The Magmatic Gas Signature of Pacaya Volcano, With Implications for the Volcanic CO₂ Flux From Guatemala

2.1 CAVA and Guatemala

The Central American Volcanic arc is a 1,100 km-long volcanic front stretching along the continental margin associated with subduction of the Cocos plate beneath the Caribbean plate (Figure 1). The convergence rate between the Cocos and Caribbean plates gradually increases toward the southeast, from about 7 to 9 cm/yr (DeMets, 2001) (Figure 1). The age of the subducting Cocos lithosphere is approximately constant (about 26 Ma) along most part of the subduction zone, except along its south easternmost segment in central Costa Rica (17 to 25 Ma). The subducting lithosphere is divided into two segments by the Cocos Ridge, offshore of southern Costa Rica. The north-western segment is made of older and colder Cocos materials subducting at a relatively steep angle (65°) below El Salvador and Nicaragua, and at 55° beneath Guatemala (Syracuse & Abers, 2006). The Cocos Ridge is generated by separation of the Cocos and Nazca Plates and is consequently younger and warmer (Chan et al., 2002). According to Carr (1984) and Case et al. (1990), crustal thickness decreases from Guatemala (50-40 km) to Nicaragua (30 km), increases across Costa Rica and western Panama and then again thins in central Panama.

Variations in subduction parameters, including dip-angle, crustal thickness, extent of sediment under-plating, faulting style and subduction erosion (Carr, 1984, Carr & Feigenson, 1990; Patino et al., 2000, Walker et al., 2001), control the degree of source melting and determine a variable sedimentary fluid influx from the subducting Cocos plate. These in turn determine large along-arc geochemical variations in Central American lavas (Carr & Feigenson, 1990; Leeman et al., 1994; Morris et al., 1990), with trace element proxies for slab fluids (such as Ba/La, U/Th, and Sr/Ce ratios) being highest in Nicaragua, and lowest in Costa Rica (Sadofsky et al., 2008). The same geochemical tracers indicate that melts beneath Guatemala are generated by mixing of a Costa Rica MORB-like melt component with Nicaragua-like slab fluid-rich melts (Sadofsky et al., 2008).

Volcanic gases do also exhibit significant along-arc variations. Fischer et al. (2002) showed that N isotopes and N₂/He ratios in geothermal fluids vary along the volcanic arc, and indicate lower slab fluid contributions in Costa Rica than Guatemala and Nicaragua. Similar variations have also been observed for He and C isotopes (Di Piazza et al., 2015; Leeuw et al., 2007; Robidoux et al., 2017; Shaw et al., 2003), pointing to higher influx and recycling of C-rich sediments beneath Nicaragua, and a minimal slab contribution in the volcanic output in Costa Rica, in spite of similar sedimentary inputs at the trench (Freundt et al., 2014). Barnes et al., (2009), studying chlorine isotope variations along the Central American volcanic front, similarly demonstrated mantle-like Cl signatures in the northernmost (Guatemala and El Salvador) and southernmost (Costa Rica) ends of the front, and sediment- and/or serpentinite-derived signature in Nicaragua. Consistent results were obtained by (Aiuppa et al., 2014) who explored the along-arc variations in volcanic gas CO₂/S ratios in the Costa Rica-Nicaragua segment of the arc.

Four different morphotectonic segments exist in Guatemala that result from the rotation of upper crustal blocks in northern Central America (Figure 1). Zone I is un-affected by Neogene extensional tectonics, while extensional regime in Zone II resulted in large volcanic rock productivity. Zones I and II are separated by the Atitlan caldera complex. The structural line between zone II and zone III is the Misto fault, which also corresponds to the western margin of the Guatemala graben (Figure 1), where the Amatitlan caldera complex is located. Zone III is a basin-and-range style structure and the principal movement is eastward, while Zone IV is characterized by arc-parallel pull apart-basins (Burkart & Self, 1985). In southern Guatemala, both E-W crustal extension (as evidenced by the existence of a series of N-S trending grabens (Burkart & Self, 1985); Figure 1) and right-lateral strike-slip movements (in the Jalpatagua Fault Zone, JFZ) are observed. Pacaya volcano lies at the intersection between the Guatemala graben and the JFZ (Cameron et al., 2003).

2.2 Pacaya Volcano

Pacaya volcano, one the most active volcanoes in Central America, is a large volcanic complex on the southern fault zone of the Amatitlan Caldera (Wunderman & Rose 1984), only 20 km south of the centre of Guatemala City, the Guatemala capital. The Pacaya volcanic complex comprises an ancestral andesitic strato-volcano, a series of rhyodacitic to andesitic lava domes, and the modern Pacaya basaltic composite volcano (2,552 m high) (Conway et al. 1992). Eggers (1971) subdivided the activity of Pacaya into three stages. Phase I was characterized by growth of a large, ancestral andesitic cone; this old volcano was largely eroded. The onset of phase II (about 0.16 Ma) was marked by an increase in the volcano's explosivity, with the eruption of voluminous pumice-bearing deposits, and the formation of andesitic-dacitic lava domes such Cerro Grande and Cerro Chiquito (these events are identified as phase III in (Bardinzteff & Deniel, 1992). Phase III started ∼400 B.P. (divided in initial, historical, and modern sub-phases) and produced primarily basaltic lava flows through to the present day. These lavas are mostly porphyritic olivine-bearing basalts exhibiting no significant temporal chemical or petrographic variation (Bardinzteff & Deniel, 1992; Eggers, 1971). A series of collapse events affected ancestral Pacaya between 400 and 2,000 years B.P., forming a large collapse amphitheatre and a debris avalanche that traveled 25 km S-SW. Historical Pacaya encompasses the mid sixteenth century to 1861. Before this, Pacaya was possibly active in 1585, from ca. 1651 to 1678, and again in 1775 (Rose et al., 2013). After a repose period between 1860 and 1960, volcanic activity resumed in 1961 with modern Pacaya, which has since then shown nearly continuous strombolian activity and extrusion of lava from the Mackenney cone (Matías Gómez et al., 2012; Schaefer, 2012). The Mackenney cone, initially asymmetrical, began to grow in 1965 with the collapse of an amphitheatre that developed in 1961. There are not significant petrographic differences between initial, historic, and modern lavas, all showing olivine and plagioclase phenocrysts (up 45 wt. %) and microphenocrysts, and minor clino-pyroxene and magnetite microphenocrysts. Modern Pacaya lavas have progressively become slightly more mafic in composition (Bardinzteff & Deniel, 1992; Matías Gómez et al., 2012).

3.1 Multi-GAS

The composition of the Pacaya volcanic gas plume was investigated using a custom-made Multi-component gas analyzer system (Multi-GAS) (Aiuppa et al., 2005; Shinohara, 2005). Gas measurements were carried out during 14–22 January 2016, when a fully autonomous Multi-GAS was temporarily installed on Pacaya summit to measure the in-plume concentrations of CO₂, SO₂, H₂S, H₂ and H₂O. The Multi-GAS was installed on the northwest rim of Mackenney cone (Lat, −14°22'55.48"N; Long 90°36'8.87"O), downwind the main direction of plume transport (Figure 2). The Multi-GAS was powered by an external (12 V, 40 Ah) battery, and acquired data at 0.1 Hz rate during 4 measurement cycles per day, each 30 minutes long (start time of the cycles, 00:00, 06:00 and 12:00, 18:00 GMT-time). SO₂, H₂S and H₂ concentrations were measured with specific electrochemical sensors (models TD2G-1A, TC4E-1A, TE1G-1A, respectively; all from City Technology). CO₂ concentrations were measured using an on-board spectrometer (Gascard EDI030102NG), while H₂O concentrations were derived from co-acquired temperature, pressure and humidity readings, as in previous work (Aiuppa et al., 2014, 2015). All data were acquired and stored in a Campbell scientific data logger (model CR6), and post-processed with the Ratiocalc software (Tamburello, 2015) to obtain ratios between volatile couples (Table 1). Ratiocalc automates the standard procedure (Aiuppa et al., 2005; Shinohara, 2005) of inferring volcanic gas ratios by taking the gradient of the best-fit regression lines in scatter plots of co-measured gas species concentrations. The error associated with each Multi-GAS derived ratio is ≤ ±20% for CO₂/SO₂ and ≤ ±30% for H₂O/SO₂ .

3.2 UV Camera

On 16 January, we used a dual-UV camera (same as in Tamburello et al., 2012) to acquire time-series of the volcanic SO₂ flux. Observations were carried out from an observational point situated on the northern volcano's flank (14°23'26.91"N, 90°36'5.91"O) at a distance of 940 m from the crater and 2,273 m elevation a.s.l. (see Figure 2) for 4 consecutive hours. The dual UV camera system used two co-aligned cameras (JAI CM 140 GR cameras, with 10 bit digitization and 1,392 × 1,040 pixels), with optical filters centred at either 310 nm (camera 1; affected by SO₂ absorption) or 330 nm (camera 2; not affected by SO₂ absorption). Absorbance A for each camera pixel was calculated as (Kantzas et al., 2010; Tamburello et al., 2012):

(1)

where IP and IB are, respectively, dark image-subtracted plume and background sky images taken with the two co-exposed cameras. Image sequences were post-processed using the Vulcamera software (Tamburello et al., 2011) to derive time-series of integrated SO₂ column amounts (ICAs) along a cross section of the plume. These were converted into SO₂ fluxes multiplying by a time series of plume transport speed, with the average of 8.2 ± 4 m/s, derived by an optical flow algorithm integrated in Vulcamera (see D'Aleo et al., 2016, for details).

3.3 Direct Sampling

During the field campaign, we collected gas samples from two low-temperature fumaroles (80°C), located on the northwest rim of the Mackenney cone. These fumaroles were aligned along a 15–20 m radial fracture trending away from the crater. For each of the two fumaroles, we sampled two glass flasks. Due to the ongoing Strombolian activity, access to the eastern part of the inner crater, where higher temperature fumaroles are possibly present, was not possible due to safety reasons. During sampling, a stainless steel tube was inserted into the fumarolic vent until thermal equilibration with the fumarole gases. Then, a 3-way valve was connected to the tube, a 100 ml syringe and a glass flask in order to collect the gases. The collected gas samples were analyzed for their chemical composition and δ¹³C of CO₂ at INGV (Table 2). The concentrations of He, H₂, O₂, N₂, CO, CH₄, and CO₂ were determined using a Perkin Elmer Clarus 500 gas chromatograph equipped with a 3.5-m Carboxen 1000 column and 2 detectors (a thermal conductivity and a flame ionization detector, respectively), with analytical errors of < 3%. The C isotope composition of CO₂ (expressed as δ¹³C ‰ versus V-PDB) was determined using a Thermo (Finnigan) Delta Plus XP CF-IRMS, connected to a Trace GC gas chromatograph and a Thermo (Finnigan) GC/C III interface. The gas chromatograph, equipped with a Poraplot-Q column (length 30 m, i.d. 0.32 mm), kept at a constant temperature of 50°C, uses He as the carrier gas. The analytical uncertainty was ±0.1‰. Due to the high extent of air contamination, samples were not analyzed for noble gases.

3.4 Chemistry of Rocks and Olivine Crystals

In January 2016, fresh rock samples were collected on the southern flank of Pacaya volcano from outcrops of the following eruptions: Pre-1961 (Cerro Chino cinder cone and Pacaya composite cone with undivided lava flow, (Eggers, 1971); hereafter referred as pre-1961), 1961, 1975, and 2010. All samples are basaltic lava flows, according to modern Pacaya behavior. Outcrops were selected from volcanics of the recent eruptive activity that are richest in olivines (based on previous results in Bardinzteff & Deniel, 1992; Matías et al., 2012; Rose et al., 2013; Wardman et al., 2012). Olivine phenocrysts are the most suitable for studying He, Ne, and Ar in fluid inclusions (FI), because they effectively retain the small helium atoms within the crystal's structure, thus avoiding (or reducing) any diffusive effect. In addition, olivines are among the first mineral phases crystallizing from a basaltic melt. The collected rocks were initially characterized for their whole-rock chemistry and mineral chemistry of olivine host crystals. Major and trace elements in whole rocks (supporting information Tables S1 and S2) were measured at the Activation Laboratories (Canada) using inductively coupled plasma–mass spectrometry (ICP-MS), in which rock samples digested under acidic attack were analyzed by a Perkin Elmer Sciex ELAN 6000, 6100, or 9000 ICP-MS device. Duplicate measurements indicate reproducibility was generally better than 3% for major-element analyses, 5% for highly incompatible elements (i.e., Rb, Ba, U, and Th), and 7% for rare-earth elements. Accuracy was quantified by comparing measured and certified values for standard references (NIST694, DNC-1, GBW, W-2a, SY-4, and BIR-1a).

Olivine phenocrysts were then handpicked, prepared on Crystal-Bond™ resin, and polished on one side in order to have a flat surface. Major elements (supporting information Table S3) were determined using a JXA-8200 combined WD/ED (energy-dispersive) electron microprobe (EMPA) at the High Pressure and High Temperature Laboratory of INGV, Sezione di Roma 1. The analytical conditions were a beam current of 7.5 nA, accelerating voltage of 15 kV, and beam size of 5 μm with counting times of 10 and 5 s for peaks and the background, respectively.

3.5 Noble Gas Isotopes in Fluid Inclusions

Olivine crystals were separated from the four rock samples previously fragmented and sieved in the fractions of 0.5 and 1 mm, with the aim of studying fluid inclusions trapped therein. Fluid inclusions are small (from tens to few microns in diameter) bubbles of fluid (generally gas) trapped within minerals during olivine growth at magmatic conditions. This means gases trapped in fluid inclusions represent an early stage of degassing, especially in olivine crystals that crystallize early from basaltic melts. The procedure of sample preparation and analysis is described in Di Piazza et al. (2015) and Robidoux et al. (2017). In samples from 1961, 1975, and 2010 eruptions, olivine crystals were separated in the two fractions (0.5 and 1 mm), while from pre-1961 only in the 0.5 mm fraction. Aliquots of 0.2 to 0.7 g were analyzed by single-step crushing in replicate measurements for elemental and isotope composition of helium, neon, and argon in fluid inclusions hosted in olivine crystals (Table 3). These analyses were performed at the noble gas isotope laboratory of INGV, Sezione di Palermo, following internal protocol (e.g., Di Piazza et al., 2015; Rizzo et al., 2015; Robidoux et al., 2017). Helium isotopes (³He and ⁴He) and ²⁰Ne were measured separately by two different split-flight-tube mass spectrometers (Helix SFT-Thermo). The ³He/⁴He ratio is expressed normalized to Ra (where Ra is the ³He/⁴He of air; i.e., 1.39 × 10⁻⁶). The analytical uncertainty of He isotope ratio measurements was ≤2.5%. The ³He/⁴He ratios were corrected for atmospheric contamination based on the measured ⁴He/²⁰Ne ratio (e.g., Sano & Wakita, 1985) and are expressed as Rc/Ra values. Argon isotopes (³⁶Ar, ³⁸Ar, and ⁴⁰Ar) were analyzed by a multicollector mass spectrometer (GVI Argus) at an analytical uncertainty of ≤0.9% (Table 3). The uncertainty in the measured amounts of He, Ne, and Ar elements was <5%. Typical blanks for He, Ne, and Ar were <10⁻¹⁴, <10⁻¹⁶ and <10⁻¹⁴ mol, respectively, and were at least 1–2 orders of magnitude lower than values measured in the samples.

4.1 Plume Composition and SO₂ Fluxes

During our observations on 14–22 January 2016, mild Strombolian activity persisted at a small pyroclastic cone inside the MacKenney crater (Figure 2). Larger explosions, interspersed within continuous degassing and spattering activity, produced <100m high columns that deposited scoriae and bombs, mostly inside the inner crater. No major change in activity was observed during our reporting period, except for a small collapse event of the north-western sector of the crater on January 20 that produced a gray ash-rich column (INSIVUMEH, monthly report January 2016). Direct observations indicated that intra-crateric Strombolian activity was somewhat more intense on 20–21 January, when some bombs reached the outer crater rim and damaged the outer case of the Multi-GAS.

The volcanic plume was dispersed toward north/north-west (e.g., toward the Multi-GAS measurement site) throughout most part of the Multi-GAS measurement interval, allowing regular plume detection (Figure 2). The maximum gas concentrations we measured in the Pacaya plume were 120 ppm (SO₂), 550 ppm (CO₂), 1.5 ppm (H₂), and 3,000 ppm (H₂O) (this latter is the excess H₂O content above an atmospheric background of 15,000–19,000 ppm). H₂S concentrations were typically below the sensor's 13% cross sensitivity to SO₂, and we therefore infer the Pacaya plume was typically H₂S-poor (at <1 ppm level). Occasionally, due to unfavorable plume transport directions, the Multi-GAS sensed a very dilute plume and/or the gas emitted from low temperature fumaroles in the inner crater's wall, close to the measurement site.

Figure 3 is a scatter plot of pairs of co-acquired CO₂ and SO₂ concentration data. The plot demonstrates that plume CO₂/SO₂ (molar) ratios fluctuated from <1 to >10 during the study period. This spread of CO₂/SO₂ ratios is evident from comparison of CO₂ versus SO₂ sub-populations taken in two distinct measurement intervals (red: 22 January; yellow; 18 January), in which the slopes of the best-fit regression lines are 1 and 8.2, respectively. In view of this temporal variability, we calculated (using Ratiocalc) the “mean” CO₂/SO₂ ratio in individual, few minutes-long, temporal intervals, during which the plume was compositionally more constant (e.g., the obtained correlation coefficients between CO₂ and SO₂ concentrations were high: R² from 0.60 to 0.91). The obtained CO₂/SO₂ ratios ranged from 0.6 to 10.5 (Table 1), and their average (arithmetic mean) was 3.0 ± 2.7 (1 standard deviation, SD). In Figure 4a, the obtained CO₂/SO₂ ratios are plotted against the peak SO₂ concentration measured in the corresponding temporal window (see Table 1). This peak SO₂ concentration is here taken as a “plume marker,” e.g., as a proxy for the density of the plume at the measurement site. We find that CO₂/SO₂ ratios are widely scattered and typically higher (mostly between 2 and 10.5) at low SO₂ (< 20 ppm) concentrations (dilute plume conditions), while they converge to low values (between 0.5 and 1.5) at higher peak SO₂ concentrations (dense plume conditions). A similar behavior is also observed for H₂O/SO₂ (Figure 4b), in which stable and low ratios (range 2–20; average ∼ 11) are obtained in dense plume conditions (peak SO₂ concentrations > 20 ppm). We also derived plume H₂/SO₂ ratios of 0.01 to 0.3, with a cluster at 0.04 ± 0.01 in the subset of high SO₂ concentration data. Although H₂S/SO₂ ratios were not detected, we inferred a maximum ratio of 0.05 based on H₂S being below detection in all acquisitions. SO₂ was therefore the only substantial S component in the plume (i.e., SO₂ ≈ Total Sulfur, S_t).

Figure 5a is a pseudo-color image of SO₂ column densities (in ppm·m) in the Pacaya plume, obtained from processing of couples of co-acquired images taken by our UV camera system in the morning of January 16, 2016. During the measurements, the plume was nearly transparent and vertically lofting above the crater (Figure 5b). Sequences of images such as that shown in Figure 5a were obtained every 3 seconds. These were processed to obtain time-series of SO₂ ICA, using a cross section orthogonal to the plume transport direction (see Figure 5a). From these, and by using a plume transport speed of 8.2 ± 4 m/s, the SO₂ flux time-series of Figure 5c was obtained. The derived SO₂ fluxes varied largely, from ∼100 to 4,210 tons/d, consistent with visual observations (from the crater rim) of fluctuations in explosive and degassing activity at the pyroclastic cone inside the MacKenney crater. We caution that part of the large SO₂ flux temporal variability was caused by passage of clouds between ∼16:50 and ∼17:20 GMT, which partially obscured view of the crater area and resulted into reduced SO₂ fluxes (Figure 5c). Excluding this subset of values, we obtain an averaged SO₂ flux of 885 ± 550 tons/d.

4.2 Chemical and Carbon Isotope Composition of Fumarole Gases

The chemistry of gases collected from the two low-temperature fumaroles displayed a strong air contamination, as demonstrated by the dominant O₂ and N₂ composition (19.3%, 76. 8% respectively; Table 2). CO₂ concentrations were between 0.9 and 2.8%. The δ¹³C of CO₂ varied in a narrow range, from −3.7‰ to −4.0‰. Based on theoretical values in air of CO₂ and δ¹³C (CO₂≈400 ppm and δ¹³C=-8‰), we calculated a fraction of contamination in our samples. Then, we used an isotope balance to quantify the effect of air mixing and recalculate the air-corrected magmatic gases δ¹³C. This procedure yields δ¹³C of CO₂ between −3.5 and −3.9‰ (Table 2).

4.3 Chemistry of Whole Rocks and Olivine Crystals

The whole-rock composition of the studied samples is basaltic, with SiO₂ contents between 50.2 and 50.7 wt. % and total alkalis (Na₂O + K₂O) between 3.8 and 4.1 wt%. The MgO content is generally comprised between 2.9 and 4.0 wt. %, while total Fe varies between 7.9 and 9.5 wt%. These results agree well with published data on the same eruptions (Bardinzteff & Deniel, 1992; Matías et al., 2012; Rose et al., 2013; Wardman et al., 2012), and confirm the basaltic composition of most of the products erupted from the volcano since historic time. In the K₂O versus SiO₂ classification diagram (Figure 6; after Peccerillo & Taylor, 1976), Pacaya volcanics spread over the typical compositional field of the calc-alkaline magmatic series. Rock samples investigated in this work are among the most mafic ever erupted at Pacaya and in Guatemala (Figure 6). Pacaya lavas also have relatively low fluid–mobile to less-fluid–mobile trace element ratios (Ba/La = 46.23 ± 2.2; U/Th = 0.42 ± 0.04; Ba/Th = 350 ± 18) (see also Heydolph et al., 2012), and plot therefore at the lower limit of the CAVA compositional range (Sadofsky et al., 2008). These low ratios suggest limited contribution from sediments in the subducted slab (Sadofsky et al., 2008).

Olivine phenocrysts (∼10 vol.%), separated for the noble gases investigation in fluid inclusions, have euhedral to sub-hedral habitus and display a broad range of compositions at cores (Fo₆₅–Fo₉₀), with an unimodal distribution (mode at Fo₆₅–Fo₇₃) that suggests the lack of clear mixing processes.

4.4 Noble Gases in Fluid Inclusions

The ⁴He content in olivine crystals ranges between 7.8 × 10⁻¹⁴ and 2.7 × 10⁻¹³ mol/g (Figure 7), ²⁰Ne varies from 1.2 × 10⁻¹⁵ to 8.2 × 10⁻¹⁵ mol/g, while ⁴⁰Ar between 2.2 × 10⁻¹³ and 2.7 × 10⁻¹² mol/g. No systematic difference in concentrations is observed among the samples. All the data fall in the concentration range of the 1975 samples (Figure 7). The ⁴He/²⁰Ne ratio is an indicator of atmospheric contamination and ranges between 21 and 113, which is at least 66 times higher than the typical atmospheric ratio (⁴He/²⁰Ne = 0.318). In Figure 8, we plot our data in order to evaluate the extent of air contamination. To do this we assume a binary mixing between an atmospheric end-member, having ⁴He/²⁰Ne=0.318 and R/Ra=1, and two magmatic end-members, having ⁴He/²⁰Ne=1,000 and Rc/Ra of either 8.3 or 9.0. These two mixing curves practically cover the entire dataset of fluid inclusions, indicating that the magmatic end-member varies in this range of ³He/⁴He ratio (8.3–9.0 Ra, Figure 8). The ⁴⁰Ar/³⁶Ar varies from 298 to 328, which is slightly higher than atmospheric (⁴⁰Ar/³⁶Ar = 295.5). Both ⁴He/²⁰Ne and ⁴⁰Ar/³⁶Ar ratios are lower than those typical of fluids from the upper mantle (⁴He/²⁰Ne>1,000 and ⁴⁰Ar/³⁶Ar>40,000; Ozima & Podosek, 1983), but fall within the range of subduction-related volcanism worldwide and in CAVA (Di Piazza et al., 2015; Fischer et al., 2005; Hilton et al., 2002; Rizzo et al., 2015; Robidoux et al., 2017). This indicates that gases released from FIs contain an atmospheric-derived component, perhaps related to the recycling of noble gases in the subducting slab.

We selected the subset of samples having ⁴⁰Ar/³⁶Ar >300 and calculated the air-corrected ⁴⁰Ar content assuming that all the measured ³⁶Ar is atmospheric in origin:

(2)

In relation (2), ⁴⁰Ar* is the air-corrected ⁴⁰Ar, and the “m” subscript indicates “measured.” The ⁴⁰Ar* was not calculated for the 2010 eruption samples (⁴⁰Ar/³⁶Ar < 300) and for one 1975 sample (for which the ⁴⁰Ar/³⁶Ar was not determined). The recalculated ⁴He/⁴⁰Ar* varies between 2.3 and 5.5, which is within the typical production ratio of the mantle (⁴He/⁴⁰Ar*=1–5; Marty, 2012; Ozima & Podosek, 1983). Considering that He is around 10 times more soluble than Ar in silicate melts, magmatic degassing would lead to an increase of ⁴He/⁴⁰Ar* especially in the late stages of degassing (e.g., Iacono-Marziano et al., 2010). Thus, the small variability of this ratio in our fluid inclusions suggests that olivine crystals were entrapped in a small pressure range.

The ³He/⁴He ratios corrected for atmospheric contamination vary from 8.4 to 9.0 Ra (Figure 7). These ratios are within the MORB-like range (R/Ra = 8.0 ± 1; Graham, 2002), and are sensibly higher than the four data previously available results for Pacaya (Figure 7). The reported ³He/⁴He ratios in the three volcanic gases collected in the 1990s (Sano & Williams, 1996; Goff et al., 2000) vary between 2.9 and 7.9 Ra, and a single R/Ra of 6.8 ± 1.5 Ra was obtained for the only available olivine phenocrysts measurement (Poreda & Craig, 1989). Excluding the highest value measured by Goff et al. (2000) (having a ³He/⁴He ratio of 7.9 Ra), the low ratios previously measured at Pacaya reasonably reflect the role of secondary processes, including air contamination or addition of radiogenic ⁴He in the hydrothermal system.

5.1 The Chemical Signature of Magmatic Gases at Pacaya

To the best of our knowledge, the only available volcanic gas composition for Pacaya was reported in (Goff et al., 2000) for 2 directly sampled fumaroles in 1992. Sano and Williams (1996) reported on δ¹³C, R/Ra, CO₂/³He information for one high-temperature Pacaya gas, but did not quote a complete gas analysis. The results of Goff et al. (2000) indicated a volcanic gas dominated by H₂O (∼ 90 mol. %), CO₂ (∼ 4–7 mol. %), and minor sulphur species. Unfortunately however, the low temperatures (84.5 and 342°C) and the high N₂ content (of the hottest sample) implied a non-trivial role of secondary processes, such as scrubbing of reactive volatiles by wall-rocks or hydrothermal fluids upon cooling, and mixing with atmospheric fluids (Fischer and Chiodini, 2015; Symonds et al., 2001). Our 2016 Multi-GAS results therefore, that were obtained on the summit crater plume directly released by a hot, erupting open-vent (Figure 2), provide the first assessment of the “magmatic” gas signature for Pacaya.

Our results (Table 1) suggest a variable gas composition during 14–22 January 2016 (Figures 4c and 4d). We find that the CO₂/SO₂ and H₂O/SO₂ ratios both varied widely, and that the two ratios were linearly correlated (Figure 9a). Using each couple of co-acquired ratios, we recalculate the molar proportions (in mol. %) of the three major volatiles in the magmatic gas phase. Validity of this procedure is based on the assumption that the above three species make together the totality of the source magmatic gas (a condition commonly encountered in volcanic gases; Fischer & Chiodini, 2015; Symonds et al., 1994). The derived compositions, shown in the triangular plot of Figure 9b, again show large variability, ranges being 38.4–96.8 mol. % (H₂O), 1.7–36.9 mol. % (CO₂) and 0.3–29.4 mol. % (SO₂).

Several processes may have concurred to determine the observed compositional variability. One possibility is that the observed gas changes responded to variations in the volcano's degassing/activity state. However, strombolian intra-crateric activity and degassing both persisted at rather steady levels in the MacKenney crater during our study period. Except for a brief period in January 20–21, when strombolian activity was somewhat more vigorous, no major change was observed that could have explained the large diversity in gas composition. We instead propose that the observed gas variations reflected time-changing gas contributions from two distinct gas components, which we refer to as H and M (see Figure 9). Two-component mixing is fully supported by the linear compositional array in the ratio/ratio scatter plot of Figure 9a. We find that all gas compositions taken in high-SO₂ “dense plume conditions” were systematically characterized by low CO₂/SO₂ and H₂O/SO₂ ratios (Figures 4 and 9). Since SO₂ is commonly recognized as a tracer of high-temperature magmatic degassing (see reviews by Fischer & Chiodini, 2015; Oppenheimer et al., 2014), we conclude that the SO₂-rich, H₂O-CO₂-poor end-member composition M is likely to most closely approximate the Pacaya magmatic gas component. High CO₂/SO₂ and H₂O/SO₂ compositions were instead systematically observed when the plume sensed by the Multi-GAS was more dilute (<20 ppm SO₂; Figure 4). In analogy with observations made at other open-vent volcanoes (e.g., Etna; Shinohara et al., 2008), we propose that this low-SO₂ component (H) likely reflected gas contribution from lower temperature, hydrothermal fumaroles that were widespread on the crater rim/inner crater wall around the Multi-GAS site. Hydrothermal fumaroles are typically H₂O-rich (Chiodini & Marini, 1998; Fischer & Chiodini, 2015) and exhibit high CO₂/S_t ratios due to scrubbing of reactive sulfur during reactions with hydrothermal rocks/fluids (Aiuppa et al., 2014, 2017). The Pacaya fumaroles reported in Goff et al. (2000) may well be representative of this hydrothermal component.

Based on the above, we calculate our best-guess for the Pacaya magmatic gas composition by averaging all results taken in “dense plume” conditions (peak SO₂ concentration > 20 ppm). We obtain a magmatic gas with 80.5 mol. % H₂O, 10.4 mol. % CO₂, and 9.0 mol. % SO₂, thus sitting at the H₂O-poor range of the population of volcanic arc gases worldwide (Aiuppa, 2015) (Figure 9b).

This unusual H₂O-poor gas composition could partially reflect our Multi-GAS measurements actually under-estimating the real magmatic gas H₂O abundances, due to in-plume H₂O condensation during transport and dispersion, from the active degassing vent to the measurement site. Although we have no arguments to rule out this possibility completely, the Multi-GAS site was only 80 m distant from the vent, implying that a short-lived plume was actually measured (e.g., the transport time, from vent to the Multi-GAS, is inferred at < 10 s, assuming a typical transport speed of 8.2 m/s). Alternatively, our results could imply a H₂O-poor magma source at Pacaya. Walker et al. (2003) analyzed a set of glass inclusions from the 1996–1999 activity of Pacaya, and found H₂O contents of ≈ 2 wt. % (range, 1.7-2.4). This is still within the normal range of H₂O abundances in arc magmas (Wallace, 2005; Wallace et al., 2015), but below (or, at least, at the lower range of) the observed glass inclusion H₂O contents in other Guatemalan volcanoes such as Agua (2.65-3.4), Atitlan (2.34-3.3), Fuego (up to 6 wt. %), and Santa Maria (1.5-3.5) (Roggensack, 2001; Sadofsky et al., 2008; Wehrmann et al., 2011). Based on these results, Walker et al. (2003) argued that the contribution of H₂O-rich slab fluids is likely more limited (0.1-0.3 wt. %) in the Pacaya mantle source region than underneath other Guatemalan volcanoes (0.5-0.7 wt. %). Clearly, additional near-vent gas observations, and possibly new glass inclusion data, are required to better confine the magmatic H₂O abundance at Pacaya.

5.2 Constraints on the Magmatic Source

Our inferred magmatic gas composition above implies a magmatic CO₂/S_t ratio for Pacaya of ∼1.1 (± 1). This relatively low ratio, falling at the lower range of high-temperature arc volcanic gases worldwide (0.2–9.2; Aiuppa, 2015; Aiuppa et al., 2017), deserves some discussion. Recently, Aiuppa et al., (2014) analyzed the CO₂/S_t ratio signature of magmatic gases from the Costa Rica – Nicaragua segment (SNVS) of CAVA, and found distinct along-arc variations between its southern (Costa Rica, CO₂/S ratios: ∼0.5-1) and northern (Nicaragua; CO₂/S_t ratios: ∼3) branches. Based on positive correlations between magmatic gas CO₂/S_t ratios and trace-element tracers (e.g., Ba/La and U/Th) in the source magmas, Aiuppa et al. (2014, 2017) ascribed the increasing carbon-rich compositions in Nicaragua to the involvement of increasing proportions of recycled C-Ba-U-rich slab sediment fluids in magma genesis, compared to more mantle-like gas signature in Costa Rica.

Our novel results from Pacaya, with recent magmatic gas data from El Salvador (Aiuppa et al., 2017; Granieri et al., 2015), allow extending scrutiny of the gas versus trace element relation further to the north (Figure 10). Overall, these novel results corroborate the hypothesis of Aiuppa et al. (2014), indicating that, at the scale of the entire arc (CAVA), the linear correlation between gas CO₂/S_t ratios and Ba/La ratios (Figure 10a) is statistically significant (R²=0.7), and follows the best-fit regression line:

(3)

where CO₂/S_{t(m, gas)} is the mean CO₂/S_t ratio in magmatic gases for each volcano (data from Aiuppa et al., 2014, 2017; Granieri et al., 2015); (Ba/La_m,rock) is the corresponding mean Ba/La ratio derived from published whole-rock analyses (extracted from the earthchem portal; http://www.earthchem.org/portal) and 0.03 and 0.07 are best-fit regression coefficients.

We find that the new Pacaya (and El Salvador) gas data allow filling the compositional gap between Central Costa Rica (e.g., Poas and Turrialba) and Nicaragua (Figure 10a). As such, these results are suggestive for the Pacaya magmatic gas having intermediate origin between mantle-like and slab-like end-member compositions in Costa Rica and Nicaragua, respectively. A higher mantle affinity of Pacaya magmas, relative to more sedimentary-derived signature of Nicaraguan magmas, is well consistent with available trace element and isotopic signatures of CAVA volcanites (Figure 10b; Carr & Feigenson, 1990; Leeman et al., 1994; Morris et al., 1990; Patino et al., 2000; Sadofsky et al., 2008), based on evidence from Cl isotopes, low U/Th ratios and ¹⁰Be/⁹Be in volcanic rocks, similarly suggested a slightly contaminated mantle source for Guatemala.

5.2.1 C and He Isotopes

Further and even clearer evidence on the mantle affinity of Pacaya magmas come from the ³He/⁴He ratios measured in fluid inclusions hosted in olivine crystals separated from the studied rocks. In fact, ³He is typically mantle-sourced while ⁴He is crustal and comes from the radiogenic decay of U and Th. This implies that strong contribution of mantle-derived gases can be detected in terms of strong increase of ³He/⁴He up to values higher than MORB range (i.e., >9 Ra), in case of lower mantle (i.e., plume) contribution (e.g., hot spot geodynamic setting). Our measured Rc/Ra vary between 8.3 and 9.0 Ra (Figure 7), which fall at the upper limit of MORB-like range (R/Ra = 8 ± 1; Graham, 2002). More importantly, our ³He/⁴He ratios are the highest ever measured in CAVA and are among the highest ever recorded in arc volcanoes worldwide (Hilton et al., 2002; Oppenheimer et al., 2014, and references therein). Pacaya magmas thus point to a non-contaminated (by slab fluids) mantle wedge source beneath this segment of the volcanic arc, with no evidence of crustal contamination. According to Cameron et al. (2003), the southern portion of Guatemala undergoes E-W crustal extension, as evidenced by a series of N-S trending grabens (Burkart & Self, 1985). At Pacaya in particular, extensional regime is determined by the volcano laying at the intersection between the JFS and the Guatemala graben (Figure 1). Extensional forces are thought to control the degree of melting, and to facilitate rapid magma ascent, causing limited magma differentiation (Cameron et al., 2003). Modern Pacaya has erupted lavas of progressively more mafic composition, and uniquely basaltic rocks since historic time (Bardintzeff and Deniel, 1992). Another important aspect to consider is the role of decompression melting in south-eastern Guatemala. Cameron et al., (2003), based on major, trace elements and Sr isotopes in volcanic rocks, identified systematic differences between Pacaya and Agua (where decompression melting may play an important role) and other volcanoes in south-eastern Guatemala (i.e., Tecuamburro and Moyuta). They argued that Pacaya volcanics have compositions similar to back-arc cinder cones in southeastern Guatemala. These observations support the idea that the exceptionally high ³He/⁴He measured at Pacaya may reflect fast magma ascent from the mantle, limiting crustal contamination by radiogenic ⁴He. Accordingly, the ³He/⁴He signature of the mantle wedge below this segment of Guatemala is not modified by subduction-related fluids.

The corrected δ¹³C (CO₂) values of fumaroles sampled in this work are −3.5 and −3.9‰, respectively (cfr. 4.2). Considering the natural variability of the magmatic system, these values are comparable to the δ¹³C values reported by Goff et al. (2000) (-2.8‰ and −3.1‰) for two fumaroles (at 84.5 and 342°C) sampled in the early 1990s. On the contrary, Sano and Williams (1996) reported a very low δ¹³C (-6.9‰) for a high-temperature volcanic gas sample (965°C) from Pacaya. Because Sano and Williams (1996) did not report on the extent of air contamination of this sample, we cannot evaluate if the low δ¹³C was the effect of a strong air dilution. In fact, a high extent of air contamination on a magmatic δ¹³C of −3‰ would lead to more negative ratios (air δ¹³C=-8‰). Alternatively, their low δ¹³C (CO₂) could reflect a residual (degassed) magmatic source (see below). As detailed below, δ¹³C of CO₂ may suffer an isotope fractionation during magmatic degassing that leads to more negative ratios. For the above reasons, we discarded this low ratio, and we hereafter assume an average δ¹³C = −3.5 ± 0.4‰ as representative of the magmatic signature of carbon at Pacaya. We caution that, in contrast to ³He/⁴He ratios that can be directly used to constrain the origin of fluids, ¹³C/¹²C ratios are more problematic to interpret, owing to the existence of isotopic fractionations during magmatic degassing and gas-water interactions. Experiments demonstrate that, in basaltic melts, C isotopes have vapour-melt enrichment factors (ɛ_CO2-melt) of 2 to 4‰, e.g., ¹³C is preferentially extracted from melt. Because of this, the δ¹³C signature of magmatic CO₂ decreases along both open- and closed-system magma degassing paths (Javoy et al., 1978; Mattey, 1991; Trull et al., 1993). Unfortunately, no independent constraint is available at Pacaya to quantify the degassing extent of the magma source feeding our Pacaya fumaroles. In addition, a possible gas-water-rock interaction control on δ¹³C (and CO₂/³He) cannot be ruled out in our Pacaya example, in view of the low fumarole temperatures (at which scrubbing is important; Symonds et al., 2001). It follows from the above that caution is needed when interpreting δ¹³C and CO₂/³He variations at any single volcano or along an arc segment. These tracers may not necessarily provide direct access to the relative slab versus mantle contributions to magmatism (Sano & Marty, 1985). In the limit-case assumption that secondary processes are negligible, the magmatic signature in the gaseous phase of Pacaya melts would have δ¹³C of −3.5 ± 0.4‰. This is in the range of MORB-like mantle, whose inferred δ¹³C range is between −8‰ and −4‰ (Deines, 2002; Javoy & Pineau, 1991; Macpherson & Mattey, 1994; Marty & Jambon, 1987; Sano & Marty, 1995).

In order to put our novel ³He/⁴He-δ¹³C results in the context of CAVA (Figure 11), we need to make some general considerations on data selection. Detailed ³He/⁴He-δ¹³C data sets are available for CAVA in comparison to other arc segments (e.g., Fischer et al., 2007; Hilton et al., 2002; Oppenheimer et al., 2014, and references therein). However, the measured range is large for both ³He/⁴He (0.7–8.1 Ra) and δ¹³C (from −14‰ to 0‰). Large He and C variations are observed not only among different CAVA volcanoes, but also within the same volcanic system (Fischer et al., 2007). This variability is largely an effect of the disparate use of different gas (high- and low-temperature crater fumaroles, flank gases, geothermal springs, and distal gas emissions) and sample (FIs in olivine phenocrysts versus gas emissions) types. In the majority of CAVA volcanoes analyzed so far (Di Piazza et al., 2015; Fischer et al., 2005; Robidoux et al., 2017; Shaw et al., 2006), the ³He/⁴He ratios in olivine-hosted fluid inclusions are typically higher than in (or at the upper range of) volcanic gases from the same volcanic system. FIs should thus be prioritized as more reliable indicators of the magmatic/mantle signature, where available. ¹³C/¹²C ratios are currently difficult to measure in olivine-hosted fluid inclusions, except in mantle xenoliths or in phenocrysts crystallized at high depth (Correale et al., 2015; Deines, 2002; Gennaro et al., 2017). Thus, volcanic gases remain in most cases the only source of information, but should be limited to sample types showing greater magmatic affinity (e.g., high-temperature crater fumaroles should be preferred to hydrothermal or distal gas emissions).

In view of the above, our selected CAVA ³He/⁴He-δ¹³C population includes, for each volcano, the maximum Rc/Ra values measured (in FIs where available), and the mean δ¹³C of crater fumaroles only (Figure 11). This selected CAVA subset (Table 4) gives the opportunity to make some considerations on along-arc variations of these tracers. Before our study, the highest ³He/⁴He values along CAVA had been measured at Turrialba (Costa Rica), with values up to 8.1 Ra in fumarole gases (Di Piazza et al., 2015; Hilton et al., 2010; Shaw et al., 2003; Tassi et al., 2004; Vaselli et al., 2010) and up to 8.3 Ra in fluid inclusions in olivines (Di Piazza et al., 2015; Fischer et al., 2005; Shaw et al., 2006). Slightly lower ratios (up to 7.6 Ra; Fischer et al., 2015; Hilton et al., 2010; Poreda & Craig, 1989; Shaw et al., 2003), but still within the MORB range, have also been measured in crater fumaroles from the nearby Póas volcano (Fischer et al., 2015; Hilton et al., 2010; Sano & Williams, 1996; Shaw et al., 2003) (Figure 11). These high ratios have widely been interpreted as representative of the signature of the mantle wedge beneath this segment of CAVA, with no (or very limited) contribution from the subducting Cocos plate lithosphere. This MORB affinity of Costa Rican volcanism is supported by several independent geochemical indicators such as the low Ba/La, Ba/Th, and U/Th ratios in the erupted rocks (Patino et al., 2000; Sadofsky et al., 2008). The few measurements carried out in El Salvador at Santa Ana volcano gave a ³He/⁴He up to 7.6 Ra (de Leeuw et al., 2007), while in Nicaragua the ³He/⁴He ration ranged from 7.2 Ra at San Cristobal (Robidoux et al., 2017) to 7.6 at Mombacho (Shaw et al., 2003) (Figure 11). In consideration of the high ³He/⁴He we measured at Pacaya, Figure 11 strongly reinforces the hypothesis of a mantle signature with no appreciable addition of radiogenic ⁴He from the slab or the crust in SE-Guatemala, similar to Southern Costa Rica. The increased sediment input from the subducting slab beneath Nicaraguan segment of CAVA, highlighted by high Ba/La and U/Th ratios (Sadofsky et al., 2008), only leads to a minor decline of the ³He/⁴He ratio.

Table 4. Geochemical Tracers, Isotopic Carbon Composition and Rc/Ra for Selected Sub-data set of CAVA Segment

Volcano	U/Th	Ba/La	CO2/3He	δ13C‰	Rc/Ra
Pacaya	0.45	44	1.89E+10a, b	−3.46a, c	8.65c, d
Tecuamburro				−3.0e	5.96e
Zunil			1.14e+10b	−2.86b	6.32b
Guatemala	0.45	44	1.89E+10a, b	−3.33b	8.65c
Santa Ana	0.50	60	2.07E+10f	−2.55f	7.56f, g
El Salvador	0.50	60	2 .07E+10	−2.55f	7.56f
Masaya	0.80	85	2.30E+10h	−1.50h	7.23g, h
Cerro Negro	0.84	118.71	2.01E+10	−2.72b, h, i	7.38g, i
Momotombo	0.90	110	2.70E+10h	−2.60h,j	7.40d, h
Mombacho	0.67	59.19	1.78E+10h	−3.25h	7.60g, h
San Cristobal	0.90	90	3.20E+10h	−2.95h	7.20d, j
Nicaragua	0.82	92.58	2.40E+10h	−2.60b, h, i	7.60h
Poas	0.30	26	2.02E+10k, l	−3.54b, k, l	7.60k, l
Turrialba	0.30	20	1.41E+10h, m, n	−3.01h, m, n	8.30d, m
Costa Rica	0.30	23.00	1.72E+10h, m, n	−3.27h, m, n	8.30g, m
DMM	0.35	7	2.00E+09o	−4.50o	8 ± 1 Rap
Organic carbon			1.00E+13	−25	0.05
Carbonate sediments	1.63	244.36	1.00E+13	0	0.05

• Note. U/Th and Ba/La ratios are average whole-rock values calculated from rock compositions extracted from the earthchem portal (http://www.earthchem.org/portal). See also Aiuppa et al. (2014) for reference (Uncertainties in the mean values are given as error bars in Figure 10). CO₂/³He and δ¹³C are mean of fumaroles reported in Figure 11 as blue x symbol. Rc/Ra are air-corrected ³He/⁴He ratios normalized to the atmospheric ³He/⁴He ratio (Ra). The maximum value measured in each volcanic system is reported (except for end-members).
• ^a Goff et al., (2000).
• ^b Sano and Williams (1995).
• ^c This work.
• ^d Fluid inclusion.
• ^e Janik et al. (1992).
• ^f Leeuw et al. (2007).
• ^g Fumaroles. Average values for nations are the same reported in Figure 11.
• ^h Shaw et al. (2006).
• ⁱ Lucic et al. (2014).
• Allard (1983).
• ^j Robidoux et al. (2017).
• ^k Fisher et al. (2015).
• ^l Hilton et al. (2002).
• ^m Tassi et al. (2004).
• ⁿ Vaselli et al. (2010).
• ^o Sano and Marty (1996).
• ^p Graham (2002).

As for δ¹³C (Table 4 and Figure 11), the Pacaya δ¹³C signature (-3.5 ± 0.4‰) approaches the average isotope compositions of Costa Rican (δ¹³C −3.1 ± 1.3; Fischer et al., 2015; Hilton et al., 2010; Sano & Williams, 1996; Shaw et al., 2003; Tassi et al., 2004; Vaselli et al., 2010) and Guatemalan (δ¹³C = −3.3 ± 0.4‰; Goff et al., 2000; Janik et al., 1992; Sano & Williams, 1996; this work) volcanoes, and reflect high mantle contributions. Slightly more positive signatures are observed in El Salvador and Nicaragua (δ¹³C = −2.9 ± 0.6‰ and −2.7 ± 0.6‰, respectively; Allard, 1983; Sano & Williams, 1996; Shaw et al., 2003; Snyder et al., 2001) that could be compatible with an enhanced limestone contribution (δ¹³C = 0‰; Sano & Marty, 1995) with possible addition of organic carbon up to 10% in the central CAVA segment. This is shown in Figure 11 by the dashed mixing curve, which supports the hypothesis that this is a mantle feature of the subducting slab common to most of the volcanoes of CAVA.

5.3 Volatile Fluxes

Our results also set the basis for a first assessment of the total volatile flux from Pacaya. Our measured SO₂ flux on 16 January 2016 was 885 ± 550 tons/d (Figure 5c), thus falling in the middle of previously reported SO₂ fluxes for the volcano. These have fluctuated substantially since onset of the ongoing eruptive period in 1965. Published SO₂ fluxes have remained at low levels in the 1970s–1990s, from 260 tons/day (1973–1993 mean; Andres et al., 1993) to 510 tons/day (1972–1997 mean; Andres & Kasgnoc, 1998). They then apparently peaked at 1,350–1,540 in the 1999–2002 period (GVN Bullettin, 2002; Mather et al., 2006; Rodríguez et al., 2004), to decrease again in the mid- to late-2000s, from 270–400 tons/day (in 2004–2005; Bluth et al., 2007) to <120 tons/day (in 2008; Dalton et al., 2010). Our 2016 data therefore suggest intensification of degassing activity relative to the low fluxes seen in the most recent (2004–2008) published reports.

We combine our SO₂ flux results with the above best-guess for the Pacaya magmatic gas composition (80.5 mol. % H₂O, 10.4 mol. % CO₂, and 9.0 mol. % SO₂) to infer the H₂O and CO₂ fluxes for Pacaya for the first time. So doing, our SO₂ fluxes and compositions convert into H₂O and CO₂ fluxes of 2,230 ± 1,390 and 700 ± 440 tons/day for 16 January 2016. The large associated uncertainties reflect the huge temporal variability of SO₂ fluxes (62% variation at 1 SD) during our measurement interval, and underscore for the need of longer, more systematic observations to better constrain the volcano's degassing strength. We also infer a H₂ flux of ∼1.1 ± 0.7 tons/day based on a H₂/SO₂ of 0.04 ± 0.01. Based on our results, the Total Volatile Flux in January 2016 was ∼3,800 ± 2,375 tons/d (Table 5), ranking Pacaya as a medium volcanic arc emitter. For example, our measured CO₂ flux ranks Pacaya at #18 in the list of the #44 volcanoes for which CO₂ information is available (Aiuppa, 2015; Burton et al., 2013).

5.4 Implications for the Volcanic CO₂ Output in Guatemala

We use our novel results for Pacaya to attempt at a preliminary characterization of the CO₂ output from the Guatemalan segment of CAVA. Refining the arc CO₂ output, both globally and at the scale of individual arc segments, is mandatory for better constraining the CO₂ arc budget, e.g., to infer the fraction of subducted carbon (with slab sediments and altered oceanic crust) that is recycled via volcanic arc magmatism and, by difference, the fraction transported back into the deep mantle (Dasgupta, 2013; Kelemen & Manning, 2015). Owing to the lack of magmatic gas data, the Guatemalan contribution to the CAVA CO₂ output has only been indirectly established in past studies, by using either gas information from other, better studied segments of the arc (e.g., Hilton et al., 2002), or a “mean arc gas CO₂/SO₂ ratio” proxy (Mather et al., 2006).

In addition to Pacaya, Fuego and Santiaguito (Figure 1) are the two other Guatemalan volcanoes showing sustained, persistent degassing activity (Rodríguez et al., 2004). In the classic SO₂ flux compilation study of Andres and Kasgnoc (1998), the two volcanoes had reported time-averaged (1972–1997 period) SO₂ emissions of 640 and 230 tons/day, respectively. For comparison, Rodríguez et al. (2004) reported SO₂ flux data for the 1999–2002 period being twice as low (337 ± 145 and 127 ± 58 tons/day, respectively). These data confirm Pacaya as the strongest degassing source in Guatemala, but also imply a not trivial CO₂ contribution from Fuego and Santiaguito.

The volcanic gas CO₂/SO₂ ratio signature and the CO₂ output are unfortunately undetermined for Fuego and Santiaguito. However, plenty of whole-rock information is available for magmas erupted at both volcanoes. Importantly, trace-element co-variation diagrams (e.g., Figure 10b) suggest magmas erupted at Fuego and Santiaguito plot along the same compositional array of other CAVA magmas (Sadofsky et al., 2008, and references cited therein). This opens the possibility to estimate the magmatic gas CO₂/SO₂ ratio signature of Fuego and Santiaguito based on the CAVA gas (CO₂/S_t ratio) versus trace-element (Ba/La ratio) relationship of Figure 10a. Using equation 3, and with the specific Ba/La_m,rock derived in earthchem for each volcano (Fuego: 55.1 ± 10.6; Santiaguito: 52.8 ± 9.7), we derive CO₂/S_{t(m, gas)} ratios of respectively 1.5 ± 0.75 and 1.4 ± 0.75. Here, the confidence interval or delta (± 0.75), calculated from the regression line and one standard deviation (Figure 10a), is taken as a proxy for the uncertainty in predicted CO₂/S_{t(m, gas)} ratios.

We caution the above procedure is likely to provide only a rough approximation of the real magmatic gas CO₂/S_t ratio for each volcano, and that direct volcanic gas observations remain needed here. Yet, we can use our predicted CO₂/S_t ratios above (1.5 ± 0.75 and 1.4 ± 0.75) in tandem with time-averaged SO₂ fluxes (337 ± 145 and 127 ± 58 tons/day, respectively), to preliminary evaluate the CO₂ flux at respectively 347 ± 375 (Fuego) and 122 ± 121 (Santiaguito). With these numbers, and considering the Pacaya CO₂ flux determined above, the total volcanic CO₂ flux from Guatemala would be ∼1,160 ± 600 tons/d. This would respond to ∼30% of the CO₂ flux from the Costa Rica – Nicaragua Volcanic Segment (CNSV: 2,835–3,184 tons/d, (Aiuppa et al., 2014), and ∼16% of the total CAVA output (∼6,932 tons/d; Hilton et al., 2002).