Effects of intensive aquifers exploitation on groundwater salinity in coastal wetlands

The coastal plain of the Río de la Plata constitutes a large wetland which develops on the right margin of the river estuary. Anthropic activities such as intensive exploitation of groundwater carried out in the vicinity of the wetland can modify the natural hydrological regime. The aim of this work is to asses the effects of intensive aquifer exploitation in coastal wetlands using hydrogeological models. Such models allow to evaluate changes in the environmental conditions of wetland at regional level. The hydrogeological model exposed in this work shows how the intensive groundwater exploitation affects the wetland area, generating important variations both in the groundwater flows and in the salinity of the groundwater. Identification of these modifications to the environment is important to generate guidelines leading to minimize these affectations.


| INTRODUCTION
Coastal wetlands are among the most fluctuating and productive ecosystems of the world, and perform a wide range of ecosystem services of socio-economic value (Odum, 1978). These values include shoreline stabilization, sediment and nutrient retention, high primary and secondary production, habitat and food resources for terrestrial, aquatic fauna, coastal water quality buffering, biomass and biodiversity reservation, and recreation and tourism amenities (Gedan, Kirwan, Wolanski, Barbier, & Silliman, 2011;Mitsch & Gosselink, 1993;Mitsch & Gosselink, 2000). These ecosystems play an important role in wildlife conservation and can also act as sinks or sources of a wide range of substances, such as nutrients, organic matter, pollutants, and so on (Boorman, 1999). Ecological systems, however, are usually affected by human activities, which may cause a loss and degradation of their natural status, a decline of their biodiversity, an alteration of their ecological functioning, and a limitation of their ecosystem services (Brinson & Malvárez, 2002;Moiwo, Yang, Li, Han, & Yang, 2010;Yamashiki, Nakamura, Kurosawa, & Matsui, 2006).
The coastal plain of the Río de la Plata constitutes a large wetland which develops in the coastal zone on the right margin of the Río de la Plata estuary. Although numerous natural reserve areas are present in the wetland, there are also large urban sites, ports and industries along the middle estuary. The wetland sector, located in the vicinity of La Plata city, presents a hydrological functioning which depends on the contribution of water coming from rains, tidal flows and groundwater discharge. Such groundwater discharge constitutes one of the main hydrological components that sustain the wetland and it derives both from the local flows in the unconfined aquifer and from the regional flows in the underlying semi-confined aquifer (Carol, Kruse, Mancuso, & Melo, 2013). Although in this wetland sector natural reserve areas are present, urban and industrial development in other areas has affected surface and groundwater dynamics and chemistry (Santucci, Carol, Borzi, & Garcia, 2017;Santucci, Carol, & Tanjal, 2018;Vecchioli, 1998). Likewise, anthropic activities such as intensive exploitation of groundwater carried out in the vicinity of the wetland  can modify water flows affecting its natural hydrological regime . The objective of this work is to asses the continental aquifer intensive exploitation effects on groundwater salinity of adjacent coastal wetlands, through the use of hydrogeological models. This will enable us to evaluate changes in the environmental conditions of wetland at regional level, and to generate bases for the water resource management in the region.

| STUDY AREA
The study area comprises the wetland sector of the coastal plain of the Río de la Plata adjacent to La Plata city ( Figure 1). This wetland sector presents anthropized sectors and also reserve areas where ecological characteristics are preserved. In the most continental sector, the wetland is bordered by a loess plain environment which is topographically higher than coastal plain, and consists of aeolian sediments with thicknesses close to 40 m, which have been reworked by water in some sectors (Teruggi, 1957).
The area is characterized by a humid temperate climate with a mean annual precipitation of 1,061 mm (period: 1901-2002). The mean annual temperature is 16 C and, according to the soil water balance (Thornthwaite & Mather, 1957), actual evapotranspiration is 783 mm/year, infiltration 225 mm/year and runoff 53 mm/year (Kruse et al., 2004). This sector of the coastal plain is part of the lower basin that comprises a set of streams which drain the water from the adjacent loess plain to the Río de la Plata. The wetland receives the local groundwater discharge from the unconfined aquifer and also the regional discharge from the underlying semi-confined aquifer. Precipitation is the source of recharge of the unconfined aquifer, whose regional groundwater flow is towards the Río de la Plata . The recharge of the underlying semi-confined aquifer is local and indirect from the unconfined aquifer mostly occurring in the high basin sector, and its groundwater flow is also towards the Río de la Plata. The semi-confined aquifer presents high salt contents in the adjacent area to the middle Río de la Plata estuary due to a paleoseawater intrusion that is preserved at present (Santucci, Carol, & Kruse, 2016;Santucci, Carol, & Kruse, 2017).
In the loess plain sector, the unconfined aquifer is composed of silty to silty-clayed sediments with calcium carbonate concretions.
The mean aquifer thickness is 32 m and it generally wedges out towards the Río de la Plata. In the coastal plain sector the unconfined aquifer corresponds to silty to clayed sediments that alternate with lenses of fine-grained sands and marine shells (Fucks, D'amico, Pisano, & Nuccetelli, 2017). The uppermost sediments in the coastal plain consist of fine-grained sands corresponding to alluvial deposits of the levee in the vicinity of the Río de la Plata. Together, both units have an average thickness of 30 m. The unconfined aquifer overlays a 4 m average thick silty-clayed aquitard which hinders the hydraulic transmission between the unconfined aquifer and the underlying semi-confined aquifer. The semi-confined aquifer is composed of fineto medium-grained quartz sands of fluvial origin with a mean aquifer thickness of 21 m and overlie green clays.
In the loess plain, where La Plata city is located, the semi-confined aquifer is intensively exploited due to the fact that it constitutes the main source of water supply in the region. The urban population has F I G U R E 1 Location of the study area and delimitation of wetland and loess plain sectors. Distribution of the pumping and observation wells, greenhouses and urban area increased progressively during the last years having registered 694,613 inhabitants for 2010 (INDEC, 2010). This urban sprawl took place from the inner area towards the environs of the city, especially in the south region.
In a more restricted south sector, distant from the urban area and in the loess plain uplands, intensive horticultural activity is developed.
This activity is performed in small pieces of land. This generates an important production, in which supplementary flood irrigation was introduced in the 1980s and drip irrigation in the 1990s. As consequence, during the last years, land use changed from open field to greenhouses, decreasing the unconfined aquifer recharge due to the plastic coverage of the greenhouses structures and increasing the need for artificial irrigation (Delgado, Carol, Casco, & Mac Donagh, 2018). This intensive water exploitation generated both in the urban area for human supply and in the agricultural area for irrigation, changes the water balance at the regional level (Auge, 2005).

| METHODOLOGY
A numerical model of groundwater flows within the unconfined and semi-confined aquifers was developed using a three-dimensional (3-D) system modelling with the MODFLOW software package (Harbaugh, Banta, Hill, & McDonald, 2000), and the generic computer program SEAWAT (Langevin, Thorne Jr., Dausman, Sukop, & Guo, 2008) was applied to simulate the salinity changes of water in the wetland area. It allowed us to simulate the temporal variations in groundwater salinity in the wetland area at regional level (loess plain and coastal plain). In previous research, MODFLOW was applied for steady state groundwater flow simulation (the natural conditions was used as the initial condition for transient state) and transient simulations for 1940 (a prior 10-year simulation was run to achieve model calibration), 1988 and 2008 . In the modelled area, topographic curves, streams, aquifer levels and exploitation wells distribution were digitized. Groundwater recharge occurs due to the infiltration of rainfall excess and was considered as constant in all models. For the year 2008, 154 wells with an exploitation volume of 1.44 10 3 m 3 /day per well were considered. Those wells were installed in the semi-confined aquifer to urban needs supply. An actual well distribution, depth of extraction (−40 m asl) and exploitation rate were used. In addition, greenhouses for horticulture use a volume of 3.2 10 3 m 3 /day for irrigation purpose, and their location comprises 36 km 2 in the loess plain uplands ( Figure 1). The pumping system for irrigation was simulated with an uniform well distribution between the cells of semi-confined aquifer (−40 m asl), below the agricultural area.
A constant recharge of 6.1 10 −4 m/day given by the rainfall infiltration from surface to the unconfined aquifer was considered for the whole area (Kruse et al., 2004), except for the greenhouses and urban areas. The infiltration rate was set as constant because the years considered for the simulation show similar hydrological situations regarding the average conditions of the water balance . In greenhouses areas, it was considered a withdrawal of 3.8 10 −4 m/day, taking into consideration the impermeable greenhouses roofs, and 1.3 10 4 m 3 /day of groundwater extraction from the unconfined aquifer for irrigation purposes. For the urban areas it was considered a hypothetical recharge of 0 m/day. Potentiometric head for 2015 were simulated from 2008 numerical model results. For 2015, it was considered an increase of 66 km 2 in agricultural land use, and also an increase in water needs for irrigation (6 10 4 m 3 /day from semiconfined aquifer and 2.4 10 4 m 3 /day from unconfined aquifer). The increase of horticultural areas was defined through satellite images analysis.
The regional model was developed for a 915 km 2 basin, which is drained by five main streams. The modelled area was discretized into 95 x 97 cells with 440 m × 315 m in x and y directions. In the z direction, a three-layered system with a thickness ranging from 15 to 48 m in Layer 1 (unconfined aquifer), 3-18 m in Layer 2 (aquitard) and 8-47 m in Layer 3 (semi-confined aquifer) was defined. The thicknesses of the modelled aquifer units and the aquitard were adjusted according to the records of perforations and isopach maps of the area (García, Kruse, & Laurencena, 2017). Regionally, groundwater flows from the basin limit (the west limit) to Río de la Plata (to northeast). For boundary condition, a specified-head boundary of 0 m asl was considered at the northeastern limit of the basin, where the Río de la Plata is, which is a natural discharge boundary. For the unconfined aquifer, the southwestern limit was considered as no-flow boundary, because is a basin perimeter. For the semi-confined aquifer and the confined layer, the western limit was considered as specified head boundary, to account by the regional inflow from semi-confined aquifer who discharge into Río de la Plata.
The hydraulic head was specify from 14 to 25 m asl (from southwest to northwest), based on the potentiometric maps of the semi-confined aquifer . The northwest and southeast boundaries are hydraulic boundary (no-flow boundaries), for modelling purpose.
There were defined from the potentiometric maps of the unconfined and the semi-confined aquifer  as groundwater streamlines, perpendicular to Río de la Plata and parallel to the equipotential lines. It was assumed no-flow through the streamline. Constant head was considered for Río de la Plata and drains for the streamlines, existing groundwater discharge zones when the water table exceeds the terrain level.
The hydraulic parameters used in each of the modelled layers are based on the values estimated in field tests through several pumping tests (Auge, 2005;Kruse et al., 2004). As a result of these tests, in the unconfined aquifer, the hydraulic conductivities vary between 3 and 10 m/d, the average transmissivity is 310 m 2 /d and the effective porosity is 0.08. This layer was divided into two main zones: one zone coincides with the high plain area (loess sediment), and the other zone coincides with the coastal plain area (wetland).
In the aquitard, the horizontal hydraulic conductivity (Kh) range Based on the hydrodynamic numerical model, the generic computer program SEAWAT (Langevin et al., 2008) was applied to simulate the salinity changes of groundwater for steady state (natural conditions) and transient state (2015). SEAWAT is a coupled version of MODFLOW (Harbaugh et al., 2000) and MT3DMS (Zheng, 2006;Zheng & Wang, 1999) and it was used to simulate three-dimensional, variable-density groundwater flow. The variable-density groundwater SEAWAT integrates MT3DMS Transport Process to solve the solute transport equation (Langevin et al., 2008).
Boundary condition for salinity in the unconfined aquifer was defined as: (a) constant concentration of 0.5 g/L at Río de la Plata (Santucci et al., 2016) (northeast); (b) watershed boundaries with constant concentration from 0.44 g/L (northwest) to 0.86 g/L (southwest) (reference data from observation wells P21 and P23) ( Table 1) For the transport simulation in the aquifer (MT3DMS/SEAWAT software), the longitudinal and transverse dispersivity (αL, αT) were set to 3.6 m and 0.36 m lengths, respectively (Perera, Jinno, Hiroshiro, & Tsutsumi, 2008;Shoemaker, 2004), and the molecular diffusion to 1.0 10 −10 m 2 /d. Fresh water density was set to 1,000 g/L and salt water density to 1,024.5 g/L.
The fluid density at the reference concentration (DENSEREF) and the slope of the linear equation of state that relates fluid density to solute concentration (DRHODC or DENSESLP, from Langevin et al., 2003) are parameters needed to compute fluid density by the variable-density groundwater flow equation (VDF Process in SEAWAT routine). The fluid density in the simulation was calculated as a function of the specie TDS, in this case. A DENSEREF value of 1,000 indicates that the reference fluid density (freshwater in this case) at 25 C is 1,000 g/L. A DRHODC value of 1.3 indicates that the density will linearly vary between 1,000 and 1,024.5 g/L for freshwater and for saltwater, respectively. DRHODC can be estimated by dividing the density difference by the concentration difference. In this case, it will be as follows: The calibration for salt concentration in steady state was assessed on the basis of observation data of salt concentration from 11 piezometers located at the modelling area boundary, of unconfined aquifer (4 wells) and semi-confined aquifer (7 wells) ( Table 1) and on the semi-confined aquifer (12 wells) ( Table 1)

| RESULTS
At regional scale, rain water infiltration is the recharge source of the unconfined aquifer throughout the study area, while the semi-  (Figures 3 and 4).
Under natural conditions, regional groundwater flows from the basin header areas located in the southwest towards the coastal wetland and the Río de la Plata, in both semi-confined and unconfined aquifers (Figure 3a,b). Analysis of the hydraulic load differences between both aquifers (Figure 3c) shows that, in the watershed divide areas, the water level of the unconfined aquifer is higher than that of the semi-confined. In this way, the watershed divide sectors constitute preferential areas for semi-confined aquifer recharge. On the other hand, the water courses and the wetland constitute zones of discharge of the unconfined aquifer and in some of these sectors the In the loess plain area, the groundwater has low salinity both in the unconfined and semi-confined aquifers, while the salinity of the groundwater in the wetland area is quite variable (Figure 3d,e). In the wetland, the unconfined aquifer presents areas with low water salinity (areas with higher rainwater recharge) and sectors with saline water (transit zones and groundwater discharge). On the other hand, in the semi-confined aquifer the water is saline, being present as a wedgeshape (Figure 3e,f).
The settlement of La Plata city and the urban growth generated around the city, produced modifications in the potential infiltration of the substrate as waterproofing due to buildings and pavement of the streets, that tend to decrease the recharge of aquifers in the sector of the urban area. Likewise, the intensive exploitation of the semiconfined aquifer causes a cone of depression that modifies the groundwater flows of both the semi-confined and unconfined aquifers, and also modifies the hydraulic relation between the aquifers.
Besides that, the development of greenhouses, which generate waterproof areas, and the increase of pumping in preferential recharge zones of aquifers constitutes another modification to the natural hydrogeological functioning of the system. confined aquifers tends to increase, which is most evident in the semi-confined aquifer, in the sector close to the limit coastal plainloess plain. Conversely, in mainly natural sectors (southeast), the increase in salinity is not noticeable.

| DISCUSSION
Coastal wetlands have been suffering from serious degradation, alteration or loss of ecosystem services due to intense anthropogenic activities (Lemly, Kingsford, & Thompson, 2000;Newton, Carruthers, & Icely, 2012;Zhao et al., 2016), and thus, coastal wetlands are listed among the most heavily damaged of natural ecosystems worldwide (Barbier, 2011;Jones et al., 2018). Wetlands degradation associated with modifications in hydrological functioning is a problem that affects many coastal regions worldwide and it has accelerated in recent years not only owing to anthropogenic activity, but also because of climate change (Perillo, Wolanski, Cahoon, & Brinson, 2009).
Hydrological changes and the effects they produce in wetlands can be easily studied with field detail studies at local level. But at the regional level, it is necessary to acquire models that allow not only to F I G U R E 5 Salinity variation between natural condition and 2015 for (a) unconfined aquifer and (b) semi-confined aquifer identify the problems, but also to analyse their evolution over time.
The results of such models should allow to obtain the prediction in changes of coastal wetlands to human impacts, so that the adaptation strategies can be put in place (Reyes, 2009).
Groundwater flows constitute one of the main water sources that sustain coastal wetlands, being the discharge of local and regional groundwater flows a significant hydrological component in coastal areas (Custodio, 2000). The results obtained with the regional hydrogeological modelling show how the intensive exploitation of aquifers in continental areas adjacent to the coastal wetland affect the groundwater flows towards the wetland, which is a feature that had already been evidenced in previous studies .
However, the novelty of the model is that it shows that these modifications in the groundwater flows lead to changes in the groundwater salinity of both the unconfined and semi-confined aquifers in the wetland area. If salinity changes of the wetland sectors, which are affected and not affected by the intensive exploitation of groundwater, are compared, a contrasting evolution from the natural state is shown. In the wetland sector, without intensive exploitation affectation, the salinity in both the unconfined and semi-confined aquifers tends not to increase or slightly increase (southeast sector in Figure 5). This behaviour is what would be expected the wetland to expose in natural conditions over time. The groundwater flow of fresh water from the more continental sectors tends to displace the saline paleo-wedge towards the estuary of the Río de la Plata. This means that the salinity of groundwater in the semi-confined aquifer in the wetland area tends to decrease over time.
Conversely, in the sector where the groundwater flow is affected by intensive exploitation and the recharge decrease as consequence of waterproofing by greenhouses, the salinity varies. In this sector, the salinity does not show important variations in the unconfined aquifer, while in the semi-confined aquifer the salinity increases (centralnorthwest sector in Figure 5). This salinization, although mainly affects the sector close to the limit coastal plain -loess plain, also causes a salinization of the groundwater in the wetland itself. This is due to the fact that the intensive pumping of the semi-confined aquifer from the continental sector reverses the groundwater flow, causing the intrusion of the saline paleo-wedge towards the continent. This causes a groundwater salinity increase in the semi-confined aquifer in the wetland sector adjacent to the exploitation zone.
The models generated show, as it has been exposed in other regions of the world, that intensive exploitation of groundwater is one of the main causes of wetland deterioration (Bernaldez, Benayas, & Martinez, 1993;Carol et al., 2013;Cooper, D'Amico, & Scott, 2003;Cooper, Wolf, Ronayne, & Roche, 2015;Johansen, Pedersen, & Jensen, 2011;Moiwo, Lu, Zhao, Yang, & Yang, 2010;Park, Wang, & Kumar, 2020;Rochow, 1994;Serrano & Serrano, 1996;Suso & Llamas, 1993;Winter, 1988;Wise, Annable, Walser, Switt, & Shaw, 2000). The modification in the underground flows, that support the wetlands, leads to a decrease in the areas of wetlands and / or to ecological effects in them (Bernaldez et al., 1993). This problematic may further be worsened in coastal wetlands due to the saline intrusion processes that characterize littoral areas. In these, intensive pumping in sectors adjacent to wetlands or within them, not only leads to modifications in water flows, but also to their salinization.
If we consider that salinity of groundwater is one of the main factors of the physical environment determining the environmental characteristics of the ecosystems (Watson & Byrne, 2009), it is expected that changes in wetland environments slowly begin to take place.
Given that no significant alterations have been recorded in the wetland ecosystems today, the diagnosis generated from the modelling allows to alert the managers of the water resource. Until now, they have only evaluated the problems of intensive exploitation associated with water urban supply, without considering agricultural impacts on recharge and on groundwater exploitation, and the consequences of what this produces in adjacent wetland environments.
6 | CONCLUSIONS The regional-level hydrogeological model shows that the intensive exploitation of aquifers in areas adjacent to the wetland, produces variations both in the salinity and groundwater flows of the wetland. A decrease of water inputs from the regional groundwater discharge of the unconfined and semi-confined aquifers occurs as a result of the investment of the flow caused by intensive exploitation. These flow modifications lead to changes in the salinity of the groundwater within the wetland area. Flow variations showed that in the area affected by intensive exploitation, the unconfined aquifer salinity will tend to slightly vary, while in the semi-confined aquifer it will tend to increase, which is most evident in the sector close to the limit coastal plain-loess plain. These changes will modify the long-term environmental characteristics of the wetland, and hence their early identification is important for the generation of guidelines that tend to minimize these affectations.
Human activities and climate change have turned coastal wetlands into highly vulnerable environments. Wetlands monitoring and diagnosis used to mitigate these changes generally tends to evaluate the changes at the local level within the wetland area. The hydrogeological model at a regional level evidences how anthropic affectations taking place outside the wetland area can also affect its natural environment.