Hydrochemical-geophysical study of saline paleo-water contamination in alluvial aquifers

An integrated hydrochemical and geophysical study of the saline paleo-water uprising into the alluvial aquifer of the Oltrepò Pavese plain sector (Po Plain, northern Italy) is presented. This study involved hydrochemical analysis of groundwater, assessment of well logs, and one-, two- and three-dimensional electrical geophysical surveys. The studied area was selected for its characteristic hydrogeological setting. The alluvial aquifer is strongly conditioned by the presence of a buried tectonic discontinuity along which the saline waters are mainly distributed. These waters rise along the discontinuities in the bedrock and flow into the overlying alluvial aquifer. Contamination from saline waters is not spatially and vertically homogeneous within the aquifer. The spatial distribution of Na–Cl waters suggests the existence of plumes of highly mineralized waters that locally reach the aquifer, diffuse and mix with freshwaters. The saline waters show a dilution during upward migration, which is due to mixing with the shallow fresh groundwater. Highly mineralized groundwater is identified even at very shallow depth in correspondence with each plume. On the other hand, there is a lower degree of contamination in those sectors of the aquifer that are further away from the structural discontinuities and this lesser contamination generally only involves the deeper parts of the aquifer.


Introduction
Saline paleo-water uprising into aquifers is a hydrogeological process controlled by the geological and structural setting of the area, and the mechanisms are often not fully understood (Barberio et al. 2021;Conti et al. 2000;Grobe and Machel 2002;Petitta et al. 2011;Re and Zuppi 2011;Schwartz and Muehlenbachs 1979;Yechieli and Sivan 2010). While down-flowing fluids may be of meteoric origin, uprising fluids may connect the lowermost parts of an aquifer with the surface. An aquifer's contamination by saline water is caused by mixing of freshwaters with brinesexamples can be found in the European countries Belgium, Denmark, England (UK), Estonia, France, Italy and Spain (Darling et al. 1997;Desiderio and Rusi 2004;Dever et al. 2001;Hinsby et al. 2001;Bonnesen et al. 2009;Marandi and Vallner 2010;Petitta et al. 2011), and in Kansas and Texas in the USA (Stueber et al. 1998). Aquifer contamination can also be found where the fossil salt water, located far from the coastline, is the remainder of ancient marine ingressions, for example in Togo, Morocco and Vietnam (Akouvi et al. 2008;Bouchaou et al. 2009). Saline groundwater may result in environmental problems and difficulties in the exploitation and management of the aquifer, not only for drinking-water supply, but also for agricultural and industrial use. Contamination with salty water may not be spatially and vertically homogeneous within the aquifer. Intrusion and mixing of salt water with fresh groundwater of the shallow aquifer may result in different degrees of groundwater salinity within the aquifer.
Typically, an unfavorable distribution of sampling wells makes it necessary to integrate any hydrogeological and hydrochemical study with geophysical surveys (Di Sipio et al. 2006;Fadili et al. 2017;Jiraporn et al. 2020;Kouzana et al. 2010;Pilla et al. 2010;Pujari et al. 2012;Torrese and Pilla 2021). This allows definition of the general hydrogeological setting of the studied area, as well as detailed characterization of localized and restricted zones of the aquifer where the uprising phenomenon of deep saline waters occurs. This integrated and multidisciplinary approach is crucial for the understanding of this particular form of natural contamination which is often strictly controlled by the peculiar geological and structural configuration of the area and which can be influenced by several intervening factors. This paper presents an integrated hydrochemical and geophysical study of the saline paleo-water uprising into the alluvial aquifer of the Oltrepò Pavese plain sector (Po Plain, northern Italy). Here, the alluvial aquifer is strongly conditioned by the presence of the Vogherese Fault, a buried tectonic discontinuity along which the saline waters are mainly distributed , which is responsible for the uprising of deep, saline paleo-waters. This contamination prevents the exploitation of the aquifer, not only for drinking-water supply (Italian Legislative Decree, Decreto Legislativo 2 febbraio 2001, n. 31, Italian Legislature 2001), but also for agricultural (Wilcox 1948) and industrial uses.
The presence of these saline waters is known in the investigated area, as well as throughout the Po Plain area at the bottom of the Apennine mountain range front, the areas of the Po Plain that correspond to bedrock structural reliefs (Bonori et al. 2000;Toscani et al. 2007;Di Sipio et al. 2007), and the Central Apennine foredeep (Nanni and Zuppi 1986;Desiderio and Rusi 2004). Even so, their distribution and the mechanisms that control their uprising into the shallow aquifer (like withdrawals and changes of total head) are little understood. Torrese and Pilla (2021) investigated a test site of the Oltrepò Pavese plain sector; they found that the bedrock is affected by saline-water contamination in localized areas that are likely associated with structural discontinuities, which facilitate the flow towards the alluvial aquifer. They also found localized and restricted zones of salinization within the aquifer (Cameron et al. 2018).
The aims of this study are the chemical and geophysical characterization of the aquifer, definition of the extent of saline-water contamination, estimation of the depth to the freshwater/salt-water interface, monitoring of the variation in salinity over time, and the carrying out of a detailed and accurate investigation of localized zones of salinization within the aquifer.
The integration of hydrochemical and geophysical investigations aims to obtain further insights into the spatial distribution of saline-water contamination and the mechanisms underlying the upward migration of saline paleo-water, its intrusion and mixing with the fresh groundwater of the shallow aquifer.
This study included periodical sampling of groundwater to monitor the chemistry and temperature, vertical electrical conductivity and temperature logs to monitor the salt-water/ freshwater transition zone, and geophysical surveys. These involved both electromagnetic (EM) surveys over vast areas for a speedy assessment of subvertical conductive bodies connected to the uprising of high-salinity waters through structural discontinuities, and accurate one-dimensional (1D) to 3D electrical resistivity surveys for detailed investigation of the sectors where the uprising phenomena of deep saline waters occurs.

Chemical-physical characterization of groundwater
In case of high hydrochemical variability of groundwater, a hydrochemical survey (groundwater sampling from wells and chemical analysis) is typically used to distinguish different types of waters and their origin and evolution (Appelo and Postma 2005).
In coastal aquifers affected by seawater ingressions, in addition to conventional hydrochemical investigations, multiparameter well logs play a key role in the identification and characterization of the ingress of salt waters in the aquifers.
The systematic use of probes equipped with sensors for temperature (Polemio et al. 2009;Vandenbohede et al. 2014) and electrical conductivity measurements (Cotecchia et al. 1999;Smart and Worthington 2003;Di Sipio et al. 2006;Polemio et al. 2009;Tal et al. 2019;Shin and Hwang 2020) allows for the collection of significant and rapid data, albeit qualitative, on the chemical-physical characteristics of the groundwater.
During the data acquisition phase, particular attention must be paid to the descent velocity of the multiparameter probes along the water column of the wells. The descent velocity must be sufficiently low (<0.03 m/s) as to not disrupt the chemical stratification of the groundwaters (Cotecchia et al. 1999).
The Oltrepò Pavese plain sector is affected by salinewater uprising into the alluvial aquifer. This is a hydrogeological process that has common aspects with seawater ingression. For this reason, well logs were expected to be particularly useful in the Oltrepò Pavese plain sector to identify the vertical distribution of saline water into the aquifer. Moreover, the repetition of the measurements in the most significant wells during different periods of the year was expected to allow characterization of the variation over time in the distribution of saline water.

VLF-EM and electric resistivity methods
Electromagnetic and electrical resistivity methods are geophysical techniques well suited to the investigation and (partial) characterization of aquifers. The theory that underlies the very-low-frequency electromagnetic (VLF-EM) technique is well described in the literature (Paterson and Ronka 1971;Phillips and Richards 1975;Ramesh Babu et al. 2007). The VLF-EM technique is a passive method that uses radiation from worldwide ground-based military radio transmitters used for navigation operating in the VLF band (15-30 kHz) as the primary EM field.
VLF-EM has been typically used to prospect for conductive mineral deposits (Paal 1965) and for the identification of fracture zones for groundwater exploration (Jamal and Singh 2018), detection of fractures in the bedrock (Adepelumi et al. 2006), geological boundaries and shear zones/faults (Gnaneshwar et al. 2011;Parker 1980;Phillips and Richards 1975;Saydam 1981, Ramesh Babu et al. 2007Sundararajan et al. 2006), but it has been also applied to map groundwater conditions in sedimentary basins (Ohwoghere-Asuma et al. 2020) and to detect leachate plumes and groundwater pollution (Al-Tarazi et al. 2008).
Electrical resistivity methods allow characterization of subsurface materials in terms of their electrical properties. They involve the application of direct current into the ground by means of two current electrodes and measurement of the resulting voltage via two potential electrodes. The arrangement of current and potential electrodes during the measurement is dependent on the chosen electrode array. Each array provides different vertical and lateral resolution and depth of penetration (Kneisel 2006;Schrott and Sass 2008;Smith 1986;Szalai and Szarka 2008;Szalai et al. 2009). To obtain a true resistivity model of the subsurface, an inversion procedure is needed (Loke and Barker 1996). The three main methods of electrical resistivity surveys are resistivity depth sounding (RDS), resistivity profiling (RP), and electric imaging, commonly termed electrical resistivity tomography (ERT). ERT provides an image of the subsurface electrical resistivity pattern and allows for identification of subsurface structures. Its theory (Arato et al. 2014;Athanasiou et al. 2007;Buvat et al. 2013;Dahlin and Loke 1998;Daily and Owen 1991;Loke et al. 2003;Spiegel et al. 1980) and application (Cassiani et al. 2009;Daily and Ramirez 1992;Griffiths and Barker 1993;Guérin and Benderitter 1995;Guérin et al. 2004;Kuras et al. 2009;Ritz et al. 1999;Torrese 2020) are well-documented in geophysical literature.

Hydrogeological setting
The Oltrepò Pavese plain sector is geologically characterized by alluvial Quaternary deposits that cover Miocene-Pliocene marine deposits with very low hydraulic conductivity, formed by soils with a high clay content (clays, sandy clays, sandy-marls, sandstones, conglomerates, gypsum-rich marls and calcareous marls (Pellegrini and Vercesi 1995)).
The upper Quaternary deposits, deposited mainly by the action of the Po River and by the Apennine streams, represent the main water-bearing units of the area. Three different hydrogeological units can be defined within the Quaternary deposits: pre-Würmian alluvial deposits, middle-ancient alluvial deposits, and recent and present alluvial deposits (Cavanna et al. 1998;Pilla et al. 2007; Fig. 1). Only the latter two hydrogeological units are affected by the presence of sodium chloride (Na-Cl) rich waters.
The middle-ancient alluvial deposits occupy most of the Oltrepò Pavese plain sector and are formed by alternating sands and gravels, with interbedded clays or clayey silts. The recent and present alluvial deposits, distributed mainly along the Po River, were originated by the post-Würmian depositional activity of this river. The most important Apennine streams also contributed to the deposition of these alluvial deposits. Moreover, the continuous presence of a clayey silty covering, which has a varying maximum thickness of 10-15 m, in the sectors close to the Apennine margin, and a minimum of 2 m in the meandering area of the Po River, limits infiltration and influences the aquifer recharge that occurs in correspondence of the coalescent fans originating from the deposition of Apennine streams . A recharge contribution from the Po River must be excluded. In fact, groundwater flow direction (Fig. 2) is towards the River Po with the exception of occasional flooding events ). The aquifer is a single-bedded unconfined aquifer, although it becomes locally and temporarily (during some periods of the year) confined due to the presence of the clayey silty covering.
The structural setting of the Oltrepò Pavese plain is strongly conditioned by the presence of an important tectonic discontinuity, known in literature as the Vogherese Fault (Boni 1967) (Fig. 1). This fault, which is buried below a few tens of meters of alluvial deposits, has a NE-SW direction at a regional scale, from west of Casteggio to the Colle of S. Colombano (in the area of Pavia plain), passing across the confluence of the Ticino and Po rivers (Boni 1967). It is suggested that it is a normal fault with hanging wall to the NE, up to the Barbianello area. Here, it becomes an inverse fault to the south of Pinarolo, folding gradually towards the west, north of Casteggio and Voghera. Given the absence of specific recent studies on the subject, and since the definition of the fault type is outside the scope of this study, the Vogherese Fault has always been traced as a vertical fault on the cross-sections of this paper. The Vogherese Fault is responsible for the sudden deepening that affects the hydrogeological bedrock in the northern sector of the Oltrepò Pavese plain, and is also responsible for the strong variability on the thickness of the aquifer. The aquifer shows a thickness of a few meters in the southern sector, that represents the upper block, and hundreds of meters in the northern sector, the lower block (Braga and Cerro 1988;Cavanna et al. 1998;Regione Lombardia and ENI Divisione AGIP 2002;AGIP 1972; Fig. 1).
There are two deep boreholes (500 m) in the area drilled for oil exploration (AGIP 1972; Regione Lombardia and ENI Divisione AGIP 2002) (Fig. 1). The Casanova Lonati 1 borehole ( Fig. 1), which is located on the down-lifted block of the Vogherese Fault, intercepts the underlying marine deposits at a depth of 275 m. Brackish groundwater, which rises along  Cavanna et al. 1998). At the studied area, the alluvial aquifer is strongly conditioned by the presence of an important tectonic discontinuity, the Vogherese Fault, along which the saline waters are mainly distributed and which is responsible for the sudden deepening that affects the hydrogeological bedrock (sections). The uplift of the bedrock to the south-east of the Vogherese Fault (sections) is related to the Apennines uplift. Figure modified from Torrese and Pilla (2021) the fault and flows into the continental deposits, is intercepted by the borehole at 136 m. Groundwater intercepts salt waters at 400 m, as revealed by self-potential and resistivity logs, as well as formation testing. Formation tests found Na-Cl up to a concentration of 34.4 g/L in the bedrock at depths of several hundred meters. The second borehole, the Stradella 1 borehole ( Fig. 1), located on the up-lifted block of the fault, intercepts the underlying marine deposits at a depth of only 36 m.
This particular setting facilitates the phenomenon of uprising saline waters which appears most prominently in the southern sector of the plain where the aquifer is thinner. Here the Na-Cl-rich waters cannot be diluted by the more abundant calcium-bicarbonate  groundwaters. This outlined setting strongly influences the chemistry of the groundwater. The origin of the Na-Clrich waters is connected to the brines (very high-density fluids) that are remnants of evaporated marine waters in the late Messinian, trapped at the bottom of the Po plain aquifer (Conti et al. 2000;Regione Lombardia and ENI Divisione AGIP 2002). The uprising of saline waters is also facilitated by structural discontinuities localized in the bedrock of marine origin. These discontinuities represent preferential flow paths for the saline waters and facilitate the flow towards the alluvial aquifer.
The existence of these mineralised waters in the Oltrepò Pavese area has been well known since Roman times. In fact, the waters were (and still are) exploited for thermal purposes (S. Colombano al Lambro, Miradolo Terme, Salice Terme and Rivanazzano Terme are the most famous Spa centres located near the investigated area). The investigated area of the Oltrepò Pavese plain sector shows a fully flat topographical surface.

Hydrochemical investigations
The hydrogeochemical study included the periodical sampling and analysis of 146 wells drilled in alluvial deposits. These wells are mainly used for agricultural purposes and most are located close to the Vogherese Fault (Fig. 3). The maximum depth of the sampled wells typically varies between 10 and 20 m, all the wells are fully screened, and the well samples were taken at half of the maximum well depth.
The sampling was carried out between July 2007 and July 2013. Hydrogeochemical characteristics, groundwater levels, in situ temperature (T), electrical conductivity (EC), redox potential and pH were monitored. The variation in the degree of water salinity with depth and the freshwater/saltwater interface were also monitored through vertical logs of EC and T that were carried out from July 2008 on 63 wells.
Major ions were analyzed in the laboratory of Università di Pavia with a Dionex DX 120 chromatograph, while volumetric analysis was used for the determination of alkalinity. A Pasi BFK 100 hydrostatic probe was used for measuring piezometric levels, a WTW LF597 conductivity meter was used for acquiring electrical conductivity and temperature data, and a WTW pH 340/ION was used for acquiring pH Fig. 3 Investigations undertaken in the studied area and Redox potential data. Groundwater sampling from the wells was carried out with submersible electric pumps supplied with the well, a Cellai 504S peristaltic pump, or with a bailer.

Geophysical investigations
Geophysical investigations were undertaken in three separate phases, whereby the first two were preliminary investigations while the third was a more detailed investigation. During the first phase, 17 RDS's were undertaken along a cross section of the Vogherese Fault with the objective of reconstructing the geometry of the bedrock and the different hydrogeological units. These were undertaken in a subarea that is considered representative of the entire study area and were necessary for the calibration of the investigations for the following phases.
The second phase involved VLF-EM surveys carried out over vast areas for a rapid assessment of the distribution of saline waters within the aquifer, even in those areas where no wells for sampling are available. In all, 35 profiles ( Fig. 3) were undertaken over an area of approximately 150 km 2 along the Vogherese Fault, using a WADI instrument by ABEM™. Both north-south and northwest-southeast profiles were undertaken with a spatial sampling of 10 or 20 m.
The third, more detailed phase of the investigations, was carried out at test sites selected on the basis of the observations from the first two phases, and involved: • A resistivity profile, 2,160 m in length, crossing the Vogherese Fault in a N-S direction and overlapping with a VLF-EM profile (test site SR in Fig. 3 Fig. 3), with a surface grid 110 m × 30 m in size each, for an accurate investigation of localized brackish-water plumes contaminating the alluvial aquifer Each ERT profile is 470 m in length and was obtained using 48 electrodes spaced 10 m apart. Each 3D ERT involved a surface snake grid comprised of 12 × 4 electrodes spaced 10 m apart both along the X and Y axes.
A fully automatic multielectrode resistivity meter, SYS-CAL Jr. Switch-48 by IRIS Instruments (400 V max output voltage, 1200 mA max output current, 100 W max output power), was used for acquiring all electrical resistivity data.
Among the geophysical surveys carried out in the area, this paper reports the results obtained from some 2D and three-dimensional (3D) ERT surveys: a long transect (CC1-CC4 in Fig. 3) to define the hydrogeological setting of the area and detailed surveys (AA1, AA2, SG3D and SR3D in Fig. 3) to investigate localized zones of the aquifer affected by saline-water contaminations.
ERT data inversion was performed using ERTLab Solver (Release 1.3.1, by Geostudi Astier s.r.l. -Multi-Phase Technologies LLC) based on tetrahedral finite element modelling (FEM). Tetrahedral discretization was used in both forward and inverse modelling and in both 2D and 3D ERT's (even 2D models were obtained from 3D inversion). The foreground region was discretized using a 5-m cell size for all 2D and 3D ERT's, i.e., half the electrode spacing, to give the model high accuracy. The background region was discretized using an increasing element size towards the outside of the domain, according to the sequence: 1×, 1×, 2×, 4× and 8× the foreground element size.
The forward modelling was performed using mixed boundary conditions (Dirichlet-Neumann) and a tolerance (stop criterion) of 1.0E-7 for a symmetric successive overrelaxation conjugate gradient (SSORCG) iterative solver. Data inversion was based on a least-squares smoothness constrained approach . Noise was appropriately managed using a data-weighting algorithm (Morelli and LaBrecque 1996) that allows the variance matrix after each data point iteration that was poorly fitted by the model to be adaptively changed. The inverse modelling was performed using a maximum number of internal inverse preconditioned-conjugate-gradient (PCG) iterations of 5 and a tolerance (stop criterion) for inverse PCG iterations of 0.001. The amount of roughness from one iteration to the next was controlled to assess maximum layering: a low value of reweight constant (0.1) was set with the objective of generating maximum heterogeneity.
Inversion involved the application of homogeneous starting models with the average measured apparent resistivity. The final inverse resistivity models were chosen based on the minimum data residual (or misfit error).

Cross-validation of geophysical results with well logs
Vertical logs were carried out in shallow water wells, used for irrigational purposes, available at the site (Fig. 3). These wells were drilled prior to the present study, with destructive rotary or auger techniques until the top of the (impermeable) bedrock. All the wells are fully screened. Although no logs of chip samples are available and these wells do not provide accurate stratigraphic logs, they still provide information regarding the depth to bedrock below alluvial deposits, which was helpful in interpreting geophysical models. The paper shows a comparison between well logs carried out in well 30 and 3D geophysical results. This well, which is located in the southern sector with respect to the Vogherese Fault trace (Fig. 3), provided an ideal opportunity to cross-validate detailed geophysical results with ground truth. Well 30 intercepts the transition between slightly brackish groundwater and underlying moderate brackish groundwater at 11 m depth. This made it possible to define the representative resistivity range for different degrees of salinity of groundwater saturating the alluvial deposits. In this study, water salinity was classified according to these classes: freshwater with EC < 500 μS/cm, slightly brackish water with EC ranging between 500 and 4,000 μS/ cm, moderately brackish water with EC ranging between 4,000 and 8,000 μS/cm, highly brackish water with EC ranging between 8,000 and 12,000 μS/cm, salt water with EC ranging between 12,000 and 70,000 μS/cm, brine with EC > 70,000 μS/cm. The term 'saline water' is used to define a highly mineralized water and includes both brackish water and salt water.

Hydrochemical results
Two main hydrochemical facies can be identified within the Oltrepò Pavese groundwater: a Ca-HCO 3 hydrofacies, which characterises most of the groundwater of the Oltrepò Pavese alluvial aquifer; a Na-Cl hydrofacies, which locally characterises the groundwater in some sectors of the Oltrepò Pavese area along the Vogherese Fault (Fig. 5).
The Ca−HCO 3 hydrofacies has low-to-medium mineralization with electrical conductivity ranging between 800 and 1,200 μS/cm and chloride concentration between 2 and 200 mg/L. The Na-Cl waters have a high degree of mineralization with electrical conductivity that often exceeds 20,000 μS/cm and chloride concentration ranging between 200 and 12,000 mg/L. Na-Cl waters have an extremely variable degree of mineralization over the area (Figs. 5 and 6). This variability is associated with the different degrees of mixing between the shallower groundwater and the deeper saline waters (brines of the Po Plain). Evidence of this variability is represented at the surface with the thermo-mineralised springs at Salice Terme, Rivanazzano Terme and San Colombano Terme (Figs. 5 and 6). Undesirable metals, like iron, manganese, arsenic and selenium, represent another peculiarity of the Oltrepò Pavese groundwater ). These metals, which compromise the quality of the waters even more, frequently accompany the already high chloride and sodium concentrations, often above the Italian drinking-water standards -250 mg/L for chloride; 200 mg/L for sodium (Italian Legislative Decree, Decreto Legislativo 2 febbraio 2001, n. 31, Italian Legislature 2001).
The Ca-HCO 3 waters in the Oltrepò Pavese area are characterised by chloride values that do not exceed 200 mg/L; concentrations of bicarbonate vary between 300 and 700 mg/L, calcium varies between 50 and 200 mg/L, magnesium varies between 30 and 50 mg/L, and sulphates vary between 30 and 100 mg/L.
The Na-Cl groundwater has a significantly higher mineralization (EC values can be above 12,000 μS/cm and is on average between 2,000 and 5,000 μS/cm, Fig. 6). The EC values are related to the solubilised chloride and sodium, given that the concentrations of other major ions are relatively similar to those of the Ca-HCO 3 water described earlier. Well 15 at Barbaniello (Fig. 3) is the only exception.
The EC values recorded at this location, sometimes above 5,000 μS/cm, are mainly connected to the high sulphate concentrations that vary between 800 and 2,700 mg/L. The origin of this calcium-sulphate groundwater can be connected to evaporite layers within the Gessoso-Solfifera Formation which are present locally at the base of the Oltrepò Pavese aquifer .
Hydrochemical investigations have shown three areas along the Vogherese Fault where the phenomenon seems to be more intense and widespread: the area to the north of Casteggio where the chloride can exceed 10,000 mg/L; the area to the west of Barbianello where the highest concentrations (above 4,000 mg/L) of chloride were recorded, and finally, the sectors that includes Mezzanino and Albaredo Arnaboldi, where chloride concentrations can reach 3,000 mg/L (Fig. 4).
While the distribution of the Na-Cl groundwater is controlled by the trend of the Vogherese Fault at a regional scale (Fig. 4), higher variability in the distribution of the saline groundwater is observed at a local scale within the aquifer (Fig. 6). The chloride concentrations mentioned earlier are the highest measured concentrations for each of the areas. However, groundwater with lower concentrations (300-400 mg/L) was sampled in nearby wells, which is indicative of ongoing natural contamination, although with different intensities. The correlation between the chemistry of the aquifer and the trend of the Vogherese Fault is not always observed. Some wells, although close to the fault, show low mineralization of the groundwater (wells 32, 34 and 35).
The investigations that were undertaken for some sectors of the studied area allowed identification of the transition zone between the shallow fresh groundwater and the heavily mineralised deep groundwater (Fig. 7). In general, groundwater salinity starts to increase at depths of between 5 and 8 m, which can increase to depths of 10-20 m. In other cases, the depth is just below ground level, like in some subareas of the studied area where saline waters were found in the surface drainage network which is buried in the first few metres of the drift deposits . Acquired data show that groundwater above and below the transition zone does not show the same degree of mineralization. This indicates that the intrusion of saline waters within the aquifer is not evenly distributed within the study area (Fig. 8). A strong increase of salinity with depth (EC increased from 3,000 to approximately 14,000 μS/cm) was recorded in some wells (well 27), while salinity was less variable (within a few thousand μS/cm) along the vertical of other wells.
Neither the Ca-HCO 3 waters nor the Na-Cl waters show the same electrical conductivity within the different Fig. 6 Piper diagram, including electrical conductivity, from sampled waters areas of the plain. In some sectors, groundwater shows electrical conductivity values between 2,000 and 6,000 μS/ cm (wells 18, 26, 27, 36, 39, 41, 109, 113, 115, 120) even over the transition zone, in place of freshwaters. Into the transition zone, the electrical conductivity of waters varies between 2,500 and 10,000 μS/cm, depending on the area, while, below the transition zone, values between 3,000 and 23,000 μS/cm were detected-Table S1 in the electronic supplementary material (ESM).
This particular context is well illustrated by Fig. 8, which shows the distribution of wells with logs across the Vogherese Fault, their depth, and the maximum values of electrical conductivity ever measured for groundwater both at the top and at the bottom of wells. One can observe that the majority of wells characterized by the highest values of electrical conductivity (red ones) are located within a distance of 1.5 km in the southern sector and of 1 km in the northern sector from the Vogherese Fault. One can also notice the presence of a few contaminated wells very far from the fault (about 4 km), both in the northern and in the southern sector. They are probably connected to the existence of secondary tectonic discontinuities that convey deep salt-water into the alluvial aquifer. Moreover, into the most contaminated zone (<1 km), wells are not all contaminated by chloride, but some of them are characterized by low values of electrical conductivity (<1,500 μS/cm up-well; <2,000 μS/cm down-well).

Geophysical results
Phase 1 of the investigations (RDS) has shown that the hydrogeological bedrock at the base of the continental deposits is deeper in the northwestern areas and is also characterized by morphological irregularities (small troughs and/or peaks), partly shaped by tectonics and partly shaped by the Scuropasso and Versa paleo-rivers.
Phase 2 of the investigations (VLF-EM) has shown that a main NE-SW trend of high-conductivity anomalies was revealed at a large scale (Fig. 4). This trend can be correlated to the occurrence of the Vogherese Fault trace and therefore it can be correlated with the up-rise of mineralized waters along the fault zone. At least one secondary NE-SW alignment of high-conductivity anomalies Fig. 7 Electrical conductivity logs undertaken within some of the monitored wells is shown. Other NE-SW minor trends were also identified, suggesting the existence of secondary and subparallel discontinuities (Fig. 4). These results indicate that the uprising mineralised waters that originate from the Mio-Pliocene deposits contaminate the superficial aquifer in correspondence of several subparallel discontinuities oriented in a NW-SW direction.
The hydrogeological configuration revealed by phases 1 and 2 is confirmed by the more accurate and complete phase 3 of the investigations. In 2D ERT inverse resistivity models (Figs. 9 and 10), warm colours (from green to red) are associated with freshwater-saturated or brackish-water-saturated alluvial deposits (e.g., freshwater-saturated clayey-to-sandy deposits or brackish-water-saturated sandy deposits); cool colours (blue) are associated with clayey and silty cover deposits, saline-water-saturated alluvial deposits (e.g., saline-water-saturated sandy-to-gravelly deposits) or with the clayey bedrock (locally saline-water saturated). The inverse models indicate sharp and irregular contact between the alluvial aquifer and the underlying hydrogeological bedrock. The hydrogeological bedrock (3-6 Ω·m) at the base of the alluvial aquifer (20-40 Ω·m) is characterized by morphological irregularities, which are likely to have been shaped either by tectonics (Vogherese Fault zone) and/or by the paleo-river's erosion. In the southern area of the investigated transect (south of the fault) the thickness of the aquifer varies between 10 and 30 m (Figs. 9 and 10). This depth increases northward due to the Vogherese Fault ( Figs. 9 and 10). These evidences are consistent with well observations regarding bedrock depth. ERT results allowed for detailed localization and trace of the Vogherese Fault in the investigated area.
The complexity of the geometry of the areas where the high-salinity groundwater is found in proximity to the fault was also identified at the local scale within this phase of the investigations. The bedrock is affected by saline-water contamination which shows resistivity values lower than 5 Ω·m (Figs. 9 and 10). These areas of contamination are likely localized along structural discontinuities which represent preferential flow paths for the saline waters and facilitate the flow towards the alluvial aquifer.
The resistivity imaging revealed an extremely variable salinization within the aquifer (according to Torrese Low-resistivity anomalies (cool colors within the aquifer) are found in correspondence of saline-water contamination; high-resistivity anomalies (warm colors within the aquifer) are found in correspondence of contamination-free zones (Figs. 9 and 10). The variability of electrical resistivity with depth is associated with the different degrees of mixing between the shallower fresh groundwater and the deeper saline waters (brines of the Po Plain). No correlation was observed between the variation in the degree of salinity of the water and the piezometric level of the groundwater.
While the distribution of the Na-Cl groundwater is controlled by the trend of the Vogherese Fault at a regional scale (Fig. 4), higher variability in the distribution of the saline groundwater is observed at a local scale within the aquifer. Both the 2D and 3D ERT (Figs. 9, 10, 11 and 12)  Figs. 9 and 10). These zones show resistivity values ranging between 3 and 8 Ω·m. These contamination-related low-resistivity anomalies correspond to high-amplitude anomalies identified along the VLF-EM profiles.
Three-dimensional ERT's provided a detailed 3D imaging of the irregular-shaped shallow salt-water-to-brackish water plumes contaminating the alluvial aquifer (Figs. 11 and 12). Here, deep saline waters, which reach the alluvial aquifer during upward migration, diffuse and mix with the fresh groundwater of the shallow aquifer, therefore originating different degrees of groundwater salinity within the aquifer. Salt-water-to-brackish waters are associated with bulk resistivity values lower than 12 Ω·m in SG3D (Fig. 10) and lower than 9.3 Ω·m in SR3D (Fig. 12). Slightly brackish waters to freshwaters are associated with bulk resistivity values ranging between 30 and 78 Ω·m (Figs. 11 and 12).

Comparison between geophysical results and well logs
When analysing the resistivity log (EC log) undertaken on the 5th June 2013, which was used for the cross-validation of SR3D geophysical survey (Fig. 12), a transition can be found between slightly brackish groundwater (1,430 μS/cm on average, 1-10 m of depth, June 2013) and underlying moderate brackish groundwater (5,262 μS/cm on average, 11-15 m of depth, June 2013) at 11 m depth. Likewise, the temperature log (5 June 2013, Fig. 12) shows a transition between shallower relatively colder groundwater (average temperature of 12.8 °C, 3-10 m of depth, June 2013) and deeper relatively warmer groundwater (average temperature of 13.5 °C, 11-15 m of depth, June 2013) at 11 m depth.
The drop in resistivity (EC log) found at 11 m depth, along with an increase of temperature and a decrease of redox potential of groundwater, are well correlated to the presence of a salt-water-to-highly-brackish water plume (resistivity lower than 9.3 Ω·m) revealed by the geophysical model (Fig. 12a). This well transition zone occurs in a portion of the aquifer in which, based on the bulk resistivity indicated by the ERT survey, the aquifer should be characterized by clean sands deposits.
At shallower depths, the presence of slightly brackish to fresh groundwater is well correlated with higher resistivity values revealed by the geophysical model (Fig. 12b). Based on the bulk resistivity (resistivity higher than 30 Ω·m), the geophysical model suggests the presence of finer deposits saturated with slightly brackish groundwater. This suggests that the distribution of saline-water contamination areas is a hydraulic-conductivity-controlled process at the plume scale within alluvial deposits.

Discussion
The study enabled definition of the general hydrogeological setting of the investigated area, as well as detailed investigation of localized and restricted zones of the aquifer which are crucial elements for the understanding of the saline-water contamination process and the exploitation management of the aquifer.
Sharp and irregular contact between the alluvial aquifer and the underlying bedrock was observed. The bedrock is characterized by morphological irregularities, which are likely to have been shaped either by tectonics (Vogherese Fault zone) and/or by the paleo-river's erosion. Although, in the southern investigated area (south of the fault) the thickness of the aquifer varies between 10 and 30 m, the bedrock depth increases northward due to the Vogherese Fault. These evidences are consistent with well observations regarding Fig. 11 Three-dimensional ERT model (SG) which shows a saltwater-to-brackish water plume (3-12 Ω·m) contaminating the alluvial aquifer (12-47 Ω·m): deep saline waters reach the alluvial aquifer during upward migration, then diffuse and mix with the fresh groundwater of the shallow aquifer bedrock depth. Geophysical results enabled detailed localization and tracing of the Vogherese Fault at the scale of the investigated area. Other minor subparallel structural discontinuities (faults and fractures) were also identified. These minor discontinuities show geometries and directions (NW-SW) that are coherent with those of the Vogherese Fault and are therefore genetically connected to the major structural discontinuity.
The bedrock is affected by the presence of Na-Cl paleowaters which are likely localized along structural discontinuities which represent preferential flow paths for the saline waters and facilitate the flow towards the alluvial aquifer. The contamination is strictly controlled by the geological and structural configuration of the area. The morphology of the tertiary bedrock and the spatial distribution of the structural discontinuities are likely to be partially controlling the distribution of salty groundwater, originating at depth, within the alluvial aquifer.
The spatial distribution of Na-Cl waters suggests the existence of plumes of highly mineralized waters that locally reach the aquifer, diffuse and mix with freshwaters. Detailed 3D imaging revealed irregular-shaped shallow saline-water contamination areas within the alluvial aquifer.
Overall, the salt-water-saturated clayey bedrock shows resistivity values lower than 3 Ω·m, while salt-water-saturated sandy alluvial deposits show resistivity values ranging between 3 and 8 Ω·m; brackish-water saturated sandy alluvial deposits show resistivity values ranging between 8 and 12 Ω·m; slightly brackish waters to freshwaters are associated with bulk resistivity values ranging between 30 and 78 Ω·m. The interpretation of these resistivity ranges is consistent with findings from Torrese and Pilla (2021) who calibrated resistivity surveys with well logs in the same area investigated in this study. Even if these ranges of values can be considered representative for such a geological setting, it is worth underlining that the resistivity of such hydrogeological bodies does not depend on the electrical conductivity of the fluid only, but also on the porosity and clay content of the solid material. Moreover, the resistivity signature depends even on the size of the body in relation to its depth and on the contrast between the resistivity of the body and that of the surrounding rock. This is the reason why the same hydrogeological body can show slightly different resistivity values even in the same site.
Contamination from saline waters is not spatially and vertically homogeneous within the aquifer. This Fig. 12 Three-dimensional ERT model (SR) cross-validated with well logs, which shows a salt-water-to-highly-brackish water plume contaminating the alluvial aquifer: a salt-water-to-highly-brackish water plume contaminating a sandy body within the alluvial aquifer and shallow low-resistivity anomalies within the upper clayey deposits; b slightly brackish-to-freshwater contaminating clayey sandy deposits. The resistivity model shows good correlation with respect to well 30 logs which found a drop in resistivity at 11 m depth, along with an increase of temperature and a decrease of redox potential of groundwater, suggesting a well transition zone between slightly brackish to moderate brackish water nonhomogeneity is likely to be affected by different factors like the aperture of the discontinuities within the hydrogeological bedrock, and the hydraulic conductivity of the aquifer, as well as seasonal variations in terms of freshwater recharge and groundwater pumping. The observed variability of salinity and depth of the transition zone can be attributed to the distance of the monitored wells from the center of the plume.
Salt water intrusion into the alluvial aquifer has a heterogeneous distribution, and suggests the existence of plumes of highly mineralized waters that locally reach the aquifer, diffuse and mix with freshwaters. Deep saline paleo-waters show a dilution during upward migration. The hydrodynamic flow of shallow fresh groundwater would tend to relegate the highly mineralized groundwater to the lower sections of the aquifer. This interpretation can explain the different degrees of groundwater salinity detected in the course of the investigations and also the different characteristics of the transition zone into the same area. In fact, wells located on a plume show higher values of water electrical conductivity and a shallower depth of the freshwater/salt-water interface. The distribution of the plumes is influenced mostly by the presence of tectonic discontinuities into the marine substratum but, probably, also by the local hydraulic permeability of the alluvial sediments. There is a lower degree of contamination in those sectors of the aquifer that are further away from the structural discontinuities and this lower degree of contamination generally only involves the deeper parts of the aquifer.
The overall simplified hydrogeological conceptual model of the investigated area is shown in Fig. 13. This model represents the main features of the shallow aquifer as revealed by hydrochemical and geophysical investigations. It shows the plumes of high-salinity groundwater that reach the alluvial aquifer. Here, they diffuse and mix with the fresh groundwater of the shallow aquifer, thus originating different degrees of groundwater salinity within the aquifer. Highly mineralized groundwater is identified even at very shallow depth in correspondence of each plume, which is located above a structural discontinuity.  (2021) A detailed simplified hydrogeological conceptual model of the investigated area is shown in Fig. 14. This model shows that the contamination from saline waters is not spatially and vertically homogeneous even at the scale of the well field. The wells located on a plume show higher values of water electrical conductivity and a shallower depth of the freshwater/salt-water interface. Moreover, the hydrodynamic flow of shallow fresh groundwater would tend to relegate the highly mineralized groundwater to the lower sections of the aquifer. However, it cannot categorically be ruled out that in the deeper portions of the alluvial aquifer, the development of the saline plumes is also conditioned by differences in hydraulic head between the deep brine reservoir and the shallow alluvial aquifer and by local variations in hydraulic conductivity within the alluvial deposits.
The study shows that it is necessary to use a multidisciplinary approach (which includes the integration of hydrochemical and geophysical investigations within the hydrogeological assessment), when it comes to understanding extremely complex forms of groundwater natural contamination, similar to the contamination that has been identified in the Oltrepò Pavese area, where several intervening factors can influence the contamination.
Future studies may be based on the correlation between the temporal variability of groundwater salinity and the variability in the discharge of uprising salt water. These studies would allow one to verify if the uprise of salt water is likely to be induced by an increase in hydraulic head within the main Apennine groundwater following rainfall or snow melting events. This system could be in hydraulic connection with a deeper aquifer that host saline waters. The pressure transfer, through a piston flow mechanism Re and Zuppi 2011), may produce a mass transfer where saline waters are forced to rise along discontinuities and reach the shallow aquifer. Forthcoming studies based on continuous piezometric and hydrochemical monitoring of the alluvial aquifer groundwater correlated to rainfall or snow melting events in the nearby Apennines will allow one to verify the piston-flow-mechanism-based hypothesis.

Conclusions
This paper presents an integrated hydrochemical and geophysical study of saline paleo-water uprising into the alluvial aquifer of the Oltrepò Pavese plain sector (Po Plain, northern Italy). This study involved periodical sampling of groundwater, vertical electrical conductivity and temperature logs, and geophysical surveys. These involved both EM surveys undertaken over vast areas for a speedy assessment of subvertical conductive bodies connected to the uprising of high-salinity waters through structural discontinuities and more accurate 1D-3D electric resistivity surveys for detailed investigation of the sectors where the uprising phenomena of deep saline waters occurs.
In this area, the alluvial aquifer is strongly conditioned by the presence of the Vogherese Fault, a buried tectonic discontinuity along which the saline waters are mainly distributed. These Na-Cl-rich waters rise along the discontinuities in the hydrogeological bedrock and flow into the overlying alluvial aquifer. This particular setting conditions the distribution of saline waters into the alluvial aquifer.
Contamination from saline waters is not spatially and vertically homogeneous within the aquifer. The spatial distribution of Na-Cl waters suggests the existence of plumes of highly mineralized waters that locally reach the aquifer, diffuse and mix with freshwaters. Detailed 3D imaging revealed irregular-shaped shallow saline-water contamination within the alluvial aquifer.
Deep saline paleo-waters show a dilution during upward migration, which is due to the mixing with shallow fresh groundwater. Highly mineralized groundwater is identified even at very shallow depth in correspondence of each plume, and is located above a structural discontinuity. On the other hand, there is a lower degree of contamination in those sectors of the aquifer that are further away from the structural discontinuities and generally only involves the deeper parts of the aquifer.
Forthcoming studies, based on continuous piezometric and hydrochemical monitoring of the alluvial aquifer groundwater correlated to rainfall or snow melting events in the nearby Apennine mountain range, will allow one to verify if the uprise of salt water is to be induced by an increase in hydraulic head within the main Apennine groundwater. The pressure transfer, through a piston flow mechanism, could produce a mass transfer where saline waters are forced to rise along discontinuities and reach the shallow aquifer.
The results from this study are applicable in similar hydrogeological contexts where the aquifer's contamination by saline water is caused by mixing of freshwaters with brines or where the fossil salt water, located several kilometers from the coastline, are the remainder of ancient marine ingressions.