Evaluation Study of Ammonium Removal from Groundwater by Electrodialysis: Case Study of Real Groundwater from the City of Kenitra in Morocco

Abstract In this work, we investigated the feasibility of ammonium removal by electrodialysis (ED), a well‐known electro‐membrane process, based on the selective migration of anions and cations through anion exchange membranes (AEMs) and cation exchange membranes (CEMs). ED experiments are performed using a laboratory pilot. The ion exchange membranes (IEMs) pair used is AXE/CMX, AXE as an AEMs and CMX as a CEMs. The first tests are performed with real groundwater solutions from Kenitra city (Morocco), spiked with an initial concentration of 3 mg/L NH4Cl. The results gave a specific demineralization (SD) to NH4 + ions of 84.60 %; for a demineralization rate (DR) of 80 % and a time of 60 min. The multivariate principal component analysis (PCA) presents a total inertia of 99.23 %, the majority of the variables of which are positively correlated on the C1 axis with a variance of 95.5 % than that of C2 of 3.78 %. The quality of the diluted water determined by the Legrand‐Poirier method showed that the water was aggressive and that the addition of 4.54 mg/L Ca2+ was necessary to balance the water and make it fit for human consumption.


Introduction
In Morocco, the agricultural sector occupies an important socioeconomic position; however, the evolution of farming practices is closely linked to the substantial demand for water resources needed to irrigate farmland.Previous studies have highlighted the excessive importance attached to intensive farming techniques and productivist models, which often neglect the conservation of natural resources. [1]Examples include the Ziz basin in south-eastern Morocco [2] and the Souss-Massa basin in central-western Morocco, [3] both of which are facing increasing pressure from excessive groundwater use and investment.
What's more, the country's water scarcity crisis is set to intensify, with an anticipated 80 % depletion of water resources in the coming decades, mainly due to reduced rainfall associated with global warming and overexploitation of groundwater. [4]hese problems are exacerbated by the widespread use of chemical fertilizers, particularly nitrogen-based, in agriculture, which results in significant contamination of water resources surrounding agricultural areas.In addition, the discharge of insufficiently treated wastewater is another source of pollution of water resources.It is imperative that Morocco effectively treats its wastewater before discharging it into the natural environment, to avoid further contamination of dwindling water resources and to mitigate the worsening water scarcity crisis. [5]Ammonium pollution of water resources is caused by agricultural nitrogen fertilizers and wastewater discharges.8][9] Ammonium (NH 4 ) is the acid conjugate of the acid-base couple (NH 4 + /NH 3 ), the gaseous form ammonia (NH 3 ) being more toxic than its ionized counterpart, ammonium (NH 4

+
).In aqueous media, the two forms coexist in equilibrium, with one prevailing as a function of pH and water temperature.The known effects of these compounds on human health and the environment are significant; their transformation into nitrites and then nitrates by nitrifying bacteria constitutes a risk.Consumption of nitrate-contaminated water can lead to methaemoglobinaemia, particularly in infants under six months of age, [10] and is suspected of contributing to certain forms of cancer. [11]Ammonium thus contributes indirectly to the development of methaemoglobinaemia and certain cancers, includ-ing colorectal cancer, thyroid disease and neural tube defects.In aquatic environments, high levels of ammonium promote eutrophication, leading to the disappearance of various aquatic species, including fish. [12]he use of membrane processes, notably electrodialysis (ED), to treat water containing high levels of contaminants is widely recognized. [13]ED works by removing or concentrating ionic compounds from aqueous solutions through ion exchange membranes (IEMs) under the influence of an electric field.A ED cell typically comprises alternating anion exchange membranes (IEMs) and cation exchange membranes (CEMs) separated by spacers, with electrodes located at the end of the modules.The elementary cell consists of two compartments: the diluted compartment and the concentrated compartment.By supplying an electrolyte solution AX (A + , X À ), an electric field causes anions and cations to be depleted in the dilute compartment and enriched in the concentrate compartment.
Compared with other membrane methods such as reverse osmosis, ED is considered promising due to its extended service life, its ability to operate at elevated EMIs temperatures, and its reduced susceptibility to membrane fouling and scaling.It is considered economically viable and most efficient when feed stream salinity is below 5 g/L. [14,15]ED operation depends on the development of IEMs that facilitate high water recovery without the need for phase changes, chemical reactions or additives.These advantages benefit the environment by reducing dependence on fossil fuels and chemical detergents.[18] In particular, ammonium can be recovered from the concentrate compartment and recycled as a nutrient.For example, Xia et al [19] used ED to improve the efficiency of hydrogen production from wastewater by hydrogen-producing bacteria through ammonium removal.In addition, unlike other processes, ED enables the concentrate stream to be reused as fertilizer for plant fertigation, with the aim of achieving zero liquid discharge (ZLD) to promote the circular economy and sustainable development. [20]However, research on ED has primarily concentrated on the removal of contaminants from wastewater, with limited exploration of real groundwater treatment, partly due to the complex chemistry of water. [13]One of the key advantages of ED is its capability to remove a wide range of ionic compounds from water.Nevertheless, excessive demineralization of water can render it aggressive for human consumption.Therefore, achieving balanced water at the conclusion of the treatment process is crucial.
The Legrand-Poirier method is a well-established and widely utilized approach for assessing water quality.This method involves representing water quality using a curve known as the calco-carbonic equilibrium curve of water.The representation plane comprises two axes: the Ca 2 + concentration on the abscissa and the total mineral carbon (TMC) or total CO 2 concentration on the ordinate.The equilibrium curve divides the plane into two regions: aggressive water and scaling water.Through computer calculations of the calco-carbonic equilibrium conditions and graphical representations, the evolution of water under the influence of treatments can be visualized. [21]e aim of this study is to assess the feasibility of removing ammonium from groundwater using ED with AXE membranes as anion exchange membranes (AEMs) and CMX membranes as cation exchange membranes (CEMs).The treated water is intended for human consumption, with the aim of mitigating the loss of water resources compromised by high levels of pollution.As the natural concentration of ammonium in groundwater is generally lower than in wastewater, the initial ammonium concentration chosen for this study is based on previous assessments of the quality of ammonium-polluted groundwater and surface water, as indicated in the references.
The quality of the water obtained (dilute) by ED will be assessed using the Legrand-Poirier method, which evaluates water aggressiveness, scaling potential or equilibrium, as a function of the demineralization rate (DR).Statistical analysis, in particular Principal Component Analysis (PCA), is used to verify the results.PCA is a statistical technique used to explore and reduce the dimensionality of data, enabling a better understanding of relationships and patterns within the data set.

Description of the ED Pilot Plant
In this study, experiments were carried out using a ED pilot plant (TS-2-10) supplied by Eurodia Co (France), equipped with NEO-SEPTA AXE membranes as anion exchange membranes (AEMs) and NEOSEPTA CMX membranes as cation exchange membranes (CEMs) manufactured by Tokuyama Co (Japan).A total of 22 membranes were used, including 12 CMX and 10 AXE.The experimental set-up included three solution tanks, each with a capacity of 2 L, designed to hold dilute, concentrate and electrode rinse solutions.By applying a continuous electrical potential, ammonium ions migrate to the cathode, exit the dilute compartment, cross the cation exchange membrane and are blocked by the anion exchange membrane in the concentrate compartment.Schematic representations of the TS-2-10 pilot plant and the membrane stack exchange process are shown in Figures 1 and 2, respectively.
The characteristics of the TS-2-10 ED pilot plant used in this study correspond to those described in the work of Elmidaoui et al [22] Figure 1.Schematic of the TS-2-10 electrodialyzer with the three tanks containing the dilute, concentrate and electrode rinse solutions, the pumps and the membrane stack.
while the properties of the membranes used are detailed in the study by Addar et al. [23]

Analyses Methods
The ED operations are performed with real groundwater solutions from Kenitra city (Morocco), spiked with an initial concentration of 3 mg/L NH 4 Cl.The use of real water allows us to control the actual effect of ammonium removal in the presence of other cations present in the water, enabling us to obtain results closer to reality than is often difficult to achieve when working with synthetic solutions.The characteristics of the feed water are presented Table 1.
During the testing process, water samples were periodically extracted from the dilute stream, and ion concentrations were determined through analytical methods.A conductivity probe was immersed in the dilute water to continuously monitor conductivity throughout the process.At each demineralization rate reached (20 %, 40 %, 60 % and 80 %), water conductivity, time, and corresponding current intensity were recorded.
The alkali ions K + and Na + were measured spectrometrically using an industrial flame emission spectrometer model PFP7 JENWAY.NH 4 + and SO 4 2À ions were determined spectrometrically using a UV/VIS spectrometer model UV1600 at wavelengths of 655 nm and 650 nm, respectively.NO 3 À ions were measured potentiometrically using selective electrodes and an ionometer model Hach SensionTM + MM340.pH levels were measured using a Jenway 3510 pH meter electrode, and electric conductivity (EC) was measured using a WTW Inolab level 1 conductivity cell.Other parameters such as HCO 3 À , T-Alkalinity, P-Alkalinity, hardness, Ca 2 + , and Mg 2 + were determined using standard methods as previously described. [24]Other parameters such as time, demineralization rate (DR) and specific demineralization rate (SD) of each ion are also monitored.DR and SD are expressed by equations ( 1) and ( 2) respectively: Where C d is the solute concentration in dilute (mg/L) and C f the solute concentration in the feed water (mg/L) of the ions considered.
The statistics carried out in this study are multivariate principal component analysis (PCA) using the jmp version 13 software.

Feasibility of Eliminating Ammonium Ions by ED
To assess the feasibility of ammonium removal via electrodialysis (ED), the following operating conditions were employed: * The feed volume was standardized across all three compartments (dilute, concentrate, and electrodes).
* The dilute and concentrate compartments received identical raw solutions, while the electrode compartment was supplied with a solution consisting of water enriched with NaCl.This ensured that the electric conductivity of the electrode compartment remained higher than that of the dilute and  * The flow rate was determined based on a prior study by Elmidaoui et al [25] which optimized nitrate removal from groundwater using the same TS-2-10 pilot plant.The optimal flow rate identified in that study was 180 L/h.
* An applied voltage of 10 V was utilized.
These operating conditions were selected to evaluate the efficacy of ammonium removal from the initial solution via ED. Figure 3 show respectively the variation of NH + and other cations in the water, driven by their concentrations, higher valences (e. g., Ca 2 + and Mg 2 + ), ionic mobility, hydration energy, and hydration size. [17,26,27]Conversely, Yang et al [14] suggest that NH 4 + tetrahedral geometry confers it with high selectivity against spherical cations (e. g., Mg 2 + and Ca 2 + ).
* The similarity in polarity and hydraulic radius between NH 4 + and water molecules. [28]The chemical composition of the CMX membrane, featuring fixed sulfonic groups and exhibiting non-selective permeability to all ions, from highly charged to less charged species. [29]The inverse relationship between solution EC and ion removal duration, with lower EC solutions requiring more time for ion removal, resulting in a lower ammonium removal rate.
NH 4 + removal is observed to increase as competitive cations are depleted from the water over time.These findings suggest potential enhancements in NH 4 + removal through optimization of electrodialysis (ED) operating conditions.Future research should explore optimal ED parameters, such as applied voltage, current intensity, flow rate, and choice of ion exchange membrane pair, to further improve NH 4 + removal efficiency.Variations in EC and pH as a function of DR for dilute water are shown in Figure 4.
EC decreases linearly with the DR, which is explained by the fact that the EC of a solution is affected by the equivalent concentration and equivalent conductivity of the ions present in the solution. [17,30]As the ions are removed, the conductivity of the solution decreases and the DR increases.As for the pH, it increases slightly with the DR from an initial value of 7.53 to a final value of 7.88.Variations in pH can affect the effectiveness of ED, but the impact is less significant. [31]The pH determines the calco-carbonic balance of water, its variation depends on the ionic strength, alkalinity and CO 2 concentration, according to the equation (3). [32] Where ɛ is expressed as a function of the ionic strength μ of the solution e ¼ ffi ffi m p 1þ ffi ffi m p and pK 1 the carbonic acid acidity constant.To investigate the selectivity of AXE and CMX membranes, Figure 5 gives respectively the SD of anions for AXE (figure 5a) and the SD of cations for CMX (Figure 5b).
Figure 5a shows the variation of the specific demineralization (SD) of the anions (SO 4 2À , NO 3 À and HCO 3 À ) as a function of the demineralization rate (DR).The SD of each anion increases in proportion to the DR, reaching 84.16 % removal for SO 4 2À , 87.23 % for HCO 3 À , and 81.98 % for NO 3 À when the DR reaches 80 %.In particular, SO 4 2À is eliminated first and quickly when the DR is between 0 and 40 %, then stabilizes from 60 % DR.This behavior can be attributed to the AXE membrane used, which is a non-selective membrane allowing all ions to pass through, from the most charged such as SO 4 2À to the least charged. [24]In addition, SO 4 2À ions have a high hydration energy and are preferentially exchangeable in AEMs due to their higher valence. [33]This is in line with the results of Tahaikt et al [24] who studied sulfate removal with two anion exchange membranes AXE and ACS and concluded that sulfate removal was greater  with the AXE membrane.For NO 3 À and HCO 3 À , removal rates are initially similar up to 20 % DR, after which HCO 3 À removal decreases between 20 % and 50 % DR before increasing again between 60 % and 80 % DR.This trend can be attributed to the lower hydration energy and higher ionic mobility of NO 3compared to HCO 3 À . [23]Despite the higher concentration of HCO 3 À in the feed water, its removal only increases as NO 3 À concentrations decrease in dilute water.The transport order of ions is as follows: SO 4 2À > NO 3 À > HCO 3 À .On the other hand, ion transport through an IEMs is influenced by thermodynamic and kinetic factors that could explain why one ion is eliminated before another.These factors affected by the distinct physicochemical characteristics of the ions, including their ionic radius, charge density, hydrated radius and hydration free energy. [33]ubita et al [34] pointed out that ED membrane selectivity between anions can be improved by increasing the ion affinity difference, increasing the AEMs thickness and decreasing the AEMs charge density.
Figure 5b depicts the variation of specific demineralization (SD) of cations (NH 4 + , Ca 2 + , Mg 2 + , Na + and K + ) with demineralization rate (DR).The removal of divalent cations Ca 2 + and Mg 2 + increases proportionally with DR, reaching 82.84 % for Ca2 + and 82.64 % for Mg 2 + at a DR of 80 %.NH 4 + removal progressively increases with DR, reaching 84.60 % at a DR of 80 %.Na + removal commences at a DR of 40 %, following the removal of Ca 2 + and Mg 2 + , achieving only 42.45 % removal at a DR of 80 %.Due to its low concentration in water and small hydrated radius, K + is not removed.The order of ion transport is as follows: Ca 2 + > Mg 2 + > NH 4 + > Na + > K + .The CMX membrane is a non-selective membrane that allows all cations to pass through, from the most charged ions to the least charged ions. [35]Despite the higher initial concentration of Na + compared to NH 4 + in water, NH 4 + is removed prior to Na + .This phenomenon is attributed to the smaller hydration radius, lower hydration energy, and higher ionic mobility of NH 4 + across the CMX membrane compared to Na + (Table 2). [17,36,37]n Additionally, Aliaskari and Schäfer [40] point out that the quality of the feed water, including contaminant concentration, pH and salinity, significantly influences contaminant removal during the ED process.For example, several studies have shown that nitrate removal is particularly high in the presence of NaCl, which can be explained by the high diffusion and ionic mobility of nitrates, as well as their competitive removal with other anions such as chloride, bicarbonate and fluoride. [25,41,42]Furthermore, the removal of arsenic(V) by ED in brackish water is relatively low due to the ionic characteristics of the other anions present, particularly chloride. [13][45] However, research indicates that anions with higher mobility and diffusivity, such as chloride, nitrate and hydroxide, compete with fluoride and hinder its removal, thus prolonging the treatment time of the ED process. [40]igure 6 shows the variation of removal of hardness and T-Alkalinity as a function of DR.
Hardness and T-Alkalinity removal increases with the demineralization rate (DR), reaching 87.23 % and 82.93 % respectively at a DR of 80 %.Furthermore, the removal of hardness and T-Alkalinity has an impact on the calco-carbonic balance of the treated water, as this balance depends on the concentrations of Ca 2 + and HCO 3 À .These ions represent the

Ion
Valence Hydrated radii (nm) [38] Hydration energy (kJ/mol) [39] Mg water's hardness and T-Alkalinity respectively.Consequently, when Ca 2 + concentration decreases, hardness decreases, and when HCO 3 À concentration decreases, alkalinity decreases.It is therefore essential to assess the quality of the treated water obtained.

Statistics Analysis
Descriptive statistics play a crucial role in the synthesis and initial understanding of data, providing an essential basis for more advanced statistical analyzes and informed decisionmaking in various fields Table 3. PCA explores the distribution of variable concentrations and their correlations during the ED process, from input to output, marking the best performance.
Figure 8 shows the Biplot projection of variables and individuals along the C1 and C2 axes.
Over time, we observe very high raw concentrations at time t 1 = 0 min, and at t 2 , t 3 , t 4 up to t 5 = 60 minutes, the concen-  trations of the physico-chemical parameters indicated on the right of the graph progressively decrease, as shown by the performance of each treated element.From the right to the left of the graph, we observe a decreasing concentration gradient, which explains the good performance of the ED process used, increasing the DR with time (t 1 to t 5 ).Points 1 to 5 represent lines considered as individuals characterizing the concentration gradient.

Water Quality
After the characterization of the dilute solutions obtained by ED, the quality of the water is evaluated based on calcocarbonic equilibrium study according to the method of Legrand-Poirier for all the demineralization rate.The equilibrium curves are plotted in Figure 9 and the dilute obtained for all DR (20 %, 40 %, 60 % and 80 %) are located (red points) according to the calco-carbonic equilibrium curves.It appears that all the dilute solutions obtained by ED are located in the aggressive zone of the curve and the higher the DR, the more aggressive the water.At DR of 20 %, the point is located in the aggressive zone close to the equilibrium curve, which means that a small  amount of reagent will be sufficient to restore the water's equilibrium.Conversely, at DR of 80 %, the point moves away more and more from the equilibrium curve and consequently, the requested amount of reagent will be significant.Then, using a calculation program, the amount of calcium to be added for each DR is calculated for all samples to reach the equilibrium curve.Table 4 gives the amounts of Ca 2 + to be added as well as the water evaluation after re-balancing.

Conclusions
This study evaluates the feasibility of ammonium removal by electrodialysis (ED) using real groundwater solutions from the city of Kenitra in Morocco, initially spiked with 3 mg/L NH 4 Cl.The results demonstrate efficient ammonium removal, reaching a final concentration of 0.5 mg/L corresponding to a demineralization rate (DR) of 80 % after 60 minutes of ED treatment.Since the cation exchange membrane used lacks selectivity for ammonium, ED treatment also results in the removal of other cations present in the diluted water, albeit to varying degrees.Similarly, anions are removed depending on the anion exchange membrane used.ED-treated water complies with Moroccan standards for ammonium content.However, the Legrand-Poirier method, used to assess water quality after ED treatment compared with DR, indicates aggressive water characteristics.Consequently, the addition of calcium is deemed necessary to balance the water and make it fit for human consumption.These preliminary results are promising and suggest that ED could serve as an effective and environmentally-friendly technology for removing ammonium from drinking water.This approach could mitigate the health risks associated with the consumption of ammonium-contaminated water, while promoting sustainable management of water resources.Future efforts will focus on optimizing the ED operating parameters specifically for ammonium removal, and carrying out further assessments on groundwater naturally polluted by ammonium.With regard to the disposal of ammonium concentrate, and in line with zero liquid discharge (ZLD) targets, the preferred strategy is to explore its reuse potential for fertigation in agriculture.

Figure 2 .
Figure 2. Selectivity of AXE and CMX membranes.a) SD of anions as a function of demineralization rate.b) SD of cations as a function of demineralization rate.

Figure 3 .
Figure 3. Ammonium concentration and EC in dilute vs time.

Figure 4 .
Figure 4. Variation of pH and EC in dilute as a function of demineralization rate.

Figure 5 .
Figure 5. Selectivity of AXE and CMX membranes.a) SD of anions as a function of demineralization rate.b) SD of cations as a function of demineralization rate.

Figure 7 4 +,
shows the correlation circle characterized by a distribution of variables along the two main axes C1 and C2; on the C1 axis, a gradient of decreasing concentration is observed on the right-hand side of the graph, showing high concentrations of the physico-chemical parameters studied: SO 4 Na + , Mg 2 + , Ca 2 + , EC, T-Alkalinity and Hardness, which are positively correlated on the C1 axis with time (T 0 = 0 at high concentrations).During the ED process, their concentrations will decrease towards the left of this graph at the time (T 5 = 60 minutes) when the demineralization rate of DR reaches its maximum.

Figure 6 .
Figure 6.Removal of T-Alkalinity and hardness as a function of demineralization rate.

Table 1 .
Characteristics of the feed water.
4 + concentration and EC in dilute as function of time.Both parameters exhibit a progressive decrease over time, converging to NH 4 + concentration of 0.5 mg/L and EC of 137.2 μS/cm at the 60-minute mark.At 28 minutes, NH 4 * Ionic competition between NH 4

Table 2 .
Properties of ions in aqueous solution at 25 °C.

Table 4 .
Evaluation of water quality after Ca 2 + addition.