Introduction

Geothermal development consists in harvesting the energy naturally generated by hydrothermal activity. Geothermal exploration is therefore greatly facilitated by the presence of obvious surface markers of hydrothermal activity such as hot springs or fumaroles. While there are plenty of efficient methods for exploring geothermal systems displaying such evidence, less visible systems prove hard to uncover, thus restraining the development of geothermal energy worldwide. Yet, with the acceleration of climate change from the use of fossil fuels, geothermal energy represents a necessary renewable and local alternative. To address this issue, the scientific community in all Earth-sciences related disciplines has been developing exploration strategies for almost two decades, specifically targeting “hidden” or “blind” geothermal systems, i.e., with no obvious surface evidence of hydrothermal activity1,2,3,4,5,6,7. For high-enthalpy systems and conventional geothermal use, the first milestone is to determine whether or not there is hydrothermal activity, as it represents the base condition. So far, few approaches have been proposed using water geochemistry. Indeed, the widely used geochemical tools in non-blind systems rely on the study of hot springs and fumaroles, absent in blind systems. Yet, small hydrothermal components are often highlighted in groundwater studies in volcanic environments8,9,10,11,12,13,14,15,16,17,18,19,20,21,22 and could therefore theoretically constitute an exploration tool. In most cases the quantification of the hydrothermal contribution is made using anionic mass budgets (Cl, SO4, HNO3) but also trace elements concentrations and isotopic ratios that are typically enriched (B, Li, …) or depleted (Se, Ge) or fractionated (δ7Li, δ11B, δ34S, δ13C, δ30Si) in the hydrothermal fluids compared to the “cold” waters. Only in the recent exploration campaigns in the islands of Hawai’i, unconventional geochemical tools were used on cold groundwaters, whose compositions result mostly from non-hydrothermal processes, which along with other methods proved efficient to evidence the existence of hidden geothermal systems3. In these ocean island volcanoes, it is assumed that strong lateral groundwater flows conceal the rise of hot fluids and that to some extent, these hot fluids are diluted in a wider hydrogeological network. While bulk geochemical data for these “mixed” groundwaters fall into the range of expected values for “pure” groundwaters, by comparing the temperatures, Cl/Mg concentration ratios and SiO2 concentrations between groundwaters samples harvested at the scale of the islands, Lautze et al.3 were able to decipher geochemical anomalies associated with hydrothermal activities in some samples. Ferguson23 showed that these waters also displayed anomalous He content and R/Ra ratios. These studies touch upon the possibilities of using geochemical markers of hydrothermal activity in cold groundwaters as an exploration tool for the first stage of geothermal exploration in blind systems, which is the approach we develop here on a larger scale.

We chose the Piton de la Fournaise volcano (Réunion Island) as our natural laboratory. This ocean island volcano displays no visible surface indicators of hydrothermal activity despite an ongoing frequent volcanic activity. As in other blind systems, geophysical studies as well as studies on the geochemistry of soil gas have been conducted for geothermal exploration, which concluded to the possible presence of hydrothermal systems at different locations in the volcanic edifice24. Additionally, in unrelated studies, observed peculiarities in the Radon content and the isotopic composition (δ13C and δ11B) of several cold groundwater samples of the massif have tentatively been attributed to the presence of magmatic fluids25,26,27. Therefore, to date, the presence of a hidden geothermal system on this volcano, although strongly suggested, is still questioned, making it an ideal candidate for the development of new exploration strategies such as the study of the chemistry of cold waters, as a complementary approach to the other methods. Here, we acquired a large dataset both in terms of samples and analyzed parameters to try and find markers of a hydrothermal signature and fit them in a hydrogeological circulation model. We sampled 80 cold springs, groundwater wells, rivers and streams on the volcanic massif (Fig. 1A) and analyzed them for major and trace elements concentrations and δ11B and δ13C isotopic ratios.

Fig. 1: Locations and main geochemical features of Piton de la Fournaise waters.
figure 1

A Map of the different water samples of Piton de la Fournaise collected for this study and (B) Na/Cl vs. Na/NO3 molar ratios and electrical conductivity (µS/cm) of Piton de la Fournaise waters, showing that water rock interaction, anthropogenic contamination and seawater contamination are the main mineralization processes. Insert: ©2023 Google Earth.

Full size image

We show that high temperature water-rock interaction and/or dissolution of magmatic gases can be evidenced in Piton de la Fournaise waters by the relative enrichments in As, Rb, P, V, Mo, Sb, F, Li, SO4, HCO3 and B, as well as by the isotopic ratios δ11B and δ13C. With these geochemical markers of hydrothermal activity, we built a unique index that measures the relative probability of the presence of a hydrothermal component in waters. The spatial distribution of waters with the most evidence of hydrothermal activity indicates that a caldera fault system surrounding the Plaine des Sables region is a preferential pathway for rising hydrothermal fluids. These results are consistent with previous geophysical exploration studies which identified a dense body and a conductive layer as a potential geothermal resource in this region24,28,29.

Results and discussion

Identifying geochemical markers of hydrothermal activity

At Piton de la Fournaise, the aquifer system is unconfined and constituted by stacked lava flows30. Lavas are transitional between the tholeiitic and alkaline fields and have a range of compositions from picrite to trachyandesite, with a large majority of aphyric basalt31,32,33. At a regional scale, permeability is decreasing from the surface to the depth of the edifice, the deeper units being older and thus more affected by secondary mineralization processes. These oldest units act as a wall for interconnected groundwater flows of regional extent with waters going from the inner to the outer part of the edifice. Preferential pathways or isolated circulations can result from local discontinuities in the stacked lava flows, such as permeable paleo-valleys and less permeable paleo-soils30,34,35. Because the aquifer system is interconnected and of regional extent, we expect that the presence of a hydrothermal system would result in a dilution of thermal waters into the aquifer system, as in Hawai’i3.

There are no resurgences of waters displaying a strong thermal component at Piton de la Fournaise, which was confirmed by our extensive fieldwork. From their interaction with hot rocks, hot fluids, and magmatic gases, thermal waters are hot and/or acidic and/or with a high total dissolved solid content. Piton de la Fournaise waters wave low temperatures, ranging between 8.5 and 21.1 °C, and neutral pH values ranging between 6.5 and 8.6. Most samples also display low electrical conductivities, although five (TUN1, TUN2, TUN3, SPH3, SPI3) display values superior to the others (266 to 685 µS/cm against a mean of 129 ± 100 µS/cm, Supplementary Data 1 and Fig. 1B).

The mineralization of rainwaters, which infiltrate the volcanic edifice, is mostly influenced by sea spray and thus have a Na/Cl ratio close to that of seawater, with however low Na and Cl concentrations36. From this initial composition, as water circulates through the ground, its composition evolves according to different mineralization processes. In Fig. 1B, the comparison of Na/Cl and Na/NO3 molar ratios gives three trends corresponding to three common mineralization processes: water-rock interaction, seawater contamination, and anthropogenic contamination. As the water residence time in the aquifers increases, water-rock interaction increases the Na concentration through the dissolution of silicate bearing minerals36, giving the first trend. Additionally, seawater may contaminate samples collected at river mouths and in wells close to the coast (SPH2, SPH3, SPH4), increasing both Na and Cl concentrations36, giving the second trend. Finally, the presence of urban areas and agricultural lands located on the catchment areas of several sampled watersheds can lead to the infiltration of artificially chlorinated waters from the domestic water supply, of organochlorine compounds and muckspreading used for agriculture and of wastewaters. These anthropogenic pollutions result increased Cl and NO3 concentrations in waters, giving the third trend. The electrical conductivity of Piton de la Fournaise waters (blue to red scale, Fig. 1B), which is a proxy for their total dissolved mineral content, increases along each trend. This means that those three processes contribute the most to the geochemical composition of the waters. Here, we are searching for evidence of mineralization associated with hydrothermal processes. Seawater contamination and anthropogenic contamination are not hydrothermal processes. Water-rock interaction can occur at low temperature, but it is facilitated by heat and acidity, and can thus be increased by hydrothermal activity. However, from the data presented in Fig. 1B alone, it is not possible to decipher whether water-rock interaction evidenced here is associated with hydrothermal activity.

For a broader understanding of the geochemical composition of our samples, and in order to highlight a possible hydrothermal contribution, we performed a principal component analysis (PCA) on a large geochemical dataset of major and trace element compositions (variables) of our water samples from Piton de la Fournaise (individuals).

Among all of the analyzed elements, the variables retained for interpretation are:

  • Major elements (except HCO3 for which not all samples had available data),

  • Trace elements that can originate from hydrothermal activity Li, B, Br, F, P, Sb, Mo, Rb, V, As, Sr,

  • SiO2 as a maker of water rock interaction due to the weathering of silicate minerals30,37,

  • NO3 as a marker of anthropogenic contamination.

We did not include HCO3 in the PCA because it requires a homogeneous dataset and HCO3 concentrations had not been measured for some samples.

Using elemental concentrations for the PCA gives larger weights to the elements with the highest concentrations. Since in our case the hydrothermal component may not constitute the main geochemical feature, and since most elements associated with hydrothermal activity are trace elements, we chose to rather study the enrichment in each element compared to the others for each sample. To obtain this enrichment, we normalized the data by dividing each elemental concentration by the sum of the concentrations of the major cations (Ca, Mg, Na, K).

The results of the analysis are presented in Fig. 2A, B, which display the contribution of each chemical element to the two principal components, and the distribution of water samples on these components, respectively.

Fig. 2: Principal component analysis of the major and trace elements composition of Piton de la Fournaise waters.
figure 2

Distribution of (A) variables and (B) indices for components 1 and 2, representing 58.86% of the total variance.

Full size image

The first component explains 44.98% of the total geochemical variability. The elements contributing the most to this component are F, P, Sb, Mo, Rb, V, K, As, and to some extent SiO2, Na and Li. F, P, Sb, Mo, Rb, V, K, As plot close together and define a pole. These chemical elements are found in magmatic gases, their condensates and associated deposits38,39,40,41, as observed at Piton de la Fournaise42,43, which can dissolve in waters. But in our case, F, P, Sb, Mo, Rb, V, K and As plot close to SiO2, which is a marker of water-rock interaction, so we preferably explain their presence by a high temperature interaction with differentiated rocks44,45,46,47,48, and thus by hydrothermal activity. By comparison, at the other volcano of La Réunion, Piton des Neiges, thermal springs display high concentrations in F, Rb, K, As and SiO2 but not in V and P, and there is no available data for Mo and Sb49. Elements that can form oxyanions such as P, V, Mo and Sb have a higher solubility in oxidizing environments47,50,51,52. Therefore, they are more likely to be present in an unconfined aquifer such as in Piton de la Fournaise, especially at the top of the water column, where the conditions are more oxic.

The second component explains 13.88% of the total geochemical variability. The elements contributing the most to this component are Br, Cl, NO3, Na and to some extent, B and SO4. Br, Cl and NO3 plot close together and define a second pole which we infer to be anthropogenic and/or seawater contamination.

A third pole can be identified on the negative end of both components, which corresponds to an enrichment in Mg, Ca and Sr. In the aquifers of Réunion Island, Ca and Mg are inferred to come from low temperature interaction with basaltic rocks30, which we infer to be the main mineralization process for this third pole. However, an enrichment in magmatic gases can cause the pH to decrease and facilitate water-rock interaction. Therefore, an enrichment in Mg, Ca and Sr could also be a knock-on effect of hydrothermal activity.

Na plots close to the circle and in-between the components 1 and 2, meaning it is contributing to both.

Elements Li, B and SO4 are poorly represented on the two main components resulting from the PCA analysis (Fig. 2A), indicating that they originate from multiple sources. Sulfur species, Li and B are found in magmatic gases at Piton de la Fournaise53,54 that could contribute to a hydrothermal system. Li and B can also come from high temperature interaction with differentiated rocks45,55,56,57. At Piton des Neiges volcano, previous studies show that several thermal springs and rivers streams contain high concentrations of Li, B, and SO4 (up to 137.1 µmol/l, 1545.4 µmol/l and 8.28 mmol/l, respectively) with low δ7Li and δ11B values, which can be associated with hydrothermal activity49,58,59. At Piton de la Fournaise, as expected, Li, B, and SO4 concentrations are several orders of magnitude lower than those of Piton des Neiges thermal springs. However, some samples display higher concentrations in these elements (up to 0.22 µmol/l, 14.92 µmol/l and 0.09 mmol/l, respectively) compared to all samples (median values: 0.02 µmol/l, 0.48 µmol/l and 0.01 mmol/l, respectively), which could correspond to a small hydrothermal contribution. Regarding the Li content of Piton de la Fournaise waters, it should be noted that Henchiri et al.58 analyzed the composition in Li at two locations, the mouth of Rivière Langevin and the mouth of Rivière de l’Est. Samples were taken at the same locations for our study. The two studies show that the Li concentration is low: 0.032 and 0.009 µmol/l, respectively, in Henchiri et al.58, and 0.021 and 0.025 µmol, respectively, in our study (Supplementary Data 1). Therefore, we do not consider these datapoints to be representative of the enriched Li pole with a possible hydrothermal input which we identified. Indeed, their δ7Li values do not indicate a contribution from a hydrothermal source.

The hydrothermal contribution in boron can come from the dissolution of the host rock, of a magmatic gas phase or of fumarolic condensates or sublimates. Little fractionation is expected during these processes60, and therefore, in any case the δ11B value should be close to that of ocean island basalts (OIB). There are no available δ11B values for Réunion basalts, but other OIB unaltered basalts display δ11B values between −12 and −3‰61. Hydrothermal alteration or contamination by subducted slabs can increase these values. At Piton des Neiges volcano, where hydrothermal alteration is evidenced in the host rock62, thermal springs have an isotopic ratio δ11B comprised between 0.2 and 3.9‰ (this study26). At Piton de la Fournaise, where there is less hydrothermally altered rocks62,63,64, the δ11B value of the hydrothermal contribution should be lower, close to the basalt value. Additionally to a possible hydrothermal contribution, in groundwaters and river waters, boron is controlled by water-rock interaction at low temperature, by processes fractionating boron isotopes (mainly the sorption onto mineral and organic surfaces and coprecipitation of boron in secondary minerals, all favoring incorporation of isotope 10B, and leaving the waters enriched in 11B, so with higher δ11B65), but also by the biological productivity, by the influence of sea spray in the infiltrated rainwater and by mixing with seawater. All of these other contributions are expected to result in natural waters displaying a higher δ11B isotopic ratio than that of thermal waters (>12‰65). Therefore, the isotopic ratio δ11B can be used to discriminate the hydrothermal contribution from these other sources of B in groundwaters and river waters66,67,68. Only anthropogenic contamination would possibly bring B with low δ11B signatures, either issued from urban/industrial uses or from fertilizers69,70,71. Measured δ11B values for Piton de la Fournaise waters range from −5.0 to 42.7‰ in agreement with previous values by Louvat et al.58 (from 3.7 to 44.2‰). Nine samples out of the 28 for which δ11B was analyzed display low δ11B values (typically <12‰): TUN1, TUN2, TUN3, RL4, RL1, RE4, RR3, RL5 and RL3. These samples come from preserved natural environments, and do not show anthropogenic contamination according to their position in the principal component analysis plot (Fig. 2). Therefore, these waters are interpreted as being strongly influenced by a hydrothermal system.

CO2 is the most abundant magmatic gas72 and can dissolve in waters. Considering the pH (7.6 ± 0.4) and temperature (16.2 ± 2.6 °C) of the waters of Piton de la Fournaise, this dissolved carbon should be mostly in the form of HCO3. The HCO3 content can thus be an indicator of the presence of magmatic degassing. HCO3 concentrations of Piton de la Fournaise waters range between 0.20 and 4.77 mmol/l. This upper value is in the range of Piton des Neiges thermal waters (3.1 to 42.7 mmol/l49), indicating that there is a probable magmatic source of carbon at Piton de la Fournaise. However, part of the dissolved carbon can also come from a biogenic source. The isotopic ratio δ13C of the total dissolved inorganic carbon (TDIC) can be used to discriminate between a biogenic and a magmatic source of carbon. At Piton de la Fournaise, magmatic gaseous CO2 is expected to have a δ13C value comprised between −0.5 and −8‰73,74,75. Dissolution of this gaseous carbon in waters does not affect the isotopic ratio, but there is an isotopic fractionation if this CO2 is lost through degassing from the water, resulting in higher values of δ13C in the residual waters49,74,76. A biogenic contribution (through the respiration of micro-organisms in the soils and vegetation C use) results in lower values of δ13C in waters (δ13C ≤ −20 in Réunion77). For Piton de la Fournaise waters, we were only able to get δ13C isotopic ratios for 54 out of 82 samples, the remaining samples having too low TDIC (see methods section). δ13C values range between −21.4 and 0.6‰ (Supplementary Data 1). Fourteen samples display δ13C values in the range of that of magmatic CO2 and 3 have a clear biogenic signature. For the others, we infer that the carbon content is a mix between both sources. In Fig. 3, we represented the δ11B and δ13C values for the 25 samples for which both parameters were analyzed. Three samples display δ11B and δ13C in the range of the expected values for B and C originating from magmatic or hydrothermal activity, and several appear to tend toward this pole.

Fig. 3: δ11B and δ13C isotopic ratios of Piton de la Fournaise waters compared to a thermal water sample from Piton des Neiges.
figure 3

The expected ranges for boron and carbon originating from hydrothermal activity are also represented by the red field14,34,49,53,55,56,57.

Full size image

Piton de la Fournaise hydrothermal system revealed

From our geochemical study, we can retain several geochemical markers of hydrothermal activity: the relative enrichments in As, Rb, P, V, Mo, Sb, F, Li, SO4 and B compared to other elements, the concentration in HCO3 and the isotopic ratios δ11B and δ13C. These markers are found in several samples and in different proportions. In this section, we use these markers to compare the hydrothermal contributions in the different samples.

Each of the markers (Y) can be converted to a linear scale from 0 to 1 (index Yi) measuring the relative likelihood of the presence of a hydrothermal component (Table 1). Depending on the elements, additional operations can be made to account for the influence of the dilution of the hydrothermal fluid and of seawater and anthropogenic contaminations.

Table 1 Geochemical markers of hydrothermal activity converted as indexes measuring the likelihood of the presence of a hydrothermal component.
Full size table

The enrichments in As, Rb, P, V, Mo, Sb and F, Li, SO4 and B are obtained by dividing the elemental concentrations by the sum of the major cations, which approximates the total mineral content. This corrects for both dilution of hydrothermal fluids and seawater contamination, as both affect the global geochemical composition of waters. SO4 and B concentrations can be affected also by anthropogenic contamination, which only partially affect the global geochemical composition. Therefore, here, we correct for anthropogenic contamination by also dividing the SO4 and B concentrations by the NO3 concentration (Table 1).

HCO3 concentration can be affected by dilution. However, this compound is inferred to originate from CO2 dissolution, which promotes mineral dissolution and in turn increases the total mineral content. Hence, HCO3 concentration cannot be corrected for dilution.

While elemental concentrations of B and HCO3 give a relative probability of the presence of a hydrothermal component of each water sample compared to the others, isotopic ratios δ11B and δ13C can be considered as absolute linear scales. The maximum likelihood of the presence of a hydrothermal component corresponds to the highest value for δ13C and to the lowest value for δ11B, resulting in an inverse scale (Table 1).

We use these indexes Yi to build a global hydrothermal index (HI = Hydrothermal Index, Eq. (1)). All indexes Yi are normalized between 0 and 1, weighted and summed (Table 1). From the result of the PCA, we know that As, Rb, P, V, Mo, Sb and F are representative of a unique hydrothermal process. Therefore, their 7 indexes Yi are summed, and each is given a weight of 1/7. Other indexes are each given a weight of 1. The hydrothermal index is then computed with the following equation:

$${{{{{{\rm{HI}}}}}}}=\frac{\frac{{{{{{{{\rm{As}}}}}}}}_{i}+{{{{{{{\rm{Rb}}}}}}}}_{i}+{{{{{{\rm{P}}}}}}}_{i}+{{{{{{\rm{V}}}}}}}_{i}+{{{{{{{\rm{Mo}}}}}}}}_{i}+{{{{{{{\rm{Sb}}}}}}}}_{i}+{{{{{{\rm{F}}}}}}}_{i}}{7}+{{{{{{{\rm{SO}}}}}}}}_{4,i}+{{{{{{\rm{B}}}}}}}_{i}+{{{{{{{\rm{Li}}}}}}}}_{i}+{{{{{{{\rm{HCO}}}}}}}}_{3,i}+{{\delta }^{11}{{{{{\rm{B}}}}}}}_{i}+{{\delta }^{13}{{{{{\rm{C}}}}}}}_{i}}{n}$$
(1)

where n corresponds to the number of terms of the numerator. δ11B and/or δ13C were not analyzed for all samples. As a result, n equals at least 5 and at most 7. We then normalized the hydrothermal index between 0 and 1, giving the number HInorm.

The spatial distribution of the values of the normalized hydrothermal index HInorm in Piton de la Fournaise waters is represented in Fig. 4, along with major geological features that can constitute preferential pathways for hydrothermal fluids or heat sources: caldera faults78, rift zones79 dense bodies and conductive layers9. The waters most likely to have a hydrothermal component (HInorm > 0.5) are all located on or close to the caldera faults surrounding the Plaine des Sables region, except for four of them (wells SPH2, SPH3, SPH4 and spring RL15). However, these four samples drain an aquifer which can be considered continuous from the Summit/Plaine des Sables region down to the shore30,80, meaning that their geochemical composition could be inherited from upstream. This spatial distribution indicates that the caldera faults act as preferential pathways for rising hydrothermal fluids. In La Plaine des Sables region, located at the center of this caldera fault system, a dense body and a conductive layer have already been identified as a possible geothermal resource (Fig. 4). Our study provides additional evidence in favor of this interpretation. Therefore, la Plaine des Sables can be considered the best target for geothermal exploration in terms of geothermal potential. This region has exceptional landscapes and biodiversity, and as such is part of a national park and is a UNESCO World heritage site, protected by strict environmental regulations, which any geothermal project will need to account for.

Fig. 4: Spatial distribution of the normalized hydrothermal index, indicating the relative probability of the presence of a hydrothermal component in the waters of Piton de la Fournaise.
figure 4

Geological, structural and geophysical features are also represented: caldera faults78, rift zones79 dense bodies and conductive layers24,28,29.

Full size image

Our study demonstrates that the geochemistry of cold waters is an efficient way to address the first stage of geothermal exploration in blind systems, which is to determine whether or not there is hydrothermal activity generating energy at depth, and it proves to be complementary to other approaches. Regardless of the extent of the contribution of hydrothermal activity to the geochemical composition of waters, as long as it exists, the study of relevant elemental and isotopic ratios can highlight it. Spatializing this diffuse hydrothermal signal points to geographical and structural targets for geothermal exploration.

Methods

The groundwaters and surface waters of Piton de la Fournaise were sampled during the dry season in 2020 and 2021, when the water tables are at a low level, in an effort to minimize mixing with newly infiltrated rainfall. These campaigns yielded a total of 82 water samples (from 80 sampling points) to be analyzed for major and trace elements concentrations and δ13C and δ11B isotopic ratios. The pH, temperature, and electrical conductivity (EC, expressed at 25 °C) of each sample were measured in situ. All water samples were filtered at 0.2 µm on site and the containers were rinsed three times with filtered water. Samples for the analysis of major, trace elements and B isotopes were collected in LDPE bottles. Samples for the analysis of major cations and trace elements were acidified with nitric acid (HNO3 0.6 N) on site. Samples for the analysis of total dissolved inorganic carbon content (TDIC) and C isotopes were collected in Labco 12 ml Exetainer vials with no headspace. All samples were stored at a temperature of 6 °C before shipment.

All analyses were performed at the Institut de Physique du Globe de Paris (IPGP). Major anions, fluoride and nitrate concentrations were determined by ion exchange chromatography. Major cations and trace element concentrations were determined using inductively coupled plasma mass spectrometry (ICP-MS). The accuracy for major cations and anions concentrations is <5%, and <20% for trace elements, as inferred from replicate measurement of solutions of known compositions. The TDIC and carbon isotope ratios of the TDIC were determined together by gas chromatography and isotope ratio mass spectrometry (GC-IRMS), after releasing the TDIC as CO2 by H3PO4 acidification in a vial from an aliquot of 5 ml81. For the analysis with the GC-IRMS, a finite amount of gas is pumped inside the vial and sent into the analysis circuit. This step is repeated several times and the TDIC and δ13C are computed from the results of several runs. This method proved not adapted for the analysis of our samples due to a too low content of TDIC. With an aliquot of just 5 ml, the available gaseous CO2 in the vial sampled by the GC-IRMS was too low and the analysis did not yield results for several runs and for several samples. We obtained reliable results for only 51 out of 82 samples. For these samples, HCO3 concentrations were computed from TDIC, pH and temperature, using the temperature dependent acid dissociation constants from Mook and Koene, (1975)82. For the other samples, HCO3 concentrations were estimated for an ionic balance equal to 0%. δ13C are reported relative to PDB standard with a 1 SD reproducibility, ±0.05‰. Boron isotope ratios were determined by MC-ICP-MS with a direct injection nebulizer (d-DIHEN83), after boron extraction through ion chromatography84. δ11B are reported relative to NIST SRM951 standard with a 2 SD reproducibility, between 0.1 and 0.4‰.