistency

The Consistency Index (CI) and Consistency Ratio (CR) are calculated to ensure the pairwise comparisons are logically consistent. If the CR is below 0.1, the comparisons are considered consistent [34]. The matrix is considered consistent because the Consistency Ratio (CR) is 0.0319 (approximately 3.2%), which is below the AHP threshold of 0.1 (10%). The CR is calculated using the Formula (2):

$$CR={\displaystyle \frac{CI}{RI}}$$

(2)

where CI is the consistency index and RI is the random index based on the number of criteria (for n = 7, RI = 1.32) [34].

The CI, calculated as (3), is:

$$CI={\displaystyle \frac{{\lambda }_{max}-n}{n-1}}$$

(3)

where ${\lambda }_{max}$ = 7.253 and $n=7$ is the number of criteria, indicating that ${\lambda }_{max}$ is close to $n$, which suggests logical consistency.

Formula (4) for ${\lambda }_{max}$, which represents the largest eigenvalue, is calculated by averaging the weighted sums of the matrix columns.

$${\lambda }_{max}={\displaystyle \frac{1}{n}}\sum _{i=1}^n{\displaystyle \frac{{\left(A\cdot W\right)}_i}{W_i}}$$

(4)

where $A$ is the pairwise comparison matrix, $W$ is the priority vector (weights), $n$ is the number of criteria, ${\left(A\cdot W\right)}_i$ represents the $i^{th}$ element of matrix $A$ multiplied by the priority vector $W$, and $W$ is the $i^{th}$ element of the priority vector.

The normalization and calculation process in this analysis systematically assigns priority weights to various criteria in the AHP model, providing insight into their relative importance for assessing groundwater recharge potential. Compared to existing literature, several key points of alignment and distinction emerge. First, lithology is identified as the most critical factor with a weight of 0.3605, highlighting its significant role in groundwater recharge potential studies [85]. Second, land use/land cover (LULC), with a weight of 0.2493, is recognized as another significant contributor, emphasizing its role in land management models [49]. Third, slope holds a moderate influence with a weight of 0.1742, as seen in several studies [85]. Finally, less influential factors include lineament density (0.0881), rainfall (0.0596), drainage density (0.0402), and soil type (0.0280), reflecting similar rankings across comparative analyses [86]. These weights provide a structured foundation for more informed groundwater management decisions.

2.6. Fuzzy Analytic Hierarchy Process (FAHP) Calculations

The fuzzy Analytic Hierarchy Process (FAHP) is an extension of the traditional Analytic Hierarchy Process (AHP) designed to address the inherent uncertainty and subjectivity in decision-making processes. While AHP provides a systematic framework for decision making through pairwise comparisons, it often assumes that decision makers can express preferences with absolute certainty [34]. However, in real-world scenarios, preferences may be imprecise or uncertain. Fuzzy AHP incorporates fuzzy logic to handle this uncertainty by using fuzzy numbers instead of crisp values for pairwise comparisons [35,36].

2.6.1. Fuzzy Pairwise Comparison Matrix

In the FAHP, triangular fuzzy numbers (TFNs) are used for pairwise comparison instead of crisp values to reflect the natural imprecision and uncertainty involved in decision making [35,36]. This ensures that the experts’ opinions are captured as a range of values rather than as crisp numerical inputs, giving a more flexible and realistic representation of individual preferences.

Triangular fuzzy numbers (TFNs) are defined as $\tilde{a}=(l,m,u)$, where [35,36]:

l: Lower bound represents the pessimistic estimate of preference.

m: Most likely value represents the neutral or medium estimate.

u: Upper bound represents the optimistic estimate of preference.

To maintain consistency with the traditional AHP, the fuzzy scale used in FAHP is derived from Saaty’s fundamental scale, as shown in

Table 5. Triangular fuzzy numbers (TFNs) for crisp values in fuzzy pairwise comparison matrix.

Crisp Value	1	2	3	4	5	6	7	8	9
TFN	(1, 1, 1)	(1, 2, 3)	(2, 3, 4)	(3, 4, 5)	(4, 5, 6)	(5, 6, 7)	(6, 7, 8)	(7, 8, 9)	(8, 9, 9)

[34,87]. This mapping allows for a structured yet flexible approach to capturing expert judgments. The triangular fuzzy numbers (TFNs) sequence represents a continuous and overlapping progression, where each TFN increases by one unit in its modal value. The narrow ranges (e.g., from 1 to 3, 2 to 4, etc.) indicate that the uncertainty is relatively small, as the possible values are tightly clustered around the modal value.

Using these fuzzy numbers, a crisp pairwise comparison matrix can be converted into a fuzzy matrix by replacing each crisp value with its corresponding TFN. This transformation is shown in

Table 6. Conversion of crisp pairwise comparison matrix to fuzzy matrix using triangular fuzzy numbers (TFNs).

Criteria	Lithology	LULC	Slope	Lineament Density	Rainfall	Drainage Density	Soil Type
Lithology	(1, 1, 1)	(1, 2, 3)	(2, 3, 4)	(4, 5, 6)	(5, 6, 7)	(6, 7, 8)	(7, 8, 9)
LULC	(1/3, 1/2, 1)	(1, 1, 1)	(1, 2, 3)	(3, 4, 5)	(4, 5, 6)	(5, 6, 7)	(6, 7, 8)
Slope	(1/4, 1/3, 1/2)	(1/3, 1/2, 1)	(1, 1, 1)	(2, 3, 4)	(3, 4, 5)	(4, 5, 6)	(5, 6, 7)
Lineament Density	(1/6, 1/5, 1/4)	(1/5, 1/4, 1/3)	(1/4, 1/3, 1/2)	(1, 1, 1)	(1, 2, 3)	(2, 3, 4)	(3, 4, 5)
Rainfall	(1/7, 1/6, 1/5)	(1/6, 1/5, 1/4)	(1/5, 1/4, 1/3)	(1/3, 1/2, 1)	(1, 1, 1)	(1, 2, 3)	(2, 3, 4)
Drainage Density	(1/8, 1/7, 1/6)	(1/7, 1/6, 1/5)	(1/6, 1/5, 1/4)	(1/4, 1/3, 1/2)	(1/3, 1/2, 1)	(1, 1, 1)	(1, 2, 3)
Soil Type	(1/9, 1/8, 1/7)	(1/8, 1/7, 1/6)	(1/7, 1/6, 1/5)	(1/5, 1/4, 1/3)	(1/4, 1/3, 1/2)	(1/3, 1/2, 1)	(1, 1, 1)

, where fuzzy comparisons between criteria such as lithology, land use and land cover (LULC), slope, lineament density, rainfall, drainage density, and soil type are established.

2.6.2. Fuzzy Geometric Mean, Normalization, and Defuzzification

Once the fuzzy pairwise comparison matrix is established, FAHP applies a series of calculations to derive crisp criteria weights, allowing for a final prioritization of the decision factors. The following steps are used:

-

Fuzzy geometric mean calculation: For each criterion i, the fuzzy geometric mean (GM_i) [87] was computed using the lower ${(l}_{ij})$, middle ${(m}_{ij})$, and upper ${(u}_{ij})$ bounds of the fuzzy pairwise comparisons by (5):

$$G{M}_i\mathrm{ }=\mathrm{ }\left({\left(\mathrm{ }\prod _{j=1}^n{l}_{ij}\mathrm{ }\mathrm{ }\right)}^{{\displaystyle \frac{1}{n}}},\mathrm{ }{\left(\mathrm{ }\prod _{j=1}^n\mathrm{ }{m}_{ij}\mathrm{ }\right)}^{{\displaystyle \frac{1}{n}}},\mathrm{ }{\left(\mathrm{ }\prod _{j=1}^n{u}_{ij}\mathrm{ }\mathrm{ }\right)}^{{\displaystyle \frac{1}{n}}}\right)$$

(5)

-

Normalization of fuzzy geometric means: Each $G{M}_i\mathrm{ }$ was normalized to obtain fuzzy weight [87] (Wi) with (6):

$$W_i={\displaystyle \frac{G{M}_i}{{\sum }_{i=1}^{k}G{M}_i}}$$

(6)

-

Defuzzification to crisp weights: The centroid method converted fuzzy weights $W_i=\left(l_i,m_i,u_i\right)$ into crisp values C_i [88,89], given by (7):

$$C_i={\displaystyle \frac{l_i+m_i+u_i}{3}}$$

(7)

This method is widely used in decision-making applications [90], as it effectively converts fuzzy numbers into a representative crisp weight while preserving the overall distribution of the fuzzy set.

-

Final normalization: Crisp weights, shown in Equation (8), were scaled to sum to 1 [91]:

$$C_i^{\prime }={\displaystyle \frac{C_i}{{\sum }_{i=1}^k{C}_i}}$$

(8)

-

Priority weights and consistency check: Final weights, given in

Table 7. Final priority weights and consistency check for criteria.

Criteria	Lithology	LULC	Slope	Lineament Density	Rainfall	Drainage Density	Soil Type
Weight	0.3574	0.2548	0.1774	0.0851	0.058	0.0395	0.0279

, were derived by normalizing $C_i^{\prime }$.

To calculate ${\lambda }_{\max }$, the following Formula (9) was used [92]:

$${\lambda }_{\max }=\sum _{i=1}^n{\displaystyle \frac{{\left(A\cdot W\right)}_i}{W_i}}$$

(9)

where $A$ is the matrix operating on the ratio of $W_i$ and $W_j$, and $W_i$ represents the local priority weights of the criteria. The matrix $A$ is given [34] by (10):

$$A=\left({\displaystyle \frac{W_i}{W_j}}\right)=\left(\begin{array}{cccc}{\displaystyle \frac{W_1}{W_1}}& {\displaystyle \frac{W_1}{W_2}}& \dots & {\displaystyle \frac{W_1}{W_n}}\\ {\displaystyle \frac{W_2}{W_1}}& {\displaystyle \frac{W_2}{W_2}}& \dots & {\displaystyle \frac{W_2}{W_n}}\\ ⋮& ⋮& \ddots & ⋮\\ {\displaystyle \frac{W_n}{W_1}}& {\displaystyle \frac{W_n}{W_2}}& \dots & {\displaystyle \frac{W_n}{W_n}}\end{array}\right)\hspace{1em}\forall i,j\in (\mathrm{1,2},\dots ,n)$$

(10)

If the Principal Eigenvalue ${\lambda }_{max}$: 7.442; Consistency Index ($CI$): 0.0737; and Consistency Ratio ($CR$): 0.0559 < 0.1, then the consistency is acceptable.

The priority weights calculated indicate that lithology is the most important factor in potential recharge, followed by LULC and slope, while soil type is the least important factor. These weights were used to prepare the groundwater potential zone.

3. Results and Discussion

The maps for the various datasets, including soil type, lithology, slope, LCLU, drainage density, lineament density, and effective annual rainfall, are presented below.

3.1. Maps of Factors Influencing Aquifer Recharge

It is important to clarify that a single criterion does not determine the proportion of each recharge class (low, moderate, high). Every criterion contributes a different amount based on its weight in the model. This weighted approach ensures that all the factors and their interactions are covered, and the final classification becomes more representative and accurate.

3.1.1. Soil Type Analysis

Figure 2a and Figure 3a illustrate the distribution of soil types and reclassified soil types in three classes: 1, 2, and 3. The FAOSOIL codes [47,64] describe different soil types across Algeria, each with varying potentials for groundwater recharge, which is influenced by their permeability and infiltration characteristics. The Jc (Calcaric Fluvisols) soil type, covering an area of 443.56 km², has low groundwater recharge potential due to its low permeability and slow infiltration rates. Similarly, Bc (Chromic Cambisols), which spans 165.16 km², exhibits low permeability and retains water at the surface. In research by [93], the authors highlighted that soil permeability directly impacts groundwater recharge, as areas with low infiltration rates significantly restrict aquifer replenishment. On the other hand, Bk (Calcaric Cambisols), which covers a larger area of 552.54 km², offers moderate groundwater recharge potential.

Finally, the Lo (Orthic Luvisols), covering 45.12 km², stands out with high groundwater recharge potential due to its loamy texture that allows for balanced infiltration and water retention, making it the most favorable soil type for groundwater recharge among the four. Empirical research supports this, as [94] found an empirical relationship between soil resistivity and permeability. This relationship showed that areas with higher resistivity, such as those with sandy and gravelly soils, exhibited higher permeability and recharge rates compared to finer-textured soils like silt-dominated ones.

Table 8. Classification of soil types according their groundwater recharge potential.

FAOSOIL Code	Soil Type	Area (km2)	Infiltration and Permeability	Groundwater Recharge Potential	Score
Jc	Calcaric Fluvisols	443.56	Low permeability, compact, slow infiltration, surface water retention	Low	1
Bc	Chromic Cambisols	165.16	Low permeability, retains water at surface, slow infiltration	Low	1
Bk	Calcaric Cambisols	552.54	Moderate to low permeability, caliche layer may obstruct infiltration	Moderate	2
Lo	Orthic luvisols	45.12	Moderate permeability, balanced infiltration and retention	High	3

classifies these soil types in increasing order of their groundwater recharge potential.

3.1.2. Aquifer Lithology Analysis

The lithology map and its attribute table, Figure 2b and Figure 3b, and

Table 9. Classification of lithology types according their groundwater recharge potential.

Lithology Code	Lithology Type	Infiltration and Permeability	Groundwater Recharge Potential	Relative Score
mt	Metamorphic rocks	Low permeability, compact, limited water movement [95,96]	Low	1
sc	Carbonate sedimentary	High permeability, potential for karst systems [69,96]	High	3
sm	Mixed sedimentary	Moderate permeability, depends on composition [69,96]	Moderate	2
ss	Sedimentary rocks	Moderate to low permeability, sandstone vs. shale areas [69,96]	Moderate	2
su	Unconsolidated sediments	High permeability, loose, porous [69,96]	High	3

illustrate the distribution of geological units across the region. Siliciclastic sedimentary rocks (ss) dominate, covering 56.47% of the area, while unconsolidated sediments (su) occupy 13.89%, primarily in the central region. Carbonate sedimentary rocks (sc) cover 9.78%, mainly in the south, and metamorphic rocks (mt) are confined to 15.06% of the northern part.

Mixed sedimentary rocks (sm) form a smaller, localized feature, covering 4.79%. Table 9 shows the relative score and potential recharge attribute at each lithological class. It presents a brief description of each score choice. These classifications align with global groundwater recharge assessments, confirming that lithology is a fundamental factor in groundwater storage and sustainability [62].

3.1.3. Slope Analysis

The slope varies from flat to steep, with values ranging from 0 to approximately 59.56 degrees. The flatter northern and central areas, indicated by lighter shades, as shown in Figure 2c and Figure 3c, offer high potential for groundwater recharge due to slower surface runoff, allowing more water to infiltrate the ground. In contrast, the steeper southern and eastern regions, shown in darker shades, have lower recharge potential, as rapid runoff in these areas limits water infiltration. The slope classification in

Table 10. Groundwater recharge potential by factor.

Factor	Range/Class	Relative Score	Potential	Area (km2)	Percentage (%)
Soil Type	-	1	Low	608.7	50.46
	-	2	Moderate	552.53	45.8
	-	3	High	45.13	3.74
Lithology	-	1	Low	15.48	1.28
	-	2	Moderate	585.39	48.41
	-	3	High	608.82	50.32
Slope	0–10°	3	High	190.32	15.75
	10–25°	2	Moderate	395.49	32.73
	>25°	1	Low	622.39	51.51
Effective Rainfall (mm)	189.05–257.48	1	Low	458.53	38.02
	257.48–325.9	2	Moderate	442.79	36.71
	325.9–394.33	3	High	304.92	25.28
LULC	Built-up	1	Low	232.87	19.3
	Bare Soil	2	Moderate	8.28	0.69
	Forest/Shrubland	3	High	201.4	16.69
	Agricultural Land	3	High	763.88	63.32
Drainage Density (km/km2)	0–0.004	3	High	15.88	1.32
	0.004–0.008	2	Moderate	157.62	13.07
	0.008–0.011928	1	Low	1032.69	85.61
Lineament Density (km/km2)	0–0.00092	1	Low	1077.19	89.29
	0.00092–0.00184	2	Moderate	122.08	10.12
	0.00184–0.002758	3	High	7.07	0.59

(0–10°, 10–25°, >25°) is based on standard geomorphological criteria commonly used in terrain analysis and land stability studies. Specifically, slopes of 0–10° are considered gentle and typically favorable for construction and agriculture, 10–25° are moderate slopes where erosion risks increase, and >25° are steep slopes associated with high erosion potential and instability [97,98]. This study confirms that flat terrains facilitate higher infiltration rates, whereas steeper slopes experience rapid runoff, reducing groundwater replenishment. This study also highlights that topography influences recharge potential, with northern slopes sometimes exhibiting higher recharge than southern ones. Globally, northern slopes exhibit higher recharge zones compared to their southern counterparts. Table 10 shows the relative score and potential of recharge used in reclassification.

3.1.4. Rainfall Analysis

The map illustrates the spatial distribution of effective annual rainfall across the region, ranging from 189.06 mm to 394.32 mm. Rainfall is higher in the mountainous areas due to orographic lifting, where air masses are forced to rise, cool, and condense, resulting in greater precipitation [99]. In contrast, rainfall is lower in the plains. However, evapotranspiration is significantly higher in the mountainous areas, surpassing 400 mm/year, while in the plains, evapotranspiration is lower, around 300 mm/year [100]. This difference in evapotranspiration further influences the effective rainfall. The maximum annual precipitation over the twenty years of study was 860.03 mm, and the minimum was 535.83 mm. Given that a spatial average was calculated for the study area for precipitation, we then selected the maximum and minimum values. The interannual average was 677.33 mm. This contrast gives rise to a distribution of effective rainfall. The map in Figure 2g and Figure 3g illustrates the spatial distribution of effective annual rainfall across the region, ranging from 189.06 mm to 394.32 mm. The regions of high effective rainfall (325.90–394.32 mm), shown in dark blue, are typically found in lower elevation areas, especially in Larbaa, Bougaraa, the Bouinan plains, and near the littoral plains. The relative score associated was 3 (Table 10). Increased rainfall can lead to rapid groundwater recharge through macropores and preferential pathways, particularly where these features facilitate fast infiltration [101]. Moderate effective rainfall (257.48–325.90 mm), depicted in medium blue, is found in regions of moderate elevation. These areas exhibit moderate recharge potential with a score of 2. Regions with low effective rainfall (189.06–257.48 mm), shown in dark blue, particularly in the southern part of the watershed, correspond to the higher mountainous areas of Radjimi, Ighil Ouassel, Ouled Ali in Medea province, and their surroundings. These mountainous areas have a recharge potential score of 1. Reduced effective rainfall in these areas, combined with higher evapotranspiration, limits groundwater recharge, affecting overall water availability [102].

3.1.5. Land Use/Land Cover (LULC) Analysis

Figure 2d and Figure 3d illustrate the spatial distribution of land classes and their impact on groundwater recharge. The land classification analysis shows that Fields-Agricultural Land dominates the region, covering 63.32% and playing a significant role in groundwater recharge by facilitating water infiltration (Table 10). Built-up areas (19.30%) are concentrated around Algiers and have low recharge potential due to impermeable surfaces. As found by [103], urbanization restricts direct infiltration and significantly reduces groundwater recharge by replacing permeable surfaces with impervious built-up land. Forest-shrubland (16.69%) also contributes greatly to groundwater recharge alongside agricultural fields by allowing rainwater to infiltrate the soil. Afforested regions and natural forests generally promote higher infiltration rates and contribute to groundwater replenishment [67,73]. Gebere et al. [104] found that land use and land cover changes, including deforestation, have reduced groundwater recharge by altering infiltration and runoff patterns.

Studies highlight that vegetated areas, including afforested regions and natural forests, generally promote higher infiltration rates and contribute to groundwater replenishment [105]. However, land use and land cover changes, mainly deforestation and urban expansion, have been found to reduce groundwater recharge by altering infiltration and runoff patterns, emphasizing the importance of sustainable land management. Additionally, increasing built-up areas and the loss of natural land cover have led to declining high recharge potential zones, reinforcing the need for green urban planning to mitigate these effects [74]. Bare Soil represents only 0.69% of the area and has moderate recharge potential (Table 10). While bare soil has some infiltration capacity, its recharge efficiency is typically lower than vegetated areas due to the lack of vegetation cover and reduced infiltration, as highlighted by [74], who emphasized the impact of land cover changes on groundwater recharge potential. The spatial distribution highlights urban development in the north and agriculture in central and southern areas, emphasizing the need to protect natural and agricultural lands for sustaining groundwater resources. The classification achieved 98.43% accuracy and a Kappa coefficient 0.9704, confirming its reliability. Accurate LULC classification enhances groundwater resource management by identifying areas with high and low recharge potentials [74].

3.1.6. Drainage Density Analysis

Areas with higher drainage density (shown in green, Figure 2e and Figure 3e) correspond to well-developed stream networks, which are prone to higher runoff. In contrast, lower-density areas (in red) have fewer drainage channels, promoting better water infiltration. Drainage density significantly influences runoff potential, and high drainage density leads to greater surface runoff and reduces groundwater recharge, while low drainage density areas facilitate better infiltration [106,107].

The data from Table 10 show the following:

-

Low drainage density (0–0.004 km/km²) covers 1.32% of the area (15.88 km²). These areas, shown in red, have fewer drainage channels, with a groundwater recharge potential with a relative score of 3.

-

Moderate drainage density (0.004–0.008 km/km²) spans 13.07% of the area (157.62 km²). These areas exhibit some surface runoff but still allow for moderate infiltration, leading to moderate recharge potential with a relative score of 2.

-

High drainage density (0.008–0.011928 km/km²) constitutes 85.61% of the area (1032.69 km²). These regions have a denser drainage network, leading to more significant surface runoff and limited water infiltration, resulting in a low recharge potential with a relative score of 1.

3.1.7. Lineament Density (Ld) Analysis

The lineament density map of the Algiers watershed (m/m²) is presented in Figure 2f and Figure 3f. It can be observed that:

Low lineament density (0–0.00092 km/km²) covers 89.29% of the area (1077.19 km²), represented by lighter colors on the map. These areas are assigned a relative score of 1 (Table 10).

-

Moderate lineament density (0.00092–0.00184 km/km²) constitutes 10.12% of the area (122.08 km²), with a relative score of 2.

-

High lineament density (0.00184–0.002758 km/km²) makes up only 0.59% of the area (7.07 km²), shown in darker red. These regions receive a relative score of 3 (Table 10).

The southern and central regions, with higher lineament density, are likely the most favorable for groundwater recharge. In contrast, the northern region, with lower density, may have reduced recharge capacity [108].

3.2. Reclassified Potential Recharge Map

3.2.1. AHP-Based Potential Recharge Map

The AHP method classified the study area into three potential recharge zones with the following distribution:

-

Low Recharge Potential (1.29–1.83): This class covers 38.14 km², predominantly in the northwestern and some central parts of the study area. These areas contain Chromic Cambisol and Calcaric Fluvisol soils, known for their low permeability, which retain water and hinder infiltration. This aligns with the findings of [109], which noted that areas with low permeability soils often exhibit reduced groundwater recharge potential, contributing to water retention rather than infiltration [109]. Low permeability soils, such as clay and compacted materials, have tiny pores that limit infiltration and slow water movement into the subsurface. These soils retain water within their pore spaces, delaying its percolation into deeper layers. Due to their low hydraulic conductivity, water accumulates near the surface, increasing runoff and reducing recharge potential [109].

-

Moderate Recharge Potential (1.83–2.38): Covering the largest portion of the study area at 634.93 km², this class occupies parts of the central, southern, and northern regions, where intermediate recharge capacity is observed. The presence of sedimentary rock soils with moderate permeability is consistent with studies showing that such geological formations support moderate recharge due to their balanced porosity and permeability. The authors of [110] demonstrated in their GIS-based AHP study that sedimentary rock areas tend to have intermediate groundwater recharge zones because of their porosity and fracture density.

-

High Recharge Potential (2.38–2.92): Represented by 528.95 km², these zones are mainly in the northern and some central-southern regions. Here, unconsolidated sediments dominate, which is known for high permeability and facilitates rainwater infiltration into aquifers. This is similar to the findings by [111], who used the AHP method to highlight the role of unconsolidated soils and geological lineaments in enhancing recharge in high permeability zones in the Panj Amu River Basin, Afghanistan.

3.2.2. FAHP-Based Potential Recharge Map

The FAHP method, using a frequency ratio approach, produced a similar recharge potential classification, with only slight variations (

Table 11. Recharge zone reclassification by AHP vs. FAHP.

Potential	Reclassified Recharge	Area (AHP) (km2)	Area (FAHP) (km2)
Low	1.29–1.83 (AHP), 1.50–2.02 (FAHP)	38.14 (3.18%)	38.17 (3.17%)
Moderate	1.83–2.38 (AHP), 2.02–2.54 (FAHP)	634.93 (52.82%)	630.71 (52.47%)
High	2.38–2.92 (AHP), 2.54–2.92 (FAHP)	528.95 (44.01%)	533.14 (44.35%)

):

-

Low Recharge Potential (1.50–2.02): Covering 38.17 km², this low recharge zone closely aligns with the AHP low recharge regions.

-

Moderate Recharge Potential (2.02–2.54): This class includes 630.71 km², similar to the AHP result, and is primarily located in the central regions.

-

High Recharge Potential (2.54–2.92): Comprising 533.14 km², this high potential zone mirrors the AHP distribution with minor pixel count and area differences.

3.3. Comparison and Interpretation

Both methods show consistent patterns, with only slight differences in pixel counts and areas across recharge potential classes. The FAHP method reduces the area slightly for moderate recharge zones, from 634.93 km² to 630.71 km², and raises the area’s high potential from 528.95 km² to 533.14 km². Approximately 44% of the study area is identified as having high recharge potential, while moderate recharge zones cover nearly 52%, followed by low recharge zones at around 3%. Despite these minor variations, the overall classification remains stable, indicating that the differences between FAHP and AHP are negligible. This aligns with [112], which also observed slight differences in the spatial distribution of recharge zones when applying AHP and FAHP. For instance, the findings of [112] using the AHP model classified the Kinnerasani Watershed as 17.76% low potential, 72.79% moderate potential, and 9.45% high potential. In comparison, FAHP resulted in a slight decrease in the moderate category (71.07%) with an increase in high potential (10.69%). These slight variations suggest that while FAHP refines classification by incorporating uncertainty, it does not substantially alter the final recharge potential distribution, confirming that both methods are equally reliable for groundwater potential assessment, with almost the same accuracy.

The absence of statistically significant differences between the AHP and FAHP methods can be attributed to several factors. Both methods rely on expert judgment to determine the relative importance of factors influencing groundwater recharge potential. Although FAHP integrates fuzzy logic to address uncertainty, the triangular fuzzy numbers (TFNs) used in FAHP were similar to the crisp values used in AHP, particularly for the thematic layers considered. Despite the incorporation of fuzziness, the models produced comparable groundwater potential zone maps, confirming their reliability in this context [113]. This situation is found, for example, in project selection, where both methods provided similar results when the criteria were well-defined and the data relatively certain [114]. The thematic layers, such as slope, drainage density, land use, and lineament density, show well-defined patterns in the study area. Given the certainty of these data, the addition of fuzzy logic did not alter the spatial distribution of groundwater recharge zones, resulting in similar outputs between AHP and FAHP [113,115].

3.3.1. Spatial Distribution of Recharge Potential

The spatial distribution of groundwater recharge potential, as depicted in both the AHP and FAHP maps, shows clear patterns across the study area. High recharge zones, represented by dark blue areas, are concentrated in the central and northern regions (Figure 4). These areas benefit from favorable conditions: permeable soils, gentle slopes, and likely higher rainfall, making them ideal for groundwater replenishment. Moderate recharge zones, depicted in medium blue, with 52% of the area, reflect a balance of factors: moderately permeable soils and slopes offering reasonable capacity for groundwater recharge. These zones are widespread and indicate that a significant portion of the region holds the potential for moderate groundwater storage.

In contrast, low recharge zones, shown in light blue, are rare, covering only 3% of the area. They are sparsely distributed in the extreme northwest and central east, where impermeable soils limit groundwater recharge potential.

These findings align with previous studies using GIS-based AHP and FAHP methods to map potential zones in diverse hydrogeological settings. For instance, in [113], it was demonstrated that nearly one-third of the Erbil Basin consisted of moderate-to-high recharge zones due to similar factors: lithology, soil permeability, and rainfall patterns [113]. Likewise, in the Vaigai Upper Basin, Tamil Nadu, [116] found that moderate recharge potential was dominant, particularly in regions with gentle slopes and alluvial deposits [116].

3.3.2. Implications for Water Resource Management

The consistent results of the AHP and FAHP methods highlight their reliability in groundwater management. High recharge zones can be prioritized for interventions like artificial recharge to maximize water resource availability, while low recharge areas may require surface water management strategies. The vast areas of moderate recharge should promote sustainable water management and balanced groundwater extraction throughout the region.

The practical implications of these findings are highlighted in several case studies where GIS-AHP models have been used in groundwater conservation planning. A study conducted in the Shinile watershed, Eastern Ethiopia, demonstrated that AHP-based assessments effectively identified high potential groundwater recharge zones, which were subsequently validated with borehole data, confirming the reliability of the approach [117].

Similarly, research conducted in southwestern Nigeria confirmed that AHP-FAHP methods provide valuable and actionable insights for policymakers to optimize groundwater use [118]. The results of this study reinforce these conclusions, indicating that such models can effectively guide sustainable groundwater management practices across various hydrogeological contexts.

3.4. Validation of Results

To ensure the recharge potential map’s accuracy and reliability, the results were validated by comparing the estimated recharge zones with 30 deep well flow measurements across well fields in several circumscriptions: Birtouta, Kheracia, Ouled Chebel, Saoula, Sidi Boukhris, Sidi Moussa, Tessala El Merdj, and Bir Khadem.

The boreholes are closely located, as shown in Figure 5. Based on their geographical distribution, four distinct groups can be identified, each outlined in Figure 5. Each borehole is characterized by a technical data sheet, but only the following parameters are considered: x and y coordinates, altitude, operating flow rate, static water table level, and dynamic water table level. The close proximity of these boreholes strongly suggests that they draw from the same aquifer, leading to potential interferences between different extraction sites. This makes conducting pumping tests more challenging, particularly when multiple boreholes are operating simultaneously [119]. The interference between closely located boreholes can significantly affect the accuracy of aquifer characterization, as highlighted in [120], which suggests that minimizing interference is crucial for reliable results in borehole testing.

The pumping flow rates range from 17 to 111 m³/h. Therefore, if 30 boreholes are simultaneously extracting water within this range, the actual classification of these areas falls between “high” and “moderate,” effectively ruling out the “low” classification. Furthermore, local hydraulic authorities typically do not undertake drilling projects unless their field assessments indicate a high groundwater potential. This further supports the reliability of the obtained results, as shown in Figure 5, where all boreholes are located in areas predicted as either “high” or “moderate”.

As a result, the validation criterion used in this study is the difference between the dynamic water level (DNL) and the static water level (STL) within the borehole, commonly referred to as borehole drawdown, denoted as Δ:

$$\Delta =DNL-STL$$

This $\Delta$ value indicates the rate of aquifer renewal, where a small Δ suggests rapid replenishment, indicating a high groundwater potential [121]. Conversely, a larger Δ implies slower water renewal in the well, signifying a lower aquifer potential. In fact, [121] showed that smaller drawdown values are typically associated with higher groundwater potential, as rapid replenishment is linked to more efficient aquifer systems. Additionally, [122] indicated that smaller drawdowns correlate with higher aquifer renewal rates and overall groundwater. According to

Table 12. Real classification of borehole drawdown compared with predicted classification.

Class of Drawdown	Drawdown Δ	Operating Flow Rate (m3/h)	Real Class	Predicted
Moderate (2)	51.8	60	1	1
Moderate (2)	12–22	18–107	19	14
High (3)	7–12	21–111	10	7

, a drawdown between 0 and 12 m is classified as high potential, meaning the area falls under Class 3 (high recharge potential). Notably, the smallest recorded drawdown is 7 m. A drawdown exceeding 12 m is categorized as moderate (Class 2).

One borehole, with a drawdown of 51.8 m, is an outlier in Table 12. Since 29 out of 30 boreholes exhibit drawdowns between 7 and 22 m, this particular value appears anomalous. However, it is still classified within the “moderate” category.

The comparison between the predicted classification of wells and the observed actual classification in terms of drawdown shows that out of 30 actual data points, 22 were correctly predicted, resulting in an accuracy rate of 73.3%.

A detailed analysis using the area under the receiver operating characteristic curve (AUC-ROC) curve approach generated the graph in Figure 6. The actual and predicted well classifications were converted into binary values to plot this ROC curve and compute the AUC. This graph, produced using a Python script, reveals a true positive rate (TPR) of 0.7 and a false positive rate (FPR) of 0.25 at the optimal classification threshold. The area under the ROC curve (AUC) is 0.72, indicating that the AHP/FAHP model has moderate to good performance, significantly surpassing random classification (AUC = 0.5) [123].

Similar applications of AUC-ROC for validating predictive models in groundwater studies have been demonstrated in previous works, for instance, in [124], where AUC values were used to assess the accuracy of groundwater potential maps generated using various models like FAHP (AUC = 65.1%), FR (Frequency Ratio)—AUC = 71.1%, and WOE (Weights of Evidence)—AUC = 71.4%. They suggest that the FAHP model, which showed a lower AUC than the FR and WOE models, could be improved by refining the selection of conditioning factors and incorporating additional spatially explicit data to better capture the complex relationships between the predictors and groundwater potential. Similarly, [50] validated their GPRZ model using AUC-ROC, with an AUC of 78%, indicating moderate to high accuracy in identifying groundwater recharge zones in the Manyara fractured aquifer. The AUC values in their study underscore the effectiveness of the AUC-ROC analysis in validating groundwater models and demonstrate its application in large-scale, semi-arid regions.

Several factors may explain the observed discrepancies between the actual and predicted classification, including the intrinsic limitations of the AHP/FAHP model’s criteria, uncertainties in the dataset, and potential subjectivity in the class definitions [43]. As noted by classification [80,81], the AHP and FAHP models are subject to limitations in the selection and weighting of influencing factors, which can contribute to discrepancies in groundwater recharge zone classification [80,81]. Additionally, local hydrogeological variations and resolution limitations in the mapping process could explain these differences. Therefore, further data refinement is necessary to improve the model’s predictive accuracy [125].

Although preliminary, this analysis highlights the value of the AHP/FAHP multi-criteria approach for initial groundwater predictions. The use of existing open-source data from online sources, such as Digital Elevation Models (DEMs), lithological maps, and soil types, demonstrates that for broad and preliminary assessments, these datasets are sufficient, as similarly suggested by [80,81] in their case study on groundwater recharge potential mapping. However, higher-resolution and locally sourced data are essential for strategic groundwater resource management to improve accuracy, reliability, and decision making.

4. Conclusions

Both AHP and FAHP methods effectively delineate groundwater recharge potential zones with nearly identical results, making both methods reliable for hydrological studies. In the Algiers watershed, the AHP method classifies 3.18% of the area (38.14 km²) as low recharge potential, 52.82% (634.93 km²) as moderate, and 44.01% (528.95 km²) as high recharge potential. Similarly, the FAHP method classifies 3.17% (38.17 km²) as low potential, 52.47% (630.71 km²) as moderate, and 44.35% (533.14 km²) as high potential. Given the close similarity in outcomes, AHP may be preferred for this analysis due to its simplicity and faster implementation. However, FAHP could become more valuable when dealing with complex or uncertain datasets. In this study, where both methods yielded nearly identical results, AHP was sufficient for practical applications. These findings provide critical insights for sustainable groundwater management, particularly in areas experiencing water scarcity. To optimize groundwater replenishment, management strategies should focus on enhancing recharge in flatter regions through methods such as Managed Aquifer Recharge (MAR) and soil retention techniques while implementing runoff control measures like reforestation in steeper areas to reduce surface runoff. Additionally, reducing groundwater extraction and promoting the restoration of natural vegetation along watersheds will further improve water retention and groundwater recharge. Continuous monitoring of groundwater levels and water quality will support informed decision making and the adjustment of strategies as needed. Although this study focuses on surface data for groundwater recharge assessment—except for lithology and soil type—it acknowledges the fundamental influence of subsurface characteristics, such as aquifer permeability, storage capacity, and depth to groundwater. Due to the scarcity of detailed regional hydrogeological data, an indirect approach is adopted by integrating surface parameters that serve as proxies for subsurface conditions. For instance, soil type and land cover are incorporated as indicators of infiltration capacity, while slope and drainage density reflect the potential for surface water retention and percolation. The lack of high-quality and accurate direct subsurface data is recognized as a limitation of this approach; however, the methodology remains applicable for preliminary assessments and decision making in data-scarce environments. Improving the resolution and precision of input datasets is essential for more reliable groundwater resource management. Indeed, the overall alignment between the predicted recharge potential and observed well classifications based on drawdown supports the reliability of the AHP/FAHP-based assessment. The model correctly classified 22 out of 30 wells, achieving an accuracy of 73.3%, with an AUC of 0.72, indicating moderate to good predictive performance. However, discrepancies remain. Several factors may contribute to these discrepancies, including intrinsic limitations in the AHP/FAHP model’s criteria, dataset uncertainties, and potential subjectivity in class definitions. Additionally, local hydrogeological variations and resolution limitations in the mapping process could further explain these differences.

These findings confirm that the AHP/FAHP-based recharge potential mapping serves as a valuable tool for groundwater resource assessment and management. However, further methodological improvements—such as integrating higher-resolution and calibrated local hydrogeological data—are necessary to enhance predictive accuracy and ensure more reliable classification. The availability of additional field measurement data from boreholes would enhance accuracy and better guide the development of the AHP/FAHP. In conclusion, the recharge potential map represents a valuable tool for guiding groundwater management and planning. This study highlights the strengths of AHP and FAHP in groundwater studies and identifies methodological improvement opportunities. By addressing these gaps, future studies can achieve even greater precision, thereby enhancing the sustainable management of water resources in the Algiers watershed and beyond.