Oxido-reduction sequence related to flux variations of groundwater from a fractured basement aquifer (Ploemeur area, France)

Abstract This paper focuses on the chemical evolution of water during the exploitation of a fractured aquifer in a NO3-rich agricultural environment. During a ten year period, both production rate and chemical parameters were continuously measured in tap water obtained from a deep-water plant in Brittany, France. Changes in SO 4 2 - and NO 3 - were observed after pumping was initiated. Nitrate concentration decreased during the first 200 days and then stabilized at ∼5 ± 1 mg/L, while SO 4 2 - concentration increased rapidly over this period and then showed a steady state increase (0.01 mg/L/day). These changes are attributed to the development of equilibrium between the physical flow parameters and the chemical kinetics of autotrophic denitrification processes that occur in the pyrite-bearing fractures. The chemical characteristics of the groundwaters collected in 18 wells located around the site allow identification of two different areas. One is weakly influenced by pumping and is characterized by high NO 3 - concentrations and a short residence time. The second area is directly related to the main pumped well, and characterized by reduced NO 3 - levels combined with an increased SO 4 2 - production, resulting from the denitrification processes in the pyrite-bearing fractures. Over the last few years, a SO 4 2 - increase unrelated to denitrification has been recorded in some wells. Based on the NO 3 - , SO 4 2 - and Fe concentrations, this is attributed to oxidation of S minerals, coupled to FeIII reduction. Exploitation of the aquifer has led to a rapid transfer of the waters within the deep fractures. Their high velocities strongly control the chemical parameters and have led to a redox sequence that has promoted S oxidation, coupled with (1) O2, (2) NO 3 - , and (3) Fe reduction.

Oxido-reduction sequence related to flux variations of groundwater from a fractured basement aquifer (Ploemeur area, France)

Introduction
Over the past few decades, discharges of NO À 3 from drains or shallow wells have been increasing due to anthropogenic inputs (Jordan et al., 1997).
Correlations between anthropogenic inputs, fertilizer application, cultivation of N-fixing crops, and groundwater and river water pollution have been studied by numerous authors (e.g. Hill, 1978;Osborne and Wiley, 1988;Mason et al., 1990;Jordan and Weller, 1996). The spatial distribution of groundwater quality is not only related to groundwater contamination aspects. It also depends on water pathways, residence time, and thus on the soil and aquifer hydraulic properties (Grambell et al., 1975;Spalding and Exner, 1993;Brenner and Mondok, 1995;Martin et al., 2004). In regions of crystalline rocks such as Brittany in northwestern France, tap water is traditionally supplied from shallow wells or surface water reservoirs. Interest in production requirements from deep-water sources has increased because of growing concern about agricultural groundwater contamination problems. However, the silicate bedrock contains small aquifers that are usually located in the fracture system of the basement. Both large porous fractures and a high connectivity are required for the development of a large and sustainable water resource. Although fractured systems commonly do not provide large water flow, large fractured zones enhance their permeability . Understanding the aquifer behavior is crucial for its management, basically to prevent any possible contamination from the anthropogenic surface activities after several years of exploitation. Understanding the physics and the chemistry of fractured aquifers illustrates several problems that are mainly related to fluid flow heterogeneity and their relation with biochemical processes. Both NO À 3 consumption due to biogeochemical processes and long-term changes in NO À 3 and SO 2À 4 have been reported in several aquifer studies (Howard, 1985;Kö lle et al., 1985;Mariotti, 1986;Mariotti et al., 1988;Frind et al., 1990;Postma et al., 1991;Smith et al., 1991;Engesgaard and Kipp, 1992;Korom, 1992;Bö hlke and Denver, 1995;Puckett et al., 1999;Puckett and Cowdery, 2002;Pauwels et al., 2001;Koenig and Liu, 1996). Understanding both NO À 3 transport through basement fractures and its consumption due to biogeochemical processes within the fractures is a key issue in evaluating the long term effects of N-application on groundwater quality. Nitrate is available from natural sources (nitrification) or as a constituent in fertilizers. Aquifer redox conditions that control NO À 3 removal and all other redox reactions may be described by characteristic redox species (C, N, O, S, Fe, Mn) (Champ et al., 1979;Berner, 1981;Postma et al., 1991;Stumm and Morgan, 1981). When N removal is associated with SO 2À 4 production the process involved is biochemical autotrophic denitrification (Stumm and Morgan, 1981;Mariotti, 1986;Rö delsperger, 1989;Pauwels et al., 2000). Even, when low contents of organic material are present either in dissolved or solid forms, chemolithotrophic bacteria such as Thiobacillus denitrificans may use reduced S compounds such as pyrite to reduce NO À 3 and produce SO 2À 4 , gaseous N 2 and Fe II (Rö delsperger, 1989;Frind et al., 1990;Postma et al., 1991;Bö ttcher et al., 1990). The water chemistry evolution of redox-dependant species depends on both the availability of the electron donors in the solid phase or in the groundwater, and the presence of electron acceptors. Electron acceptors such as O 2 , NO À 3 , and potentially SO 2À 4 are usually transported by water. The potential electron donors that are thermodynamically able to reduce NO À 3 are organic matter, pyrite and Fe(II)-silicates (Postma et al., 1991;Chapelle et al., 1995;Tesoriero et al., 2000;Bö hlke et al., 2002;Puckett and Cowdery, 2002). In natural systems, Mn(III, IV) and Fe(III)oxides are subject to reductive dissolution. For instance, dissolution of Fe(III)-oxides is markedly enhanced by reducing agents (Hering and Stumm, 1990). Therefore, MnO 2 and Fe(III)-solids may also be associated with the redox processes, although they are solid oxidizers.
In fractured rocks with very low matrix permeability, such as crystalline rocks, fluid flow is often very heterogeneous and located in few fractures (Neretnieks, 1985;Bour and Davy, 1997). Incompatible reactions may occur juxtaposed in the same system, depending on the permeability of the medium (Moncaster et al., 2000). Pumping from a fractured aquifer will force waters through fractures toward the wells and thus change the fluid velocity and possibly the water flow path. In this case, it is difficult to apply classical hydrogeological theories that are based on an equivalent porous medium.
In this paper, new information is provided on the changes observed in groundwater chemical composition, particularly NO À 3 , SO 2À 4 and Fe contents, which are related to the development of a deep water reservoir in a fractured system at Ploemeur in Brittany, NW France. This paper reviews: (1) the main processes that influence groundwater quality; and (2) the possible relationships that may exist between the chemical composition changes of the groundwater and the water fluxes induced by pumping development.

Geological and hydrogeological settings
The deep-water plant of Ploemeur has been exploited as the principal source of tap water for a medium sized city (15,000 inhabitants) since June 20th 1991. Fig. 1(a) shows the study site and a location map of the wells. The annual water production is about 10 6 m 3 . The net recharge is about 0.7 m/a (Touchard, 1999). The site is located on a 2.5 km 2 watershed, with an average elevation of 25 m above sea level. This area includes three types of land uses: Lorient domestic airport, farms, and the town of Ploemeur. Sixty percent of the watershed surface is dedicated to agricultural land use such as dairy livestock production and cropland (mostly vegetables and corn). Overall, about 50 kg N/ha of fertilizer was applied to the soil across this area in 1995 (Roussel and Gallat, 1996).
The geology of this part of southern Brittany is characterized by the South-Armorican Shear Zone. This structure is a NE trending shear zone that extends for several hundred kilometers along the southern border of French Brittany. The development of this structure was accompanied by intense tectonic activity that strongly fractured the formations. In the study area, there is a flat-lying contact between massive granites and surrounding mica schists. This contact zone consists of alternate subhorizontal units mainly made up of granites, micaschists or pegmatites. These formations are strongly weathered at the surface down to a depth of about 35 m. Below the weathered zone, the rocks show variable degrees of fracturing. The main fracture zones correspond to a major hydraulic path and coincide with a pegmatite-rich area ( Fig. 1(b)). The production zone extends perpendicular to the Southern Armorican Shear Zone along a 20°N striking fault that crosscuts the older shear structures.
Eighteen boreholes were drilled in the study area in 1990 to a depth of 70-125 m to find a new reservoir to satisfy the local need for tap water. Three of these boreholes are now pumped wells (solid square in Fig. 1(a)) and the others are used as observation wells. Two shallow boreholes, MF1 and MF2 (35 m depth), are also used as observations wells in the weathered domain ( Fig. 1(a)). The local geology was studied from drill cuttings obtained from the 18 boreholes ( Fig. 1(a)), and from cross sections exposed along the sea-shore. The schematic representation of six boreholes is presented in Fig. 1b. Tap water has been extracted from one main pumped well, PE since June 1991. During the first 417 days, the tap-water production rate was close to about 80 m 3 /h and then increased to about 120 m 3 /h with the addition of two new pumping wells, F29 and F31. After 417 days, the production was maintained at around 120 m 3 /h with peaks of up to 170 m 3 /h during summer (Fig. 2).
Predicting flow paths and velocities in a fractured medium is difficult due to the unknown complex geometry of the fracture network and the wide range of densities, lengths, and apertures of the fractures (Neretnieks, 1985). Since 1991, when exploitation began, a series of piezometric measurements at different intervals have been made in the 18 wells to construct the 3D piezometric surface of the aquifer. The maximum drawdown depression is located in the pumped well zone and its extension has increased with time, indicating that the local recharge does not balance the pumped volumes.
Constant-rate pumping tests have been performed primarily to study hydraulic properties of the fractured zone on different scales. The analyses focused specifically on the scale dependence of the hydraulic parameters. The area directly connected to the pumped wells, which extends along the 20°N striking fault, was about 250 · 10 4 m 2 . Transmissivity remained constant regardless of distance to the pumping well with values around 1.5 · 10 À3 m 2 /s. The storage coefficient, however, decreased by one order of magnitude with the distance to the well pumped during the hydraulic test. A complete interpretation of these results, incorporating a generalized version of the Theis/Jacob model to non integral flow dimension (Barker, 1988), is presented elsewhere (Le Borgne et al., 2004).

Data collection and experimental methods
Two different geochemical surveys were performed at the Ploemeur site after groundwater exploitation began: (1) In the PE pumping well, production rate and electrical conductivity (EC) were recorded, and NO À 3 , Cl À , SO 2À water plant. Five chemical analyses were also made for the major and trace elements in PE between 02/19/91 and 02/06/95. (2) In 15 wells (which correspond to 29 sites at different depths in the either pumped or observation wells) in the Ploemeur catchment, water samples were collected over 7 field campaigns (from June 1996 until December 2003).
(1) Long-term monitoring: Chemical analyses of NO À 3 , Cl À , SO 2À 4 concentrations, and Total alkalinity every 10 days and the 5 chemical analyses of PE were performed by the CGI laboratory (''Centre de Génie Industriel'', Guidel, France) according to French standard methods, which comply with the French laws on drinking water quality.
(2) Geochemical sampling: The 7 sampling campaigns were carried out in June 1996, September 1996, February 1997, May 1997, October 2001, December 2002and December 2003. A two-holed automatic groundwater sampler was used to collect water samples at predefined depths in observation wells. Sampling depth corresponds to the location of water inflows identified during drilling and further observed through diagraphy. The sampling depth is indicated in Table 1. For the pumped wells, water was directly sampled at the pump tap. A minimum of two samples were collected at each location. One sample (250 mL) was collected for major ion analyses. This sample was divided into two sub-samples and filtered on site through a 0.2 lm mesh sieve. All samples for cation and trace element analyses were acidified with 2 mL of 2 N ultrapure HNO 3 and stored in PTFE bottles, rinsed twice with double-distilled acid and in ultrapure water.
Water temperature, pH, and salinity-conductivity were measured in the field at each location. Total alkalinity (Alk.) was determined by H 2 SO 4 (0.04 N) titration within a few hours after collection. Electrical conductivity and temperature were measured using a WTW Conductivity Meter LF196, equipped with a TetraCon 96-1.5 electrode. The pH was measured using a WTW pH 320 instrument, equipped with a pH-combined electrode with integrated temperature sensor, SENTEX 97T. Major ions and selected trace elements were analyzed, several days after each field experiment. NO À 3 , Cl À , SO 2À 4 , and Br À concentrations were measured using ion-chromatography (DIONEX DX-100), and SiO 2 and F À by colorimetry (HACH DR-4000 spectrophotometer). Calcium 2+ , Mg 2+ , Na + , K + , Fe total , Al total , Mn 2+ , Sr 2+ and Ba 2+ concentrations were measured using an inductively coupled plasma atomic emission spectrometer (ICP-AES) (ISA Jobin-Yvon JY 70 Plus) in the LARAAH laboratory (Laboratoire de Recherches Appliquées Atmosphère-Hydrosphère) for the 1996 and 1997 campaigns, and for the 2001-2003 campaigns using an inductively coupled plasma mass-spectrometer (ICP-MS) (HP-4500) in the Geosciences Laboratory (Bouhnik-Le Coz et al., 2001). All resulting cationanion charge balances are within 10%, with most samples lying within 5%. Precision based on triplicates is better than 3%, except for alkalinity and SiO 2 analyses which were 8% and 5%, respectively.  Table 1 Chemical data for wells from Ploemeur site During the May 1997 campaign, a selected set of samples was collected for 3 H analyses. Tritium was analyzed at the ''Centre de Recherches Géodynamiques'' at Thonon-les-Bains by an electrolytic enrichment and the standard method of liquid scintillation counting, with a precision of 0.5 tritium units (TU).

Evolution of the water quality in the pumping well
The data set allows the distinction of two types of changes: (1) long-term variation of chemical composition; (2) short-term variations affecting SO 2À 4 and NO À 3 concentrations and alkalinity, which are linked to rapid changes in the production rate.
(1) Long-term changes: The quality of the groundwater extracted at PE is of the Na-Cl type and has evolved since exploitation began (Table 2). Between 1991 and 2003, electrical conductivity increased by up to 90%, in response to increase of concentrations of all major ion species except NO À 3 . The increases of Ca 2+ + Mg 2+ and SO 2À 4 +Alk were 91% and 199%, respectively, and those of Na + + K + and Cl À were 64% and 97%, respectively. Furthermore, a 92% decrease in NO À 3 concentration was also observed. As illustrated in Fig. 2, the SO 2À 4 and NO À 3 variations through time are inversely correlated after pumping initiation. During the initial 158 days, the rate of decrease in NO À 3 was similar to the increase in SO 2À 4 at 0.06 and 0.07 mg/L/day, respectively. After 158 days, the decline in NO À 3 concentration ended and stabilized at about 5 ±1 mg/L, whereas the SO 2À 4 concentration was still increasing. After 417 days, the increase in SO 2À 4 concentration also slowed down and reached a steady increase of about 0.01 mg/L/day.
(2) Short-term variations: Some of the short-term variations were observed at the onset of pumping, whereas others occurred after a change in the pumping rate in the pumped well (PE). Such variations are particularly observable consequent to increase of pumping rate from 80 to 120 m 3 /h after 417 days of exploitation by bringing two new wells (F29 and F31) into production. This induced a decrease in SO 2À 4 concentration by 8.0 mg/L along with (continued on next page) a small increase in NO À 3 concentration by 2.0 mg/L. During the following days, the rate of NO À 3 consumption was once more similar to the SO 2À 4 production rate (0.06 and 0.07 mg/L/day, respectively). Chloride concentration also shows some variation when pumping rate increases, but the effect is smaller and shorter than for SO 2À 4 and NO À 3 . Table 1 presents the chemistry of the groundwaters sampled in the Ploemeur area. Sampling of the wells around the pumping site allows the spatial chemical variations to be characterised. Three groups can be distinguished on the basis of their NO 3 -SO 4 content (Fig. 3).

Vertical and spatial distribution of the water quality
Group I is composed of almost NO À 3 -free waters with high SO 2À 4 =NO À 3 ratios and low 3 H content (<5 TU). Wells MF1, MF2 and F36 belong to this group. These wells have characteristics similar to the PE pumping well described above.
Group II, which includes wells F6 and F20, is characterized by an extremely high NO À 3 content (>100 mg/L), a low SO 2À 4 =NO À 3 ratio, and a high 3 H content (7 TU in F6).
Group III waters have high NO À 3 contents (10 to 60 mg/L) and SO 2À 4 =NO À 3 ratios and 3 H contents similar to group II. Wells F9, F19, F30, F34 and F35 belong to this group (8.5 TU in F9, and F19).
The chemistry of the groundwater shows a clear depth zonation during the 96 and 2001 campaigns (Table 1), which correlates with the distribution of the major inflows described above (Fig. 1). The chemical composition of well F28 (Fig. 4), shows large vertical variations that are related to different water inflows over the range of 30-50 m and in the deeper part of the borehole (60-80 m). The NO À 3 -SO 2À 4 ratios in all the samples (for the 6 campaigns from 1996 to 2002, Table 1) have a good linear correlation (R 2 = 0.66) if the extreme point of Units are mg/L unless indicated. The resulting cation-anion charge balances for most samples were within 4%. b.d., below detection limit. Detection limit of NO À 2 and NH þ 4 is 0.01 mg/L and 1.0 lg/L for Fe total and Al total .
December 2002 is omitted. This clearly indicates that encountered waters result from a mixing of two end-members. The first end-member is encountered in the upper part of the borehole and clearly belongs to group I (low NO À 3 and high SO 2À 4 , low Cl À and high Na + /Cl À and SiO 2 ) exemplified by wells MF1 and MF2. The second end-member is encountered in the lower part of the borehole (below 70 m). It does not show any of the characteristics of the group II wells with anthropogenic influence, F6-F20 (high NO À 3 , Cl À , low SiO 2 concentrations and Na + /Cl À ratio), and seems to be chemically similar to the group III samples (the western F9, F19, F30, F34 and F35 wells).
The two different groups (I and II) can also be distinguished on the basis of other elements. Group I has relatively high Na + /Cl À ratios, ranging mainly from 0.5 to 1.2, high alkalinity (40-76 mg/L) and SiO 2 concentrations (26-41 mg/L) with respect to Group II for which Na + /Cl À ratios vary between 0.5 and 0.8 and range of alkalinity is between 12.2 and 24.4 mg/L. SiO 2 content are 12-22 and 17-29 mg/L, respectively. The two groups are also characterised by different chemical carbonate equilibria (Table 3). The pCO 2 values computed for group I are in the range expected for waters recharged in soil rich in CO 2 (0.9-3.2%). Group II shows higher values (2.1-8.7%), out of the range for soil-equilibrated waters. The mineral saturation indices (SI) for the waters were computed using PHREEQC (Parkhurst and Appelo, 1999). Calcite saturation indices (SI) show that none of the waters have achieved calcite equilibrium, but group II has much lower SI than group I (À3.9 to À3.0 and À2.7 to À1.5, respectively).
Several wells have shown a major change in their chemical evolution during the studied years. Fig. 5 highlights the difference between groups I and II in their NO À 3 and SO 2À 4 evolution. Group III (F19-F34-F35-F9-F30) displays very stable concentrations for all the elements analyzed during the sampling period (1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003). Group II (F6 and  F20 wells with anthropogenic influence) always shows very high but decreasing NO À 3 concentrations (Fig. 5). These wells represent the western part of the aquifer where agricultural influence is important. The decrease observed in NO À 3 concentration suggests a recent lower anthropogenic pressure in this zone. Group I (F36, MF1, MF2 wells) shows a clear increase in the SO 2À 4 =NO À 3 ratio during the last sampling campaigns. The range of this increase varies among the wells by a factor of 2 for MF2 to about a factor of 10 for F36 and MF1. The evolution described for F36, MF1 and MF2 wells is also seen in the pumped well PE.
A hydrochemical synthesis of the vertical and spatial concentrations is presented in Fig. 6. Anthropogenic sources are spatially distributed with high loads in the eastern zone (group II). Within that area, flows of anthropogenic influenced waters Computations were performed using the PHREEQC thermodynamic data base (Parkhurst and Appelo, 1999). ratios such as those observed in group II, which originate mainly from the weathered part of the aquifer, along the N20 and N150 fault (as observed in the upper part of F28).
(3) Waters from the PE pumping well which evolved during exploitation from the group III composition towards high SO 2À 4 =NO À 3 ratios, although lower than those of group II.

Origin of the chemical groups and mixing processes
All the wells sampled at the Ploemeur site show limited variations of some major element concentrations, either in space or time (Cl À concentration has an average value of 58 ± 10.5 mg/L, SiO 2 : 28.6 ± 6.6 mg/L, K + : 4.6 ± 1.8 mg/L, Na + : 40.5 ± 6.2 mg/L and Mg 2+ : 12.6 ± 8.0 mg/L). Only NO À 3 and SO 2À 4 and, to a lesser extent, alkalinity and Ca 2+ content show great amplitude variations. As seen in Fig. 3, three groups of water have been distinguished based only on their NO À 3 and SO 2À 4 contents. The high NO À 3 content in group II waters indicates high anthropogenic loads from agriculture. Moreover, group II waters have high 3 H concentrations. The present 3 H background level in precipitation is in the order of 5-12 TU (IAEA, 2004). The high 3 H concentrations are thus interpreted as an indication of a recent recharge and the short residence time of groundwater. The lower Na + /Cl À ratios of group II with respect to that of groups I and III may be related to agricultural sources such as fertilizers (KCl) or cattle manure. Wells MF1, MF2 and F36 (group I) have high SO 2À 4 and low NO À 3 concentrations but 3 H contents are lower. The 3 H contents, closer to the present background level, could indicate a very recent recharge or a mixing between recent water and waters older than 1950. The occurrence of a sample showing tritium below detection limit and low 14 C activities (Touchard, 1999) strongly supports the second hypothesis. It indicates that these waters are less influenced by anthropogenic sources, because recharge occurred partly before agricultural developments of the study area. This interpretation is also supported by the negative-correlation between the 3 H data and SiO 2 concentration which is indicative of a higher water-rock interaction degree due to a longer residence time. The group III waters, which show no temporal variation of concentrations and moderate NO À 3 concentrations as compared to group II are less influenced by contamination than group II. They are thought to represent the ''original'' composition of the aquifer water, before pumping started inducing mixing with the weathered compartment and a chemical evolution of the pumped water.
The evolution of the pumping well (PE) might also be partly related to time-mixing processes of waters of different types, independently of the chemical evolution of components such NO À 3 and SO 2À 4 . The changes in the Na + /Cl À ratio and SiO 2 contents suggest an evolution of the pumped groundwater evolved from group I (MF1-MF2) waters towards group II (F6-F20) waters, but with chemical modification of their NO À 3 and SO 2À 4 contents through biochemical processes (see below).
The decrease of NO À 3 concentration with depth is due partly to mixing with older waters and redox processes such as denitrification. Despite the occurrence of water with NO À 3 concentration, varying from <1 up to 150 mg/L, a vertical relationship between the chemical composition of groundwater and the structure of the aquifer does not apply at Ploemeur. Groundwaters of wells F6 and F20 (group II), which are located in the eastern part of the pumping site in a topographic high, are highly contaminated by NO À 3 even though the sampling depth lies in the fissured part of the aquifer (tubing in F20 is open from 11 to 82 m and, from 10 to 50 m in F6). In contrast, wells MF1 and MF2 (group I), with low NO À 3 concentrations, are short wells (<35 m) located entirely in the weathered part of the aquifer. Moreover, although well F28 has different water quality according to sampling depth, the vertical stratification differs from that expected. At low depth, the waters from the upper part of well F28 (<50 m) have very low NO À 3 content and are representative of group I, whereas with depth, NO À 3 concentrations increase and the deeper inflows (>70 m) belong to group III.
Actually, water type occurrence is somewhat related to geographic location, induced by the tectonic setting of the area. Despite a variable intensity of anthropogenic impact in the eastern and western parts of the study area, NO À 3 contaminated waters showing short residence times and low SO 2À 4 concentration are encountered in both the weathered and in the deep fractured compartment. Conversely, waters with low NO À 3 content and long residence time can be found along the N20 (and conjugate N150) fault axis, either in the upper weathered part or in the fractured zone of the aquifer, close to the pumping well.
Furthermore, the pumping seems to induce a change in the type of the waters produced from the fault related group I (MF1-MF2) towards the chemically modified group II type (F6-F20), which is influenced by anthropogenic activity.

Evidence for in situ nitrate and iron reduction processes
As shown above, NO À 3 and SO 2À 4 are the species with the most variable concentration at the Ploemeur site. Within such a context, these species are mainly affected by redox microbial processes. Stumm and Morgan (1981) describe the sequence of microbiologically mediated redox processes (Table 4). This sequence has already been observed in natural environments (Postma et al., 1991;Appelo and Postma, 1993;Chapelle et al., 1995;Tesoriero et al., 2000;Bö hlke et al., 2002;Puckett and Cowdery, 2002). In those cases, when organic matter has been consumed or when it has low reactivity, reduced minerals such as pyrite may serve as electron donors. In the anoxic zone, NO À 3 is then  concentration. According to the sequence described above, the redox conditions should remain oxidizing because high concentrations of NO À 3 indicate that NO À 3 is not being used as an electron acceptor. This hypothesis is confirmed by relatively low Fe and Mn concentrations. On the other hand group I waters have high SO 2À 4 concentrations and a lack of NO À 3 . In view of the present agricultural activities in the area, lack of contamination must be dismissed even if contribution of waters recharged before 1950 has been evidenced. Observation of reduced inorganic S minerals (pyrite) along the fracture planes in the granite (N20 and N150 fault axis) of the Ploemeur area and increasing SO 2À 4 concentration indicate that autotrophic denitrification must be active and partly explain the lower NO À 3 concentrations. Low concentrations of organic matter have been measured in the Ploemeur area at all depths (Touchard, 1999), which agrees with the results from similar watersheds (Martin, 2003). It can be considered that in the Ploemeur site denitrification can be mainly autotrophic and only to a lesser extent heterotrophic. An electron balance has been used and the results for the groundwaters collected during the September 1997 campaign have been published in Touchard, 1999. The conclusions were that in the hydrosystem of Ploemeur, the principal e À donors were CH 2 O and FeS 2 . The contribution of CH 2 O was comparable in all the studied groundwater samples (groups I, II, and III). The influence of the oxidation process of FeS 2 was not important in the two groups I and III but it was predominant in the group II.
In the Ploemeur site, denitrified waters have higher pH and alkalinity, as well as higher Ca 2+ and Mg 2+ content than NO À 3 contaminated waters. However, no increase in pCO 2 is noticed ( Table 3). The combining of calcite and/or a magnesian mineral dissolution, induced by the acidity generated by Fe (II) oxidation might have caused such a chemical evolution.
The chemical composition of the water in the PE well, 6 months before pumping (Fig. 2a) is very close to group III presented in the results section. It seems that this water type represents most of the aquifer ''original'' composition before pumping started, at least for the pumping site and the eastern part of the investigated area. At the onset of the pumping and 417 days later, when the pumping rate is increased by the addition of pumping in F29 well, a rapid evolution of NO À 3 and SO 2À 4 is observed. After about 200 days, the evolution slows down and a steady increase of SO 2À 4 (0.01 mg/L/day) is observed for several years, while NO À 3 concentration remains constant. Fig. 7 compares the evolution of NO 3 -SO 4 concentration of pumped water with theoretical autotrophic denitrification of waters containing initially 22, 50 and 120 mg/L NO À 3 . When pumping starts the NO 3 -SO 4 relation is close to the stoichiometry of the autotrophic denitrification Fig. 7. SO À 4 NO 3 diagram and chemical pathways. Black curves represents autotrophic denitrification for an initial NO À 3 content of 22, 50 and 120 mg/L.