Journal of Hydrology 519 (2014) 23–33

Contents lists available at ScienceDirect

Journal of Hydrology

journal homepage: www.elsevier .com/ locate / jhydrol

Infiltration processes in karstic chalk investigated through a spatialanalysis of the geochemical properties of the groundwater: The effect ofthe superficial layer of clay-with-flints

http://dx.doi.org/10.1016/j.jhydrol.2014.07.0020022-1694/� 2014 Elsevier B.V. All rights reserved.

⇑ Corresponding author.E-mail address: daniele.valdes_lao@upmc.fr (D. Valdes).

Danièle Valdes a,⇑, Jean-Paul Dupont b, Benoît Laignel b, Smaïl Slimani c, Célestine Delbart d,e,a

a UPMC Univ Paris 06, UMR 7619 METIS, Case 105, 4 Place Jussieu, F-75005 Paris, Franceb Morphodynamique Continentale et Côtière UMR CNRS 6143, Université de Rouen, 76821 Mont-Saint-Aignan Cedex, Francec Ecole d’ingénieurs en Agro développement à l’international (ISTOM), 32 Boulevard du Port, 95000 Cergy, Franced Univ. Paris Sud, Laboratoire IDES, UMR8148, Orsay F-91405, Francee CEA, DAM, DIF, F-91297 Arpajon, France

a r t i c l e i n f o

Article history:Received 25 October 2013Received in revised form 30 June 2014Accepted 1 July 2014Available online 9 July 2014This manuscript was handled by CorradoCorradini, Editor-in-Chief, with theassistance of Barbara Mahler, AssociateEditor

Keywords:ChalkGroundwaterInfiltrationClay-with-flintsSpatial correlations

s u m m a r y

In the Paris Basin in Upper Normandy (France), the chalk plateaus are covered with thick deposits of loessand clay-with-flints, from a few meters to approximately 40 m thick locally. A perched groundwater issometimes observed in the superficial layers in which evapotranspiration processes seem to occur.

This study’s objective was to understand the effects of the thick clay-with-flints layers on the infiltra-tion processes. To achieve this, we adopted a spatial approach comparing the maps of the geochemicalproperties of the Chalk groundwater and the maps of the thickness of clay-with-flints.

The French national groundwater database, ADES (Accès aux Données des Eaux, BRGM), provided themean geochemical properties in the Chalk aquifer of Upper Normandy. This database was used to preparemaps of the environmental tracers: Ca2+, HCO3

�, Mg2+, Cl�, Na+, NO3�, and SO4

2. The data are spatially wellorganized.

Using principal component analysis (PCA), these maps were compared with the maps of the thicknessof clay-with-flints. A focus on the coastal basins (northern Upper Normandy) shows a very strong spatialcorrelation between the maps of clay-with-flints thickness and all of the maps of the major ions. Thethickness of clay-with-flints is negatively correlated with the autochthonous ions (HCO3

� and Ca2+) andis positively correlated with the allochthonous ions (Cl�, Na+, SO4

2�, and NO3�).

These results highlight that the thickness of clay-with-flints controls recharge. Two types of infiltrationprocesses are proposed: (1) Thicker clay-with-flints allows storage in the perched groundwater, whichallows evapotranspiration, resulting in high concentrations of allochthonous ions and a decrease in thedissolution potential of water and low concentrations of autochthonous ions. The infiltration of theperched groundwater is thus delayed and concentrated. (2) Thinner clay-with-flints causes the infiltra-tion to be more diffuse, with low evapotranspiration and thus low concentrations of allochthonous ionsin the Chalk groundwater; more, there is more dissolution and higher concentrations of autochthonousions in the Chalk groundwater.

� 2014 Elsevier B.V. All rights reserved.

1. Introduction

The Cretaceous Chalk aquifer is the primary groundwaterresource in Upper Normandy (France) and has been demonstratedto have karst characteristics (Calba, 1980; Rodet, 1992).

The quality of this Chalk groundwater is generally good; how-ever, its geochemical properties vary spatially and temporally.These temporal variations may be related to climatic impacts, suchas rain events, which often result in contamination problems, gen-erally associated with an increase in turbidity in wells and springs(Dussart-Baptista et al., 2003; Mahler et al., 2008; Massei et al.,2003; Valdes et al., 2005, 2006). Chalk aquifer groundwater faciesare primarily composed of Ca–HCO3, but the ions vary spatiallythroughout the Upper Normandy region. The spatial heterogeneity

24 D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33

of the groundwater geochemistry is generally related to the heter-ogeneity of external factors, such as atmospheric inputs (Meybeck,1983), land use, and domestic sewage (Negrel and Petelet-Giraud,2005; Sherwood, 1989).

Valdes et al. (2007) have already presented a spatial study of thegeochemical properties of the Chalk groundwater in the Paris Basinat a regional scale. They focused on the Eure department in south-ern Upper Normandy. In this area, the Chalk aquifer is drained bythe Seine River, then oriented from south to north perpendicularto the structural settings (anticlines, faults). They showed that, inspite of its karst properties, the large-scale geochemical propertiesof the Chalk groundwater are strongly spatially organized. Actually,they observed a continuity of the geochemical properties in theChalk groundwater. To better understand this spatial distribution,they compared these geochemical data to the physical propertiesof the aquifer, in particular aquifer thickness (representing aquifergeometry) and piezometric level (representing aquifer flow). Theyobserved that (1) the degree of mineralization (principally com-posed of Ca2+ and HCO3

� ions) increased along the flow direction,corresponding to an increase in the chalk dissolution rate alongthe flow path, and (2) the steepest mineralization gradients wererelated to an increase in the Mg/Ca ratio, evidence of longer resi-dence times and corresponding to zones where aquifer flow capac-ity is limited because of a decrease in the thickness of the flowsection (anticlines or faults). These results highlighted the domi-nant role played by the geometry and structural context in control-ling Chalk groundwater geochemistry.

These chalk plateaus are covered with thick deposits of loess andclay-with-flints, from a few meters to approximately 40 m thicklocally (Laignel et al., 2002). The flow behaviors of karst aquifers(quality and quantity) are characterized by recharge (diffuse/con-centrated), storage (vadose/phreatic), and flow (diffuse/concen-trated) (Ford and Williams, 1989; Smart and Friederich, 1986;White, 1988). Many studies have investigated the infiltration andthe flow in the unsaturated chalk zone (Brouyère, 2006; Iresonand Butler, 2011; Ireson et al., 2009; Price et al., 2000; Van denDaele et al., 2007), and the infiltration through clay-with-flints isnot well understood: in the past, the clay-with-flints layer was con-sidered a homogeneous and relatively impermeable layer. Later,these ideas changed, however. More recently, Klinck et al. (1998)presented heterogeneous measurements of hydraulic conductivityof the clay-with-flints in the UK: 1) in the laboratory the values varyfrom 10�8 m s�1 to 10�5 m s�1 depending on the site measured andthe depth; (2) in situ (trial pit infiltration capacity) the values varyfrom 4 � 10�5 m s�1 to 4 � 10�9 m s�1. The effect of the thickness ofthis layer remains unknown. It can be assumed that infiltration pro-cesses differ between zones with a thick layer of clay-with-flintsand zones where the chalk almost outcrops. Does this layer concen-trate the infiltrated water? Does it protect the aquifer? Can a thicklayer be considered an impervious layer?

The objective of this study was to understand the effect of thethickness of the clay-with-flints layers on the infiltrationprocesses. To achieve this, we adopted a spatial approach, whichconsists of comparing the maps of the geochemical properties ofthe Chalk aquifer and the maps of the thickness of clay-with-flintsusing principal component analysis (PCA). We then propose aninterpretation of the role played by clay-with-flints in the infiltra-tion processes in the Chalk aquifer.

1.1. Study area

This study focuses on the Chalk aquifer of the Western ParisBasin within the Upper Normandy Region in France (Fig. 1a). Thereis little urbanization in the area, and land use is primarilyagricultural. The climate is temperate and maritime, with averagetemperatures of 10–12 �C and average annual rainfall varying

spatially from 600 to 1000 mm. From a geomorphological pointof view, the regional topography is characterized by large plateausof moderate elevations: <300 m above sea level (asl), deeplyincised by narrow valleys draining the Chalk aquifer.

1.1.1. Geological settingsThe chalk plateaus are composed of Cretaceous Chalk formations

(from Cenomanian to Campanian), with the exception of a zonenorth of the Seine, which is Jurassic, and a zone in the southern partof the study area, where the Cenomanian has a sand facies (‘‘PercheSands’’) (Fig. 1a). The chalk formations are covered with Cenozoicclay-with-flints, Quaternary loess (Lautridou, 1985) and Tertiarydeposits (Fig. 1a and b). The clay-with-flints layer resulted fromthe weathering of the chalk during different periods of the Cenozoic(Laignel et al., 1998a) and is from 0 to 40 m thick (Fig. 1a). The thick-est clay-with-flints is located in the southern and western portionsof the study area. The Quaternary loess deposits are eolian in originand range from 0 to 5 m thick. The Tertiary deposits occur as infillon weathered surfaces (pockets) in some places and provide contin-uous cover in other places (in the eastern portion of the study area,near the Seine River, Fig. 1a); they are composed of clayey sands andclays with up to 90% smectite content, with the smectite oftencontaining Mg2+ and SO4

2� (Laignel, 2003).

1.1.2. Hydrological and hydrogeological settingsAlthough the clay-with-flints formation is relatively imperme-

able, the saturated zone in the Chalk aquifer is lower than the baseof this overlying formation. The Chalk aquifer is considered to bean unconfined aquifer. The thickness of the saturated aquifer variesfrom a few meters to more than 300 m (Valdes et al., 2007). Theaquifer is karstified to some degree (Massei et al., 2003; Rodet,1999; Valdes et al., 2006), and the groundwater flow velocitiesare extremely heterogeneous because fissures and conduits pro-vide underground drainage routes for the highly localized trans-port of water at velocities from 50 to 300 m/h (Calba et al.,1979). During rain events, concentrated infiltration of surfacewater may result in a rapid increase in turbidity at springs andwells connected to the surface via karst systems.

The major drainage axis is the Seine River; the hydrologic net-work is typical of a karst zone, with primarily low-order streams(Hauchard et al., 2002) because most of the flow is subsurface.The Chalk aquifer can be divided into three parts (Fig. 1a): (1)the SEA basins: the coastal basins where the direction of flow istoward the sea (La Manche); (2) the SEINE RIGHT basins, whichdrain into the right bank of the Seine River; and (3) the SEINE LEFTbasins, which drain into the left bank of the Seine River.

Perched groundwater is sometimes observed at some sites, par-ticularly in the clay-with-flints layer a few meters above the soilsurface (Jardani et al., 2006).

2. Data and methods

2.1. Data

The French national groundwater database, ADES (Accès auxDonnées des Eaux Souterraines, BRGM), provided the mean turbid-ity and the mean geochemical properties for approximately 300springs and wells in the Upper Normandy Chalk aquifer (Fig. 2a).The geochemical part of the study focuses on the environmentaltracers: Ca2+, HCO3

�, Mg2+, Cl�, Na+, NO3�, and SO4

2 (Table 1). Sam-pling was conducted from 1995 to 2005. For each sampling site,the number of measurements ranged from 2 to 23, with a meanof five measurements per site (the distribution is presented inFig. 3a). Those few samples with an charge balance error greaterthan 10% were removed. For each site, the mean concentration

Fig. 1. (a) Map of the study area and limits of the basins: spatial definition of the three areas: the coastal basins (SEA basins), the basins that drain into the Seine River (SEINERIGHT basins for the right bank and SEINE LEFT basins for the left bank): projected coordinates system: Lambert 2 étendu, (b) map of the thickness of clay-with-flints –projected coordinates system: Lambert 2 étendu, and (c) a sketch of the geological section of the Western Paris Basin (by Quesnel, 1997; Laignel et al., 2004).

D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33 25

was calculated. Fig. 3b and c gives the distribution of mean Ca2+

and NO3� concentrations, respectively, for the 300 sites to estimate

the spatial variation of the geochemical properties. The geochemi-cal properties of the groundwater for each sampling site are notconstant over time: seasonal or long-term variations may occur.In order to assess the temporal variation for each site (dependingon the number of measurements), the standard deviations (in spiteof the low number of measurements per site) were calculated (seeFig. 3d and e for Ca2+ and NO3

� concentrations, respectively).Approximately 5% of the sampling sites were not used when thestandard deviation of the geochemical measurements was judgedto be too large with respect to the mean (Fig. 3c and e). These casescorrespond to (1) measurement errors or (2) sites connected tokarst systems, which may be disturbed by surface water contribu-tions during rain events (Valdes, 2005; Valdes et al., 2006). To theextent possible, the data represent the groundwater quality with

no surface water contribution. For the remaining data, for eachgeochemical parameter, the distribution of the mean concentrationcompared to the distribution of the standard deviation for each siteshows that the spatial differences observed between sites aregreater than the temporal differences in geochemistry.

The mean turbidity values provided by the ADES database willbe used to approach the karstic properties of the aquifer.

The maps of the thicknesses of clay-with-flints and loess fromLaignel et al. (1998b) and Quesnel (1997) were used.

An artificial well located in the clay-with-flints was studied atthe Bouville site (Fig. 2a). The water level, electrical conductivity,and chloride concentration were measured from February toAugust 2007 at a variable time step, ranging from 1 week to2 months (the chloride concentration was measured from Marchto June 2007 only). Meteo-France provided the rain data at theBOOS station (Fig. 2a) at a daily time step.

Fig. 2. Maps of the sampling sites and of the geochemical properties (mean values per sampling site) of the Chalk groundwater – projected coordinates system: Lambert 2étendu: (a) Chalk groundwater sampling sites, the Bouville site and the Boos rain station, (b) Chalk groundwater HCO3

� concentration, (c) Chalk groundwater Cl�

concentration, and (d) Chalk groundwater Mg2+ concentration.

26 D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33

2.2. Methods

PCA was performed on the regionalized geochemical variables(Ca2+, HCO3

�, Mg2+, Cl�, Na+, NO3�, and SO4

2�). PCA is a multivariatestatistical technique used for data reduction of large data sets. Thismethod is commonly used for environmental studies with a high

degree of temporal variation (Ben Othman et al., 1997; Eisenlohret al., 1997) or spatial variation (Reisenhofer et al., 1998; Valdeset al., 2007; Wang et al., 2001).

The geochemical and turbidity data were interpolated usingordinary kriging (Goovaerts et al., 1993) in ArcGIS using an omni-directional variogram and mapped with a 2000 m lag.

Table 1Mean and standard deviation of the geochemical properties of the Chalk groundwa-ter: major ions and the thicknesses of the superficial layers: loess and clay-with-flints: CWF (for the entire study area: Upper Normandy).

Mean SD Unit

[Ca] 106.1 9.4 mg L�1

[HCO3] 298.4 30.5 mg L�1

[Mg] 5.9 2.9 mg L�1

[Na] 10.7 2.3 mg L�1

[Cl] 21.3 4.5 mg L�1

[NO3] 28.9 9.0 mg L�1

[SO4] 13.4 4.0 mg L�1

Loess thickness 1.7 1.0 mCWF thickness 11.7 7.2 m

D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33 27

3. Results and interpretation

The first phase of this study examines the Upper Normandyregion (the entire area presented: SEA basins + SEINE RIGHTbasins + SEINE LEFT basins, and the second phase focuses on theSEA basins given that the processes in these basins are easier tounderstand because they are not disturbed by Perche Sands orTertiary deposits.

3.1. Upper Normandy region

3.1.1. Geochemical spatial distributionsGroundwater of Chalk aquifers is primarily a Ca–HCO3 facies,

but the other ions vary throughout the Upper Normandy region.The mean concentrations and standard deviations are presentedin Table 1 and the matrix correlation in Table 2.

Three maps ([HCO3�], [Cl�] and [Mg2+]) representing the primary

spatial variations in these ions are presented in Fig. 2.The HCO3

� concentration map (Fig. 2b) is similar to the Ca2+ map(Valdes et al., 2006) and represents autochthonous mineralization(ions originating from within the aquifer). The lowest concentra-tions are located in the southern portion of Upper Normandy, in

Fig. 3. (a) Distribution of the number of measurements per site, (b) distribution of the meconcentration calculated for each site, (d) distribution of the standard deviation of Ca2+ coNO3� concentration calculated for each site.

the upstream area of the SEINE LEFT basins. Low concentrationsare found in the western portion of the area north of the SeineRiver.

The Cl� concentration map (Fig. 2c) is quite similar to the Na+,SO4

2�, and NO3� maps (Valdes et al., 2006) and represents allochth-

onous mineralization (ions originating from the surface). The high-est concentrations are located south of the Seine River and in thewestern portion of the area north of the Seine River.

The Mg2+ concentration map (Fig. 2d) shows that the highestconcentrations are located under the Tertiary deposits in the east-ern part of the study area near the Seine River.

3.1.2. PCA of the geochemical dataPrincipal component analysis was performed from the raster

(with a 2000 m lag) of the geochemical data and of the superficiallayers: the thickness of clay-with-flints and the thickness of loess(Fig. 4 and Table 3). The results are very similar to those foundby Valdes et al. (2007) south of the Seine River. The structure ofthe factorial space is strong with two factors explaining 68% ofthe total variance. Factor 1 (F1) (48% of the variance) is interpretedas representing the origin of the ions. F1 is positively correlatedwith the autochthonous ions (HCO3

� and Ca2+) and negatively cor-related with the allochthonous ions (Cl�, Na+, SO4

2�, and NO3�). Fac-

tor 2 (F2) (20% of the variance) may be interpreted as representingthe global mineralization of the groundwater, with a high correla-tion with Mg. The relationships among the thicknesses of thesuperficial layers are not clear. The thickness of clay-with-flintshas a low correlation with the two PCA factors. The spatial hetero-geneity of the groundwater geochemistry is related to many fac-tors described by Valdes et al. (2007): large flows from thePerche Sands in the south explain the low mineralization in thewater there, Tertiary deposits explain the highest Mg concentra-tion, and the structural elements (anticlines and faults) perpendic-ular to the groundwater flow play an essential role.

There are too many interacting factors modifying thegroundwater geochemistry to clearly show the role played by thethickness of the superficial layers (in particular the thickness of

an of Ca2+ concentration calculated for each site, (c) distribution of the mean of NO3�

ncentration calculated for each site and (e) distribution of the standard deviation of

Table 2Correlation matrix of the maps of the geochemical properties of the Chalk groundwater: major ions and the thicknesses of the superficial layers: loess and clay-with-flints: CWF(for the entire study area: Upper Normandy). Significant results (P < 0.05) are given in bold type.

Ca HCO3 Mg Na Cl NO3 SO4 Loess CWF

Ca 1 0.89 �0.12 �0.31 �0.44 �0.37 �0.37 0.60 �0.68HCO3 1 0.16 �0.29 �0.49 �0.53 �0.31 0.51 �0.69Mg 1 0.31 0.23 0.34 0.61 �0.27 �0.05Na 1 0.91 0.47 0.42 �0.16 0.24Cl 1 0.64 0.47 �0.26 0.36NO3 1 0.58 �0.39 0.28SO4 1 �0.51 0.15Loess 1 �0.38CWF 1

Fig. 4. Factorial space of the PCA of the geochemical data of the Chalk groundwaterand the aquifer properties for Upper Normandy.

Table 3Variable loadings (%) for the two main factors F1 and F2 of the PCA performed on thegeochemical properties of the Chalk groundwater: major ions and the thicknesses ofthe superficial layers: loess and clay-with-flints: CWF for the entire study area: UpperNormandy.

Variable loadings F1 F2

Ca 14.8 10.9HCO3 14.4 16.6Mg 2.9 27.6Na 10.4 8.5Cl 14.5 3.4NO3 13.1 3.4SO4 10.9 11.9Loess 10.0 0.9CWF 9.0 16.8

28 D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33

clay-with-flints). Further analysis focused on the SEA basins seemsto be relevant to examine the role of clay-with-flints because thiszone is simple, the chalk formations are covered only with a vari-ably thick clay-with-flints layer (no Perche Sands and no Tertiarydeposits), and the structural features are parallel with the ground-water flow.

3.2. SEA basins

3.2.1. Geochemical spatial distributionsThe spatial distributions of the geochemical content in the

Chalk groundwater and the thicknesses of the superficial layersin the SEA basins are presented in Fig. 5. The eastern and westernportions of the SEA basins contrast significantly. The western por-tion of the SEA basins is characterized by the substantial thickness

of clay-with-flints, high concentrations of allochthonous ions, andlow concentrations of autochthonous ions, whereas the oppositeis observed in the east. The loess thickness is less spatially orga-nized, with thicknesses slightly greater in the western portion ofthe SEA basins.

The correlation matrix shows very high correlations (positive ornegative) between the thickness of clay-with-flints and thegeochemical content (Table 4).

3.2.2. PCA of the geochemical dataA second PCA was performed on the SEA basin data. The struc-

ture of the factorial space is very strong, with two factors explain-ing 83% of the total variance (Fig. 6 and Table 5). F1 represents 70%of the variance and is positively correlated with the allochthonousions and negatively with the autochthonous ions. The location ofthe thickness of clay-with-flints in the factor space is notable:the thickness of clay-with-flints is strongly related to high allo-chthonous contents and low autochthonous contents. F2 accountsfor only 13% of the variance and is primarily correlated with loessthickness. This parameter has low correlation coefficients with thegroundwater contents (Table 4).

3.2.3. Spatial distribution of the karst properties of the Chalk aquiferA map of the karst properties of the Chalk aquifer would be very

helpful, especially a map of the karst surface data that includes, forexample, the location of swallow holes. Unfortunately, we do nothave this kind of data for the Upper Normandy region. We havelocal data, but not a homogeneous (spatial) data set.

Failing that, we propose using the mean turbidity, which pro-vides a rough estimation of the degree of connection of the sam-pling site to the surface. The spatial distribution of the meanturbidity is presented in Fig. 7. The western portion of the SEAbasins is characterized by the highest ‘‘karst connection’’ (highestturbidity), whereas the opposite is observed in the east.

3.2.4. Perched groundwater in the superficial layerPerched groundwater is observed in the clay-with-flints. The

water level, electrical conductivity and chloride concentration ofthe perched groundwater at the Bouville site (Fig. 2a) are presentedin Fig. 8 with the rainfall time series recorded at the Boos Stationfrom February to August 2007.

During the winter, the water level of the perched groundwateris high (almost through the soil), the mineralization is low (lowelectrical conductivity), and the chloride concentration is approxi-mately 15 mg L�1. Beginning in April, the level decreases, and min-eralization increases (electrical conductivity varies from less than400 lS/cm at 25 �C at the beginning of April to approximately550 lS/cm at 25 �C in June); the chloride concentration increasesas well (from 17 mg L�1 in March to 28 mg L�1 in May). Becausechloride is a conservative ion, this increase may be related toevapotranspiration. We can assume that evapotranspirationprocesses in the perched groundwater will affect the other

Fig. 5. Spatial distribution of the geochemical properties of the Chalk groundwater and the Chalk aquifer properties in the coastal basins: SEA basins – projected coordinatessystem: Lambert 2 étendu.

Table 4Correlation matrix of the maps of the geochemical properties of the Chalk groundwater: major ions and the thicknesses of the superficial layers: loess and clay-with-flints: CWFfor the coastal basins (SEA basins). Significant results (P < 0.05) are given in bold type.

Ca HCO3 Mg Na Cl NO3 SO4 Loess CWF

Ca 1 0.89 �0.53 �0.71 �0.73 �0.71 �0.76 �0.25 �0.80HCO3 1 �0.48 �0.57 �0.66 �0.84 �0.86 �0.18 �0.90Mg 1 0.73 0.67 0.68 0.56 0.47 0.59Na 1 0.96 0.67 0.73 0.46 0.60Cl 1 0.77 0.84 0.40 0.71NO3 1 0.91 0.24 0.91SO4 1 0.26 0.87Loess 1 0.28CWF 1

Fig. 6. Factorial space of the PCA of the geochemical data of the Chalk groundwaterand the Chalk aquifer properties for the coastal basins (SEA basins).

Table 5Variable loadings (%) for the two main factors F1 and F2 of the PCA performed on thegeochemical properties of the Chalk groundwater: major ions and the thicknesses ofthe superficial layers: loess and clay-with-flints: CWF, for the coastal basins (SEAbasins).

Variable loadings F1 F2

Ca 11.9 3.1HCO3 12.1 12.0Mg 8.9 13.4Na 11.5 9.2Cl 13.0 2.6NO3 13.3 2.6SO4 13.60 3.0Loess 2.70 49.1RS 13.0 5.0

D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33 29

autochthonous ions, inducing an increase in their concentrationsduring dry periods.

The increasing water level coupled with decreasing mineraliza-tion in June and July are related to the significant rainfall at thistime (250 mm from 11/06/07 to 23/07/07).

Fig. 7. Spatial distribution of the Chalk groundwater turbidity (mean values persite) in the coastal basins: SEA basins – projected coordinates system: Lambert 2étendu.

30 D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33

4. Discussion

The origin and spatial distribution of allochthonous ions aregenerally attributed to atmospheric inputs, to anthropogenic fac-tors such as land use or domestic sewage (Meybeck, 1983;Negrel and Petelet-Giraud, 2005; Sherwood, 1989) and/or to theposition of the river valley (Feast et al., 1998; Hiscock, 1993).

Feast et al. (1998) have shown that the superficial layers of gla-cial deposits on top of chalk in the UK may have an impact on deni-trification processes and thus on nitrate concentrations in the Chalkgroundwater: they found high levels of nitrate (>15 mg/L NO3–N) inthe unconfined valley areas, whereas in confined regions the nitratelevels were low (<0.04 mg/L NO3–N). They showed denitrificationoccurring within the overlying glacial deposits, providing a mecha-nism for naturally improving groundwater quality. Later Rivettet al. (2007) showed that denitrification within the Chalk matrix,although geochemically possible, may not occur, as bacteria arepotentially excluded by the narrow pore throats. They concludedthat denitrification is unlikely to lead to very significant mitigationof the nitrate problems manifested in these aquifers. In our study inthe SEA basins, the NO3 concentration values varied from 15 to42 mg L�1 with the highest values located in the area with thethickest clay-with-flints layer (Fig. 5). This shows there is little orno denitrification, which suggests that the Chalk beneath theclay-with-flints is well oxygenated and consequently that clay-with-flints is permeable (probably a heterogeneous permeability

Fig. 8. Time series of rainfall at the Boos station and water level, electrical conductivity a2007 to July 2007).

with preferential pathways as described by Jardani et al. (2006))and may confirm that the Chalk aquifer is really unconfined evenunder a thick layer of clay-with-flints.

Orban et al. (2010) presented a spatial study of nitrate and tri-tium concentrations in the Chalk groundwaters in the Geer basin ofeastern Belgium. The geological settings of this area are similar tothe Upper Normandy region: the chalk is covered by a maximum of10 m of flint conglomerate resulting from the dissolution of thechalk, several meters of Tertiary sand deposits, and Quaternaryloess from 2 to 20 m thick. They showed that the spatial distribu-tion of the contamination in the Geer Basin is linked to the hydro-dynamic conditions prevailing in the basin, more precisely to theage and mixing of the groundwater, and not to the spatial patternsof land use or to local hydrodispersive processes. Knowledge ofthese hydrodynamic and recharge processes is therefore essential.

The recharge processes in Chalk aquifers have been widelystudied. There are three types of recharge: by karst conduits, frac-tures, or matrix. Most studies agree that the dominant recharge isensured by the transfer through the matrix, with a velocity lessthan 1 m a�1 (Smith et al., 1970; Barraclough et al., 1994; Hariaet al., 2003; Mathias et al., 2005; Ireson et al., 2009). The preferen-tial transfers through the fractures and the karst conduits are fas-ter, with velocities as much as a few hundred meters per hour(�600 m/h in the Paris Basin (Nebbache, 1999) and �250 m/h inthe London Basin (Maurice et al., 2006)). The primary controls onpreferential recharge to Chalk aquifers are the characteristics ofrainfall events, in terms of duration and intensity, the physicalproperties of the near-surface, and the antecedent soil moisturenear the surface (Ireson and Butler, 2011).

4.1. Recharge processes controlled by the thickness of clay-with-flints

The 1998 report published by Klinck et al. presents a review ofthe UK studies on the effect of superficial layers on recharge pro-cesses. They conclude that ‘‘the presence of a continuous clay-with-flints cover will reduce the vulnerability of the underlyingChalk aquifer to pollution by a significant extent due to reductioninfiltration rate and of some contaminant species concentrationsthough precipitation, complexation, cation exchange, sorptionreaction and subsequent degradation’’.

In the present study, we observed strong spatial correlationsbetween the thickness of clay-with-flints map and all of the maps

nd chloride concentration in the perched groundwater at the Bouville site (February

Fig. 9. Conceptual models of infiltration (a) with thick clay-with-flints and (b) with thin clay-with-flints.

D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33 31

of major ions (autochthonous and allochthonous) in the Chalkgroundwater, which highlights the significant role played by thethickness of clay-with-flints. The following proposes a conceptualinterpretation of the infiltration and evapotranspiration processesoperating in the unsaturated zone controlled by the clay-with-flints thickness.

Our results in the SEA basins show that thick clay-with-flints isassociated with the highest values of allochthonous ions ground-water concentrations (rCWF/autochthonous ions from 0.60 to 0.91,Table 4), contrary to the conclusions of Klinck et al. (1998). Differ-ent factors can affect concentrations of Cl�, Na+, NO3

�, and/or SO42�

ions in groundwater: anthropogenic activities such as humanactivities on the catchment or agriculture (Edmunds et al., 2003;Edmunds and Smedley, 2000; Kloppmann et al., 1998; Negreland Petelet-Giraud, 2005; Widory et al., 2004) and atmosphericinputs (Meybeck, 1983). There is clear evidence that these factorsplay a role in the study area, but they cannot explain all of ourresults. The spatial distributions of allochthonous ions (Cl�, SO4

2�,Na+, and NO3

�) are very well correlated with one another(R > 0.67, Table 4), although these ions have different origins(Valdes, 2005). This suggests that the allochthonous ions are allcontrolled by a common process. The results at the Bouville siteallow us to assume that this process is evapotranspiration. A thickclay-with-flints layer would have perched groundwater in theclay-with-flints in which evapotranspiration processes mayincrease the allochthonous ion concentrations.

The spatial correlations between the concentrations of autoch-thonous ions (HCO3

� and Ca2+) and the thickness of clay-with-flintsare highly negative (Table 4) (i.e., the lowest concentrations ofautochthonous ions occur where the superficial layer is thickest),suggesting that the rates of the dissolution processes are lowestwhere the clay-with-flints is thickest. In this thickest part ofclay-with-flints, the infiltration processes would most likely bemore concentrated within the input karst (allowing lesswater–rock interaction) and be delayed (because of the residencetime in the perched groundwater). Moreover, the spatialcomparison between the clay-with-flints, geochemistry and tur-bidity (Figs. 5 and 7) gives us new highlights on the infiltrationprocesses:

– Thick clay-with-flints is associated with the highest turbidityvalues: the Chalk aquifer is probably more connected to the sur-face via karst systems allowing concentrated infiltration duringintense rain events via developed karst conduits (resulting inturbidity at the outlets, Fig. 7) and delayed infiltration via theperched groundwater in the clay-with-flints.

– Thin clay-with-flint is associated with the lowest turbidity: theChalk aquifer is less connected to the surface, the karst is prob-ably less developed with small conduits or karstified fractures).It can be deduced that under a thin or nonexistent clay-with-flints layer, the infiltration seems to be more diffuse with morewater–rock interaction than in the first case.

These results and interpretations allow two models of rechargefrom the surface to the Chalk groundwater to be proposed, depend-ing on the thickness of the clay-with-flints layer (Fig. 9):

� In the case of thick clay-with-flints (Fig. 9a), the storage in theperched groundwater increases the rate of evapotranspiration.The infiltration of the perched groundwater is delayed and con-centrated. These processes involve high concentrations of allo-chthonous ions and low concentrations of autochthonous ions.Moreover, during intense rain events, surface runoff and directconcentrated infiltration occurs through the sinkholes.� In the case of thin clay-with-flints (Fig. 9b), the infiltration is

more diffuse, allowing less evapotranspiration than in the firstcase. These processes involve low concentrations of allochtho-nous ions, more this would induce more significant water–rockinteraction and thus more dissolution and so high concentra-tions of autochthonous ions.

In the past, before the study of Klinck et al. (1998) in which theymeasured the hydraulic conductivity of the clay-with-flints vari-able from 10�9 m s�1 to 10�5 m s�1, clay-with-flints had long beenconsidered a homogeneous, relatively impervious layer, whichsuggests either that this layer protects the Chalk aquifer or in con-trast that it concentrates the surface water, suggesting that theChalk aquifer is highly vulnerable. Using self-potential measure-ments, Jardani et al. (2006) provided evidence of preferential path-ways and storage zones in clay-with-flints and showed thepresence of sinkholes and crypto-sinkholes, which cause muchgreater permeability than classical clays. It is likely the presenceof the flints in this layer that induces great heterogeneity, allowingpreferential flows.

5. Conclusion

In this study, we have conducted a thorough spatial investiga-tion of groundwater. The ability to map the data (geochemical datain this case) and to compare them with other spatial data brings outevidence of relationships between the various parameters andenables us to speculate on the processes occurring in these

32 D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33

underground layers. Currently, many environmental databaseswith spatial properties exist but have not been sufficientlyexploited: the spatial analyses and correlations are a good way tostudy them.

The use of the French groundwater database demonstrates thatthe geochemical properties of the Chalk aquifer in Upper Nor-mandy are spatially well organized. Two types of groundwaterare spatially observed, corresponding to the highest concentrationsof autochthonous ions (HCO3

� and Ca2+) and the highest concentra-

tions of allochthonous ions (NO3�, SO4

2�, Mg2+, Na+, and Cl�).The comparison of the geochemical maps of the Chalk ground-

water with the maps of the thicknesses of the superficial layers,especially the thickness of clay-with-flints, highlights the strongcorrelations between them. The greatest thicknesses of clay-with-flints are related to (1) the highest concentrations of allochth-onous ions and the lowest concentrations of autochthonous ions,and (2) the most elevated turbidity values related to a well-devel-oped karst system. These results allow us to propose two types ofinfiltration process:

– Thicker clay-with-flints allows storage in the perched ground-water, which allows evapotranspiration, resulting in high con-centrations of allochthonous ions and is related to lowconcentrations of autochthonous ions. The infiltration of theperched groundwater is thus delayed and concentrated viadeveloped karst.

– Thinner clay-with-flints causes the infiltration to be more dif-fuse, but with less evapotranspiration than in the first case(no perched groundwater near surface allowing evapotranspira-tion) and thus low concentrations of allochthonous ions in theChalk groundwater. Moreover, the less developed karst underthin clay-with-flints causes more significant water–rock inter-action and thus more dissolution and higher concentrations ofautochthonous ions in the Chalk groundwater.

One of the questions of this study was ‘‘Does the clay-with-flintconcentrate the infiltrated water?’’ Our results seem to show thatthe answer is yes. The map of the turbidity in the groundwatersis spatially correlated to the clay-with-flints thickness, which rein-forces this idea. To give more robustness to these conclusions, wewould have to build a detailed map of the karst surface features: anexhaustive map of the swallow holes would be highly valuable.

Another question was ‘‘Does the clay-with-flints protect theaquifer?’’ The answer given by this study’s results is actually thatit does not particularly protect the aquifer, this layer is related toa high autochthonous concentration in the groundwater; however,it would delay the infiltration with a temporal storage in theperched groundwater, except during intense rain events. This leadsto another question: what are the respective proportions for directinfiltration through sinkholes to the karst system during intenserain events and for water stored in the perched aquifer and infil-trated later?

These results are very important for the protection of groundwa-ter and land use policies. Many vulnerability estimation methodsexist for carbonate groundwater (Edmonds, 2008; Plagnes et al.,2010), and the protection of the aquifer is often the most importantcriterion. The role played by the superficial layers in the infiltrationprocesses is crucial and needs to be examined in depth.

References

Barraclough, D., Gardner, C.M.K., Wellings, S.R., Cooper, J.D., 1994. A tracerinvestigation into the importance of fissure flow in the unsaturated zone ofthe British Upper Chalk. J. Hydrol. 156 (1–4), 459–469.

Ben Othman, D., Luck, J.-M., Tournoud, M.-G., 1997. Geochemistry and waterdynamics: application to short time-scale flood phenomena in a small

Mediterranean catchment: I. Alkalis, alkali-earths and Sr isotopes. Chem.Geol. 140 (1-2), 9–28.

Brouyère, S., 2006. Modeling the migration of contaminants through variablysaturated dual-porosity, dual-permeability chalk. J. Contam. Hydrol. 82 (3–4),195.

Calba, F., 1980. Hydrogéologie du karst crayeux du pays de Caux (France). Etude dedeux bassins. Université de Rouen, Rouen, pp. 189.

Calba, F., Charrière, G., Conrad, G., Lefebvre, D., Rodet, J., 1979. Relations entre ledéveloppement du karst de la craie, la dynamique et la qualité des eauxsouterraines du Pays de Caux. Bull. Trimest. Soc. Géol. Normandie. Amis. Mus.Havre. 66 (4), 45–68.

Dussart-Baptista, L., Massei, N., Dupont, J.-P., Jouenne, T., 2003. Transfer of bacteria-contaminated particles in a karst aquifer: evolution of contaminated materialsfrom a sinkhole to a spring. J. Hydrol. 284 (1–4), 285–295.

Edmonds, C., 2008. Improved groundwater vulnerability mapping for the karsticChalk aquifer of south east England. Eng. Environ. Prob. Karst, Nat.Anthropogenic Hazards Karst Areas 99 (3–4), 95.

Edmunds, W.M., Smedley, P.L., 2000. Residence time indicators in groundwater: theEast Midlands Triassic sandstone aquifer. Appl. Geochem. 15 (6), 737–752.

Edmunds, W.M., Shand, P., Hart, P., Ward, R.S., 2003. The natural (baseline) qualityof groundwater: a UK pilot study. Sci. Total Environ. 310 (1–3), 25–35.

Eisenlohr, L., Bouzelboudjen, M., Kiraly, L., Rossier, Y., 1997. Numerical versusstatistical modeling of natural response of a karst hydrogeological system. J.Hydrol. 202 (1–4), 244–262.

Feast, N.A., Hiscock, K.M., Dennis, P.F., Andrews, J.N., 1998. Nitrogen isotopehydrochemistry and denitrification within the Chalk aquifer system of northNorfolk, UK. J. Hydrol. 211 (1–4), 233–252.

Ford, D., Williams, P., 1989. Karst Geomorphology and Hydrology. Unwin Hyman,London, pp. 601.

Goovaerts, P., Sonnet, P., Navarre, A., 1993. Factorial kriging analysis of springwatercontents in the Dyle river basin, Belgium. Water Resour. Res. 29 (7), 2115–2125.

Haria, A.H., Hodnett, M.G., Johnson, A.C., 2003. Mechanisms of groundwaterrecharge and pesticide penetration to a Chalk aquifer in southern England. J.Hydrol. 275 (1–2), 122–137.

Hauchard, E., Laignel, B., Delahaye, D., 2002. Proposition d’un nouveau schemastructural du Nord-Ouest du bassin de Paris reposant sur l’analyse fractale desreseaux de thalwegs et les donnees recentes de la geologie regionale:proposition of a new structural map of the northwestern Paris Basin, basedon the fractal analysis of the talweg networks and the new results of regionalgeology. C.R. Geosci. 334 (4), 295–302.

Hiscock, K.M., 1993. The influence of pre-Devensian glacial deposits on thehydrogeochemistry of the Chalk aquifer system of north Norfolk. UK. J.Hydrol. 144 (1–4), 335.

Ireson, A.M., Butler, A.P., 2011. Controls on preferential recharge to Chalk aquifers. J.Hydrol. 398 (1–2), 109–123.

Ireson, A.M., Mathias, S.A., Wheater, H.S., Butler, A.P., Finch, J., 2009. A model forflow in the chalk unsaturated zone incorporating progressive weathering. J.Hydrol. 365 (3–4), 244.

Jardani, A., Dupont, J.P., Revil, A., 2006. Self-potential signals associated withpreferential groundwater flow pathways in sinkholes. J. Geophys. Res. 111,B09204. http://dx.doi.org/10.1029/2005JB004231.

Klinck, B.A., Hopson, P.M., Lewis, M.A., Macdonald, D.M.J., Inglethorpe, S.D.J.,Entwisle, D.C., Harrington, J.F., Williams, L., 1998. The hydrogeological behaviorof the clay-with-flints of southern England. British Geological Survey TechnicalReport WE/97/5.

Kloppmann, W., Dever, L., Edmunds, W.M., 1998. Residence time of Chalkgroundwaters in the Paris Basin and the North German Basin: a geochemicalapproach. Appl. Geochem. 13 (5), 593–606.

Laignel, B., 2003. Caractérisation et dynamique érosive de systèmesgéomorphologiques continentaux sur substrat crayeux. Exemple de l’Ouest duBassin de Paris dans le contexte nord-ouest européen. Livret I: Axes etDéveloppement de la Recherche. HDR Thesis, Université de Rouen, pp. 138.

Laignel, B., Quesnel, F., Meyer, R., Macaire, J.-J., 1998a. Relations quantitatives entreles craies à silex et les formations résiduelles a silex de l’ouest du bassin deParis. Geodin. Acta 11 (4), 171–181.

Laignel, B., Quesnel, F., Lecoustumer, M.-N., Meyer, R., 1998b. Variability of the clayfraction of the clay with flints of the western part of the Paris Basin. C. R.l’Academie. Sci., Ser. II A – Earth Planet. Sci. 326 (7), 467–472.

Laignel, B., Quesnel, F., Meyer, R., 2002. Classification and origin of the clay-with-flints of the western Paris Basin (France). Z. Geomorphologie 46 (1), 69–91.

Laignel, B., Dupuis, E., Rodet, J., Lacroix, M., Massei, N., 2004. An example ofsedimentary filling in the chalky karst of the Western Paris Basin:characterization, origins and hydrosedimentary behavior. Z. Geomorphologie48 (2), 219–243.

Lautridou, J.-P., 1985. Le cycle périglaciaire pléistocène en Europe du Nord-Ouest etplus particulièrement en Normandie. PhD Thesis, Université de Caen, Caen, pp.908.

Mahler, B.J., Valdes, D., Musgrove, M., Massei, N., 2008. Nutrient dynamics asindicators of karst processes: comparison of the Chalk aquifer (Normandy,France) and the Edwards aquifer (Texas, USA). J. Contam. Hydrol. 98(1–2), 36.

Massei, N., Wang, H.Q., Dupont, J.P., Rodet, J., Laignel, B., 2003. Assessment of directtransfer and resuspension of particles during turbid floods at a karstic spring. J.Hydrol. 275 (1–2), 109–121.

Mathias, S.A., Butler, A.P., McIntyre, N., Wheater, H.S., 2005. The significance of flowin the matrix of the Chalk unsaturated zone. J. Hydrol. 310 (1–4), 62–77.

D. Valdes et al. / Journal of Hydrology 519 (2014) 23–33 33

Maurice, L.D., Atkinson, T.C., Barker, J.A., Bloomfield, J.P., Farrant, A.R., Williams, A.T.,2006. Karstic behavior of groundwater in the English Chalk. J. Hydrol. 330 (1–2),63.

Meybeck, M., 1983. Atmospheric inputs and river transport of dissolved substances.Dissolved loads of rivers and surface water quantity/quality relationships. In:Proc. Hamburg symposium, 1983, Published by International Association ofHydrological Sciences, IAHS Publication 141, Editors Webb B.W., pp. 173–192.

Nebbache, S., 1999. Communautés et contamination bactérienne associées à laturbidité des eaux d’un aquifère karstifié en domaine rural. Approchepréliminaire dans la bioremédiation des nitrates. PhD Thesis, Université deRouen, Rouen, pp. 148.

Negrel, P., Petelet-Giraud, E., 2005. Strontium isotopes as tracers of groundwater-induced floods: the Somme case study (France). J. Hydrol. 305 (1–4), 99–119.

Orban, P., Brouyère, S., Batlle-Aguilar, J., Couturier, J., Goderniaux, P., Leroy, M.,Maloszewski, P., Dassargues, A., 2010. Regional transport modeling for nitratetrend assessment and forecasting in a Chalk aquifer. J. Contam. Hydrol. 118 (1–2), 79.

Plagnes, V., Kavouri, V., Huneau, F., Fournier, M., Jaunat, J., Pinto-Ferreira, C., Leroy,B., Marchet, P., Dörfliger, N., 2010. PaPRIKa, the French multicriteria method tomap the intrinsic vulnerability of karst water resource and source – twoexamples (Pyrénées, Normandy). In: Andreo, B. et al. (Eds.), Advances inResearch in Karst Media. Springer, Berlin (Heidelberg), pp. 323–329.

Price, M., Low, R.G., McCann, C., 2000. Mechanisms of water storage and flow in theunsaturated zone of the Chalk aquifer. J. Hydrol. 233 (1–4), 54–71.

Quesnel, F., 1997. Cartographie numérique en géologie de surface – application auxaltérites à silex de l’ouest du Bassin de Paris. PhD Thesis, Université de Rouen,BRGM Ed., Orléans, pp. 255.

Reisenhofer, E., Adami, G., Barbieri, P., 1998. Using chemical and physicalparameters to define the quality of karstic freshwaters (Timavo river, North-Eastern Italy): a chemometric approach. Water Res. 32 (4), 1193–1203.

Rivett, M.O., Smith, J.W.N., Buss, S.R., Morgan, P., 2007. Nitrate occurrence andattenuation in the major aquifers of England and Wales. Q. J. Eng. Geol.Hydrogeol. 40 (4), 335.

Rodet, J., 1992. La craie et ses Karsts. CNEK-Groupe Seine, Centre deGéomorphologie du CNRS, Caen, p. 560.

Rodet, J., 1999. Tectonic network as the initial factor of karstification of the chalklimestones in the Perche hills (Orne, Normandy, France). Geodin. Acta 12 (3–4),259–265.

Sherwood, W.C., 1989. Chloride loading in the South Fork of the Shenandoah River,Virginia, USA. Environ. Geol. (Hist. Arch.) 14 (2), 99–106.

Smart, P.L., Friederich, H., 1986. Water movement and storage in the unsaturatedzone of a maturely karstified carbonate aquifer, Mendip Hills, England. In:Proceedings of the Environmental Problems in Karst Terranes and theirSolutions Conference, KY, USA, pp. 59-87.

Smith, D.B., Wearn, P.L., Richards, H.J., Rowe, P.C., 1970. Water Movement in theUnsaturated Zone of High and Low Permeability Strata using Natural Tritium.Isotope Hydrology. Atomic Energy Agency, Vienna, 73–87.

Valdes, D., 2005. Variabilité spatiale du fonctionnement hydrologique de l’aquifèrekarstique de la craie (Département de l’Eure, France). PhD Thesis, Université deRouen, Rouen, pp. 288.

Valdes, D., Dupont, J.-P., Massei, N., Laignel, B., Rodet, J., 2005. Analysis of karsthydrodynamics through comparison of dissolved and suspended solids’transport. C. R. Geosci. 337 (15), 1365–1374.

Valdes, D., Dupont, J.-P., Laignel, B., Massei, N., Ogier, S. and Leboulanger, T. 2006.Spatial variability of the geochemistry of a Chalk aquifer at a regional scale: roleof the structural context (Haute-Normandie, France). In: Colloque InternationalGestion Des Grands Aquifères 30 Mai–1er Juin 2006, Dijon,France = International Symposium Aquifers Systems Management: 30 May–1June 2006, Dijon, France. Oral Communication.

Valdes, D., Dupont, J.-P., Massei, N., Laignel, B.T., Rodet, J., 2006. Investigation ofkarst hydrodynamics and organization using autocorrelations and T-[Delta] Ccurves. J. Hydrol. 329 (3–4), 432–443.

Valdes, D., Dupont, J.-P., Laignel, B., Ogier, S., Leboulanger, T., Mahler, B.J., 2007. Aspatial analysis of structural controls on karst groundwater geochemistry at aregional scale. J. Hydrol. 340 (3–4), 244–255.

Van den Daele, G.F.A., Barker, J., Connell, L., Atkinson, T., Darling, W., Cooper, J.,2007. Unsaturated flow and solute transport through the Chalk: tracer test anddual permeability modeling. J. Hydrol. 342 (1–2), 157.

Wang, Y., Ma, T., Luo, Z., 2001. Geostatistical and geochemical analysis of surfacewater leakage into groundwater on a regional scale: a case study in the Liulinkarst system, northwestern China. J. Hydrol. 246 (1–4), 223–234.

White, W.B., 1988. Geomorphology and Hydrology of Karst Terrains. OxfordUniversity Press, New York, pp. 464.

Widory, D., Kloppmann, W., Chery, L., Bonnin, J., Rochdi, H., Guinamant, J.-L., 2004.Nitrate in groundwater: an isotopic multi-tracer approach. J. Contam. Hydrol.72 (1–4), 165–188.