Integrated Hydrochemistry and Statistical Methods to Investigate Groundwater Origin and Salinization Process in Arid Regions (Central-eastern Tunisia)

Hydrochemistry is a discipline widely used given the groundwater quantitative and qualitative reliability in the hydrogeological study. The geochemical study of groundwater in the Nadhour-Sisseb-El Alem basin aimed to characterize the water chemistry, determination of the physicochemical parameters and chemical facies well as and the mineralization processes. The Piper and Durov diagrams and scatter plots, conventional classication techniques, are applied to evaluate the geochemical processes. Samples are classied using two multivariate statistical methods, Principal Component Analysis (PCA) and Hierarchical Cluster Analysis (HCA). Waters compositions are affected by cation exchange reactions in the intercalated clay, resulting in a Na + increase, and peaks of K + , Ca 2+ and Mg 2+ . PCA analyses show that the water samples have been classied into 8 groups. The waters quality deterioration is caused essentially by; overexploitation, decreased in freshwater recharge rates, climate condition; height evaporation low precipitation, articial recharge by dam water, and irrigation return water.


Introduction
Groundwater, usually, is meteoric water stored in rock reservoirs. From their presence in atmosphere until their discharge (natural or arti cial) and during stays in the reservoir, the water inherits the mineralogical and isotopic characteristic of the crossed environments (Mustafa et al., 2016, Mouna et al., 2017 . The groundwater mineralization is affected by many factors; the natural processes (atmospheric input, the dissolution/precipitation, geology of the reservoir, groundwater recharge source and climate) and anthropogenic activities (agriculture, urbanization, industry) ( Increasing water demand, in arid and semi-arid regions due to agricultural, urban and industrial development, greatly stressed the water resources. The utility of water in drinking water, agriculture, domestic and industrial supply is related at the water quality ( (Jellalia et al., 2015). Indeed, Nadhour Sisseb El Alem region is characterized by a mean annual temperature of 20°C and a potential evapotranspiration of 1500 mm/year (INM, 2016). The mean annual precipitation is 350 mm indicating therefore a semi-arid climate (Souei, 2019). Study area is characterized by an absence of high mountains (Fig. 1). The study basin is considered among the important water reservoirs of the Central Tunisia (Fig.1). Water resources are used to satisfy the water demands of the Nadhour, Sisseb and El Alem regions and some cities of Sahel region, since the 70's.
The large expansion of irrigated area and the unfavorable climatic conditions increase the demand for groundwater. Indeed the deep aquifer is considered the main source of freshwater for different purposes In the last decades the Nadhour Sisseb El Alem deep aquifer system has been subjected to overexploitation, caused by development of illicit well, consequently, a piezometric drop of 20 m apparent (Souei, 2019;. In addition to the natural recharge, a process of arti cial recharge by water from the dam is carried out to compensate the piezometric drop (Houatmia et al Therefore, the chemical composition of the water and the mineralization process are affected by climate change and human activity. Study of the hydrochemical of water in the Nadhour-Sisseb-El Alem basin has numerous interests, in particular decrease the nancial managment costs (annual cost of deepening wells, consumption increasingly important energy), prevent the water degradation and soil salinization, and improve the speci c ow rates of wells. Knowledge of the hydrochemical process of groundwater in this area improved the use of groundwater and guiding the sustainable development of water resources and effective management.
The aim of this study was the determination of hydrochemical processes of groundwater, to clarify the relationship between the hydrogeological unit of the study area, to assess the effect of anthropogenic activity, and develop a conceptual model to support the management and development of water resources in the Nadhour Sisseb El Alem region.

General setting
The study area is characterized by a semi-arid Mediterranean climate, long dry periods, with an irregularity rainfall. The precipitation (2008)(2009) show rainfall varies between a minimum of 150mm characterizes the SW part of the basin and a maximum of 330mm record at the center-east part of the basin (Fig. 1). Generally, rainfalls not exceed 400 mm/year at the northern zone and 300 mm/year at the southern zone.
The Oligocene-Aquitanian series (Fortuna Formation) is represented by sandstone, sand and clay with silt (Yaich, 1992) (Fig. 2). The Burdigalian characterized by pedogenic red silte with root. The Langhian formed by the lumachellic calcareous bar of the Ain Grab Formation (Fig. 2).

Hydrochemical methods
Thirty-two samples were selected for chemical analysis. They were shipped to the Géoressource Laboratory (GR.), at Water Researches and Technologies Center Borj-Cedria Technopark-Tunisia, which using the standard method (Rodier, 2005).
The total hardness was determined by complexometric titration with EDTA. EDTA acts as a complexion reagent which forms soluble complexes with metal ions such as Ca 2+ and Mg 2+ . The Ca/Mg-EDTA complexes were stable at pH 8 to10 the pH of the solution during titration was maintained at 10 by addition of a buffer such as a solution of NH4OH using an Eriochrome black indicator. Ca 2+ and Mg 2+ were determined by the titration method using standard EDTA. The chloride was determined by the selective electrode method. The alkalinity and total alkalinity are determined by titration the samples with the HCl solution using a phenolphthalein indicator and a methyl orange indicator. The carbonate and bicarbonate were determined from the alkalinity. The ame photometer method is used in the determination of Na + and K + concentrations. The concentration of nitrate (NO 3 -) and nitrite (NO -2 ) is determined by colorimetry with a UV-visible spectrophotometer. Pie diagram, bivariate plots, Durov, Chadha and Piper plots are used in the determination of Groundwater facies classi cation and major processes control the global groundwater chemistry (Piper, 1944). The speciation code PHREEQC 2.8 is used in the calculation of the saturation indices of the principal mineral phases. Principal component analysis (PCA) and Hierarchical Cluster Analysis (HCA) are used to determine the classi cation of groundwater.

Statistical analysis
The graphical representation of the chemical elements and the study of characteristic ratios showed that a large number of chemical and physicochemical parameters were signi cant, taken individually (Hamzaoui et al., 2011). To explain the evolution of chemistry, most of the time one (or many), more or less strong link exists between these parameters. It was therefore interesting to try the data by statistical process, using a method of multivariate analysis as Principal Component Analysis (PCA). The intermediate correlation matrices and the projection of the variables in the space of the axes F1 and F2 were obtained with XLSTAT 2015 software.
PCA is a method of statistical study that allows the projection of observations into a space of a dimension smaller than the original space so that a maximum of information is conserved (Hamzaoui et al., 2011) (the information here is measured through the total variance of the point cloud) on the rst dimensions. In our geochemical study, each water sample is a statistical observation. The various parameters subject to analysis constitute the variables which characterize these statistical units (Cloutier et al., 2008;Hamzaoui et al., 2011). Our present study focuses on the major elements (variable) that in uence the evolution of mineralization (Davis, 1986).
The variables most contributing to the formation of an axis are those whose coordinates on this axis are close to 1 in absolute value. Similarly, the individuals contributing the most to the formation of an axis are those whose coordinates on this axis are the highest in absolute value.
The HCA is based on the Euclidean simple distance between various parameters (Ashley and Lloyd, 1978;Cloutier et al., 2008). It is used to classify water into groups. This classi cation is based on several methods in the present study we will use that of ward (Ward, 1963

Results
In this part, we will try to identify the various geochemical phenomena that can occur within the aquifers studied. We will use, for comparison purposes, all the available data and some relationships between the main major elements leading to the acquisition of the water salinity. The results of the groundwater chemical analyze are plotted on the Piper diagram (Fig. 4b) using the « DIAGRAMMES » software (Simler, 2013). Piper diagram indicates a Ca-Mg-Cl-SO4 water facies (Fig. 4b).
Durov and Chadha plots con rm that the groundwater type is Ca-Mg-Cl-SO4. The anion diagram shows that the majority of samples are characterized by a predominance of chlorides over the sulphates (Fig.  4b). The cation diagram shows the dominance of magnesium and relative dominance of sodium over the calcium (Fig. 4b). Dominance of Ca 2+ and Mg 2+ indicate that the water mineralization is affected by the inverse ion exchange process (Figs. 5 and 6). It's given by the exchange of Ca 2+ from the matrix by the Na + from the groundwater (Escolero, 2005).
Samples were plotted on a Durov diagram (Fig. 5). Fresh water is indicated by samples located in the upper left corner, old water is indicated by samples located in the lower right corner (Appelo and Postma 2005). The Chadha diagram (Chadha, 1999) show that the samples of Fresh water is located in the upper right corner, old water is shown by samples located in the lower left corner (Melloul and Goldenberg, 1998). Greater parts of samples are located along the mixing path between fresh water and old water (Fig. 6). Other samples are located outside the mixing path (Fig. 6). We determine that groundwater is dominated by brackish water (Fig. 6). We conclude that the mixing process of fresh water and old water is not the only process that affects the groundwater quality in the study area.
The Ca 2+ concentrations ranged from 32 to 541 mg/L, with an average concentration of 105 mg/L (Tab. 1). The average concentration of HCO 3 is 250 mg/L, with a maximum of 760 mg/L (Tab. 1).
The correlation between calcium and bicarbonate (Fig. 7) shows that the majorities of the samples are above the line of carbonate dissolution (slope 1). The Mg 2+ concentrations ranged from 85 to 491 mg/L, with an average concentration of 281 mg/L (Tab. 1). However, water has an excess of (Ca 2+ + Mg 2+ ) versus to HCO3 (Fig. 7).

Correlation matrix
Examination of the correlation matrix shows a weak correlation between the variables. We note the presence of only four correlations of absolute value greater than 0.5 (Tab. 2).
The proper value obtained for the factorial axes shows that there are three axes to be used (Tab. 3). The two factorial axes F1 and F2 are contribution of the total information of 49.42%, due to inertia of 27.73% and 21.69% respectively for F1 and F2. The factorial plane F1-F3 represents 37.99% of the total variance.

Discussions
This excess of Ca 2+ ions has other origin than the dissolution of carbonates, which is probably the dissolution of gypsum or the inverse ion exchange of Ca 2+ ions in favor of Na + ions (Fig. 9a)  However, this excess of (Ca 2+ + Mg 2+ ) versus to HCO3 could be the result of the basic exchange process (Fig. 9a). The presence of sulphates in water is related to the dissolution of the gypsum derived from Mio-Plio-Quaternary outcrop. The correlation between Ca 2+ and SO 4 2-shows an excess of SO 4 2-. This calcium de ciency may be related; Either to the precipitation of calcite or to the base exchange of Ca 2+ ions in favor of Na + and Mg 2+ between groundwater and the clay layers (Fig. 9a). The excess of Clversus to Na + can be reported to the in ltration of the evaporated water dam (Zammouri and Feki, 2005), which is relatively more salty (Kacem, 2008). During the in ltration of the dam water the Na + ions are adsorbed and Clions released, which explains the excess in chlorides. The Na + concentrations range from 22 to 97 mg/L, with an average of 48 mg/L (Tab.1).
The Cl/(Na + K) ratio (Fig. 10a)  potassium (KCl). The K + / Clratio (Fig. 10b) shows an excess of Cl + versus to K + thus demonstrating the mixing effect between the deep waters and the surface waters. Na + and Clions in the rural regions are habitually resulting from chemical fertilizers and animal waste (Rodvang et al., 2004).
The natural concentration of nitrate NO3 in groundwater is habitually very low (typically less than 10 mg/l) (USEPA, 1987), although nitrate concentrations is amplify among, agriculture and industry activity and the domestic e uents. Even if they are very low NO3 concentrations in groundwater can con rm the hypothesis of return ow irrigation waters (Stewart and Aitchison-Earl, 2020).
Results obtained from PCA shows there is a good correlation between total salinity and chloride, sodium, sulphate, magnesium and calcium. Other less signi cant correlations between the electrical conductivity and the bicarbonates (Tab. 3) deduced that the water salinity is due to salt formations such as gypsum, anhydrite and halite. This tendency is con rmed by the good correlation between the sulphates and the calcium of one part and between the sodium and the chloride of another part.
There is a good positive correlation between total salinity and magnesium on the one hand and total salinity and sodium on the other hand (Tab. 4). Correlation matrix shows a lower correlation between sodium and chloride. These two positive correlations between TDS/Na and Na/Cl indicate a high participation of halites in the mineralization of water. However, the low correlation between TDS/Cl indicates another source of Clthan dissolution of halite that may be due to the effect of evaporation.
Based on the results of the PCA (Fig. 11) we found that 2014 individuals "observation" are grouped into 8 groups. These eight groups describe the spatial evolution of water mineralization as one ow to the accumulation zone.
GI: This group is represented by four samples; F9, F10, F11 and F12 (Fig. 11) the rst sample is taken from wadi Nebhana, however the last three constitute the water of the western part of the deep groundwater of Sisseb. The grouping of these samples in the same pole, high mineralization (Fig. 11), indicates that Sisseb groundwaters may be coming from a common source. Almost similar concentration of the elements of this group, probably related to recharge by the salty waters of Nebhena dam.
GII: Presence of this group in this intermediate position (Fig. 11) between shallow waters (GI) and the deep waters of El Alem region (GIV) is probably due to the discharge of the Nabhena River in the El Alem plain. This shows the importance of the Nabhena River in the recharge of this aquifer.
GIII: Represents the groundwater of Sbikha region. The Sbikha aquifer is largely recharged by the rainwater. The position of this group in the diagram shows the role of the Miocene outcrops in the recharge of the aquifer (Fig. 11).
basin demonstrates that this group located at the downstream, presents the accumulation area of deep water from the basin. The waters from the group GVIII is alimented by the waters of the group GVII ow from the North and the waters from group GI derived from Nebhena dam ow from the west.
The statistical method (PCA) shows that the Sisseb region is a junction through which many came from waters represent by the three groups GI, GV and GVI. The waters of Nebhena (GI) percolation into aquifers in the Sisseb area towards the SE direction; followed the natural stream wadi Nebhana and progressively percolation towards the El Alem aquifer of the South; followed the diverted wadi Nebhana to Sbikha. The deep aquifer of El Alem is recharged by Nebhena water percolation from the North and horizontal percolating from the waters of the Sbikha deep aquifer came from the west.
The hierarchical classi cation (Fig. 12) of the samples shows that the waters of the basin are divided into two groups: -A rst group is presented by thirteen samples: S1, S2, S4, S5, S6, S7, S8, S9, S10, S11, S12, S15 and S17. This group formed the rst major groundwater ow direction between the GI, GII and GIV groups; water move from the Sidi Neji region toward the El Alem. -A second group presented by the other samples: S3, S13, S14, S16, S18, S19, S20, S21, S22, S23, S24, S25 S26, S27, S28, S29, S30, S31and S32 (Fig. 12). These groups present the second main direction of groundwater ow between groups GIII, GV, GVI, GVII and GVIII; the water moves from the upstream side of the basin, Nadhour region, towards Bled Saadia, crossing the Sisseb region. The HCA analyses con rmed the groundwater ow relationship deduced from the PCA analyses and determined the path of mineralization of water in the basin.

Mineralization conceptual model of groundwater in the Nadhour-Sisseb-El Alem basin
The Ca 2+ comes from leaching of the Jurassic, Cretaceous and Eocene carbonate outcrops and Triassic gypsum outcrops (Fig. 13). The decrease in Ca 2+ with the ow direction may be related to the precipitation of Ca 2+ of calcite and of Base Exchange.
The decrease in Mg 2+ with the ow direction (Fig. 13) is an indicator that the Mg 2+ is imported by inverse ion exchange; surface water and rainwater enrichment in Mg 2+ during the in ltration and recharge by surface waters initially loaded by the Mg 2+ (Kacem , 2008) comes from leaching of the Cretaceous and Eocene outcrops (Fig. 13). This will have a vertical enrichment during the in ltration process and lateral discharge during lateral ow (Fig. 13). In the rst phase is the dissolution of Mg 2+ by the runoff water and enrichment by the inverse ion exchange during ow and in ltration the second phase is the decrease of Mg 2+ by the Base Exchange process and the precipitation of Mg 2+ as aragonite and calcite.
Bicarbonates HCO3in water is due to the dissolving of the carbonate and carbonate-sandstone rocks which borders the region (Fig. 13). The sulphate in water is probably related to the dissolution of the gypsum derived from Triassic, Eocene and Mio-Plio-Quaternary outcrop. The excess of SO 4 2 is related to the evaporation of surface and shallow groundwater. Chloride (Cl -) is derived from dissolution of the gypsum derived from Triassic, Eocene and Mio-Plio-Quaternary outcrop. Enrichment by Clreported to the evaporation of water dam and shallow groundwater (Fig. 11). The Potassium outcome from the alteration of Eocene and Miocene clays, the dissolution of NPK arti cial fertilizers and the dissolution of chloride in potassium (KCl). Na + and Clions are generally resulting from synthetic fertilizers and animal waste. The excess of (Na + K) is related to the Base Exchange.
The recent deterioration of the waters quality caused by overexploitation, decreased fresh water recharge rates, climate condition; height evaporation low precipitation, arti cial recharge by dam water, and irrigation return water the intensive use of chemical fertilizers. Climate change and exhaustive groundwater mobilization, although supplying the water requirements in drinking water, agriculture, industry and tourism, with are the origins of depletion and deterioration of water quality and piezometric level increase in this basin.

Conclusions And Perspectives
The groundwater hydrochemical analysis in this region helped to identify and characterize the mineralization processes of groundwater. The mineralization groundwater of the aquifer system is guided by several processes the most important of is minerals dissolution/precipitation process, in a less important order the ion-exchange, and water mixing process with low intervention. The hydrogeochemical and statistical methods present the main tools used in the identi cation of these processes. Water samples in groups I, III, and VI present higher salinity compared to other groups. his shows that the water quality of groups II, IV, V, VII, and VIII is slowly getting to degradation. Processes mineralizations are accelerate by the constant increase of extraction rates and arti cial recharge with dam water, relatively most charged.
The various analyses techniques used in this study con rmed that the deterioration of groundwater quality in this basin due to the anthropogenic activities; overexploitation, arti cial recharge, water irrigation return…This anthropogenic activities are ampli ed by climate change; low precipitation, height evaporation rat.
Groundwater deterioration could be reduced by the decreasing pumping rates, limited number of well by creation of public irrigate area, management of recharge area by implantation of tree, management in the wadi bed, constructing more wells in bordure of basin, in this area water is more fresh; rapid in ltration.
Decreases the storage period of water, intended to the arti cial recharge, in dams. Mobilizes drinking water from boreholes locates near to the recharge areas.