The Total Dissolved Solids (TDS) fluctuates from 10 up to 166 mg/L in the groundwater from the extensive sandy-argillaceous terrane and from the alluvia (hereafter referred to as GWA/alt-sed) and from 46 to 218 mg/L in the groundwater from the basement (hereafter referred to as GWA/bas). The water chemistry shows a large variation in major element contents (Table 2). Chloride content in the GWA/alt-sed groundwater ranges from 50 to 4241 μmol/L and from 45 to 457 μmol/L in the GWA/bas groundwater. Though Cl⁻ ions do not have any significant lithological origin, since French Guiana is evaporite free, Cl⁻ ions in groundwater may originate mainly from rainfall recharge (sea salts, Gaillardet et al., 1997), and to a lesser extent from human activity (domestic sewage, fertilisers, etc). Chloride often behaves conservatively through the hydrological cycle and is used as an atmospheric-input reference element in many hydrosystems (Gaillardet et al., 1997). Like Cl⁻, Na⁺ shows a wide range of contents and, when compared to Cl⁻ in Fig. 2a, Na concentrations in most surface and ground waters plot above the seawater dilution line (SWDL), indicating Na⁺ excess. Most of the surface and ground waters from the GWA/alt-sed plot between the SWDL and the line have a molar Cl/Na ratio of around 0.5. Three GWA/bas ground waters (CR1, CR2 and F1-oct-98) plot with a molar Cl/Na ratio of 0.7, the hydrogeological data suggest that the alterite compartment contributes significantly to the well flow rate. The remaining GWA/bas ground waters plot with a molar Cl/Na ratio of 0.2. This divergence from the seawater dilution line reflects a large Na enrichment that is mainly related to water-rock interaction. Ca²⁺ concentrations show a large range in groundwater from about 5–10 to 999 μmol/L and fluctuate also largely in rain and surface waters when compared to Cl⁻ in Fig. 2b (all samples plot above the SWDL, with a Cl/Ca ratio of 1.42 for the surface waters indicative of a Ca²⁺ excess). The groundwater from the GWA/alt-sed also display a Ca²⁺ excess and plot with a molar Cl/Ca ratio ranging between that of SWDL and the ratio 1.42, four of them plot with a molar Cl/Ca ratio higher than 1.42. The GWA/bas groundwater plot with a molar Cl/Ca ratio ranging between 1.42 and 0.05, reflecting the larger Ca²⁺ excess. Some of them plot with a molar Cl/Ca ratio of around 1.42, close to the groundwater from the GWA/alt-sed, reflecting the fact that water partly originates from the alterite compartment. Similarly, K⁺, Mg²⁺ and Sr²⁺ excess are observed when plotted versus Cl⁻, (not shown), reflecting the weathering of aluminosilicates.

The positive correlation between HCO₃ and the sum of cations in groundwater samples (Σ⁺, r² = 0.55, n = 38) clearly indicates that the cations released by weathering are balanced by the alkalinity, in good agreement with observations in other South American basins influenced by similar climatic conditions (Freyssinet and Farah, 2000; Gaillardet et al., 1997). The bicarbonate increase during weathering originates from the atmospheric/soil CO₂, which implies a positive correlation between HCO₃ and pH in groundwater samples (r² = 0.66, n = 38) related to the water-rock interaction.

Variations in the stable-isotope composition in a catchment's water balance are mainly caused by natural variations in the isotopic composition of rainfall and through mixing with pre-existing water and the influence of evaporation (Kendall and McDonnell, 1998). The stable isotopes δ²H and δ¹⁸O of all rain samples collected in French Guiana (Négrel et al., 1997) define the local meteoric water line (LMWL), as illustrated in Fig. 3a. The δ¹⁸O and δ²H in GWA/alt-sed ground waters fall in the range −2.1 to −3.7‰ for δ¹⁸O and −7.9 to −19.2‰ for δ²H and in the range −2.9 to −4.0‰ for δ¹⁸O and −12.3 to −18.4‰ for δ²H for the GWA/bas ground waters, in close agreement with the range of surface water (Négrel and Lachassagne, 2000).

The δ¹⁸O and δ²H relationships are illustrated in Figs. 3a and b for rainwater (regression line and 95% confidence range), global meteoric water line (GMWL; δ²H = 8 δ¹⁸O + 10, Craig, 1961), surface waters and groundwater. Surface waters plot close to the LMWL in Fig. 3b and, as demonstrated by Négrel and Lachassagne (2000), most show a significant shift (close to 1.4‰) to the right of this line as result of evaporation processes. Most of the ground waters from GWA/alt-sed clearly plot close to the local and global lines. Some of these ground waters are displaced below the LMWL. This could be due either to evaporation prior or during infiltration or, as suggested by Boronina et al. (2005), because of partial evaporation from soils and dilution by subsequent recharge. GWA/bas groundwater also plot close to the LMWL and GMWL, reflecting a meteoric origin and a lack of significant evaporation during recharge or oxygen isotope exchanges between water and the rock matrix.

Infiltrating rainwater shows a mean weighted signatures for δ¹⁸O and δ²H of −4.51 and −24.23‰ respectively, corresponding to 1790.4 mm of rainfall (Négrel et al., 1997). However, mean rainwater δ-values cannot explain groundwater recharge as less negative values occur in many samples (Fig. 3b). Thus, recharge from precipitation events occurs in significant amount at least during two periods. The first period was September to January and corresponds to the rainy season with 496 mm of rainfall (27.7% of the total annual rainfall, mean weighted rain signatures input led to δ¹⁸O and δ²H of around −2.45 and −7.58‰, respectively). The second period occurred in March-April with 249 mm rainfall, which corresponds to the middle of the rainy season (13.9% of the total annual rainfall, mean weighted rain input led to δ¹⁸O and δ²H signatures of around −3.10 and −11.11‰, respectively). These two rainy periods should constitute the recharge period that would explain the least negative δ-values in the ground waters. It is worth noting that part of the δ¹⁸O and δ²H signatures in the groundwater should reflect the recharge occurring during May to July. During this period, the rainfall was 943.3 mm, (52.7% of the total annual rainfall) and the mean weighted rain input led to δ¹⁸O and δ²H signatures of around −4.24 and −20.89‰, respectively. This period of rain and thus recharge may explain the most negative δ-values in the groundwater samples. This suggests different stages of groundwater recharge with isotope signatures in close connection with the range in rainwater.

Sr isotope studies of rivers and lakes have shown that variations in ⁸⁷Sr/⁸⁶Sr and Sr contents are caused by mixing waters with different ⁸⁷Sr/⁸⁶Sr ratios and Sr contents, each of them reflecting water-rock interaction with different rock types (Chung et al., 2009; Grove et al., 2003; Oliver et al., 2003).

All the surface- and groundwater samples from French Guiana are plotted in a ⁸⁷Sr/⁸⁶Sr versus 1/Sr diagram (Fig. 4a), which is classically used to evaluate two-component mixing and end-member water compositions. This figure indicates the existence of at least three end-members. The surface waters from the Maroni catchment, reported from Négrel and Lachassagne, 2000, plot along the mixing trend between the end-members corresponding to the drainage of the Unit P metavolcanic rocks (low ⁸⁷Sr/⁸⁶Sr ratio, highest [Sr]) and that of the Unit S meta-sedimentary lithology (highest ⁸⁷Sr/⁸⁶Sr ratio, intermediate [Sr]). The low ⁸⁷Sr/⁸⁶Sr ratio would reflect the weathering of rocks such as basalt and amphibolite that are known to impart a low ⁸⁷Sr/⁸⁶Sr ratio to the waters (Dessert et al., 2001; Louvat and Allègre, 1997) while the high ⁸⁷Sr/⁸⁶Sr ratio would be related to the weathering of schists and micaschists, which deliver higher ⁸⁷Sr/⁸⁶Sr ratios to the water (Aubert et al., 2002). This compares to the results obtained on rivers draining “undifferentiated Proterozoic rocks” of the Guyana Shield (Edmond et al., 1995), which also lie on the mixing trend between the S and P end-members. The third end-member (intermediate ⁸⁷Sr/⁸⁶Sr ratio, low [Sr]) could correspond to the drainage of plutonic granitoid intrusions (δη). In the Maroni catchment, the shift of some surface waters to the right of the mixing trend between S and P end-members could reflect the input into the main stream of tributaries draining weathered granitoids. This is also illustrated by some rivers from the GWA/alt-sed (Négrel et al., 2002) that plot in the field δη in the Fig. 4a. Some ground waters from the same area plot very close to the mixing trend between S and P end-members but most of them plot with a low ⁸⁷Sr/⁸⁶Sr ratio, reflecting the influence of the drainage of volcanic rocks of the Lower Paramaca (Unit P). It is worth noting that some points of groundwater are shifted with a higher 1/Sr ratio agreeing with the rainwater samples. Therefore, it may be concluded that the shift reflects a large rainwater input for the Sr budget compared to the Sr released by water-rock interaction.

The GWA/bas groundwaters lie on the straight line corresponding to a mixing between waters that have interacted with rocks from Unit P and S. Most ground waters display the lowest ⁸⁷Sr/⁸⁶Sr ratio, reflecting the interaction with rocks having a low Sr isotopic ratio. However, the location of the deep groundwaters from Grand Santi (F1, F2) is not consistent with an interaction with volcanic rocks of the Unit P because the boreholes were drilled within granitoids. One way to explain the similarity between the drainage of volcanic rocks (Unit P) and ground waters that have interacted with granitoids is to consider that the latter are influenced by the weathering of low Rb-rich Sr phases, which would impart a low ⁸⁷Sr/⁸⁶Sr ratio to the waters. Weathering of low Rb-rich Sr phases and more constraints on the ⁸⁷Sr/⁸⁶Sr ratio variations may be tested through a diagram ⁸⁷Sr/⁸⁶Sr versus Ca/Na ratio (Fig. 4b). The use of a cation ratio rather than absolute concentrations alone, avoids variations due to dilution or concentration effects (BenOthmann et al., 1997; Chung et al., 2009). As previously detailed, when Ca is compared to Cl content (Fig. 2b), all groundwaters are indicative of a Ca²⁺ excess, more marked in the ground waters from the basement, as well as for Mg, while the Na enrichment is less. Thus, Ca/Na appears to be a good indicator of weathering processes in waters from French Guiana. The surface water from the Maroni catchment does not define a clear hyperbola that binary mixing should lead to (Langmuir et al., 1978). They appear rather to be scattered along a main trend showing a large fluctuation in the ⁸⁷Sr/⁸⁶Sr ratio with a weak Ca/Na range. This makes possible to define a first water-rock interaction process that corresponds to surficial weathering of unit P and S. Surficial weathering of Unit S shows a larger ⁸⁷Sr/⁸⁶Sr ratio without changes in the Ca/Na ratio. Rainwater field (RW in Fig. 4) shows a low Ca/Na ratio and a ⁸⁷Sr/⁸⁶Sr ratio around 0.710. Two other trends can be defined with a weak fluctuation in the ⁸⁷Sr/⁸⁶Sr ratio within a change of the Ca/Na ratio. The first trend corresponds to a water-rock interaction process from the surficial to the deep weathering of granitoids. The deeper weathering of granitoids reflected in this trend implies that the weathering of Ca-bearing phases (e.g. plagioclases and/or calcite, Pett-Ridge et al., 2009) is greater in sample MM4 than in sample CRM2. The second trend corresponds to a water-rock interaction process from the surficial to the deep weathering of the unit P with a weak fluctuation of the ⁸⁷Sr/⁸⁶Sr ratio (around 0.704) associated with a large increase in the Ca/Na ratio. As for the first water-rock interaction processes, the shift towards a relatively high Ca/Na ratios can be explained by more extensive weathering of Ca-bearing phases with a low ⁸⁷Sr/⁸⁶Sr ratio, such as plagioclase and/or calcite. High Ca/Na ratios generally relate to carbonate weathering (Gaillardet et al., 1997), which is also accompanied by increase in the Mg/Sr and Ca/Sr ratios (Rengarajan et al., 2009). The increase in the Ca/Na ratios during the weathering of the unit P is accompanied by an increase of the Ca/Sr ratios (up to 500) not withstanding that this value is the lowermost value of carbonate weathering (Meybeck, 1986). On the other hand, large Ca/Na ratios were observed during basalt weathering (up to 3, Dessert et al., 2001; Raiber et al., 2009; Rengarajan et al., 2009). Such large Ca/Na ratios were accompanied by Ca/Sr ratio around 500 (Dessert et al., 2001). Thus, we can conclude that the trend in the GWA/bas ground waters from the unit P may reflect a more intense weathering of Ca-bearing phases like plagioclases as no evidence of calcite has been reported in basalts.

Using the ⁸⁷Sr/⁸⁶Sr and Ca/Na ratios of the extreme samples of the surficial and deep weathering of the unit P, which can be considered as end-members, a mixing line was calculated and reported in Fig. 4b. The first end-member corresponds to the surficial weathering with a ⁸⁷Sr/⁸⁶Sr and Ca/Na ratios of around 0.703 and 0.4. The latter represents the value classically used for the weathering of silicate rocks (Gaillardet et al., 1997; Louvat and Allègre, 1997; Rengarajan et al., 2009). The second end-member shows a slightly higher ⁸⁷Sr/⁸⁶Sr ratio (i.e., 0.705) and a greater Ca/Na ratio (i.e., 5). The calculation led to a divergence from the surficial weathering end-member of around 10% for F5, 25% for F2 and M4 and 50–60% for L1bis and L2.

Since the development of the method by ICP-MS to determine the dissolved REE concentrations (Johannesson and Lyons, 1995; Stetzenbach et al., 1994), many studies deal with the behaviour of the REE in continental waters (Andersson et al., 2001; Gaillardet et al., 1997; Janssen and Verweij, 2003; Viers et al., 2000 and references therein) but, in contrast, neodymium isotopes have not been extensively used in hydrogeological studies. Some data are available on the compositions of river water (Andersson et al., 2001; Goldstein and Jacobsen, 1987; Steinmann and Stille, 2006; Tricca et al., 1999) and very little information is available for saline waters (Négrel, 2006) and ground waters (Tricca et al., 1999; Viers and Wasserburg, 2004). The contents of Sm and Nd, as well as for the other dissolved REEs vary greatly. The dissolved Nd content in the surface waters varies from 20 to 102 ng/L with a low TDS (10–20 mg/L), in agreement with values reported by other studies (Amazon: Gaillardet et al., 1997; Cameroon: Viers et al., 2000; Viers and Wasserburg, 2004), but broadly independent of other parameters such as total dissolved solids and pH. In the groundwater from the GWA/alt-sed, the dissolved Nd content varies from 39 to 752 ng/L together with a broad range in the TDS (10 up to 166 mg/L). In the GWA/bas groundwater, the dissolved Nd contents are lower and vary from 3 to 63 ng/L with TDS values ranging from 46 to 218 mg/L. These ranges agree with other studies in groundwater (Oliva et al., 1999; Viers and Wasserburg, 2004). As usually stated in terrestrial surface processes and in granite-gneiss rocks, Sm displays similar behaviour to Nd (Nd = 4.9 × Sm + 6.5, r² = 0.98, n = 30), because of the similarity in their physical and chemical properties (Gaillardet et al., 1997; Goldstein and Jacobsen, 1987; Viers and Wasserburg, 2004). The isotopic composition of dissolved Nd in surface and ground waters from French Guiana ranges from ¹⁴³Nd/¹⁴⁴Nd = 0.511377 ± 9 × 10⁻⁶ to 0.512162 ± 8 × 10⁻⁶ corresponding to a range of ɛNd(0) from −9.2 to −24.6. The lowest ɛNd(0) are observed in the GWA/alt-sed groundwater whereas the highest are for the GWA/bas groundwater. The small differences between ɛNd(0) values for both the dissolved and suspended phases found within the same rivers suggests that the Nd isotopic composition can be used as an indicator of the weathered parent rock (Goldstein and Jacobsen, 1987, 1988; Tricca et al., 1999). Mixing processes and water–rock interaction have both contributed to the observed ɛNd(0) (Négrel et al., 2001).

The similarity between the ¹⁴³Nd/¹⁴⁴Nd ratios in waters and related bedrock can be postulated now according to the recent works by Aubert et al. (2001). Comparing the Nd and Sr isotopic composition in minerals and waters, they demonstrated that the Sr and Nd isotopic characteristics of the major mineral phases of a granite clearly shows the strong influence of plagioclase and phosphate minerals (e.g. apatite) on the isotopic composition of the spring and stream waters. This was confirmed on other granite massif (France) by Négrel (2006). Thus, comparison of the ɛNd(0) values with the corresponding Sm/Nd ratios (e.g. ¹⁴⁷Sm/¹⁴⁴Nd ratios represented with the propagate errors) yields more information about the origin of the REE in the waters from French Guiana (Fig. 5). With the exception of the samples GUY99-11 and F1, they are positively correlated (ɛNd(0) = −43.3 + 233.8 × ¹⁴⁷Sm/¹⁴⁴Nd; r² = 0.64; n = 16) with low ¹⁴⁷Sm/¹⁴⁴Nd ratios for the river water and rather high ratios for the ground waters.

Most of the ground waters have ¹⁴⁷Sm/¹⁴⁴Nd ratios significantly higher than the average continental crust value of 0.105 (Allègre and Lewin, 1989). The ɛNd(0) versus the ¹⁴⁷Sm/¹⁴⁴Nd ratio in the surface water and GWA/bas groundwater, illustrated in Fig. 5, are consistent with that of the parent rocks (Delor et al., 2001; Gruau et al., 1985). If the ɛNd(0) is mostly controlled by the weathering of plagioclase and phosphate phases, as suggested by Aubert et al. (2001) and Négrel (2006), they should be shifted towards higher ɛNd(0) and Sm/Nd ratios and thus would plot on the right of the parent rocks field, which is not observed. The ɛNd(0) versus the ¹⁴⁷Sm/¹⁴⁴Nd ratio for the groundwater from the GWA/alt-sed plot between the values measured in the parent rocks and that of suspended matter from the Amazon Basin (Allègre et al., 1996; Goldstein et al., 1984). This suggests for some groundwater a possible influence of sedimentary deposits in the coastal area that originate from the Amazon. The other groundwater from the same area agrees with the field of parent rocks, suggesting that Nd originates from the weathering of the bedrock.