The Chaitén rhyolite lava dome : Eruption sequence , lava dome volumes , rapid effusion rates and source of the rhyolite magma *

We use geologic field mapping and sampling, photogrammetric analysis of oblique aerial photographs, and digital elevation models to document the 2008-2009 eruptive sequence at Chaitén Volcano and to estimate volumes and effusion rates for the lava dome. We also present geochemical and petrologic data that contribute to understanding the source of the rhyolite and its unusually rapid effusion rates. The eruption consisted of five major phases: 1. An explosive phase (1-11 May 2008); 2. A transitional phase (11-31 May 2008) in which low-altitude tephra columns and simultaneous lava extrusion took place; 3. An exogenous lava flow phase (June-September 2008); 4. A spine extrusion and endogenous growth phase (October 2008-February 2009); and 5. A mainly endogenous growth phase that began after the collapse of a prominent Peléean spine on 19 February 2009 and continued until the end of the eruption (late 2009 or possibly earliest 2010). The 2008-2009 rhyolite lava dome has a total volume of approximately 0.8 km3. The effusion rate averaged 66 m3s-1 during the first two weeks and averaged 45 m3s-1 for the first four months of the eruption, during which 0.5 km3 of rhyolite lava was erupted. These are among the highest rates measured world-wide for historical eruptions of silicic lava. Chaitén’s 2008-2009 lava is phenocryst-poor obsidian and microcrystalline rhyolite with 75.3±0.3% SiO2. The lava was erupted at relatively high temperature and is remarkably similar in composition and petrography to Chaitén’s pre-historic rhyolite. The rhyolite’s normative composition plots close to that of low pressure (100-200 MPa) minimum melts in the granite system, consistent with estimates of approximately 5 to 10 km source depths based on phase equilibria and geodetic studies. Calcic plagioclase, magnesian orthopyroxene and aluminous amphibole among the sparse phenocrysts suggest derivation of the rhyolite by melt extraction from a more mafic magmatic mush. High temperature and relatively low viscosity enabled rapid magma ascent and high effusion rates during the dome-forming phases of the 2008-2009 eruption.

Prior to 2008, Chaitén caldera was considered an inactive volcano, consisting of a circular 2.5 km diameter collapse caldera containing a pre-historic intracaldera lava dome surrounded by a circular moat.Formation of the caldera and associated eruption of rhyolite tephra took place between 9,000 and 10,000 years ago (Naranjo and Stern, 2004;Lara et al., 2013, this volume;Watt et al., 2013, this volume).The pre-historic lava dome is composed of high-silica rhyolite (obsidian and variably devitrified rhyolite lava).Obsidian pebbles from an archeological site dated at ~5,600 radiocarbon years are attributed to the lava dome (Stern et al., 2002).However, other deposits on the flanks of the caldera are dated at ≤3ka (Lara et al., 2013, this volume;Watt et al., 2013, this volume) and contain rhyolite that is geochemically similar to that of the lava dome.Consequently, the exact age of the last eruption of rhyolite at Chaitén prior to 2008 is uncertain, but may have been historical (Lara et al., 2013, this volume).
Overviews of the 2008-2009 1 eruption and its impacts are given in Carn et al. (2009), Muñoz et al. (2009b), Lara (2009) and Major and Lara (2013, this volume).Petrologic and geodetic data that constrain the initial ascent rate and pathway for the rhyolite magma are presented in Castro and Dingwell (2009) and Wicks et al. (2011).Halogen geochemistry and partitioning of elements during late degassing are given by Lowenstern et al. (2012), and tephra distribution from the 2008 eruptions is described by Watt et al. (2009) and Alfano et al. (2011).Additional information on the eruption, its deposits and impacts is presented in this issue of Andean Geology.

Photogeology and Geographic Information System (GIS) analysis
We used a combination of oblique photographs from aircraft (Figs. 2 and 3), helicopter-assisted fieldwork by an international team in 2010 (Pallister et al., 2010) and commercial satellite imagery (Fig. 4) to map the development of the Chaitén lava dome during the 2008-2009 eruptive sequence (Fig. 5).Volumes of dome lobes were estimated during the first 4 months of the eruption while their geometries were relatively simple.We calculated areas and thicknesses of the dome lobes by a combination of visual inspection of aerial photos (Fig. 6) and GIS analysis of georectified satellite images (e.g., Fig. 4).Volumes were then calculated using the most appropriate geometries (e.g., ½ hemispheres, or width multiplied by length and height for approximately rectangular solids).
In addition, the total 2008-2009 dome volume was calculated by differencing post-and pre-eruption Digital Elevation Models (DEMs) (Fig. 7).Three pre-eruption DEMs were constructed using: 1.The 1:50,000-scale Chaitén topographic map (Sheet 4245-7230 of the Instituto Geográfico Militar de Chile, 1997 edition); 2. Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images; and 3. Shuttle Radar Topography Mission (SRTM) images.Two post-eruption DEMs were constructed using data from: 1.An airborne LIDAR survey conducted by Digimapas Chile between 15 October and 7 December 2009 and provided by the Regional Government of Los Lagos region and 2. Overlapping digital oblique photographs using the photogrammetric method of Diefenbach et al. (2012).The oblique aerial photographs were shot from a helicopter by the first author on 24 January 2010 using a digital single-lens reflex camera, while circling the caldera rim at 1,530 m and 2,740 m elevations.Control points for the oblique photographs were established using benchmarks from the topographic map and by using a combination of GPS locations and altimeter readings.A boundary shapefile ('cookie cutter', Fig. 7) of the area of dome growth was used to clip each of the three pre-eruption DEMs ('Base surface', Fig. 7) and the two new lava dome DEMs ('New surface', Fig. 7).The elevation difference ('difference map', Fig. 7) between corresponding cells yielded an estimate of the total volume of extruded lava.Each of the resulting six differencing iterations gave similar results (within ±10%); however, owing to radar shadows in the SRTM DEM and the low spatial resolution of the ASTER DEM, we regard the most reliable volumes as those derived by differencing the 1997 topographic map DEM and either the photogrammetric oblique photo DEM or the LIDAR DEM.tion).This is also a minimum volume as it does not include parts of pre-historic dome that are buried by moat deposits.

Geochemistry and petrology
Rock samples were collected from the 2008-2009 deposits within and near the caldera during two weeks of helicopter-assisted field work in January 2010.Sample descriptions, locations and analytical data are given in Table 1.Powdered whole-rock samples were analyzed for major-elements by X-ray fluorescence techniques at the Washington State University GeoAnalytical Lab in Pullman, Washington.Trace-elements in the same powders were analyzed by inductively coupled plasma mass spectrometry at the same laboratory.Analytical procedures and precision are discussed at http://www.sees.wsu.edu/Geolab/note.html.
Electron microprobe analyses of the minerals in the Chaitén rhyolite were done on the JEOL 8900 at the United States Geological Survey in Menlo Park, California, and results from representative analyses are listed in Table 2.We analyzed both sparse phenocrysts in polished thin sections of rock samples as well as grains in mineral separates (prepared to provide more representative suites of minerals from the crystal-poor Chaitén rhyolite).Microprobe standards included a variety of well characterized natural minerals (Tiburon albite, sodalite, barite, Wilberforce apatite) and simple elemental oxides (MgO, TiO 2 , MnO 3 ) that are in standard use at the Menlo Park facility.In this work we focus on data that bear on the origin and viscosity of the rhyolite magma.Data and interpretations concerning partitioning of halogen gases are given in a separate paper (Lowenstern et al., 2012).

Eruptive sequence
On the basis of observations and photographs by the Servicio Nacional de Geología y Minería (SERNAGEOMIN, 2008;Basualto et al., 2009), reports by residents of the area and our geologic mapping (Fig. 5) we divide the 2008-2009 eruption of Chaitén into five phases.1. Explosive phase.This phase began late on 1 May 2008 (local time) or early on 2 May 2008 (UTC) and had a volcanic explosivity index (VEI) between 4 and 5 (Carn et al., 2009;Watt et al., 2009;Alfano et al., 2011) 3) that exited the caldera through the pre-existing Caldera Creek drainage to the south and flowed down the Chaitén River to within 3 kilometers of Chaitén town (SERNAGEOMIN, 2008;Duhart et al., 2009;Pallister et al., 2010;Major et al., 2013, this volume).The collapse eroded into the adjacent areas of the composite lava dome, producing a deep reentrant in its western face (Fig. 3 and dotted line in Fig. 5).In the days to weeks following the 19 February collapse, additional headward collapses within the reentrant led to collapse of the spine.5. Endogenous growth phase.This phase began with lava extrusion following the 19 February 2009 spine collapse, but then shifted to mainly endogenous growth, characterized by magma intruding beneath and inflating previously erupted lobes ('Post-Feb.2009 lobe and endogenous growth', Fig. 5).
The February collapse-generated reentrant was filled in by new lava, and as endogenous growth took place earlier phases of the dome were thrust outward and away from the vent area (Fig. 5).This phase of endogenous growth gradually slowed but persisted until the eruption effectively ended by late 2009 or earliest 2010.

Photogrammetry and GIS
The total volume of extruded lava erupted in 2008-2009 is ~0.8 km 3 .Our estimate of the lava effusion rate for first two weeks is 66 m 3 s -1 (Fig. 6) and for the first four months the rate averaged 45 m 3 s -1 (value calculated from a polynomial fit to the volume-time data in Fig. 8).We estimate relative errors of ±20% in our rate calculations.The cumulative volume for the first four months of the eruption is ~0.5 km 3 , i.e. 5/8 of the total volume of 0.8 km 3 determined by photogrammetry.The trend in figure 8 indicates that the rate of extrusion had decreased to near-zero by the end of September 2008.Consequently, to reach the full effusive volume of 0.8 km 3 eruption a second period of renewed extrusion and endogenous growth is suggested.We believe that this renewed period of growth took place during late 2008 and early 2009, a period of spine growth, spine collapse and renewed extrusion that filled the February 2009 collapse scar.Over the entire duration of the 2008-2009 eruption (May 2008-January 2010), the averaged lava eruption rate was ~16 m 3 s -1 .We note that this overall rate is a minimum as it does not include the volumes of PDC deposits from the December 2008 and February 2009 dome collapses (Fig. 5).
An independent photogrammetric analysis of oblique photographs was conducted by Valenzuela (2011), who determined an effusion rate of ~27 m 3 s -1 for the first five months (12 May-30 October 2008) of the eruption.Our estimate for the same period based on the polynomial fit in figure 8 is 30 m 3 s -1 , within error of Valenzuela's estimate.
We calculate the minimum volume of Chaitén's pre-historic intracaldera lava dome by subtracting a horizontal plane constructed at the lowest point of the caldera moat from the pre-eruption DEM, yielding 0.5 km 3 (Fig. 7).This is a minimum volume for the pre-historic dome, as it does not include parts of the dome that were buried below fragmental moat deposits.Together, the two domes constitute about 20-40% of the 3.5-7 km 3 collapse volume of the prehistoric caldera, which we estimate from the DEM using average and maximum heights of the rim..192x -8.69) and converting from units of m 3 d -1 to m 3 s -1 yields an average effusion rate of 45 m 3 s -1 for the first four months of lava dome eruption.

Petrography and geochemistry
Here we provide a brief summary of the petrography and geochemistry of rhyolite erupted during 2008-2009 in the Chaitén area.We focus on petrologic and geochemical factors that affect eruption rates.We also compare the 2008-2009 compositions to those erupted during prehistoric time in the Chaitén area.Table 1 lists locations, descriptions and major and trace-element chemistry of the samples.Additional details of the petrology of these rhyolite samples are given in Castro and Dingwell (2009) and Lowenstern et al. (2012).
All of the 2008-2009 Chaitén rhyolites are phenocryst-poor (~0.1 vol.% phenocrysts) with mainly plagioclase and subordinate amounts of orthopyroxene and oxide minerals.Biotite or amphibole phenocrysts are also present in some thin sections.Mineral separates show that the 2008-2009 rhyolite also contains accessory zircon, apatite and pyrrhotite.We found hornblende but no biotite as phenocrysts in mineral separates from lump pumice that was collected on 20 May 2008 from a beach at Auchemo, ~25 km southwest of Chaitén.This pumice had apparently been carried west during flooding caused by heavy rainfall after the initial week of eruptions, but prior to May 20 and it is interpreted as a product of the initial early-May 2008 Plinian eruption (A. Lockhart, USGS;personal communications, 5 September 2008 and15 October 2012).The plagioclase crystals are subhedral to euhedral, though even well-formed crystals have slightly rounded edges.In general, the mafic minerals retain more euhedral crystal shapes.Though the amphiboles in the pumice sample from Auchemo are not euhedral, they lack the marginal reaction rinds that are generally attributed to dehydration accompanying decompression (Rutherford et al., 1998;McCanta et al., 2007) or possibly to heating (De Angelis et al., 2012).
Rhyolite pumice and obsidian lapilli occur together in the May 2008 Plinian deposits, but the abundance of water and the degree of microlite crystallization are different in the two lapilli types (Lowenstern et al., 2012).Pumiceous glass is nearly anhydrous with <0.2% wt.% H 2 O, whereas glass from two studied obsidian lapilli contained ~0.8 wt.% H 2 O (analyses by Fourier transform infrared spectrometry).Other samples of the 2008-2009 rhyolite lava are visibly more crystalline; backscatter electron image maps reveal that they contain abundant microlites of plagioclase, orthopyroxene and oxide.There are no hydrous microlites (biotite or amphibole) in the Chaitén samples.Silicate melt inclusions are common in all studied phases of the rare phenocrysts.The melt inclusions are predominantly glassy and contain small (5-15 μm) shrinkage bubbles that make up an estimated 1 to 5 volume percent of the inclusions.Castro and Dingwell (2009) and Lowenstern et al. (2008) reported that plagioclase grains in the 2008 pumice are mainly An [40][41][42][43][44][45] ; however, we also found rare cores as calcic as An 81 (Table 2).Orthopyroxenes are typically En 50-55 and amphiboles are more aluminous and more calcic (Table 2) than expected for crystals forming from high-silica rhyolite magma (c.f., Table 4 of Coombs and Gardner, 2004).In addition, experiments failed to produce amphiboles at reasonable temperatures (775-850°C) and water concentrations (4-6 wt.%) from the whole-rock pumice erupted at Chaitén (Castro and Dingwell, 2009;written communication T. Sisson and J. Lowenstern, 2008).

Whole-rock chemistry and comparison of 2008-2009 and prehistoric lava
The whole-rock compositions of the 2008-2009 eruptive products from Chaitén are strikingly similar to and overlap completely with the pre-historic dome (Table 1 and Muñoz et al., 2009b) and they contain the same suite of phenocryst minerals (mainly plagioclase, orthopyroxene, and oxides).The entire suite of Chaitén rhyolites (pre-historic and 2008-2009) are remarkably uniform in composition with 75.36±0.028%SiO 2 , 14.0±0.019%Al 2 O 3 , 0.26±0.006%MgO, 1.47±0.005%CaO, 4.25±0.011%Na 2 O and 3.04%±0.004%K 2 O (Table 1).The relatively low alkali content for a rhyolite creates a slightly peraluminous (0.74 wt.% corundum) and orthopyroxene-bearing normative rock composition.Trace-elements are also remarkably similar among the 2008-2009 and pre-historic rhyolite samples.Strontium concentrations are relatively high (~145 ppm) and exceed Rb (~110 ppm) and the rare-earth element (REE) patterns show broadly concavedownward patterns with minor Eu-anomalies (Fig. 9).Due to the low abundance of phenocrysts, glass compositions in all the Chaitén rhyolites are essentially identical in composition to those of the whole-rock samples.

Origin of the rhyolite magma
The 2008-2009 rhyolite's normative composition plots close to that of relatively low pressure (100-200 MPa) minimum melt in the haplogranite system (Fig. 10), consistent with the pressure estimates based on phase equilibria (Castro and Dingwell, 2009) and model depths from geodesy (5-9 km; Wicks et al., 2011).These data, as well as radiogenic Sr, Nd and Pb isotopic ratios (Muñoz et al., 2012), favor a shallow crustal origin for the rhyolite magma.Calcic plagioclase cores, relatively magnesian orthopyroxene and the presence of relatively calcic and aluminumrich amphibole suggest that the rhyolite was derived by melt extraction from a more mafic (diorite to granodiorite) magmatic mush at these shallow crustal depths, an origin also favored for voluminous rhyolite erupted at Alaska's Katmai-Novarupta caldera system in 1912 (Hildreth and Fierstein, 2012).The REE patterns for the Chaitén rhyolites in figure 9 are similar to other 'wet oxidized rhyolites' (Bachman and Bergantz, 2008) interpreted as melt extracts from granodioritic crystal mushes in which plagioclase crystallization is delayed and amphibole and titanite are retained in the mush (conditions that reduce the extent of the Eu-anomaly and result in high Sr relative to Rb in the rhyolite extract).Castro and Dingwell (2009) reported mineral assemblages in Chaitén pumice similar to those we report here.Based on Fe-Ti oxide thermometry and equilibria experiments for the last-crystallized phases, they determined a magmatic storage temperature of ~800ºC at a pressure of 120-200 MPa (a depth range of about 5-10 km).Furthermore, using experimental calibration of plagioclase rim overgrowths, they estimate decompression rates of >40 MPah -1 (>0.5 ms -1 ) from a depth of >5 km during a period of ~4 hours preceding the initial eruption.Several percent of H 2 O was lost during this ascent, as indicated by the low residual H 2 O contents of the 2008-2009 pumice.Castro and Dingwell (2009) note that the viscosity of the Chaitén magma was low (~10 6 to 10 8 Pa s) -an order of magnitude lower than critical values required for glass transition and autobrecciation.These data indicate that the Chaitén rhyolite was very fluid and that it ascended rapidly to near-vent level, losing several percent H 2 O enroute and undergoing the explosive fragmentation that powered the initial Plinian columns.

Rapid ascent and rapid rates of lava effusion
Our extrusion rate estimates of 66 m 3 s -1 for the first two weeks and 45 m 3 s -1 for the first four months of the eruption at Chaitén are among the highest for historical lava-dome-forming eruptions (Fig. 11).High rates of extrusion are generally attributed to The ChaiTén rhyoliTe lava dome :erupTion sequenCe, lava dome volumes, rapid effusion raTes... combinations of low magma viscosity, high conduit pressure and large vent size, and in turn, magma viscosity and extrusion pressure are related to gas, crystal and bubble content and to magma ascent velocity at shallow levels (Melnik and Sparks,2009).
As reviewed above, we know from the experimental work of Castro and Dingwell (2009) and the low phenocryst content of the rock samples that the Chaitén rhyolite was a hot, low viscosity magma.The explosive nature of the 1-6 May 2008 eruptions, presence of tubular pumice (indicative of gas streaming, probably at conduit margins) and high eruption columns (Carn et al., 2009) indicate rapid gas loss during the initial explosive phase of the eruption.In contrast to the comparably rapid but fissure-fed extrusion of rhyolite at Cordón Caulle in 1960 (Lara et al., 2004), we know from visual observations at Chaitén that the initial vents for the 2008 eruption were circular and relatively small and that they penetrated and did not significantly destroy or excavate the pre-historic dome (SERNAGEOMIN, 2008;Major et al., 2013, this volume).Rapid effusion at Chaitén cannot be explained as a result of a large (i.e., fissure) vent.The presence of small circular vents and lack of disruption of the pre-historic dome suggest high-strength wall rocks surrounding the shallow conduit, resulting in high-strength vent nozzles.We attribute the initial explosive Plinian phase of the eruption to high pressure and explosive fragmentation of the hot, low-viscosity Chaitén magma as it accelerated through the small high-strength vents.The high pressure and small vent sizes contributed to the high altitude of the initial eruption columns.The presence of pumice with tube-shaped vesicles and hydrous obsidian lithic fragments in the initial PDC and tephra deposits (Lowenstern et al., 2008;2012) indicates retention of relatively gas-rich magma to shallow crustal levels.We suggest that the initial Plinian phase of the eruption tapped the most gasrich vanguard magma in the conduit and underlying magma reservoir.The subsequent transitional, lava flow, spine and endogenous phases of the eruption then tapped progressively gas-depleted magma.Decreasing extrusion pressure (and gas content) is indicated by: 1. the transitional phase of the eruption between 11 May 2008 and the end of May 2008, during which simultaneous and rapid extrusion of lava and low-level explosive ash columns took place (e.g., Fig. 6); and 2. the progressively decreasing lava extrusion rates over the course of the eruption (Fig. 8).Overall, we propose that the remarkably rapid rates of rhyolite lava effusion at Chaitén were a consequence of the high temperature and low viscosity of the crystalpoor magma, coupled with high ascent rates.The high ascent rates allowed retention of volatiles and consequent high magma pressure in the reservoirconduit system, which in turn sustained the high rates of effusion.

FIG. 3 .
FIG. 3. Oblique aerial photographs of Chaitén dome taken in late February 2009 showing: a.An overview of elongate collapse scar and PDC deposit of 19 February 2009; b.Close-up view of late 2008 -early 2009 spine; c.View of caldera looking northwest showing late 2008 -early 2009 spine towering over dome.

FIG. 5 .
FIG. 5. Geologic map of lava dome within Chaitén caldera, as it appeared in January 2010.Labels give age ranges for individual lobes and PDC deposits emplaced in 2008 (08) and 2009 (09).Talus slopes (Talus) cover buried contact of 2008-2009 lava, which is indicated by dashed line.Dotted line indicates former position of February 2009 PDC fan, which originated by collapse of September 2008-February 2009 spine, a remnant of which remains near center of the dome complex.Line decorated with half circles marks a detachment-faulted boundary produced by post-February 2009 endogenous growth.Detachment faulting during endogenous growth shifted late 2008 to early 2009 lava (08-09 lobe) to the northwest.Decorated dashed lines within area of post-February 2009 lobe and endogenous growth are normal fault scarps produced in response to collapse and lateral spreading to the west.Dashed circle near center of figure represents approximate position of main vent for May-September 2008 dome lobes.Dash-dot lines indicate areas of forest damage resulting from early May 2008 pyroclastic surges.

FIG. 8 .
FIG. 8. Cumulative volume of newly erupted Chaitén lava dome as calculated during the period between 11 May and 30 September 2008.Estimated errors of ±20% are assigned to volumes.Initial value of zero volume is set to 11 May 2008, as first indication of dome extrusion was on 12 May 2008.Horizontal error bars indicate ranges of dates for observations used in estimations of volumes.A polynomial equation is fit to the data and shows a slowing of extrusion rate with time.Applying this equation (y = -0.026x 2 +7.192x -8.69) and converting from units of m 3 d -1 to m 3 s -1 yields an average effusion rate of 45 m 3 s -1 for the first four months of lava dome eruption.
FIG. 9. Chondrite-normalized rare-earth element diagram, modified from Lowenstern et al. (2012), comparing abundance patterns for 9 samples from the 2008-2009 eruption to a sample of the pre-historic Chaitén dome and to the field of 'wet oxidized rhyolites' of Bachman and Bergantz (2008).

TABLE 1 . WHOLE-ROCK MAJOR AND TRACE ELEMENT COMPOSITIONS DETERMINED BY XRF AND ICP-MS.
The ChaiTén rhyoliTe lava dome :erupTion sequenCe, lava dome volumes, rapid effusion raTes...

TABLE 2 . REPRESENTATIVE ELECTRON MICROPROBE ANALYSES OF CRYSTALS IN CHAITÉN RHYOLITES.
Sample numbers refer to samples described inTable1.Samples CH-4, MT2, and C10D2 are characteristic crystal-poor rhyolite lava from lobes of the indicated dates.Spot # refers to individual analysis identification number.Mineral: p=plagioclase; o=orthopyroxene; a=amphibole; b=biotite; Ab, Fs=albite component in plagioclase, Ferrosalite in orthopyroxene; Or, En=orthoclase component in plagioclase, enstatite in orthopyroxene; An, Wo=anorthite component in plagioclase, wollastonite in orthopyroxene.Dashes indicate not analyzed or below detection limit.