Transfer of water and contaminants in the Chalk unsaturated zone – underground quarry of Saint-Martin-le-Nœud Open Access

The Chalk aquifer is a major resource for water supply in France, Belgium, England, Germany and The Netherlands (Lapworth et al. 2015). In France, where the Chalk of the Paris Basin covers an area of approximately 110 000 km², the aquifer provides close to 12 billion m³ a⁻¹ of water (Lallahem 2002). For several decades, this resource has been threatened by high anthropogenic pressure in the form of intensive agriculture using fertilizers and pesticides, which have caused degradation of the groundwater quality (Brouyère et al. 2004; Baran et al. 2008; Orban et al. 2010). The contamination of drinking water by nitrate and atrazine is recognized as a crucial issue for human health (Comly 1945; Espejo-Herrera et al. 2016).

In the Paris Basin, despite a reduction in the use of nitrates and a ban on atrazine in France in 2003, their concentrations in Chalk groundwater can exceed the authorized French standards for drinking water production of 50 mg l⁻¹ and 0.1 µg l⁻¹ for each individual pesticide (El Gaouzi et al. 2013), while the concentration of nitrate in the unconfined Chalk aquifer is increasing over the majority of the territory. The temporal trend of nitrate content between 2010 and 2017 was positive for 83% of the Paris Basin area with 47% having a trend higher than 0.25 mg l⁻¹ a⁻¹ (Public data from the French ‘Ministère de la Transition Ecologique’).

Most studies of Chalk groundwater deal with the hydrogeochemical responses at the outlets of the aquifer: the springs (natural outlet) or the wells (anthropic outlet) (Kloppmann et al. 1994; Valdes et al. 2005, 2006; Stuart and Smedley 2009; Orban et al. 2010; Ascott et al. 2016; Cao et al. 2020). The observations made in these studies concern processes both in the saturated zone (SZ) and the unsaturated zone (UZ), and it is difficult to separate them in order to understand specifically the processes occurring only in the UZ of the Chalk aquifers.

The underground quarry of Saint-Martin-le-Nœud (Fig. 1) is an exceptional site for studying the transfer of water and contaminants in the UZ of Chalk. It permits entry to the aquifer to analyse directly the chemical evolution and the flow variability of groundwater as it reaches the water table.

This underground quarry was dug in Chalk at the limit between the UZ and the SZ and is located in the Upper Cretaceous Chalk layer of the Paris Basin. The water table is intercepted in the deeper parts of the quarry, in the form of underground lakes. The water percolating at the ceiling of the quarry is interstitial solution circulating through the UZ. This set-up gives direct access to the water at the base of the UZ and the top of the SZ, and the behaviour of both compartments can therefore be studied. The quarry is dug to a maximum depth of 31 m and a layer of clay-with-flints (CwF) of varying thickness covers the Chalk. The surface is used as agricultural land on which nitrate fertilizers have been applied since the 1950s and atrazine pesticides from 1977 to 2003.

The first hydrogeological investigations in the quarry started in 2012 with measurements of the underground lakes and percolating water along the quarry, including: time series of percolating flow, lake water levels, temperature and electrical conductivity, as well as discrete measurements of major ions, pesticide concentrations and tritium content (Barhoum 2014; Barhoum et al. 2014; Chen 2019; Chen et al. 2019a, b).

The long-term study of Saint-Martin-le-Nœud provides a better understanding of and new insights into the Chalk transfer processes into the UZ, mainly those concerning water and contaminant transfer. Insight has been obtained into the following:

1. The local variability in groundwater chemistry upon reaching the water table. Spatial and temporal heterogeneities of Chalk groundwater chemistry have been widely studied. The chemical variability of water is explained by factors such as local variation in rock mineralogy (Edmunds et al. 2003; Kloppmann et al. 1998) or structural controls (Valdes et al. 2007), and human pollution especially due to agricultural activities (Kloppmann et al. 1994; Hakoun et al. 2017), as well as mixing between waters of different residence times and origins including rapid fissure water and matrix water (Gillon et al. 2010). However, most of the studies are performed at large scale (not local scale) and typically concern concentrations beneath the water table, rather than at the point at which UZ groundwater reaches the water table.

2. The balance of fracture v. matrix flow. The Chalk is a complex carbonate rock with double porosity: matrix and fractures – eventually a triple porosity in karst areas with the presence of conduits (Barker and Foster 1981). The matrix is characterized by a low permeability of approximately 10⁻⁸ m s⁻¹ and a high total porosity from 20 to 40%, while the fractures are characterized by a higher permeability in the range of 10⁻⁵–10⁻⁶ m s⁻¹ (Megnien 1978; Bloomfield 1996) and a low porosity of less than 1%. Water transfer processes occur in both the matrix and fractures. A two-way transfer process of water and contaminants characterizes the Chalk aquifer: ‘piston flow’ (transfer of energy only: the old water is displaced by a new input of water) and ‘direct transfer’ (transfer of water) occurring within fractures and fissures. The water velocity is much lower in the piston-flow case than in the direct-transfer case. The transfer processes and the proportions and exchanges between the matrix and fissures have been widely studied but remain difficult to define. Most studies consider that the transfer mainly occurs through the matrix (Mathias et al. 2005; Cooper et al. 2006). The transfer through fractures may be active under specific conditions, such as high rainfall intensity and high soil moisture (Haria et al. 2003; Ireson et al. 2009; Ireson and Butler 2011).

3. The transport of contaminants. Some models have been developed to simulate nitrate and atrazine transfer in the Chalk aquifer (Brouyère 2006; Gutierrez and Baran 2009; Orban et al. 2010; Smith et al. 2010; Wang et al. 2012; Ascott et al. 2016), but the lack of knowledge of the processes occurring in the UZ remains a limitation. A few studies have directly observed nitrate profiles in the Chalk UZ, obtaining transfer velocities of the order of 0.5–1 m a⁻¹, and therefore estimating the arrival of nitrates in the water table in the next few years, at most a few decades (Foster et al. 1982; Gutierrez and Baran 2009; Surdyk et al. 2021). Atrazine is more complex to study because of its susceptibility to degradation and sorption processes. This contaminant moves with the water by advection but can be affected, mainly in the UZ, by several other processes depending on its physico-chemical and soil properties (Chefetz et al. 2004). Moreover, due to degradation processes, the main degradation product of atrazine, deethylatrazine (DEA), was the most frequently detected substance in Chalk groundwater of the Paris Basin over the period 2012–17 (AESN 2020). According to Issa and Wood (1999) and Johnson et al. (2000), the degradation processes mainly occur in the soil, but they could also occur in the UZ at a lower rate. The need for improved knowledge on the contaminant transport processes is a major issue in hydrogeology.

4. The role of the clay-with-flints cover on the water and contaminant transfer in the Chalk aquifer. The CwF results primarily from the weathering of Chalk (Laignel et al. 1999). It may be thick (up to 40 m in Upper Normandy) or thinner (a few metres) and even absent in the Paris Basin (Valdes et al. 2014). In the past, this layer was considered to be impervious. Later, Klinck et al. (1998) measured in situ the hydraulic conductivity of CwF and found greatly variable values that may be high locally: from 10⁻⁴ to 10⁻⁹ m s⁻¹. Using self-potential measurements, Jardani et al. (2006) provided evidence of preferential pathways and storage zones in CwF and showed the presence of sinkholes and crypto-sinkholes, which cause much higher permeability than that of classic clays. At the regional scale of Upper Normandy, Valdes et al. (2014) showed a thick CwF layer above karstic Chalk, which may lead to perched groundwater and local infiltration via developed karst inducing locally preferential pathways with concentrated infiltration, while in the case of a thin CwF layer, the infiltration is more diffuse.

The Saint-Martin-le-Nœud underground Chalk quarry belongs to the French National Observatory Service of Karst – (namely, SNO Karst; Jourde et al. 2018), part of the national research infrastructure eLTER-France OZCAR (the French network of Critical Zone Observatories; Gaillardet et al. 2018). It is located in the Haut-de-France region in the northern part of the Paris Basin (5 km south of Beauvais, 70 km north of Paris, Fig. 1). The quarry was exploited for Chalk blocks by the room-and-pillar method from the twelfth to the eighteenth centuries, and is made up of corridors, galleries and rooms. It is about 1200 m long and 150 m wide (Fig. 2).

The quarry belongs to the eastern part of the Bray anticline where the Chalk strata dip approximately 12° NNE (Fig. 1b, c). It was dug into the Upper Cretaceous Chalk layer dating from the Coniacian and Turonian (Fig. 1c). In this study area, the Chalk aquifer may be considered as a double-porosity aquifer with matrix and fractures, as field observations give no evidence of karst conduits, only minimal dissolution in some fractures. The total porosity of the Chalk in the quarry has been measured as approximately 40% (Lafrance 2016), the fracture porosity is estimated at around 1% (Megnien 1978). A layer of CwF covers the quarry Chalk with a variable thickness from nearly 0 to about 10 m (Barhoum et al. 2014).

At a depth of 18–31 m (Barhoum 2014), the quarry gives direct access to the limit between the UZ and the SZ. The lowest parts of the cavity lie at the water table, forming a series of underground lakes and water percolates at the ceiling of the quarry. The study focuses on 16 sites: the lake and associated percolation areas (Fig. 2). The list of the sites (name and three-letter site code) is given in Table 1.

The percolating water comes directly from the UZ and flows from the ceiling of the lowest parts of the quarry, most often above the lakes, and forms 4–5 cm-long stalactites. Percolation is variable in time and space. This percolating water is spatially and temporally variable and may be permanent throughout the year in some places (Barhoum et al. 2014).

The quarry is located under a topographic ridge along a SSE/NNW axis, which is the major axis of the anticline (Fig. 1b, c). The underground lakes are aligned in low-lying areas, along the same orientation. The lake water levels correspond to the Chalk piezometric levels given by the hydrogeological map of Beauvais (BRGM 1969), and their temporal variability is very similar to that of the Chalk water table recorded in Beauvais (data exported from the public website ‘Portail ADES’ http://www.ades.eaufrance.fr/). The piezometric map of the Chalk (data provided by SIGES, BRGM) is shown in Figure 1b. Due to the absence of observation wells, there is no piezometric information in the proximity of the quarry. However, the piezometric map indicates that the water table is drained to the NE by the Thérain River. Furthermore, studies injecting dye tracer carried out in 2010 (INERIS 2010) and recently in 2021 in the underground lakes have shown that the lakes also flow toward the Thérain River and no water exchange was observed between the lakes. A cross-section of the piezometric level is proposed in Figure 1c. Between the quarry and the Thérain River, the hydraulic gradient is rather high (about 2.5%), which is quite coherent with the hydraulic gradient of approximately 2% computed along the C–D profile in a small topographic dome similar to that of the quarry (Fig. 1b). In the SW of the quarry, the piezometric gradient is unknown. Transposition of the piezometry of the area around the C–D profile to the quarry area allows us to propose a flow toward the anticline. Thus, the underground lakes are probably on or near the piezometric ridge and are supplied almost exclusively by groundwater flowing vertically through the UZ. The lake waters are therefore assumed to result from a mixture of percolating water dripping from the ceiling of the quarry or water passing directly through the chalky pillars and the walls in the quarry to the water table.

The climate is oceanic, with an average annual temperature of 11°C and an average annual rainfall of 700 mm (data from 1981 to 2010 provided by Météo-France at the Beauvais station). The evapotranspiration ranges from 450 to 500 mm a⁻¹ (Chen 2019).

The soil has been cultivated (corn, wheat and rape) since at least 1950. The agricultural practices (type of cultivation, fertilizers, and pesticide inputs) since the 1970s are known thanks to the farmer and owner of the field who recorded all this information in a notebook. In the field above the quarry, atrazine was used from 1977 to 2003 and nitrogen fertilizers have been widely used in the area since the 1950s (Le Noë et al. 2017).

Firstly, the UZ thickness was determined with an underground GPS survey in 2012 (Barhoum et al. 2014). Secondly, the CwF layer was characterized using geophysical measurements at the surface above the quarry. Electrical or electromagnetic methods energize the subsurface in order to estimate electrical conductivity or its inverse, electrical resistivity. CwF thickness is related to this conductivity because of the electrical properties of clay and water. The electrical conductivity of the soil can be used as a proxy for the CwF layer thickness (Barhoum et al. 2014). The higher the electrical conductivity (or the lower the electrical resistivity), the greater the thickness of the CwF.

Electromagnetic induction (EMI) allows one to map near-surface electrical variation over the entire quarry (McNeill 1980). Two geophysical surveys were carried out: in September 2013 using an EM31 (Barhoum et al. 2014) and in May 2019 using a CMD explorer.

Two-dimensional (2D) electrical resistivity tomography (ERT) is a direct current electrical method investigating the distribution of lateral and vertical electrical resistivity (Griffiths and Barker 1993). It can provide deep 2D imaging of the bulk resistivity (Daily and Owen 1991). The ERT profiles were carried out using a SyscalPro resistivity meter in May 2019.

The EMI map (Fig. 3a) shows a large spatial variability of the surface apparent electrical conductivity with values ranging from 1 to 70 mS m⁻¹ (from 1000 to 14.3 Ω m in terms of apparent electrical resistivity). The most conductive areas correspond to higher clay content in the soil and thus wetter subsurface. On the other hand, the less conductive areas are interpreted as having a thinner surface layer of CwF (eventually no clay). The electrical conductivity was determined for each of the 16 sites of the quarry (Table 1).

According to the differences in UZ characteristics, CwF thickness and UZ thickness (Table 1), the quarry can be divided into three areas and the sites separated into three groups:

• Group 1 (G1): CAS, CHA, BLE, PIL, JAR and LIT. These sites have the highest UZ thickness (around 30 m) with a thin layer of CwF.

• Group 2 (G2): NAT, TSA, STA, LUC, VOU, NIC and PI2. These sites have a high UZ thickness (around 28 m) with a thick layer of CwF.

• Group 3 (G3): SQU, CAR and PED. These sites have the lowest UZ thickness (around 20 m) with a thin layer of CwF.

Measurements and sampling of the percolating water and the underground lakes of the 16 sites were performed. The study of the lakes began in 2012 and that of the percolating waters during 2014–15. Field campaigns were conducted approximately every 2 months from 2012 to 2020.

The 16 lakes were sampled and electrical conductivity, temperature and water level were measured in situ. In several lakes, time series of the water level, electrical conductivity and temperature were recorded using CTD (conductivity, temperature, depth)-Diver probes.

Sampling from each of the 16 percolation sites was achieved using 250 ml beakers installed under the quarry ceiling. Percolating volume, electrical conductivity and temperature were measured in situ. In several sites, time series of percolating flow were recorded using rain gauges. Due to technical problems and equipment failure, the time series may have gaps in the records.

From April 2016 to September 2017, seven complete campaigns were carried out: 12 April 2016, 6 June 2016, 23 June 2016, 5 September 2016, 9 January 2017, 27 February 2017 and 5 September 2017. The dataset ‘April 2016–September 2017’ is composed of 94 samples for the lakes and 97 for the percolating water for the 16 sites. Data are available on the water level of the lakes, percolating flow, and geochemical analysis of major ions (including nitrate) and pesticides for both lakes and percolating waters.

Major ions (Ca²⁺, K⁺, Na⁺, Mg²⁺, $NO3−$⁠, Cl⁻, $SO42−$⁠) were analysed using ion chromatography (Dionex ISC 3000). The accuracy is 1 mg l⁻¹ with 5% uncertainty. These analyses were performed for all samples. Alkalinity (⁠$HCO3−$⁠) was measured by acid titration. It was performed for all lake samples and for about half of the percolation samples. Calcite precipitation was observed in some of the beakers sampling the percolation water, and alkalinity was not measured for these samples. When alkalinity was available, the ionic balance was calculated.

Atrazine and its metabolite deethylatrazine (DEA) were analysed through solid phase extraction LC-MS/MS (liquid chromatography mass spectrometry) (Agilent Technology) using only 2 ml water samples (Blanchoud et al. 2020). The quantification limit was 20 ng l⁻¹ with 20% uncertainty. These analyses were performed on lake and percolating water, in April 2013 at the Bleu and Pedro sites only, from April 2016 to September 2017 at all 16 sites, and thereafter only at eight selected sites including Bleu, Lucarnes and Pedro. The atrazine degradation ratio (DAR) is defined as DAR = [DEA]/[Atrazine]. It is an indicator of the atrazine degradation processes. The higher the DAR, the greater the degradation processes.

The tritium content in groundwater may be used to estimate the residence time of the water. It was determined using the electrolytic enrichment and liquid scintillation counting methods with a standard deviation measured for each sample (Thatcher et al. 1977) at the University of Avignon Emmah Laboratory. Tritium analyses were performed for ten sites including Bleu, Lucarnes and Pedro for the percolation water in April 2016 and for the lakes in March 2016. The tritium content in groundwater may be used to estimate the age of the water using piston flow models (Maloszewski and Zuber 1996; Chen et al. 2019a).

The geochemical and hydrodynamic data collected for groundwater (lakes and percolating water) are presented; first, the spatial approach, for the 16 sites based on 2016–17 data, and second, the temporal approach, on the longest time series, i.e. 2012–20, for three sites, one per group: Bleu for Group 1, Lucarnes for Group 2 and Pedro for Group 3 (Figs 2 & 3, Table 1).

The data concern both percolation water and underground lakes. The spatial variation of water quality as well as of the hydrodynamics of percolating water has already been presented and discussed by Chen et al. (2019a).

The main statistics of the geochemical parameters – mean, minimum, maximum and standard deviation (SD) – for the 16 lakes and the 16 percolation waters of the ‘April 2016–September 2017’ dataset are shown in Table 2.

The physical and hydrochemical characteristics (normalized lake level – mean centred, electrical conductivity of water, Ca²⁺, $NO3−$⁠, Cl⁻, Mg²⁺, atrazine, DEA concentrations and the DAR) of groundwater (UZ and SZ) are presented in Figure 4 for the 16 sites along the quarry.

Three main points can be highlighted:

• The percolating water and lake water have different characteristics: the mean electrical conductivity of the percolating water is 483 ± 159 µS cm⁻¹, while that of the lake water is 554 ± 77 µS cm⁻¹. At some of the sites, the percolating water is locally more mineralized than the associated lake (e.g. Lucarnes), whereas at other sites the lake water is more mineralized than the percolation water (e.g. Bleu).

• Some water reaches extremely high levels of agricultural contamination, with the highest contamination measured in percolation water. For example, the maximum nitrate concentration in percolating water for Carrefour is 145 mg l⁻¹ and the atrazine concentration in Niches is about 15 000 ng l⁻¹.

• Spatially, the geochemical properties of groundwater are different. For percolating water, the electrical conductivity varies from 249 to 825 µS cm⁻¹. For lakes, it varies from 276 to 768 µS cm⁻¹. Even if the range of values is similar between lakes and percolating water, the spatial variability is greater in percolating water than in lakes (the mean SD of all the parameters in Table 2 is 87% for percolating water and 60% for lakes). This spatial variability was observed for all chemical elements. In a more detailed way, the three groups can be characterized as follows.

  • Group 1, the highest UZ thickness (around 30 m) with a thin layer of CwF: The annual water level variation of the lakes is 0.60 m on average. The lake water is more mineralized than the percolating water. All chemical species are more concentrated in the lake water than in percolating water. Agricultural water contamination is low in percolation water. In the lakes, average concentrations (mean of all the samples) are 33.8 ± 4.2 mg l⁻¹ of nitrate, 231 ± 462 ng l⁻¹ of atrazine and 1013 ± 1318 ng l⁻¹ of DEA; in the percolating water, average concentrations are 21.2 ± 10.4 mg l⁻¹ of nitrate, 28 ± 28 ng l⁻¹ of atrazine and 58 ± 95 ng l⁻¹ of DEA. Lake Jardin has the highest contamination of Group 1 with a maximum nitrate concentration of 42 mg l⁻¹, a maximum atrazine concentration of 2100 ng l⁻¹ and a maximum DEA of 4200 ng l⁻¹. Chloride and magnesium concentrations are low (around 15 and 2 mg l⁻¹, respectively). The DAR of Group 1 is about 3 for lakes and percolation water.

  • Group 2, a high UZ thickness (approximately 28 m) with a thick layer of CwF: The annual water level variation of the lakes is 0.43 m on average. Very low variations for Lake Lucarnes and Lake Voutes (0.11 and 0.17 m, respectively, during the 2016–17 period) are observed. All the water in Group 2 is highly contaminated, with a generally higher contamination in the percolation water than in the lakes. In the lakes, average concentrations are 39.4 ± 9.3 mg l⁻¹ of nitrate, 263 ± 296 ng l⁻¹ of atrazine and 1621 ± 2099 ng l⁻¹ of DEA; in the percolating water, average concentrations are 48.6 ± 22 mg l⁻¹ of nitrate, 2191 ± 3396 ng l⁻¹ of atrazine and 6639 ± 11 913 ng l⁻¹ of DEA (mean and standard deviation are calculated for all the samples of percolating and lake water of the group). Atrazine and DEA have very high concentrations, especially in percolation water, with a maximum for atrazine at Niches of 15 200 ng l⁻¹ and a maximum for DEA at Voutes of 52 000 ng l⁻¹. DAR is the highest in this group, especially in the lakes, with a maximum of 42 for Lucarnes and 21 for Voutes. Chloride concentrations are high with a maximum in percolation water of 54 mg l⁻¹ for Lucarnes. The highest magnesium concentrations were observed for Group 2 with an average of 2.9 mg l⁻¹.

  • Group 3, the lowest UZ thickness (around 20 m) with a thin layer of CwF: The water level variation of the lakes is 0.88 m on average. The water of this group is highly contaminated, with higher concentrations in the percolation water. In the lakes, average concentrations are 52.5 ± 30.9 mg l⁻¹ of nitrate, 287 ± 281 ng l⁻¹ of atrazine and 901 ± 720 ng l⁻¹ of DEA; in the percolating water, average concentrations are 105.2 ± 30.3 mg l⁻¹ of nitrate, 1233 ± 1460 ng l⁻¹ of atrazine and 2773 ± 2534 ng l⁻¹ of DEA. The percolation water of Lake Carrefour (the lowest UZ thickness of the 16 sites) has the highest concentrations of nitrate (144 mg l⁻¹) and chloride (59 mg l⁻¹). Pesticide concentrations are rather high with maxima in percolation water of 4800 ng l⁻¹ for atrazine in Pedro and 12 800 ng l⁻¹ for DEA in Carrefour. In spite of these high concentrations, the DAR is rather constant (around 3) for both the lakes and the percolating water.

The tritium content in the underground lakes (23 March 2016) and in the percolating water (12 April 2016) are presented in Figure 5. The mean tritium content is 9.2 ± 1.1 UT in the lakes and 12.1 ± 1.8 UT in the percolating water. Groundwater sampled in 2016 with concentrations of >10 UT contains a large fraction of water infiltrated in the 1960s.

The lower tritium content in the lakes is evidence that the lake water is composed of younger water than the percolating water. Chen (2019) deduced that the lakes are a mixture of matrix and fissure flow coming from the UZ, while the percolating water is more representative of the matrix water (described in the Discussion).

The temporal approach is presented for three sites: Bleu, Lucarnes and Pedro, respectively belonging to Groups 1, 2 and 3 with different UZ properties (Figs 2 & 3, Table 1). Firstly, the geochemical and hydrodynamic variations are presented from 2012 to 2020 (Fig. 6) and, secondly, a focus on January 2019 to May 2020 is presented for hydrodynamic data (Fig. 7).

Figure 6 shows the rainfall time series at Beauvais station and the hydrodynamic and geochemical variations from 2012 to 2020 for Bleu, Lucarnes and Pedro.

On the Bleu and Pedro sites (Groups 1 and 3 having a thin CwF cover), there is a discontinuity in the percolation season with a shut-off period every year. Percolation starts between January and March and stops between May and August depending on the year. On the Lucarnes site (Group 2 with a thicker CwF cover), percolation is much more continuous. There are small peaks of high-frequency variations in the percolation flow.

The time series of the lake levels correspond with the associated Chalk piezometric variations (data available on ADES database, https://ades.eaufrance.fr), they show annual variations for the three sites, with low levels between December and February and high levels between May and July depending on the site and year. Moreover, the annual lake level amplitude is different on the three sites with an annual variation of about 1 m at Bleu and Pedro (Groups 1 and 3), while it is only about 0.10 m at Lucarnes (Group 2).

The electrical conductivity measurements show seasonal variations of up to 100 µS cm⁻¹. The electrical conductivity of water shows that the ionic content of the groundwater varies between the three sites and that the percolation water and lake water are different: for Bleu, the percolation water is less mineralized than the lake water (perco = 350 µS cm⁻¹, lake = 550 µS cm⁻¹), while for the Lucarnes site, the percolation water is more mineralized than the lake water (perco = 750 µS cm⁻¹, lake = 530 µS cm⁻¹). For Pedro there is very little difference between the two types of water (perco = 680 µS cm⁻¹, lake = 645 µS cm⁻¹). These geochemical differences between lake water and percolation water are found for all major ions as well as for pesticides.

Contamination of quarry groundwater by agricultural input is highly variable, with some water having little contamination (e.g. percolating water at Bleu with $NO3min−=10mgl−1$ and no trace of pesticide). Some other sites reach extremely high levels of contamination: for example, the percolation water of Lucarnes reaching 1000 ng l⁻¹ for atrazine and 5000 ng l⁻¹ for DEA, and the lake water and percolation water at the Pedro site with respective nitrate concentrations from 100 to more than 120 mg l⁻¹.

The concentrations of major ions (Fig. 6) show a long-term evolution between 2012 and 2020, but this evolution depends on the type of water (percolation or lake) and on the site.

• For Bleu (Group 1), between 2014 and 2020, the chloride concentration of the lake water increased from about 12 to 17 mg l⁻¹, and for percolation water it increased from 8 to 22 mg l⁻¹ between 2016 and 2020; the nitrate concentration also increased from 22 to about 40 mg l⁻¹ for the lake water between 2014 and 2020 and from 10 to more than 30 mg l⁻¹ for percolation water between 2016 and 2020.

• For Lucarnes (Group 2), the chloride concentration of the lake was stable at around 25 mg l⁻¹, while it decreased in percolation water between 2016 and 2020 (from 55 to 35 mg l⁻¹). The nitrate concentration of the lake was around 50 mg l⁻¹, with a slight tendency to increase, while the concentration in percolation water decreased from around 75 to 55 mg l⁻¹ between 2016 and 2020. A slight positive trend toward increase may also be observed in the concentration of calcium in the lake.

• For Pedro (Group 3), between 2016 and 2020, the chloride concentration of the lake was relatively stable at around 40 mg l⁻¹, while it decreased from 45 to 40 mg l⁻¹ in percolation water. The nitrate concentration shows a slight decreasing trend around 90 mg l⁻¹ for the lake water between 2012 and 2020, while for percolation water it increased from about 85 to 105 mg l⁻¹ between 2016 and 2020. Calcium concentration does not show any long-term change, except for Pedro lake water, in which calcium increased from 100 to more than 140 mg l⁻¹ between 2012 and 2020.

Pesticide concentrations show very high variations in lakes and percolation water. For the Bleu site, the maxima in the lake for atrazine (140 ng l⁻¹) and DEA (600 ng l⁻¹) were observed in January 2017 and the minima for atrazine (20 ng l⁻¹) and DEA (50 ng l⁻¹) were observed in January 2018. The pesticide concentrations in percolation water are below the limit of detection. For the Lucarnes site, lake and percolation water contained pesticides at high concentrations, with the maxima of atrazine and DEA in the lake at 550 and 4000 ng l⁻¹, respectively, while for percolation water it is 1000 ng l⁻¹ for atrazine and 8000 ng l⁻¹ for DEA. The temporal evolutions of atrazine and DEA concentrations were not correlated, nor do they seem to follow the water level or percolation flow. The Pedro site is also highly contaminated with pesticides, with atrazine maxima of 600 and 350 ng l⁻¹ and DEA maxima of 3000 and 1700 ng l⁻¹ in percolation and lake water, respectively. For the Pedro site, pesticide concentrations in the percolation water and the lake water appear to be correlated with each other but are not directly related to the lake water level or percolation rate.

The temperature measured in the lakes increased with a trend of about 0.05°C per year. This is consistent with the change in global mean surface temperature (IPCC 2018). If this trend were to remain constant until 2100, global warming by 2100 would be in the order of 5°C. Lake Lucarnes and Lake Pedro do not show any high-frequency variation of temperature, while sub-annual variations were observed for Lake Bleu.

The high-frequency time series of hydrodynamic parameters (percolation flow, lake water level), temperature and electrical conductivity of the lakes, for the three sites Bleu, Lucarnes and Pedro, are available between January 2019 and May 2020. Figure 7 shows that the three sites have different behaviours.

For Bleu and Pedro (Groups 1 and 3), the percolation is not permanent. The percolation for Bleu stopped from July 2019 to the end of December 2019, and for Pedro from May 2019 to January 2020. The percolation periods are associated with high lake levels, with an annual variation during 2019–20 of about 1.20 m for Bleu and 0.9 m for Pedro. The time series for Lucarnes (Group 2) showed a permanent percolation that never stopped. The seasonal variation of the lake level (about 0.15 m between January 2020 and April 2020) still showed a slight seasonality.

Variations in temperature and electrical conductivity also show different behaviours between sites (Fig. 7). For Bleu, the electrical conductivity and the lake temperature increased slightly (about 20 µS cm⁻¹ and 0.2°C) at the beginning of the recharge, suggesting a piston effect bringing old water. In March 2020, there was a peak in percolation, a higher increase in the lake level, associated with a decrease of about 0.05°C in temperature and a decrease of about 50 µS cm⁻¹in electrical conductivity. This suggests a recharge by relatively recent water. For Lucarnes, the increase in lake level is associated with an increase in conductivity of about 80 µS cm⁻¹, followed by a decrease of more than 120 µS cm⁻¹ beginning in April 2020. The temperature shows only an increasing trend without annual or infra-annual variation. For Pedro, electrical conductivity and temperature varied very little, with only a slight increase in electrical conductivity (about 15 µS cm⁻¹) at the beginning of the recharge.

The quarry data are very numerous, 16 sites, two types of water (lake and percolation) per site and, for each of them, many parameters, some constant in time (UZ characteristics), and others variable over time (geochemistry, hydrodynamic parameters).

A good way to summarize all these data is to perform a principal component analysis (PCA) of the 16 lakes and the 16 percolation waters for the seven campaigns of 2016–17. This is a multivariate statistical technique used for the reduction of large datasets. This method is commonly used for environmental studies with a high degree of temporal (Ben Othman et al. 1997; Eisenlohr et al. 1997) or spatial variation (Wang et al. 2001; Valdes et al. 2007, 2014).

with H_lake the lake water level, $Hlake¯$ and SD (H_lake) the mean and the standard deviation, respectively, of the lake level (calculated on 94 samples from the 16 sites).

PCA was performed on the UZ characteristics and spatiotemporal data of the groundwater: lake water and percolation water of the 16 sites (dataset ‘April 2016–September 2017’, presented in Fig. 4).

The PCA was performed on 191 individual samples (94 samples of lake water and 97 samples of percolation water) with 11 active parameters: groundwater electrical conductivity, concentrations of major ions (K⁺, Na⁺, Ca²⁺, Mg²⁺, Cl⁻, $SO42−$⁠, $NO3−$⁠), pesticides (atrazine and DEA) and I_hydro, and three supplementary parameters: DAR, UZ thickness and CwF thickness.

The structure of the PCA space is strong with three principal components explaining 63.5% of the total variance (Fig. 8). The factor loadings for PC1, PC2 and PC3 are presented in Table 3 and the PC1–PC2 and PC1–PC3 spaces in Figure 9.

Principal Component 1: PC1 (36.5% of the variance) is positively and strongly correlated with electrical conductivity, and the concentrations of Ca²⁺, Cl⁻, $SO42−$and $NO3−$⁠, positively but more weakly correlated with the concentrations of pesticides, and negatively correlated with UZ thickness. The good correlation between Cl⁻, $SO42−$ and $NO3−$ is a common observation in Chalk groundwater (Valdes et al. 2006, 2014): Cl⁻ and $SO42−$ are often used in fertilizers, along with $NO3−$⁠. The good correlation with Ca could be explained (1) by the increase in carbonate dissolution associated with nitrate inputs or (2) by the farmer's use of fertilizers that combine calcium and nitrate. PC1 may be interpreted as a contamination factor, i.e. in the quarry, the thicker the UZ, the less contaminated the groundwater and, inversely, the thinner the UZ, the more contaminated the groundwater.

Principal Component 2: PC2 (15.9% of the variance) is positively and strongly correlated with the concentrations of K⁺ and Mg²⁺, positively but more weakly correlated with CwF thickness, and negatively but weakly correlated with I_hydro. Laignel et al. (1999), Barhoum et al. (2014), and Valdes et al. (2014) showed that K⁺ and Mg²⁺ can be brought to the Chalk groundwater by the CwF layer, as these clays are mainly composed of smectite, which can contain Na⁺, K⁺ and Mg²⁺ (Deconinck et al. 2005). PC2 may be interpreted as the effect of the CwF thickness on the groundwater: the thicker the CwF layer, the higher the concentrations of Mg²⁺ and K⁺ and the lower the hydrodynamic variations (e.g. low variation in lake level and percolation rate).

Principal Component 3: PC3 (11.1% of the variance) is positively and strongly correlated with the concentration of atrazine and DEA, especially due to values measured in the percolation of Group 2. CwF thickness and Mg also contribute positively but more weakly. PC3 may be interpreted as the contamination by the axis of pesticides. Even if the variance of this axis is low (11.1%), PC3 gives evidence of the relationship between the CwF thickness and the pesticide concentration: the higher the CwF thickness, the higher the pesticide contamination.

The individuals are plotted in the PC1–PC2 and PC1–PC3 spaces (Fig. 9). The distribution of the individuals is consistent with the three groups: Group 1 is in the negative part of PC1 (low contamination) and has low PC2 values (low impact of the CwF); Group 3 is globally in the positive part of PC1 (high contamination) and has low PC2 values (low impact of the CwF); Group 2 is very dispersed in the PC1–PC2 space. The range of contamination is substantial (samples from not contaminated to highly contaminated) and some samples are highly impacted by CwF. The PC1–PC3 space distinguishes the percolating water of Group 2 from other types of water that are heavily contaminated with atrazine and DEA.

The surface inputs of the UZ are (1) rainwater, which is homogeneous at the surface above the quarry, (2) chemical elements from Chalk dissolution or clay–water interaction and (3) anthropogenic inputs used for agriculture, which are considered to be homogeneous on the fields above the quarry (slight differences exist in the quantity of atrazine and nitrate from one field to another). The differences in the quality and hydrodynamics of the groundwater observed in the three groups show that the transfer processes differ depending on the properties of the UZ, among which are the CwF thickness and the UZ thickness. The role of these two parameters will be discussed in the following section.

In Figure 9, the distribution of lake water differs from that of percolating water: the lakes of the three groups are centred in the middle of the factorial space, whereas the percolation water is more scattered. The scattering of individuals representing the percolation water is greater than that of lake individuals.

The underground quarry of Saint-Martin-le-Nœud allows for the investigation of the chemical composition and dynamics of groundwater over a small area of the Chalk aquifer (0.18 km²) in the Paris Basin. The differences observed between the various sites and between lake water and percolating water are surprising if we consider the size of the quarry. The variations may be interpreted in terms of differences in the transfer processes. They also show that the proportion of water flowing through fractures or the matrix may differ.

The Chalk of the study area is described as a double-porosity medium with fractures and a matrix (Barhoum et al. 2014). Many studies have focused on the transfer processes, but the proportion of water in the fractures to that in the matrix, as well as the related velocities, remains difficult to assess. Most studies consider that recharge occurs mainly through the matrix (Mathias et al. 2005; Cooper et al. 2006; Ireson and Butler 2011). Smith et al. (1970) estimated that transfer through the matrix accounts for 70–85% of the total recharge. Some authors consider that transfer through fractures only occurs under specific conditions, such as high rainfall events and high soil moisture (Haria et al. 2003; Ireson and Butler 2011). Other studies observed that flow occurs mainly through fractures in the winter (Keim et al. 2012).

Rapid responses of the water table to rain events are often observed in the Chalk aquifer, which can result from a direct transfer from the surface (infiltration and transfer of rain water from the surface to the water table through the fractures), but also from a ‘piston effect’ mechanism: in this case, the water stored in the matrix of the UZ of Chalk is pushed by pressure and goes down slowly without mixing with overlying recent water (Downing et al. 2005; Ireson et al. 2006).

This section is mainly based on the paper by Chen et al. (2019a), who estimated the age of the percolation water in the Saint-Martin-le-Nœud quarry, thus allowing for the estimation of the transfer velocities in the Chalk UZ based on nitrate, atrazine and tritium (Figs 4 & 5) as dating tracers. Tritium is an inherent constituent of the water molecule, and thus its transfer time corresponds to that of water. Nitrate and atrazine are dissolved molecules, and thus their transfer may be different from that of water, especially for atrazine, which may be delayed by sorption (Coquet 2003). For nitrate and even more for atrazine, the transfer time would be greater than the transfer time of water.

Transfer time was estimated as follows: in oxidizing conditions, the absence of nitrate contamination provides evidence of water infiltrated before 1950. Tritium content up to 10 UT gives evidence of water infiltrated in the 1960s and the presence of atrazine or metabolite demonstrates water infiltrated between 1977 and 2003. Transfer velocities were computed for each environmental tracer (nitrate, atrazine and tritium) by dividing the estimated transfer time by the depth of each site.

The water collected at the ceiling of the quarry, at the base of the UZ: about 1–2 m above the water table, has a residence time of roughly a few decades. A summary of the results for the three groups is given in Table 4. The estimated tritium/water velocities range from 0.37 to over 0.72 m a⁻¹. The three tracers have velocities that may differ from each other, demonstrating that there is actually a mixture of water with different velocities. The results are slightly different for the three groups, with the maximum velocity for Group 2. The transfer velocities are then higher for a thicker CwF layer.

The vertical hydraulic gradient (i) may be estimated at 1, as the pore space is almost all filled by water held by the narrow pore throats. The effective porosity n_e is around 0.4. According to Chen et al. (2019a), a range of velocities from 0.1 to 1 m a⁻¹ is chosen for the calculation. The calculated hydraulic conductivity ranges from 1.3 × 10⁻⁹ to 1.1 × 10⁻⁸ m s⁻¹. These values of hydraulic conductivity clearly correspond to the matrix hydraulic conductivity and are consistent with the literature (Megnien 1978; Price 1987; Bloomfield 1996).

The lake water is a mixture of percolating water (visible at the quarry ceiling) and water flowing in the unsaturated Chalk through the walls and the pillars of the quarry (not visible). Even if the spatial variation of the water quality between the lakes shows the same trend as the percolating water, the lake water has a quality different from the associated percolating water (Figs 4,, 5,, 6 & 9). The water flowing through the pillars and walls of the quarry is therefore not only composed of matrix water, but also of fracture water that is recent water.

Two geochemical arguments allow us to assume that the lake water is younger than the percolating water sampled from the roof above the lake.

• For the lakes in Group 1, there is little contamination in the percolation while some lakes are slightly more contaminated with nitrate or atrazine (Fig. 4), showing the contribution of more recent water.

• The tritium content is lower in the lakes than in the percolating water (mean, 9.2 ± 1.1 UT in the lakes; mean, 12.1 ± 1.8 UT in the percolating water, Fig. 5), giving evidence of a proportion of fracture flow.

The time series of the lake water level combined with those of temperature and electrical conductivity values (Fig. 7) are helpful to determine whether this velocity corresponds to direct transfer or to a piston effect.

The temperature variation of Lake Bleu gives evidence of direct transfer from the surface, meaning a contribution of recent water flowing throughout the fractures. For Lake Lucarnes and Lake Pedro, the temperature variation is lower: The contribution of recent water from direct transfer is probably lower for these sites.

For the three sites – Bleu, Lucarnes and Pedro – the electrical conductivity value increases as the water level increases, giving evidence of a piston effect process: more mineralized water is brought to the lake corresponding to old water pushed with pressure. Nevertheless, a large decrease in electrical conductivity was observed in March 2020 for Lake Bleu at the moment of maximum recharge, indicating surface water infiltration. A small decrease was also observed in Lake Pedro at the end of February 2020, showing that direct transfer contributed a small proportion of water. These direct and rapid transfers, even if they seem to contribute only little to the recharge, are proof of fracture flow at the sites.

To summarize, the water table is mainly composed of ‘old’ water resulting from transfer through the matrix at a velocity of about 1 m a⁻¹, but with a small contribution of rapid transfer through the fractures at an average velocity of approximately 60–80 m a⁻¹.

The concentrations of agricultural contaminants are very high in groundwater (with the maxima of nitrate concentration at 144 mg l⁻¹, atrazine at 15 200 ng l⁻¹ and DEA at 52 000 ng l⁻¹; Table 2) compared to regional studies on Chalk groundwater. For $NO3−$⁠, El Gaouzi et al. (2013) observed concentrations between 34 and 49 mg l⁻¹ in boreholes or springs in the Paris Basin. Sorensen et al. (2015) found nitrate concentrations ranging from 30 to 60 mg l⁻¹ for the Chalk aquifer in the UK. For pesticides, Baran et al. (2008) observed concentrations of atrazine between 50 and 100 ng l⁻¹ and of DEA between 100 and 200 ng l⁻¹ in the Paris Basin from 2000 to 2006. In the Chalk aquifer in Belgium, Hakoun et al. (2017) measured the highest concentrations of atrazine and DEA at less than 200 ng l⁻¹.

At Saint-Martin-le-Nœud, the concentrations of nitrate, atrazine and DEA vary spatially along the quarry both in lakes and percolating water. According to PCA (Fig. 9), the first factor explaining these variations is UZ thickness. The higher the UZ thickness, the lower the contaminant concentrations.

Figure 10 presents the nitrate concentration from 2014 to 2019 in the percolation water at different depths (=UZ thickness). It highlights the nitrate profile with a nitrate peak at about 15–20 m deep. The nitrate peak descends slowly with a rate of 0.3 to 0.5 m a⁻¹ (Chen et al. 2019a) depending on the site. Concentrations have increased for most points, and sometimes intensely (e.g. nitrate concentration at Bleu increased by more than 30 mg l⁻¹ in 5 years, Figs 6 & 10). For these sites, the concentrations will probably continue to increase, as shown in Figure 6, until they reach the peak over 140 mg l⁻¹. Only two sites show a different evolution: Carrefour is the most superficial of all the sites with only 18 m of UZ; the percolating water is therefore younger. The peak of nitrate probably occurred around or before 2014. The Lucarnes UZ thickness is substantial (27 m) but has the highest transfer velocity in the UZ. Therefore, the nitrate peak has probably already been reached here as well. In the future, the nitrate peak will continue to go down. With a very simple approach, considering a velocity of 0.5 m a⁻¹, the peak could reach a depth of about 35 m in 2050. The water table in the Paris Basin is approximately 50–60 m deep and, in this context, the nitrate peak could reach the water table around 2090. This approach is raw since diffusion phenomena are not taken into account. Moreover, a single velocity is considered, whereas it would be necessary to take into account the heterogeneity of the transfer velocities according to the properties of the UZ. Nevertheless, a strong increase in nitrate concentrations in groundwater can clearly be expected in the coming decades. The very high levels of contamination observed in the quarry can be explained by the configuration of the site. The percolation and lake water come exclusively from the agricultural area and the water is taken from the upper groundwater table without dilution. This high level of agricultural contamination is rarely observed in boreholes and springs. However, it can be assumed that the concentrations found in the UZ matrix (from percolation) are representative of the storage of contaminants in the Chalk UZ in agricultural areas.

The transfer and the degradation of atrazine residues through the Chalk aquifer UZ in Saint-Martin-le-Nœud has already been published in Chen et al. (2019b). A summary of the results is presented here, supplemented by additional observations on percolation.

Firstly, atrazine transfer velocities from the soil surface toward the water table were estimated (Table 4) to be in the range of 0.54–2.15 m a⁻¹ (Chen et al. 2019a).

Secondly, the spatial variations of atrazine and DEA concentrations are very high for both lake and percolation water and are related to the spatial variations of UZ thickness and the thickness of the CwF layer, as follows:

1. For the UZ thickness, in the case of Saint-Martin-le-Nœud at present, the deepest groundwater is globally less contaminated. More generally, the transfer time is directly related to the depth of the water table: the deeper the groundwater, the older the water.

2. For the thickness of the CwF layer, the relationship with pesticide concentrations is more complex. In our case, the third principal component PC3 of the PCA (Table 3, Fig. 9) shows a correlation between the CwF thickness and the pesticide concentration: the higher the CwF thickness, the higher the atrazine contamination. Furthermore, this type of agricultural soil is favourable for growing maize where atrazine was used. The UZ thickness of Group 1 is quite similar to that of Group 2, whereas the thickness of CwF is higher in Group 2. The concentration of pesticide is very low for Group 1, whereas it is very high and variable for Group 2. Chen et al. (2019a, b) deduced that the transfer velocities are the highest under a thick layer of CwF. Moreover, Chen et al. (2019b) showed that the degradation processes (expressed by the DAR) are the greatest for a thick CwF layer. The degradation of atrazine is a biological process, occurring mainly in the soil (Johnson et al. 2000). Chen et al. (2019b) hypothesize that the CwF layer allows for the formation of a ‘perched water table’, leading to atrazine staying near the surface for extended periods of time, resulting in the degradation processes. The role of the CwF layer in transfer processes is detailed in the next section.

In karst Chalk of Upper Normandy (France), Jardani et al. (2006) and Valdes et al. (2014) studied the CwF layer. Firstly, Jardani et al. (2006) observed a perched water table at the surface, particularly in CwF layers, and they also showed that this layer is heterogeneous with preferential pathways. Secondly, Valdes et al. (2014) showed that more preferential pathways exist under a thick CwF layer. They used the map of turbidity in Chalk groundwater to provide a rough estimation of the degree of connection of the water table to the surface and they compared it with the map of CwF thickness. A good correlation shows the more karstic areas (i.e. a higher degree of connection between the surface and water table) correspond to the higher CwF thickness. This can be explained by the process of CwF formation: the CwF results from the weathering of Chalk (Laignel et al. 1999), and thus a thick layer of CwF indicates that a significant amount of Chalk dissolution has occurred allowing for the presence of larger karst conducts or larger fractures. To summarize, Valdes et al. (2014) established that a thick CwF layer allows for storage of perched groundwater in the UZ in which biogeochemical processes can occur, followed by a delayed and concentrated infiltration via preferential pathways. The exact location of this water storage is difficult to define. Jardani et al. (2006) observed water storage and water movement in the CwF (allowed by CwF heterogeneities), but also in the overlying loess at their study site.

The Chalk of Saint-Martin-le-Nœud is clearly less karstified than the Chalk of Upper Normandy, but similar transfer processes may be advanced. The results presented here show that in the case of a thick CwF layer:

1. Infiltrated water is stored in the near surface, forming a sort of ‘near-surface perched groundwater’, that could be located in the CwF layer and in the covering soil. This is supported by the following arguments:

   • The farmer indicates that he cannot sow wheat on this plot (root rot due to the high water content of the soil).

   • The greatest pesticide degradation processes measured in the area with thick CwF indicate a long residence time near the surface.

   • The permanent percolation flow and the low variation in the lake level may be explained by the storage of water in a ‘near-surface perched groundwater’, which will continuously and slowly discharge into the Chalk UZ.

2. The infiltration may be concentrated. This is supported by the previous arguments and the geometry of the CwF (Fig. 3) that shows the presence of depressions that could facilitate concentrated infiltration with potential preferential pathways, as described by Jardani et al. (2006).

3. The transfer velocity in the UZ is higher. This is supported by the velocities deduced from the environmental tracers nitrate, atrazine and tritium, which are highest under a thick CwF in our study site.

The Saint-Martin-le-Nœud underground quarry is a unique site for the study of water and contaminant transfers in the Chalk aquifer. Firstly, it gives access to both the UZ at different depths and the SZ (the water table) directly. Secondly, the UZ characteristics are spatially variable at the scale of the quarry, which allows improved understanding of the role of the UZ thickness and CwF layer in the geochemical and hydrodynamic behaviour of groundwater. Thirdly, the detailed monitoring from 2012 onwards allows study of the longer-term evolution of the Chalk groundwater.

This study provides new insights into the following points presented in the introduction

1. The local variability in groundwater behaviours upon reaching the water table

   Chalk groundwater has highly variable hydrodynamic behaviour and geochemical properties, although the measuring points are very close to each other. The groundwater, here supplied exclusively by the UZ, may have little contamination, or very high concentrations of nitrate and atrazine, over small spatial scales.

2. The balance of fracture v. matrix flow

   The Chalk groundwater is mostly old (a few decades in this study) and, therefore, water transfer processes occur mainly in the matrix. At Saint-Martin-le-Nœud, the estimated water velocities in the Chalk matrix range from 0.37 to over 0.72 m a⁻¹ and the estimated hydraulic conductivity ranges from 1.3 × 10⁻⁹ to 1.1 × 10⁻⁸ m s⁻¹. However, a small part of recent water flowing through the fractures is highlighted. The proportion of water flow between the matrix and fractures remains difficult to determine, although this question may be addressed thanks to a detailed analysis of the hydrodynamic chronicles (time series) and the addition of other dating tracers.

3. The transport of contaminants

   The percolating water has shown storage of agricultural contaminants in the Chalk UZ, which percolates slowly toward the water table through the matrix via the piston effect, at speeds similar to the water velocity or below. It can be assumed that the deep groundwater, today not contaminated, will have a large increase in nitrate, atrazine and DEA concentrations in the future and for many years. Moreover, other contaminants used later, such as other pesticides (metolachlor, acetochlor, dimethenamid, etc.), pharmaceutical products (antibiotics, hormones, etc.), or other industrial contaminants are probably stored today in the Chalk UZ and will likely reach the Chalk water table in the future.

4. The role of the CwF cover on the water and contaminant transfer in the Chalk aquifer

   The CwF layer, often considered protective in the past, plays a role in the hydrodynamics and the quality of the groundwater. This layer allows a perched water table to be created near the surface, which makes the degradation processes of pesticides possible. This perched water table then flows to the unsaturated Chalk by more concentrated infiltration through preferential pathways with higher transfer velocity and continuous recharge throughout the year. This mainly explains the high variations of pesticide concentrations for a thick CwF layer. Furthermore, the DAR results demonstrate that degradation processes occur where CwF is thick.

The authors thank the SNO Karst (CNRS) for the integration of the project into their research network. They also thank the Conservatoire d'Espaces Naturels de Picardie for access to the quarry. The authors would like to thank the reviewers and the Editor for their helpful comments.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

DV: conceptualization (lead), data curation (equal), funding acquisition (lead), investigation (equal), methodology (lead), project administration (lead), supervision (lead), writing – original draft (lead), writing – review & editing (lead); NC: data curation (equal), formal analysis (equal), investigation (lead), validation (equal); MD: data curation (supporting), investigation (equal), methodology (supporting); CM: conceptualization (equal), methodology (equal); HB: data curation (equal), methodology (equal), validation (equal); RG: data curation (supporting), methodology (supporting), supervision (supporting); JG: methodology (supporting); FA: data curation (equal), formal analysis (equal); RN: data curation (equal), investigation (equal); EA: data curation (equal); MR: data curation (supporting), formal analysis (supporting), investigation (supporting); CF: data curation (equal); PG: conceptualization (supporting); PR: conceptualization (supporting), supervision (supporting).

The authors gratefully acknowledge funding provided to DV by IPSL (ID0E6UBI3597, ID0EZFCI3599), Conseil Régional, Île-de-France (ID0EM4BI3598) and CNRS (EC2CO).

All data generated or analysed during this study are included in this published article (and its supplementary information files)