Gas Leakage From Shallow Ponding Magma and Trapdoor Faulting at Sierra Negra Volcano (Isabela Island, Galápagos)

We report on new volcanic gas composition results acquired in October 2017 at Minas de Azufre, a persistent fumarolic field topping the resurgent Sierra Negra caldera, in the Galápagos archipelago. Our results indicate that the Minas de Azufre fumaroles are moderately hydrous (52–64 mol.% H2O) and rich in CO2 (35–46 mol.%), with total sulfur (ST) being 21–35 times less abundant than CO2. SO2, the most abundant S species, is released at an average rate of 19 ± 9 tons/day. Using a volatile saturation model that provides the composition of magmatic gases at equilibrium with western Galápagos basaltic melt (48 wt. % SiO2) in the 400–0.1 MPa pressure range, we infer that Minas de Azufre fumarolic emissions consist of a mixture of (a) magma‐derived gases coexisting with a melt at ∼50–60 MPa and (b) shallow meteoric water. We thus propose that the fumaroles are supplied by outgassing of magma stored in a ∼2 km deep sill‐like reservoir underneath the caldera floor, and that the trapdoor fault system at the western margin of the resurgent caldera block acts as a preferential pathway for magmatic gas ascent and surface discharge. Our results thus suggest that, in contrast to the majority of the volcano‐hosted hydrothermal systems worldwide, Minas de Azufre releases a relatively pristine magmatic gas.

and Ethiopian (Erta Ale: Sawyer et al., 2008;de Moor et al., 2013;Zelenski et al., 2013) hot spots. This paucity of data reflects the relatively infrequent eruptions and remote locations of many/most hot-spot volcanoes.
The western portion of the Galápagos archipelago, in the eastern Pacific (Figure 1), has recently been the theater of some of the most recurrent hot-spot related volcanic eruptions (Bell et al., 2021;Kurz et al., 2014;Vasconez et al., 2018). With its 10 major volcanic islands (and 21 subaerial Holocene volcanoes in total; Harpp et al., 2014), the Galápagos archipelago is thought to represent the surface expression of an upwelling mantle plume (Geist et al., 1988; see also Harpp & Geist, 2018 for a recent review) that intersects the Nazca plate ∼1,000 km off the west coast of Ecuador. Seismic imaging suggests that the plume originates from depths greater than 410 km (Hooft et al., 2003) and that its center is rooted today beneath Fernandina and Isabela islands (Villagómez et al., 2007), in the western part of the archipelago (Figure 1). These two western islands of Galápagos are thus relevant sites to investigate the chemistry and degassing of plume-sourced magmatic volatiles. Up to now, however, the chemistry of magmatic gases during an eruption on Fernandina or Isabela has not yet been measured, mostly because of the remoteness of the area and challenging access to eruption sites.
So far, the only accessible source of information on Galápagos volcanic gases has been the long-lived fumarolic field of Minas de Azufre (Colony & Nordlie, 1973). This persistent degassing area occurs within the summit caldera of the Sierra Negra shield volcano  that constitutes the southern part of Isabela island ( Figure 1). Giggenbach (1996) and Goff et al. (2000) were the first to report a high SO 2 /H 2 S ratio in fumarolic gas emissions at Minas de Azufre, which, in combination with detectable HCl, was taken as the signature of a shallow magmatic source (Goff et al., 2000). Based on oxygen/hydrogen isotopic compositions of the fumarolic steam Goff et al. (2000) and Taran et al. (2010) inferred that the magma-supplied volatiles were extensively mixed with shallow meteoric fluids upon ascent toward the surface. However, in the late 1990s, assessing the pristine (pre-mixing) composition of the magmatic end-member and its source depth beneath the fumarolic field remained hampered by limited knowledge of the structure of the Sierra Negra plumbing system , as well as by the lack of quantitative volatile saturation models allowing to predict the pressure-related composition of H-C-O-S fluid in equilibrium with western Galapagos magma.
Here, we present novel information for the chemical composition and emission rate of volcanic gases released by the Minas de Azufre fumarolic system, which we acquired during a field work conducted in early October 2017 in the framework of the thirteenth CCVG-IAVCEI gas workshop (https://ccvg.iavceivolcano.org/workshops/workshop-2011-2017.html). We take advantage of the improved geophysical knowledge of the subsurface structure of Sierra Negra summit caldera (Amelung et al., 2000;Bell et al., 2021;Chadwick et al., 2006;Jónsson et al., 2005), as well as of new volatile saturation models that incorporate the complex sulfur behavior in magmas (Moretti et al., 2003;Moretti & Papale, 2004), to propose that degassing at Minas de Azufre is sustained by magmatic gas leakage from magma ponding at ∼2 km depth beneath the summit caldera floor.

Sierra Negra Volcano
Sierra Negra is a large, frequently erupting basaltic shield volcano  that occupies the southern part of Isabela island (Figure 1). Sierra Negra is, after Fernandina further to the west (Figure 1), the historically most active volcano of the Galápagos hot-spot . Its two most recent eruptions occurred in 2005  and 2018 (Bell et al., 2021;Vasconez et al., 2018).
The subaerial (60 km long and 40 km wide) edifice of Sierra Negra has been modeled by protracted effusive activity over the last 7 kyr . The erupted volcanics are MgO-rich (∼4-∼7 wt %) basalts with transitional alkaline to tholeiitic affinity whose enriched Light Rare Earth Element (LREE) and incompatible trace-element compositions and radiogenic Sr and Pb signatures indicate their derivation from 5% to 15% partial melting of a garnet-bearing peridotitic plume-related source White et al., 1993). This enriched "mantle plume" affinity, further corroborated by noble gas systematics Kurz & Geist, 1999), contrasts with the MORB-like signatures of magmas erupted in the central and eastern segment of the archipelago (Gibson et al., 2012;Harpp & White, 2001). Upper Field). We distinguish field sub-segments (a and b) for both the LF and UF. The fumarolic field is located on a Trapdoor Fault System (TDF) at the margin of the sinuous ridge (Goff et al., 2000;. The top morphology of Sierra Negra volcano is characterized by a ENE-WSW elliptical (10 × 7 km) summit caldera that has recently undergone several episodes of collapse, upheaval, and deformation (Amelung et al., 2000;Bell et al., 2021;Geist et al., 2008;Jónsson et al., 2005). The caldera hosts in its center a 14 km-long C-shaped sinuous ridge (Figure 1), interpreted as the product of a long-lived caldera floor resurgence process . Deformation and resurgence are thought to be controlled by a Trapdoor Fault (TDF) mechanism, of which the sinuous ridge would be the surface expression, with the eastern caldera floor acting as hinge zone (Amelung et al., 2000;Jónsson et al., 2005). Uplift/subsidence of the TDF block are interpreted as caused by magma pressure change inside a flat, sill-like magma reservoir emplaced at only ∼2 km depth beneath the caldera floor (Bell et al., 2021;Chadwick et al., 2006).

Minas de Azufre
The Minas de Azufre fumarolic field is located on a steep scar at the western margin of the sinuous ridge ( Figure 1). A series of relatively narrow benches and "moats" in the fumarolic field itself and on the adjacent caldera floor supports that the Minas de Azufre fumarolic field stands right on the main faulted boundary of the resurgent block, thereby suggesting gas ascent and fumarolic discharge are controlled by the geometry of the TDF system (Goff et al., 2000). In their 1995 survey, Goff et al. (2000) reported the presence of two fumarolic clusters (max. temperature of 208°C) marking the edges of two closely spaced N-S trending faulted blocks. Additional measurements were taken during short-lived campaigns in 2004 (Taran et al., 2010) and 2006 (Padrón et al., 2012).

Data
Our gas investigations were realized on 6 October 2017 under clear sky and dry weather conditions. We used a portable Multi-Component Gas Analyzer System (Multi-GAS; Aiuppa et al., 2005;Shinohara, 2005) to perform a walking traverse through the Minas de Azufre fumarolic field along the path shown in Figure 1 (bottom panel). Two fumarole clusters, thereafter referred to as Lower Field (LF) and Upper Field (UF), were actively degassing during our survey (Figure 1), with respective discharge temperatures of 139 and 273°C (Hidalgo et al., 2014). Both fields were composed of several tens of fumarolic vents and, for simplicity, we subdivide each of them into 2 subcategories (a and b) based on their location ( Figure 1; Table 1). The fumarolic field in between UF and LF (Middle Field, MF; see Figure 1) was weakly fuming. Note. The molar ratios in the four subsegments of the Minas de Azufre fumarolic field (LFa, LFb, UFa, UFb) are derived from the gradients of the best-fit regression lines in the gas versus SO 2 scatter plots of Figure 2. Mean molar ratios for the fumarolic field are calculated by averaging the entire data set (all data) or the Upper Field (UF) results only (as this contributes most of the emission budget). Molar fractions are calculated assuming that H 2 O, CO 2 , SO 2 and H 2 S make up the totality of the gas phase, which is a reasonable approximation (see text). Gas fluxes are derived by scaling the average fumarolic compositions to the UV-Camera derived SO 2 flux of 19 tons/day.

Table 1 Gas Compositions and Fluxes
During our walking traverse the Multi-GAS was operated in a backpack and fumarolic gas effluents were pumped in (at 1.2 l/m rate) through the handheld inlet tubing positioned ∼50 cm above the vents. Each fumarolic vent was monitored for a few minutes, so that the entire field (LF-UF) was covered in ∼80 min. Our Multi-GAS device included a Gascard nondispersive infrared (NDIR) spectrometer from Edinburgh Sensors (for CO 2 ) and two specific electrochemical sensors (both from City Technology) for SO 2 and H 2 S (see Lages et al., 2020, for a description of the most recent Multi-GAS configuration, and for details on sensor's accuracy and repeatability, calibration ranges and gas standards). The system also measured temperature (T) and relative humidity (Rh; with a KVM3/5 Galltec-Mela sensor), which were converted into H 2 O concentrations using the Arden Buck equation (Tamburello, 2015). Data were acquired at 1 Hz synchronously from all sensors and stored on a Campbell CR6 datalogger. The obtained Multi-GAS data set is illustrated in Figure 2.
Synchronously to the Multi-GAS walking traverse, we operated a portable dual UV-camera system (described in Aiuppa et al., 2015), positioned at the base of the fumarolic field ( Figure 3a), to determine the SO 2 emission rate from the overall fumarolic field. The UV-camera, powered by a 12-V battery and commanded via a portable PC using the Vulcamera software (Tamburello et al., 2011), acquired sets of images at 0.5 Hz for ∼100 min. The instrument was equipped with two JAI CM-140GE-UV cameras sensible to UV radiation and fitted with two distinct bandpass filters centered at 310 nm (strong SO 2 absorption) and 330 nm (no SO 2 absorption) (compare panels B and B' in Figure 3). During post-processing, sets of co-acquired images (e.g., B and B') were combined (using Vulcamera) to obtain sequences of "absorbance" images (panel B' in Figure 3); these were converted into slant column amount (SCA) images using calibrations derived from calibration cells. Integration of SCA images along a cross section perpendicular to the plume transport direction yielded time series of Integrated Column Amounts (ICAs) which, multiplied by the plume speed, result in the SO 2 flux time series shown in Figure 3c. The plume speed and its temporal variations ( Figure 3c; uncertainty, ±5%) were derived by tracking the motion of plume gas fronts in image sequences, using an optical flow algorithm (see Delle Donne et al., 2019). Uncertainty in the resulting SO 2 flux is assessed at ±30%

Results
The Multi-GAS data set, illustrated in Figure 2, provides a snapshot of the spatial heterogeneity of volcanic gas composition throughout the Minas de Azufre fumarolic field. The scatterplots compare sets of co-acquired (a) H 2 S versus SO 2 , (b) CO 2 versus SO 2 and (c) H 2 O versus SO 2 concentration couples. Colors of the symbols identify measurements taken in distinct portions of the fumarolic field and thus respectively refer to subfields LFa, LFb, UFa and UFb (see Figure 1). For each of these subfields, strong correlations between SO 2 and other gas species are observed (Figure 2), implying relatively constant gas/SO 2 ratios. These ratios (listed in Table 1 and all given on a molar basis) are derived from the gradients of the best-fit regression lines shown in Figure 2 (with equations and regression coefficients). Gray dots identify measurements taken in the walking path between LF and UF (in the MF area; Figure 1), in which correlations between the different gas specifies are somewhat weaker. As degassing in the MF area was very sluggish, these measurements more likely represent distal and variable mixing of the plumes issuing from the UF and LF fumarolic fields, which were occasionally dispersed tens of meters away from the vent(s) by erratic change in wind direction and mixed in the background atmosphere. The gray dots consistently overlap the compositions of UF and LF fields. Our results confirm previous indications (Giggenbach, 1996;Goff et al., 2000;Padrón et al., 2012;Taran et al., 2010) that SO 2 is the prevalent sulfur gas species in Minas de Azufre fumaroles (Table 1). However, compared to these earlier studies based on direct fumarole sampling at only a few vents, the rapidity of Multi-GAS sensing allowed us to analyze several tens of degassing vents in sequence and hence to capture with great detail the spatial heterogeneity of H 2 S/SO 2 ratios in the field. We find significant compositional differences between the LF and UF fumarolic fields: compared to UF, the LF exhibits lower H 2 S/SO 2 ratios (0.11-0.12 vs. 0.29-0.31; Figure 2a), higher CO 2 /SO 2 (34-39 vs. 31-35; Figure 2b) and CO 2 /S T (S T = SO 2 + H 2 S) ratios (27-32 vs. 21-24; Table 1). LF also displays slightly higher H 2 O/SO 2 ratios (50-62) than UF (36-44; Figure 2c), while H 2 O/CO 2 ratios are similar at both fields (1.3-1.8 and 1.1-1.6, respectively; Table 1).
Gas molar fractions (Table 1) were calculated by assuming that H 2 O, CO 2 , SO 2 and H 2 S make up the totality of the gas phase (other species, such as N 2 and HCl, were found to occur at minor levels in Minas de Azufre fumaroles; Goff et al., 2000;Taran et al., 2010). Our results demonstrate a volcanic gas phase containing 52-64 mol% of H 2 O, 35%-56% of CO 2 , and much lower amounts of SO 2 (1.0%-1.4%) and H 2 S (0.1%-0.4%). Figure 3 shows a temporal record of the SO 2 flux, as derived from our ∼100 min-long UV-camera recording. The absorbance images displayed in the figure clearly identify the volcanic plume (indicated by green-yellow tones on blue background) gently lofting above the sinuous ridge's upper rim, and then dispersing to the right of the camera's field of view (e.g., southward). The plume speed (Figure 3c) was relatively constant at 0.9 ± 0.3 m/s. The computed SO 2 flux varies from 6.2 to 71 tons/day, with a mean at 19 ± 9 tons/day. Qualitative inspection of the images suggests that the SO 2 flux was mainly supplied by degassing from the UF.
The fluxes of other volatiles (H 2 O, CO 2 , and H 2 S) are quantified by combining the time-averaged SO 2 flux with the mean composition of the Minas de Azufre fumaroles (Table 1). We provide two separate estimates in Table 1 when using the average gas composition from either the entire field (LF + UF) or the UF only (the largest SO 2 flux source, see above). Under these two conditions we assess the H 2 O, CO 2 , and H 2 S fluxes at 219-263, 395-442 and 2-3 tons/day, respectively (Table 1).

Discussion
Owing to its remoteness, the Minas de Azufre fumarolic field has only rarely been targeted by volcanic gas investigations in the last decades. Our 2017 results indicate a fumarolic gas composition broadly consistent with previous reports in 1990 (Giggenbach, 1996), 1995 (Goff et al., 2000), 2004 (Taran et al., 2010) and 2006 (Padrón et al., 2012;Figure 4). More specifically, in a H 2 O-CO 2 -S T triangular diagram (Figure 4a) our data plot on the CO 2 -rich side of the compositional domain of within-plate/rift magmatic gases (Aiuppa, 2015) and overlap with the composition of the 2004 samples reported by Taran et al. (2010). In contrast, our results indicate a less hydrous gas phase than reported by Goff et al. (2000) and Padrón et al. (2012) (Figure 4a).  (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006). The compositional domains of within-plate/rift and arc magmatic gases are from Aiuppa (2015), while crosses are a selection of hydrothermal gases (Chiodini & Marini, 1998). Representative compositions of magmatic gases from intraplate/continental rift contexts are from Aiuppa (2015) Table 2), under redox conditions of either ΔNNO = −1.5 (model 1) and ΔNNO = −0.75 (model 2). Small circles (red/black) on the model lines identify model gas compositions at specific pressures (indicated in MPa). The red arrow indicates mixing between uprising magmatic gas separated from the magma at 60 MPa and a pure meteoric water component; (b) 0.5CO 2 -3SO 2 -5H 2 S triangular plot. The colored areas refer to the gas classification fields of Stix and de Moor (2018) except for the light green area labeled as "S loss, scrubbing or deep magmatic"?, that can refer to the compositional field of either (i) S-depleted, "scrubbed hydrothermal gases, or (ii) CO 2 -rich, S-poor deeply equilibrated magmatic gas (see text). Model lines 1 and 2 are as in (a). The dashed gray lines, referred as "H 2 S/SO 2 re-equilibration," describe the evolution of the uprising magmatic gas while its SO 2 /H 2 S ratio re-equilibrates (at constant CO 2 /S T and redox at ΔNNO = −0.75) during decompression from 50 to 60 MPa (magmatic sill) down to 0.1 MPa. Intraplate-continental rift magmatic gases (white circles) range from C-rich to C-poor, depending on the C-enriched or C-depleted signature of their mantle source (Aiuppa et al., 2021).
In the CO 2 -SO 2 -H 2 S classification diagram of Stix and de Moor (2018) our gas results in 2017 cluster, together with the 1990-1995-2004-2006 samples, in a relatively narrow area of CO 2 -rich and H 2 S-poor compositions (Figure 4b). This area, originally classified as "S loss, scrubbing" by Stix and de Moor (2018), would correspond to that of volcanic gas that experienced substantial subsurface loss of reactive sulfur species during hydrothermal reactions (e.g., reactions with meteoric/hydrothermal fluids and subsurface rocks) and consequently reached the surface with enhanced CO 2 /S T ratio. However, considering the high SO 2 /H 2 S ratio of Minas de Azufre fumaroles (Table 1), it is equally possible that the measured high CO 2 /S T ratios instead reflect a direct gas supply from magma stored at a depth where poorly soluble CO 2 is preferentially exsolved and degassed relatively to S (e.g., Aiuppa et al., 2007). This latter hypothesis of a deep magmatic signature, rather than the bearing of hydrothermal interactions, was privileged by Goff et al. (2000) but remained untested owing to the lack (at that time) of a quantitative C-O-H-S degassing model relevant to Sierra Negra magmas. In the following we attempt at resolving the magmatic versus hydrothermal nature of the Minas de Azufre fumarolic gas using a volatile saturation model.
Using a volatile saturation model requires independent knowledge of input variables such as the melt composition (major elements, volatiles, redox conditions), temperature and initial pressure. This information is typically derived from pre-eruptive magma storage conditions recorded in primitive melt inclusions (e.g., Aiuppa et al., 2007Aiuppa et al., , 2010Spilliaert et al., 2006). Unfortunately, melt inclusion information is not available neither for  Koleszar et al., 2009 andGleeson et al., 2022). The composition of Galápagos submarine glasses (all data from Peterson et al., 2017) are also shown by crosses (red, Fernandina group; green, Sierra Negra group; blue, eastern islands-Depleted Mantle group). All model calculations were made using the volatile saturation model of Moretti et al. (2003). Solid red line: Model 1; Solid black line: Model 2. Model run starting conditions are summarized in Table 2. (b) Same as (a) but comparing CO 2 versus S contents in model runs with melt inclusion data. (c) Model-predicted pressure dependence of magmatic gas concentrations (given in mass fractions) at equilibrium with the melt in the 400 to 0.1 MPa pressure range; (d) Same as (c) but for the molar gas ratios of the magmatic gas phase. Model 2 produces higher equilibrium SO 2 /H 2 S ratios, reflecting the more oxidising conditions considered in this run.
Sierra Negra volcano nor for any other volcano on Isabela island. We therefore initialise our model calculations using major element compositions and volatile contents measured in olivine-hosted melt inclusions from the nearby Fernandina volcanic island (Figures 5a and 5b; data from Koleszar et al., 2009). The validity of this indirect approach is justified by the following arguments: (a) the close proximity of Fernandina and Isabela islands, right above the centre of the Galápagos plume hot spot (Hooft et al., 2003;Villagómez et al., 2007); (b) the plume-related noble gas signature measured for both islands, in contrast to the MORB-like signature for the eastern islands (Kurz & Geist, 1999;Kurz et al., 2009;Taran et al., 2010); (c) the trace-element and radiogenic (Sr-Nd-Pb) isotope compositions of erupted volcanics from both islands which indicate a common plume-related mantle source component, enriched in incompatible trace elements (ITE) and radiogenic Sr (Gibson et al., 2012;Harpp & White, 2001;White et al., 1993), and (d) the similarity of H 2 O and S contents measured in submarine basaltic glasses dragged in the proximity of the two islands (Peterson et al., 2017; Figures 5a and 5b). Instead, a less enriched mantle source has been shown to prevail under the eastern part of the archipelago (Harpp & White, 2001;Peterson et al., 2017), as represented by the H 2 O-ITE-poor melt inclusions from Santiago island (Koleszar et al., 2009) Table 2. Model runs (1-2) were performed using as parental melt composition (major and volatile elements) the average composition of the two most primitive and undegassed melt inclusions from Fernandina island analysed by Koleszar et al. (2009). This parental melt has H 2 O (1.0 wt%) and S (0.15 wt%) contents (Figures 5a and 5b) that match the upper H 2 O and S concentrations measured in Sierra Negra submarine basaltic glasses (0.498-1.15 wt% and 0.089-0.159 wt%, respectively; Peterson et al., 2017), but is definitely richer in CO 2 (0.4 wt%) than the submarine basalts, erupted under a lower confining pressure. Runs were performed by simulating magma decompression from an initial pressure of either 400 MPa (Model 1) or 200 MPa (Model 2) down to atmospheric conditions (0.1 MPa), under isothermal (temperature kept constant at 1,300°K) and closed-system conditions (i.e., gas and melt continuously re-equilibrating during the decompression path). Equilibrium compositions of the melt and the coexisting gas phase were calculated at each step throughout the decompression path. In Model 1 redox conditions are fixed at 1.5 log units below the Nickel Nickel Oxide (NNO) buffer (ΔNNO = −1.5), in order to match the oxygen fugacity (∼10 −11.4 bar) for Fernandina magmas inferred by Koleszar et al. (2009) when modelling the sulphur solubility at sulfide saturation. In Model 2, we assume more oxidised (ΔNNO = −0.75) redox conditions in order to account for recent finding that mantle plumes are generally oxidised (Moussallam et al., 2019). These more oxidised conditions are also more consistent with the high SO 2 /H 2 S signature of Minas de Azufre fumaroles (see below) and with Fe oxidation state measurements that point to redox conditions ranging between ΔNNO = −0.6 and ΔNNO = 0 in glass samples from the Galapagos Archipelago (Peterson et al., 2015). Figure 5. Figures 5a and 5b show that the modelled melt-volatile compositions reproduce quite well the range of volatile contents measured in Fernandina's melt inclusions analysed by Koleszar et al. (2009), especially for CO 2 and S (Figure 5b). In Figure 5a the trends from both Models 1 and 2 plot towards the high H 2 O range of the melt inclusion population, suggesting that many of the melt inclusions may have trapped dehydrated melt flushed by deeper sourced CO 2 -rich bubbles (Caricchi et al., 2018). The modelled equilibrium gas compositions (Figures 5c and 5d) are similar in both model runs (1 and 2) and highlight a CO 2 -rich gas phase prevailing at high to moderate pressure and evolving to H 2 O-richer (>50 wt%) "" −0.75 −10.6 "" "" "" "" "" "" "" "" "" "" "" ""

Results of our model calculations are illustrated in
Note. Parental melt composition (major and volatile elements) is represented by the most primitive olivine-hosted melt inclusions on Fernandina island (Koleszar et al., 2009) (Figure 5d). Modelled CO 2 /S T and H 2 O/CO 2 ratios are negatively correlated (Figures 5d and 6) and evolve quite similarly in the two model runs (Model 2 resulting in slightly higher H 2 O/CO 2 ratio).

Origin of the Minas de Azufre Gas
Comparing our results with the modeled equilibrium gas compositions during magma decompression, as attempted in Figures 4 and 6, offers quantitative clues on the origin of the Minas de Azufre fumarolic gases. We find that the measured gas compositions (this study and previous ones) plot to the right (toward more H 2 O-rich compositions) of the modeled degassing trends. This observation agrees with H-O isotopic evidence that most of the fumarolic steam is non-magmatic and rather derives from meteoric water vaporised in the subsurface (Goff et al., 2000;Taran et al., 2010).
Instead, two lines of evidence support a magmatic origin of both CO 2 and S. First, the high prevalence of SO 2 over H 2 S in the fumaroles (Figure 4b) points to little sulfur scrubbing during hydrothermal reactions in the subsurface since SO 2 is much more soluble than H 2 S in liquid water and would thus be preferentially removed through such reactions (e.g., Stix & de Moor, 2018). Instead, the measured mean H 2 S/SO 2 ratio of 0.2-0.3 (Table 1), corresponding to SO 2 /H 2 S of 3-5, is identical to the ratio of the modeled equilibrium magmatic gas phase at near surface conditions (∼0.2 at 0.1 MPa pressure; Figure 4b) for ΔNNO = −0.75 (Model 2). Second, the measured CO 2 /S T ratios of Minas de Azufre fumaroles are relatively constant irrespective of H 2 O/CO 2 variations ( Figure 6). This strongly suggests no or limited S scrubbing while the rising magmatic gas interacts with infiltrating meteoric water; otherwise, CO 2 /S T would decrease as H 2 O/CO 2 increases.
Hence, we propose that the CO 2 /S T ratios of the fumarolic gas are scarcely (if any) affected by hydrothermal S processing, and may rather reflect the conditions of gas separation from the magma body underlying the caldera floor. According to the modeled pressure dependence of CO 2 /S T ratio in the gas phase at equilibrium with decompressing magma (Figure 5d), the mean CO 2 /S T ratio of 28 for the whole fumarolic field (Table 1) constrains pressures of 79 ± 14 (Model 1) and 63 ± 10 MPa (Model 2) for the ultimate gas-melt equilibrium prior to gas separation from the magma body. At that point we favor Model 2 as being more representative of the magma conditions at Sierra Negra since its more oxidized redox conditions well reproduce the SO 2 -rich signature of the Minas de Azufre fumarolic gas, whereas Model 1 (ΔNNO = −1.5) does not (Figure 5b). In this framework, the CO 2 /S T ratio of 23 for the UF (Table 1), the portion of the fumarolic field that contributes the main fraction of the gas output (Figure 3), would yield a slightly lower degassing pressure of 51 ± 2 MPa. In summary, we propose that the Minas de Azufre fumarolic gas is best interpreted as being supplied by the outgassing of magma stored at about 50-60 MPa pressure. Such a magma/gas separation pressure corresponds to a source depth of ∼1.9-2.2 km. We emphasize that this independent estimate nicely fits with the ∼2 km depth of the sill-like magma reservoir underlying the caldera floor as inferred from both geodetic and petrological data (Bell et al., 2021 and references therein).
Our model calculations (Model 2) show that, at 60 MPa pressure, the equilibrium magmatic gas phase would consist of ∼78% CO 2 , ∼18% H 2 O and ∼4% S T . Comparing with the "mean" composition of Minas de Azufre fumaroles (40% CO 2 , 58% H 2 O and 1.2% S T ; Table 1), simple mass balance calculation implies that the fumarolic fluid in 2017 can well be explained by the shallow mixing of magmatic gas upraising from 60 MPa with meteoric water in respective mass proportions of 75%-60% and 25%-40% ( Figure 6). The more hydrous compositions measured in 1995 (Goff et al., 2000) and 2006 (Padrón et al., 2012) imply greater mixing proportions of meteoric water (from ∼50% to >80%; Figure 6). An even higher extent of meteoric dilution was inferred from the stable isotope ratios of fumarolic steam in 2006 (Taran et al., 2010).   (Taran et al., 2010), whilst it may have contributed to the higher CO 2 /S T ratios measured in H 2 O-enriched gas samples in 1995 (Goff et al., 2000(Pádron et al., 2012.

UF Versus LF Chemical Diversity
The UF and LF fumarolic fields slightly differ in their H 2 S/SO 2 and CO 2 /S T ratios (Table 1; Figures 2, 4, and 6).
Because their H 2 O/CO 2 ratios are indistinguishable (Table 1), these compositional differences cannot be explained in terms of distinct magmatic versus meteoric mixing proportions (see Figure 6). Instead, the two fumarolic fields differ in their spatial position and maximum emission temperature, 139°C at LF and 273°C at UF as measured in 2017. As its name actually means, the Minas de Azufre fumarolic field is characterized by abundant deposits of elemental sulfur that have been economically exploited (Colony & Nordlie, 1973). The lower H 2 S/SO 2 ratios and somewhat higher CO 2 /S T ratios at LF may thus reflect partial sulfur precipitation favored by the lower exit gas temperature in this field. Elemental sulfur deposition/remobilization occurs via the reaction (Giggenbach, 1987): which consumes twice more H 2 S than SO 2 and can thus modify significantly the H 2 S/SO 2 ratio of the residual gas. In Figure 7 we tested the saturation state of Minas de Azufre fumaroles with respect to the deposition of elemental sulfur, using thermodynamic data from Giggenbach (1987) for the equilibrium constant of reaction (1). We find that while the hotter UF fumaroles plot in the undersaturated field, the LF fumaroles do plot to the left of the saturation line at 0.1 MPa, implying that elemental sulfur is actively deposited at the discharge conditions, in agreement with field observations. Therefore, we propose that near-surface elemental sulfur deposition controls the (minor) LF versus UF compositional differences in terms of H 2 S/SO 2 and CO 2 /S T ratios (Table 1). Enhanced sulfur precipitation from LF fumarolic gases, together with their lower discharge temperatures, can be explained by their more peripheral location relative to the main gas upstream (forced by the morphology of the TDF system; Figure 8, see below), favoring more effective gas cooling and air dilution.

Magma Budget
If, as our data suggest, the bulk SO 2 emissions from Minas de Azufre are magma-derived and are not or little affected by hydrothermal scrubbing (except for some minor elemental sulfur deposition at LF, see above), then the UV-camera based SO 2 flux can be used to quantify the rate of magma supply to and degassing from the shallow (∼2 km deep) Sierra Negra magma reservoir. In our calculations, we consider a magma density of 2750 kg/m 3 and a 5% crystal content . We assume that the magma supplying the sill contains 0.15 wt.% S (Table 2). At the reservoir conditions (50-60 MPa), our model calculations imply ∼0.14 wt. % S is still dissolved in the melt (Figure 5b). Hence, ∼100 mg of S can be outgassed per kg of magma supplied to the sill-like reservoir. From these numbers, we infer that a magma supply (degassing) rate of 3.6·10 4 m 3 /day (or 0.4 m 3 /s) is required to account for the SO 2 output of 19 tons/day from Minas de Azufre (Table 1).
One can safely consider the SO 2 flux of 19 tons/day measured in 2017 as representative for the long-term degassing regime at

Conceptual Model and Conclusions
Our observations and models are synthetized in the interpretative conceptual model of Figure 8. We propose that the Minas de Azufre fumarolic field at Sierra Negra volcano is supplied by magmatic gas leakage from the ∼2 km deep magmatic sill identified beneath the caldera floor (Bell et al., 2021). According to our calculations, deeper derived magma is supplied at an average rate of ∼0.4 m 3 /s to the sill during volcano quiescence. This deep magma supply leads to gas bubble accumulation in the shallow sill during inter-eruptive periods, as indicated by gravity changes (Vigouroux et al., 2008). Seismic and geodetic evidence (Bell et al., 2021) show that both magma and gas progressively accumulate in the sill, leading to gradual pressure build-up, caldera inflation, and ultimately dyke emplacement shortly prior and during eruption. However, as our novel data highlight, magmatic gas leakage from the sill persists during volcano quiescence, sustaining the fumarolic gas discharge at Minas de Azufre. This implies that the TDF system, located on the western side of the sinuous ridge at the margin of the resurgent block (Figure 8), remains open enough to act as a preferential pathway for the magmatic gas leakage feeding the fumarolic field. We estimate that, along with ∼19 tons of SO 2 , ∼395-442 tons of magmatic CO 2 and ∼65 tons of magmatic H 2 O (calculated from the H 2 O/S mass ratio of ∼3.5 in the magmatic gas phase at 50-60 MPa) are daily transferred from the sill to the fumarolic system. Since the measured H 2 O flux from Minas de Azufre reaches ∼220-265 tons/day, we infer that ∼150-200 tons/day of vapourized meteoric water get admixed to the rising magmatic gas, which easily accounts for the prevalent meteoric isotopic composition of the emitted steam (Goff et al., 2000).
One corollary of our interpretation that the Minas de Azufre fumaroles are sourced by degassing of the shallow magmatic sill emplaced beneath Sierra Negra caldera is that the H 2 S/SO 2 ratio of the magmatic gas fully  Bell et al., 2021) and our interpretative framework of magmatic gas leakage through the western trapdoor fault (TDF), variably diluted by meteoric water, that controls the chemical composition and mass output of volcanic gas at Minas de Azufre fumarolic field. The rate of magma replenishment of the sill-like reservoir is inferred from the mean SO 2 output we measured in 2017 and the mass fraction of exsolved sulfur available in this reservoir (at 50-60 MPa pressure). See the text for discussion.
re-equilibrates along its ascent from the 50-60 MPa initial pressure to atmospheric discharge at 0.1 MPa (Figure 4b). As a matter of fact, the emitted gases have H 2 S/SO 2 ratios that closely match those calculated at 0.1 MPa (model 2), whereas the modeled gas phase equilibrated at 50-60 MPa is richer in H 2 S (see the "H 2 S/SO 2 re-equilibration lines" in Figure 4b). In the near surface, and especially at the colder LF vents, the H 2 S/SO 2 ratio can additionally be partially altered by elemental sulfur deposition (Figure 7).
The magmatic CO 2 /S T signature of the Minas de Azufre volcanic gas that we bring to light is relatively unique. Indeed, the majority of caldera-hosted fumarolic systems (e.g., Campi Flegrei; Chiodini et al., 2016;Yellowstone;Bergfeld et al., 2011;Hurwitz & Lowenstern, 2014) tend to exhibit S-depleted hydrothermal gas signatures, owing to extensive loss of magmatic S through scrubbing and deposition in the subsurface. We note that our gas measurements in October 2017 were performed only ∼9 months before the onset of the June-August 2018 eruption at Sierra Negra, when both seismicity and ground deformation were entering an escalating phase (Bell et al., 2021).Therefore, it is possible that the heightened seismicity/ground uplift contributed to more efficient TDF-mediated transport of magmatic volatiles from the sill at that time. Instead, the comparatively higher H 2 O content of Minas de Azufre fumaroles in 1995 (Goff et al., 2000) and 2006 (Padrón et al., 2012;see Figures 4 and 6) suggests enhanced meteoric dilution of the upraising magmatic gas, and thus enhanced possibility of its hydrothermal alteration, during the post-eruptive phases or quiescent periods. In these phase more remote from eruptions, the reduced magmatic gas supply is supported by ground subsidence . However, the relatively steady "magmatic" CO 2 /S T ratio observed by Taran et al. (2010) in 2004 ( Figure 6), one year before the October 2005 eruption , implies that the gas conditions encountered in October 2017 are not exceptional. We conclude that the unusually shallow depth (∼2 km) of magma storage beneath the Sierra Negra caldera provides favorable conditions for a rapid transit of magmatic gases to the surface, thereby preventing their hydrothermal alteration. We therefore suggest that implementing a continuous gas monitoring at Minas de Azufre could be highly beneficial to improve volcanic hazard assessment and eruption forecasting at Sierra Negra volcano.

Data Availability Statement
All processed data generated or analysed during this study are included in Tables 1 and 2. The raw Multi-GAS data set (used to generate Figure 2) and the processed UV Camera data set (used to generate Figure 3) are also publicly available in the EarthChem data library (ECL) repository (https://www.earthchem.org/ecl/) (Aiuppa et al., 2022). The CO 2 -H 2 O-SO 2 -H 2 S-melt saturation and degassing code (Moretti et al., 2003;Moretti & Papale, 2004) is publicly available for download at https://github.com/charlesll//chosetto (doi: https://doi. org/10.5281/zenodo.5554941). The CHOSETTO.exe program (for Windows®) is available together with instructions to run it and relevant information on licensing and credits to be given.