Comptes Rendus Géoscience Sciences de la Planète

. Since May 2018, an unexpected long and intense seismic crisis started o ff shore Mayotte (Indian Ocean, France). This ongoing seismic sequence is associated with the birth of a newly discovered submarine volcano 50 km east of Petite Terre. Here, we investigate the indirect impact of this crisis on the stability of an atypical ecosystem located in Mayotte, the Dziani Dzaha Lake. This lacustrine systempresentedphysical,chemicalandbiogeochemicalcharacteristicsthatweredistinctfromthose classicallyobservedinmodernlakesorseawater,e.g.highsalinity(upto70psu),lackofnitrate,sulfate content below 3 mM, and permanent water column anoxia below ca. 1.5 m depth (2012–2017 period). Based on three surveys conducted in 2020 and 2021, we documented an episode of water column oxygenation, a significant pH decrease and an impressive change in the carbon isotope signatures. These preliminary data suggest that the functioning of biogeochemical cycles in the Dziani Dzaha Lake is impacted by increased CO 2 inputs and the changes in the lake mixing dynamics, which is an indirect consequence of the ongoing seismo-volcanic crisis.


Introduction
Mayotte Island is one of the four main volcanic islands of the Comoros Archipelago located in the 300 Pierre Cadeau et al. the Comoros Archipelago were known for their diffuse and moderate seismicity with few historically felt events. The ongoing seismic sequence is thus unexpected and remarkable with regard to its duration and intensity (i.e. earthquakes of magnitude 5.9 on 15 May 2018). After the first two months of intense seismicity at the beginning of the crisis, seismicity still persisted a long time with numerous small earthquakes [Bertil et al., 2019], and ground deformation was observed [Lemoine et al., 2020]. A new submarine volcano was discovered at 50 km east of Petite Terre [Feuillet et al., 2019, and significant subsidence of Mayotte Island (up to 20 cm) was attributed to the drainage of a deep magma chamber near the coast . This long-term seismic sequence and the associated new active volcano offshore Mayotte led the scientific community to investigate further the seismo-volcanic hazard in this area [Feuillet et al., 2019, Cesca et al., 2020, Darnet et al., 2020, Lemoine et al., 2020, Berthod et al., 2021, Saurel et al., 2022. This seismo-volcanic crisis has been monitored since the beginning of May 2018; numerous studies were performed and many others are in progress to gain further information about the deep volcanic system, to better constrain the seismo-volcanic hazard for Mayotte and its population (e.g. seismicity, slope instability, tsunami, island subsidence or even new volcanic activity onshore).
On the Petite Terre Island in Mayotte, one of the latest eruptive events proposed at around 7 and 4 ky is probably at the origin of the Dziani Dzaha Lake formation [Zinke et al., 2003]. For the last decade, this atypical lacustrine system has been extensively studied [Leboulanger et al., 2017, Cellamare et al., 2018, Gérard et al., 2018, Milesi et al., 2019, Aucher et al., 2020, Jovovic et al., 2020, Sala et al., 2022. Based on its water chemistry, the lake was most likely initially filled with seawater , and has since evolved as a closed system. Before the seismovolcanic crisis, the physical, chemical and biological features of this ecosystem were far from those observed in a modern oceanic or lacustrine environment (details in Study site section). The ongoing seismo-volcanic crisis offshore Mayotte could have a significant impact on the stability and the functioning of this ecosystem through the reactivation of fractures in the crater/basement, the infiltration of fresh/sea water and/or the increase in magmatic bubbling and degassing through its waters [Liuzzo et al., 2021].
In order to investigate the impact of the perturbations related to the growth of the submarine volcano offshore Mayotte on this lacustrine system, we conducted three field trips in 2020 and 2021 on the Dziani Dzaha Lake. We documented the evolution of the physical and chemical properties in the water column (e.g. [O 2 ], pH, salinity, T), as well as the carbon and nitrogen isotopic signatures of dissolved inorganic carbon (DIC) and suspended particulate matter (SPM)) within it (i.e. δ 15 N SPM , δ 13 C DIC , δ 13 C SPM ). The comparison of these results with those obtained before the seismo-volcanic crisis illustrates the response of this lacustrine system to this kind of geological perturbations.

Study site
The Dziani Dzaha Lake is a shallow maar lake located on the Petite Terre Island of Mayotte ( Figure 1). This lake is small (i.e. 2.36 × 10 5 m 2 ), close to sea level and separated from the nearby seashore by a 220-m-thick crater wall [Rouwet et al., 2021]. The lake watershed is restricted to the volcanic crater (i.e. ∼1 km in diameter and 50-100 m height). The average depth of the Dziani Dzaha Lake is about 3 m with a narrow depression reaching 18 m depth (Figure 1), likely related to the chimney of the phreatomagmatic eruption at the origin of this lake. Magmatic gases bubble through the water column in several locations , Liuzzo et al., 2021 mostly in shallow areas of the lake, with one bubbling site in the central part for which a strong increase in the bubbling activity has been observed since 2020 compared to our last observations from 2012 to 2017 (Figure 1).
From 2010 to 2017, i.e. before the beginning of the seismo-volcanic crisis in May 2018, the lake water column showed an atypical combination of physical and chemical features for a modern lacustrine system [e.g. Sarazin et al., 2021]. Salinity measured was between 35 and 70 psu (practical salinity unit), nearly twice the usual salinity of seawater. The alkalinity was also very high, with values ranging from ca. 0.1 to 0.2 M (nearly 100 times the alkalinity of seawater), and pH ranged between 9 and 9.5. The water column was periodically stratified with a sharp halocline at ca. 2 m depth about half of the year, driven by the salinity decrease in surface waters as a result Picture of the lake taken in October 2014, and (D) Bathymetric map of the lake, with blue circle representing the water column sampling location and the red circles representing the identified bubbling areas at the water-air interface (modified from Leboulanger et al. [2017] and Cadeau et al. [2022]). of increasing precipitation during the rainy season. A permanent chemocline was also present at ca. 14 m depth in the narrow depression . Importantly, the lake was permanently anoxic below ca. 1.5 m depth despite seasonal variations in its water column structure, and was strongly euxinic (sulfides H 2 S/HS − concentration ranging from 2 to 6 mM) below 2 m depth when the halocline is present (during the stratified period), or only below a 14 m depth when it is not [during the non-stratified period, Sarazin et al., 2021]. Details on the evolution of the physical and chemical parameters according to stratification are available in previous works [e.g. Leboulanger et al., 2017, Cadeau et al., 2020. All our field trips done since the beginning of the crisis were conducted during nonstratified periods (i.e. from August to December). Before the crisis, these non-stratified periods were characterized by the absence of halocline (except at ca. 14 m depth). As a result, most physical, chemical and biological parameters were constant with depth down to the deep chemocline at 14 m depth, except for the dissolved oxygen that was only present down to a maximum of 1.5 m depth depending on the photosynthetic activity. The fact that O 2 saturation values fluctuate widely (from 0 to 350%) is related to large nycthemeral fluctuations in the balance between the activities of photosynthesis (which produces O 2 ) and respiration (which consumes it) in this biologically very active lacustrine system [e.g. Leboulanger et al., 2017, Cadeau et al., 2020. The salinity was close to 60 psu, the alkalinity close to 0.14 M, the pH close to 9.2, the [SO 2− 4 ] content close to 3 mM, [H 2 S] content close to 0.1 mM, and a relatively homogeneous diversity of cyanobacteria, bacteria, and archaea was observed along the water column . Below the 14 m deep chemocline, the salinity and alkalinity increased to 70 psu and 0.2 M, respectively, the pH decreased to a value close to 9, no SO 2− 4 was observed while H 2 S accumulated up to 6 mM. Our previous work has demonstrated that the characteristics of non-stratified seasons have remained stable since 2010 [e.g. Leboulanger et al., 2017, Bernard et al., 2019.

Materials and methods
The isotope signatures in this study were acquired on samples from surveys conducted in November 2020, September 2021 and December 2021. A nomenclature consistent with previous publications was used to identify each sample taken during the different surveys. For example, for the sample "DZ21-9 2m," "DZ" indicates the Dziani Dzaha Lake, "21-9" indicates the year (2021) and month (September) of the survey, "2m" refers to sampling depth of the water column station called "18m" as shown on Figure 1.
Water samples were collected using a horizontal 1.2 L Niskin bottle along a vertical profile at the 18 m depth station for each survey. For each sampling depth, samples were taken in 500 mL plastic bottles. Each sample was then filtered onto precombusted Whatman GF/F glass fiber filters (0.7 µm) and quartz Millipore AQFA (size cut-off ca. 3 µm) for δ 13 C POC and δ 15 N PON analysis, and a 12 mL Labco Exetainer glass tube was filled with the filtrate and fixed with 0.2 mL of saturated HgCl 2 solution for DIC and δ 13 C DIC analyses. The GF/F filters with SPM were decarbonated by exposure to concentrated HCl fumes in a desiccator for 12 h to avoid loss of material.

Physical and chemical in situ measurements
In situ profiling were performed with a CTD multiparameter EXO2 probe (temperature, conductivity/salinity, pH, ORP, O 2 , turbidity, chlorophyll, phycocyanine (PC) and fDOM) and a turbidity AQUAlogger 210 TYPT Aquatec probe. The pH electrode was calibrated using pH 4.00, 7.01 and 10.00 NBS-NIST buffers; O 2 optode was calibrated at 100% saturation in wet air; salinity sensor was calibrated with a 34.271 psu seawater. Other sensors have not been calibrated and the corresponding parameters are used as relative values; nevertheless, most of these sensors offer quite stable response (especially T, turb., chl., PC, fDOM), at the exception of oxidationreduction potential (ORP) electrode which can only be used as relative values. In addition, two O 2 NKE SDOT optodes were deployed at 0.25 m depth from 10/09 to 13/12 2021 near the central gas bubbling zone (at ca. 1-3 m from the outlet, respectively), and a water level probe (OTT Orpheus Mini) was installed in September 2021 in the NO part of the lake.

Carbon and nitrogen isotope measurements
DIC and δ 13 C DIC were measured using an AP2003 mass spectrometer. Phosphoric acid (100% H 3 PO 4 ) Figure 2. Vertical profiles of physical and chemical parameters recorded during each survey including salinity (psu), temperature (°C), pH and dissolved oxygen (%). The grey profiles correspond to previous data obtained before 2018 from Cadeau et al. [2022].
was injected into 12 mL Labco Exetainer tubes that were then flushed with helium. For each sample, 0.1 mL water was sub-sampled and transferred into the acidified Labco Exetainer tubes. The samples were agitated for 15-24 h so that CO 2 in the water was in equilibrium with the CO 2 in the headspace. The CO 2 in the headspace gas was then separated by gas chromatography with helium as a carrier gas and analyzed using the AP2003 mass spectrometer. Solutions presenting a range of DIC were prepared using sodium hydrogen carbonate for DIC calibration. Three internal carbonate standards with known isotopic compositions (Across, Merck and Rennes II) were used for isotopic composition calibration. The standard carbonate powder was loaded into 12 mL Labco Exetainer tubes with 0.1 mL distilled water before being flushed with helium and then analyzed in the same way as DIC. DIC values are expressed in mol·L −1 with a reproducibility of ±0.01 mol·L −1 , and δ 13 C DIC values are expressed in relative to Vienna Pee Dee Belemnite (VPDB) with a reproducibility of ±0.1 (1σ).
The δ 13 C POC and δ 15 N PON of the decarbonated filters were measured using a Flash EA1112 elemental analyzer coupled to a Thermo Finnigan Deltaplus XP mass spectrometer via a Conflo IV interface (Thermo Fisher Scientific, Waltham, MA, USA). The decarbonated samples were loaded into tin capsules and heated to 1200°C in a combustion tube with a mixture of chromium oxide and silver cobalt oxides. The combustion gases were carried by helium through a reduction column and a gas chromatography column to separate CO 2 from N 2 and from the other gases. CO 2 and N 2 were successively injected into the mass spectrometer for isotopic analysis. For the isotope calibration, four internal standards of organic-rich soil or sediment were analyzed in the same way. δ 13 C POC values are expressed as relative to VPDB with a reproducibility of ±0.2 (1σ) and δ 15 N PON are expressed as relative to air with a reproducibility of ±0.3 (1σ).

Physical and chemical features of the water column
In September 2020, salinity was constant down to 14 m depth with an average value close to 55.6 psu and increased below this depth up to 59.7 psu (Figure 2, Table 1). pH was constant down to 14 m depth with an average value of 9.1 and decreased below this depth down to 8.6. Temperature decreased in surface waters from 34°C to 30.3°C at 1.5 m depth, and remained constant with depth down to the lake bottom. In September 2021, salinity, pH and temperature significantly decreased in the water column.  Figure 2) that are related to the high photosynthetic activity similar to before the volcanic crisis. Surprisingly however, the deeper part of the lake, which was anoxic all year long before the crisis was oxygenated at 20% of O 2 saturation. In December 2021, the temperature decreased from about 32°C in surface waters to 30°C at about 2 m depth. Temperature then remained constant down to 14 m and increased to values close to 30.5°C below 14 m depth. Salinity was constant down to 14 m depth with an average value close to 58 psu and increased below this depth up to 60.2 psu. pH was constant with depth throughout the water column with an average value close to 8.5. Dissolved oxygen was only present in surface water down to 0.8 m; below this depth the water column remained completely anoxic.
Physical and chemical features of the water column in September 2020 were similar to those observed during the surveys conducted before the seismo-volcanic crisis offshore Mayotte (Figure 2), except for the salinity which was already lower than those observed previously. In contrast, salinity and pH values were significantly lower in September 2021 compared to the pre-seismo-volcanic crisis values observed, whereas dissolved oxygen was distributed in the water column, which had never been observed before. In December 2021 the pH still remained as low as in September 2021, the salinity increased by 2.8 psu compared to September 2021 (up to about 5 psu in the deeper part of the lake), and the water column was anoxic again below 1.5 m depth. In addition, based on O 2 optode placed in the central part of the lake, monitoring of dissolved oxygen content indicates that oxygenation of the entire water column lasted for several months between September and December 2021 (Figure 3).

Carbon and nitrogen isotopic signature in the water column
In the Dziani Dzaha Lake water column, the DIC concentrations ranged from 0.15 to 0.21 M with an aver-  Table 2). An increasing trend with depth was observed below the deep chemocline at 14 m depths in November 2020 from 0.15 to 0.21 M. In September and December 2021, the DIC concentrations decreased between 6 and 12 m depths. Overall, these concentration data show some variability but remain around an average value close to 0.2 M throughout the water column, which is similar to the DIC concentrations observed before the seismo-volcanic crisis ( Figure 4). The DIC isotopic compositions (δ 13 C DIC ) ranged between 12.2 and 13.0 with an average value of  12.6±0.2 in November 2020, between 9.6 and 9.9 with an average value of 9.8 ± 0.1 in September 2021, and between 9.3 and 9.8 with an average value of 9.5 ± 0.1 in December 2021 (Figure 4). The δ 13 C DIC was constant with depth during these three surveys. The δ 13 C DIC observed in November 2020 was similar to the δ 13 C DIC measured before 2018 (i.e. average value of 12.1 , Figure 4), while the isotopic signatures observed in September and December 2021 were depleted in 13 C by almost 3 compared to the δ 13 C DIC measured before 2018.
The particulate organic carbon (POC) isotopic signatures (δ 13 C POC ) ranged from −14.1 to −10 with an average value of −13 ± 1.5 in November 2020, from −18.8 to −18 with an average value of −18.5± 0.2 in September 2021, and from −18 to −16.9 with an average value of −17.8 ± 0.3 in December 2021 ( Figure 4). In November 2020, the δ 13 C POC was constant with depth from the surface water to the deep chemocline at 14 m depth with an isotopic signature close to −14 , and significantly increased below it to −10 . These isotopic signatures and this isotopic pattern with depth are similar to the δ 13 C POC observed before 2018 (i.e. average value of −14 , Figure 4). In contrast, the δ 13 C POC in September and December 2021 were constant with depth throughout the whole water column and strongly depleted in 13 C compared to the δ 13 C POC observed before 2018.
The particulate organic nitrogen (PON) isotopic signatures (δ 15 N PON ) ranged from 7.8 to 10.7 with an average value of 8.6 ± 1 in November 2020, from 7.4 to 7.8 with an average value of 7.7 ± 0.1 in September 2021, and from 7.5 to 8.4 with an average value of 8 ± 0.2 in December 2021 ( Figure 4).
Overall, during these three surveys the δ 15 N PON values were close to 8 and constant with depth throughout the water column, except in November 2020 when a significant isotopic enrichment was observed below 14 m depth, up to a value of 10.7 . This isotopic enrichment is similar to the isotopic enrichments observed before 2018 (δ 15 N PON values up to 13 ), and contrasts with the constant isotopic values observed in September and December 2021 (Figure 4).

Physical and chemical features in the water column
The evolution of the physical and chemical parameters observed in the Dziani Dzaha Lake water column during the surveys conducted after the beginning of the seismo-volcanic crisis illustrates significant modifications of the water column characteristics. Indeed, before 2018, one of the main features of this lacustrine system was its great stability (since it was first studied in 2010), with a permanent anoxia of the water column below ca. 1.5 m depth [sometimes shallower depending on photosynthetic activity in surface water, e.g. Sarazin et al., 2021]. The oxygenation of the whole water column, as well as a significant decrease in salinity and pH values, had never been observed. These represent major perturbations of this lacustrine system. Two distinct processes related to the seismo-volcanic crisis may be invoked to explain these observations: (i) an impact of seawater or freshwater infiltration into the lake in response to the subsidence, and (ii) an intensification of volcanic degassing into the lake (with physical, chemical and biological impacts).
Water chemistry of the Dziani Dzaha Lake and the ocean are very different. A contribution of an oxygenated, more neutral and less salty water from the ocean (or from groundwater) into the lake could have a consistent impact with what is observed in the Dziani Dzaha Lake water column in September 2021. However, firstly, after infiltration into bedrock, ocean or groundwater is unlikely to remain welloxygenated. Moreover, given the very high organic matter content in the water column [Leboulanger et al., 2017, Cadeau et al., 2020, an oxygen supply from episodic water inputs into the lake would probably be quickly consumed through oxygenic organic matter degradation. Secondly, to significantly change the chemistry of the whole water column in the Dziani Dzaha Lake, water inputs would have to be significant, which seems unrealistic. Considering seawater and freshwater inputs with a salinity of 35 psu and 0 psu (Salinity input in (1)), respectively, and a lake salinity before the seismo-volcanic crisis near 63.5 psu (Salinity pre-2020 in (1)), the impact of water input on salinity can be tested using mass balance equations as follows: V post-2020 × Salinity post-2020 = V input × Salinity input +V pre-2020 × Salinity pre-2020 . (1) For instance, based on water volume calculations [e.g. Cadeau et al., 2022], an increase of the water level of about 20 cm (corresponding to the subsidence of Mayotte Island) would represent a water input of about 7.2% of the total water volume (V input in (1)). From this, the calculated salinities considering seawater or freshwater inputs are close to 61.4 psu or 58.9 psu, respectively. In September 2021, salinity was close to 55 psu. According to the mass balance equation, to obtain a calculated salinity of 55 psu it would be necessary to consider a freshwater input representing 12.7% of the total water volume. In addition, the origin of potential freshwater inputs is difficult to constrain, because it could also be related to an increase of precipitation water inputs as observed in surface waters during the period when the lake is stratified [e.g. Sarazin et al., 2021]. Although oceanic or freshwater inputs cannot be dismissed, it seems unrealistic to assume that this process alone could be responsible for all the modifications observed in the physical and chemical features in 2020 and 2021, but could at least in part explain the salinity decrease. Before the beginning of the seismo-volcanic crisis offshore Mayotte, several bubbling areas were identified at the water-air interface of the Dziani Dzaha Lake. The volcanic origin of these degassing into the atmosphere was confirmed by the CO 2 carbon isotopic composition, which was typically of magmatic origin [around −3 , Milesi et al., 2020, Liuzzo et al., 2021]. Yet, since a large part of these degassing was ebullitive, the contribution of this carbon source to the lake water column was difficult to estimate [Cadeau et al., 2020]. In addition, degassing areas were mainly located in the shallow parts of the lake (<2 m depth, Figure 1), with only one weakly bubbling area in the central part of the lake. An intensification of volcanic degassing through the Dziani Dzaha Lake water column related to the seismicity associated with the submarine volcano growth offshore Mayotte could impact both the oxygenation state and the pH of the water column. Indeed, an intense and permanent degassing in the central part of this lacustrine system, as observed during the surveys conducted in 2020 and 2021, could generate a convective plume that would significantly increase the mixing of the water column, favoring oxygenation events throughout the water column. Based on [O 2 ] data from two O 2 NKE SDOT optodes deployed at 0.25 m depth between September 10th and December 10th 2021 near the central plume, long-term water column oxygenation is highlighted (Figure 3). From 10th September to 7th December, day-night cycles of oxygenation are observed, which reflect the relative variations of photosynthesis and respiration activity in this highly productive lacustrine system [e.g. Leboulanger et al., 2017]. As the optodes are under the central plume influence, water comes from the bottom part of the lake (ca. 5 m depth at this point), which suggests the presence of dissolved oxygen at depth at least part of the day (O 2 decreases according to the respiration activity at night). Since magmatic gases do not contain O 2 , the only explanation for the presence of O 2 in the deeper part of the lake (at least at 5 m depth) would be a more active water column mixing (i.e. which was never observed before 2018) introducing O 2 from the surface waters to the deep waters. This increase of the mixing dynamics probably results from the water uplift induced by the more active central degassing in the absence of strong salinity gradient. After December 7th, these day-night cycles of oxygenation stopped, indicating a reduction of the water column mixing dynamics. We assume here that the water uplift due to the central degassing zone was not strong enough to fight the onset of the water column stratification. This is consistent with an increase in precipitation water inputs in late November and early December leading to a rise of the lake level (Figure 3), and the beginning of haline stratification. In addition to the water column mixing, this intense and long-term central 311 plume could also increase the dissolution of volcanic CO 2 in the lake waters, decreasing the pH. Although this process might explain both the dissolved oxygen content increase (i.e. physical mixing of the water column) and the pH decrease (i.e. dissolution of CO 2 ), the volcanic CO 2 contribution to the dissolved inorganic pool in the lake is difficult to constrain and it seems unrealistic to propose that the significant pH decrease is entirely related to it. In addition, pH value represents the balance of many processes, such as photosynthesis/respiration or sulfate reduction, which may also have changed.
Based on the three surveys conducted in 2020 and 2021, physical and chemical properties in the Dziani Dzaha Lake water column have already been impacted by the seismo-volcanic crisis offshore Mayotte. It is probably related to a combination of a dilution by precipitation water (explaining the salinity decrease), an infiltration of low salinity groundwater and an intensification of volcanic degassing (explaining both the oxygenation of the whole water column, through a mixing process, and at least a part of the pH decrease). These results provide a first assessment of the extent to which the Dziani Dzaha Lake functioning is impacted by the consequence of the birth of this submarine volcano.

The Carbon biogeochemical cycle
Since the beginning of the seismo-volcanic crisis offshore Mayotte, the three surveys conducted in the Dziani Dzaha Lake indicate: (i) similar but more variable DIC concentrations, and (ii) a significant depletion in 13 C since 2021 of about 3 (Figure 4). These results could be explained by (i) an input of a carbon source isotopically lighter than the DIC pool, or (ii) a modification in the respective influence of the processes that regulate the carbon cycle in the Dziani Dzaha Lake.
Two main parameters related to the seismovolcanic crisis offshore Mayotte need to be considered to discuss the potential impact of an isotopically lighter carbon source in the lake: the subsidence of Mayotte and the seismicity affecting these islands for 4 years now. This combination of parameters might favor both water input through the bedrock and an increase of volcanic CO 2 input in the lake. Based on the chemical and physical parameters (e.g. salinity, pH, [O 2 ]), no seawater input has been identified since the study of the Dziani Dzaha Lake in 2010 and until 2018. In contrast, volcanic CO 2 inputs have already been highlighted by previous studies [Cadeau et al., 2020, mostly thanks to the observation of several bubbling areas at the water/air interface. The contribution of volcanic CO 2 to the Dziani Dzaha Lake carbon cycle was difficult to constrain, and CO 2 dissolution in the water could be limited by a mainly ebullitive degassing through the water column. In addition, since 2018 the activity of the degassing area located in the central part of the lake has strongly increased, which must have increased water column mixing and CO 2 dissolution. These carbon sources present a significantly lighter δ 13 C value compared to the δ 13 C DIC values in the Dziani Dzaha Lake, i.e. close to 0 and −3 for seawater DIC and volcanic CO 2 , respectively [Cadeau et al., 2020]. Considering the strong modifications of the physico-chemical parameters observed in the lake since 2018 (e.g. [O 2 ], pH, Figure 2), both water inputs and increased CO 2 inputs through old fractures in the volcanic basement are strongly suspected. However, considering the very high DIC concentrations in the lake (i.e. about 100 times seawater concentrations), it is unlikely that these two sources alone are responsible for the 13 C depletion observed in September and December 2021. The impact of these sources can be tested using isotopic and mass balance equation as follows: X post-2020 × δ 13 C DIC-post-2020 = X input × δ 13 C DIC-input + X pre-2020 × δ 13 C DIC-pre-2020 . (2) With X representing the amount of DIC considering the concentrations and water volumes, and δ 13 C the calculated or measured carbon isotopic compositions of DIC. Based on the (2), considering the concentrations and isotopic signatures of the lake and ocean DIC (i.e. 200 mM and 12 for the lake, and 2 mM and 0 for the ocean), an input of seawater equivalent to 7.2% of the lake volume as mentioned previously would modify the lake δ 13 C DIC by less than 0.1 . The effect of an increase of volcanic CO 2 dissolution with a δ 13 C of −3 would have a similar impact.
As mentioned previously, water input or most likely an intensification of CO 2 volcanic input through the central plume have the ability to strongly and quickly impact the oxygenation state of the water column, which is a key parameter in the regulation of the carbon cycle by exercising a strong control on the organic matter degradation pathways.
Previous studies have shown that the strongly positive δ 13 C DIC in the Dziani Dzaha Lake was the consequence of organic matter degradation by methanogenesis associated with methane degassing into the atmosphere, which results from the complete sulfate consumption in an anoxic water column [Cadeau et al., 2020[Cadeau et al., , 2022. In contrast to other lacustrine systems in which methanogenesis has been suggested to explain 13 C enrichment observed in the sediment record or in the water column [Talbot and Kelts, 1986, Valero-Garcés et al., 1999, Gu et al., 2004, Zhu et al., 2013, Birgel et al., 2015, the Dziani Dzaha Lake exhibits similar carbon isotope signatures in the lake waters and in the sediment, as well as a long-term steady state [Cadeau et al., 2020]. Oxygenation of the water column necessarily modified the carbon cycle functioning by limiting the methanogenesis activity in favor of an aerobic degradation of the organic matter, and by fostering methane oxidation in the water column, hence limiting methane degassing into the atmosphere. Using the isotopic and mass balance equation described below, the impact of methane oxidation in the water column can be tested with the following equation: δ 13 C DIC-post-2020 = X CH 4 -pre-2020 × δ 13 C CH 4 -pre-2020 + X DIC-pre-2020 × δ 13 C DIC-pre-2020 . ( With X representing the amount of DIC or CH 4 considering the concentrations and water volumes, and δ 13 C the calculated or measured carbon isotopic compositions of DIC or CH 4 . Based on the (3) and considering a methane concentration close to 2 mM (i.e. hundred times lower than DIC which is of 200 mM) and its isotopic signature close to −65 , as observed in the water column before 2018 [e.g. Cadeau et al., 2020, methane oxidation would decrease the δ 13 C DIC by only 0.8 .
This illustrates that punctual and complete methane oxidation in the water column (i.e. simulating an extreme oxygenation episode) would have a limited impact on δ 13 C DIC .
Although such punctual perturbation has a limited impact on the δ 13 C DIC , a modification of the Dziani Dzaha Lake steady state related to seismovolcanic crisis perturbations could easily explain the δ 13 C DIC decrease. Based on the steady state box model proposed in Cadeau et al. [2020], a decrease in methane degassing into the atmosphere of about 10% of methane generated, a decrease of about 10% of the amount of organic matter degradation through methanogenesis, or a combination of both, would be sufficient to explain 3 decrease in δ 13 C DIC as observed in September and December 2021. The δ 13 C DIC significantly depleted in 13 C is thus most probably the consequence of the limitation in both organic matter degradation by methanogenesis and methane degassing to the atmosphere related to the oxygenation of the water column. Based on the three surveys conducted in 2020 and 2021, the carbon cycle in the Dziani Dzaha Lake seems to be significantly impacted by the seismovolcanic crisis offshore Mayotte most likely because of the intensification of volcanic CO 2 degassing into the lake, resulting in water column oxygenation and modifications of the organic matter degradation pathways within it. Future work is needed to determine the spatial extent and the persistence of these processes, as well as determine whether the carbon cycle is temporarily or permanently modified.

The nitrogen biogeochemical cycle
The particulate organic nitrogen isotope signatures (δ 15 N PON ) observed since 2018 are very similar to those observed before the beginning of the seismovolcanic crisis offshore Mayotte, i.e. close to 8 and constant with depth ( Figure 4). The only difference is the absence of significant isotopic increase below 14 m depth for the samples collected in September and December 2021.
In a previous study, we have suggested that the isotopic enrichment observed in the δ 15 N PON evolution in the deeper part of the water column or in the surface sediments results from a 15 N-enriched ammonium (NH + 4 ) assimilation, which itself results from NH + 4 dissociation to ammonia (NH 3 ) associated with a strong isotopic fractionation under basic and anoxic conditions [Cadeau et al., 2021]. As also shown for other modern basic lacustrine environments  and sedimentary records of past lacustrine environments [Stüeken et al., 2020[Stüeken et al., , 2019[Stüeken et al., , 2015, pH and oxygenation state are thus major parameters in the regulation of the nitrogen cycle in the Dziani Dzaha Lake.

313
The oxygenation state of the water column controls the speciation of dissolved nitrogen into ammonium versus nitrate. Water column oxygenation, as observed in September 2021, should prevent any NH + 4 accumulation by inducing its oxidation. In addition, even when the water column remains anoxic as observed in December 2021, the significant decrease in pH can be predicted to limit the dissociation of NH + 4 into NH 3 . Based on the average pH value of 8.5 and on the isotopic fractionation associated with the NH + 4 /NH 3 dissociation reaction [Li et al., 2012], only about 15% of the NH + 4 dissociates into NH 3 , resulting in an isotopic enrichment of the residual NH + 4 of only about 7 . These predicted amounts of dissociated NH + 4 and isotopic enrichment of residual NH + 4 are significantly lower than those estimated before the seismo-volcanic crisis event [i.e. about 45-72% of NH + 4 dissociated with an isotopic enrichment of residual NH + 4 comprised between 20 and 32 , Cadeau et al., 2021]. The perturbations related to the seismo-volcanic crisis 2018 should therefore have had a significant impact on the δ 15 N PON in the Dziani Dzaha Lake.
Surprisingly however, nitrogen isotope signatures are quite similar to those observed before the beginning of the ongoing crisis. The only difference is that in September and December 2021, δ 15 N PON values are constant over the whole water column, including the deeper part of the water column, while up until 2020 a marked isotopic increase was observed in the deep part of the water column. These constant nitrogen isotopic signatures observed in September and December 2021 are probably related to the assimilation of NH + 4 only slightly enriched in 15 N as a consequence of a less basic condition in the deeper and NH + 4 -rich part of the water column. We predict that this minor change in the δ 15 N PON pattern represents the premise of bigger changes, should the lower pH and the periods of oxygenation persist or aggravate in the future.

Conclusion
Since the beginning of the lake study in 2010 and until the beginning of the seismo-volcanic crisis in 2018, the Dziani Dzaha Lake presented an atypical combination of physical, chemical and biological features, as well as a high stability. Among them, the permanent anoxia of the water column and the pH exerted a strong control on the carbon and nitrogen cycles through: (i) an organic matter degradation by methanogenesis coupled to methane degassing into the atmosphere following the quantitative sulfate consumption in the water column, and (ii) a 15 N-enriched NH + 4 assimilation in the deeper part of the lake related to the dissociation of NH + 4 to NH 3 under basic conditions. Three years after the onset of the seismo-volcanic crisis, the submarine volcano growth and its associated perturbations are already having an impact on the biogeochemical functioning of the lake. Through water infiltration into the lake or/and an increase of volcanic CO 2 degassing into the lake, the water column oxygenation and the strong decrease of the pH value have: (i) modified the organic matter degradation pathways by limiting both the methanogenesis activity and methane degassing into the atmosphere, which is illustrated by a significant decrease in the δ 13 C DIC , and (ii) limited the NH + 4 dissociation into NH 3 and the isotopic enrichment of the residual NH + 4 , which is illustrated by the absence of δ 15 N increase with depth. These results constitute a first assessment of the present-day impact of the seismo-volcanic crisis on the lake. Although the perturbations affecting the Dziani Dzaha Lake are difficult to constrain now, studying this lacustrine system is an exciting opportunity to explore the resilience of these ecosystems to such geological disturbances.

Data availability
All the data generated and analyzed in this study are available in the paper.