Magnesium isotope systematics of the lithologically varied Moselle river basin, France

Transcription

1 Available online at Geochimica et Cosmochimica Acta 72 (2008) Magnesium isotope systematics of the lithologically varied Moselle river basin, France Agnès Brenot *, Christophe Cloquet, Nathalie Vigier, Jean Carignan, Christian France-Lanord CRPG, Centre de Recherches Pétrologiques et Géochimiques, 15 rue Notre Dame des Pauvres, Vandoeuvre lès Nancy, France Received 24 October 2007; accepted in revised form 27 July 2008; available online 20 August 2008 Abstract Magnesium and strontium isotope signatures were determined during different seasons for the main rivers of the Moselle basin, northeastern France. This small basin is remarkable for its well-constrained and varied lithology on a small distance scale, and this is reflected in river water Sr isotope compositions. Upstream, where the Moselle River drains silicate rocks of the Vosges mountains, waters are characterized by relatively high 87 Sr/ 86 Sr ratios ( ). In contrast, downstream of the city of Epinal where the Moselle River flows through carbonates and evaporites of the Lorraine plateau, 87 Sr/ 86 Sr ratios are lower, down to Magnesium in river waters draining silicates is systematically depleted in heavy isotopes (d 26 Mg values range from 1.2 to 0.7&) relative to the value presently estimated for the continental crust and a local diorite ( 0.5&). In comparison, d 26 Mg values measured in soil samples are higher (0.0&). This suggests that Mg isotope fractionation occurs during mineral leaching and/or formation of secondary clay minerals. On the Lorraine plateau, tributaries draining marls, carbonates and evaporites are characterized by low Ca/Mg ( ) and low Ca/Sr (80 400) when compared to local carbonate rocks (Ca/Mg = 29 59; Ca/Sr = ), similar to other rivers draining carbonates. The most likely cause of the Mg and Sr excesses in these rivers is early thermodynamic saturation of groundwater with calcite relative to magnesite and strontianite as groundwater chemistry progressively evolves in the aquifer. d 26 Mg of the dissolved phases of tributaries draining mainly carbonates and evaporites are relatively low and constant throughout the year (from 1.4& to 1.6& and from 1.2& to 1.4&, respectively), within the range defined for the underlying rocks. Downstream of Epinal, the compositions of the Moselle River samples in a d 26 Mg vs. 87 Sr/ 86 Sr diagram can be explained by mixing curves between silicate, carbonate and evaporite waters, with a significant contribution from the Vosgian silicate lithologies (>70%). Temporal co-variation between d 26 Mg and 87 Sr/ 86 Sr for the Moselle River throughout year is also observed, and is consistent with a higher contribution from the Vosges mountains in winter, in terms of runoff and dissolved element flux. Overall, this study shows that Mg isotopes measured in waters, rocks and soils, coupled with other tracers such as Sr isotopes, could be used to better constrain riverine Mg sources, particularly if analytical uncertainties in Mg isotope measurements can be improved in order to perform more precise quantifications. Ó 2008 Elsevier Ltd. All rights reserved. 1. INTRODUCTION Magnesium and calcium are intimately linked to the carbon cycle. Both are related to ocean ph, and the weathering * Corresponding author. Present address: BRGM, 3 avenue C. Guillemin, BP 39009, Orléans, France. address: a.brenot@brgm.fr (A. Brenot). of Ca and Mg silicates is a significant long-term sink of atmospheric CO 2 (Walker et al., 1981; Berner et al., 1983). Quantifying the sources of Ca and Mg in the dissolved phases of rivers, as well as the relative proportions of silicate and carbonate weathered is thus of primary importance for understanding atmospheric CO 2 consumption rates (Galy and France-Lanord, 1999; Millot et al., 2003) /$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi: /j.gca

2 Magnesium isotope systematics on the Moselle river basin 5071 Many studies use Ca/Na, Ca/Sr ratios and the Sr isotope composition of rivers and sediments to distinguish stream water solute fluxes derived from silicate and carbonate weathering (e.g. Wadleigh et al., 1985; Galy et al., 1999; English et al., 2000). This approach assumes that carbonate dissolution occurs congruently, and/or that dissolved Ca, Sr and Mg behave conservatively during transport. However, some studies document significant excesses in rivers of dissolved Mg and Sr relative to Ca compared to the chemical composition of limestones present in the basin (Kotarba et al., 1981; Sarin et al., 1989; Galy et al., 1999; Jacobson et al., 2002). Most of these studies suggest that an excess of dissolved Mg and Sr may result from the recrystallization of calcite or gypsum from saturated surface waters. In such environments, dissolved Ca displays non-conservative behaviour. Consequently, the chemical compositions of rivers that drain carbonates can be significantly different from the source rocks. Similarly, as a result of weathering incongruency, the geochemical composition of rivers draining silicates does not reflect the compositions of rocks drained at the basin scale. In mixed lithology basins it is therefore difficult to identify signatures induced by silicate weathering alone. Over the last few years, new potential isotopic tracers of silicate weathering, generally based on elements that are highly enriched in silicate rocks, such as Li, Mg and Si, have been developed (e.g. Huh et al., 1998; Georg et al., 2006). Also, studies of small monolithological basins have been undertaken (e.g. Louvat and Allègre, 1997; Ziegler et al., 2005). New generation inductively-coupled plasma mass spectrometers (ICP-MS) have allowed the development of precise isotopic measurements for various elements (e.g. Luais et al., 1997; Marechal et al., 1999; Halicz et al., 1999; Rehkamper and Halliday, 1999; Galy et al., 2001; Rouxel et al., 2002; Beard et al., 2003; Cardinal et al., 2003; Wombacher et al., 2003). This has provided a new means of studying geological and environmental processes. Magnesium isotope studies of terrestrial materials are still at an early stage (Galy et al., 2002; Chang et al., 2003; Carder et al., 2004; De Villiers et al., 2005; Tipper et al., 2006a,b). Galy et al. (2002) and Tipper et al. (2006a) reported significant differences in the Mg isotopic compositions of silicate ( 0.6 to 0&) and carbonate rocks ( 4.5& to 1.1&), highlighting the strong potential of Mg isotopes as a lithological source tracer. However, until now, few data have been available for river waters, and the existing data are mainly derived from large rivers draining mixed lithology basins (Tipper et al., 2006a,b). In this study, we test the potential of Mg isotopes to trace lithological sources and weathering processes by analyzing river waters, rocks and soils of the Moselle River basin (Northeastern France). This small basin is remarkable for its well-constrained and varied lithology, with silicate rocks upstream and carbonate/evaporite rocks downstream. We have analyzed major and trace elements and Sr and Mg isotopic compositions for the dissolved phases of the main rivers. Soils and parent rocks, representative of the main lithologies drained by these rivers, were also analyzed. 2. HYDROLOGY AND GEOLOGY The Moselle River basin (surface area = 3080 km 2 ) is located in northeastern France (Fig. 1). The Moselle River flows from its source in the Vosges mountains to its outlet at Pont Saint Vincent, 150 km downstream. The basin displays strong contrasts in lithology and topography on a small scale. Upstream of the city of Epinal, silicate rocks form the Vosges mountains (at altitudes between 1300 and 400 m), whereas downstream of Epinal, a carbonate platform corresponds to the Lorraine plateau ( m) (Fig. 1). Mean annual rainfall varies from 1730 mm in the Vosges mountains down to 1000 mm on the Lorraine plateau. The mean annual temperature is 6 C in the Vosges and 10 C on the Lorraine plateau, with an annual amplitude of 16 C. Forests are dominant in the Vosges mountains, while on the Lorraine plateau agricultural soils predominate. Near its source, the Moselle River flows over gneisses, granites, microgranites and greywacke schists of the Vosges mountains. Based on chemical composition and mineral abundances, Nédeltcheva et al. (2006) have subdivided these granitoids into three principal types (1) Type I: K-Feldspar and muscovite granites; (2) Type II: Plagioclase predominating over K-feldspar and muscovite granite (main granitoid type of the sub-basin ending at Maxonchamp (Fig. 1), representing 18% of the surface area); (3) Type III: Biotite and hornblende bearing granites (main granitoid type of the Moselotte sub-basin, representing 45% of the surface area). This part of the basin is characterized by low substratum permeability, and groundwater systems are limited to Quaternary moraines and granitic sands (Dadi, 1991). Thus, most of the precipitation water is transferred to the Moselle River by surface runoff (Jung, 1927; Gagny, 1959; Dadi, 1991). In contrast, the Permian and Buntsandstein sandstones located in the central part of the basin correspond to higher permeability layers and to the main aquifer system of the area, and inputs to the Moselle River mainly occur by infiltration (Périaux, 1961; Dadi, 1991). In the silicate part of the basin, the main tributaries of the Moselle are the Moselotte River (18% of the Moselle discharge at Epinal), which flows mainly over granites, and the Vologne River (33% of the Moselle discharge at Epinal), which also drains some sandstones. The diversity of silicate rocks is entirely integrated by the Moselle River at the city of Epinal, the location of the major lithological boundary. Downstream of Epinal, the Moselle flows over carbonate and evaporite formations of the Lorraine plateau, which can be divided into three sedimentary units: (1) the Muschelkalk formation, comprising marls, dolomites, limestones and some gypsum deposits; (2) the Keuper formation, comprising marls with gypsum and anhydrite layers, and (3) the Lias and Bajocian formations with marls and limestones. The Madon River, the main Moselle tributary downstream Epinal, drains marls, limestones and some evaporites (Fig. 1). The other tributaries are the Durbion, which mostly drains the Muschelkalk formation, and the Euron, draining mainly evaporite layers from the lower Keuper formation, as reflected in its specific and

3 5072 A. Brenot et al. / Geochimica et Cosmochimica Acta 72 (2008) Fig. 1. Geology, sampling points and river flow monitoring stations of the Moselle River basin (Nancy location: 48.42N; 06.12E). constant S isotope signature over the course of the year (Brenot et al., 2007). River flows are recorded continuously at some gauging stations and are available from the Banque HYDRO, databasis of the DIREN: French Regional Directory in Environment ( (see monitoring stations on Fig. 1). At Epinal, the annual discharge of the Moselle River is 39 m 3 /s, which represents 70% of the annual river discharge of the Moselle at Pont Saint Vincent, located downstream in the carbonate plateau (Fig. 1). The contribution of the Durbion and Euron (0.95 m 3 /s) to the Moselle in terms of water masses is relatively small compared to the Madon (10.7 m 3 /s) Sampling 3. MATERIALS AND METHODS The Moselle River and its main tributaries were sampled 10 times, all points for each sampling within a single day, between February 1999 and January The sampling strategy was mainly based on river flow and lithological

4 Magnesium isotope systematics on the Moselle river basin 5073 diversity, and samples were collected upstream and downstream of the confluence of the main tributaries (Fig. 1). Whenever possible, the sampling points were located close to Water Agency river flow monitoring stations. River water was sampled from bridges, in the middle of each channel, and filtered through a 0.20 lm nylon Millipore membrane and stored at 4 C prior to chemical and isotope analyses. Samples for cation concentration and isotope measurements were acidified with distilled HNO 3. ph values were measured directly in the field. Soils and parent rocks representative of the main lithologies (granite, limestone, dolostone, gypsum evaporite) drained by the rivers were also sampled (Fig. 1). Leaching experiments were performed on a few sedimentary rocks, in order to dissolve the calcium carbonate phase only. 12 ml of 0.3 M HCl for 20 mg of rock powder was used, and leachates were separated by centrifugation, for Ca/Mg and Mg isotope measurements Major and trace element analyses Major cation concentrations were measured in acidified water samples by ICP-AES IRIS Thermo Elemental at the SARM (French National Facilities, Nancy). Uncertainties were better than 2%. Sr content was analyzed by ICP-MS SCIEX/Perkin Elan 6000, following the procedure reported in Yeghicheyan et al. (2001). Corresponding uncertainties were better than 10%. Accuracy and reproducibility were monitored by repeat analyses of SLRS-4 reference material (international reference for river water, NRC-CNRC, Canada). Analyses of anion concentrations were performed at the LIMOS laboratory (Nancy). Cl,NO 3 and SO 4 2 were measured by ion chromatography on non-acidified water samples using a DIONEX TM series 4000I instrument with an AG-9HC/AS-9HC column and conductivity detection associated with an anion self-regenerating suppressor. The eluent used was 9 mm Na 2 CO 3 with a rate of 1.5 ml/min. Uncertainties were better than 0.5% for all the anions measured. Alkalinity was determined using charge balance calculations. Rock and soil samples were fused with LiBO 2 and subsequently dissolved in diluted HNO 3. Major elements were analyzed by ICP-AES as reported by Govindaraju and Mevelle (1987). Related uncertainties were better than 2%. Trace elements were analyzed by ICP-MS, following the method reported in Carignan et al. (2001). Uncertainties were better than 6% for Sr and better than 15% for Th. Accuracy and reproducibility were monitored by repeat analyses of rock reference materials (Carignan et al., 2001) Sr and Mg isotope analyses For Sr and Mg isotopic analyses, between a few microliters (for rivers flowing on carbonates) and 60 ml (for rivers flowing on silicates) were evaporated in Teflon Ò beakers. The residues were then dissolved in 1 N HCl and centrifuged. The residue after centrifugation was treated with HF, evaporated and re-dissolved in 1 N HCl with boric acid until complete dissolution, prior to Sr or Mg chemical separation. For Mg isotopic analysis, a few mg of powdered rocks and soils were dissolved with HNO 3 and HF. After complete dissolution, the solution was evaporated at 80 C in a Teflon Ò beaker and the residue dissolved in 10 N HCl Strontium isotope measurements The standard chemical separation technique used Eichrom Sr-Spec TM resins (Horwitz et al., 1992), following a method inspired by Pin and Bassin (1992). Procedural blank levels were lower than 300 pg, which is negligible in relation to total sample Sr and the precision of the isotopic measurements. The Sr isotopic compositions were measured using a Finnigan Mat 262 thermal ionization mass spectrometer either in static mode or in dynamic mode and normalized to 86 Sr/ 88 Sr = Analyses of the international reference material NIST-NBS987 measurements yielded 87 Sr/ 86 Sr = ± (2r, N = 20) in static mode and 87 Sr/ 86 Sr = ± (2r, N = 4) in dynamic mode. Uncertainties for individual 87 Sr/ 86 Sr measurements were (2SD) Magnesium isotope measurements For Magnesium isotopic measurements by MC ICP MS (multiple-collector inductively coupled plasma mass spectrometry), separation by ion chromatography is a prerequisite in order to avoid interferences and matrix effects. Galy et al. (2001) showed that the presence of Ca, Al and Na can induce an important instrumental isotopic fractionation of more than 1& relative to a pure Mg solution. Several authors (Galy et al., 2002; Chang et al., 2003) have developed a technique for Mg separation with a yield of 100%, and which is suitable for high precision analysis of Mg isotopes by MC ICP MS. Following the same principle, a method modified from James and Palmer (2000), using an 8.5 cm tall bed of AG50W-X12 cation exchange resin, was developed. After an initial pass through the column, Mg was completely separated from Na and Ca, and at this stage, the Mg fraction contains 15% of the K, 1.5% of the Al and 30% of the Mn of the sample. Water samples, dissolved in 0.5 ml 1 N HCl, were passed twice through this cation exchange resin. Magnesium fractions eluted from river water samples consistently had K/Mg ratios lower than 0.025, sufficient to rule-out any possible matrix effect (Galy et al., 2001). Solid samples, dissolved in 10 N HCl, were passed first through an anion exchange resin (AG1 X8) for eliminating Fe, and then six times through the cation exchange resin. The total procedural blank, after separation by ion chromatography and measured by ICP-AES, was between 4 and 22 ng, representing a maximum Mg contribution of 0.09%. We verified that our separation technique did not induce any significant isotopic fractionation by using synthetic solutions with compositions typical of silicate-draining and carbonate-draining river waters (SMS and SMC, respectively) (see Table 2 for corresponding compositions). Magnesium was taken from the NIST SRM980 CRPG reference solution (an internal isotope reference solution obtained at CRPG by dissolution of the NIST SRM980 reference material). In addition, we analyzed four geological reference

5 5074 A. Brenot et al. / Geochimica et Cosmochimica Acta 72 (2008) Table 1 MC ICP MS operating conditions during Mg isotope measurements in hard extraction Parameters Special facilities Teflon Ò nebuliser Flow rate = 100 ll/mn Cyclonic chamber Inductively coupled plasma RF power 1350 Plasma Ar flow rate 13.6 l/mn Intermediate Ar flow rate 1 l/mn Mass spectrometer Sampler cone Nickel Skimmer cone Nickel Hexapole collision cell He collision gas; flow rate = 10 ml/ mn Data aquisition parameters Faraday collectors L3: tailing peak of 23 Na; L2: 24 Mg; used Axial: 25 Mg; H2: 26 Mg; H3: 27 Al One-peak zeros 120 s acid blank acquisition procedure subtracted on line Standard-Sample 25 cycles of 10 s each; acquisition 6 9 brackets/sample Cleaning procedure HNO N for 60 s and HNO N for 5 min materials: a river water (SLRS-4, NRC-CNRC river water reference, Canada), a carbonate rock (CAL-S, CRPG reference limestone, Yeghicheyan et al., 2003), a basalt (BEN) and a diorite (DRN), both reference rocks from Geostandards ( CAL-S and DRN are from the Lorraine plateau and the Vosges mountains, respectively. Magnesium isotope ratios were measured using an Isoprobe MC ICP MS (ex-micromass, now IsotopX) at CRPG. Instrument operating conditions and set-up are described in Table 1. A standard-sample bracketing technique was used to correct for instrumental mass bias (Galy et al., 2001). 27 Al was monitored during measurement in order to verify that no aluminium was present in the Mg fraction, as well as mass All Magnesium fractions were dissolved in 0.05 N HNO 3 and adjusted to match (within ±10%) the Mg content of the bracketing reference solution (3.5 ppm). Acid blank signal intensity for 26 Mg, 25 Mg and 24 Mg never exceeded 1% of the sample signals and were systematically distributed. Since the NIST SRM980 reference material has been shown to be heterogeneous (Galy et al., 2003; Carignan et al., 2004), all results are reported relative to DSM3 (Dead Sea Magnesium, a reference material distributed by Cambridge University) in the conventional delta notation: Table 2 d 26 Mg and d 25 Mg values relative to DSM3 and their associated uncertainties (i.e. twice the standard deviation, 2r) for reference materials and fully replicated samples Sample Session d 26 Mg 2r d 25 Mg 2r N Reference material DSM3 1: March ,0 0,09 0,0 0,08 2 2: May ,1 0,12 0,0 0,14 3 3: October ,0 0,13 0,1 0,28 3 4: December ,1 0,0 1 5: January ,1 0,13 0,0 0,11 4 6: May ,0 0,1 1 Camb-1 1: March ,8 1,4 1 3: October ,7 0,41 1,4 0,27 3 5: January ,7 0,14 1,3 0,01 2 6: May ,7 0,05 1,4 0,13 6 SLRS-4 6: May ,0 0,25 0,5 0,15 3 Cal-S 7: January ,3 0,18 2,2 0,16 8 BEN 6: May ,1 0,24 0,0 0,07 3 DRN 8: January ,5 0,04 0,3 0,04 4 SRM980 CRPG All sessions 4,0 0,14 2,1 0,05 18 SRM980 CRPG through chemistry SRM980 CRPG pure 4: december ,0 0,40 2,1 0,14 2 SMS-a * 4: December ,8 0,12 2,0 0,03 2 SMS-b * 4: December ,1 0,30 2,0 0,06 3 SMS-c * 4: December ,9 0,07 1,9 0,03 3 SMC ** 1: March ,1 0,41 2,0 0,31 2 River water full replicate AB0903-3a 5: January ,0 0,23 0,5 0,25 3 AB0903-3b 5: January ,2 0,26 0,6 0,15 4 AB0903-4a 5: January ,0 0,05 0,5 0,07 2 AB0903-4b 6: May ,8 0,15 0,4 0,13 3 * SMS Synthetic solution with NIST-SRM980 CRPG, imitating the composition of typical river water on silicates: Mg = 1: Na = 7 : K = 1.1: Ca = 3.9 (molar). ** SMC Synthetic solution with NIST-SRM980 CRPG, imitating the composition of typical river water on carbonates: Mg = 1: Na = 5: K = 0.5: Ca = 5: Li = 0.5 Sr = 0.2 (molar).

6 Magnesium isotope systematics on the Moselle river basin 5075 Mg Mg sample δ Mg= ð1þ Mg Mg DSM 3 Mg Mg sample δ Mg = Mg Mg DSM 3 1 ð2þ 100 Delta values for the NIST SRM980 CRPG reference solution are d 26 Mg DSM3 = 4.00& (±0.14&) and d 25 Mg DSM3 = 2.05& (±0.05&), based on 18 measurements of DSM3. Errors for d 26 Mg and d 25 Mg are given as twice the standard deviation (2SD) throughout the text. 2SD were 0.14& and 0.10& during d 26 Mg measurements of DSM3 and Camb-1 (n = 4), respectively (not passed through the chemistry). External uncertainty, estimated using pure and doped SRM passed through the chemistry, was 0.22& (2SD) for d 26 Mg and 0.14& (2SD) for d 25 Mg (Table 2). Delta values for the Cambridge-1 pure Mg solution (internal reference material of Cambridge University) were d 26 Mg DSM3 = 2.73& (±0.10&, 2r) and d 25 Mg DSM3 = 1.38& (±0.07&) (Table 2), in close agreement with published values (d 26 Mg DSM3 = 2.60 ± 0.14&, 2.57 ± 0.13& and 2.58 ± 0.14&, obtained by Tipper et al. (2006a); Pearson et al. (2006) and Galy et al. (2003), respectively). d 26 Mg DSM3 measured for SLRS-4 is 0.98 ± 0.25&, for Cal S is 4.31 ± 0.18& and for BEN is 0.14 ± 0.24&. These values are all consistent with values given in Wombacher et al. (2006) ( 0.97 ± 0.05&, 4.38 ± 0.31& and 0.41 ± 0.19&, respectively). Linear regression of all the data in a d 25 Mg 0 vs. d 26 Mg 0 diagram (62 samples, Fig. 2), taking into account the uncertainty for each sample, yields a slope of ± 0.023, in agreement with theoretical and published values (e.g ± from Galy et al., 2001; ± fromyoung and Galy, 2004) Dissolved elements 4. RESULTS Concentrations in rivers upstream of Epinal (silicate area) In the silicate part of the basin, river ph ranges between 6.2 and 7.4. Concentrations of dissolved elements in the Moselle River and its main tributaries (Moselotte and Vologne Rivers) range between 30 and 103 lmol/l for Mg, 19 and 90 lmol/l for K, 57 and 523 lmol/l for Si and 0.15 and 0.53 lmol/l for Sr (Table 3, Brenot et al., 2007), with alkalinities of less than 850 lmol/l. Two small isolated streams of an upper catchment (Grosse Pierre and Pont Martin, sampling points #21 and #22) yield concentrations of Mg (27 and 34 lmol/l), K (14 and 7 lmol/l) and Si (69 lmol/l) that are within the ranges of the larger rivers (Table 3, Brenot et al., 2007). Concentrations of Fig. 2. Magnesium three-isotope plot showing the 62 samples analyzed (SRM980 CRPG through chemistry, Camb-1 (Cambridge-1), Cal-S, river waters, soils and rocks) and their associated 2r errors. The slope of the regression line and associated error (0.509 ± 0.023) were calculated using Isoplot linear regression of Ludwig (2001); All measurements were made relative to SRM980 CRPG, and then renormalized to DSM3.

7 Table 3 Geochemical analyses for the Moselle River and tributaries sampled between 1998 and 2003 No. River Location Distance to the source (km) Surface drained (km 2 ) Collection date Cl (lmol/l) Maxomchamp /02/ /06/ /06/ Epinal /07/ ?/02/ /02/ /06/ /06/ Girmont 87 26/02/ /06/ Portieux /07/ ?/02/ /02/ /06/ Bayon /07/ ?/02/ /02/ /06/ Messein /07/ ?/02/ /02/ /06/ Pont St 148, /07/ Vincent?/02/ /02/ /06/ Le Pont Vosges 06/12/ Martin 22 Grosse Vosges?/05/ pierre 2 Moselotte St Amé /07/ /02/ /06/ /06/ Vologne Jarmenil /07/ /02/ /06/ /06/ SO 4 (lmol/l) NO 3 (lmol/l) Alk (lmol/l) Ca (lmol/l) Mg (lmol/l) K (lmol/l) Na (lmol/l) Si (lmol/l) 5076 A. Brenot et al. / Geochimica et Cosmochimica Acta 72 (2008)

8 Magnesium isotope systematics on the Moselle river basin Durbion Pallegney 93 26/02/ /06/ Euron Froville ,9 26/02/ /06/ /07/ Madon Pont St Vincent?/02/ /02/ /06/ Madon Haroué 26/02/ /06/ Gitte Tatignécourt 15/03/ /05/ /09/ /11/ Uncertainties were better than 0.5% for anions and better than 2% for major cations. <L.D. Lower than the detection limit. dissolved Ca in this two streams (29 and 32 lmol/l) are lower than in the Moselle River ( lmol/l) but the range documented for dissolved Ca in small streams in the Vosges mountains is relatively large ( lmol/l; Probst et al., 2000; Nédeltcheva et al., 2006). In contrast with other major elements, concentrations of Na ( lmol/l) and Cl ( lmol/l) are much higher in the Moselle River and its tributaries than in the two streams studied (42 and 27 lmol/l for Na; 33 lmol/l for Cl), or in other Vosgian streams documented by Probst et al.(2000)and Nédeltcheva et al. (2006). This strongly suggests that significant Cl and Na are acquired from anthropogenic inputs, most probably from NaCl road-salts. Indeed, the total amount of Cl and Na spread per year as road-salt in the basin would represent 80% and 40% of the annual flows of Cl and Na, respectively in the Moselle at Epinal. Thus Cl concentrations cannot be used in this area to correct the dissolved river load for atmospheric inputs (Stallard, 1980; Stallard and Edmond, 1981; Meybeck, 1983; Négrel et al., 1993). The atmospheric contribution was therefore calculated using the composition of mean open field precipitation sampled in the Vosges mountains (Cl = 15 lmol/l; Probst et al., 2000), assuming 38% evapotranspiration (calculated for the silicate part of the basin using Meteo France data and Bénichou and Le Breton (1987)). According to these calculations, atmospheric inputs contribute less than 15% of the dissolved Mg and less than 20% of the dissolved Sr in river water Concentration in rivers of the Lorraine plateau (carbonate dominated area) Alkalinities of rivers draining the Lorraine plateau are greater than for the silicate area, and range between 3300 and lmol/l, with ph ranging between 7.2 and 8.5. Taking into account the local composition of mean open field precipitation, atmospheric inputs were calculated to represent less than 10% of dissolved Mg and less than 6% of dissolved Sr in river water. All samples were collected from across the entire watershed within a single day, allowing for a direct comparison of the dissolved concentrations. An increase in the concentrations of Ca ( lmol/l), Mg ( lmol/l), and Sr ( lmol/l) in the Moselle River between Epinal and Pont Saint Vincent (the outlet of the studied area) is systematically observed for all sampling days, and can be related to the dissolution of carbonates and evaporites present on the Lorraine plateau (Brenot et al., 2007). Indeed, tributaries (Durbion, Euron, Gitte, Madon) draining mainly sedimentary units of the Lorraine plateau display very high concentrations of SO 4 2 ( lmol/l), Ca ( lmol/l), Mg ( lmol/ l), and Sr (5 94 lmol/l) (Table 3, Brenot et al., 2007). Cl contents ( lmol/l) are also significant, especially for the Madon tributary where dissolution of halite, present in the local lithology, is a relatively important source of Na and Cl (Brenot et al., 2007). For a sampling campaign in February 2002 of the Madon River along its course, the Ca, Mg and Sr contents of the Madon River also increased significantly downstream, from 21 to 4175 lmol/l for Ca, from 30 to 2023 lmol/l for Mg and from 0.1 to 27 lmol/l for Sr. The Madon River chemistry before its confluence with the

9 5078 A. Brenot et al. / Geochimica et Cosmochimica Acta 72 (2008) Moselle (at Pont Saint Vincent) seems therefore to be buffered by weathering of carbonates and evaporites, when compared with the chemical signal inherited from Buntsandstein sandstones upstream Sr and Mg isotopic compositions The section of the Moselle River and its tributaries that flow over silicate rocks yield radiogenic Sr isotopic compositions, with 87 Sr/ 86 Sr ratios ranging between and The two streams from the upper catchment (Grosse Pierre and Pont Martin) yield more radiogenic compositions ( and ). The Sr isotopic composition of the Moselle River is less radiogenic downstream of Epinal (down to ), where the river also drains carbonates, marls and evaporites of the Lorraine plateau. All tributaries draining only these formations (Durbion, Euron, Gitte, Madon) have low 87 Sr/ 86 Sr (from to ). Thirty four water samples from the Moselle River basin were analyzed for Mg isotopic composition (Table 4). An overall variation of 0.9& in d 26 Mg is found, representing 4.1 times the estimated external uncertainty (Section 3.3.2). Magnesium dissolved in the Moselle River and tributaries in the silicate part of the basin display d 26 Mg values higher than 0.9& except for two values (d 26 Mg values included in the range 1.2& to 0.7&). The streams with the most radiogenic Sr isotopic compositions also yield the highest d 26 Mg values (d 26 Mg = 0.7&, Table 4). For comparison, other small basaltic and granitic rivers also display d 26 Mg greater than 0.9&, ranging between 0.86& and 0.31& (Tipper et al., 2006b). De Villiers et al. (2005) reported highly depleted d 26 Mg, between 3.8& and 2.0&. for rivers draining granite, sandstone and lava in Swaziland (d 26 Mg values were recalculated relative to DSM3, based on reported seawater analyses in this study and seawater value (d 26 Mg DSM3 = 0.8&) published in Young and Galy (2004) and Carder et al. (2004)). At Epinal, the Moselle River d 26 Mg is constant throughout the year ( 0.8&). From Epinal to Pont Saint Vincent, on the Lorraine carbonate plateau, the d 26 Mg of the Moselle River ranges from 0.8& to 1.4&. For comparison, rivers draining Himalayan mixed lithology catchments display d 26 Mg values ranging between 1.8& and 0.8& (Tipper et al., 2006a). In larger rivers, d 26 Mg values range between 1.7& (for the Mackenzie) and 0.5& (for the Nile) (Tipper et al., 2006b). The Madon River, draining carbonates and evaporites, displays relatively constant and homogeneous d 26 Mg, ranging between 1.6& and 1.4&. The Euron, draining mainly evaporites, similarly displays low and relatively constant d 26 Mg (from 1.4& to 1.2&). All tributaries draining carbonate and evaporite formations display d 26 Mg values lower than 1.1& Rocks, soils and leachates Chemical and Mg isotopic compositions were determined for a Vosgian diorite (DRN), sedimentary rocks, and associated soil samples (Table 5). Soils developed on silicates have greater LOI (loss on ignition) values ( %) than their related parent rocks ( %). Since Ca contents are broadly the same in rocks and soils (Table 5), this suggests enrichment in secondary phyllosilicate mineral phases in the soils. Mg/Al ratios are broadly similar in silicate rocks and soils ( ). Soils developed on limestones display a strong depletion in carbonate phases as indicated by significantly lower Ca contents (CaO = 15.2% and 1.0% for two soil samples, and 50.4% and 53.0% for their respective parent rocks) (Table 5). These soils are also greatly enriched in silica (SiO 2 = 63.7 and 62.8%) relative to their parent limestones (SiO 2 = 5.3% and 2.8%). This indicates that limestones contain some silicate phases, mainly phyllosilicate minerals, and that, during weathering, the calcium carbonate phase is preferentially dissolved. In contrast, Mg contents are broadly similar in source limestones (MgO = 1.2% and 0.6%) and their associated soils (MgO = 1.7 and 1.1%). Ca/Mg ratios are also similar in sedimentary rocks and in acid leachates (not shown). The DRN Vosgian diorite displays a d 26 Mg of 0.53&, close to published values for a High Himalayan paragneiss and a loess ( 0.42& and 0.6&, respectively; Young and Galy, 2004; Tipper et al., 2006a). One gypsum sample was analyzed, for which d 26 Mg is 0.8&, identical to the present-day seawater value of d 26 Mg = 0.8& (Young and Galy, 2004; Carder et al., 2004; De Villiers et al., 2005). d 26 Mg is 0& for the soil developed on granites. This is within the range obtained for Himalayan soils ( 0.11& to 0.02&, Tipper et al., 2006a). Two limestones, both characterized by low Si, Mg and Al concentrations, display significantly different d 26 Mg values ( 4.5& and 1.0&). The dolomitic limestone (R4, 12.4% MgO, Table 5) yields d 26 Mg of 1.4&. These values are within the published range for similar rocks (Galy et al., 2002; Carder et al., 2004; De Villiers et al., 2005). Soils developed on limestones are enriched in heavy isotopes relative to their source rocks, with d 26 Mg of 0.6& and 1.3&. Acid leachates, aimed at preferentially dissolving carbonates from the limestones and the dolomitic limestones, display d 26 Mg similar, within analytical uncertainty, to the corresponding rocks (Table 5) Silicate Vosgian rivers 5. DISCUSSION In the silicate (Vosgian) part of the basin, river waters are characterized by 87 Sr/ 86 Sr ratios ( ) that are intermediate between values for mean local precipitation ( and ; Probst et al., 2000; Chabaux et al., 2005) and values for mineral separates and granites from the Vosges mountains ( , Bonhomme, 1967; France-Lanord, 1982; Probst et al., 2000; Aubert et al., 2001; Aubert et al., 2004). Fig. 3 is a plot of Sr isotopic compositions vs. Mg/Sr molar ratio for rivers flowing over silicates compared to values reported for mineral separates, bulk bedrock and initial 87 Sr/ 86 Sr ratio of local granites. As described in Section 4.1.1, atmospheric inputs represent a maximum of 20% of the dissolved Sr in these rivers. The dominant input of Sr in the silicate part of the

10 Magnesium isotope systematics on the Moselle river basin 5079 Table 4 Sr concentrations (Uncertainties 10%) and Sr and isotopic composition for the Moselle River and tributaries sampled between 1998 and 2004 No. River Location Collection date d 26 Mg (&) 2r d 25 Mg (&) 2r N Sr (&) 87 Sr/ 86 Sr 2r/ Moselle Maxomchamp 11/03/02 ** 0,19 0, /09/02 ** 0,31 0, /01/03 ** 0,7 0,3 0,4 0,3 3 0,20 0, /05/03 ** 0,9 0,3 0,4 0,2 2 0,26 0, /06/03 0,30 0, /09/03 ** 0,9 0,3 0,4 0,2 3 0,33 0, Epinal 08/07/98 0,43 0, /03/02 ** 0,25 0, /09/02 ** 0,38 0, /01/03 ** 0,8 0,3 0,5 0,2 2 0,23 0, /05/03 ** 0,8 0,4 0,4 0,4 3 0,31 0, /06/03 0,37 0, /09/03 ** 0,8 0,3 0,4 0,2 3 0,44 0, Girmont 12/03/02 ** 0,33 0, /09/02 ** - 0,64 0, /01/03 ** 0,28 0, /05/03 ** 0,50 0, /09/03 ** 0,83 0, Portieux 18/09/02 ** 0,9 0,3 0,4 0,2 2 1,17 0, /01/03 ** 0,8 0,3 0,5 0,2 2 0,44 0, /05/03 ** 1,1 0,3 0,5 0,2 2 1,09 0, /09/03 ** 1,4 0,3 0,7 0,2 2 1,52 0, Bayon 08/07/98 1,3 0,3 0,7 0,2 2 3,05 0, /09/02 ** 2,75 0, /01/03 ** 1,17 0, /05/03 ** 2,11 0, /09/03 ** 2,91 0, Messein 08/07/02 1,4 0,3 0,7 0,2 2 7,06 0, ?/02/01 2,49 0, /03/02 ** 3,54 0, /01/03 ** 1,0 0,3 0,5 0,2 2 2,14 0, /05/03 ** 1,3 0,3 0,7 0,2 3 4,69 0, /09/03 ** 1,3 0,3 0,7 0,2 2 6,37 0, Pont St Vincent 08/07/98 14,37 0, ?/02/01 3,86 0, /03/02 ** 3,54 0, /09/02 ** 14,36 0, /01/03 ** 8,16 0, /05/03 ** 36,74 0, /09/03 ** 23,24 0, Le Pont Martin Vosges 06/12/04 *** 0,7 0,3 0,4 0,2 2 0,10 0, Grosse pierre Vosges?/05/01 0,7 0,4 0,3 0,3 2 0,08 0, St Amé 11/03/02 ** 0,16 0, /09/02 ** 0,26 0, /01/03 ** 0,8 0,3 0,4 0,2 2 0,15 0, /05/03 ** 0,24 0, /06/03 0,25 0, /09/03 ** 0,7 0,3 0,3 0,2 3 0,28 0, Jarmenil 18/09/02 ** 0,38 0, /01/03 ** 0,9 0,3 0,5 0,2 3 0,28 0, /05/03 ** 1,1 0,3 0,6 0,2 3 0,42 0, /06/03 0,43 0, /09/03 ** 1,2 0,3 0,6 0,2 4 0,53 0, Durbion Pallegney 12/03/02 ** 7,60 0, /09/02 ** 13,01 0, /01/03 ** 4,55 0, /05/03 ** 15,84 0, /09/03 ** 18,98 0, Euron Froville 12/03/02 ** 1,3 0,3 0,6 0,2 1 46,70 0, /09/02 ** 79,00 0, /01/03 ** 1,4 0,3 0,7 0,4 3 32,22 0, /05/03 ** 1,2 0,4 0,6 0,2 3 71,56 0, (continued on next page)

11 5080 A. Brenot et al. / Geochimica et Cosmochimica Acta 72 (2008) Table 4 (continued) No. River Location Collection date d 26 Mg (&) 2r d 25 Mg (&) 2r N Sr (&) 87 Sr/ 86 Sr 2r/ Madon 08/07/98 41,17 0, Pont St Vincent?/02/01 8,17 0, /09/02 ** 40,27 0, /02/03 20,02 0, /05/03 ** 39,02 0, /06/03 18,58 0, /09/03 ** 1,5 0,3 0,7 0,2 3 47,02 0, Madon Haroué 18/09/02 ** 44,08 0, /01/03 ** 1,4 0,3 0,7 0,3 5 10,70 0, /05/03 ** 1,6 0,3 0,7 0,2 2 42,99 0, /09/03 ** 1,5 0,3 0,7 0,2 2 55,83 0, Gitte Tatignécourt 15/03/03 1,3 0,3 0,6 0,2 2 17,13 0, /05/03 1,1 0,3 0,6 0,2 2 24,07 0, /09/03 1,2 0,3 0,5 0,2 2 32,25 0, /11/03 1,4 0,4 0,7 0,3 4 15,07 0, Major concentrations of these samples are reported in Table 3 and in Brenot et al. (2007). <L.D. Lower than the detection limit. ** Major elements concentrations published in Brenot et al. (2007). Moselle River basin is therefore the weathering of silicate rocks. The positive correlation between 87 Sr/ 86 Sr and Mg/ Sr ratios reflects preferential weathering of Mg-rich and radiogenic mineral phases such as biotite ( 87 Sr/ 86 Sr = , Bonhomme, 1967; Probst et al., 2000; Aubert et al., 2001) (Fig. 3). A similar correlation is observed between 87 Sr/ 86 Sr and K/Sr, which also supports the preferential weathering of other mineral phases such as K-feldspars ( 87 Sr/ 86 Sr = , Probst et al., 2000; Aubert et al., 2001). The range of 87 Sr/ 86 Sr shown by the silicate rivers of the Moselle basin might be best explained by the disparities in silicate mineral compositions and abundances within the sub-basins (Fig. 3). Nevertheless, according to the literature, local precipitations are expected to display intermediate Mg/Sr molar ratio and 87 Sr/ 86 Sr values between local rocks and separated minerals (Fig. 3). Thus mixing line in Fig. 3 could also be partly due to mixing with precipitation. The magnesium isotopic compositions of Vosgian rivers (d 26 Mg = 1.2& to 0.7&) are significantly lower than the value estimated for the continental crust and Vosgian diorite ( 0.5& this study, Young and Galy, 2004; Tipper et al., 2006a) (Fig. 4). In contrast, the d 26 Mg of silicate soil samples are enriched in heavy isotopes, with values close to 0& for both the Moselle and Himalayan soils (Tipper et al., 2006a) (Fig. 4b). This suggests that Mg isotopes fractionate during silicate weathering, resulting in the preferential release of light isotopes into the aqueous phase. This observed fractionation of Mg isotopes between source rocks and dissolved phases can not be an artifact due to anthropogenic inputs of dissolved Mg in river systems. Indeed the previous study of Brenot et al. (2007) considered that only road salt contributions may alter the natural geochemical signature of river water in the silicate (Vosgian) part of the basin, where forests are dominant. Potential contributions of Mg from road salt leaching would represent less than 4% of the annual flux of dissolved Mg in the Moselle at Epinal. Such a contribution would generate a maximum decrease of 0.04& for d 26 Mg values in river water. This is not significant compared to the difference of values observed between source rocks and dissolved phases (d 26 Mg rock-water between 0.4& and 0.7&). A number of different processes may be responsible for the observed Mg isotopic fractionation between source rocks and dissolved phases, such as secondary mineral formation, mineral leaching, isotope exchange, uptake by plants, or organic litter decomposition. In the simplest case, it might be assumed that Mg isotopes do not fractionate during mineral leaching, and that only clay formation significantly fractionates Mg isotopes. Following a Rayleigh law, soil d 26 Mg (0&) associated with d 26 Mg in waters ranging between 1.1& and 0.7&, would be consistent with d 26 Mg clay-water isotope fractionation ranging between 0.4& (for <40% of Mg uptake by clays) and 0.8& (for 60% of Mg uptake by clays). However, the Mg/Sr molar ratios of the dissolved phases are broadly similar to the Mg/Sr ratios measured in local rocks and soils. The Mg/Al ratios are also similar in silicate rocks and soils. This suggests that little Mg is involved in clay neoformation in the Vosges. Moreover, in river waters, the highest 87 Sr/ 86 Sr values are associated with the highest d 26 Mg values (Table 4). Mg and Sr leaching from source rocks and minerals therefore also appears to be a key process, leading to Mg isotope fractionation. Following a Rayleigh law, the corresponding d 26 Mg rock-water is estimated to range between 0.4& and 0.7& (with a corresponding fraction of Mg of less than 70% released in the aqueous phase). River waters have d 26 Mg lighter than host rocks. Black et al. (2006) have shown that plant material prefers light Mg isotopes, thus uptake by plants is certainly not a dominant process explaining isotopic fractionation. More data and experiments are needed in order to determine and calibrate the processes that fractionate Mg isotopes. However, overall, these data show that d 26 Mg of silicate rocks cannot be used directly as an end-member for natural waters, and highlight the potential of Mg isotopes as a tracer of processes such as clay formation, weathering intensity or recycling by plants at a watershed scale.

12 Table 5 Major, trace element concentrations and Mg isotopic signatures of selected rocks and soils No. Location Nature d 26 Mg (&) R1a Col de Grosse pierre 2r d 25 Mg (&) 2r SiO 2 Al 2 O 3 Fe 2 O 3 MnO MgO CaO Na 2 O K 2 O TiO 2 P 2 O 5 LOI ** Total Sr CIA *** (molar (ppm) ratio) Granite * 0,2 0,50,0 0,2 63,7 13,9 4,4 0,1 4,1 2,7 2,5 6,5 0,7 0,5 0,9 100, ,46 R1b Soil on granite 0,5 0,40,2 0,3 63,3 13,9 4,3 0,1 2,4 1,0 1,4 6,3 0,7 0,5 6,3 100, ,57 R1c Soil on granite 0,4 0,30,1 0,2 59,6 14,3 5,6 0,1 4,9 2,8 1,7 5,8 0,9 0,6 3,8 100, ,18 R2a Zainvillers Granite 0,5 0,30,5 0,2 70,4 15,4 2,1 <L.D. 0,6 0,4 1,5 7,5 0,4 0,2 1,5 99, ,84 R2b Soil on granite 0,0 0,20,0 0,1 57,3 12,8 3,1 <L.D. 0,6 0,1 0,9 4,8 0,4 0,3 20,0 100, ,04 R3a Adoncourt Limestone * 1,0 0,1 0,5 0,3 5,3 1,3 0,7 0,0 1,3 50,4 0,1 0,4 0,1 <L.D. 40,8 100,4 351 R3a-leached a 1,2 0,1 0,7 0,1 R3b Marly soil 0,6 0,20,2 0,1 63,7 13,9 4,1 0,1 1,7 15,2 0,2 3,9 0,6 0,1 17,4 99,8 89 R4 Girmont Dolomitic 1,4 0,2 0,7 0,1 16,3 5,1 3,3 0,4 12,4 25,2 0,1 1,7 0,2 0,1 35,6 100,4 138 limestone R4-leached a 1,3 0,3 0,7 0,1 R5a Zincourt Limestone * 4,5 0,3 2,4 0,1 2,8 0,6 <L.D. <L.D. 0,7 53,1 <L.D. 0,3 <L.D. <L.D. 42,3 99,7 528 R5b Clayey soil 1,3 0,3 0,6 0,4162,8 14,1 6,9 0,4 1,2 1,1 0,4 3,7 0,7 0,4 8,8 100,4 105 R6 Circourt Gypsum 0,8 0,2 0,4 0,1 2,9 0,6 0,1 <L.D. 0,6 32,4 <L.D. 0,1 <L.D. <L.D. 60, R7 Bayon Clayey marl 63,8 13,4 3,8 0,0 2,2 2,6 0,1 6,2 0,8 0,4 6,4 99,7 50 R7-leached a 0,4 0,20,3 0,3 R8a Rozelieure Dolostone 2,1 0,6 0,3 0,0 20,8 30,5 <L.D. 0,2 <L.D. <L.D. 45,7 100,1 101 R8a-leached a 1,5 0,2 0,7 0,2 R8b Marly dolomitic soil 42,6 9,5 3,8 0,1 7,3 10,4 0,3 3,0 0,5 0,2 21,7 99,3 88 R8c Dolostone weathered 0,7 0,2 0,5 0,0 21,3 30,3 <L.D. 0,1 <L.D. <L.D. 46,7 99,7 106 R8d Dolostone weathered 3,7 1,2 0,6 0,1 20,3 28,5 0,1 0,3 <L.D. <L.D. 44,9 99,5 93 R8e Calcareous dolomitic soil 21,8 6,8 2,5 0,1 13,9 19,2 0,1 1,9 0,3 <L.D. 33,3 99, Madon at Pont bed load sediments 28,5 2,7 12,1 1,0 28,4 0,2 0,6 449 Saint Vincen Uncertainties were better than 2% for major elements and better than 6% for Sr. <L.D., Lower than the detection limit. Fe2O3 = 0.1; MnO = 0.03; Na 2 O = 0.15; TiO 2 = 0.03; P 2 O 5 = * Samples submitted to full replicates. ** LOI, Lost in ignition, volatilized elements (mass percent) loss after heating at 980 C. *** CIA, chemical index of alteration using molecular proportions: CIA = [Al 2 O 3 /(Al 2 O 3 + CaO + Na 2 O+K 2 O)] 100. a Leaching experiments consist in 12 ml 0.3 HCL for 20 mg rock powder and separation of leachates by centrifugation. Magnesium isotope systematics on the Moselle river basin 5081

13 5082 A. Brenot et al. / Geochimica et Cosmochimica Acta 72 (2008) Fig Sr/ 86 Sr vs. Mg/Sr molar ratio for rivers flowing over silicates compared to values reported for mineral separates, bulk bedrock and initial 87 Sr/ 86 Sr ratio of local granites (from Granite des Crêtes and Granite de Brézouard documented by (a) Bonhomme (1967); (b) France-Lanord (1982); (c) Probst et al. (2000); (d) Aubert et al. (2001)). Biotite 87 Sr/ 86 Sr = ; albite 87 Sr/ 86 Sr = ; apatite 87 Sr/ 86 Sr = ; bulk granite 87 Sr/ 86 Sr = ; initial 87 Sr/ 86 Sr = Simulated silicate weathering endmember of Probst et al. (2000) = Mean open field precipitation 87 Sr/ 86 Sr = and , as documented by (c) Probst et al. (2000) and (e) Chabaux et al. (2005), respectively. 87 Sr/ 86 Sr of mean throughfall precipitation = , as reported in (c) Probst et al. (2000) Mg excess in rivers draining carbonates Rivers draining the Lorraine plateau formations display high alkalinities and high Ca contents, mainly attributed to carbonate weathering. However, Ca/Mg molar ratios ( ) are significantly lower than those of local carbonates (29 59 for limestones, Fig. 5). Similarly, the range of Ca/ Sr is lower (80 to 400) than estimated for carbonate rocks (370 to 2200), and a positive correlation between Ca/Sr and Ca/Mg can be highlighted in the Moselle basin waters (Fig. 5). Low Ca/Mg ratios in dissolved river loads have already been documented by Meybeck (1984) for waters draining marls and evaporites of the Marnes irisées Inférieures formation (Upper Keuper formation) of the Lorraine plateau. Ca/Mg ratios for these waters are similar to, or even lower than, Ca/Mg ratios for rivers draining silicate bedrocks. Usually, carbonate dissolution is considered to be a congruent process, with no chemical fractionation. Consequently, Ca/Mg and Ca/Sr ratios are expected to be similar in the dissolved phase and corresponding carbonate solid phase, as shown by our leaching experiments (Table 5). The observed dichotomy in the Moselle basin highlights a significant Mg excess and, to a lesser extent, a Sr excess (relative to Ca). The Mg excess is even more important if we consider that dissolved Ca is also partially derived from the dissolution of evaporite (gypsum), as shown in Brenot et al. (2007). In the literature, only a few cases of Mg and Sr excess relative to Ca are highlighted and explained (Kotarba et al., 1981; Sarin et al., 1989; Galy et al., 1999; Jacobson et al., 2002). On a more global scale, most carbonate rivers display low Ca/Mg ratios, between 0 and 5, with 87 Sr/ 86 Sr values between and (Fig. 6; Sarin et al., 1989; Petelet et al., 1998; Gaillardet et al., 1999; Galy, 1999; Roy et al., 1999; English et al., 2000; Jacobson et al., 2002; Millot et al., 2003). First, we consider whether silicate minerals present as impurities in limestones could explain this Mg excess. Indeed, these waters also display 87 Sr/ 86 Sr values that are slightly more radiogenic than pure Triassic carbonates ( 87 Sr/ 86 Sr 0.707, Koepnick et al., 1990; Martin and Macdougall, 1995; Korte et al., 2003), and this difference is best explained by some contribution from silicate material, either from the Vosges mountains (for the Moselle River), or present within the carbonate formations. For the Moselle River, simple calculations show that Sr originating from the weathering of the silicate phase would represent at most 16% of the dissolved Sr (assuming that phyllosilicates are inherited from erosion of the local Paleozoic granite, and using the mean composition of the Moselle River at Epinal, 87 Sr/ 86 Sr = ). Moreover, the extremely low Ca/Mg ratio measured for dissolved phases of the Lorraine plateau is difficult to explain by a significant contribution from the detrital phase of marls and carbonate formations. Indeed, it is unlikely that phyllosilicates have the capacity to weather easily, compared to calcite, and to release significant amounts of Mg into water. On the Lorraine plateau, dominated by agricultural land use, Mg-fertilizers could potentially modify natural Ca/Mg ratios and the d 26 Mg signature of river water. The Mg-fertilizers that are potentially used in this area are derived from dolostones and thus they display low Ca/Mg ratio (1.05 for R8a, Table 5) and they are expected to have d 26 Mg signatures similar to local dolostone rocks