Spatial characterization of the seawater upconing process in a coastal Mediterranean aquifer (Plana de Castellón, Spain): evolution and controls

In this contribution, we describe the formation and evolution of the upconing process in a Mediterranean coastal aquifer. The study area has experienced severe salinization over the last 40 years because of intensive exploitation of groundwater. We used historical and current records of piezometric levels and chloride concentrations to trace the development of the salinization of the aquifer. We defined the 3D shape of the saline wedge from the spatial distribution of chloride concentrations and vertical well logs of electrical conductivity using monitoring network data. Upconing first appeared in the early 90s and has continued until the present day. In this study, we examined the intensity of the upconing process. Dry periods and the associated increases in pumping caused the advance of seawater intrusion. The sharp reduction in groundwater withdrawals over the last 10 years has caused the saline wedge to move backwards, although the ongoing pumping and the climate conditions mean that this retreat is quite slow.


Introduction
Saltwater intrusion is a serious problem for aquifers in many countries in Africa (Steyl and Dennis 2010;Bouderbala 2015), Australia (Werner et al. 2005), Asia (FAO 1997;Liu 2004;Parck et al. 2012;Pratheepa et al. 2015), North America (Cardoso 1993; Barlow and Reichard 2010), and the Caribbean (Morell et al. 1997;Boschetti et al. 2015). In Europe, seawater intrusion affects a large number of coastal aquifers, particularly along the Mediterranean coast of countries such as Spain (Gómez et al. 2003;Renau-Llorens 2010), Italy (Barrocu 2003), Greece (Petalas and Lambrakis 2006), and Turkey (Günay 2003). The problem is mainly the result of inadequate management of water resources resulting from the high density of wells, their proximity to the coastline, and, in some cases, inadequate well design and construction.
Groundwater salinization in coastal aquifers arises from a range of sometimes overlapping factors. In most cases, the predominant process is seawater intrusion that is caused by intensive exploitation. Other associated mechanisms also frequently contribute to salinization, such as regional flows with high salinity (Vengosh et al. 1999), natural mixing processes between seawater and continental water (Fleury 2005), mobilization of connate water (Lambrakis and Marinos 2003;Stoecker et al. 2013), and urban, agricultural, and industrial discharges (Custodio 2010; Mondal et al. 2011).
Seawater intrusion can generally be categorized into one or more types: horizontal and upward movement of the interface, downward leakage of brackish or saltwater from surface water (such as estuarine environments), or upconing beneath individual wells or well fields (Cheng and Ouazar 2003). Saltwater upconing is a widespread problem in many coastal and inland aquifers around the world (Zakhem and Hazef 2007;Kouzana et al. 2009;Khairy and Janardhana 2013;Rey et al. 2013;Cai et al. 2014) and is a topic of great interest. In general, studies have mainly been restricted to examinations of, and controls on, local saltwater upconing below a pumping well. Numerous analytical and numerical methods have been employed to determine the position of the sharp interface and the critical pumping rate (Chandler and McWhorter 1975;Garabedian 2013).  applied a finite element model to determine the maximum permissible discharge rate in an inland aquifer. A number of computer codes have been developed that consider hydrodynamic dispersion in the upconing process and simulate the combined density-dependent flow and salt transport (Diersch et al. 1984;Zhou et al. 2005;Paster and Dagan 2008). Various studies have demonstrated that numerical modelling is a successful tool for investigating upconing in field applications (Aliewi et al. 2001;Paniconi et al. 2001;Marandi and Vallner 2010;Cai et al. 2014). Because of the challenges associated with making field-based measurements of salt transport dynamics below the pumping well, saltwater upconing research has also been developed at the laboratory scale (Oswald et al. 2002;Werner et al. 2009).
A considerable proportion of the knowledge on the upconing process is based on laboratory and mathematical modelling. To date, however, there have been few field-based studies that report detailed measurements of upconing processes. Recent studies have reported that electrical resistivity profiles can be used to capture and define the shape of the upconing and seawater intrusion (Rey et al. 2013;Kura et al. 2014). However, direct observations of saltwater upconing and additional investigations are needed to give an improved understanding of the 3D nature of seawater upconing in real-world systems (Werner et al. 2013).
In this paper, we have described the evolution of seawater intrusion in the southern part of the Plana de Castellón aquifer over the last 42 years. We used direct field observations (chloride concentration and electrical conductivity [EC] well logs) to define the extent of the upconing process, including its genesis, evolution, and geometry. This approach may be used (a) to establish the relationship between climate conditions, pumping regime, and seawater intrusion; (b) model seawater upconing processes, and (c) to plan management activities, such as artificial recharge, in such an area (García-Menéndez et al. 2015).

Site description and groundwater management
The study area is in the southern part of the Plana de Castellón aquifer on the Spanish Mediterranean coast, where groundwater is extensively exploited (Rambleta Area) (Figure 1).
The Plana de Castellón is an alluvial plain that covers an area of 490 km 2 . There is intensive agricultural, industrial, and tourist activity on the plain, and the area has a population of around 300,000 inhabitants. It is bounded to the north by the Sierra del Desierto de las Palmas, to the southwest by the Sierra del Espadán, and to the east by the Mediterranean Sea. consecutive years or more in which the rainfall is below the annual mean, are relatively frequent; extreme rainfall events with more than 200 mm over a 24-h period are also frequent.
There is a permanent watercourse (Mijares River) in the northern half of the Plana de Castellón that supplies water for agriculture in this part of the aquifer. In contrast, groundwater is the main source of urban, agricultural, and industrial water in the southern half of the Plain.
Since the 1960s, the study area has supported significant agricultural activity. Before 1970, the amount of groundwater pumped in the Rambleta was between 4 and 6×10 6 m3/year. The production wells were shallow (25-40 metres) but the flow rates sometimes exceeded 50 L/s. Likewise, there were numerous wells along the coastal strip (north of Moncófar), which provided a further 4×10 6 m3/year. Additional wells were constructed in the Rambleta to extend the irrigated area and to compensate for the gradual closure of coastal wells where salinity had increased rapidly. These new wells were generally between 40 and 75 metres deep. As a consequence, wells were densely packed into an area of barely 2 km 2 , and were only between 100 and 200 m apart. In the mid-nineties, the amount of groundwater pumped for agriculture in this area increased to 21×10 6 m 3 /year and an extra 3×10 6 m 3 /year was abstracted to supply a drinking water plant.
As detailed below, this intensive exploitation provoked a widespread drop in the piezometric surface and encouraged the progression of the saline front. Even now, there is ongoing degradation of groundwater quality and almost all of the wells in the coastal area and many of the wells in the Rambleta have been abandoned. As a result, some of the production, including the drinking water supply, has been moved to adjacent aquifers.
The groundwater pumping has now decreased to around 14×106 m3/year, almost half of the peak production, for various reasons. The irrigation system has been modernized from a flooding to a drip-feed system, because of which the unit application rate has decreased from 8,500 m 3 /hectare/year to 5,200 m 3 /hectare/year; infrastructure was constructed; 15 % of the croplands were abandoned, and treated wastewater, amounting to 2.5×10 6 m 3 /year, has been incorporated into the irrigation network. The UDA formation is 85-m thick and comprises polygenic gravels, clays, sands, and sandstones.

Geological and hydrogeological characterization
Permeability is due to intergranular porosity (2-8 %) and transmissivity values are ~300 m 2 /day, with a maximum of 1,000 m 2 /day. The highest yields are from the wells in the Rambleta Area, which have specific flows of between 6 and 10 L/s/m. The impermeable base is 20-m thick, and comprises a layer of clays with gravels.
The LDA is composed of limolites and clays with layers of sandstones and polygenic conglomerates. Permeability is attributable to the porosity of the scarce sand and conglomerate layers, so this formation is of less hydrogeological importance than the UDA. The LDA is between 50 and 100 m thick, increases towards the coast, and is sometimes absent where the Mesozoic substratum (MS) rises closer to the ground surface because of the NNE-SSW direction fault (parallel to the coastline).
The MS underlies all the above-mentioned formations. It comprises limestones, marls, dolostones (Muschelkalk facies), orthoquartzitic sandstone and limolites (Buntsandstein facies), and marls with gypsum (Keuper facies), which have undergone intense folding and fracturing. In general, they form a paleorelief that has been fossilized by the overlying sediments. The MS formation includes two aquifers corresponding to Muschelkalk dolostones and Buntsandstein sandstone. In these settings, there may be hydraulic connections with the overlying permeable formations.

Materials and methods
Detailed geological knowledge of the Rambleta Area was assembled from a comprehensive inventory of wells, boreholes, and lithological columns, and from measurements of electrical resistivity tomography, collected in several campaigns.
The general characteristics (coordinates, elevation, depth, design features, etc.) of 115 wells were recorded, from which 33 lithological columns were compiled.
A computer (Terrameter SAS4000 ABEM) with an imaging system (LUND) and an interelectrode spacing of 15 was used for the geophysical surveys. A geophysical inversion software package (RES2DINV v.3.46b) was used to convert the field data, and the robust and constrain inversion algorithm was used to obtain the resistivity profiles. We produced a total of six profiles, four of which were 1,200-m long and two that were 900-m. Additional induced polarization measures were carried out in four of the profiles.
Historical data of piezometric levels and chloride concentrations were obtained from the monitoring network of the Geological Survey of Spain (IGME), which was initiated in the early of the groundwater ranged from 1,500 to 4,500 µS/cm, and the water densities were below 1000.77 kg/m 3 and did not vary significantly. As such, piezometric corrections were not needed because the error associated with the head measurements was generally less than 0.02 (Post et al. 2007).
Water was sampled from a depth of 5 m below the water level with a discrete interval sampler (Solinst 425) and a controlled opening mechanism. EC, pH, and temperature were recorded in situ with a waterproof pH/conductivity/temperature portable meter (Eutech Instruments PC650). Further samples for chloride analysis were stored in polyethylene bottles. Chloride concentrations were determined by titration with silver nitrate, following the method of Mohr.
Vertical measurements of EC and temperature were logged bimonthly in all monitoring wells from December 2012 to October 2014 using a 100-m long well logger (TLC Solinst).
Most of the historical data were from samples collected by pumping, which means that there is a difference between the historical and the recent data. Samples from pumping describe the quality of the part of the aquifer that is influenced by the pump, while samples from a depth of 5 m below the water level describe the quality at that specific depth. This can, in some cases, lead to quantitative differences in the chloride concentrations. In this study, the differences were not sufficient to significantly alter the trends in salinity.
The groundwater contour lines and the chloride concentration contour maps were created using geographic information system (GIS) software (ArcGIS 10.1). Contour maps were generated using the spatial interpolation Kriging method, which was included in the GIS software (ArcToolbox), and then were modified to take account of hydrogeological criteria using ArcMap editing tools. A 1:25,000 IGN map was used as the topographic base.

Piezometric behaviour
Because of the groundwater exploitation and climate, piezometric levels have changed markedly since 1970. These changes are reflected in the piezometric history of Wells 25 and 30 (Figure 4), which are located close to the coastline and in the Rambleta Area, respectively. The trend in both cases is quite similar but the amplitude of the oscillations differ. Both wells show periods of rising and falling levels, which correspond quite well with wet and dry periods, respectively, and also with the exploitation regime. The influence of climate and exploitation generally overlap.
They may have different weights and may vary, depending on the length of the dry or wet periods and the pumping intensity.
Piezometric levels were clearly below sea level between 1973 and 1987 and also from 1993 to 2001, with some measurements dropping to 5 m below sea level (b.s.l.). In contrast, the highest levels were from 1972 to 1973 and from 1987 to 1993, when they exceeded 4 m a.s.l. In the current period (from 2012 to 2014), the depth to the water level is above sea level, and is between 0.5 and 3 m a.s.l.

Hydrochemical behaviour
There are two main hydrogeochemical water types in the study area: the calcium chloride type and the calcium-magnesium sulfate type. The chloride water corresponds with a mixture of freshwater and seawater as a consequence of seawater intrusion, while the sulfate water is associated with regional flows from the Triassic limestones and dolomites of the Mesozoic substrate (Fidelibus et al. 1992;Escrig et al. 1993;Giménez 1994;Morell et al. 1996;Giménez and Morell 2008;Renau-Llorens 2010;Gménez-Forcada and Vega-Alegre 2015).
In the early 1970s, salinization followed the usual pattern of aquifers influenced by seawater intrusion, with a progressive decrease from the coast towards the interior. However, in 1972 ( Figure 5), a zone with chloride concentrations above 500 mg/L was visible 2 km inland (north of Moncófar). At this time, the chloride concentrations in the Rambleta Area were less than 200 mg/L.
In the Moncófar area, the water quality in the wells deteriorated significantly, so that, by 1984, the chloride concentrations exceeded 2,000 mg/L ( Figure 6). Also, concentrations above 500 mg/L were detected 5 km inland, thereby influencing the eastern part of the Rambleta (Well 19).
In 1995, concentrations were still very high, but the chloride concentrations had dropped to below 2,000 mg/L in the north of Moncófar (Figure 7), and an area appeared between this area and the Rambleta in which the concentrations were below 750 mg/L. In contrast, concentrations In 2012 ( Figure 9) the upconing still persisted over a wide area, and had chloride concentrations ranging from more than 500 mg/L to slightly more than 750 mg/L. Even though the concentrations were lower than those recorded previously, the salinization was still severe. At the same time, the freshwater zone in the north of Moncófar became wider, with concentrations between 100 and 300 mg/L.  At 0 m a.s.l. (Figure 12), the salinities were highest, and above 3,500 S/cm, in the coastal strip and in the Rambleta area. There is a body of fresh water with EC values of 1,000 S/cm to 2,000 S/cm between these two zones that clearly separates the area of lateral movement of the intrusion (coastal area) and the upconing area.
At 5 m b.s.l. (Figure 13) the distribution of salinity in the Rambleta is quite similar to that at 0 m a.s.l. but the intermediate band of fresh water is narrower and minimum EC values exceed 2000 S/cm. Likewise, in the coastal strip, the water is more saline (up to 4,000 S/cm).
The EC values at 10 m b.s.l. exceed 4,500 S/cm along the coastal strip ( Figure 14), but are generally between 3,000 and 4,000 S/cm in most of the study area. It is worth noting that the inland zone with high salinity does not exactly correspond with the Rambleta area, but is slightly displaced to the east, probably because the saline front has moved towards the sea.
Finally, at 15 m b.s.l. (Figure 15), the areas with EC values above 4,000 S/cm are linked, and extend from the coast to the Rambleta Area. Figure 15 highlights the elevated and roughly homogeneous salinity in the deeper part of the aquifer.
The geometry of the salt body reconstructed from Figures 12, 13, 14, and 15, clearly shows that there is an area of higher salinity in the Rambleta (Figure 16). We added a hydrogeochemical cross-section (A-A´) to represent the shape of the seawater intrusion ( Figure 16). The horizontal advance of saline water moving inland from the sea and the vertical upconing in the Rambleta Area are clearly visible. There are two saline cones, a principal cone and a secondary cone. This suggests that the saline cone was broader and is now moving towards the sea, which is consistent with the end of the groundwater exploitation at Wells 6 and 32 in the central part of the Rambleta. Conversely, the persistence of the secondary saline cone could be related to the pumping that continues in the western boundary (Well 2).

Conclusions
We used historical and recent data from a dense monitoring network to accurately define the geometry of the seawater intrusion at the study site. We reconstructed the dynamics of seawater intrusion and the development of upconing from historical chloride concentrations and piezometric measurements.
Because we used historical data, we had to take a number of methodological precautions; for example, we needed information about the construction characteristics of the wells (construction depth, depth into the aquifer, grids situation), and we assessed the representativeness of the samples based on the collection method. It should be noted that, where the salinity is high, piezometric corrections should be applied to account for differences in the density of water.
We evaluated the influence of the recharge and exploitation regimes on the fresh water-saline water balance by studying the development of the chloride concentrations and groundwater levels over a period of 42 years.
The dry periods that frequently occur in Mediterranean climates and the consequent increases in pumping cause severe declines in piezometric levels and progression of the saline front.
Conversely, wet periods tend to restore the situation. This is the expected response, and can be applied to aquifers with similar characteristics. However, in this case, the normal situation has been made more complex by the ongoing groundwater exploitation. Dry and wet periods occurred regularly over the period when there were intensive changes in the pumping regime.
During the first half of the study period  pumping continually increased from 10×10 6 to 24×10 6 m 3 /year, while in the second half, pumping gradually decreased from a maximum of 24×10 6 to 14×10 6 m 3 /year.
Chloride contour maps for different dates have shown how salinity has developed in different parts of the study site and have indicated higher chloride concentrations in the Rambleta, the area of most intense pumping. They have also allowed us to follow the emergence of an upconing process in the early 90s and the subsequent retreat of the saline front toward the coast.
However, the movement of the saline cone is slow and the recovery is subject to the variable climate regime and the still significant pumping in the area.
Thirty four vertical logs were made of EC (December 2013) to characterize the 3D nature of the saline cone. Our results show that this method can be used to effectively determine the geometry of the upconing process.
The balance of the saline front is unstable and is strongly affected by boundary conditions (pumping regime and variations in recharge). An adequate observation network should be established to ensure that EC profiles can be recorded at least every 6 months on an ongoing basis in such areas.