Unrest at the Nevados de Chillán volcanic complex : a failed or yet to unfold magmatic eruption ?

New eruptive activity at volcanoes that have been long quiescent poses a significant challenge to hazard assessment, as it requires assessment of how the situation may develop. Such incipient activity is often poorly characterised as most quiescent volcanoes are poorly monitored, especially with respect to gas geochemistry. Here, we report gas composition and flux measurements from a new vent at the onset of eruptive activity at the Nevados de Chillán volcanic complex (Chile) in January-February 2016. The molar proportions of H2O, CO2, SO2, H2S and H2 gases are found to be 98.4, 0.97, 0.11, 0.01 and 0.5 mol % respectively. The mean SO2 flux recorded in early February 2016 during periods of eruptive discharge amounts to 0.4–0.6 kg s−1. We show that magmatic gases were involved in this activity, associated with a sequence of eruptions. Tephra ejected by the first blast of 8 January are dominated by lithic fragments of dacitic composition. By contrast the tephra ejected from a subsequent eruption contains both lithic fragments of dense dacite, and a fresher, sparsely vesicular material of basaltic andesite composition. By October 2017, the ejected tephra was again dominated by dense dacitic lithic material. Together with seismic and ground deformation evidence, these observations suggest that a small intrusion of basaltic to andesitic magma at shallow level led to the explosive activity. Our serendipitous survey, right at the onset of eruptive activity, provides a valuable window into the processes of reawakening of a dormant volcano.


Introduction
Nevados de Chillán, a large volcanic complex built in the Southern Volcanic Zone of the Chilean Andes is formed along a 12 km northwest-trending ridge (Figure 1. It is considered one of the most hazardous volcanoes in Chile due to the proximity of the resort towns of Las Trancas and Termas de Chillán-approximately 10 and 5 km away from the active crater, respectivelywith permanent populations of 1600 rising to 30,000 during the high season. Whilst Holocene activity is represented by widespread pyroclastic flow and tephra fall deposits around the volcano, lahars associated with snow melt are considered to be the greatest potential hazard [Orozco et al. 2016]. In particular, the Las Trancas valley is covered with repeated sequences of lahar deposits separated by paleosols and intercalated with centimetre-thick pyroclastic flow deposits [Carrasco and Andrés 2012]. The latest hazard map shows high lahar threats extending more than 50 km away from the complex [Orozco et al. 2016  been the most active. This subcomplex comprises four cones, Nuevo, Arrau, Viejo and Chillán (from north to south; Figure 1). Viejo is the oldest stratocone, it was active from ca 9.3 ka to 2270 BP and is composed of interstratified lavas and pyroclastic units that include prominent densely welded andesite and dacite agglutinate layers [Dixon et al. 1999]. Chillán cone is located south-west of Viejo and partially overlaps it. It is a dacitic cone dominated by lavas intercalated with pyroclastic deposits, and its last eruption took place in 1883 [Brüggen 1948] possibly associated with the collapse of the southern flank of the edifice [Naranjo et al. 2008]. Nuevo and Arrau are dacitic lava cones that formed on top of the older volcán Democrático. Nuevo formed from 1906 to 1943 while Arrau formed from 1973 to 1986 [Deruelle 1977;Dixon et al. 1999;Naranjo et al. 1994]. From August to September 2003, a series of lowmagnitude explosive events generated gas and ash columns 400 to 500 m high, leaving a 64 m double crater in the saddle between the Nuevo and Arrau cones [Naranjo and Lara 2004]. On 29 January 2009, the Volcanic Ash Advisory Centre of Buenos Aires reported a small ash column rising 500 m above the volcanic complex. This eruption could not be confirmed by subsequent field observations, which found instead that an unnoticed eruption must have occurred between January and August 2008, producing a new lava field termed Volcán Sebastián, 1 km northeast of the Arrau cone [Naranjo and Moreno 2009].

The 2015-present unrest
In December 2015, the Observatorio Volcanológico de Los Andes del Sur (OVDAS), part of the Servicio Nacional de Geología y Minería (SERNAGEOMIN), changed the warning level from green to yellow following an increase in seismicity observed during the month prior. On 8 January 2016 at 17:56 local time, an eruption occurred at Nevados de Chillán producing a small column of ash. On 13 January 2016, the authors observed a new vent, 30 m across ( On 29 January 2016, a second vent, 25-30 m wide, opened, followed by a third in early February 2016 along a NNE trend Figure 2B-C). The proximity of these vents precluded unambiguous identification of which was responsible for the frequent small eruptions. On 9 May, 8 August and 1 September 2016, three comparatively larger eruptions (in terms of plume height) occurred, producing ash clouds up to 2000 m above the vent and resulting in a widening of the 8 January vent Figure 2D). A fourth vent opened in October 2016 (Figure 2E), eventually merging with the first and third in March 2017 to produce a large crater more than 100 m across. This is the likely source of the activity occurring at the time of writing (December 2017). From March 2016, incandescence has been sporadically reported during night-time eruptions, and sustained weak incandescence was reported between 25 and 31 March 2017 [SERNAGEOMIN 2017]. All eruptions have been characterised by columns of gases and ash of low height (<2 km) with most of the erupted material deposited within 1 km of the crater.

In-situ gas measurements
Gas composition data were obtained on 13 January 2016, five days after the first eruption. The data were collected using a portable "Multi-GAS" instrument [Shinohara 2005] deployed a few meters downwind of the first vent, which was actively degassing at the time (S36°52'1.64"; W71°22'40.18"; Figure 2A). The instrument incorporated SO 2 , H 2 S and H 2 electrochemical sensors. The SO 2 and H 2 sensors have calibration ranges of 0-200 ppmv while the H 2 S sensor has a calibration range of 0-100 ppmv. A non-dispersive infrared sensor was used for CO 2 and calibrated for 0-10,000 ppmv with an accuracy ±2 %. A relative humidity (Rh) sensor (Galltec) was used to measure H 2 O, providing a measuring range of 0-100 % Rh with an accuracy of ± 2 %. The conversion from relative humidity to water mixing ratio was made following Buck [1981] and using the following equation: where H 2 O is the absolute water mixing ratio in ppmv, T is temperature in°C, Rh is relative humidity in %, and P is atmospheric pressure in mbar. The gas temperature used in this equation is measured in real time by the Multi-GAS, the pressure is also measured by the Multi-GAS and the average during the measurements is used. All sensors were housed inside a weatherproof box, with the ambient air sampled via Teflon tubing connected to a HEPA filter fed through an inlet in the box and circulated via a miniature 12V rotary pump through the sensors. An on-board data-logger captured measurements at a rate of 1 Hz. The complete system was powered by a small (6 Ah) 12V LiPo battery. Similar systems have now been deployed at many volcanoes and the system used here is the same as that reported in Moussallam et al. [2016].
All sensors were calibrated in the laboratory at INGV Palermo (in October 2015), with target gases of known amount. The differences in response time for the different sensors were corrected by finding the lag times from correlation analysis of the various time series (see Moussallam et al. [2014] for sensors response time). Post processing was performed using Ratiocalc [Tamburello 2015]. Below we report H 2 O, H 2 and CO 2 mixing ratios after correction for mean ambient air mixing ratios (measured by the Multi-GAS directly prior to entering the plume). The measured H 2 S mixing ratio is corrected for a laboratory-determined cross-sensitivity with SO 2 gas (amounting to 16%).
The ultraviolet cameras were coupled to Pentax B2528-UV lenses, with focal length of 25 mm (FOV 24°), and 10 nm full width at half maximum (FWHM) bandpass filters were placed immediately in front of each lens one filter was centred at 310 nm (Asahi Spectra XBPA310) where SO 2 absorbs and the other at 330 nm (XBPA330) outside the SO 2 absorption region [Kantzas et al. 2010;Mori and Burton 2006]. Image acquisition and processing were achieved using Vulcamera [Tamburello et al. 2011]. Every image acquired is saved in a 24 bit Portable Network Graphics (png) file with lossless compression. A set of four SO 2 calibration cells were used (94, 189, 475 and 982 ppm·m) to calibrate the apparent absorbance [Kantzas et al. 2010]. Two parallel sections in our data series, perpendicular to the plume transport direction were used to derive plume speed (ranging from 3 to 7 m s −1 , with an average of 6 m s −1 ). The data processing was carried out following the protocols outlined Kantzas et al. [2010].

Tephra analysis
Two tephra samples were collected on 13 January 2016 at S36°52'8.1097"; W71°22'38.1298" and S36°52'24.6853"; W71°22'41.7828". Both samples were collected on snow that was clean before the 8 January eruption and hence originate from ash fall produced by the first eruption. The first sample consists predominantly of fine ash (< 63 µm), while the second consists predominantly of coarse ash (< 2 mm). A third tephra sample was collected on 3 February 2016 at: S36°53.291' W071°23.606' from a patch of snow that was clean on 13 January 2016. The sample hence originates from ash fall from an eruption that occurred between 14 January and 3 February 2016. A fourth tephra sample was collected on 11 October 2017 at S36°54.808' W071°29.574' directly from ash fall in the village of Las Trancas. Given that the samples were collected opportunistically, and in very small amounts, no attempt to quantify grain sizes was attempted.
Tephra samples were analysed for bulk composi-    nA, and a defocussed beam of 10 µm. To mitigate for Na loss, Na was measured first in the analytical sequence, at reduced count times (10 s on peak; 5 s background) at a fixed peak position. Major elements were standardized against rhyolitic (VG568) and basaltic (VGA-99) glasses and pure oxides of Ti, Mn and Cr [Jarosewich et al. 1980]. Analytical spots were chosen to avoid microlites, thus capturing evolved interstitial glasses.

Gas composition
We obtained 12 minutes of high quality Multi-GAS measurements. Data acquisition time was restricted due to the unstable nature of the vent area (requiring rope access), and the need to limit time exposed in a hazardous area. Multi-GAS measurements are presented in Figure 4, which shows four scatter plots of CO 2 ; H 2 O; H 2 S and H 2 vs. SO 2 mixing ratios in the plume emitted from the first vent. The strong positive co-variations observed between SO 2 and the other detected volatiles confirm a single, common, volcanic origin. The gas/SO 2 molar plume ratios were obtained from the gradients of the best-fit regression lines. Scatter plots yield the following molar ratios: CO 2 /SO 2 of 9.

SO 2 flux
The ultraviolet cameras were deployed at distances of 2.5 km (03 February 2016), 2.7 km (31 Jan. 2016), 3 km (13 Jan. 2016), 5 km (02 February 2016) and 8 km (30 Jan. 2016) from the source. Weather conditions during these measurements were generally good, with long periods of clear sky, except on 3 February, which was cloudy. Results obtained over these 5 days of measurements indicate a negligible SO 2 flux outside eruptive episodes. For instance, on 13, 30 and 31 January there were no eruptions during the observation periods, and the recordings did not register any SO 2 release, despite the generally short distances between the UV cameras and the active crater. Eruptions did occur during measurement periods on 2 and 3 February. However, due to heavy cloud on 3 February, only the data collected on 2 February are useful. On that day, UV camera measurements were taken downwind of the active crater after most of the ash in the eruptive plume had deposited. Figure 5 shows the SO 2 emission rate obtained during two eruptions on 2 February. The emissions fluctuate between 0.1 and 1 kg s −1 with a mean value of 0.5 kg s −1 . These fluctuations of SO 2 emission rate are associated with observable successive pulses during the eruptive discharge period. High resolution videos of eruptions on 2 and 3 February are given in the supplementary material.

Tephra composition
Bulk tephra compositions are given in Table 3. Clear differences can be seen between the bulk composition of the tephra emitted during the first eruption on 8 January and that of the subsequent ash fall (occurring between 13 January and 3 February). In the total alkalisilica diagram, 8 January tephra are dacitic, while the subsequent tephra are andesitic Figure 6A). When examined by backscattered electrons (BSE), both samples of the 8 January tephra are similar, and consist of glassy, dense material with 30-40 % plagioclase microlites ( Figure 6B). In all samples, some of the dacitic fragments are bounded by corroded vesicles containing vapour-phase SiO 2 polymorphs ( Figure 6D). Microstructural phase identification of these SiO 2 crystals is beyond the scope of this work, but many of them show the "fishscale cracking" that is diagnostic of α-cristobalite that has undergone volume contraction during the β − α transition [Horwell et al. 2013]. These same ash fragments also have devitrified groundmass containing SiO 2 polymorphs ( Figure 6D) that we have not structurally identified, but that are similar to groundmass cristobalite in the holocrystalline cores of effusive silicic lava bodies [Schipper et al. 2015].
The small amount of cristobalite (and SiO 2 assumed to be cristobalite) is all intimately bound to glass and/or other crystal phases, and is therefore unlikely to pose any respiratory hazard at Chillán [Horwell et al. 2012], but it is rather an indicator that at least some of the dacitic ash fragments is derived from lava bodies extensively degassed at low pressure, under slow-cooling conditions [Schipper et al. 2017]. Interstitial glass in the dacitic material is rhyolitic ( Figure 6A). Similar materials are present in the subsequent ash fall, but with an additional component that is moderately vesicular, with fewer microlites, and with interstitial glass of andesitic composition ( Figure 6). No alteration products and/or hydrothermal minerals were observed in any of the ash samples.

Magmatic-gas propelled phreatic eruptions
The composition of the gases emitted from the first 2016 vent at Chillán was measured five days after the first eruption. Subsequent images of the summit area show that the vent was later buried by tephra and that passive outgassing had ceased (Figure 2). Assuming a gas mixture at equilibrium, following Giggenbach [1987] and Giggenbach [1996] and using the thermodynamic data of Stull et al. [1969], the gas-melt equilibrium temperature and oxygen fugacity (f O 2 ) can be calculated using:    [1991]. The value of f H 2 O used here is 0.98 given that at 1 bar the fugacity of a gas is equal to its partial pressure and that P H 2 O = (P tot × n H 2 O )/n tot = 0.98 bar.
The dominance of SO 2 over H 2 S and the high equilibrium temperature strongly support a magmatic origin for the gas emitted by the first 2016 vent. The computed equilibrium temperature of 856°C is much higher than the temperature at which scrubbing of magmatic gases by hydrothermal systems is expected to be significant [Gerlach et al. 2008;Symonds et al. 2001], Table 2 -X/SO 2 molar and mass ratios measured by Multi-GAS and gas composition of the plume at Chillán volcano. Error are expressed as the standard error of the regression analysis and subsequent error propagation.  Table 3 -ICP-AES bulk composition of tephra samples. The first two samples were collected on 13 January and originate from ash fall produced by the first eruption on 8 January 2013. The third sample was collected on 3 February 2016, the eruption date producing this ash fall is unknown but constrained between 13 January and 3 February 2016. The fourth originate from an ash fall during an eruption on 11 October 2017. Note the significant difference in composition between the third sample and the others. giving confidence that the reported gas composition has not been affected by secondary processes other than cooling. The equilibrium temperature, whilst a minimum estimate of the magmatic temperature, is consistent with equilibrium with a dacitic to basaltic andesite magma at depth. The clear magmatic composition of the gas at the very onset of eruptive activity implies that the hundreds of small eruptions that have occurred at Chillán since, are not driven by a hydrothermal system but are instead propelled by pressurised magmatic gases.

Gas
A large hydrothermal system with hot springs, fumaroles and hot grounds is present at the Nevados de Chillán Volcanic Complex. The most significant fumaroles are located near the village of Termas de Chillán, where water derived from hot springs is used for the spa of the same name. Other nearby locations with hydrothermal activity include the Valle Hermoso and Aguas Caliente sectors. In all locations, fumaroles and hot springs have maximum temperatures around 90°C [Berríos Guerra 2015]. Our observations suggest that this large hydrothermal system was largely bypassed by the recent activity.

A shallow level magmatic intrusion
At the onset of eruptive activity, a single dacitic component was present in the erupted tephra ( Figure 6). This probably originates from the Arrau lava cone, made of glassy block-flow medium-Si dacites [66.6-67.6 % SiO 2 : Dixon et al. 1999]. In the following days, a second basaltic-andesite component appeared in the erupted tephra. It can be difficult to identify juvenile components in the material from small eruptions that may incorporate a variety of lithics [Pardo et al. 2014], and the probable origin of this second component cannot be readily ascertained since basaltic andesites are well-represented in the stratigraphy of Nevados de Chillán. However, these are much more common at Cerro Blanco than the currently active Las Termas subcomplexes, and we consider it less likely that these fragments are lithics recycled from the preexisting volcanic edifice-especially considering they disappeared from the mix of ejected material by October 2017. Although the fragments are vesicular and appear unaltered ( Figure 6C) they cannot be confirmed as juvenile. What is clear is that the gas emitted during the repeated eruptions originates from exsolution from a melt at high temperature (> 850°C). These elevated temperatures are consistent with the incandescence observed during night-time eruptions, reported intermittently by OVDAS-SERNAGEOMIN since March 2016. These observations suggest the presence of magma at a shallow level within the edifice. This is corroborated by seismicity recorded by OVDAS-SERNAGEOMIN with long period (LP) events located at depths less than 5 km. The significant tremor registered in the first few days of the new activity suggests magma or fluids motion at that time. Several periods of increased tremor have been recorded since, many temporally associated with explosions and incandescence ( Figure 3). Altogether, these observations suggest that the current eruptive activity was triggered by the rise and shallow emplacement of magma, accompanied by exsolution of volatiles triggering explosions that opened a new vent and ejected fragments of the cone. The activity since might have been sustained by periodic small-scale recharge and/or gas exsolution from the cooling and crystallising magma (Figure 7). The surficial expression of the intrusion, with several craters aligned along a straight NNE line ( Figure 2B-C), is consistent with diking. The size of the intrusion is unknown, but we note the lack of evidence for deformation that has been monitored by GPS and tiltmeter stations from OVDAS-SERNAGEOMIN and using interferometric synthetic aperture radar observations (unpublished data). Given the evidence for a shallow intrusion, this tends to suggest it has a comparatively small volume.

Scenarios for evolution of the eruption
With nearly two years since the onset of activity and with continued daily to monthly eruptive events at the time of writing (December 2017), it remains pressing to address the evolution of this episode of activity. However, our conceptual model for the intrusion (Figure 7) provides little basis for prognostication of how the episode will unfold. Nevertheless, we can pose three scenarios: The first is a gradual or abrupt end of the unrest with the intrusion stalling at shallow level without a magmatic eruption. This "failed magmatic eruption" scenario [Moran et al. 2011] is statistically likely, considering that globally, most recorded periods of phreatic eruptions at volcanoes are not followed by magmatic eruptions [Barberi et al. 1992]. Examples of such "failed eruptions" include the 1979-1982 unrest at Mt Ontake, Japan [Oikawa 2008] and the 2006-2007 unrest at Fourpeaked volcano, USA [Gardine et al. 2011].
The second scenario considers a transition to a magmatic eruption. In this case, the intrusion does reach the surface and lava is extruded either effusively or explosively ( (Table 3) and interstitial glass within grains.
[B]-[C] BSE images of tephra fragments showing the dacitic material that dominated the 8 January blast material and 11 October material [B], and the andesitic component that is also present in the February blast material [C]. Components are marked glass (gl), plagioclase (pl), pyroxene (px), and oxides (ox).
[D] BSE image of SiO 2 polymorph-bearing dacite from 8 January. Note that ash grain is bounded by a SiO 2 -bearing vesicle, around which the walls are corroded. One SiO 2 grain shows fishscale cracking that is diagnostic of cristobalite. Expanded view with high-contrast colour balance shows devitrified groundmass, in which the darkest grey phase is an unidentified SiO 2 polymorph.
clude the 1980 eruption of Mt St Helens, preceded by two months of earthquakes and frequent phreatic explosions [Lipman and Mullineaux 1981] and the 1990-1995 eruption at Unzen volcano, where seismicity and phreatic eruptions escalated over a year, culminating in the extrusion of a lava dome and generation of pyroclastic flows [Nakada et al. 1999].
Between these two possibilities lies a third scenario in which activity could continue at a low level for several years. Such prolonged activity, induced by shallow magmatic intrusions, accompanied by magmatic degassing but with limited expulsion of juvenile tephra or lava, has characterized eruptions of Turrialba

Conclusions
In 2016, we measured the composition of gases emitted at the onset of the ongoing (at the time of writing in December 2017) eruptive episode at Nevados de Chillán. We also measured the SO 2 flux emission during eruptive discharge and the composition of tephra associated with these explosive events. The main conclusions we derive from this study are: 1. Right from the onset, eruptive activity was driven by magmatic gases, although ejecta were dominated by recycled material from the edifice.

2.
A shallow magmatic intrusion is likely to be the trigger for the current unrest and may be periodically recharged.