1. Introduction
The extensive rural and urban development in coastal areas strongly impacts the coastal environment. This partly occurs via subterranean interaction between land and the sea, which involves a bi-directional flow, including seawater intrusion [
1,
2] and Submarine Groundwater Discharge (SGD, [
3,
4,
5]). The latter may cause deterioration in water quality of the shallow sea, leading to algae blooms [
6,
7,
8,
9]. In some coastal areas, there are surface reservoirs (e.g., fish ponds) above the aquifer which can make the hydrogeological conditions more complex. Worldwide aquaculture has been annually increasing by 8.7% over the past 40 years, the fastest-growing food-producing sector [
10]. One of the key environmental concerns about aquaculture is water degradation, due to the discharge of effluents with high levels of nutrients and suspended solids into adjacent waters, causing eutrophication, oxygen depletion and siltation [
11]. In the study area, it was shown that the phreatic unit beneath the fishponds area has indeed been deteriorating. It is rich in phosphate, ammonium, Total Organic Carbon (TOC) and manganese, and characterized by strong reducing conditions [
12].
Seawater intrusion (SI) is also a global concern, exacerbated by increasing demands for freshwater in coastal zones and predisposed to the influences of rising sea levels and changing climates. Open questions still exist, in particular about the transient SI processes and timeframes, and the characterization and prediction of freshwater–saltwater interfaces over regional scales and in highly heterogeneous and dynamic settings [
13].
Numerical solutions have become standard tools for analyzing aquatic systems, which involve a bi-directional flow [
14]. Two different modeling approaches have been used in the literature to represent the freshwater–saltwater relationship, one of which is sharp interface approximation [
15,
16,
17], while the other includes a transition zone between freshwater and seawater [
18,
19,
20,
21]. Numerical solution with the FEFLOW cod is a common tool to study the salt water interface movement [
22]. Yet, extensive measurement campaigns are necessary to accurately delineate interfaces and their displacement in response to real-world coastal aquifer stresses, encompassing a range of geological and hydrological settings [
13]. Analytical studies of tide-induced groundwater fluctuations in confined coastal aquifers that extend under the sea have been studied in References [
23,
24,
25] and others. Other studies [
24,
26] suggested a general analytical solution that showed that the tidal fluctuation in an inland borehole decreases with the confined layer roof length under the sea until a certain distance offshore, wherefrom it remains constant due to the loading effect.
The direct discharge of groundwater to the sea was quantitatively studied by several authors [
27,
28,
29]. A common way of assessing the discharge to the sea is the use of numerical simulations and analytical solutions of the water flow regime [
30,
31]. This is aided by geophysical [
32] and the use of chemical and isotopic tracers, including radon and radium isotopes [
5,
6,
33]. Seepage meters, which are deployed on the seabed, provide direct local measurements of fluxes [
34].
The objective of this study was to investigate the effect of the marine and terrestrial boundary conditions on the hydraulic connections between the sea and a multi-layered coastal aquifer and its resultant seawater intrusion, using numerical simulations and field observations.
2. Hydrogeological Background of the Study Area
The study area extends from Mt. Carmel to the sea at the southern Carmel coastal plain, Israel (
Figure 1). It is covered by Quaternary Nilotic sands with a total thickness of about 40 m, which are underlain by Late Cretaceous carbonate basement [
35]. Located on its eastern part are the Taninim springs at +3 m asl, which are the northern discharge of the Late Cretaceous Yarkon–Taninim carbonate Aquifer (
Figure 1). These springs are the principal source of water for fishponds in this area. The spring water is brackish (1500–2000 mg L
−1 Cl), which is due to the mixing of fresh groundwater with saline water that exists in the western part of the Late Cretaceous aquifer [
36]. The Late Cretaceous carbonates change from limestone and dolomite facies in the Mt. Carmel area on its eastern side to a chalky facies on its western (seawards) side [
35,
37], which prevents the direct flow of groundwater from the Late Cretaceous aquifer to the sea and forces the water to discharge in the Taninim springs through Quaternary sediments (
Figure 2). Some of this water flows westward through the Pleistocene sediments and plays an important role in the water balance of the coastal Quaternary Aquifer [
12,
38,
39,
40].
The thickness of the coastal Quaternary aquifer in the study area is 40 m (
Figure 3a). It includes three sub-aquifers (
Figure 3a). The shallow sub-aquifer is a phreatic Holocene sandy unit (unit A) with a thickness of several meters, which is separated from the underlying Pleistocene calcareous sandstone (locally called Kurkar) aquifer by a clay unit. The Pleistocene Kurkar is subdivided by another clay layer at an approximate depth of 20 m into two confined units, B and C, which in turn are underlain by the Late Cretaceous chalks. The coastal area is characterized by two N–S oriented Kurkar ridges on its western side, with the western one often partly or completely submerged [
41,
42]. In the study area, fish ponds occupy the area between the eastern ridge and the sea (
Figure 1 and
Figure 3a), with an annual infiltration to the aquifer of about 9 × 10
6 m
3year
−1 [
40].
The confined units B and C, which contain brackish water (1500–2000 mg L
−1 Cl), have been extensively exploited since 2005 as raw water for a desalination plant and for the fish ponds. The pumping in this area currently reaches ~20 × 10
6 m
3year
−1. Although exploitation is very high, groundwater level in these units is quite stable, and seawater intrusion is very limited [
43]. Geochemical mass balance indicates that more than 85% of the water in units B and C derives from the underlying Late Cretaceous aquifer [
12], while the ponds contribution is no more than 3 × 10
6 m
3year
−1 [
12]. This suggests that most of the ponds’ losses flow through unit A to the sea.
3. Field Methods
The relations between the aquifer and the sea were studied in two coastal sites, 1.5 km apart (
Figure 1). Groundwater level was mainly studied in boreholes drilled at 45–70 m from the sea to a depth of 10–40 m (wells 71-1, 71-2, 71-3, 72, 73,
Figure 1), each opened to a different unit, A, B or C. We also used data from a few observation boreholes located farther inland (1500–600 m from the sea).
TDEM (Time Domain Electro Magnetic) measurements were conducted to locate the fresh–saline water interface. Surveys in coastal aquifers [
44,
45,
46,
47] show that this method is feasible for detecting the interface due to the very high resistivity contrast between the sea water saturated portion of the aquifer (1–2 ohm-m) and the fresh water part, which is characterized by values more than 10 ohm-m [
45]. The survey was conducted using the Geonics EM-67 TDEM system (GEONICS Ltd., Mississauga, ON, Canada). The TDEM method utilizes a step-wise current flowing in a squared transmitter loop that produces a transient secondary electromagnetic field in the subsurface. Depending on the required exploration depth, the loop size varies between tens by tens meters to hundreds by hundreds meters. In the described survey, the loop size was 25 × 25 m. The electromagnetic field induced into the ground produces a secondary response, which is picked up on the surface by a receiver coil. The data were analyzed using two different inversion methods, smooth inversion, and sharp boundary inversion. The combination of solutions usually improves both the accuracy and reliability of decryption. The TDEM survey was conducted along two E–W traverses, one in the southern site and the second in the northern site (
Figure 1). In each profile, three loop measurements were carried out between the shore and approximately 80 m inland. The TDEM survey and the interpretation of the data were conducted by the Geophysical Institute of Israel.
CHIRP sub bottom profiler was used for mapping of the shallow sub-seafloor sediments. The seismic survey was conducted with the “Adva” boat of the IOLR (Israeli Oceanography and Limnology Research), which also carried a Trimble AgGPS 132–receiver, an Odom Echotrack MKIII depth meter, and Oceanographer navigator. The depth meter was calibrated by Bar check and sound wave velocity meter in water, SV-2000 Smart sensor, which belongs to AML company (Hong Kong, China). The velocity of sound waves in the submarine sand and clay was taken as 1700 m/s, while for the sea water, it is 1530 m/s [
48]. The CHIRP survey and the interpretation were conducted by Arik Golan from IOLR.
Electric conductivity (EC) profiles were conducted in the monitoring wells 71-1, 71-2, and 71-3 (locations in
Figure 1) using a Robertson Geologging (RG) profiler.
Water level was measured by Schlumberger divers with 5 min frequency in wells 1, 2, 71-1-3, and 73. Sea Level was measured at Hadera MedGloss station, 2.5 km from the shore by IOLR (Israeli Oceanography and Limnology Research).
Slug tests were conducted for the determination of the hydrological conductivity in the aquifer. In each well, we conducted five tests, including three nitrogen injections, slug in and slug out. The slug tests were conducted in the monitoring wells 1, 2, 71-1, 71-2, 72, and 71-3. Results are presented in the
Supplementary Materials, while the interpretation follows the procedure given in [
49].
Point dilution test with Uranine solution was implemented in well 71-3, which penetrates the shallow phreatic aquifer. The tracer, with a concentration of 1000 ppb was injected uniformly along the hole, and time series concentrations were measured by a Fluorometer (Turner Design, San Jose, CA, USA). The rate of decrease in tracer concentration was used to calculate the horizontal groundwater velocity in the well by the procedure described in Reference [
50]. Results are given in The
Supplementary Materials.
Interference recovery test was conducted by analyzing the level recovery in wells 71-2 (unit B) and 71-1 (unit C) after stopping pumping in the nearby (within 50 m) well 211 (
Figure 1). Test results were used to calculate the transmissivity and the storativity of the aquifer, using the solution of Reference [
51] for confined aquifers.
Radon (222Rn) was sampled in both seawater and groundwater, using pre-vacuumed 4.5 L glass bottles. Seawater was sampled close to the seafloor in coastal water up to 700 m offshore (maximum water depth of 5 m) and groundwater was sampled from wells (the 71 series) using a peristaltic pump. 222Rn activities were measured by alpha-scintillation Lucas cells, using a modified emanation technique. In this method, helium is circulated and bubbled through the sampling bottle, resulting in the extraction of 222Rn. The radon is trapped on activated charcoal (at −90 °C) and then transferred to the Lucas cell.
Numerical simulations were conducted with the 2D FEFLOW program, which is a finite element code, allowing the examination of both flow and salt transport. To deal with the specific process in a precise way, a very dense net of triangular elements (~100,000) was built (element size range from ~1 m to 10 m). The model configuration included seven layers, in accordance with the geological section of the southern Carmel coastal plain: five layers in the Quaternary section and two in the underlying Late Cretaceous (
Figure 3b). The conceptual model and model calibration were based on field measurements, presented and discussed in
Section 4, including seabed geology (provided by the CHIRP survey), the TDEM survey, EC profiles, hydrological field experiments, and groundwater levels. The length of the model section is 7 km, from the sea on its west to the Carmel Mountains on its eastern end.