New insights into the magmatic-hydrothermal system and volatile budget of Lastarria volcano, Chile: Integrated results from the 2014 IAVCEI CCVG 12^th Volcanic Gas Workshop

The chemical and isotopic composition of volcanic gases can provide insights into subvolcanic conditions and specifically can be used to identify the presence of shallow magma and/or hydrothermal aquifers. These data can complement and refine geophysical models whose methods alone are unable to distinguish different types of crustal fluids responsible for seismic tomography anomalies and sources of crustal inflation. Discrete volcanic gas measurements collected at Lastarria volcano (northern Chile) found evidence for a well-developed hydrothermal system between May 2006 and June 2009 from direct fumarole samples (Aguilera et al., 2012), and shallow magma with minimal hydrothermal influence in November 2012 from in situ plume sampling techniques (Tamburello et al., 2014). These apparently discrepant findings could be attributed to: (1) spatial variability among the individual sampled fumaroles and the bulk plume composition; (2) mixing, dilution, and chemical processing (e.g., condensation of steam) of the volcanic gases with ambient air for the in situ measurements (Tamburello et al., 2014); (3) short-term excursions from baseline gas compositions due to localized changes in stress; and/or (4) a change in the volcanic-hydrothermal system between the two measurements periods, which can be attributed to both exogenous (e.g., precipitation, weather) and endogenous (e.g., volcanic, stress) changes (Zimmer et al., 2017). In November 2014, the International Association of Volcanology and Chemistry of the Earth’s Interior (IAVCEI) Commission on the Chemistry of Volcanic Gases (CCVG) hosted its 12^th volcanic gas workshop in Chile and conducted a two-day comprehensive volcanic gas sampling campaign at Lastarria volcano. The aims of these workshops are to test, compare, validate, and refine new and existing measurement and analysis techniques to accurately quantify and characterize volcanic degassing. The results from previous workshops were summarized by Giggenbach and Matsuo (1991) and Giggenbach et al. (2001). Here we present integrated results from this community-wide effort to provide a synoptic characterization of Lastarria’s volcanic gas emissions. Specifically, in this manuscript we: (1) integrate and compare gas composition and flux measurements collected using a variety of gas measurement techniques; (2) use our volcanic gas measurement results to confirm a change in Lastarria’s magmatic-hydrothermal system; and (3) calculate an updated total degassing budget using a combination of remote sensing, diffuse degassing, and in situ gas composition measurements.

Lastarria volcano (25.168°S, 68.507°W, 5706 m elevation) is located in South America’s Central Volcanic Zone on the Chilean-Argentinian border, ∼250 km southeast of Antofagasta, Chile (Fig. 1). It is part of the larger Lastarria–Cordón del Azufre complex, known colloquially as Lazufre. Within Lazufre, the Lastarria volcanic complex comprises three different volcanic structures including: (1) Negriales lava field, constituted by an andesitic-to-dacitic lava flow succession (400 ± 60 to 116 ± 26 ka); (2) Espolón Sur, consisting of andesitic lava flows (150 ± 50 ka); and (3) Lastarria sensu stricto (herein referred to as Lastarria volcano), a compound stratovolcano formed by successions of andesitic lava flows and domes, andesitic pyroclastic flows, and avalanche deposits, with ages ranging from 260 ± 20 to <2.5 ka (Naranjo, 2010; Fig. 2). Lastarria volcano contains five overlapping summit craters and four actively degassing fumarole fields, two of which were measured in this study (Fig. 2). No historical eruptions from Lastarria volcano have been reported or observed, with its recent volcanic activity limited to persistent degassing, which was first observed in the early 19^th century (González-Ferrán, 1995). However, Naranjo (2010) found evidence for profuse sulfur fumarolic activity prior to its 2.5 ka eruption, which continued after the eruption, forming the current four fumarolic fields and extensive sulfur flows.

Lastarria’s edifice has been built on a basement of Pliocene- to Pleistocene-aged lavas from a diverse set of stratovolcanoes and lower Pleistocene dacitic ignimbrites (Naranjo and Puig, 1984; Naranjo and Cornejo, 1992; Naranjo, 2010). The continental crust in this region is estimated to be ∼60–70 km thick (James, 1971). The earthquake foci defining the Benioff zone and indicating the top of the subducting slab are located ∼140 km beneath the volcanic complex (Barazangi and Isacks, 1976, 1979). These boundaries are used to estimate the mantle wedge thickness beneath Lastarria’s crust at ∼70–80 km.

Large-scale crustal inflation centered between Lastarria and Cordón del Azufre was first detected by interferometric synthetic-aperture radar (InSAR) data beginning in ca. 1997–1998 (Pritchard and Simons, 2002) and has made this region the target of international multidisciplinary research (Pritchard et al., 2018). Following these initial findings, Froger et al. (2007) analyzed InSAR data collected between 2003 and 2005 and identified two inflation sources beneath Lastarria, including a deeper source located at 7–15 km depth and a shallow source ∼1 km below the summit. They speculated that both sources were related, with the deeper source representing magmatic intrusion and the shallow source representing pressurization of a hydrothermal system (Froger et al., 2007). Ruch et al. (2009) similarly observed a shallow inflation source at a depth of ∼1 km between 2000 and 2008. Continued inflation of the 7–15 km source has been observed through 2010 at an average rate of 3 cm/yr by Pearse and Lundgren (2013) and at a reduced rate of <1.5 cm/yr between 2011 and 2016 by Henderson et al. (2017). Updated geodetic modeling has redefined the deep inflation source as a shallowly dipping ∼20 km × 30 km sill centered at 8 km depth (Pearse and Lundgren, 2013). No new deformation studies able to discriminate changes in the shallow inflation source have been conducted, such that it is currently not known if the shallow source has continued to inflate since 2008.

Aguilera et al. (2012) collected volcanic gas samples from Lastarria’s main fumarole fields during campaigns in May 2006, March 2008, April 2009, and June 2009. Aguilera et al. (2012) then used the chemical composition and temperature of the volcanic gases, along with geochemical tools including geothermometers and isotopic tracers, to infer the presence of crustal magma degassing and a discontinuous shallow hydrothermal system. The geochemical interpretations of Aguilera et al. (2012) are consistent with the InSAR interpretations by Froger et al. (2007) indicating two deformation sources: a relatively deep magmatic source that is supplying the high-temperature magmatic gases, and a shallow boiling hydrothermal system within Lastarria’s edifice that is supplying the low-temperature gas emissions. More recently, Spica et al. (2015) used high-resolution seismic tomography from data acquired during campaigns in 2008 and 2012 to identify two shallow low-velocity zones located at depths of <1 km and between 3 and 6 km centered beneath Lastarria’s edifice. Spica et al. (2015) interpreted these low-velocity zones to represent the previously proposed shallow hydrothermal system and a potentially fluid-rich region overlying the ∼8-km-depth magma source.

In November 2012, in situ plume composition and remote SO₂ flux measurements collected by Tamburello et al. (2014) indicated high fluxes of SO₂ and HCl gases consistent with shallow magma degassing and minimal hydrothermal scrubbing, in contrast to previous observations by Aguilera et al. (2012). Then between 2013 and 2015, Zimmer et al. (2017) collected continuous temperature, precipitation, and CO₂ measurements from three of Lastarria’s high-temperature fumaroles to evaluate the influence of external processes on fumarole output. They found that major precipitation events caused significant temperature decreases at individual fumaroles over time scales of hours to days, but in most cases the fumarole then returned to its baseline temperature. They also found that changes in barometric pressure did not correlate with CO₂ output from high-temperature fumaroles.

Lastly, Carn et al. (2017) used temporal and spatial averaging techniques with Ozone Monitoring Instrument (OMI) satellite observations of SO₂ to calculate annual SO₂ emissions from Lastarria. These emissions ranged from 59 to 134 kt/yr between 2005 and 2015, and while some temporal variability in degassing is evident, the only temporal trend observed is a slight increase in SO₂ flux between 2010 and 2015.

During the CCVG gas workshop, participants conducted a comprehensive volcanic gas sampling campaign at Lastarria volcano on 21 and 22 November 2014. These measurements were collected during optimal (clear sky) weather conditions within the dry season, with no major precipitation events having occurred in the previous weeks (Zimmer et al., 2017). The main objectives were to: (1) collect direct fumarole samples and in situ gas composition measurements from the summit and flank fumarole fields; and (2) measure volcanic outgassing of SO₂ and CO₂ using ground-based remote sensing and diffuse soil degassing techniques, respectively. Direct fumarole sampling and in situ plume measurements targeted fumaroles located in fields 1 and 3 as defined by Aguilera et al. (2012) (Figs. 2, 3), and referred to throughout this document as the lower (field 1) and upper (field 3) fields, respectively. The selected fumaroles correspond with the most representative high-temperature and high-flux fumaroles of the four fumarolic fields, and have exhibited notably stable temperatures and fluxes over time, based on previous work by Aguilera et al. (2012). Additionally, the lower and upper field fumaroles are the main contributors to the bulk plume and for that reason were targeted in this study. Diffuse soil degassing targeted the region around the lower field (Fig. 3). SO₂ camera measurements were collected north of the edifice to obtain a perpendicular view of the plume, while differential optical absorption spectroscopy (DOAS; Platt and Stutz, 2008) traverse measurements were collected downwind (east) of the edifice underneath the plume (Fig. 3).

Direct fumarole samples for analysis of chemical and/or isotopic composition were collected by various teams including: University of New Mexico (USA; UNM); University of Florence (Italy; UNIFI); Istituto Nazionale di Geofisica e Vulcanologia Palermo (Italy; INGV-Palermo); Observatorio Vulcanológico y Sismológico de Costa Rica–Universidad Nacional (Costa Rica; UNA); French Atomic Energy Commission–Saclay (France; CEA); and McGill University (Canada; MU). A single fumarole from each field was targeted such that direct comparison of results from multiple groups would be possible. At each target fumarole, a 1-m-long, 2.5-cm-diameter titanium tube was inserted into the fumarolic vent and connected either through quartz Dewar tubes or through a glass adapter and Tygone and Silcone tubing to the sampling flasks. Traditional gas sampling techniques using pre-weighed evacuated bottles containing a caustic 4–5 N NaOH solution connected directly to the sampling line were used to quantify the major and trace element composition of fumarolic gases including H₂O, CO₂, SO₂, H₂S, HCl, HF, N₂, O₂, Ar, H₂, CH₄, CO, and He following the methods of Giggenbach and Goguel (1989). Additionally, the UNIFI group’s bottles also contained 0.15 M Cd(OH)₂ to directly precipitate H₂S as CdS, enabling separate analyses of H₂S and SO₂ as described by Montegrossi et al. (2001). A total of six and three Giggenbach-style samples were collected from the lower and upper field fumaroles, respectively. With the exception of samples collected by INGV-Palermo on 22 November, all other Giggenbach samples were collected on 21 November. The MU group collected two fumarolic gas samples from the lower field on 21 November and four from the upper field on 22 November by syringe from the end of the sampling line and immediately transferred the sampled gas into evacuated bottles containing HgCl₂ (to prevent bacterial growth) for later analysis of CO₂ concentration and carbon isotope composition. Additionally, the INGV-Palermo group collected six dry gas samples from the lower field on 22 November by connecting a condensing system to the sampling line and allowing the gas to pass through the system prior to collection in two-stopcock bottles. The dry gas was sampled specifically to quantify He, H₂, O₂, N₂, CO, CH_4, and CO₂, as well as carbon isotope composition. Lastly, the group from CEA employed copper tubes connected to the sampling line to collect three dry-gas fumarolic gas samples from the lower field on 21 November for helium isotope analysis following the methods of Jean-Baptiste et al. (1992). Details of the various analytical methods employed are described in the following paragraphs.

Evacuated bottles containing absorbing (alkaline) solution were analyzed by each group in laboratories at UNM, UNIFI, UNA, and INGV-Palermo primarily following the methods of Giggenbach and Goguel (1989). Non-condensed (i.e., head-space) gases (e.g., N₂, O₂, CO, H₂, He, Ar, Ne, CH₄, and light hydrocarbons) were analyzed by gas chromatography. Absorbed gases including CO₂, total S (S_t = SO₂ + H₂S), HCl, and HF were quantified after oxidation and neutralization as dissolved SO₄^2–, Cl^–, and F^– by ion chromatography. Groups from UNM and UNIFI additionally quantified both SO₂ and H₂S in the alkaline solution following the methods of Giggenbach and Goguel (1989) and Vaselli et al. (2006), respectively. The H₂O content was determined by all groups by weighing the bottle before and after sampling, taking into account the amount of CO₂ and acidic species absorbed in alkaline solution. More details of the instruments and methods employed by individual groups can be found in Mitchell et al. (2010) and de Moor et al. (2013) (UNM and UNA), Vaselli et al. (2006) (UNIFI), and Sortino et al. (1991) (INGV-Palermo). The analytical uncertainties for acidimetric titration, gas chromatography, and ion chromatography are estimated at <10% for all groups.

All fumarolic measurements of ¹³C/¹²C are presented in comparison to the reference ¹³C/¹²C composition of Vienna Peedee belemnite (V-PDB; ¹³C/¹²C = 11,180.2 × 10^–6; Chang and Li, 1990) using standard delta notation. INGV-Palermo dry gas samples were analyzed by a Perkin Elmer Clarus 500 gas chromatograph equipped with a 3.5 m Carboxen 1000 column and double detector (hot-wire detector and flame ionization detector) with analytical errors <3%, while the carbon-isotope composition of CO₂ was measured using a dual-inlet mass spectrometer (Delta Plus, Finnigan), with an analytical uncertainty of ∼0.1‰. The MU samples were analyzed in the geochemistry laboratory at MU in Montreal, Canada, on a Picarro G1101-i cavity ring-down spectrometer (CRDS). The standard deviation of the isotopic CO₂ values is <0.5‰. Additional details of the methods employed can be found in Lucic et al. (2015) and Malowany et al. (2015).

CEA samples for simultaneous helium and neon isotope determination were analyzed at the Saclay Nobel Gas Facility using a MAP215-50 mass spectrometer (Jean-Baptiste et al., 1992, 2010). The measured ³He/⁴He ratios (R) are presented in comparison with the same ratio in air (R_A = 1.39 × 10^–6; Mamyrin and Tolstikhin, 1984) as R/R_A. Typical ⁴He and ²⁰Ne blanks were 5 × 10^–10 and 3 × 10^–10 cm³, respectively (standard temperature and pressure), and analytical uncertainties on R/R_A and helium mixing ratios were ±0.05 and ±0.1 ppmv, respectively.

Two types of in situ plume measurements were carried out by three groups at the upper and lower fumarole fields on 21 and 22 November including continuous, high-temporal-resolution major species gas composition (H₂O, CO₂, SO₂, H₂S) measurements using a multi-component gas analyzer system (MultiGAS), and average acid gas composition (S_t, HCl, HF, HBr) sampled over a discrete time period using alkaline trap methods.

MultiGAS measurements of major species plume composition were collected from the lower fumarole field by a group from the University of Palermo (Italy; UNIPA). The MultiGAS instrument contains a LICOR LI-840A infrared analyzer for simultaneously quantifying CO₂ and H₂O; and City Technology electrochemical sensors 3ST/F to quantify SO₂, EZ3H to quantify H₂S, and EZT3HYT to quantify H₂. Measurements were collected for ∼2 h between 12:24 and 14:33 UTC on 22 November 2014 in “mobile” mode by UNIPA. During this time period, three transects were conducted while walking through the fumarole field. Molar gas ratios for the full data set were calculated by conducting linear interpolations of each measured gas species with respect to SO₂ over the sample period using RatioCalc software (Tamburello, 2015). No barometric pressure corrections were applied to the SO₂ and H₂S data due to the ineffectiveness of these corrections at high altitudes (Peter Kelly, 2017, personal commun.).

Two styles of alkaline trap sampling techniques were employed, both of which involve pulling plume air through an alkaline solution to absorb acid gases. Samples were collected on 21 November from the upper field and on 22 November from the lower field. The group from INGV-Palermo used Drechsel bottles containing ∼100 ml of alkaline solution (1 M NaOH) and a pump to pull the volcanic gas–atmospheric mixture through the solution at a 1 L/min flow rate (e.g., Roberts and McKee, 1959; Liotta et al., 2012; Rizzo et al., 2013; Wittmer et al., 2014). This method enabled quantification of S_t and HCl. This sampling system was installed downwind of fumarolic gases in an area that appeared free of soil gas emissions. A group from the University of Heidelberg (UH) used a Raschig tube (e.g., Levin et al., 1980)–style alkaline trap made in house (see Wittmer et al., 2014), which was filled with 50 ml of a 1 M NaOH solution to enable quantification of S_t, HCl, HF, HBr, and HI. The volcanic gas–atmosphere mixture was pumped through the tube using a GilAir Plus pump (Sensidyne) set at a constant 4 L/min flow rate. Following collection, the samples collected by both the UH and INGV-Palermo teams were stored, prepared, and analyzed according to the methods of Wittmer et al. (2014) at laboratories within INGV-Palermo. All samples were analyzed using ion chromatography (Thermo Scientific Dionex ICS1100 and ICS5000) to quantify sulfur, chlorine, and fluorine content, and those of the UH group were also analyzed by inductively coupled plasma–mass spectrometry (ICP-MS; Agilent 7500 CE) to quantify the bromine and iodine content.

Two styles of ground-based remote sensing were employed during the field campaign to measure SO₂ column densities (ppmm [parts per million × meter]) across Lastarria’s plume, allowing SO₂ and total volatile fluxes (t/d) to be calculated. These methods include: (1) stationary SO₂ camera measurements, and (2) zenith-looking DOAS traverse measurements (Fig. 3). Both methods utilize scattered solar radiation and the SO₂ absorption spectrum within the ultraviolet (UV) wavelength region to quantify the amount of SO₂ within an atmospheric column. On 21 November, SO₂ camera measurements were collected by a group from UNIPA, while walking traverse measurements were collected around the crater rim downwind of the upper field fumaroles by the UNA group. On 22 November, SO₂ camera measurements were repeated from the same site by groups from UNIPA and the Geological Survey of Japan (AIST), while vehicle-based DOAS traverses were conducted downwind of the plume on the east side of the edifice by groups from the Instituto Geofísico–Escuela Politécnica Nacional (Ecuador; IG-EPN) and the University of Alaska Fairbanks (UAF). More details of these measurements and their expected errors are described below.

On 21 and 22 November, SO₂ camera measurements were collected from a location north of the edifice at a distance of ∼7 km from the plume (Fig. 3) and with near-perpendicular views of the plume. On both days, the plume was hugging the ground such that SO₂ from the entire plume was not quantified and total calculated fluxes are likely underestimated. The group from AIST used an Apogee Alta F260 camera, while the group from UNIPA used a Jai CM-140GE camera. Both groups employed double camera systems with filters centered at 310 and 330 nm, to enable SO₂ column density to be quantified (Kern et al., 2015). Absorbance of UV radiation by SO₂ gas was converted to column densities using a combination of integrated DOAS measurements and calibration cells. Plume speed was calculated using the acquired plume imagery and feature tracking software (e.g., ImageJ with the MtrackJ plugin; Abràmoff et al., 2004; Taddeucci et al., 2012) to manually track the plume’s main features.

On 21 November between 19:56 and 20:16 UTC, continuous SO₂ camera measurements of the bulk plume were collected by UNIPA at a 1 s sample rate. The light intensity data were corrected for light dilution effects (Mori et al., 2006; Kern et al., 2010) using the following procedure. First the reflectivity of a rock within the image field of view was assumed to be zero, and then the average measured UV intensity of the rock was subtracted from the 310 and 330 nm filtered data. On 22 November, continuous SO₂ camera measurements were collected by the AIST group between 14:01 and 14:16 UTC. Measured column densities were corrected for atmospheric dilution using a different procedure than described above. Specifically, coincident DOAS spectral data and SO₂ imagery were used following the methods of Campion et al. (2015) and Kazahaya et al. (2013) to calculate a range of light dilution causing the intensity data to be underestimated by ∼50%–60%. The SO₂ fluxes were corrected to compensate for this underestimation. We note that on both days, a ground-hugging plume geometry and the transport of the lower field plume in front of the edifice prevented the entire plume from being measured from the SO₂ camera site, such that these fluxes may be underestimated by ∼10%–30% based on visual observations. Considering this factor in addition to standard errors in SO₂ column density and plume speed, we estimate a total error in these calculated SO₂ fluxes to be ∼50%.

DOAS traverse measurements were made to quantify absorption of UV light by plume SO₂ gas by the UNA, IG-EPN, and UAF groups using Ocean Optics (S2000 and USB2000 model) spectrometers. SO₂ was calculated from the measured absorption spectra using clear (SO₂-free) sky and dark (instrument noise) spectra measured in the field, along with: (1) laboratory spectra of SO₂ and O₃, a Ring spectrum to account for inelastic (Raman) scattering, and a fifth-order polynomial to account for broadband extinction, through application of a spectral fitting routine integrated in NOVAC mobile DOAS software (Zhang and Johansson, 2009; Galle et al., 2010) (UNA and IG-EPN); and (2) field measurements of high- (1305 ppmm) and low-concentration (482 ppmm) calibration cell spectra integrated in a FLYSPEC system (Horton et al., 2006; Businger et al., 2015) (UAF). Spectral data were analyzed between ∼310 and 325 nm (UNA and IG-EPN) and between 319 and 330 nm (UAF). In all cases, coincident GPS measurements were collected to obtain measurement time and location. Plume speed was estimated from windspeed using: (1) handheld anemometer measurements from the Lastarria crater rim at plume height by the UNA group, and (2) modeled wind data from the U.S. National Ocean and Atmospheric Administration (NOAA) for Lastarria’s summit altitude and approximate sample collection time by the IG-EPN and UAF groups. In all cases, measured SO₂ column densities were integrated over the plume width and multiplied by plume speed to calculate SO₂ fluxes (t/d). The calculated standard error for the DOAS traverse measurements based on the error in the DOAS fit and windspeed is ∼30%.

Two walking traverses on the upper field crater were made by UNA to measure SO₂ column densities and calculate SO₂ mass fluxes from the condensed upper field fumarolic plume on 21 November 2014 between 16:57 and 17:49 UTC. Then on 22 November, three vehicular-based DOAS traverses were conducted downwind and ∼1 km beneath Lastarria’s translucent plume by UAF and IG-EPN in the same vehicle between 14:09 and 15:24 UTC (Fig. 3).

Soil CO₂ flux measurements were carried out at Lastarria on 21 November by groups from the Research Institute for Volcanology and Risk Assessment (University of the Azores, Portugal; IVAR) and the University of Perugia (Italy; UNIPE). A total of 139 measurements based on the accumulation chamber method (Chiodini et al., 1998) were made surrounding the lower fumarolic field (Fig. 3). Soil temperature at 15 cm depth was also recorded in 91 of the sampled sites. For reference purposes, six “background flux” measurements were performed ∼5 km northwest of the lower fumarole field, near the SO₂ camera site (Fig. 3).

A soil CO₂ flux map was generated using the discrete soil CO₂ flux measurements made over an area of 38,725 m². The soil CO₂ flux and temperature maps were generated following the sequential Gaussian simulation (sGs) method (Deutsch and Journel, 1998; Cardellini et al., 2003). As the sGs method needs a multi-Gaussian distribution of the data, which implies a normal distribution of the values, original data were transformed into a normal distribution using the “nscore” algorithm by Deutsch and Journel (1998). Furthermore, the normality of the bivariate cumulative distribution function of any pairs of values, the second condition required by the multi-Gaussian model (Deutsch and Journel, 1998), was checked. Experimental variograms were computed and the derived variogram models were used in the sGs to create 100 realizations of the flux and temperatures grids (Supplemental Fig. S1¹). The E-type estimate maps, i.e., the maps of the “expected” value at any location obtained through a pointwise linear average of all the realizations, are shown in Figure 4. The total CO₂ released by soil diffuse degassing was computed by multiplying the “expected” values of each grid cell by the cell surface.

The total gas composition as measured using Giggenbach bottles from sampled fumaroles on both the upper and lower fields comprised primarily H₂O (87.3–96.4 mol%), CO₂ (2.0–9.12 mol%), SO₂ (0.69–1.50 mol%), H₂S (0.51–0.85 mol%), and HCl (0.15–0.54 mol%) (excluding one outlier thought to be contaminated during analysis) (Table 1). The atmospheric components Ar (2.78 × 10^–5 to 5.4 × 10^–4 mol%), and O₂ (0–0.024 mol%) composed a relatively minor proportion of the emitted gases in most samples, suggesting minimal air contamination. H₂, a common gas in moderate- to high-temperature volcanic and hydrothermal emissions, was present at relatively high abundance (0.0067–0.18 mol%). The reduced carbon species made up a relatively minor component of the gas emissions (∼1.2 × 10^–6 to 2.47 × 10^–5 mol% CO, 1.9 × 10^–5 to 2.2 × 10^–4 mol% CH₄). The remaining gas species analyzed include HF (0.002–0.01 mol%), N₂ (0.028–0.19 mol%) and He (1.5 × 10^–5 to 1.47 × 10^–4 mol%) (Table 1).

Good agreement was observed in δ¹³C-CO₂ between both upper and lower field fumaroles_, and between the CRDS and mass spectroscopy analyses. All measurements converge around a mean value of –2.5‰ with a standard deviation of 0.3‰ (Table 2).

The Lastarria calculated R/R_A values from the three collected samples were highly similar, with a mean value of 5.12 and a standard deviation of 0.005. ⁴He/²⁰Ne ratios (between 159.9 and 199.1) were used to correct for minor contamination by atmospheric air assuming a purely atmospheric origin for neon (Hilton, 1996). The resulting air-corrected mean R_c/R_A value is 5.13 with a measured helium concentration in the range of 11.9–13.3 ppm.

Major species gas mixing ratios (error ≤5%) measured by the MultiGAS from the lower fumarole field, representing emissions from numerous coalesced fumaroles (referred here as the plume), ranged from 400 to 800 ppm CO₂, 0.1 to 160 ppm SO₂, and 0.1 to 60 ppm H₂S. After subtracting local atmospheric background for both H₂O and CO₂, the calculated molar gas ratios give H₂O/CO₂ of 6.2, CO₂/SO₂ of 2.2, and H₂S/SO₂ of 0.21. When gas composition is normalized to only include H₂O, CO₂, and S_t, a gas composition containing ∼78.1 mol% H₂O, 12.7 mol% CO₂, and 9.1 mol% S_t is observed. Error in the molar gas ratios is estimated to be 20% based on experimental results of de Moor et al. (2016a).

The compositions from the two main degassing sites measured by alkaline traps were composed primarily of S_t (0.18–27.9 mmol/L) and HCl (0.21–5.48 mmol/L) (millimoles of the measured species per liter of air sampled), with minor quantities of HF (0.13–0.49 mmol/L), HBr (0.0022 mmol/L), and HI (0.003–0.007 mmol/L). Slightly lower relative HCl/S_t ratios were observed from the lower field emissions (HCl/S_t = 0.12–0.29) in comparison with the upper field (HCl/S_t = 0.20–0.47) (Table 3). Singular (Raschig tube) samples from each field were used to calculate an HF/S_t ratio of 0.009 and HBr/S_t of 1.5 × 10^–4.

On 21 November, plume speed from Lastarria’s upper field crater was measured using a handheld anemometer at 14 m/s. These measurements were used along with the DOAS walking traverse measurements that day to calculate SO₂ fluxes from Lastarria’s upper field only, ranging from 294 to 486 t/d, with an average value and 1σ standard deviation of 390 ± 136 t/d. SO₂ camera measurements by the UNIPA group acquired ∼2 h later were used to calculate apparent plume speeds of 5 m/s based on feature tracking within the SO₂ imagery. The final SO₂ camera-derived fluxes for the bulk plume, corrected for light dilution, ranged from 417 to 916 t/d, with an average value and a 1σ standard deviation of 615 ± 87 t/d (Table 4).

On 22 November, AIST SO₂ camera measurements and feature tracking methods were used to calculate an average bulk plume speed of 9 m/s. After incorporating the light dilution correction, the final calculated AIST SO₂ camera fluxes between 14:01 and 14:16 UTC ranged from 664 to 1719 t/d with an average value and 1σ standard deviation of 1094 ± 193 t/d (Fig. 5). The expected error for these measurements is ∼50%, resulting in a mean SO₂ flux and error range of 1094 ± 547. Plume speed was calculated for that same day by the UAF and IG-EPN groups using NOAA modeled wind data to be 9.7 m/s. Calculated SO₂ fluxes for UAF and IG-EPN, respectively, for the three traverses acquired between 14:09 and 15:24 UTC were: 759 and 903 t/d for traverse 1, 1033 and 997 t/d for traverse 2, and 1453 and 1334 t/d for traverse 3 (Fig. 5). The average flux and 1σ standard deviation for UAF and IG-EPN measurements respectively, were calculated to be 1082 ± 350 t/d (UAF) and 1078 ± 226 (IG-EPN)(Table 4). The upward-looking viewing geometry and closer distance to the plume minimizes errors due to light dilution and scattering (Mori et al., 2006; Kern et al., 2012), such that no corrections were done in these cases, and these measurements are assumed to be our most accurate. Then between 15:46 and 16:05 UTC, continuous SO₂ camera measurements were collected from the same SO₂ camera site (Fig. 5) by the UNIPA group. Plume speed for this time period was estimated to average 7 m/s based on plume feature tracking of the SO₂ imagery. Following the light intensity correction, the calculated SO₂ fluxes ranged from 170 to 765 t/d, with an average value of 455 ± 90 t/d. The mean SO₂ flux (±1σ standard deviation) calculated by combining the traverse and SO₂ camera measurements on 22 November yields a value of 604 ± 296 t/d.

The CO₂ released from Latarria’s edifice as measured during the discrete soil degassing survey on 21 November was calculated by integrating the average CO₂ flux values estimated from sGs over the measurement area (∼38,725 m²). These analyses yield a combined diffuse CO₂ flux of ∼5 t/d. Dividing the calculated CO₂ output by the measurement area gives an average degassing flux from the sample region of 1.29 × 10^–4 t/d/m².

Throughout this study, a variety of gas sampling and/or remote sensing measurements were employed to quantify the composition and flux of volcanic gases from Lastarria volcano. As previously mentioned, inherent differences in the individual sampling methods prevent direct comparison of these data sets. Specifically, fumarole sample composition collected at a single emission source will differ from the bulk plume composition (a mixture of volcanic gases from numerous fumaroles and air) measured using in situ techniques such as MultiGAS and/or alkaline traps. While acknowledging these differences, we briefly compare complementary measurements and discuss in general variations among the data sets to better evaluate the limitations in our results and implications for our data interpretations.

In general, fairly good agreement in direct fumarole sample gas analysis collected in Giggenbach bottles was observed for measurements collected by various groups from the same fumarole (Tables 1 and 3). When normalized gas concentrations from Table 1 are evaluated, variations in individual gas species that, at times, exceed the analytical uncertainties (<10%) are observed. Variations in measured gas compositions from the same fumarole have been observed at previous CCVG gas workshops, and mechanisms to account for these variations have been described in detail by Giggenbach and Matsuo (1991) and Giggenbach et al. (2001). The main factors contributing to these differences include: (1) variations in samples due to condensation or entrainment of water in the sample line and/or the outgassing itself, and (2) variations in analytical methods (Giggenbach et al., 2001). In particular, we note that variations in sulfur content and speciation (Tables 1 and 3) between groups from UNIFI and UNM are likely due to the different analytical methods employed (e.g., Vaselli et al., 2006; Giggenbach and Goguel, 1989). Because slight differences in the concentration of one measured gas species can have large effects on the normalized composition, we consider evaluation of gas ratios to be more useful for comparison purposes. When major species gas ratios determined by the various groups for the lower field fumarole samples are compared, we find good agreement, with the following average ratios and 1σ standard deviations found: H₂O/CO₂ = 15 ± 3, H₂O/S_t = 43 ± 8, CO₂/S_t = 2.9 ± 0.3, and HCl/S_t = 0.17 ± 0.04 (Table 3). When we consider the major species gas ratios directly sampled from the lower-temperature upper field fumarole, we find somewhat larger variations with the following average ratios and 1σ standard deviations: H₂O/CO₂ = 38 ± 14, H₂O/S_t = 71 ± 13, CO₂/S_t = 1.9 ± 0.38, and HCl/S_t = 0.20 ± 0.04 (Table 3). While variations in gas composition among the different sample collection groups can most likely be attributed to differences in sampling and analytical techniques (Table 1), general trends in gas composition can be identified to characterize the overall gas source. When the major species gas ratios for the lower and upper fumaroles are compared, the upper field fumarole appears to be enriched in H₂O, SO₂, and HCl relative to the lower field, which is relatively enriched in CO₂ (Table 3).

We next compare the average major species gas ratios measured by the MultiGAS to those measured through direct fumarole sampling from the lower field. The average MultiGAS ratios observed of H₂O/CO₂ = 6.2 ± 0.7, H₂O/S_t = 11.1 ± 2.2, and CO₂/S_t = 1.8 ± 0.3 are all notably lower than the average observed ratios measured by direct fumarole sampling from the lower field (H₂O/CO₂ = 15 ± 3, H₂O/S_t = 43 ± 8, and CO₂/S_t = 2.9 ± 0.3), which can be explained in part through a relative depletion in H₂O compared to other gases for the bulk plume measured by the MultiGAS (Table 3). The observation of lower H₂O/CO₂ in MultiGAS measurements compared to direct fumarole samples has been previously observed in other volcanic gas studies, and has been attributed to condensation of water vapor in the plume before reaching the instrument (e.g., Allard et al., 2014; Lopez et al., 2017). In this case, a similar (e.g., Allard et al., 2014) or slightly higher CO₂/S_t ratio would be expected in the MultiGAS data compared to the fumarole sample compositions as depletion of H₂O due to condensation may also involve some absorption of SO₂ by water vapor. At Lastarria, however, depletions in both H₂O/CO₂ and CO₂/S_t are observed in the MultiGAS data relative to the fumarole samples, consistent with a higher proportion of sulfur gas measured by the MultiGAS. To explain this observation, we therefore conclude that at least two processes must be contributing to the variations in gas composition observed by the MultiGAS versus direct fumarole samples. Because the direct fumarole samples represent gases emitted from a single point source, while the MultiGAS measurements represent the composition of the bulk plume, it is likely that spatial heterogeneities in gas composition within the fumarole field may be partially contributing to these observed differences (e.g., Tamburello et al., 2015). We speculate that the MultiGAS measurements of CO₂/S_t may better represent Lastarria’s bulk plume composition, including the highest-temperature, SO₂-rich fumaroles, as fumarole temperature and SO₂ composition are found to directly correlate (e.g., Aiuppa et al., 2005; Aguilera et al., 2012; Tamburello et al., 2015). We further speculate that the direct fumarole analyses of H₂O/CO₂ composition may best represent Lastarria’s gas composition that has not been modified by steam condensation. We propose that future CCVG workshops conduct more detailed spatial sampling of fumarole fields by both MultiGAS and direct fumarole sampling methods to further constrain the differences between direct sampling and in situ gas measurements.

When the alkaline trap results are compared among the two collection groups, fair agreement in HCl/S_t was observed between the two sampling methods employed, with mean and 1σ standard deviations of 0.21 ± 0.07 and 0.37 ± 0.14 for the upper and lower fields, respectively (Table 3). The HCl/S_t ratios derived from the alkaline trap methods agree fairly well with those calculated from the direct fumarole sampling, with the alkaline trap measurements having consistently higher HCl/S_t (Table 3).

Variations in calculated SO₂ fluxes from Lastarria volcano that are within the assumed error are observed both within sampling periods by individual groups, as well as when measurements among the different groups are directly compared. Our results suggest fairly good agreement among the measurement techniques and somewhat consistent degassing over the study period. Unfortunately, the relatively large errors due in part to the ground-hugging plume geometry, which has increased the error in UV camera measurements to ∼50%, do not enable us to discriminate potential volcanogenic variations in gas fluxes that have been observed at other volcanoes (e.g., Tamburello et al., 2013).

On 21 November, UNA DOAS traverse measurements from the upper field were collected 2 h prior to UV camera measurements by UNIPA. The DOAS traverse measurements yielded an average SO₂ flux of 390 ± 136 t/d, while the SO₂ camera results of the bulk plume yielded an average flux of 615 ± 87 t/d (Table 4). The condensed nature of the plume during walking traverses, which likely increased the occurrence of multiple scattering, suggests that these DOAS SO₂ fluxes may be overestimated (e.g., Kern et al., 2010). Similarly, the ground-hugging plume geometry likely resulted in underestimated SO₂ fluxes from the SO₂ camera. Using these measurements, we calculate that <60% of the total SO₂ flux may have been supplied by the upper field fumaroles, however due to the limitations described above, we expect that the actual proportion of SO₂ supplied by the upper field is quite a bit lower.

On 22 November, SO₂ camera measurements by AIST temporally overlap with a single DOAS traverse made by UAF and IG-EPN at ∼14:10 UTC (Fig. 5). The most optimal viewing conditions (e.g., Kern et al., 2012) were experienced during these DOAS traverses as characterized by the translucent plume, vertical viewing geometry, and close location to the plume (estimated at <1 km), such that we expect these flux measurements to be the most accurate. Good agreement, well within the measurement error, was observed between the two groups conducting simultaneous DOAS traverse measurement for the three traverses conducted, with average and 1σ standard deviations of 1082 ± 350 by UAF and 1078 ± 226 by IG-EPN (Table 4; Fig. 5). We expect that the slight variations between the calculated fluxes were due to differences in SO₂ retrieval algorithms. This traverse, which took ∼20 min to complete, was coincident with 6 min of AIST 1 Hz SO₂ camera measurements. During the overlapping time period, the minimum, maximum, and mean SO₂ camera–derived fluxes were 664, 1384, and 1038 t/d (Fig. 5). The mean SO₂ flux and 1σ standard deviation calculated by the three groups of 1082 ± 350 (UAF), 1078 ± 226 (IG-EPN), and 1038 ± 161 (AIST) are in excellent agreement.

To accurately identify the temporal evolution of Lastarria’s magmatic-hydrothermal system, continuous measurements of Lastarria’s bulk plume composition over an extended time period are needed. Unfortunately, this and previous ground-based studies are limited to temporally discrete gas composition measurements from a select number of fumaroles or the bulk plume (Aguilera et al., 2012; Tamburello et al., 2014), and to continuous measurements of a single gas species (CO₂ and SO₂) from either a discrete number of fumaroles (Zimmer et al., 2017) or the bulk plume averaged over a year-long period (Carn et al., 2017), which limits the interpretations that can be made. The longest time-series data set for Lastarria’s gas emissions comes from temporally and spatially averaged OMI satellite SO₂ detections averaged over year-long periods between 2005 and 2015 (Carn et al., 2017). During this time, SO₂ emissions were persistently detectable, yet variable in strength, with annual mean masses ranging from ∼60 to 135 kt (Carn et al., 2017). When these values are compared to ground-based measurements from Tamburello et al. (2014) and this study (extrapolated over a year to facilitate comparison), we find that mean satellite-derived annual SO₂ masses are lower than those extrapolated from ground-based measurements (Fig. 6). These discrepancies could be explained by temporal variability in emissions that are not captured in our discrete ground-based campaigns and/or systematic biases among these measurement techniques. For example, a ground-hugging plume geometry, common at volcanoes with high winds, would hinder detection by satellite and likely cause an underestimation in measured mass. While the accuracy of the satellite observations may require further investigation, temporal variations outside a 1σ standard deviation are evident, suggesting that temporal variations are captured by these observations. Based on these observations, we can conclude that continuous SO₂ emissions of significant quantity have been supplied by Lastarria volcano for at least the past 10 yr and appear to have increased between 2010 and 2015. When we consider the temporal variations between the ground-based SO₂ flux measurements collected in 2012 (884 ± 779 t/d, mean ± standard deviation; Tamburello et al., 2014) and those presented here (604 ± 296 t/d), the relatively high standard deviations and estimated measurement error (50% for SO₂ camera measurements in this study) unfortunately prevent us from identifying temporal changes in SO₂ flux (Fig. 6).

In addition to the somewhat low temporal resolution of our observation set, other factors that can impede interpretations from being made include: (1) spatial variations in the fumarole field, (2) potential biases due to different measurement techniques, and (3) variations in apparent gas composition due to changes in meteorological conditions or regional stresses. Here, by combining direct fumarole samples from a discrete number of fumaroles with bulk plume composition measurements, we eliminate the potential interpretation biases due to spatial variations and technique biases due to using only one of the above methods. Additionally, we consider it unlikely that meteorological changes such as increased precipitation events significantly modified our gas composition measurements in 2014 based on observations by Zimmer et al. (2017). Specifically, those authors found that large precipitation events have only short-term (hours to days) effects on temperature and gas composition at Lastarria, with these parameters quickly returning to their baseline values. No significant precipitation events were observed in the weeks preceding our November 2014 field campaign (Zimmer et al., 2017), such that we do not expect the elevated H₂O in 2014 to be meteoric in origin. We note that isotopic measurements of water from Lastarria fumaroles by Aguilera et al. (2012) showed the water to be largely magmatic in origin, further supporting a magmatic source to Lastarria’s water.

Here we evaluate changes in Lastarria’s gas composition in recent years and evaluate potential trends with respect to location (i.e., fumarole field), temperature, and sampling method employed (Fig. 7). First, we consider only the normalized concentrations of the following gas compounds: (1) CO₂, total S (S_t), and H₂O; and (2) CO₂, HCl, and S_t. These compositions are presented in ternary diagrams often used to depict the composition of gases from magmatic-hydrothermal systems (e.g., Giggenbach et al., 1990; Symonds et al., 2003; Shinohara et al., 2011; Aiuppa et al., 2015; Fischer and Chiodini, 2015), where clear temporal differences in gas composition can be seen that are well outside of the scatter due to different measurement techniques (Fig. 7). Specifically, in both ternary diagrams, the compositions of gas measured from 2006 to 2009 are dominated by water and/or CO₂, reflecting a more typical hydrothermal gas composition as previously described by Aguilera et al. (2012). Then in 2012 and 2014, the gas composition contains a significant portion of acidic gases (S_t and HCl), relative to the more hydrothermal gases, consistent with shallow degassing magma in the absence of significant scrubbing by hydrothermal water. The scatter in gas composition observed in 2014 is likely due to different sampling and/or analytical techniques, while the apparent clustering observed in both ternary diagrams in 2006–2009 has been attributed to fumarole outlet temperature (Aguilera et al., 2012). In Figure 7A, the MultiGAS data show lower relative H₂O content with respect to fumarole sample data, which may be due in part to steam condensation, however further analyses are needed to fully explain the compositional differences observed between these techniques.

In Figure 8, we plot gas ratios to evaluate Lastarria’s gas composition measured over the five discrete study periods: May 2006, March 2008, April–June 2009 (Aguilera et al., 2012), November 2012 (Tamburello et al., 2014), and November 2014 (this study). To minimize the influence of shallow processes on gas composition, we plot compositional data from high-temperature (>120 °C) gases only, using the nomenclature of Aguilera et al. (2012). Because no major difference in gas composition was observed between the two fumarole fields during our campaign in 2014 (Table 3), or among these fields in 2006–2009 for the high-temperature fumaroles (Fig. 7), we include measurements from both fields in Figure 8. The composition of fumaroles sampled in 2006, 2008, and 2009 show some scatter, likely due to spatial variations, but the measured gas compositions differ generally from those seen in 2012 and 2014. As we move forward in time from 2006 to 2014, distinct compositional changes can be seen, while some differences due to different sampling techniques are observed. In general, the following changes are observed: (1) the H₂O/CO₂ ratio increases over the five sample periods, with a maximum value observed in 2014; (2) the CO₂/S_t and H₂O/S_t ratios are high between 2006 and 2009 and then low in 2012 and 2014, consistent with enrichment in S gas in recent years; and (3) the S_t/HCl ratios measured by direct sampling appear to be similar throughout the study period, though some temporal variations are observed that may be attributed to different sampling techniques. The simultaneous incremental increase in H₂O/CO₂ along with a decrease in CO₂/S_t and H₂O/S_t from 2006 to 2014 all are consistent with an increase in the proportion of gases emitted with higher solubilities either in the magmatic melt phase and/or water. If we assume that these discrete observations are representative of long-term variations, these changes over time can be explained by four possible scenarios: (1) magma has ascended to relatively shallow depths and has undergone decompression-induced degassing; (2) a stalled, shallow magma body has undergone crystallization-induced degassing; (3) the hydrothermal system has become less voluminous through heating and drying, a decrease in precipitation events, and/or a change in the local stress regime; and/or (4) the hydrothermal system has become increasingly more acidic, such that SO₂ and HCl are no longer being scrubbed as efficiently by water (Symonds et al., 2001; Tamburello et al., 2015; Capaccioni et al., 2016; de Moor et al., 2016b). These scenarios are illustrated in Figure 9.

We next consider each discrete gas observation to determine its consistency with the proposed scenarios. First, an increase in H₂O/CO₂ is consistent with scenarios 1 and 2 (decompression- and crystallization-induced degassing, respectively), but cannot be readily explained by scenarios 3 and 4 (depletion and acidification of hydrothermal water, respectively), where a decrease in H₂O relative to CO₂ and/or no change would be expected. The temporal increase in S_t compared with H₂O and CO₂ over the study period is consistent with all four scenarios, due to the relatively high magma and water solubilities of S species. The fairly consistent S_t/HCl ratios observed from the direct samples in all five campaigns could also be consistent with all four scenarios, though interpretations are less clear due to the complicated solubility behavior of HCl in both water and melts (e.g., Webster et al., 2015). If the temporal variations observed by all measurement techniques are representative of long-term trends and not artifacts of spatial variability and/or differences in measurement technique, than the low S_t/HCl composition observed in 2012 would be most consistent with decompression- or crystallization-induced degassing scenarios based on previous empirical and modeling results of S_t/HCl behavior in volcanic systems (Edmonds et al., 2001; Lesne et al., 2011). Considering these volcanic gas observations, we favor scenarios 1 and 2 as the most likely models to explain subvolcanic conditions at Lastarria volcano during our study period. We next consider existing geophysical constraints to better constrain a subsurface model to explain our observations.

Numerous geophysical studies over the past 20 yr indicate the presence of two shallow crustal inflation sources beneath Lastarria’s edifice, presumed to represent magma and magmatic-hydrothermal fluids, respectively. Specifically, InSAR geophysical observations by Pearse and Lundgren (2013) found continued inflation of the ∼8-km-depth deformation source beneath the Lazufre region between 2005 and 2010, and again by Henderson et al. (2017) through 2016. To sustain this continuous inflation, we assume that magma has been continuously supplied to this ∼8-km-depth crustal storage region from deeper within the crust through 2016. Additionally, InSAR observations by Ruch et al. (2009) indicate inflation of a <1 km source between 2000 and 2008, and seismic tomography results of Spica et al. (2015) identify the presence of low-velocity zones also at <1 km depths below Lastarria between 2008 and 2012. These authors presume that the shallow inflation (Ruch et al., 2009) and spatially coincident low-velocity zone (Spica et al. 2015) represent magmatic volatiles and/or hydrothermal waters, though we argue that shallow magma may be another viable option that would produce the same geophysical signatures. As with the deeper inflation source, we presume that a discrete or continuous supply of magma or volatiles was necessary to produce inflation of the shallow source between 2000 and 2008. No targeted observations of this shallow source are available since 2008, so it is unclear if this shallow source has continued to deform in recent years.

We now consider our gas observations in the context of these geophysical interpretations to constrain a viable subsurface model for Lastarria during our study period. First, the CO₂-rich composition of Lastarria’s high-temperature fumaroles between 2006 and 2009 is consistent with an influx of a deep CO₂-rich magma and a permeable conduit enabling volatiles to reach the surface and outgas in advance of their host magma. The additional changes in high-temperature gas composition observed between April 2009 and November 2014 indicate a change in the magmatic-hydrothermal system during this time. In the absence of updated geophysical data targeting the shallow (<7 km depth) region beneath Lastarria, we simply describe the recent deformation that would be expected for the four scenarios (Fig. 9) if observations were available. Scenario 1 would require continued (since 2009) magma recharge and shallow inflation to explain the measured gas composition and to be consistent with the observed low-velocity zone (Spica et al., 2015) and the pre-2009 deformation signal (Ruch et al., 2009). In this scenario, an increase in SO₂ flux would also be expected, which is not clearly seen in the existing satellite and ground-based data as previously described. Scenario 2 could be explained by shallow inflation prior to 2009, followed by shallow deflation after 2008 as the shallow magma crystallizes and degasses (assuming other parameters are equal). Scenarios 3 and 4 can only be explained if the magmatic gases supplied by the ∼8 km magma source condensed and became trapped beneath an impermeable layer to cause the shallow inflation prior to 2009. Because persistent degassing has been observed from Lastarria throughout historical time and quantified with high measured proportions of magmatic gases since 2005 (Carn et al., 2017; Aguilera et al., 2012; Tamburello et al., 2014), a permeable conduit facilitating degassing likely prevails at this volcano, which would make scenarios 3 and 4 less likely. If a change in stress were to cause the sealing of former gas pathways and contribute to gas and/or fluid accumulation in the shallow source, then a decrease in SO₂ emission rates during this time period would be expected, which was also not clearly observed in satellite data by Carn et al. (2017). Considering these factors, we think scenarios 3 and 4 are less likely than the other scenarios to explain observations prior to 2009. Since 2009, a near-surface deflation signal would be expected to support scenario 3, in the case that the hydrothermal system was depleted between 2009 and 2016. No deformation change of the shallow source since 2009 would be expected to support scenario 4, acidification of the hydrothermal system. These interpretations are summarized in Table 5 and Figure 9.

Considering the combined observations in both the gas and deformation data sets, we consider scenarios 1 (decompression-induced degassing) and 2 (crystallization-induced degassing), both involving degassing of a shallow magma source, to be the most likely to explain the recent geochemical and geophysical observations at Lastarria volcano. Additional deformation studies that target shallow depths could help further refine and discriminate between these two scenarios.

We can quantitatively test the validity of scenario 1 by estimating magma degassing depths from the normalized CO₂/H₂O gas composition observed in 2006–2009 and in 2014 using the Lastarria melt composition and volatile solubility trends following the methods of López et al. (2013). The required input parameters to calculate magma degassing depths (assuming degassing due only to decompression) include: initial melt CO₂ and H₂O concentrations; melt composition and temperature at storage depths; degassing pathways (open system versus closed system); and the final gas composition. Due to petrologic analyses of prehistoric Lastarria eruptive products by Stechern et al. (2017), many of these parameters are already constrained or can readily be calculated. Specifically, Stechern et al. (2017) calculated melt H₂O concentrations of up to 6.0 wt% based on amphibole analyses from Lastarria eruptive products and models by Ridolfi et al. (2010) and Ridolfi and Renzulli (2012). They further constrained a rhyolitic melt composition, melt storage pressures of ∼200 and 300–550 MPa, and melt temperatures of 840–1060 °C. Using values near the upper bounds calculated by Stechern et al. (2017) of 478 MPa and 1025 °C to represent the deep portion of the crustal storage region, we calculate the amount of CO₂ that could be dissolved in Lastarria melt containing 6 wt% H₂O using VolatileCalc software (Newman and Lowenstern, 2002) to be 1400 ppm. These composition, temperature, and pressure values are used as our initial conditions in the following exercise.

The goal of this exercise is to evaluate how the gas composition changes during decompression ascent from 478 MPa (4780 bars) to the surface. We consider two end-member degassing paths including: (1) open-system degassing where gases readily separate from their host magma and ascend through the conduit and are immediately lost to the system, and (2) closed-system degassing conditions where the gases travel with their host magma and maintain their equilibrium composition with the melt after separation and outgassing. We note that due to the presumed high magma viscosity for a rhyolitic melt, we expect purely open-system degassing conditions to be unlikely at Lastarria volcano and would expect closed-system degassing to dominate until depths at which a permeable conduit is achieved, at which point open-system degassing could commence. In the closed-system degassing case, we consider a concentration of 1 wt% exsolved volatiles (Fig. 10), which is a reasonable value for arc volcanoes (e.g., Edmonds, 2008), and assume that Lastarria’s degassing behavior likely represents a combination of the open-system and the closed-system (with 1 wt% exsolved volatiles) end members. Next, we calculate the pressure at which a rhyolitic melt would maintain the observed normalized gas compositions of ∼85% H₂O (2006–2009) and ∼95% H₂O (2014), with all other factors being equal. These lines of equal composition are referred to as isopleths, and when compared with the magma degassing pathways, can be used to constrain magma degassing depths (Fig. 10). Finally, lines of equal pressure (isobars) for a rhyolitic magma at the assumed 1025 °C temperature can be calculated to enable the intersection of the degassing pathways and gas compositions to be used to estimate magma degassing pressures (i.e., depths). The results of this exercise are shown in Figure 10. The intersection of the ∼85 mol% H₂O isopleth (2006–2009) with the open-system and the 1 wt% exsolved gas closed-system degassing pathways occurs at ∼270 MPa (∼10.5 km) and 170 MPa (6.6 km), respectively. This suggests that based on the observed gas chemistry and our model assumptions, for the time period between 2006 and 2009, the degassing magma beneath Lastarria’s edifice likely resided at a mean depth between 10.5 and 6.6 km. This depth range is consistent with the depth of the InSAR-derived mid-crustal storage region throughout this period. The intersection of the ∼95% H₂O isopleth (2014) and the open-system and the 1 wt% exsolved gas closed-system degassing pathway occurs at ∼230 MPa (∼9.0 km) and 50 MPa (1.2 km). This suggests that in 2014 the mean depth of Lastarria’s degassing magma likely resided between 9 and 1.2 km. While this most recent range of magma degassing depths is large, it is consistent with magma ascent since 2009. If Lastarria’s 2000–2008 near-surface inflation source is magma, and has continued to inflate since 2008, then the more-shallow range in magma degassing depths calculated here would be consistent with scenario 1. The relatively high S_t/HCl ratios observed at Lastarria in 2014 are also suggestive of shallow magma (e.g., Edmonds et al., 2001), supporting scenarios 1 and 2, though better constraints on the initial volatile composition from melt inclusion data and improved volatile solubility models that include HCl are needed to test this hypothesis.

Within subduction zone settings, the primary sources of volatiles to arc magmas include: (1) the subducted slab, consisting of altered oceanic crust and overlying marine sediments, (2) the mantle wedge, and (3) the crust. When the isotopic composition of carbon and helium within measured gas samples from arc volcanoes are compared with the end-member compositions of the potential volatile sources, constraints on volatile source can be made (e.g., Sano and Marty, 1995). Specifically, the δ¹³C-CO₂ composition of mid-ocean ridge basalt (MORB), which is commonly used as a proxy for upper mantle material, is estimated to be –6.5‰ ± 2‰. The δ¹³C-CO₂ composition of carbonate rocks is 0‰, while that of organic sediment is <–20‰ (Sano and Marty, 1995). The δ¹³C-CO₂ composition emitted from Lastarria gases in 2014 had a mean value of –2.5‰, which suggests a mixture of carbon sourced primarily from subducted or crustal carbonates and the mantle. The δ¹³C-CO₂ composition can be used with the CO₂/³He composition of the measured gases and the three-component mixing model of Sano and Marty (1995) to calculate the proportion of gases sourced from these three possible end members (Table 2; Fig. 11). Using the average measured CO₂ and He compositions in mole percent from Table 1 along with the R_c/R_a composition measured by CEA (this study) and the known ³He/⁴He composition of air, we calculate a CO₂/³He for 2014 of 2.2 × 10¹⁰ (Table 2). This value is within the typical range of arc volcanoes worldwide (Hilton et al., 2002). Using these values, we conduct the three-component mixing calculation and find that the 2014 volcanic gas samples appear to be sourced from 7% mantle-derived volatiles, 87% carbonate-derived volatiles, and 6% organic sediment-derived volatiles (Table 2). This composition has a slightly lower ratio of CO₂/³He compared with the mean gas samples collected by Aguilera et al. (2012), which are calculated to be sourced from ∼2%–7% mantle material, ∼89%–91% carbonate, and 4%–9% organic sediment (Fig. 11). The 2014 composition would suggest a slightly greater proportion of mantle-derived volatiles and lower proportion of carbonate volatiles compared to previous years, however spatial variations in individual fumaroles sampled, analytical uncertainties in the measurements, and chemical fractionation during magma degassing (e.g., Barry et al., 2014) can also influence the gas composition and are likely affecting the calculated volatile source proportions. Regardless of these variations, it is clear that Lastarria’s volatiles are sourced in large part from carbonate rocks. We note that the calculated volatile-source proportions observed at Lastarria are similar to those of other Central Andean volcanoes including Lascar (Tassi et al., 2009) and Tacora (Capaccioni et al., 2011) volcanoes in northern Chile, suggesting that they have a similar crustal or subducted-slab carbonate source. Because the basement rocks within the Lastarria region are mostly composed of Paleozoic intrusive rocks and Tertiary–Quaternary volcanic rocks (Naranjo, 1992; Naranjo and Cornejo, 1992; Mamani et al., 2008), we can assume that the carbonate and organic volatile signatures are derived from the subducted slab, indicating that the subducted slab is the main volatile source at Lastarria. This interpretation is consistent with the findings of Aguilera et al. (2016), who infer a subducted sediment source for condensed gases sampled from Lastarria in 2014.

Constraints on total volatile flux can be calculated by combining SO₂ flux measurements, diffuse soil CO₂ flux measurements, and gas composition measurements. We assume that our combined average SO₂ flux obtained on 22 November 2014 of 604 ± 296 t/d gives us the most complete and accurate view of Lastarria degassing during our two-day study period. We note that this value is somewhat lower than the average SO₂ flux of 884 ± 779 t/d measured by Tamburello et al. (2014) two years prior (November 2012), however the high standard deviation in the calculated SO₂ fluxes calculated by Tamburello et al. (2014), which exceeds measurement error, indicates high temporal variability in degassing that may not be accurately captured during our brief sample campaigns. Considering this, we calculate a range in possible total volatile fluxes considering the minimum, maximum, and mean SO₂ flux measured (Table 6).

Next we consider the contribution of diffuse CO₂ soil degassing to Lastarria’s total volatile flux. Diffuse soil CO₂ degassing measurements were collected over a discrete area near the lower fumarole field to constrain degassing from the sample region. While measurements were collected from this one site only, we assume that diffuse degassing persists at similar rates at other locations around the edifice, acknowledging the limitation of this assumption due to variations in substrate composition and permeability. Previous studies have shown correlations to exist among diffuse degassing, hydrothermal alteration, and elevated surface temperatures at degassing volcanoes (e.g., Chiodini et al., 2001, 2005; Viveiros et al., 2010; Bloomberg et al., 2014). Considering this and the general agreement between regions of higher soil CO₂ fluxes and soil temperature in the sampled area (excepting the central area where the soil temperature was not measured due to technical problems; Fig. 4, black dots), we conduct the following extrapolation exercise to better constrain the diffuse degassing contribution to Lastarria’s total outgassing. First, we identify a region of expected elevated diffuse degassing based on locations where both Google Earth (Digital Globe [http://www.earth.google.com, accessed 5 June 2017]) imagery showed hydrothermal mineralization (Fig. S2²) and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) nighttime thermal infrared data acquired on 19 August 2014 at 03:15 UTC show elevated surface temperatures (Figs. 2 and 12). We ignored regions with apparent elevated thermal temperatures due to differences in emissivity and daytime heating (e.g., the lava flow in the north central part of the image). The selected area is 733,200 m² and includes the four fumarole fields outlined by dashed black lines in Figure 12 and by white solid lines in Figure S2 (^{footnote 2}). We note that elevated CO₂ fluxes were measured in our soil survey that do not correspond with elevated thermal temperatures or hydrothermal mineralization as seen in Figure 12, which emphasizes that this remote method is a conservative approximation only. We then assume that diffuse degassing over these selected areas is occurring at the average measured soil CO₂ flux rate of 1.29 × 10^–4 t/d/m² (measured area shown by solid black rectangle in Fig. 12) to calculate an extrapolated diffuse CO₂ flux of 95 t/d. Acknowledging that our measured flux area represents only a small fraction of the Lastarria edifice, we consider the measured flux of 5 t/d to be the minimum possible diffuse CO₂ flux for Lastarria volcano and consider the extrapolated value calculated here to represent our best estimate of total diffuse degassing, while further acknowledging that a greater flux is also possible. When the extrapolated diffuse soil CO₂ flux is compared to the calculated average plume CO₂ fluxes (Table 6), we calculate that diffuse CO₂ degassing from Lastarria most likely makes up ∼6% of total CO₂ degassing (plume degassing + diffuse degassing). This value is similar to what has been observed at other volcanoes with summit plumes including Mount Asama, Japan, which emits ∼12% of its total CO₂ diffusively (Morita et al., 2016), and Vulcano, Italy, which releases ∼20% of its CO₂ diffusively (Inguaggiato et al., 2012).

Next we calculate an average gas composition using the direct and in situ measurements collected in 2014 (Tables 1, 3) and assume that this result best represents the bulk plume composition from Lastarria. We use the resulting gas ratios with respect to SO₂ and the SO₂ flux of 604 t/d to calculate Lastarria’s total flux of the major gas species including H₂O, CO₂, H₂S, HCl, and HF (Table 6). When the calculated plume flux (assuming the average plume composition) and diffuse degassing fluxes are combined, we estimate the most reasonable (average) total degassing flux from Lastarria volcano of ∼12,400 t/d or ∼4.5 Mt/yr. This estimated total volatile flux agrees very well with that calculated by Tamburello et al. (2014) of ∼13,500 t/d, even though the fluxes of individual gas species and plume composition have changed. Acknowledging the differences in gas ratios calculated from the various gas measurement techniques, we also use the minimum and maximum X/SO₂ ratios observed, where X is one of the other primary gas species, to calculate lower and upper bounds of total volatile outgassing from Lastarria’s plume during our 2014 field campaign. Through this exercise, we calculate a minimum total volatile flux of ∼4150 t/d or ∼1.5 Mt/yr, and a maximum total volatile flux of ∼38,300 t/d or ∼14 Mt/yr. Next, we repeat the total volatile flux calculations using the minimum and maximum SO₂ fluxes measured on 22 November along with the mean gas ratios to better evaluate how variations in SO₂ may influence our total volatile flux calculations. The results are presented in Table 6B and show minimum and maximum total volatile fluxes of ∼3600 t/d (∼1.3 Mt/yr) and 35,100 t/d (∼12.8 Mt/yr), respectively. The highly similar minimum and maximum fluxes calculated using these two methods suggest that the range in possible total volatile fluxes presented in Table 6 well captures the observed variability in outgassing at Lastarria volcano. The wide range in gas composition and flux values observed and used in these calculations highlights the challenges in accurately constraining total degassing from active volcanoes.

New measurements of volcanic gas composition and flux from Lastarria volcano were acquired during two days in November 2014 and are used to compare and evaluate current gas measurement techniques, present updated constraints on Lastarria’s magmatic-hydrothermal system and total degassing budget, and provide further insights into proposed subsurface geophysical models. A qualitative comparison of the various gas sampling and measurement techniques is conducted, and finds good general agreement (within error) among most techniques. The largest variations were seen between direct and in situ (MultiGAS) gas composition measurements, and we recommend that future work aim to resolve the source of these discrepancies. The composition of Lastarria’s volcanic gases measured in 2014 was similar to that of 2012, while notably different than measured in 2006–2009. If these observed compositional differences are representative of long-term trends, then a distinct change in Lastarria’s magmatic-hydrothermal system likely occurred between 2009 and 2012. The 2014 measured gas composition contained relatively high proportions of H₂O and SO₂ and relatively low proportions of CO_2, and is consistent with degassing of shallow magma. Four possible subsurface scenarios are discussed to explain the apparent temporal variations in measured gas composition, including: (1) decompression-induced degassing due to magma ascent within the shallow crust; (2) crystallization-induced degassing of a stalled magma body; (3) depletion of the hydrothermal system due to heating, changes in local stress, and/or minimal precipitation; and/or (4) acidification of the hydrothermal system. Of these scenarios, options 1 and 2 are most likely due to the observed increase H₂O/CO₂ in 2014 that is not consistent with scenarios 3 and 4. When these scenarios are evaluated in the context of the existing geophysical models, all scenarios are feasible, and would likely produce different shallow deformation signals that could help discriminate the most likely subsurface scenario. We recommend that InSAR analyses be conducted on imagery acquired since 2008 to identify recent deformation changes and constrain which of the four scenarios best describes Lastarria’s subsurface system. Lastly, we use the average SO₂ flux calculated on collected on 22 November 2014 of 604 ± 296 t/d, the average plume composition measured through direct sampling and in situ measurements, and extrapolated diffuse degassing measurements to estimate a total degassing budget from Lastarria of ∼12,400 t/d, which is similar to that previously estimated in 2012.

The authors would like to thank the organizing committee of the CCVG 12^th Gas Workshop who made this project possible, especially Julio Cortés, Cristóbal González, Manuel Inostroza, and Susana Layana for their invaluable support during the field campaign. TL would like to acknowledge funding that supported the collection and analysis of this project from the U.S. National Science Foundation (NSF) PLUTONS project (award 0909254), the NSF GeoPRISMS Program (award 1250148), and the UAF Geophysical Institute. CS and FV would like to acknowledge funding that supported the survey from Fundo Regional da Ciência e Tecnologia (Program M3.3.a from 2014), Governo Regional dos Açores. FV is supported by a post-doctoral grant from Fundação para a Ciência e Tecnologia, Portugal (SFRH/BPD/100032/2014). ALR and ML thank INGV-Palermo colleagues, who provided laboratory support in preparing and analyzing dry gas and alkaline solution samples. NB was supported by a post-doctoral grant from the volcanoes and Atmosphere in Magmatic Open Systems (VAMOS) research unit, University Mainz, Germany. This manuscript was significantly improved by discussions with Matt Pritchard and thorough reviews by Jose Naranjo and two anonymous reviewers.