3.1. Gas composition

With the exception of one CH₄-rich bubbling spot in the Dziani lake (DZE), all studied samples of the Mayotte gas seeps are CO₂-dominated. The Mayotte noble gas show a more variable composition for the Dziani Dzaha Lake than for the Airport tidal flat (BAS; Tables 1a–1e), with ⁴He ranging between 8.2 and 113 ppm at BAS and between 42.5 and 1139.5 for the Dziani Dzaha Lake. The ²⁰Ne values for the BAS zone range from 0.1 to 7.7 ppm and between 0.8 and 18.6 for the Dziani Dzaha Lake. Finally, the ⁴⁰Ar is in the range 55.2–3346.6 ppm in BAS and 281.4–2055.1 for the Dziani Dzaha Lake. Some of the samples in both Dziani and BAS areas have a significant air contamination showing concentrations of N₂ and O₂ up to 76.4% and 20.4% respectively (Tables 1a–1e, Figure 2a).

The BAS bubbling gases from the tidal flat show a CO₂ dominated composition with concentrations up to 98.8% and a variable concentration in CH₄ ranging between 16 and 5291 ppm with the lowest value related to an air-contaminated sample (C₂a₂-glass). The BAS gases show low concentrations in H₂ and CO, ranging from below detection limit to 11 ppm for H₂ and to 18 ppm for CO (higher values for H₂ and CO in Tables 1a–1e refer to samples collected in steel samplers, which we do not consider reliable, as discussed in Section 4.3). At the Dziani Dzaha Lake, CO₂ is also the dominant gas species in two of the three analyzed spots (DZW, DZN), with values up to 97.3%, while CH₄ is variable between 3897 and 81,900 ppm, except in the third spot (DZE) where CH₄ is the dominant gas. The H₂ and CO are generally present in low concentrations.

The chemical composition of the Mayotte gases plotted in a N₂, He and Ar ternary diagram (Figure 2a) follows a mixing trend between a He-rich component and an atmospheric component (air or air-saturated water—ASW). The bubbling gases from both areas at Mayotte show a variable degree of contamination by an atmospheric end-member, and its contribution is probably slightly higher for air than for ASW at least in BAS. On the whole, the He–Ar–N₂ variability falls within a typical compositional range of gases emitted in intraplate or extensive tectonic settings. The two dominant mixing sources appear to be atmospheric and MORB-type mantle, and they are distinct from typical subduction-related gases, which have higher N₂/Ar ratios due to the N₂ excess released to the wedge by subducting sediments [e.g., Barry Hilton 2016]. With regard to the helium variability, the similarity between Dziani Dzaha Lake and BAS is interesting, as there seems to be no appreciable difference in the air-MORB mixing trends. This suggests that shallow processes do not significantly affect helium abundances at Petite Terre.

The chemical composition in relation to the plot of CO₂–CH₄–He (Figure 2b) highlights that low temperature gas seeps of Petite Terre have a general high CH₄ concentration. A relatively higher abundance of CH₄ in Dziani than in BAS tidal flat is likewise evident. Since the beginning of the seismo-volcanic crisis, the CO₂-rich bubbling spots of the two sites show comparable abundances of CH₄, and Dziani still hosts the only CH₄-dominated bubbling area. The INGV-1 sample in Dziani is the most enriched in methane (Figure 1c); however, it should also be noted that this specific sample is also affected by air contamination.

3.2. Systematics of noble gas isotope ratios, CO₂ and CH₄ isotopes

Tables 1a–1e report the isotopic compositions of noble gases, CO₂ and CH₄ in the sampled gases from the two bubbling areas of Petite Terre. Regarding the measured ³He/⁴He ratios, these vary between 5.45 and 7.5 R_a in BAS, and between 5.3 and 6.8 R_a in Dziani Dzaha Lake (Figure 3). It is worth noting that the range of R/R_a values is very similar between BAS and Dziani, although it should be noted that less data is available for Dziani and that they are only related to the 2020 campaign. In both BAS and Dziani, a few samples show a clear air contamination as recorded by relatively high content of N₂ and O₂. It is possible that these samples underwent some issues during the sampling operations or during the storage and transport to the laboratory that fractionated the ³He/⁴He, (for example MAR-1 Figure 3) leading us to exclude them for further discussion. The majority of samples do not show significant air contamination as indicated by both the chemistry of gases and the noble gas systematics (Tables 1a–1e). For instance, the ⁴He/²⁰Ne ratios is at least 2 orders of magnitude higher than the ratio in air (0.318) reaching 1663 in the BAS (this study) and up to 2750 in the survey carried out by BRGM in 2008 [Sanjuan et al. 2008], while is reaches up to 2122 in the Dziani Dzaha Lake.

The ⁴⁰Ar/³⁶Ar values range from the atmospheric ratios up to 434 in BAS, and up to 468 in the Dziani Dzaha Lake. The few air contaminated samples have lower ratios, like MAR-1 and MAN-2 in BAS site (⁴⁰Ar/³⁶Ar = 290 and 308; and ⁴He/²⁰Ne = 1.07 and 43.59, respectively), and INGV-01 in the Dziani Dzaha Lake (⁴⁰Ar/³⁶Ar = 306 and ⁴He/²⁰Ne = 33.3).

With the exception of a few air-contaminated samples, the difference between R/R_a and R_c/R_a is almost negligible.

The ⁴He/⁴⁰Ar^∗ ratio of BAS gases ranges between 1.3 and 1.7, with a general overlap of values between different sampled pools and sampling periods, and they are within 1.1 and 1.5 in Dziani Dzaha Lake, irrespective of the variable CH₄/CO₂ ratios. These values fall within the typical range of mantle production ratio [⁴He/⁴⁰Ar^∗ = 1–5; Marty 2012] and magmatic values from other geodynamic settings [e.g., Reunion hot spot, Boudoire et al. 2018; Eger Rifţ intra-plate in Europe, Bräuer et al. 2011; Etna intra-plate, Paonita et al. 2012; Kolumbo arc volcano in Greece, Rizzo et al. 2019].

The ⁴He/⁴⁰Ar^∗ variability is commonly used to track magmatic degassing processes due to the ∼7–10 times lower solubility in silicate melts of Ar than He [e.g. Burnard 2001; Paonita et al. 2012; Barry et al. 2014; Boudoire et al. 2018]. ⁴He/⁴⁰Ar^∗ ratios at Petite Terre do not show systematic variations as a function of location or time. Liuzzo et al. [2021] suggest that this homogeneous signature reflects that all bubbling spots are related to a single degassing source and pressure, likely related to magmas stored close to the mantle-crust underplating depth (15–20 km depth), as observed in other magmatic systems [e.g., Reunion, Boudoire et al. 2018]. This depth range corresponds to the shallower part of the magmatic plumbing system below Mayotte, where evolved basanitic magma differentiate to form phonolitic melts [Berthod et al. 2021a, b; Foix et al. 2021].

The C-isotope compositions of CO₂ (δ¹³C_{CO 2}) at BAS gases vary from − 5.7‰ and − 3.5‰, where the most negative ratios are those measured in samples from MAN-1 and MAN-2 a sampling pool with relatively low gas flux and located close to a dense area of Mangrove trees on the airport flat. At Dziani Dzaha Lake the variability is much wider with δ¹³C_CO2 spanning between − 0.9 and − 6.3‰ (data from this work), where the most negative value is that of the sample INGV-01 from the CH₄-rich plume (DZE). Interestingly, previous data collected in 2016 from Milesi et al. [2019] show an even higher variability with ¹³C_CO2 between − 0.3 and − 20.1, where the most negative correspond to G2 sample in Milesi et al. (2019) collected in the same area of DZE. Figure 4a shows the δ¹³C_CO2 distribution and highlights a significant separation of the C isotope signature between the two bubbling areas, where Dziani is systematically more positive than BAS (with the exception of G2 not represented in the plot) and where the δ¹³C_CO2 in BAS are confined to a narrow range of about 2 delta units.

On the contrary, the distribution of the C-isotope composition of CH₄ (δ¹³C_{CH 4}) (Figure 4b) partly overlaps in BAS and Dziani areas, with the BAS site showing again a fairly narrow range of variability, (from − 25‰ to − 10‰). At BAS, the least negative values are found in the MAN low-flux pool located close to the mangrove area (MAN-1 and MAN-2 samples) that is considered affected by gas–water dissolution by Liuzzo et al. [2021]. The site of the Dziani Dzaha Lake shows a wider variation in δ¹³C_CH4 fluctuating from minimum values of more than − 65‰ (DZE area) to maximum values of − 11‰ (DZW area). The different isotopic variability of carbon in a δ¹³C_CO2 and δ¹³C_CH4 space is evident in Figure 5, where the two markers are linearly and negatively correlated (R² = 0.7) in the BAS site and positively correlated in the Dziani Dzaha Lake. Interestingly our new dataset on Dziani Dzaha Lake clearly shows more negative δ¹³C_CH4 values with respect to that of Milesi et al. [2020]. Regarding hydrogen isotopic values in CH₄ (δD_{CH 4}), our samples yielded a δD of − 118.1‰ and − 137.8‰ V-SMOW, respectively in BAS and − 124‰ and − 184‰ V-SMOW in Dziani.

4.1. CO₂ origin and gas–water interaction

In order to assess the origin of CO₂ rich gas emissions in Mayotte, we followed the approach used in Sano and Marty [1995] and we extended the dataset from the BAS area discussed in Liuzzo et al. [2021] with new gas sampling made in 2020 in BAS and in the Dziani Dzaha Lake. In order to discriminate the possible contributions from various sources (magmatic, organic and marine sediments), we considered the isotopic variability of carbon in CO₂ versus CO₂/³He ratio (Figure 6). The two mixing curves were modeled between the local mantle endmember resulting from the average values of our data (in which we only considered data that were not modified by secondary processes), data from literature [CO₂/³He = 5.0 × 10⁹ and δ¹³C = −4.3‰, Liuzzo et al. 2021] and an organic (δ¹³C = −25‰) and limestone endmember (δ¹³C = 0‰) from Hoefs [2015]. For both organic and limestone carbon endmembers, a value of CO₂/³He = 1.0 × 10¹³ is assumed. Finally, in order to evaluate the secondary processes of gas–water interaction, we have considered data corrected for air only for samples having N₂ < 22%, following the approach of Liuzzo et al. [2021].

The distribution of the data for the two degassing areas at Petite Terre shows clear differences. With the exception of the MAN-1 and MAN-2 samples, the BAS area shows little variability in bulk chemistry and C isotopy and are reasonably interpretable as being related to an outgassing process from a deep magmatic source. Conversely, the gas samples from the Dziani Dzaha Lake area have a larger scattering.

Gas–water interaction effects previously highlighted for some samples from the BAS area by Liuzzo et al. [2021] likely play an even more important role in the Dziani Dzaha Lake. We modeled four possible Rayleigh fractionation curves by assuming gas dissolution in water under equilibrium conditions (Appendix A, Equation (.3)). However, whereas in the case of the BAS zone a simple dissolution process can be assumed (MAN 1-2 curve Rf-1) [Liuzzo et al. 2021], in the case of Dziani, a single dissolution step cannot reproduce the measured large scattering. Moreover, when considering the typical parameters of the Dziani Dzaha Lake (pH = 9 and temperature of 36 °C from Milesi et al. [2020]) the corresponding curve Rf-2 clearly does not fit with any Dziani Dzaha Lake data. The pH value of up to 9 in Diazni Dzaha Lake can potentially facilitate calcite precipitation. However, the precipitation of calcite, by subtracting the heavier isotope, would produce a more negative δ¹³C component in the gas phase (resulting in a curve that would overlap with Rf-2), but in our case the trend of the values measured in the lake is the opposite. In order to fit the Dziani Dzaha Lake gases in the range of temperatures of 32–36 °C, much more acid pH of about 5.2–5.7 are needed (Rf-3 path). If, on the other hand, fractionation in equilibrium with the lake’s pH of about 9 is considered, the temperature required to obtain a good fit on the data (Rf-4 path) should be about 102 °C. However, these high temperatures are not realistic, nor have such temperature anomalies ever been found in any of the surveys performed in the lake before or after the beginning of the seismic crisis. Modeling seems to suggest that simple fractionation does not occur in the water column of the Dziani Dzaha Lake, on the other hand if the fractionation process occurs in the aquifer and does not undergo major changes in the lake, the Rf-3 model is likely the most plausible.

Nevertheless, the variability of He/CO₂ versus R_c/R_a shown in Figure 6b, evidences that a partial dissolution in water did play a role in modifying the composition of gas bubbling streaming through the Dziani Dzaha Lake, and appears to be notably extreme in the CH₄-rich INGV-1 sample. In particular, this sample is collected in correspondence of the deepest portion of the lake (Figure 1c), and as already noted for sample G2 in Milesi et al. [2020], may have been significantly affected by the gas transfer and residence in the higher water column.

The current dataset does not permit to exclude that gaseous emissions from Dziani Dzaha Lake undergo gas–water interaction processes under non-equilibrium conditions and/or the occurrence of other fractionation processes than those here modeled. Among these, isotope exchange in Dziani between δ¹³C_CO2 and δ¹³C_TDIC could lead to the less negative isotopic values measured at the Dziani Dzaha Lake with respect to BAS, if we consider that water lake reaches DIC concentrations up to 0.2 mol/L and δ¹³C_TDIC up to + 13‰ [Cadeau et al. 2020]. The acquisition of more data on both exsolved and dissolved gases in the future is thus needed in order to better constrain the processes controlling the variability of the carbon isotopic signature in the Dziani Dzaha Lake. All of these effects need to be evaluated and eventually filtered out in order to calculate the thermobarometric conditions of the hydrothermal system feeding the gas seeps (see next section), as it has been recognized in other studies of hydrothermal gases [Capasso et al. 2005; Gilfillan et al. 2009; Rizzo et al. 2019]. In fact, besides δ¹³C isotopic signature, the CO₂/³He, He/CO₂, CH₄/CO₂ ratios can also be potentially modified by oxydation/reduction reactions and/or by gas–water interaction in which CO₂ dissolves preferentially with respect to the other species.

4.2. CH₄ origin and δ¹³C versus δ²H variability

For the samples collected in the 2019–2020 campaigns, it was possible to define the isotopic signature of carbon and hydrogen in methane in both BAS and the Dziani Dzaha Lake sites and compare them with previous data of Dziani Dzaha Lake from Milesi et al. [2020]. The results are plotted in the classification diagram proposed by Schoell [1980] (Figure 7). It must be stressed that distinguishing between methanogenesis processes of biological origin and thermogenic processes at the origin of CH₄ [Mazzini et al. 2011; Schoell 1980; Welhan 1988] is complicated by the possible mixing between endmembers with different isotopic signatures [Taran et al. 2010a, b], or by the occurrence of oxidation processes [e.g., Batista Cruz et al. 2019].

Although with some relative variability, the isotopic signature of methane in BAS falls in the field of abiogenic origin, together with sample G7, acquired in 2016 in this area by Milesi et al. [2020]. This confirms that outgassing in the BAS tidal area originates from a homogeneous and unique source, particularly if we also take into account the stable and low variability of R_c/R_a and δ¹³C‐CO₂ values (Figures 3–5). The situation in the Dziani Dzaha Lake is more complex. The CH₄-rich G₂ sample collected near the deepest part of the lake corresponds to a gas that has clearly undergone a reduction process by microbial methanogenesis, as already highlighted in Milesi et al. [2020]. Our INGV-01 sample from this area (Tables 1a–1e), records a possible strong interaction in water or in the sediments (Figure 6b). The isotopic characteristics of methane for this sample can be consistent with a thermogenic origin or with mixing between abiogenic and biogenic endmembers. In the Dziani Dzaha Lake, the isotopic signature of the CH₄-rich bubbling is thus distinct from that of the CO₂-rich bubbling (ex. G3), which have abiogenic signature similar to the samples in BAS, but with a less negative proportion of δ²H. The remaining samples from the 2020 campaign in Dziani have a signature intermediate between abiogenic and thermogenic fields (DZN 1-2; OVPF-5; DZW 1-4). The variability of methane isotopic characteristics in distinct outgassing areas in Dziani thus appears to be much more complex than measured in BAS, which is confirmed also by the wider variability in δ¹³C signature in CO₂ (Figures 4, 5). This again suggests that several processes may operate in the lake, including abiotic degassing of gases, microbial production and oxidation of methane, and partial dissolution of CO₂ in water, all of which contribute to varying degrees to the greater chemical and isotopic variability compared to the more homogeneous BAS area.

It is therefore clear that further data are needed to better constrain the origin of methane in the Dziani Dzaha Lake, and we are aware that the debate on methane origin is open within the scientific community, in particular in order to justify its possible abiotic origin assuming a “magmatic” or “late magmatic” origin [Etiope and Sherwood Lollar 2013]. However, answering this question is not trivial, considering that the recent submarine eruption only 50 km off-shore from Petite Terre represents by far the largest known submarine eruption until now and that intense seismicity occurs at variable distance from the island [Feuillet et al. 2021; Berthod et al. 2021a, b; Foix et al. 2021]. In this context, a further clue is given by the isotopic signature of helium involved in the outgassing process at Petite Terre, which can be useful in discerning and assessing the deep origin of the gas. Therefore, we considered the isotopic variability of methane by comparing it with the percentage of mantle-related helium, following the approach indicated in Etiope and Sherwood Lollar [2013] (Figure 8). In our case, the Petite Terre gases fall within the area where magmatic CH₄ inputs are clearly recognized (EPR—East Pacific Rise, Socorro, Lost City [Proskurowski et al. 2008; Taran et al. 2010a; Welhan and Craig 1983]). This leads us to conclude that input of gas of magmatic origin contributes to the general outgassing at Petite Terre, and specifically even regarding a not negligible contribution of methane, which in turn, particularly in the Dziani Dzaha Lake area, has undergone further transformation processes by microbial activity.

4.3. Equilibrium temperature of hydrothermal gases

The elemental composition of the gases of Petite Terre shows a general low concentration of H₂ and CO in both BAS and Dziani Dzaha Lake areas. Therefore, although H₂ and CO are considered useful geo-indicators for equilibrium temperature and pressure in the hydrothermal system, we did not consider these species to be suitable for thermobarometric purposes in the study of Mayotte fluids. On the contrary, the amounts of CO₂ and methane (Tables 1a–1e) are high enough to have good analytical precision in both concentration and isotopic data. Therefore, we used the approach adopted by Liuzzo et al. [2021], assuming that in the hydrothermal system an equilibrium is attained between the dominant species H₂O–H₂–CO₂–CO–CH₄, in which the formation of methane is favoured by the decreasing temperature from the reaction (.4) indicated in Appendix A.4. For this system, the temperature has been calculated assuming a condition of thermal equilibrium between CH₄ and CO₂ by (.5) in Appendix A.4 as proposed by Giggenbach [1992]. Moreover, in order to further constrain the possible evidence of recent input of deep fluids in the Mayotte hydrothermal system, we evaluated the thermal equilibrium in combination with their isotopic signatures based on their δ¹³C isotopic fractionation factor between CO₂ and CH₄. To this aim, we combined the temperatures obtained from (.5) with the temperatures calculated using (.6) proposed by Bottinga [1969], valid for temperatures ranging between 0–700 °C (Appendix A.4), following the approach proposed in Ono et al. [1993], and recently applied in the Comoros area by Liuzzo et al. [2021], finally obtaining (.7). On this basis, the curves of thermal equilibrium were modeled assuming that both chemical and isotopic equilibrium is maintained with a fixed δ¹³C_C02 representative of the possible range of magmatic signature; here we considered a range of δ¹³C_C02 magmatic signatures from − 4‰ to − 8‰ when coupling (.5) and (.6).

The results are shown in Figure 9, which allows to extend the preliminary study made by Liuzzo et al. [2021], taking into account recent gas sampling in 2020 at both BAS and Dziani Dzaha Lake areas of Petite Terre. The data considered are those in which there are no obvious secondary variations and/or gas–water interactions.

The methane-rich gases at the Dziani Dzaha Lake are certainly conditioned by the microbial activity in lacustrine waters and sediments; the range of variability of δ¹³C_(CH4) and δD is high, and the values are generally scattered. The variability of δ¹³C_(CO2) is less than for δ¹³C_(CH4); however, δ¹³C_(CO2) of gases bubbling through the Dziani Dzaha Lake show a wider range of variability than BAS. As noted in the previous sections, this leads us to conclude that outgassing in Dziani Dzaha Lake is probably affected by several physicochemical processes and are related to several sources. Their isotopic variability is related to variable degrees of mixing between organic and abiotic components [Milesi et al. 2020]. Figure 9 also shows that isotopic equilibrium is not reached between CO₂ and CH₄, as all sampled gases at Dziani Dzaha Lake have variable δ¹³C_(CH4) values above the calculated equilibrium curves. Bacterial oxidation of CH₄ likely controls isotopic fractionation, determining an increase in the isotopic ratio in the residual methane [Baker and Fritz 1981; Coleman et al. 1981; Horita 2001], and this process is expected to be significant in the Dziani Dzaha Lake, where an extensive microbial activity has been well documented.

Regarding the BAS area, the range of variability of δ¹³C_(CH4) is consistent with an abiogenic source [Schoell 1980], and the data also show moderate variability in their isotopic signature. Therefore, a single degassing source produces the bubbling observed at BAS. In the 2020 samples, however, a significant shift of the methane toward heavier isotopic concentrations, as observed in Liuzzo et al. [2021] is still evident. Although a carbon isotopic fractionation of methane cannot be excluded in BAS area as in Dziani, Liuzzo et al. [2021] have shown that isotopic signature of the hydrothermal gases in the BAS area likely records a quenching effect. In this interpretation, CO₂ and CH₄ are considered initially in isotopic equilibrium during outgassing of the deep (mantle level) magmatic source. In this interpretation, isotopic disequilibrium is a consequence of fast ascent of the gases to shallow crustal layers, with little time for isotopic re-equilibration. Such a quenching effect is expected because of the much faster rate of re-equilibration (about 100 times) of the elemental reaction with respect to the isotopic one [Giggenbach 1982]. Our new dataset confirms that isotopic disequilibrium between CO₂ and CH₄ previously shown in 2018–2019 samples, is still well recorded by the 2020 samples. An important consequence of the disequilibrium is that the actual temperatures could be higher than those calculated from the equilibrium obtained from (.5), as discussed by Liuzzo et al. [2021]. This is because, having numerically simulated a thermal isotopic-chemical equilibrium between CO₂, CH₄ and their d δ¹³C, the corresponding δ¹³C values of methane would be in equilibrium if they intersected the corresponding curve. However, the δ¹³C(CH₄) values would have to shift to the left to meet the equilibrium curve and thus would have even higher temperatures.

The occurrence of isotopic disequilibrium implies that the hydrothermal system at BAS may have recently received new input of deep-hot CO₂ and CH₄-rich gases possibly before or at the beginning of the volcanic activity in 2018.

This hypothesis is further supported by the time evolution of the equilibrium temperatures (calculated with (.5)) over years in both the BAS and Dziani areas (Figure 10). In the BAS area, a well-defined trend of cooling can be observed between 2018 and 2021. The equilibrium temperature of gases collected in the Dziani Dzaha Lake follows the same trend even if the dataset is smaller and the gases have been variably affected by secondary processes. The cooling trend recorded by equilibrium temperatures mimics that of progressive decrease of the seismic activity and magma extrusion rate over the same time period [Bachèlery et al. 2019; Berthod et al. 2021a, b; Cesca et al. 2020; Feuillet et al. 2021; Lemoine et al. 2020; REVOSIMA 2019; REVOSIMA Monthly Bulletin, December 2021]. It is therefore reasonable to assume that a deep fluid input, which was somehow related to the most intense phase of submarine eruptive activity, may have reached the outgassing areas of Petite Terre, resulting in the initial increase in equilibrium temperature and isotopic disequilibrium in the hydrothermal system, which then over time fell steadily, as the seismo-volcanic crisis declined. The existence of this link is expected as most of the seismicity and the deep sources of magma feeding the distal submarine eruption are located close to Petite Terre (5–15 km) [Foix et al. 2021] and the involved volumes of magma are huge and CO₂-rich [Feuillet et al. 2021; Berthod et al. 2021a, b].