Introduction

Groundwater salinization is a major concern in arid and semi-arid regions. In fact, freshwater stored in coastal aquifers is particularly susceptible to degradation due to its proximity to seawater in addition to its vulnerability to numerous salt waste derived from industrial activities and higher population densities. It was shown that density contrast has a significant effect on the spatial distribution of contamination in coastal aquifers (Simmons et al. 2001). Understanding the dynamics of interaction between saltwater bodies and fresh groundwater could allow better strategies to control and well manage the stressed aquifers. The best management of salinized aquifers suggests a deep knowledge of this process. Traditional hydrodynamic and hydrochemical investigations are not sufficient to understand the saltwater dynamics. Thus, more efficient techniques are needed to identify and clearly delineate the freshwater–saltwater interactions, such as modeling procedure (Saidi et al. 2013).

Density-driven groundwater flow models are being increasingly used to solve scientific, engineering and water resources management issues such as saltwater intrusion, contaminated site remediation and fresh groundwater discharge into oceanic water bodies (Guo and Langevin 2002). Several simulations models of density-dependent groundwater flow were developed like SUTRA (Voss 1984), MOCDENS-3D (Sanford and Konikow 1985), FEMWATER (Lin et al. 1997), CODESA-3D (Lecca 2000) and SEAWAT (Guo and Langevin 2002). SEAWAT code has shown efficiency in simulating several related variable density groundwater flow problems like seawater intrusion into coastal aquifers (e.g. Qahman and Larabi 2006; Jin et al. 2009; Cobaner et al. 2012), saltwater intrusion from ephemeral rivers (Bauer et al. 2006), submarine groundwater discharge (Langevin 2003), brine transport (Mastrocicco et al. 2012) and groundwater flow near salt domes (Bakker et al. 2004).

In Wadi Al Ayn and Daroufa plain, which is located in CapBon peninsula in northeast of Tunisia (Fig. 1), a sudden increase in groundwater salinity was registered since 2002. Nevertheless, the origin and processes of groundwater salinization are still poorly understood. Salinization of groundwater in this zone is a spatially and temporally variable phenomenon which is depending from different salt sources, especially the infiltrated oilfield brine near Wadi Al Ayn and the intruded seawater in the coastal part near Wadi Daroufa. It constitutes one of the rare cases of multi-salt contamination sources and it opens perspective to observe the system reaction to such complex pollution (Chekirbane et al. 2013). The study area was subject of a multi-disciplinary study including hydrodynamic, hydrochemical and geophysical investigations which constitute important inputs for the numerical simulation. The objective of the present study is to simulate the historical trends of saltwater distribution in the groundwater flow system of Wadi Al Ayn and Daroufa plain and to predict its future dynamics under some proposed countermeasures for remediation.

Fig. 1
figure 1

The location map of the study area showing the sample locations and the regional geology (extracted and modified from the geological map of Nabeul Prefecture)

Full size image

Study area

The study area is the plain of Wadi Al Ayn and Daroufa, a coastal floodplain located in the southeastern part of CapBon peninsula, northeastern of Tunisia (Fig. 1) which is known as one of the most touristic, industrial and agricultural productive zones in Tunisia with a contribution of 15 % (CCI CapBon 2011). Since 2000s, farmers were claiming a decreasing in their production because an increased groundwater salinity. In fact, some water wells were abandoned and consequently farming activities of their owners were stopped leading to an additional jobless and financial problem in the region. The study area is situated in the vicinity of the Zinnia oilfield concession which has been exploited since 1994. Oil production is accompanied by water with high salinity called oilfield brine. In fact, petroleum-bearing formations usually contain brine (altered seawater trapped in sediment pores). Large volumes of brine usually accompany oil and gas production. High chloride concentrations commonly characterize oilfield pollution of groundwater. Brine can be toxic to crops, corrosive to metal, and unsafe to drink (Hudak and Wachal 2001). Since the beginning of exploitation of Zinnia oilfield in 1994 until its end in late 2009, the separated brine from oil are directly rejected via a pipeline in the sandy bed of Wadi Al Ayn without any treatment (Chkirbene 2008). Therefore, the groundwater flow system of Wadi Al Ayn and Daroufa plain might be under a stress caused by overexploitation as well as contamination from multiple salinization sources.

Hydrogeologic characterization of the groundwater flow system of Wadi Al Ayn and Daroufa

Geological settings

The outcropping formations in the oriental coastal plain of CapBon are mainly Mio-Plio-Quaternary sediments (Ben Salem 1992). The simplified lithostratigraphic column of Wadi Al Ayn plain is mainly composed by Quaternary, Upper Miocene and Middle Miocene.

Figure 1 shows the geological map of the study area. The upstream region of Al Ayn plain, western Somâa City, is composed by the Upper Miocene which corresponds to Tortonien geological age known in Tunisia as Somâa formation, a detrital continental sediments mainly composed by an alternation of sand and conglomerates with frequent clay lenses resulting from the erosion of the top of the anticline of Abderrahmen Mountain (Colleuil 1976).

The lower part of Middle Miocene is composed of lenticular sandstones and marls with lignite levels belonging to the geological age of Serravalian–Tortonian, called in Tunisia as Saouaf formation (Kouzana et al. 2010). In contrast to the neighboring northern area, Pliocene outcrops do not exist in the study area and that was believed because of the widespread of erosion during the Miocene orogeny (Ennabli 1980). Sedimentary discontinuity is characterizing the deposit of Quaternary sediments above Upper Miocene formation (Ben Salem 1992). The coastal consolidated sandy dunes from Tyrrhenian are outcropping in northern Tazarka city with a width of 1–2 km, an altitude which can reach 30 m and a thickness varying between 10 and 50 m (Zghibi et al. 2011). However, these dunes are disappearing near Daroufa region where they are buried beneath the outcropping Wadi alluvia and found in a depth around 100 m as it is shown in an established longitudinal cross section A–A′ along Wadi Al Ayn and Daroufa (Fig. 2) based on a performed geophysical survey in April 2011 (Chekirbane 2013) and some available boring data.

Fig. 2
figure 2

Geological cross section A–A′

Full size image

Quaternary outcrops consist mainly in encrusted limestone largely extended and rich in calcite, silica, sometimes in gypsum and alumina and frequently colored by iron salts (Kouzana et al. 2010). They are also represented by the Holocene deposits which are formed by recent alluvia observed near Wadi Al Ayn, Wadi Daroufa, Sebkha Al Gharbya and current dunes and beaches.

Hydrogeological settings

The regional hydrogeology of the coastal oriental aquifer of CapBon was studied by numerous previous works. Several studies focused mainly in understanding the flow and hydrological budget of the regional groundwater flow system (Rouatbi 1968; Ennabli 1980; Rekaya 1989; Jemai 1996; Khlaifi 1998; Paniconi et al. 2001; Kerrou 2008; Zghibi et al. 2011; Trabelsi et al. 2013). Some others have dealt with characterizing the qualitative degradation of the aquifer caused by seawater intrusion (Kouzana et al. 2009; Slama 2010; Ben Hammouda et al. 2011). All the mentioned authors focused exclusively on regional groundwater flow; there was no previous works which studied small scale groundwater flow and contamination. The hydrogeological basin of Wadi Al Ayn covers an area of 20 km2 located in the southwest of the large coastal plain of CapBon which has a total area around 400 km2.

In Al Ayn plain, the Quaternary sediments and the top layers of the Upper Miocene are hosting a significant shallow aquifer used principally for irrigation and drinking purposes. The first deposits are mainly composed of Wadi alluvia and the latter consist of Somâa formation with sand and conglomerate. The frequent presence of clay lenses interbedded in the sandy layers of Somâa formation (Upper Miocene) turns the aquifer in some locations into a multi-layered system.

Piezometric behavior

The groundwater level map of Wadi Al Ayn plain was established based on 34 observation wells in April 2011 (Fig. 3).

Fig. 3
figure 3

Groundwater contour map and the spatial distribution of the electrical conductivity in the study area (April 2011). (1) Underflow type following under the Wadi bed, (2) recharged area from the Wadi water, (3) negative groundwater head = seawater intrusion

Full size image

Groundwater in the study area is characterized by a shallow flowing type mostly controlled by topographical changes as it is shown for instance in contour 15 and 40 m following, respectively, changes in topographical contours 30 and 60 m.

In the upstream region near Somâa City, the groundwater table depth does not exceed 10 m of depth in the alluvial deposits of Wadi Al Ayn where it constitutes an underflow aquifer.

The main flow direction is south-east following the shape of the Wadi. In the middle part of the plain, near the location of AYR2 and AYR3, the major flow directions are respectively south-east and north-east with a hydraulic gradient of nearly 5 %. In these locations, the aquifer is hydraulically connected to the hydrological network; the Wadi water, mainly composed of non treated wastewater and oilfield brine may be infiltrated and spread according to the main flow direction which can explain the high salinity measured in some wells in the Wadi vicinity (Fig. 3). The downstream region, near Wadi Daroufa, is characterized by lower hydraulic gradient because of the flat topography. However, an inversion of the groundwater flow gradient is registered and the contour zero is appearing at 2 km inland. Closer to the shore line, groundwater contour becomes negative; a probable seawater intrusion is occurring in this area which can explain the high salinity of some neighboring wells. This phenomenon is frequent in the eastern coastal aquifer of CapBon since the 1990s and was reported by several authors. In fact, in Korba area (located 10 km in the northeastern of Wadi Al Ayn watershed), seawater intrusion is occurring (Paniconi et al. 2004; Kouzana et al. 2010; Zghibi et al. 2011).

Methodology

Numerical modeling of density-dependent groundwater flow: governing equations

Variable density groundwater flow and transport have recently been one of the important topics in hydrological sciences. Consequently, a large number of numerical models of salt transport have been developed in order to cope with the diversified environmental problems such as seawater intrusion in exploited coastal aquifers, leakage from landfills and disposal of industrial wastes in alluvial aquifers.

SEAWAT in conjunction with VISUAL MODFLOW graphical user interface was chosen as the basis for numerical simulation because of its ability to solve density-driven groundwater flow in porous media. It is a coupled version of MODFLOW (Harbaugh et al. 2000) and MT3DMS (Zheng and Wang 1999; Zheng 2006) designed to simulate three-dimensional, variable-density groundwater flow and multi-species transport (Langevin et al. 2008). Variable density flow (VDF) process uses MODFLOW methodology to solve the variable density groundwater flow equation and the MT3DMS for solving the transport equation.

The VDF process solves the following form of the variable-density groundwater flow equation Eq. (1):

$$\nabla \cdot \left[ {\rho K_{\text{o}} \left( {\nabla h_{\text{o}} + \frac{{\rho - \rho_{\text{o}} }}{{\rho_{\text{o}} }}\nabla z} \right)} \right] = \rho S_{\text{s,o}} \frac{{\partial h_{\text{o}} }}{\partial t} + \vartheta \frac{\partial \rho }{\partial C}\frac{\partial C}{\partial t} - \rho_{\text{s}} q_{\text{s}}^{ '}$$
(1)

where ρ, ρ ο, fluid densities [M L−3] in flow field and at the reference (freshwater); h o hydraulic head [L] vs. reference; K o freshwater hydraulic conductivity [L T−1]; S s,o specific storage [L−1] vs. equivalent freshwater head; C salt concentration [M L−3]; \(\vartheta\) porosity [unitless]; t time [T]; x i is the orthogonal coordinate [L]; \(q_{s}^{{^{\prime } }}\): source or sink [T−1] of fluid of density ρ s [M L−3].

In the case of VDF, head (h) is not only varying according to pressure and elevation changes but it depends also from the density (ρ). Thus, the concept of equivalent fresh water head (h f) is introduced to take in account the difference in density while calculating the heads (2):

$$h_{\text{f}} = \frac{P}{{\rho_{\text{r}} g}} + Z$$
(2)

where ρ r is the reference density, usually taken as the reference of fresh groundwater at reference chloride concentration [M L−3], and P is the pressure [M L−1 T−2].

The flow equation is solved jointly with solute transport Eq. (3):

$$\frac{{\partial \left( {\vartheta C} \right)}}{\partial t} = \nabla \cdot \left( {\vartheta D \cdot \nabla C} \right) - \nabla \cdot \left( {qC} \right) - q_{\text{s}}^{{^{\prime } }} C_{\text{s}}$$
(3)

where C s salt concentration [M L−3] of source/sink flux; D hydrodynamic dispersion coefficient tensor [L2 T−1]; q specific discharge [L T−1]. For VDF, fluid density is assumed to depend only from solute concentration since the pressure and temperatures effects are ignored (Langevin et al. 2003). Thus, an equation of state is used by SEAWAT to convert solute concentration to fluid density:

$$\rho = \rho_{\text{o}} + \frac{\partial \rho }{\partial C}C$$
(4)

where ∂ρ/∂C is the linear slope of density over concentration.

Model discretization, boundary conditions and model parameters

Domain discretization

A finite-difference mesh in SEAWAT was used to discretize the groundwater flow system of Wadi Al Ayn and Daroufa plain (Fig. 4). The model domain is composed of 102 rows, 143 columns, and 5 layers. The grid spacing is 60 by 30 m while the layer thicknesses are variable with maximum depth reaching 86 m below mean seal level according to the geophysical survey that we performed in the area.

Fig. 4
figure 4

Finite-difference grid and boundary conditions

Full size image

Boundary conditions

Constant head and a concentration of 35 g/L were specified to the model cells along the coast. The effect of tidal fluctuation was not taken into consideration because it is assumed to be limited to a short distance from the shore line (Narayan et al. 2007; Jin et al. 2009; Cobaner et al. 2012). A constant concentration linearly ranging from 67 to 4.5 g/L was also imposed for the cells of Wadi Al Ayn during the period 1994–2009 (oilfield brine infiltration). The lateral boundaries are assumed to be no-flow boundaries because the flow lines are parallel to them as it is previously shown in the groundwater contour map (Fig. 3).

The use of general head boundary (GHB) in some location is dictated by the influence of recharge from some adjacent and neighboring formations and they are consequently interpolated from measured heads along the boundaries.

Model parameters

Zoned hydraulic conductivity maps are created using geological information from field description, geophysical surveys and previous studies (Chekirbane 2013) (Fig. 5). Values are averaged for each zone and the mean horizontal hydraulic conductivity (Kx = Ky) is assigned to every zone. The vertical hydraulic conductivity (Kz) is assumed to be the tenth of the horizontal one. The calibrated horizontal hydraulic conductivity is estimated to 5 × 10−5 m/s for current dunes and beaches formations located near the coastal part, 7 × 10−5 m/s for the recent Wadi alluvia, 3.4 × 10−6 m/s for limestone crusts, 8 × 10−6 m/s for Marine Quaternary formation and 10−5 m/s for Somâa formation. The other model parameters including specific storage, porosity and dispersivity are presented in Table 1.

Fig. 5
figure 5

Hydraulic conductivity cross sections: a top layer, b longitudinal cross section, c upstream cross section (Wadi Al Ayn) and d downstream cross section (Wadi Daroufa)

Full size image
Table 1 Aquifer parameter values after calibration
Full size table

The considered recharge in the model consists of the infiltrated rainfall and the percolated water from Wadi Al Ayn. Ennabli (1980), Paniconi et al. (2001) and Kerrou (2008) considered the recharge of the eastern coastal aquifer of CapBon by infiltration of precipitation to be less than 10 % of the average annual rainfall. The percolated water from Wadi was estimated based on the measured inflow and outflow in 2011. In fact, the total calculated inflow is 29.788 L/s and the infiltrated quantity calculated by water budget method is 9.01 L/s, which mean that the Wadi contribution in aquifer recharge is 30.24 %. Nazoumou (2002) estimated the infiltration from Wadi in semi-arid area in central Tunisia to 30 % of the annual runoff. According to the mentioned study and the results of the present study, the total percolated water from Wadi Al Ayn for 2011 is 0.308 Mm3.

The existing groundwater abstraction data is globally covering the entire eastern aquifer of CapBon. Specific data for the alluvial aquifer of Wadi Al Ayn and Daroufa does not exist. Thus, for the extended period from 1980 to 2003, the pumping rate of 52 wells was proportionally extracted from the total amount of groundwater abstraction. Nevertheless, the pumping rate for 2010 was calculated based on field interviews performed in the study area (Table 2) and the monitored groundwater level in five monitoring exploited wells. The estimated 2010 value is more precise than the other year values since the data of every single well is treated separately than the total amount is obtained by simple algebraic sum. Owing to the high density of dug wells in the study area, it was difficult to collect a complete pumping rate data from all the exploited wells. In the coastal part of the aquifer domain (Daroufa area), 11 pumping wells are identified and added to the model, where we assumed their pumping rate to be the same as the monitored well AW18 which is located in the same area.

Table 2 Estimated groundwater abstraction from the groundwater flow system of Wadi Al Ayn and Daroufa
Full size table

Initial conditions and time discretization

A steady state simulation was performed based on the earliest stable hydrological state of the eastern coastal aquifer of CapBon. The results of groundwater level as well as salt distribution from the steady state simulation were considered as initial conditions for transient simulation.

The equilibrium period was determined after running the performed SEAWAT model for 100 years and the idea is to determine the required time for the seawater fringe to reach the nearest TDS monitoring well to the coast line (AW20). The predicted TDS in the monitoring well AW20 reached 5,805 mg/L (observed TDS in November 2011) after 62 years (Fig. 6). It means that the equilibrium between seawater and fresh groundwater was reached 62 years before 2011, which is the year 1949. Therefore, when concerning the time discretization in the developed VDF model of the groundwater flow system of Wadi Al Ayn and Daroufa, the starting date was chosen as 1949. The chosen date seems to be reasonable since the early study of the entire aquifer of the eastern coastal aquifer of CapBon does not detect any seawater intrusion based on the piezometric map of 1962 (Ennabli 1980).

Fig. 6
figure 6

Time discretization of the numerical model

Full size image

Accurate solution of the variable-density flow problem requires a specific time stepping. The temporal discretization scheme in SEAWAT model is a combination of MODFLOW and MT3DMS time discretizations. Time steps are derived from stress periods and they are further divided into transport steps. SEAWAT solves flow and transport equations (1) and (3) for each time step using explicit or implicit approaches described in Langevin et al. (2003). When concerning the present study, the flow and transport equations are solved implicitly using generalized conjugate gradient (GCG) solver. The length of time step is depending from Courant number (Cr) which should be less than or equal to one. It denotes the ratio of the advective distance during one time step to the spatial discretization (Eq. 5):

$${\text{Cr}} = \frac{v\Delta t}{\Delta l}$$
(5)

where v [L T−1] is the velocity, ∆t [T] is the time step and ∆l [L] is the grid size in the direction of v. The higher the Cr, the longer is the time step and hence a large number of iteration is needed for the solution convergence (Al-Maktoumi et al. 2007).

For the present model, the Courant number was specified as 0.75 in approximation with the heterogeneous and anisotropic aquifer materials, allowing three quarters of the flow and transport over a distance equal to one grid spacing.

Model calibration

Model calibration was performed using trial-and-error approach. The gap between observed and simulated heads and TDS was minimized by adjusting the hydraulic conductivities as well as dispersivities of the lithostratigraphic units. During the calibration, a total of 36 observed hydraulic head values measured in 2011 are used (Fig. 7) in addition to head time series of five wells continuously monitored during the period 2011–2012. An overall correlation coefficient of 0.98 and a root mean square error of 7.79 values are obtained after finishing the model calibration, indicating a reasonably good match between the observed and simulated heads.

Fig. 7
figure 7

Observed versus simulated heads

Full size image

When concerning the TDS calibration, only the salt concentration measured in wells located in the vicinity of Wadi Al Ayn or near Daroufa plain were mainly considered (13 wells) (Fig. 8). The correlation coefficient and root mean square error are respectively 0.47 and 3,703. Furthermore, salt concentration and the spatial extension of the saltwater distribution were also compared with the iso-2D resistivity pseudosections and they show a satisfying agreement; the lower the resistivity of the aquifer horizon, the higher is the salt concentration of its water (Fig. 9a).

Fig. 8
figure 8

Observed versus simulated TDS

Full size image
Fig. 9
figure 9

a Comparison between the simulated hydraulic head and salt concentration of April 2011 and the observed salt concentration converted from TDEM data of April 2011. b 3D distribution of the simulated salt concentration of April 2011

Full size image

Simulation results

Retracement of the extent of saltwater plumes

The developed variable density groundwater flow and salt transport model was temporally discretized into two periods: the first one started from 1949 until 1994 where the system is only subject of a unique stress; the seawater intrusion in Daroufa area, near the coast and a second period extended from early 1994 until late 2009 when a second stress, in addition to seawater intrusion, was added: the infiltrated oilfield brine into the alluvial aquifer beneath Wadi Al Ayn (Fig. 6).

The obtained results are satisfyingly reproducing the salt distribution situation observed in April 2011 (Fig. 9a). The model shows that the longitudinal extent of the oilfield brine plume (TDS >10 g/L) is around 2 km and the seawater intruded the aquifer for an inland distance near 1 km. The 3D distribution of salt in the aquifer gives an idea about the transversal extent of the saline water plume (Fig. 9b) which shows that the oilfield brine plume has its largest transversal extent in the middle part of Wadi Al Ayn plain and the maximum of seawater intrusion is registered beneath the middle and downstream part of Wadi Daroufa. That could be explained by the sufficiently porous media constituting the wadis beds and their neighboring alluvial sediments.

An acceptable agreement was found between salt concentration simulated by SEAWAT model and a pseudo-2D-section of resistivity determined based on TDEM data performed in April 2011 (Chekirbane et al. 2012; Chekirbane 2013). Therefore, the numerical model can be used for predictive simulations.

Future exploitation scenarios of the groundwater flow system of Wadi Al Ayn and Daroufa

Once calibrated, the developed numerical model can serve to predict the future behavior of the groundwater flow and salt concentration under different scenarios and constraints.

Scenario 1

Under this scenario, we suppose that no countermeasures are taken to control the depression cone of groundwater and the saltwater plumes. The recharge and groundwater abstraction are considered the same as the values of the year 2010.

The results of scenario 1 (Fig. 10) show that the oilfield brine plume found in upstream of the aquifer needs at least 5–10 years to be naturally reduced to less than the half of its actual size, while in the coastal part of the aquifer the seawater–freshwater interface can reach an inland extent of 1.3 km with a TDS more than 10 g/L. Therefore, the aquifer situation will be aggravated and unsustainable. Thus, another scenario including some countermeasures is needed to remediate the actual situation of the groundwater flow system of Wadi Al Ayn and Daroufa.

Fig. 10
figure 10

Prediction of saltwater without any remediation countermeasure (scenario 1)

Full size image

Scenario 2

This scenario is based in some remediation countermeasures which are the application of artificial recharge of the aquifer with treated wastewater and optimized pumping rates of the wells exploiting the shallow aquifer according to crop water needs.

The artificial recharge will be specially applied to remediate the coastal part of the aquifer near Daroufa area (Fig. 11).

Fig. 11
figure 11

Proposed location of the artificial recharge basins and TDS observation well

Full size image

Four percolation basins of 30 m × 50 m dimension are proposed to infiltrate a daily amount of 0.25 m of treated wastewater per basin during the period extended from 2014 to 2044. The percolated water can be provided from a neighboring wastewater treatment plant (Maamoura plant) which is programmed to be installed near Wadi Daroufa with an expected treatment capacity of 8,435 m3/day (CRDA of Nabeul 2004). The measured TDS of the existing wastewater in Wadi Al Ayn is around 1.5 g/L. Thus, the predictive simulation of the seawater intrusion remediation was performed based on the above-mentioned parameters.

In parallel to the proposed project of artificial recharge, the pumping rates from the exploitation wells was optimized according the natural water needs of the available crops in the study area using FAO method (Allen et al. 1998) (6, 7):

$${\text{ET}}_{\text{crop}} = {\text{ET}}_{ 0} \times {\text{Kc}}$$
(6)
$${\text{ET}}_{0} = K_{\text{pan}} \times E_{\text{pan}}$$
(7)

where ETcrop is the crop evaporation or crop water needs (mm/day), Kc the crop factor, ET0 the reference evapotranspiration (mm/day), K pan the pan coefficient = 0.45–1.1 for Sunken Colorado pan, E pan the pan evaporation (mm/day).

When concerning tomato, potato and citrus plantation, the optimized and applied pumping rates are close, contrary to the strawberry case where the farmers are irrigating it with more than three times of its optimal water requirement (Fig. 12). That could be explained by the fact that strawberry is a relatively new crop in the study area and the farmers are lacking experience on its practice comparing with the ordinary and traditional plantations (tomato, potato, and citrus).

Fig. 12
figure 12

Comparison between the applied and optimized pumping rates

Full size image

Optimal artificial recharge conditions

The efficiency of the artificial recharge is usually function of the quantity and quality of the water used for recharge. A sensitivity analysis of the 3D VDF model for the amount of percolated water and its TDS was performed (Fig. 13). From the viewpoint of quantity, the model was tested for two recharge amounts: 1 m per day for four basins and 0.5 m per day for two basins which is respectively equivalent to a capacity of 6,000 and 3,000 m3/day. The mentioned amount is limited because of the capacity of the recharge water source which may be the wastewater treatment plant of Maamoura with a capacity of 8,435 m3/day.

Fig. 13
figure 13

Sensitivity analysis of the model for recharge quantity and quality

Full size image

When concerning the recharge water quality, TDS of 1,500, 2,500, 3,500, and 5,000 mg/L was tested for the two proposed amounts. A fictive well located 100 m far from the coast line near Wadi Daroufa (Fig. 11) is used to observe the variation of predicted TDS versus time.

The model shows sensitivity for both amount and quality of recharge water. However, the sensitivity for the amount is much more important than quality. Necessary time to remediate the aquifer from salinization, especially the seawater intrusion, is shorter for high recharge water amount with the lowest TDS. Therefore, the most efficient scenario of artificial recharge can be based on percolating 0.25 m/day/basin of water having a TDS of 1,500 mg/L.

Proposal of an adaptive management plan

The artificial recharge with treated wastewater seems to be an efficient countermeasure solution to stop the seawater intrusion (Fig. 14). Its immediate effect starts to be observed just after 3 years of application. A period between 5 and 10 years is recommended to remediate the seawater intrusion in the coastal part of the aquifer and to have an acceptable recovery of the seawater–fresh groundwater equilibrium.

Fig. 14
figure 14

Prediction of saltwater evolution under the artificial recharge scenario (scenario 2)

Full size image

When concerning the plume of oilfield brine, a period of 10 years is sufficient to reduce its longitudinal extent to less than 50 % of its actual dimension. The simulation results are predicting that a plume core of 10 g/L with an extent around 250 m will remain in a depth ranging from 30 to 40 m under AW3 well after 30 years of exploitation. Enhance of the natural remediation of saltwater plume derived from the oilfield brine might be possible by installing a prohibited pumping perimeter near AW3 well and by controlling the discharges into the wadis.

Conclusion

The hydrodynamic, hydrochemical, and geophysical data served to build a conceptual model and they constituted the input of a numerical model constructed with VISUAL MODFLOW and SEAWAT code to retrace the salt dynamics and predict its behavior under remediation scenario. The effect of oilfield brine and seawater intrusion in groundwater salinization were successfully reproduced and confirmed by the 3D numerical model. The oilfield brine plume needs at least 5 years to be naturally reduced to less than the half of its actual size, while the seawater–fresh groundwater interface can reach an inland extent of 1.3 km with a TDS more than 10 g/L if no countermeasures will be taken until the next three decades.

The artificial recharge with treated wastewater seems to be an efficient countermeasure solution to reduce the effect of seawater intrusion especially in the coastal part of the aquifer. Its immediate effect starts to be observed just after 3 years of application. A period between 5 and 10 years of artificial recharge is recommended in order to remediate the seawater intrusion in the coastal part of the aquifer and to have an acceptable recovery of the seawater–fresh groundwater equilibrium. When concerning the plume of oilfield brine, a period of 10 years is sufficient to reduce its longitudinal extent to less than 50 % of its actual dimension. Enhancing its natural remediation might be possible by installing a prohibited pumping perimeter near AW3 well and by protecting the Wadi’s water from any kind of contaminant which may reach the saturated zone of the alluvial aquifer.