Figure 1. Location of the study area.

The study area is located in central western Senegal, in the departments of Mbour and Fatick. It encompasses part of the districts of Sindia, Sessene, Tattaguine, and Fimela, covering an area of approximately 986 km². It is located between longitudes 17˚02'12.6"W and 16˚35'06.7"W and latitudes 14˚27'11.3"N and 14˚35'13.9"N. It is bordered to the west by the Atlantic Ocean, to the north by the districts of Fissel and Notto, to the east by Niakhar and Diakhao, and to the south by Niodior (Figure 1).

The geology of the study area is part of the Senegal-Mauritania sedimentary basin, which is well known through numerous oil and water drilling investigations. These investigations have led to lithostratigraphic reconstruction and the identification of deposit environments. The study area is characterized by the presence of Quaternary and Continental Terminal geological formations consisting of detrital soils (laterite, sand, and clay) with a predominance of clay and sand. In the study area, the Eocene is heavily eroded and is represented only by its lower terms, where it consists of the following levels:

¬ A layer consisting of clayey limestone, marl, and clay, often enriched with phosphate or silica, which lies directly on top of the Paleocene formations ([58]);

¬ A clayey or marly layer with some very frequent limestone intercalations in the upper part;

¬ A layer consisting of alternating limestone and marly limestone, found mainly in the Ngazobil area ([72]). The Lower and Middle Eocene are characterized by the presence of phosphate levels ([73]). However, the base of the Lower Eocene consists of gray marl and clay with intercalations of flint ([52]).

The Paleocene, in the study area, has some very distinctive features. It outcrops at Mbour and consists of limestone, clayey limestone, and marl, which are frequently enriched with flint, glauconite, and phosphate ([52]) and sinks towards the east and southeast, where it is covered by Eocene, Continental Terminal, and Quaternary formations. It is characterized by homogeneous facies of limestone and marly limestone, often with shell deposits ([44]). These characteristics are indicative of shallow oceanic deposition with benthic microfauna dominated by ostracods ([15]). The base of the Paleocene consists of hard limestone and gray marly limestone that may be sandy ([59]). In contrast, the Middle and Upper Paleocene are consistently formed of numerous sandy limestone levels of varying fineness and rich in shell debris ([58]). Phosphate deposits mark the upper limits of the Paleocene limestones in the Thiadiaye and Fatick areas ([44]).

The Paleocene-Eocene boundary is marked by the sudden disappearance of microfauna and macrofauna and the appearance of facies that are very poor in fossils and more distinctly marine ([58]). The transition from Paleocene limestones to Lower Eocene marls is often marked by the deposition of a marl-clay layer containing flint, particularly in the northeastern part of the local area ([48]).

¬ In the hydrogeology context, two aquifer systems are found in the study area:

a shallow aquifer system consisting of the Miocene-Pliocene, Quaternary, and Upper Ypresian aquifers, which are mainly tapped by traditional wells to meet the domestic and agricultural water needs of the population. The Upper Ypresian aquifer is formed of marly limestone, while the Miocene-Pliocene and Quaternary aquifers are formed of clayey sand. The Eocene limestone aquifer is distinct from or associated with the Miocene-Quaternary aquifer.

¬ A deep aquifer system, mainly tapped by modern boreholes and wells, and consisting of the deep aquifer of the Paleocene and Maastrichtian. Due to falling water levels or salinization observed in some areas of the shallow aquifers, the Paleocene aquifer is currently the most tapped. The latter is relatively thin and lies on Maastrichtian sandstone-clay sediments. It consists of limestone, clayey limestone, and marl with flint, glauconite, and phosphate ([52]). Following a change in facies, its base, mainly composed of marl, consists of shell limestone in the west and marl in the northwest, east, and south ([68]). It has the following hydraulic characteristics: transmissivity varies from 5 × 10⁻⁵ m²/s in Joal to 10⁻² m²/s in Louli Mbafaye ([57]), while storage coefficients range from 1 × 10⁻⁴ to 7 × 10⁻². Hydraulic conductivity values range from 6.6 × 10⁻⁶ to 2.0 × 10⁻² m/s. In the Miocene-Pliocene-Quaternary aquifer, the average values for hydraulic conductivity and effective porosity are approximately 1.5 × 10⁻⁴ m/s and 20%, respectively.

A sampling campaign was carried out from May 30 to June 1, 2023, during which 42 samples were collected from various sampling points, including 15 boreholes and 27 wells. These sampling points were carefully selected to ensure coverage of the entire study area (Figure 2). The samples collected at each sampling point were filtered using a 0.45 µm pore membrane to remove suspended solids, then divided between two labeled polyethylene bottles that had been washed two or three times with the water to be collected in order to avoid contamination. Samples intended for cation analysis were acidified to lower the pH of the water below 2 by adding a 0.05 N hydrochloric acid solution in order to diminish the bacterial activities, while those intended for anion analysis were not acidified. After sampling, the collected samples were stored in coolers and then transported to the laboratory for chemical analysis. Electrical conductivity, pH, and temperature were measured in the field using a WTW multi 350i portable multi-parameter device. Chemical analyses of the collected samples were performed at the Geo-Environment Laboratory of the University of Lille in France. Anions were analyzed using Dionex-100 ion chromatography, while cations and some metals were determined using inductively coupled plasma atomic emission spectrometry (ICP-AES). Carbonates and bicarbonates were analyzed by titrimetry using a 0.05 N sulfuric acid solution. The reliability of the chemical analyses was checked by calculating the ion balance error for each water sample using equation 1. The results indicate that three samples have ion balance error values above the acceptable limit of ±10% ([78]; [81]) and therefore, 39 out of 42 samples were ultimately considered in the data processing.

Figure 2. Sampling points in the study area.

Multivariate statistical analysis is a method that allows a large dataset to be reduced and organized, with minimal loss of critical information, into groups with similar characteristics ([19]; [62]; [66]). It is also used to determine similarities between water samples and water parameters ([10]; [43]). In this study, 12 chemical parameters of water, pH, CE, HCO 3 − , Cl − , SO 4 2 − , F − , N O 3 − , Ca 2+ , Mg 2+ , Na + , K + , and TDS were used to perform Principal Component Analysis (PCA), Ascending Hierarchical Analysis (AHA), and a correlation matrix. However, the data were first standardized using Equation (2) to put them on the same uniform scale ([46]) in order to facilitate their interpretation and comparison.

where Z i is a standardized variable, X i is the raw variable, X m is the mean of the variable, σ i is the standard deviation.

3.2.1. Correlation Matrix

The correlation matrix is an effective statistical tool commonly used in data analysis to highlight relationships between different parameters. Pearson’s correlation analysis is widely used to assess and determine the strength of a linear relationship between two variables ([61]). A correlation is considered perfect when the correlation coefficient (r) between two variables is equal to +1 or −1 ([43]). According to [23], the strength of the correlation between variables can be divided into three levels: very high (if 0.8 ≤ r ≤ 1), high (if 0.6 ≤r < 0.8), and low (if r< 0.6). The correlation matrix is determined according to Equation (3) ([75]):

where r is the correlation coefficient, n is the number of variables in the sample, x i and y i are the two variables to be analyzed, x ¯ and y ¯ are the mean values of x and y, σ x and σ y are the standard deviations of x and y.

3.2.2. Principal Component Analysis (PCA)

Multivariate statistical analysis is a technique commonly used in numerous hydrochemical studies to reduce data dimensionality ([24]; [26]) and is essential for obtaining data on the relationships between hydrochemical variables ([56]). This method involves transforming the initial variables into a new coordinate system called principal components. In principal component analysis, the variable with the greatest variation is designated as the first principal component ([75]). The second variable with the next highest variance is identified as the second principal component, and this pattern continues successively. Taking into account the loading values, we observe a categorization into three distinct levels: strong (>0.7), moderate (0.7 - 0.5), and weak (0.5 - 0.3) ([8]; [12]). This classification shows the influence of each variable on the different principal components. Principal components with an eigenvalue greater than or equal to one (1) are considered significant and are therefore retained ([22]; [63]).

3.2.3. Hierarchical Cluster Analysis (HCA)

Hierarchical cluster analysis is also a multivariate statistical method that allows variables or water samples to be classified into groups called clusters based on their similarity and dissimilarity to other groups ([30]; [33]). It is the main method for identifying relatively homogeneous clusters based on measured data ([11]). It is the most commonly used clustering method, which identifies the initial relationship between any sample and all data relating to water samples ([33]). In this study, Euclidean distance was used as the measurement distance, allowing the proximity between different samples to be assessed. Furthermore, Ward’s method was adopted to form more or less homogeneous clusters and link them together. This method evaluates the distance between clusters while minimizing the sum of the squares of the distances between two clusters formed at each stage of the classification ([17]). Previous studies ([18]; [80]) have shown that the use of hierarchical cluster analysis, combining Euclidean distance as a distance measure and Ward’s method as a linkage rule, leads to the formation of more distinctive groups or clusters. The samples with the greatest similarity are first grouped together, and so on. This process results in the construction of a dendrogram, which is a hierarchical visual representation of the similarity relationships between the data. The number of clusters is determined based on the phenon line ([17]; [42]), and changing the location of this line on the dendrogram affects the number of clusters resulting from the classification. The Euclidean distance can be computed using Equation (4) ([33]):

whereiand k are two different water samples, j is the mode of the variables, and v is the value of the variables. The smaller the Euclidean distance D, the more similar the water samples are. The value D quantitatively indicates the degree of similarity between the water samples.

Water quality index (WQI) is an important parameter generally computed to assess the overall water quality in a given area based on several water quality parameters. It is widely used as a tool for assessing and comparing the suitability of water for different uses, such as consumption ([78]; [3]), irrigation, and the state of aquatic ecosystems ([3]). It considers several chemical parameters and assigns them a weight ([60]). In the present study, the water quality index was computed using eleven (11) parameters: pH, HCO 3 − , Cl − , SO 4 2 − , F − , NO 3 − , Ca²⁺, Mg²⁺, Na + , K⁺, and TDS. The calculation of the WQI involved the following steps ([7]; [47]):

Π The first step involves calculating the unit weight (Wn) of each water quality parameter. This step quantifies the effect of each parameter on overall water quality. The unit weight is computed using Equation (5).

where S n is the permissible standard value according to [76], for a water quality parameter (Table 1). K is the proportionality constant derived from Equation (6).

Table 1. Relative weight of chemical parameters.

Π The second step defines the specific quality index for each quality parameter. These indices reflect the relative concentration and individual impact of the different parameters. The specific quality index ( q n ) is computed using Equation (7).

whereV_n is the observed value of each parameter in each sample, S_n is the acceptable limit for each variable, and V_id is the ideal value of the i^th parameter in pure water. The ideal value for all parameters is 0, except for pH, which has an ideal value of 7 ([47]).

Π The third step involves calculating the Water Quality Index (WQI) using Equation (8).

∑ i = 1 n W n = 1

Table 2.Classification of groundwater quality based on WQI.

Water quality is classified into five (5) categories ([4]; [7]; [35]): excellent, good, poor, very poor, and unfit for human consumption (Table 2).

Descriptive statistics for the physicochemical parameters are represented in Table 3. The pH values range from 6.89 to 8.2 with a median value of 7.8, indicating that the water is neutral to slightly alkaline. In the study area, all pH values are within the acceptable pH limits set by the [76]. Electrical conductivity is the ability of an aqueous solution to conduct electrical current ([38]) and is an important parameter for assessing water quality ([74]). It provides information on the ionic concentration of water. In this study, electrical conductivity ranged from 127 to 5800 µS/cm, with an average of 1524.4 µS/cm and a median of 1107 µS/cm. Approximately 26% of water samples exceeded the WHO permissible limit of 1500 µS/cm. TDS values range from 85.1 mg/L to 3886 mg/L, with average and median values of 1021.4 mg/L and 741.7 mg/L, respectively. Around 26% of water samples exceed the WHO limit. Chloride concentrations in water range from 17.5 to 1556 mg/L, with mean and median values of 267.3 mg/L and 129 mg/L, respectively. NO 3 − content varies from 0.5 to 425 mg/L, with an average of 54.6 mg/L and a median of 24.2 mg/L. Sulfate concentrations range from 1.8 to 587 mg/L, with mean and median values of 69.6 mg/L and 37.6 mg/L, respectively. HCO 3 − concentrations range from 18.3 to 457.5 mg/L, with mean and median values of 330.5 mg/L and 366 mg/L, respectively. Ca²⁺ varied from 13.5 to 353 mg/L, with mean and median values of 105.5 mg/L and 79.1 mg/L, respectively. Magnesium (Mg²⁺) concentrations range from 0.7 to 188 mg/L, with mean and median values of 44.8 mg/L and 45.5 mg/L. For Na⁺ and K⁺ ions, their concentration ranges vary from 13.4 to 609.4 mg/L and 0.1 to 23 mg/L, respectively, with associated mean and median values of 116.4 mg/L and 65.8 mg/L for sodium, and 4.7 mg/L and 2.3 mg/L for potassium. The order of abundance of ions in groundwater is defined as follows: HCO 3 − > Cl − > SO 4 2 − > NO 3 − for anions, Na⁺ > Ca²⁺ > Mg²⁺ > K⁺ for cations.

Table 3. Descriptive statistics of physicochemical parameters.

Determining the chemical facies of water is an essential step in understanding and identifying the geological and geochemical factors that influence the hydrogeochemical characteristics of groundwater ([25]). The Piper trilinear diagram reveals four main hydrochemical facies that together describe the evolutionary trends of groundwater in the study area. The Ca-HCO₃ type is dominant (54% of samples) and characterizes wells and boreholes tapping relatively less mineralized waters, typically associated with recharge areas or shorter residence times in carbonate-rich strata. Mixed Ca-Mg-Cl waters (26%) and Ca-Cl waters (10%) represent more evolved compositions, indicative of enhanced interaction with the aquifer matrix and progressive salinization along flow paths.

Figure 3. Piper trilinear diagram of groundwater samples.

Na-Cl waters account for 10% of the samples and mark locally more saline conditions, which may result from a combination of halite dissolution, ion exchange, and, in some sectors, the influence of older, more stagnant groundwater (Figure 3). The transition from the Ca-HCO₃ facies found in recharge areas to the mixed Ca-Mg-Cl facies may be related to the dissolution of carbonate rock minerals (calcite, dolomite) and/or the alteration of silicates during water infiltration or along the flow paths. The occurrence of the mixed Ca-Mg-Cl facies in this study may be associated with a mixing phenomenon observed in areas where rock dissolution is combined with the diffusion of brackish water. This mixed Ca-Mg-Cl facies changes to a Na-Cl facies in certain areas, particularly near the sea. This transition to the Na-Cl facies may be due to the ion exchange process that occurs during marine intrusion, but also to the evaporation process in areas where the aquifer is shallow. The coexistence of these four facies within a relatively small area suggests that the aquifer system is hydrogeochemically heterogeneous, reflecting the superposition of shallow and deep flow systems, lithological variability, and spatially variable anthropogenic inputs.

The main natural processes that may lead to variations in the chemical composition of water are evaporation, precipitation, and water-rock interaction ([37]), which are often determined using the Gibbs diagram (1970). The Gibbs diagram highlights the relationship between total dissolved solids (TDS) and the ratio ( Cl − / ( Cl − + HCO 3 − ) ) firstly and the ratio ( Na + / ( Na + + Ca 2 + ) ) secondly. This diagram is widely used in numerous studies ([29]; [32]; [65]; [67]) around the world in different geological environments to identify the main natural factors controlling water mineralization. In this study, the plot of chemical data on the Gibbs diagram shows that the majority of water samples (74%) are located within the water-rock interaction zone (Figure 4), suggesting possible mineral dissolution through chemical weathering ([51]). However, 23% of the samples are located in the evaporation zone, indicating an influence of this process on water chemistry (Figure 4). Thus, it may be argued that water-rock interaction is the primary mechanism governing the variation in groundwater chemical composition, followed by evaporation as the secondary mechanism.

Figure 4. Gibbs plot for mechanisms controlling groundwater chemistry.

BivariateDiagram

Relationships between major ions are often used to understand hydrogeochemical evolution processes and the mechanisms controlling groundwater chemistry ([83]). Mineral dissolution is an important process in aquifers that can lead to changes in the chemical composition of groundwater. For example, the dissolution of halite leads to an equivalent release of Na⁺ and Cl⁻ into groundwater ([83]). When halite dissolution occurs, the points lie on the 1:1 line on the Na⁺ vs. Cl⁻ graph. If the points on the Na⁺ vs. Cl⁻ graph are above the 1:1 line, this indicates an excess of Na⁺, which may result from silicate weathering or the base ion exchange process between Na⁺ and Ca²⁺ or Mg²⁺ ([53]; [83]). However, if the points lie below the 1:1 line, this indicates chloride enrichment in the water, which may be due to natural phenomena such as marine intrusion ([14]) or anthropogenic pollution. Figure 5 shows that most points lie below the 1:1 line, suggesting a sodium deficit relative to chloride. However, some points lie on the 1:1 line, indicating equal concentrations of sodium and chloride in the water, which could result from the dissolution of halite. Some points lie slightly above the 1:1 line, indicating an excess of sodium in the water relative to chloride, which could be related to ion exchange processes or silicate weathering.

Figure 5.Binary diagram Na⁺ vs Cl⁻.

IonExchange

Ion exchange is a significant process that can alter the chemical composition of groundwater ([85]). In order to assess the ion exchange process, the bivariate diagram ((Ca + Mg) − (HCO₃ + SO₄)) vs. ((Na + K) − Cl) is commonly used. If ion exchange plays an important role, then a linear relationship with a slope of −1 is observed between the parameters. In this study, Figure 6 shows that the majority of points lie on or near a straight line with equation y = −0.9343x − 0.2034 with a slope of −0.93, which is very close to −1, and a correlation coefficient of r² = 0.85. This observation suggests that the ion exchange process is important and plays a crucial role in controlling the variation in the chemical composition of groundwater.

Figure 6. ((Ca + Mg) − (HCO₃ + SO₄)) vs ((Na + K) − Cl).

The influence of the ion exchange process was also checked by calculating the chlorine-alkaline indices (CAI). These are important tools commonly used to evaluate ion exchange reactions between groundwater and the aquifer matrix ([82]). Ion exchanges are described by reactions 9 and 10, while ion exchange indices are calculated using Equations (11) and (12) ([83]).

If CAI > 0, this indicates reverse ion exchange, where Na⁺ and K⁺ ions present in groundwater are trapped by the aquifer matrix, while Ca²⁺ and Mg²⁺ ions are released into the groundwater. Conversely, if CAI < 0, this indicates direct ion exchange, where Na⁺ and K⁺ ions are released into the water while Ca²⁺ and Mg²⁺ ions are fixed to the matrix ([9]; [47]).

In this study, 77% of groundwater samples displayed positive CAI, indicating reverse ion exchange where K⁺ and Na⁺ ions are fixed by the aquifer matrix while Ca²⁺ and Mg²⁺ ions are released into the groundwater. In contrast, 23% of the samples showed negative CAI, indicating direct ion exchange where the matrix fixes Ca²⁺ and Mg²⁺ ions and releases Na⁺ and K⁺ ions. These observations suggest that the reverse ion exchange process predominates in the groundwater of the study area (Figure 7).

Figure 7. Chloro-alkaline indices diagram CAI-I vs CAI-II.

Figure 8.(Ca + Mg) vs (HCO₃ + SO₄).

Weathering of carbonate, silicate, and sulfide minerals and dissolution of evaporites are commonly the main lithogenic sources of ions in water. The binary diagram (Ca + Mg) versus (HCO₃ + SO₄) can be used to study the relative importance of ion exchange and weathering processes ([28]). This relationship highlights the relative contributions of calcium, magnesium, bicarbonate, and sulfate ions to the chemical composition of groundwater ([41]). Figure 8 shows that most samples fall above and around the 1:1 line, indicating that, in addition to the weathering of carbonate minerals, the excess Ca²⁺ and Mg²⁺ in groundwater could be related to the reverse ion exchange process. However, some samples lie below the 1:1 line, suggesting that silicate weathering also contributes to the variation in the chemical composition of groundwater.

SaturationIndex

The extent to which water-rock interactions influence the chemical composition of groundwater depends on the magnitude of mineral dissolution and precipitation phenomena, which are determined by the saturation index (SI) ([31]). The SI is an essential parameter in geochemistry that describes the quantitative deviation of water from the equilibrium of a specific dissolved mineral ([45]; [20]). Widely used in the study of geochemical processes, the saturation index is determined using Equation (13) ([29]; [83]).

where K I A P represents the ionic activity product for a mineral and K S P is the solubility product constant for the mineral.

When IS < 0, this indicates a solution that is undersaturated with respect to the mineral, promoting its dissolution. Conversely, if IS > 0, this indicates a solution that is supersaturated with respect to the mineral, and precipitation prevails. When IS = 0, this means that the mineral is in equilibrium with the solution ([83]). In this study, PHREEQC software was used to calculate saturation indices. Owing to the geological context of the study area, which is mainly composed of limestone and marl-limestone, the dissolution of these rocks may be the main source of groundwater mineralization. Bicarbonates in groundwater generally originate from the dissolution of carbonate minerals in aquifers ([6]). The dissolution of carbonate minerals is described by Equations (14) - (16) ([32]; [36]).

Results of the saturation index calculations are presented in Table 4.

Saturation indices (SI), calculated with respect to calcite, dolomite, halite, and fluorite, further constrain the extent of dissolution and precipitation processes. Mean SI values for calcite (3.15), dolomite (5.97), and fluorite (2.48) are positive, indicating that groundwater is generally supersaturated with respect to these minerals and that precipitation or partial equilibrium conditions are likely to occur along flow paths. In contrast, halite exhibits a negative mean SI (−3.88), consistent with persistent undersaturation and ongoing dissolution where halite is present in the geological formations.

Given the carbonate dominated lithology of the study area, these results are consistent with an initial phase of carbonate dissolution during recharge, followed by a progressive approach to equilibrium and local precipitation of calcite and dolomite in more evolved waters. The supersaturation with respect to fluorite suggests that fluoride concentrations in groundwater are controlled by complex interactions among fluorite dissolution-precipitation, pH-dependent sorption-desorption reactions, and competition with other anions such as bicarbonate and hydroxide.

Table 4. Statistics of mineral saturation index.

Results of the water quality index are shown in Table 5. These results show that the majority of groundwater samples (64%) (F2, F4 - F6, F8 - F10, F12, F14 - F16, F18, P1, P6 - P10, P17, P18, P21, P23, P24, P26, and P27) are unfit for human consumption (class E). However, 10% of the groundwater samples (P2, P14, P15, and P16) had excellent water quality (class A); 13% of the groundwater samples (F1, F3, P11, P20, and P25) had good water quality (class B); 5% of groundwater samples (P4 and P2) have poor water quality (class C); 8% of groundwater samples (P3, P5, and P22) have very poor water quality (class D).

Spatial distribution of water quality indices (Figure 9) shows that unsuitable water for human consumption is spread across the entire study area, with the exception of certain parts, particularly in the northwest, south, and at certain locations such as Keur Yerim (P2), Ndiol Khokhane (P11) and Roff (P25), where the water is of excellent quality and good quality, respectively, for the latter two points. In the area, the decline in water quality is mainly linked to the high fluoride concentrations recorded in several localities and, to a certain extent, to nitrate concentrations, which can have adverse effects on human health. In fact, consuming water polluted with nitrate can lead to diseases such as methemoglobinemia, thyroid problems, and cancer ([40]; [69]), while consuming water polluted with fluoride can cause dental and skeletal damage ([49]). In addition, the high proportion of samples falling into the “unfit for drinking” class is comparable to results obtained in other semi-arid basins ([21]; [27]; [84]) where WQI is used to evaluate groundwater quality, although the relative contribution of each ion may differ depending on land use and lithology. This underlines the need to not only report WQI values, but also to interpret them in the light of the underlying hydrogeochemical processes and human pressures.

Table 5. Classification of groundwater samples based on water quality index.

Figure 9. Spatial distribution of water quality index in the study area.

CorrelationAnalysisofPhysicochemical Parameters

Pearson’s correlation analysis (Figure 10) highlights strong, statistically significant positive correlations between EC, TDS, and most major ions, confirming that these species jointly contribute to groundwater mineralization. In particular, EC correlates strongly with Cl − , Na⁺, Ca²⁺, Mg²⁺, SO 4 2 − , NO 3 − and K⁺, underlining that increases in salinity reflect both natural water-rock interaction and anthropogenic inputs. A very strong correlation between pH and HCO 3 − indicates that bicarbonate is the main driver of groundwater alkalinity and suggests extensive carbonate dissolution and CO 2 − water interactions in the unsaturated and saturated zones. This alkalinity promotes fluoride enrichment in groundwater ([2]), as evidenced by the positive correlations between F⁻ and pH (r = 0.61; p < 0.001), as well as F⁻ and HCO 3 − (r = 0.67; p < 0.001). There is also a strong positive correlation between Na⁺ and Cl − concentrations (r = 0.95; p < 0.001). This association suggests a common origin for these two ions in groundwater ([53]). Furthermore, the negative correlation between calcium (Ca²⁺) and bicarbonates ( HCO 3 − ) (r = −0.21; p > 0.05) suggests that calcite (CaCO₃) is not the main source of Ca²⁺ in groundwater. Bicarbonates could result from the dissolution of carbonate rocks by carbon dioxide (CO₂) rich waters. The positive correlation between nitrate ( NO 3 − ) and calcium (Ca²⁺) (r = 0.62; p < 0.001), as well as between nitrate and chloride (r = 0.66; p < 0.001), indicates a possible anthropogenic origin linked to agricultural activities or domestic waste ([71]). Finally, the positive correlation between sodium (Na⁺) and potassium (K⁺) (r = 0.84; p < 0.001) suggests a possible common origin for these ions, which could be related to silicate weathering ([50]). Potassium (K⁺) is significantly correlated with chloride Cl⁻ (r = 0.75; p < 0.001), indicating that a higher concentration of K⁺ can lead to an increase in chloride concentration ([16]).

Figure 10. Pearson correlation matrix of physicochemical parameters.

PrincipalComponent Analysis

PCA extracted two principal components with eigenvalues greater than 1, together explaining 80.54% of the total variance in the dataset, thereby providing a compact yet robust representation of the controlling processes. The results of the principal component analysis are presented in Table 6, which shows the eigenvalues, variances, and cumulative variances. The first component (PC1) shows high positive loadings on EC, TDS, Na⁺, Cl⁻, Ca²⁺, Mg²⁺, SO 4 2 − and NO 3 − , and can be interpreted as a “mineralization-anthropogenic” factor that integrates both natural salinities built up and superimposed contamination by nitrate and other ions of anthropogenic origin. The second component (PC2), with strong loadings on pH, HCO 3 − and F⁻, represents an “alkalinity-fluoride” factor, specifically discriminating fluoride-rich, alkaline waters from the rest of the dataset (Figures 11-13). Alkalinity, which is linked to the bicarbonate enrichment of water, promotes anion exchange between F⁻ and OH⁻, with the release of F⁻ and the fixation of OH⁻ in the aquifer matrix, thereby leading to an increase in fluoride concentration. This explains the strong positive correlation between pH and fluoride, but also between F⁻ and HCO 3 − .

Table 6. Eigenvalues, variance and cumulative variance of the selected components.

Figure 11.Variables’ contribution to the first two principal components.

Figure 12.Correlation matrix between variables and principal components.

Figure 13.Variables projection on the factorial plane (Dim1 - Dim2).

HierarchicalCluster Analysis

Hierarchical cluster analysis classifies water samples into two (2) distinct clusters based on their chemical characteristics. Cluster 1: includes boreholes F12, F15, F16, F18, P15, P18, and well P22 (Figure 14) and represents 18% of groundwater samples. It is characterized by highly mineralized water with high average values for TDS (2642.5 mg/L), electrical conductivity (3944 μS/cm), sodium (361 mg/L), chloride (919.7 mg/L), calcium (231.2 mg/L), sulfate (222.3 mg/L), and nitrate (180.6 mg/L) (Table 7). These chemical parameters, combined with TDS and electrical conductivity (EC), reveal that cluster 1 is linked to groundwater mineralization, which may result from natural processes such as water-rock interaction, but also from anthropogenic pollution.

Cluster 2: includes boreholes F1 - F6, F8 - F10, F14, and wells P1 - P12, P14, P16, P17, P20, P21, P23 - P27, and represents 82% of water samples (Figure 14). Cluster 2 is associated with moderately mineralized, alkaline waters with a dominant mixed HCO 3 − facies. It includes water samples whose physicochemical parameters have lower average values than those in cluster 1, with the exception of bicarbonates, fluoride, and pH, which have average values of 332.8 mg/L, 3.2 mg/L, and 7.8 mg/L, respectively (Table 7).

Table 7. Physicochemical characteristics of each cluster.

Figure 14. Dendrogram generated through hierarchical cluster analysis of groundwater samples.