Explaining the groundwater salinity of hard-rock aquifers in semi-arid hinterlands using a multidisciplinary approach

ABSTRACT Shallow crystalline groundwater in the semi-arid hinterland of Ceará is brackish or saline with mixed chloride or sodium chloride facies. Very few hydrochemical data are available for the area and the drivers behind this salinity are not clearly identified. In this study, extensive field data collection was performed to provide new information about the hydrogeological functioning and the salinization processes, through the implementation of piezometric, hydrogeochemical, isotopic (18O, 2H) and multitracer dating (14C, 3H, CFC, SF6) monitoring. Piezometric and isotopic data evidence fast flow circulation processes and a high contribution of evaporated surface water to aquifer recharge. Multitracer dating shows the groundwater is essentially composed of seasonal vertical infiltration flows that mix with older waters stored in the aquifer. Chemical analyses suggest the groundwater, originally low mineralized, has become progressively saltier due to leaching of salts that were evapoconcentrated in either surface waters or the unsaturated zone during drier periods.


Introduction
Groundwater resources in semi-arid and hardrock areas may be subject to salinity issues that deteriorate water quality for drinking water supply or agricultural activities (Rabemanana et al. 2005, Basavarajappa and Manjunatha 2015, Sreedevi et al. 2021, Banyikwa 2023).Understanding the processes behind the groundwater salinization is crucial to develop sustainable water resource management.In the semi-arid area of northeastern Brazil (NEB), groundwater from crystalline rocks represents a strategic water resource for rural populations (MMA 2007), especially during drought periods.Indeed, this region is frequently affected by intense drought events, and rainfall occurs during a short period of the year with a highly irregular distribution in space and over time (Pontes Filho et al. 2020, Ribeiro Neto et al. 2022).This great climate variability and the consequent vulnerability to drought have led to the construction of a dense network of small and mid-/large-sized surface water reservoirs which constitute the primary source of water supply in the state of Ceará.However, demand pressures added to prolonged droughts have boosted the exploitation of the groundwater contained in the crystalline rocks (Da Silva et al. 2007).
The presence of aquifers in hard rocks is related to the degree of weathering and fracturing of the rock, considering their very low primary porosity and permeability (Wyns et al. 2004, Lachassagne et al. 2011).These chemical and physical weathering processes lead to the in situ development of saprolite (unconsolidated alterite) and of a sub-horizontal permeable fractured layer several tens of metres thick below the saprolite, due to the stress generated by the swelling of certain minerals such as biotite (Wyns et al. 2004, Lachassagne et al. 2021).The fractured layer is characterized in the first few metres by a dense fracturing decreasing with depth, down to the base of the weathering profile, which results in a decrease in hydraulic conductivity and connectivity between fractures with depth (Lachassagne et al. 2021).Even if plutonic and metamorphic rocks have a low aquifer potential, they represent a valuable water resource in comparison with the scarce surface water.This is why many rural communities in Ceará had no alternative but to drill wells in the crystalline rocks during the last severe drought from 2012 to 2016.
Groundwater from the Ceará crystalline aquifers is characterized by a high and heterogeneous salinity, mostly above the drinking water limit and up to several tens of g L −1 (Burte 2008, Araújo 2017).The origin of the salinity and the salinization processes of groundwater in semi-arid NEB have been sporadically studied since the 1960s, and the mechanisms behind the groundwater salinity are still debated (Matsui 1978, Frischkorn and Santiago 2000, Silva 2003, Britto Costa et al. 2006, Araújo 2017).Paradoxically, these waters present short residence times (Salati et al. 1979a, Santiago et al. 2000), implying that the strong salinity observed cannot be explained by a long duration of water/rock interaction.In addition, chloride is often the prevailing anion, which is not consistent with water-rock interactions in crystalline rocks (Shand et al. 2007, Almeida et al. 2008).The crystalline aquifers of Ceará have been little studied compared with sedimentary or alluvial aquifers, which represent more important water resources in terms of water storage capacity.Consequently, there are very few data available for characterizing the hydrogeological functioning of such aquifers (Fundação Cearense de Meteorologia e Recursos Hídricos [FUNCEME] 2007).
This multidisciplinary study is intended to provide new interpretation elements regarding the Ceará fractured crystalline aquifers through an extensive field data collection work and the simultaneous acquisition of piezometric, hydrochemical and isotopic data and multi-tracer dating ( 18 O, 2 H, 3 H, 14 C, CFC, SF 6 ).Piezometric variations were used to observe the aquifer reactivity and to estimate recharge through the use of the water table fluctuation (WTF) method (Healy and Cook 2002).Analyses of the water stable isotopes ( 18 O, 2 H) in the different compartments of the water cycle (rain, surface water and groundwater) were performed to identify groundwater recharge processes and the origin and organization of groundwater flows within aquifers (Ahmed et al. 2019, Babaye et al. 2019).Water residence times were determined using 14 C and 3 H radioactive tracers, and refined with the first-ever CFC (Chlorofluorocarbons) and SF 6 analyses performed in the Ceará state.Water quality data were evaluated using Piper diagrams, chemical ratios including the Cl − /Br − molar ratio, and principal component analysis (PCA) to identify groundwater facies, the potential sources of ions and the dominant geochemical processes.
The objectives of this study were to (1) assess the hydrodynamics and hydrogeological functioning of the crystalline aquifers and (2) assess the origin of the groundwater salinity, in order to contribute to the knowledge about groundwater salinity in semi-arid hard-rock areas.

Geography and land uses
The study area is included in the Banabuiú watershed, whose eponymous river drains, from west to east, an area of 19 647 km 2 .It is located in the centre of the Ceará state, approximately 180 km from Fortaleza (capital city) and the Atlantic coast (Fig. 1).The hydrogeological investigations were carried out in four sub-basins of the Banabuiú watershed: Forquilha (214 km 2 ), Pirabibú (127 km 2 ), and Vista Alegre (550 km 2 ), all three located in the municipality of Quixeramobim; and Ibicuitinga (286 km 2 ), mainly located in the municipality of Ibicuitinga (Fig. 1).These are rural areas with low human density (about 25 inhab.km−2 ), inserted in the Caatinga biome, whose socioeconomic development is mainly based on livestock (extensive farming) and agriculture (MME 1998).

Climate
The semi-arid climate has been relatively stable in the study area over the last 13 Ma (Peulvast et al. 2008).It is characterized by a strong water deficit due to a low rainfall regime (< 800 mm a −1 ), a high potential evapotranspiration (PET) rate (> 1700 mm a −1 ), an irregular rainfall distribution and recurrent drought periods (Martins andVasconcelos Júnior 2017, Pontes Filho et al. 2020).The mean annual rainfall is about 700 ± 275 mm in the study area (1912-2021 period) with minimum and maximum values of 154 and 1441 mm, respectively.The wet season extends from December to July, and 75% of the annual precipitation falls between February and May.The four rainiest months, locally called "quadrachuvosa," result from the movement of the Intertropical Convergence Zone (ITCZ), which is formed by the confluence of the northeast and southeast trade winds, and is characterized by strong upward air movements and heavy rains in the north of NEB (Da Silva et al. 2017).Consequently, the rain generally falls intensely in the region during this period.
The monthly PET is always higher than the monthly precipitation, except in March and April (Supplementary material Fig. S1).As a result, these two months are theoretically the most likely to produce the annual runoff and groundwater recharge.However, the amount of rainfall contributing to runoff or recharge cannot be directly deduced from a monthly water balance in such a climatic context; taking into account the daily scale is required.The region regularly suffers from droughts lasting several years.During the last prolonged drought (2012-2019), the 2012-2016 period was the most severe 5-year drought ever recorded in the area (with a 40% decrease in annual rainfall).The average annual temperature is 27.2°C, with monthly mean extrema of 25.7°C and 28.5°C (Supplementary material Fig. S1).Monthly relative humidity varies between 41.6% (July) and 85.9% (February), with an annual average of 61.0%.

Geology and geomorphology
The crystalline rocks constitute 96.5% of the Banabuiú watershed area.They consist of various types of Precambrian gneisses and migmatites (3.2 Ga to 541 Ma), associated with plutonic and metaplutonic rocks of predominantly granitic composition (INESP 2009).The various lithostratigraphic units have a NE-SW orientation (Supplementary material Fig. S2).The remaining 3.5% are composed of Tertiary-Quaternary sediments that appear in scattered patches and are interpreted as fluvial-lacustrine deposits (Da Silva et al. 2003), but also of alluvial and colluvial-eluvial Quaternary sediments deposited along watercourses in low-slope areas.
Elevation of the Banabuiú watershed ranges from 1154 m asl (W) to 30 m asl (E).The landscape is dominated by hinterland depressions (78% of the basin area, called Sertaneja Depression and represented by a flat and slightly undulating relief.The Sertaneja Depression was formed from a large pediplanation process that produced the current surface of the Precambrian rocks, corresponding to the lower parts of the basin (< 300-400 m asl).In opposition, elevations above 300-400 m asl correspond to the reliefs formed by the residual massifs (inselbergs, mountains or plateaus), evidence of the action of weathering and erosion over millions of years.

Hydrological and hydrogeological setting
In the crystalline zones of the study area, surface runoff is of Hortonian type due to the low thickness (<1 m) and low permeability of soils (Burte 2008).All rivers in Ceará are naturally intermittent (Santiago et al. 2001).Nascimento (2016) estimated that 90% of the surface water flows during the "quadra-chuvosa," between February and May, although with a great inter-annual variability.River discharge is generally intense, short and irregular (Burte 2008).River recession time is short (Santiago et al. 2001).The intermittence of water courses and the recurrence of drought pressed public policymakers to promote the construction of dams to guarantee the water supply during the periods of water stress (Santiago et al. 2001, Burte 2008).As a result, the Banabuiú watershed hosts numerous dams, with 17 299 reservoirs over 20 m in length being registered by the Research Institute of Meteorology and Water Resources -FUNCEME (Freitas Filho and Carvalho, 2021).Among them, 1415 reservoirs have an area greater than 5 ha (INESP 2009).In addition to these dams, temporary ponds of various size form in the topographic lowlands when it rains (Araújo 2017).These reservoirs are subject to high evaporation rates and, consequently, to significant issues of salinization and eutrophication.
In the semi-arid NEB, the saprolite is generally thin or nonexistent (MMA 2007) because of erosion.This allows a direct connection between the surface drainage system and the fractured layer.In the study area, the median thickness of the saprolite layer is about 9 ± 6 m, while the underlying fractured layer is about 50 m thick (Kreis 2021).In Ceará State, the median depth of boreholes is about 60 m and the piezometric level is generally met at a depth of 10 m (Da Silva et al. 2007), which implies that groundwater is mainly present in the fractured layer.The boreholes in the crystalline aquifers in NEB are characterized by low discharges (< 1 to 3 m 3 h −1 ) and specific capacities of less than 1 m 3 h −1 m −1 (Asomaning 1992, MMA 2007, Feitosa 2008).Transmissivity is low, between 1.10 −6 and 7.10 −5 m 2 s −1 (Manoel Filho 1996).As in most crystalline aquifers (Lachassagne et al. 2021), there is no regional flow (Santiago et al. 2000), which means that the natural outflow from these aquifers feeds the rivers or alluvial aquifers, at the scale of small local sub-basins, when evapotranspiration does not recover all flows (Burte 2008).In Ceará, in addition to these groundwater flows from crystalline aquifers to the alluvial ones, recharge of alluvial aquifers mostly comes from diffuse infiltration of rainfall and localized infiltration from the rivers (FUNCEME 2007).The reversal of flow between the rivers and the alluvial aquifers, which prolongs their flow, can be observed locally at the end of the rainy season (FUNCEME 2007).

Materials and methods
Comprehension and characterization of the hydrogeological and hydrochemical functioning of complex hydrosystems require a multidisciplinary approach.Each method provides clues that, when they converge, combine to efficiently deepen the knowledge of the system (e.g.Maréchal et al. 2014).For this study, investigations were developed along two main axes.On one hand, we studied the subterranean hydrodynamics based on the implementation of a piezometric monitoring network and the realization of a multi-tracer sampling programme ( 18 O, 2 H, 3 H, 14 C, CFC, SF 6 ).On the other hand, we studied the hydrochemical composition and evolution of the groundwater through monitoring of electrical conductivity (EC) and the analysis of major, minor and trace ions.Results of the first axis allow us to characterize groundwater recharge, circulation processes and water residence times, while results of the second axis provide information about the types of salts involved and their evolution over time.All these elements give insights into the processes that lead to groundwater salinization and salinity dynamics.

Characterization and quantification of aquifer recharge from piezometric monitoring
Monthly and hourly piezometric data were collected from March 2018 to December 2019 over a network of 56 boreholes (monthly monitoring), and from November 2016 to December 2019 in four of these boreholes (hourly monitoring).Existing piezometric data were also exploited, including 93 measurements generally taken quarterly (monthly for the borehole FOR-Y22) between 2010 and 2016 in seven of the boreholes used in the 2018-2019 monthly monitoring, and 73 sporadic piezometric measurements carried out between August 2016 and March 2018 in the 56 boreholes of the monthly monitoring.Monthly piezometric measurements were taken manually, while hourly piezometric measurements were performed with Rugged TROLL®100 pressure sensors (absolute pressure measurement) and a Rugged TROLL® barometer, branded Aquatroll, from Insitu Inc.The pressure measurement accuracy is ± 0.1% of the full scale (± 3 cm).The locations of the monitored boreholes are shown in Fig. 1.
Piezometric data were collected to characterize the reactivity of the aquifer in response to rainfall and to estimate the aquifer recharge through the use of the WTF method.The WTF method was applied as an exploratory approach.It requires that the specific yield (Sy) is known and that the piezometric level rise is only due to the vertical infiltration of the recharge through the unsaturated zone (USZ).The recharge rate can be inferred from the seasonal water fluctuation: where R is the recharge rate, ΔWL the variation of the piezometric level, Δt the time period and Sy the specific yield (Healy and Cook 2002).In the absence of any direct measurement of Sy in the study area, we used values from the literature, as discussed later.

Characterization of recharge processes and surfacegroundwater relationships from water stable isotopes ( 18 O, 2 H)
Stable isotope (δ 18 O, δ 2 H) analyses were performed on: (1) Rainwater, sampled monthly between January and July (the period concentrating 95% of the annual precipita- The location of the raingauges is given in the Supplementary material (Table S1) and the location of the sampled borewells and wells is plotted in Fig. 2. Groundwater samples were collected after 30 min of pumping or after the stabilization of the physicochemical parameters (pH, temperature and EC) to ensure good representativeness.Daily precipitation of less than 1 mm day −1 was not sampled.All samples were packaged in 20 mL transparent plastic bottles, hermetically closed with an inner cap and a plastic lid, and stored in a dark refrigerated container.Analyses were performed at the HydroSciences Montpellier (HSM, France) water stable isotope laboratory (LAMA).Results are given in delta permil units (δ, ‰) vs Vienna Standard Mean Ocean Water (V-SMOW), with an analytical error of ± 0.08‰ for δ 18 O and ± 0.8‰ for δ 2 H.

Groundwater multi-tracer dating ( 14 C, 3 H, CFC and SF 6 )
Groundwater residence times were obtained from the sampling of three borewells in November 2009 and seven borewells in February 2018 for 14 C and 3 H analyses.The 14 C analyses were completed with a δ 13 C analysis to model the water's radiocarbon apparent age by differentiating the biogenic sources of carbon.This dating was complemented by analyses of CFC and SF 6 collected from 10 borewells in June 2019, whose results represent the first CFC and SF 6 measurements ever carried out in Ceará.Unfortunately, five boreholes sampled for 14 C and 3 H measurements were no longer available to be sampled again for the CFC and SF 6 analysis in 2019; they were replaced with other boreholes.The groundwater δ 13 C, 14 C and 3 H analyses were carried out at the Avignon University (France) Mediterranean Environment and Agrohydrosystems modelling (EMMAH) laboratory.The 14 C results are expressed as a percentage of modern carbon (pMC), with pMC = 100% corresponding to post-1950s infiltrated waters.The analytical error is about 0.5 pMC.The δ 13 C results are expressed in delta permil units (δ, ‰) vs Vienna Peedee Belemnite (V-PDB), with an analytical error of ± 0.1‰.The 3 H results are expressed in tritium units (TU), with an analytical error which varies from 0.3 to 0.6 TU.
Analyses of CFC-11, CFC-12, CFC-113 and SF 6 water concentrations were performed at the Condate Eau -Earth Sciences and Astronomy Observatory (OSUR) analytical platform of the University of Rennes (France).Analyses are within 3% accuracy for CFCs, and within 5% for SF 6 .The location of the sampled borewells is given in Fig. 2.

Hydrochemical analyses (EC, major ions, minor ions and trace elements)
EC monitoring was conducted simultaneously with the monthly piezometric monitoring to evaluate the dynamics of groundwater salinity over time.Major ion (Ca 2+ , Mg 2+ , Na + , K + , HCO 3 ) sampling campaigns were carried out in December 2017 (end of the dry season) and June 2018 (end of the wet season) in 27 and 38 borewells, respectively.Major ion analyses were performed at the Ceará Water and Sewerage Company (CAGECE) (Fortaleza, Brazil) Product Quality Control Management (GECOQ) laboratory laboratory, in accordance with methodologies recommended by the American Public Health Association (APHA, 2012).To improve the understanding of the geochemical sources at the origin of the groundwater salinity, a sampling campaign for major ions (Ca 2+ , Mg 2+ , Na + , K + , Cl − ), minor ions (Br − ), and trace elements (Li, B, Al, Si, P, S, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, As, Se, Rb, Sr, Y, Mo, Ag, Cd, Sn, Sb, Cs, Ba, La, Ce, Pr, Nd, Sm, Eu, Tb, Gd, Dy, Ho, Er, Tm, Yb, Lu, Tl, Pb, Th, U) was implemented on 20 borewells in June 2019.Samples were analysed at the Environmental Trace Element & Isotope Analysis (AETE-ISO) platform of Hydrosciences Montpellier (HSM) platform (Montpellier, France).
Piper diagrams were used to determine the hydrogeochemical facies and the effects of the seasonal recharge on the ionic composition, while the ionic ratios (binary plot investigations) were calculated to identify potential sources of salts.Cl − /Br − molar ratio was used as an indicator of the source of the chloride ions (Cartwright et al. 2006, Alcalá andCustodio 2008).The PCA method was used on chemical data to strengthen the identification of the processes controlling groundwater chemistry.PCA statistically quantifies the relationship between variables and the contribution and significance of each variable.It is a multivariate statistical analysis method that transforms correlated variables into uncorrelated principal components (PC) that represent the directions of maximum variance in the data.The projection of the PC on factorial axes allows for the identification of patterns and relationships between the variables by defining which variables contribute the most to the variance in the data.The goal of this technique is to summarize the maximum amount of information contained in the mass of data on such factorial axes.

Evolution of piezometric levels and recharge rate assessment
Piezometric levels in the crystalline aquifers are relatively shallow, between 0.4 and 21.4 m below the ground surface, with a mean value of around 10 m.Their contrasting temporal variations confirm that there is no regional flow.The severe 5-year drought of 2012-2016, marked by a 40% decrease in the annual rainfall, caused a general lowering of the piezometric level, by up to 7 m (Fig. 3).Even during these dry years, the monthly monitoring of the borewell FOR-Y22 highlighted a recharge of the crystalline aquifer, except in 2016.The rainier conditions between 2017 and 2019 (annual rainfall only 10% below the 1912-2021 average) allowed the water stock to be reconstituted.This underlines the dependence of recharge on climatic fluctuations, a dependence that is common in such low Sy aquifers (Dewandel et al. 2010).
Generally, the seasonal recharge of the crystalline aquifers occurs between February and May ("quadra-chuvosa" period), and mostly in March and April.The hourly monitoring carried out between November 2016 and December 2019 (Fig. 4) highlighted that recharge does not directly depend on the annual or monthly rainfall amounts.Moreover, a gradual increase of the groundwater level is observed throughout the rainy season (Fig. 4), even if a strong reaction of the piezometric level is observed in less than 24 h after significant rain events (> 50-60 mm).However, a decrease in the piezometric level is almost never observed following such events during the rainy season.The seasonal recharge depends on the distribution and quantity of daily rainfall and is favoured by the storage of surface runoff in ponds, as observed in other semi-arid areas (Leduc et al. 1998).During the dry period, water levels decrease due to the absence of recharge, while the aquifer discharge continues through the river baseflow or evapotranspiration processes.

Assessment of aquifer recharge rates
In semi-arid areas, the determination of the amount of recharge using conventional water balance and/or hydraulic methods is tricky (Ahmed et al. 2019).Moreover, the use of the monthly scale for water balance calculations is not adapted to semi-arid regions with highly variable and often intense rainfall events.Thus, water balance calculations based on monthly data, as attempted by Araújo (2017) for example (yielding maximum recharge rates of 12% of the annual rainfall), are not realistic.Furthermore, a maximum soil retention capacity of 100 mm is generally used for this type of calculation, and this value is likely to be overestimated: aquifer recharge can occur with a monthly rainfall of about 60 mm, as observed for instance in the PIR-P14 borehole in January 2018 (Fig. 4).In this study, recharge rate estimates were then approached using the WTF method and based on a multi-event time series analysis (Healy and Cook 2002, Crosbie et al. 2005, Maréchal et al. 2006).It is used here only as an exploratory approach, considering that the Sy data are based on the bibliography and that the hourly piezometric monitoring suggests the influence of a sub-horizontal flow coming from localized infiltration zones such as ponds and reservoirs.The assessment of the aquifer recharge was performed from the hourly piezometric monitoring of the boreholes PIR-P14 and IBI-P24.Only the rainfall that fell during the period when the piezometric level rose was considered.Manoel Filho (1996) is the only author who provided estimates of the Sy in Ceará, from 24 boreholes, with values varying from 0.0022% to 0.28%.Unfortunately, the Sy values were calculated from single-well production tests, without any distant piezometer, which makes these results unrepresentative.The choice of 0.3% for Sy was based on the existing literature about the fractured layer in similar crystalline aquifers from other regions of the world (Dewandel et al. 2010, Mizan et al. 2019) and in coherence with the recharge rate determined with the 3 H tracer (Kreis 2021).With a Sy of 0.3%, the recharge of the crystalline aquifer is around 1 to 3% of the annual rainfall (i.e.around 8 to 22 mm a −1 ), as detailed in Supplementary material Table S2.These results are consistent with those from other semi-arid regions of the world: for Scanlon et al. (2006) in their synthesis of about 140 sites in all continents, the average recharge was between 0.1 and 5% of the long-term average annual precipitation.

Recharge processes and surface-groundwater relationships inferred from the isotopic composition ( 18 O, 2 H) of rainfall, groundwater and surface water
The daily monitoring of rainwater between 2018 and 2019 showed δ 18 O values ranging from −12.97‰ to +2.27‰ and δ 2 H values ranging from −98.5‰ to +20.2‰, and evidenced a high variability of the isotopic composition of rainfall, typical of dry tropical regions with strong seasonal contrasts (e.g.Taupin et al. 1997).The annual amount-weighted mean isotopic composition of rainfall calculated from this dataset is −3.21‰ for δ 18 O and −12.0‰ for δ 2 H (n = 212).Only 9% of the daily rainwater samples showed marked evaporation values with deuterium-excess (d-exc) lower than +8.5‰.Conversely, rainwater samples from the monthly monitoring over the 2011-2019 period revealed highly evaporated values.Indeed, 33% of the monthly data showed d-exc values < +8.5‰ regardless of the amount of rainfall.Moreover, the δ 18 O vs δ 2 H plot evidenced that the monthly samples (n = 172) align along an evaporated isotopic line with a slope of 6.41 (r 2 = 0.90).The comparison of the δ 18 O values from the monthly samples with the monthly isotopic signal calculated from amount-weighted daily δ 18 O during the same period (2018)(2019) and between two stations with similar environmental characteristics highlighted that the monthly data suffered from an isotopic bias, probably due to the partial evaporation of the water stored in the water collector during the monthly storage.Nonetheless, the main features and relative variations of the monthly monitoring are consistent with those of the daily monitoring, and highlight the seasonal variability of the isotopic composition of precipitation at the monthly scale.Indeed, the monthly isotopic concentrations obtained from the daily amount-weighted δ 18 O are most depleted during the main rainy season (from February to May), particularly between March and April when δ 18 O values can decrease from −3.09‰ to −5.08‰ (Supplementary material Fig. S3).These seasonal variations are consistent with those observed in Fortaleza (Da Silva Nobre et al. 2019).
The monitoring of the groundwater isotopic composition, based on 205 samples originating from 78 distinct borewells sampled between 2009 and 2018, showed a wide range of values varying from −4.19 to +1.76‰ for δ 18 O, −25.1 to +8.0‰ for δ 2 H, and −8.6 to +11.6‰ for d-exc.Moreover, historical isotopic data obtained from the 2009-2012 quarterly monitoring showed a seasonal variation of the groundwater δ 18 O composition, which ranged from −3.90‰ to −1.03‰ over the borewells.The observation of a high spatio-temporal variability of the groundwater isotopic signal, even between two neighbouring borewells, confirms the strong lateral compartmentalization of the aquifers, implying that vertical transfers prevail over horizontal transfers, consistent with the literature about crystalline aquifers (Guihéneuf et al. 2014, Alazard et al. 2016, Lachassagne et al. 2021).It also reveals the occurrence of local recharge processes displaying the spatial variability of localized infiltration.Moreover, the seasonal variation of the groundwater δ 18 O implies rapid recharge and circulation processes within the aquifer.
Water isotopic compositions of the non-evaporated rainfall samples (i.e. with d-exc values ≥ +8.5‰) were plotted to determine the local meteoric water line (LMWL; Fig. 5).The groundwater and surface water samples were also plotted, which allows us to better understand the water's origin and the recharge processes (Fig. 5).The LMWL of Quixeramobim obtained from the monthly and daily rainfall datasets is defined by Equations ( 2) and (3), respectively: Slopes of Equations ( 2) and (3) are close to that of the global meteoric water line (GMWL), but their y-intercept value is greater (+11.3‰ and +14.1‰ vs +10.0‰) despite the proximity of the study area to the Atlantic coast, considered the dominant source of moisture.The non-evaporated precipitation record obtained from the study region in Moura (2013) or from Fortaleza in the available data of the Global Network of Isotopes in Precipitation of the International Atomic Energy Agency (IAEA-GNIP) also evidence a y-intercept value higher than 10‰, being +11.4‰ (r 2 = 98%, n = 12, 2011-2012 period) and +11.6‰ (r 2 = 96%, n = 68, 1965-1984 period), respectively.The high y-intercept value of Quixeramobim traces the mixing in rainwater of moisture from marine origin and continental recycling (Salati et al. 1979b).
Groundwater data points are scattered between the isotopic poles of rainwater and surface water (Fig. 5), suggesting the contribution of both poles to the recharge processes.Moreover, the slope of Equation ( 4) and the d-exc values of groundwater clearly show that meteoric waters undergo significant evaporation before their infiltration into aquifers.Thus, the substantial similarity between Equations ( 4) and ( 5) highlights that groundwater recharge is constituted not only by the diffuse infiltration of rainwater but also and mainly by localized infiltration of evaporated surface water.This hypothesis was confirmed by the piezometric survey and is consistent with the presence of numerous reservoirs, ponds and temporary drainage networks in the study area.
The intercept between the linear regressions of groundwater and rainfall (Fig. 5) shows an offset of about −2‰ in δ 18 O compared to the annual amount-weighted mean isotopic composition of rainfall, which typically represents the isotopic input signal for groundwater.However, in this case, the offset might be due to the natural inter-annual variability of the rainfall isotopic composition or may indicate that most aquifer recharge occurs during the intense rainfall events of the rainy season, particularly in March and April.This last hypothesis is consistent with the piezometric monitoring, and with similar research (e.g.Blavoux et al. 2013, who used another tracer: tritium).
The δ 18 O values of rainwater and surface water were used to infer the fraction of water that infiltrates indirectly in the aquifer.A δ 18 O value of −5‰ was assigned to rainwater, as it is considered that the aquifer recharge occurs mainly during the wettest month of the rainy season.Concerning the isotopic value of surface water, most of the samples were collected from large reservoirs that do not dry up during the year and that consequently present more enriched δ 18 O values (+1.78‰ < δ 18 O < +9.23‰).However, small reservoirs (−1.32‰ < δ 18 O < +1.65‰) are more prevalent in the region (Molle and Cadier 1992), and are probably more likely to lose water through leakage due to the absence of a rigorous construction method.Therefore, it was estimated that the isotopic signal of surface water undergoing infiltration may vary between −1 and +1‰.From these assumptions, it is estimated that for 67 to 32% of wells respectively, the fraction of evaporated water in the recharge is greater than 50%, and that the mean localized recharge rate varies from 65 to 44% of the total recharge respectively, depending on the assumptions made about the isotopic value of surface water.

Groundwater residence time determination
A multi-tracer approach based on the 14 C, 3 H, CFC and SF 6 contents in groundwater was implemented.Carbon-14 and tritium values ranged from 73.2 to 126.7 pMC and from ≤0.3 to 1 TU, respectively (Supplementary material Table S3).Carbon-13 delta values varied from −17.84‰ to −14.02‰, corresponding to the expected range of carbon-13 values produced by the respiration of C3 plants, which means that no geological or sedimentary input of carbon has disturbed the apparent ages that can be directly calculated from the radioactive decay law (Supplementary material Table S3).Thus, 14 C values generally higher than 100 pMC and measurable values of 3 H indicate that the waters recharging aquifers were mainly infiltrated after the nuclear tests of the 1950s.
However, the flowpath organization within fractured crystalline aquifers is complex.The CFC and SF 6 data, presented in Supplementary material Table S4 and best interpreted by a binary mixing model (BMM), underline this complexity.The results show the mixing of two types of water, including a recent pole (15 to 85% contribution) corresponding to the present-day recharge, and an older pole infiltrated before the 1960s.Consequently, the apparent age of groundwater varies from a few decades to hundreds of years, albeit with a significant contribution of post-1950s water.This implies fast flow processes, except in a few particular cases.The strong contribution of a current recharge validates the hypothesis of rapid recharge and vertical circulation processes observed with the water stable isotope data, superimposed on a scarcely circulating stock of old water (Kreis et al. 2020).The presence of older waters is a common finding in crystalline aquifers, as their deepest fractures are poorly connected to the more superficial ones (Guihéneuf et al. 2014, Alazard et al. 2016, Lachassagne et al. 2021).

Electrical conductivity
The EC values of the crystalline groundwater measured between 2016 and 2019 on 56 distinct borewells ranged from 886 to 19 310 μS cm −1 , with a median value of 2854 μS cm −1 .The mean value and standard deviation were 3561 and 3064 μS cm −1 , respectively.The large range of variation, the high standard deviation and the salinity differences observed between neighbouring wells confirm the strong spatial heterogeneity in the crystalline aquifer, already demonstrated in this paper.Moreover, a strong vertical stratification was also observed in 15 borewells, with EC increasing with the depth within the fractured layer (Kreis 2021).The absence of a salinity gradient at the sub-basin scale, associated with the absence of regional flow as demonstrated notably by piezometric data, implies that the observed salinity differences between the borewells are linked to local environmental processes.No relationship was found between the apparent age, or the stable isotopic composition, and the EC of the groundwater.However, the rainy season generally induces an increase in EC within the borewells (Supplementary material Fig. S4), which may be caused by evapoconcentration of dissolved species in the surface water bodies before infiltration, a leaching of salts stored in the USZ, or the increase of an upwards vertical flow component due to higher piezometric heads.

Major ions
The whole set of 65 analyses from 41 active borewells had an ionic balance error of less than 10%, therefore all results were used for chemical interpretation.This 10% limit is higher than the traditionally accepted standard (5%), but allowed us to significantly expand the number of borewells used for the chemical study.Furthermore, 62% of these analyses showed an ionic error of less than 5%, and 85% had an imbalance < 7%.All analyses with an ionic error between 5% and 10% had a dominant cation or anion exceeding 40% of the total cation or anion content.Thus, the 10% bias does not really affect the determination of the chemical facies.Therefore, setting the ionic balance acceptance limit at 10% did not introduce significant biases in the interpretation of the chemical facies or geochemical processes at the origin of salinization (Zakaria et al. 2015).Groundwaters are characterized by high concentrations of chloride and sodium.Chloride concentrations varied from 29 to 5126 mg L −1 , with a median value of 702 mg L −1 , while sodium concentrations ranged from 98 to 1800 mg L −1 , with a median value of 275 mg L −1 (Table 1).An excellent correlation is observed between the EC values and the chloride concentration (r 2 = 0.96), which indicates that the salinity (represented here by EC) derives essentially from chloride salts and, to a lesser extent, calcium (r 2 = 0.79), magnesium (r 2 = 0.69), and sodium (r 2 = 0.68).No relationship is observed between EC and bicarbonate, sulphate or nitrate (r 2 < 0.1).The chemical analyses highlighted that 49, 22, 11, 5 and 4% of the samples present an Na-Mg-Cl, Na-Cl, Na-Mg-Cl-HCO 3 , Mg-Cl or Na-Cl-HCO 3 facies, respectively (Fig. 6).Nevertheless, Na-Mg-HCO 3 and Na-HCO 3 facies are observed for the less salty waters (with EC < 1600 μS cm −1 ).Results show that the proportion of chloride increases with the salinity and exceeds that of bicarbonate until reaching percentage levels up to 50% of the total ionic content for the most saline waters (Supplementary material Fig. S5).
Groundwater samples collected before (December 2017) and after (June 2018) the rainy season are plotted on Piper diagrams (Fig. 6).The rainy season generally caused a shift of the points in the cation sector from the sodium pole, dominant at the end of the dry season, towards the mixed area (Fig. 6).This observation suggests, in association with the observed increase in salinity, that processes such as calcium and/or magnesium salt dissolution (as calcite, dolomite or even magnesium-chloride salts) may be involved.For the anions, the evolution of the groundwater chemical composition during the different hydrological periods is more complex, with distinct behaviours according to the wells especially in terms of chloride and bicarbonate proportions.Nonetheless, an increase in chloride ions is generally observed at the end of the rainy season compared to bicarbonates (in accordance with the EC increase).This increase is not necessarily linked with an increase in nitrate or sulphate concentrations, and suggests therefore that the seasonal increase in salinity is associated not with the recharge of contaminated water (from agriculture and/or waste water) but, as already stated, with the mobilization of more mineralized, possibly older waters.
Nitrate does not show a significant correlation with the other chemical elements.However, groundwaters show high nitrate concentrations with median and maximum values of 71 and 690 mg L −1 , respectively.The highest values of nitrate observed are explained by livestock farming as well as evapoconcentration.Indeed, in the study area, all the borewells with nitrate concentrations over 50 mg L −1 are localized near agricultural or residential areas.Thus, nitrate in groundwater may result from the use of organic fertilizers (KNO 3 ) and manure, faecal contamination from animals, lack of wastewater treatment or poor septic systems, or even from the decomposition of plant and animal matter.Natural sources may also contribute to high nitrate concentrations, through the presence of termite mounds or nitrogen-fixing trees such as acacias (Schwiede et al. 2005) in the study area.More investigations (with δ 15 N, for example) should be carried out to evaluate the main sources of the nitrate.The absence of correlation between nitrate and chloride concentrations confirms that salinity is not directly linked to anthropic pollution, even if human activities can contribute as an additional source of salts.

Minor ions and trace elements
Minor ions and trace elements were determined from a set of 20 borewells, whose EC varied from 975 to 15 780 μS cm −1 with a median value of 2392 μS cm −1 .These samples are characterized by a median concentration of silica (Si 4+ ) of 45 mg L −1 , and a relatively limited range of values between 28 and 62 mg L −1 (Supplementary material Table S5).Bromide (Br − ) concentrations ranged from 0.2 to 16 mg L −1 , with a median value of 1.0 mg L −1 .Regarding trace elements, groundwaters are mainly characterized by the presence of strontium, barium, boron and phosphorus, with median concentrations of 977, 514, 142 and 131 μg L −1 , respectively.Standard deviations are high (Supplementary material Table S5) and large differences in concentration can be observed, even between neighbouring borewells, which reflects the strong spatial variability of the aquifer lithology.
The Si 4+ present in most waters derives from aluminosilicate minerals such as feldspars and ferromagnesium silicates (Shand et al. 2007).Si 4+ values measured in this work are slightly higher than the concentrations usually found, and are suggestive of water-rock interaction processes leading to mineral hydrolysis (e.g.Pradeep et al. 2016, Marçais et al. 2018or N'guettia et al. 2019).Water evapotranspiration processes could explain the high Si 4+ values.In any case, there is no relationship between EC and Si 4+ , even for the least salty waters.Conversely, Br − does not derive from the hydrolysis of silicate mineral (Kloppmann et al. 2011), and an excellent correlation between bromide and chloride (r 2 = 0.99) was observed.The groundwater Cl − /Br − molar ratios ranged from 255 to 1217 (Supplementary material Table S6), with a mean Cl − /Br − molar ratio of 815 ± 198.All the samples with a concentration of chloride above 100 mg L −1 presented a Cl − /Br − ratio similar to or higher than the oceanic ratio of ∼650 (Drever 1997).On the other hand, the P17 borewell, which has the lowest salinity and a chloride content of around 20 mg L −1 , showed a much lower Cl − /Br − ratio (∼ 255).Thus, the Cl − / Br − molar ratio in the P17 borewell could be interpreted as reflecting that of the local rainfall (Cartwright et al. 2006).The increase of the Cl − /Br − molar ratio in the other borewells, associated with an increase of Cl concentration, is suggestive of halite dissolution during recharge (Cartwright et al. 2006, Alcalá andCustodio 2008).The fact that Cl − /Br − ratios remain relatively constant as Cl concentrations increase above 300 mg L −1 provides evidence that another factor responsible for the rise in groundwater salinity may be evapotranspiration (Cartwright et al. 2006, Alcalá andCustodio 2008).Investigation with δ 37 Cl or 36 Cl could be useful to determine the origin of chloride or to constrain processes (e.g.evapotranspiration) affecting groundwater (Cartwright et al. 2006, Kloppmann et al. 2011).
Strontium, barium, boron and phosphorus are traditionally found in silicate and carbonate minerals.Ba 2+ and Sr 2+ values in groundwaters are generally on the order of hundreds of μg L −1 (Blum et al. 2001, Shand et al. 2007), as is the case in our study.Strontium can also derive from the dissolution of evaporite minerals, and in that case may reach concentrations on the order of tens of mg L −1 .In crystalline regions, boron concentrations generally do not exceed a few μg L −1 (Blum et al. 2001), but it may also have an anthropic origin or derive from evaporite deposits.The good correlation observed between the EC and bromide (r 2 = 0.96), lithium (r 2 = 0.87) or strontium (r 2 = 0.84) suggests that part of these elements derive from the dissolution of salts, but it also reveals that their groundwater concentrations are at least partly controlled by processes leading to the evapoconcentration of the dissolved species.

Assessment of groundwater salinization processes from mass balance evaluation
Groundwater generally originates from meteoric water, which is poorly mineralized.However, despite flowing in a crystalline environment with relatively short residence times (a few decades in 80% of the borewells investigated, up to a few hundred years), the groundwater is marked by high mineralization levels and chloride concentrations, as well as the predominance of the chloride anion facies.These characteristics are not compatible with the equilibrium time required for the interaction between rock and water to take place to such extent.Indeed, groundwater from igneous or metamorphic rocks normally exhibits low mineralization (100-300 mg L −1 ), with a predominance of bicarbonate and calcium ions, or even magnesium or sodium cations, whether in temperate climates (Gustafson andKrásný 1994, Blum et al. 2001) or in semi-arid conditions (Ousmane et al. 1984, Asomaning 1992, Babaye et al. 2019).
Bicarbonate present in water is generally derived from the hydrolysis of silicate or the dissolution of carbonate minerals, while the presence of calcium, magnesium and potassium is generally attributed to the hydrolysis of plagioclases (source of Na + and Ca 2+ ), K-feldspars and/or ferromagnesian minerals such as biotite (source of K + and Mg 2+ ), amphiboles, pyroxenes or olivine (sources of Mg 2+ and Ca 2+ ), among others.The predominance of sodium ions in the crystalline aquifer groundwaters (as observed in our case or in other areas of the world) is generally explained by the presence of alkali feldspars (such as albite or anorthoclase, for example), by reverse cation exchange phenomena (replacement of Ca 2+ or Mg 2+ by Na + ) or by calcite and dolomite precipitation phenomena (precipitation of Ca 2+ and/or Mg 2+ ).
An analysis of the chemical ratios between ions was performed to identify potential sources of groundwater salinization, and the samples were sorted into 10 groups according to facies trends for the sake of clarity: groups 1 to 4 correspond to the least salty waters whose dominant or almost dominant anion is HCO 3 − , while groups 5 to 10 represent the waters whose dominant or almost dominant anion is Cl − .More details on the group constitution are given in Table 2.These analyses showed that mineral hydrolysis cannot explain the high salinities of the aquifers, but can explain the basic mineralization and the bicarbonate facies of the less salty waters.Indeed, the relationship in meq L −1 between alkalinity + SO 4 and Ca 2+ + Mg 2+ (Fig. 7(a)) underlines that groundwater samples from groups 1 to 4 are close to or below the 1:1 equiline, indicating a deficit in calcium and magnesium.This observation suggests that these ions could come from the  The P17 borewell sample (group 2), corresponding to the lowest EC value (EC = 886-1059 μS cm −1 ) is the only perfectly balanced borewell that does not present evidence of contamination by nitrate: the chloride content and a part of the sodium are derived from atmospheric inputs, while the other ions (including the other part of sodium) are essentially derived from silicate weathering and cationic exchange.Even if this sample is isotopically marked by evaporation (d-exc < −6‰), which suggests a possible concentration of the dissolved elements by evaporation, we can assume that the reference values for the crystalline groundwater of the Ceará, when not subjected to salinization or contamination issues, are characterized by a sodium-bicarbonate facies,   a total dissolved solids (TDS) content < 700 mg L −1 and chloride concentration < 30 mg L −1 (Table 3).Therefore, chloride concentration above this limit may indicate additional salinization processes.The mineral hydrolysis process cannot explain the origin of the cations for the salty groundwaters of groups 5 to 10, whose dominant ion is chloride.Indeed, groundwater samples from groups 6 and 8 (≈ 1700 < EC < 3500 μS cm −1 ) are close to (or slightly above) the 1:1 equiline (Fig. 9(b,d)), while samples from groups 5, 7, 9 and 10 (≈1100 < EC < 14300μS cm −1 ) present a strong excess in Ca 2+ and Mg 2+ (from 2 to 90 meq L −1 ) compared to bicarbonate (Fig. 9(a,b)), as well as a strong deficit in sodium compared to chloride (Fig. 9(c,d)).Results suggest that part of the mineralization originates from silicate hydrolysis, while another part is linked to other processes.Indeed, samples from groups 6 and 8 show a chemical signal that derives from the hydrolysis of silicate minerals but also from the dissolution of salts of halite (and probably of calcite and dolomite, considering the sequence of precipitation of salts).Samples from group 10 also show a signal of halite dissolution.Samples from groups 5, 7 and 9 of mixed cationic facies (generally Na + > Mg 2+ > Ca 2+ ) present a chloride excess in relation to sodium ranging from a few meq L −1 to 100 meq L −1 , and it is particularly significant for solutions above 20 meq Cl− L −1 (corresponding to solutions whose EC > 3200 μS cm −1 ).Thus, the additional source of Ca 2+ and Mg 2+ and the deficit of Na + might be explained by reverse cationic exchange phenomena between calcium and magnesium with sodium (adsorption of Na + in exchange for bound Ca 2+ and Mg 2+ ; Fig. 8(b)).Alternatively, the sodium deficit may be linked to the dissolution of more evolved salts such as MgCl 2 (or CaCl 2 from fertilizers), which is possible considering the good correlation of these cations with the chloride (r 2 > 0.7).
Similar observations were made by Laraque (1991) for the surface water of the study region: water with low mineralization presented a calcium-bicarbonate or sodium-bicarbonate facies, while water with higher mineralization mainly presented a sodium-chloride facies.The salinization of surface water was attributed to the leaching of salt derived from rainfall that was accumulated in the watershed, to the strong evaporation rates, to the inadequate sizing of dams and to the irregular use of the water contained in the latter (Laraque 1991).

Evaluating groundwater salinization processes from PCA
Two sets of PCA computations were performed to evaluate the relationships between variables and to quantify the con- tribution of each variable in the constitution of the factorial axes.The first analysis considered all major ions (Fig. 10(a)), while the second also took into account the main trace elements (Fig. 10(b)).Results show that the variance expressed by the factorial axes F1-F2 is quite significant in both cases and can explain more than 70% of the initial information.Indeed, more than 55% (Fig. 10(a)) or 75% (Fig. 10(b)) of the variance is controlled by the F1 axis, which is defined by EC, Cl − , Br − , Ca 2+ , Mg 2+ , Na + , K + , Sr 2+ and Li + .The grouping of these elements around the F1 axis implies a common origin and reflects the mineralization of water attributed to the dissolution of salts.The F2 axis, which explains 13 to 15% of the variance, is represented, on the one hand, by NO 3 − and/or SO 4 2− , which represent the impact of human activities; and, on the other hand, by HCO 3 − , Si 4+ , Ba 2+ and/or SO 4 2− , which reflect the mineralization of water due to the hydrolysis of silicate or sulphur-rich minerals.Thus, the PCA outcome supports the hypothesis that the substantial salt concentrations found in the crystalline waters may not be due to the mineral hydrolysis, but may be due to the dissolution of salts.This hypothesis is supported by the observation of saline crusts constituted by evaporite minerals (calcite, dolomite, gypsum, halite and magnesium salts as bischofite) on the banks and sediments of dry reservoirs (Laraque 1991, Araújo 2017).

Discussion
The first and unprecedented hourly piezometric data acquired in the region evidenced that the seasonal piezometric fluctuation is often continuous and inertial, even if abrupt variations were also detected in some wells (rapid vertical infiltration during the highest rain events).Similarly to other semi-arid regions in the world (Leduc et al. 1998, Babaye et al. 2019, Lachassagne et al. 2021, Goswami and Sekhar 2022), the recharge of Ceará crystalline aquifers does not directly depend on the amount of annual or monthly rainfall, but on the intensity and distribution of rainfall events, parameters that will influence both the production of surface runoff and surface water accumulation in temporary rivers and dams and the consequent localized aquifer recharge.Indeed, the isotopic investigation evidenced that groundwater recharge is mainly constituted by localized infiltration of evaporated surface water.This observation, based on 654 recent isotopic analyses, leads to a conclusion contrary to the interpretations formulated by Frischkorn et al. (1989), Santiago et al. (2001) and Moura (2013), who concluded, from a more limited set of analyses, that the crystalline aquifer recharge is due to direct (diffuse) infiltration without any sign of evaporation of the surface water.
Among the processes that can lead to a salinization of groundwater, marine intrusions and dissolution of marine evaporite rocks are ruled out, considering the distance of the study area from the sea (≈180 km) and the absence of primary evaporitic terms in the Precambrian rocks (Almeida et al. 2008).Dating of the groundwater through our multi-tracer approach ( 14 C, 3 H, CFC and SF 6 ) and the seasonal variations of stable isotopes ( 18 O and 2 H) highlighted that the circulation of groundwater is relatively fast (mean residence time from a few decades to hundreds of years, with a large contribution of post-1950s water) and corroborates the water dating obtained by Santiago et al. (2000).It also showed that the apparent age of the crystalline aquifer waters is a function of the annual renewal rate of the aquifer via vertical flows.The hypothesis of a past marine transgression is incompatible with the apparent age of the water, considering that the last marine transgression that could possibly have reached the study area, and whose witnesses could have been removed by erosion processes, dates from the Upper Cretaceous (≈ 70 Ma; Arai 1999).Furthermore, the stable isotopes are incompatible with the hypothesis of mixing with seawater, geothermal flows or fluid inclusions as potential processes of water salinization.Mineral hydrolysis of the crystalline rocks explains part of the mineralization of the water, but does not account for the significant salinization of the groundwater, contrary to what was suggested by Araújo (2017).Indeed, it was demonstrated in our study that dominant mineral hydrolysis is only associated with low mineralization (< 700 mg L −1 ) and bicarbonate facies.Therefore, brackish or saline water with chloride or mixed facies within the crystalline rocks is a consequence of other salinization processes.Chloride is not representative of crystalline rocks, and the presence of brine, fluid inclusions or evaporite rocks does not correspond to the lithological types of the studied region.Consequently, the chloride sources are external to the crystalline aquifer, as evoked by Matsui (1978), Silva (2003) andBritto Costa et al. (2006).It was evidenced in this study that anthropic contamination is not a determining factor in the processes of salinization of water, even if it can contribute to a local degradation of water quality, more specifically in terms of nitrate content.Therefore, the most likely hypothesis in this context is that the chloride source -and hence the groundwater salinity -is derived from meteoric inputs, whether in the form of wet deposition (during the rainy season) or dry deposition (during the dry season, when most winds occur).Our piezometric monitoring showed that small rainfall events are completely recycled by evapotranspiration and do not contribute to the recharge.Considering the low permeability of soils and the presence of Hortonian-type runoff, most of the chloride deposited on the soil surface may be leached by surface runoff and transferred to surface water reservoirs (rivers, dams, ponds) which are subject to partial or total evaporation.The evaporation of these surface waters can even lead to the local precipitation of evaporite minerals and saline crusts.Indeed, previous studies (Laraque 1991, Araújo 2017) evidenced the presence of evaporite minerals (calcite, dolomite, gypsum, halite) on the banks of dried reservoirs, as well as in the clays and sediments of the reservoirs (and probably in the USZ).Saline crusts also contained magnesium salts (probably bischofite MgCl 2 -6H 2 O; Araújo 2017).
The increase in EC observed after the rainy season in most groundwaters, associated with the rise in the piezometric level and the infiltration of partly evaporated surface water, suggests that the seasonal dynamics of water salinity comes from the leaching of salts from the surface.This increase varies from year to year, and suggests that the annual amount of salts leached into the aquifer must be variable and a function of the spatio-temporal distribution of daily rainfall.In terms of salinization processes, these new observations highlight the hypothesis according to which the salinity of recharge waters would be due to the accumulation of salts in localized areas, which will be progressively concentrated, in surface waters (rivers, dams), in the soil or in the USZ, due to the high evaporation rates associated with the semiarid climate, and being leached into the aquifer after more significant rainfall events.

Conclusion and perspective
In this multidisciplinary study, piezometric, hydrochemical, isotopic data and multi-tracer dating ( 18 O, 2 H, 3 H, 14 C, CFC, SF 6 ), provided new hydrogeological information, especially about the hydrogeological functioning and the salinization processes of the crystalline aquifers.The high piezometric reactivity of aquifers in response to rainfall marks the seasonal recharge, even during the driest years.The stable isotopes ( 18 O, 2 H) revealed that groundwater is systematically marked by evaporation, confirming that recharge comes from a mixing of evaporated surface waters (localized recharge in lowlands or from surface water bodies) and, to a lesser extent, diffuse recharge of rainwater.Thus, preferential infiltration from localized areas more propitious to infiltration, such as river beds, alluvium, reservoirs or local depressions, is the main pathway of recharge of these aquifers.The determination of groundwater residence times through a multi-tracer approach ( 3 H, 14 C, CFC, SF 6 ) underlined that the apparent groundwater age is recent and composed of seasonal vertical infiltration flows (direct and indirect) that mix with older waters stored in the aquifer, whose infiltration occurred before the 1960s, and which flow much more slowly.
Although the crystalline aquifer groundwaters are apparently young, they are characterized by high salinities and present essentially mixed-chloride or sodium-chloride facies.Chemical data reveal that groundwater initially exhibits a mixed-bicarbonate or sodium-bicarbonate facies resulting from the hydrolysis of the aquifer crystalline rocks and cation exchange, resulting in EC values that should not exceed 1000 μS cm −1 .However, the partial or total evapo-(transpi)ration of small rainfall events and surface waters favours the concentration of the atmospheric salts deposited onto the soil, and in the USZ, the sediments of the ponds and/or the surface waters, eventually causing their precipitation in the form of evaporitic salts, such as calcite, dolomite, gypsum, halite or even more evolved salts such as MgCl 2 (indicating the strong evaporation rates that can locally affect the waters before their infiltration).The dissolution and leaching of these evapoconcentrated salts into the aquifers during more intense rainfall events shift the groundwater facies from the bicarbonate pole to the chloride pole.Therefore, the geochemical interpretations showed that the origin of the groundwater salinization is not linked to the hydrolysis of crystalline rock minerals, but to the lixiviation of evapoconcentrated salts into the aquifer.Reverse cationic exchange processes can affect the groundwater composition, but not the degree of salinization.Moreover, anthropogenic pollution can contribute locally to the degradation of the water quality (more specifically in terms of nitrate), but has not been identified as a determining factor in the salinization processes.
In light of these conclusions, the perspectives of this study are to test with a quantitative model whether the hypothesis of atmospheric inputs concentrated locally by evapotranspiration would produce the observed salinities, and also to evaluate the possible influences of surface water evaporation or other phenomena, such as the transpiration of groundwater (from the saturated zone) by the vegetation (Herczeg et al. 2001, Favreau 2018), on groundwater salinity.Such a quantitative modelling of the salinization processes is performed in the frame of the continuation of this research (Kreis et al. 2023).

Figure 2 .
Figure 2. Location of the sampled points for the isotope and dating analyses.Only borewells that were sampled for groundwater dating have a colored background.

Figure 3 .
Figure 3. Monitoring of the groundwater levels of the crystalline aquifers between 2011 and 2019.

Figure 4 .
Figure 4. Hourly piezometric monitoring (November 2016 to December 2019) of the PIR-P14 and IBI-P24 boreholes with associated daily and annual rainfall.

Figure 5 .
Figure 5. Water isotopic compositions of rainwater, groundwater and surface water.
hydrolysis of the silicate minerals (and possibly from sulphurrich minerals for sulphate), considering the absence of primary carbonates in the crystalline aquifers.The deficit in calcium and magnesium observed for samples 144, 164, P17 and 45 is also associated with an excess of sodium in relation to chloride (Fig.7(b)), which indicates an additional source of sodium beyond that from atmospheric inputs (blue dotted line) or halite dissolution (black dotted line).Analyses suggest that the deficit in calcium and magnesium in relation to the alkalinity + SO 4 could be explained by cationic exchange phenomena with clays (Fig.8(a)), or by the albite weathering of the crystalline rocks, which generates bicarbonate ions associated with sodium ions, as described in Equation (6):

Table 2 .
Groundwater groups identified according to facies trends.

Table 3 .
Reference values for the groundwater of the crystalline rocks in Ceará that are not subject to pollution or salinization processes.