Introduction

Combining ocean color remote sensing observations (e.g., Lorthiois et al. 2012; Ody et al. 2016) with in situ sampling and laboratory analyses helps to characterize, chemically and mineralogically, suspended particles in river plumes, in particular the nature and origin of particulate organic carbon (POC). Many former studies showed correlation between organic carbon (OC) in estuaries and environmental factors, i.e., anthropic pressure or rainfall (e.g., Abril et al. 2002), and the anthropogenic impacts on blue carbon, i.e., carbon captured by marine living organisms (Kuwae et al. 2016). For this purpose, protocols have been recently reviewed by the ocean color community for sample collection, filtration, storage, analysis, and quality assurance in the scope of POC measurements appropriate for satellite algorithm development and validation (Chaves et al. 2019).

There is growing evidence of a link between climate change and the increase in extreme meteorological events such as floods and droughts (IPCC 2012). The northwestern Mediterranean Sea is known to be significantly impacted by climate change (Metzger et al. 2005; Kim et al. 2019) and is often subject to extreme precipitation events and subsequent floods (Colmet-Daage et al. 2018). It has been shown that a shift to more extreme precipitation events (more heavy rainfall and longer dry intervals) alters estuarine salinity and temperature, carries nutrients, organic matter, and contaminants, and alters the structure of phytoplanktonic and coastal microbial communities (e.g., Pollack et al. 2011; Wetz and Yoskowitz 2013; Zoppini et al. 2019). Besides, the discharge of riverine suspended particles to coastal waters often results in the alteration of their chemical composition. Characterization of carbon riverine inputs in estuaries is well documented (e.g., Cathalot et al. 2013), but not under peak flooding conditions, a less documented research field.

Organic carbon analyses are not simple, notably because organic phases are often intimately associated with inorganic phases (King et al. 1998). Carbon is commonly considered in two major forms, organic (OC) and inorganic (IC), having distinct isotopic signatures, e.g., δ13Corg = − 25 to − 30‰ for POC of Rhône River (Harmelin-Vivien et al. 2010), δ13Cinorg = 0 to 2‰ for dissolved IC in tropical waters (Cotovicz et al. 2019), and − 5 to 0‰ in Mediterranean waters close to the Rhône River mouth (Pozzato et al. 2018). The removal of IC forms is generally carried out by acid treatment (“decarbonation”), usually with hydrochloric acid (HCl), prior to POC analysis, assuming that most of this carbon is contained into calcium carbonate phases. Acidification is carried out either in direct aqueous solution or in vaporous mode (fumigation) (Komada et al. 2008). However, leaching of soluble OC may occur (Froelich 1980), as well as the release of non-carbonate inorganic matter, which may lead to an overestimation of the OC content if not properly dried (Hedges and Stern 1984). Several methods have been developed to overcome these potential biases (Nieuwenhuize et al. 1994). Inefficient carbonate removal has been reported for diluted acid solutions (0.1N), and potential alteration of OC has been reported for concentrated solutions (6N). Many protocols use diluted HCl 1N or 2N (e.g., Darnaude et al. 2004) in aqueous phase. Fumigation is often recommended (Liénart et al. 2017), but the use of more concentrated acid may induce partial loss of OC if the sample is overexposed (Komada et al. 2008). There is no consensus on the most appropriate method to use, depending of the samples studied (Brodie et al. 2011).

This work proposes to use stable carbon isotope analysis to verify the completeness of the decarbonation process for suspended particles in river plumes. Three rivers are considered owing to the different mineralogy and chemical matrix of their particulate discharge. The results obtained on natural samples collected in river plumes show a significant heterogeneity of the isotopic signatures of OC (δ13Corg) among samples, without being able to say whether this results from a real chemical heterogeneity (most likely due to different terrigenous loads) or from an incomplete sample decarbonation. This question is the purpose of the present note.

Material and methods

Study sites and sampling

Mediterranean rivers experience annual peak floods during the winter and spring seasons. The three study areas selected here are the Rhône, Var, and Paillon Rivers (south-east Mediterranean coast of France) (Fig. 1). The Rhône is a major river of the northwestern Mediterranean Sea while the Var and Paillon are typical peak flood rivers, with average freshwater discharges of 1690, 50, and 1 to 40 m3s−1, respectively, and watershed surface areas of 95,590, 2800, and 236 km2.

Fig. 1
figure 1

a Location of the three river mouths and plumes (northwestern Mediterranean Sea) selected as study areas. Locations of sampling stations in the b Rhône, c Var, and Paillon River plumes and d typical water samples collected at the mouths of the Var “V” and Paillon “P” Rivers

Full size image

Water samples were collected during flooding conditions in the Rhône (February 2016, 51 stations), Var (24 stations, November 2016), and Paillon (18 stations, November 2016) plumes. Surface waters (0–1 m depth, i.e., within the surface layer viewed by satellite ocean color sensors) were collected using horizontal 8L-Niskin bottles, from the river mouth to the offshore edge of the turbid plume. Each bottle was split in three sub-samples to triplicate measurements. In addition, 20L-samples were collected a few meters upstream the mouths of the Var and Paillon Rivers to sample large amounts of suspended particles for mineralogical determination, to be compared to results from the Rhône River (Pont et al. 2017).

Sample preparation

Known volumes of water samples (V = 0.2 to 0.6 L, depending on water turbidity) were filtered through pre-combusted (1 h at 450 °C) and pre-weighed (mass M0, in mg) 25 mm glass-fiber filters (Whatman, GF/F). The three sub-samples were filtered, resulting in three filters (triplicates) for each sampling point. Filters were then dried for 24 h at 60 °C, prior to weighing under dry atmosphere (mass M1, in mg) to obtain the suspended particulate matter concentration (SPM, in mg/L) as SPM = (M1 − M0)/V. The uncertainty associated to SPM measurements was assumed to be the standard deviation on triplicates (< 5%). The first data set for isotopic analysis was carried out after removal of carbonates by adding 20 μL HCl 2N several times, until the effervescence caused by acidification was no longer observable, according to a widely used standard protocol (King et al. 1998), here called “protocol #1.” On many filters, this effervescence was unusually low and disappeared after the second addition of acid. A decarbonation problem was suspected, and, in order to address this problem, analyses on sub-samples from the same sampling campaign were carried out using an alternative protocol characterized by more intensive acid leaching (200 to 350 μL of HCl 2N on each filter, according to the sample mass) and hereinafter referred to as “protocol #2.” This amount of acid, able to achieve decarbonation of samples weighing more than 20 mg, is in large excess relative to the limit value measured for certified marine sediments (6 μL/mg, see below).

Analyses

Particulate fractions collected on GF/F filters were analyzed by isotope ratio mass spectrometry (IRMS). Samples collected upstream the river mouths underwent mineralogical analyses.

Carbon stable isotope analysis (δ13C)

Filters were analyzed for both carbon isotopes and concentrations using an Elemental Analyzer (Elementar VarioPyroCube) coupled to an isotope ratio mass spectrometer (Isoprime 100). All isotopic data are expressed in conventional notation (Eq. 1), as the isotopic ratio normalized to the reference gas:

$$ {\updelta}_{\mathrm{sample}}=\left[\left({R}_{\mathrm{sample}}/{R}_{\mathrm{ref}}\right)-1\right] $$
(1)

where δsample is expressed in ‰, and Rsample and Rref are the isotopic ratios R = 13C/12C of the sample and the reference gas, respectively.

Certified standard IAEA-600 caffeine was used in each batch for isotopic data monopoint correction (Paul et al. 2007), according to Eq. 2:

$$ {\updelta}^{13}{{\mathrm{C}}^{\mathrm{c}}}_{\mathrm{sample}}=\left({\updelta}^{13}{{\mathrm{C}}^{\mathrm{m}}}_{\mathrm{sample}}+1000\right)\times \left({\updelta}^{13}{{\mathrm{C}}^{\mathrm{t}}}_{\mathrm{standard}}+1000\right)/\left({\updelta}^{13}{{\mathrm{C}}^{\mathrm{m}}}_{\mathrm{standard}}+1000\right)-1000 $$
(2)
$$ {\updelta}^{13}{{\mathrm{C}}^{\mathrm{c}}}_{\mathrm{sample}}:\mathrm{corrected}\ {\updelta}^{13}\mathrm{C}\ \mathrm{value}\ \mathrm{for}\ \mathrm{sample},\mathrm{in}{\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}} $$
$$ {\updelta}^{13}{{\mathrm{C}}^{\mathrm{m}}}_{\mathrm{sample}}:\mathrm{measured}\ {\updelta}^{13}\mathrm{C}\ \mathrm{value}\ \mathrm{for}\ \mathrm{sample},\mathrm{in}{\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}} $$
$$ {\updelta}^{13}{{\mathrm{C}}^{\mathrm{t}}}_{\mathrm{standard}}:\mathrm{certified}\ {\updelta}^{13}\mathrm{C}\ \mathrm{value}\ \mathrm{for}\ \mathrm{standard},\mathrm{in}{\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}} $$
$$ {\updelta}^{13}{{\mathrm{C}}^{\mathrm{m}}}_{\mathrm{standard}}:\mathrm{measured}\ {\updelta}^{13}\mathrm{C}\ \mathrm{value}\ \mathrm{for}\ \mathrm{standard},\mathrm{in}{\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}} $$

Each run of 30 samples started with two measurements of isotopically certified caffeine (IAEA-600, δ13Ctstandard = −27.771‰, SD = 0.043) and ended with two other measurements of the same certified standard. Bracketing every six samples in each batch was carried out using acetanilide (Merck KGaA), widely used to calibrate the response of the elemental analyzer and used in the laboratory as a secondary IRMS standard, corrected by caffeine δ13C measurements. Measurements of δ13C in acetanilide showed standard deviation for all batches lower than 0.1‰, and the average measured carbon mass content was 71.16%, SD = 1.16% (the theoretical value for pure compound is 71.09%). The standard deviation was higher for %C than for C/N, because of the uncertainty on the small sample weight that impacted %C calculation.

Mineralogy

After centrifuging, particle samples from bottles collected upstream the river mouths were dry-weighed, and then smoothly reduced to powder in an agate mortar. Analyses were carried out with a Siemens D5000 X-ray diffractometer (Bruker) mounted in Bragg-Brentano ϑ/ϑ geometry with a 3 to 65° scanning by steps of 0.04° during 10 s. The diffractometer was equipped with an X-ray tube fitted with a copper anticathode with a 40 kV accelerating voltage and 30 mA current. Analyses were conducted on bulk samples (including all minerals) and on separated clay samples. Results provided semi-quantitative proportions of minerals.

Results

Reference materials

The acidification protocol was first tested on reference samples. Certified marine sediment samples (MESS-2, from NRCC, Canada, %Ctotal = 2.14% ± 0.03) were analyzed without acidification, giving an averaged measured %Ctotal value of 2.17%, with a standard deviation (SD) of 0.07%, in accordance with the certified value. The average isotopic measurement for total carbon, i.e., without acidification, was δ13Ctotal_MESS2 = −18.25‰, SD = 0.06‰.

A new set of MESS-2 samples acidified with various HCl 2N volumes, covering the ranges of both protocols in microliters per milligram of sample, and sample masses were analyzed. Variability on δ13C measurements was significant for added 2N HCL volumes lower than 6 μL per mg of sample mass (Fig. 2), an amount similar to the volume proposed by Lukasewycz and Burkhard (2005).

Fig. 2
figure 2

Isotopic ratios of MESS-2 organic carbon as a function of the added volume of 2N HCL per mg of sample: δ13C (‰) = f(volume 2N HCl per mg (μL/mg)). The size of points represents the measurement reliability (0.1‰)

Full size image

This variation of δ13C depending on the amount of acid is typical of a two end-member mixing model (Loftis and Meile 2014):

$$ {\updelta}^{13}{\mathrm{C}}_{\mathrm{mix}}={\mathrm{x}}_{\mathrm{C}\mathrm{org}}\times {\updelta}^{13}{\mathrm{C}}_{\mathrm{org}}+{\mathrm{x}}_{\mathrm{C}\mathrm{inorg}}\times {\updelta}^{13}{\mathrm{C}}_{\mathrm{inorg}} $$
(3)

with 1 = Σxi, xCorg, and xCinorg are the organic and inorganic fractions of carbon, respectively.

For MESS-2 samples, according to Eq. 3:

$$ {\updelta}^{13}{\mathrm{C}}_{\mathrm{mix}}={\updelta}^{13}{\mathrm{C}}_{\mathrm{total}}=-18.25{\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}}, \mathrm{SD}=0.06{\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}}, \mathrm{and}\ {\updelta}^{13}{\mathrm{C}}_{\mathrm{org}}=-25.91{\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}}, \mathrm{SD}=0.13{\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}}, \mathrm{with}\ {\mathrm{x}}_{\mathrm{C}\mathrm{org}}=0.57 $$

Field samples

Particles from samples collected upstream the river mouths were mineralogically analyzed (Table 1). Particles in the Var River can contain more than twice the amount of clays compared to particles from the Paillon River.

Table 1 Relative abundances (%) of clays, quartz, feldspars, and carbonates (calcite and dolomite) in particles in suspension in the Paillon and Var Rivers based on X-ray diffraction analysis. Suspended particles in the Rhône River contain significant amount of clays (Pont et al. 2017)
Full size table

IRMS analyses were performed on samples collected in the river plumes. Samples acidified with the two protocols were grouped and randomly ordered for daily analysis subsets. Figure 3a displays the OC isotope ratios (δ13Corg) measured as a function of the SPM concentration for samples acidified with protocol #1.

Fig. 3
figure 3

a Carbon isotope ratios (δ13Corg) obtained using protocol #1 (standard decarbonation) as a function of SPM concentration, for the three sampled river plumes. b δ13Corg ratios obtained using protocol #1 and protocol #2, for the Rhône and Paillon Rivers, as a function of SPM concentration. The size of points represents the measurement reliability (0.5‰)

Full size image

A strong heterogeneity can be observed (Fig. 3a), with δ13Corg values decreasing seaward (towards low SPM), in particular in the Rhône (from − 16.1 to − 27.7‰) and Var (− 5 to −26‰) plumes. This suggests the presence of strong mineral matrices in river loads and the occurrence of terrigenous inputs, which are both likely to make more difficult the total removal of carbonates using the applied protocol. Significant loads of clays might reduce the efficiency of the standard decarbonation protocol and, therefore, yield shifted isotopic measurements, i.e., less negative values, if δ13Corg is different from δ13Cinorg (see Eq. 3).

The two decarbonation protocols were then applied to suspended particles from the Rhône River (clay-rich matrix) and Paillon River (negligible matrix effect) (Fig. 3b) plumes. For the Rhône, samples acidified with protocol #1 yielded heterogeneous results with less negative δ13Corg values (from − 16.1 to − 27.7‰) exhibit more homogenous isotopic ratios when acidified with protocol #2 (from − 26.9 to − 27.7‰). Moreover, the samples acidified with protocol #1, yielding the most negative δ13Corg values (generally those with a low SPM value), exhibit fairly similar isotopic ratios when acidified according to protocol #2: protocol #1 is clearly less efficient for samples close to the river mouths (i.e., with high SPM concentrations).

Discussion

During flood events, Mediterranean rivers discharge large amounts of mineral-rich SPM (mainly clays, silts, and fine sands, Table 1) in coastal waters. POC represents only a very low fraction of total SPM at the river mouths, and the POC/SPM ratio typically increases with the distance from the coast with increasing proportion of organic-rich particles, i.e., phytoplankton cells and associated detrital organic matter. This feature is common to most world river plumes (Ludwig et al. 1996; Doxaran et al. 2012), and it was observed in the three sampled river plumes. Samples collected close to the river mouths presumably contain particles characterized by a heavier mineral matrix likely to perturb the sample pre-treatment required for carbon isotope or POC measurement.

The present results (Fig. 3b) suggest that the heterogeneity of δ13Corg values observed with the standard decarbonation protocol (protocol #1) was attributable to matrix effect in high SPM or OC-poor waters. Thus, the standard decarbonation protocol might not be suitable for estuarine environments where suspended particles exhibit high amounts of clays. This hypothesis is strengthened by the relatively low shift of δ13Corg toward high values observed in the Paillon discharge: the mineral matrix effect is presumably stronger in the Rhône and Var discharges, chiefly due to the abundance of clays in their beds (see Table 1 for the Var and Pont et al. 2002 for the Rhône), which might alter δ13Corg analysis.

The mineralogical analysis of samples (Table 1) confirms the differences visually observed between water samples collected at the mouths of the Var and Paillon Rivers (Fig. 1d): there is significantly more clay in the Var than in the Paillon waters (20 vs 10%, respectively). One may thus infer that clay-rich suspended particles are likely to hinder total decarbonation. Regarding the Rhône River, mineralogical analyses were missing in the present work, but some studies have pointed out the relatively high content in illite and chlorite clay minerals in this river (Pont et al. 2002). According to dissolution kinetics studies (Morera-Chavarría et al. 2016), detrital dolomite (Ca and Mg-rich carbonate) and siderite (Fe-rich carbonate) are more resistant to acid decarbonation than detrital calcite or, by extrapolation, calcareous phytoplankton cells. The Paillon particles contain a relatively high content of carbonates, while calcite and dolomite contents are roughly similar between the Paillon and Var Rivers (Table 1). This suggests that the abundance of carbonates does not significantly perturb the decarbonation process.

It is noteworthy that, if the POC content (%POC) is only slightly affected by the classical decarbonation protocol (Lorrain et al. 2003), the impact on δ13Corg is expected to be strong and may only refer to a small fraction of the total mass composed of remaining carbonates with significantly different δ13Cinorg value.

Conclusions

The physicochemical composition of suspended particles in the mouths and plumes of three rivers during floods has been analyzed. The elemental and isotopic analyses of POC show a strong heterogeneity, which results from the protocol used to remove carbonates. This study shows that carbon isotopic variability can be used to validate the results of the OC elemental analysis and to ensure the completeness of the decarbonation protocol.

Analyses of POC in natural water samples require a compromise between efficiency of decarbonation and preservation of OC, both in nature and in concentration. The extent to which the POC content may be affected by the decarbonation method varies significantly according to (i) the mineral or chemical matrix characterizing the sample (salty water, estuarine water, sediment, clays, dolomite, etc.) and (ii) the decarbonation protocol, i.e., mainly, the type (e.g., hydrochloric, phosphoric or sulfuric) and amount of acid used. Basically, the stronger the matrix, the greater the amount of acid needed, and the greater the risk of losing OC. Most of the protocols seem nowadays suitable for measurements of OC concentration. However, this work shows that isotopic signatures may be much more sensitive to partial decarbonation. Measurements of %Corg may be biased by partial decarbonation in high SPM waters, and the consideration of δ13Corg values thus permits the determination of whether the remaining C is homogenous and actually exhibits a representative isotopic signature. Therefore, the δ13Corg heterogeneity may be used to control the quality of the POC data for estuarine environments, and the efficiency of the decarbonation protocol.