1,500 mas), its maximum width reaches 2km and no deposition is observed
due to the high gradient of the heads of both Calabozos creek (eastern branch, 0.38 m/m) and
Carmelito creek (western double branch, 0.46 m/m). Along the transitional segment, between 1,500
and 1,000 masi, both gradients decrease to0. 13 m/m and 0.17-0.21 m/m, respectively. This causes
that from 1,000 m downwards two branches derived from Calabozos creek to Carmelitocreek, tuning
the latterinto the most active riverbed. Both gradients decrease (Calabozos=0.038 m/m; Carmelito=
0.035) below 1,000ma.s.|. The main glacial valley ends at 600m asland splits into different riverbeds
following ancient lava lobes (Fig. 6). Figure 6 also shows both the distribution of remnants of debris
flow deposits that occurred during the XX century and lahar deposits accumulated during the 1971
event. Thin (< 1 m) easy removable lahar deposits are found between 1,100 and 1,500 m high,characterised by randomly distributed (< 2 m diameter) blocks in a matrix of sand and gravels.
Thicker (<2.5-3 m) lobes that cover the entire valley width (1.5 km) are present at 1,100 m, but no
wave marks were found.
Laharic lobe deposits of 0.5-1 km width were measured at El Playén (1,100-750 ma.s.l., 0.067
m/m gradient), where three debris levels of different events are observed adding up a thickness of
10m. The younger, deposited in 1971, reached up to 2 m, while the middle one (1948) ranged from
2 to 3.5 m thick. Lahar heads of the 1971 flow produced marks in the trees up to 6 m high, roughly
4 times higher than the thickness of the tail deposit (Figs. 7, 8). The sand and fine gravel matrix
supported blocks of less than 1 m diameter. The lahar overflowed the previous terraces where
abundant 15 m tall trees were placed, with the consequent covering of 2/3 of their trunks (Fig. 8).
Discharges of 10,000 to 15,000 m’/s are estimated for the 1971 lahar in this riverbed segment. The
FIG. 6. Main lahar paths along the Zanjén Seco-Carmelito valley (A). The last reactivation of the 1971 event is also
shown (B).FIG. 7. Lahar's wave (1971 event) momentum must have been high enough to overspill 1S m deep crevasses, leaving
up to 4 m high impact marks on tree trunks.
FIG. 8. Lahar deposits of 1971 eruption in the western branch of Zanjén Seco valley. The central channel is >60m wide
and the flow overspills inundate large areas of forested terrain.
flow-section deeps to 50 x 8 m from 500 m contour up to the bottom of the valley, where lavas look
‘scratched and polished by the lahar erosion.
Evidences of 400 m wide lahar overflows older than 1944 are exposed at banking curves where
blocks < 27 m* were lett, 0.75 km before the Villarrica lake shoreline (Fig. 9). A half km before the
lake coast, the Carmelito creek diverges into a eastern branch (Carmelito) and a western main fan
where the lahar fel ike a 25m high cascade (Fig. 9). In the Carmelito branch, coarse facies of the
1971 laharare exposed, while a 5 m section formed by three hyperconcentrated units characterised
the 1971 fan at the creek mouth. Field evidences indicated that downstream dilution occurred over
the course of Zanjén Seco-Carmelito riverbed and caused a complete transformation from debris flow
to hyperconcentrated flow at the thalweg. These deposits are better sorted and show a faint
stratification, so it is assumed that they accreted during significant time intervals. Sediment
remobilisation along the Carmelito fan facies repeatedly occurred after the 1971 eruption in 1973,
1975 and 1976, during heavy winter rainstorms.
Tothe eastern partof the Zanjon Seco Valley, older (1908?) and higher energy lahars flows have
left boulders (0.3 to2m diameter) spread throughout La Turbina (Pucén) and Calabozo (Agua Grande)
creeks (Fig. 6).%
FIG, 9. Map showing the main lahar flow paths at the mouth of Carmelito creek. Course evidences of events larger
than 1971, as 1908(?) andlor older, are incicated
DISCUSSION
Inthe Villarrica volcano, field evidences indicate that most lahar deposits appear to be clay-poor
(less than 5% clay/sand+silt+clay), thus they are generally induced by sudden water release (Scott,
1988). In addition to the presence of perennial glaciers and snow as water sources, the Villarrica
voleano shows steep slopes and an abrupt relief at the summit. On the other hand, the presence of
small caldera at ~2,400 m a.s.I., produces a structural slope break, which favours the subglacial
accumulation of lavas during effusive eruptions that can form small subglacial lakes. This may
eventually break out when a section of the ice cap becomes buoyant, as it was observed in 1971
with the consequent release of the trapped water. Thus, rapidly melted snow and ice are the main
triggering mechanism to generate subsequent laharic flows, while the lava emission rate suddenly
increases during the eruption turning point (Major and Newhall, 1989)
The solid fraction of the Villarrica lahars is mainly supplied by common unconsolidated debris
sources of glacial tillthat were left on the volcano flanks. These materials are loose clastic sediments
that are easily incorporated to the flux as debris with the consequent generation of alahar flow.On
the contrary, ifpeak water-discharge is reached and the amount of sediment available isnot enough
a flood would result instead of alahar. For example, during the 1984-1985 eruption no laharic event
was produced when lava emission continued for many weeks, but small floods resulted instead3
FIG, 10.Sketch map of the morphological evolution
showing the Pucén river delta progradation
from 1944 until 1995. Shaded area as reference.
The major modifications occurred after the
1948 and 1971 lahars.38
Althoughitcouldbe interpretedas.a consequence of the absence of an eruption climax, the drainage
system showed no enough sediment available to form lahars.
The laharic deposits relatedto both river systemsare quite different due to their gradient features
and flowing behaviours. While the Turbio-Pedregoso riverbeds show remarkable slope breaks,
permanent water-flow, and tribute to the Pucon river, the Zanjon Seco creek is commonly dry and
discharges straightinto the Villarrica lake. Hence. facies changes due to /ahar dilution normally takes
place along the Turbio-Pedregoso-Pucén rivers. As a consequence, a very dynamic fan delta
develops in the Villarrica lake, since 1944, after three lahar-generating eruptions in 1948-1949, 1963
and 1971 (Fig. 10)
Lava eruptions of the Villarrica volcano have taken place through the main crater or through radial
flank fissures. In fact, during the 1984 eruption, the lava emission rate reached only about 20 m/
s (Moreno e¢a/, 1985: Moreno, 1993) and all lava flows poured out from the main crater without a
lahar generation. On the contrary, in 1971, a 2 km long N10°E fissure yield two lava fountains with
a very high emission rate of ca. 500 m*/s (Moreno, 1993) that produced large volumes of lahars
(Marangunic, 1974).
The pattern directions of the fissure vents on the volcano flanks depend both on the intemal cone
structure and the magma injection together with its pathway to the surface. These variables may
dictate possible routes that lavas and lahars will take, and only a slight angle variation of the fissure
and its lava fountain will affect one or another riverbed
Chronicles of the XX century eruptions in addition to field evidences indicate that lateral or flank
fissures occurred in all of them: a. 1908 to the north (Zanjan Seco and Turbio-Pedregoso drainage
systems); b. 1948-1949 to the WNW and ESE (Chaillupén, Voipir, Molco. Turbio-Pedregoso,
Correntoso, Zanjén Seco riverbeds): ¢. 1963 to the WSW (Zanjén Seco, Huichatio, Chaillupén), d.
1964to the south (Cofaripe, Correntoso, Huichatio, Voipir) ande. 1971 tothe NNE and SSW (Turbio-
Pedregoso, Zanjén Seco, Correntoso, Voipir, Chaillupén) (Fig. 1). Hence, during future eruptions the
volume and discharge of laharic ows together with theirhazard scenario would be highly dependent
on the fissure and lava fountain direction.
CONCLUSIONS
Laharic flows descending along the Turbio-Pedregoso and Zanjon Seco River systems are the
main Villartica’s volcanic hazard that threatens Pucon town. The course of the lahars will strongly
depend on the direction that emitted lava will take during the eruption climax, together with its
magnitude in terms of effusion rate and discharge.
The directions that take the erupting fissures that cross the crater increase the hazard problem
definition. The probability thata lahar could take the Turbio-Pedregoso river or Zanjon Seco creek will
depend on a NW or a NNW fissure eruption, respectively. The resulting scenario could vary from a
catastrophic to a small effect in Pucon. The uncertainty to determine the hazard degree comes from
allthe riverbeds that drain the Villarrica volcano, until the turing pointof an effusive eruptionis reached
ACKNOWLEDGEMENTS
FONDECYT GRANT No. 1247 (HM) support the field work at the Villarrica volcano. Reviews of L.
E. Lara and D. Welkner greatly improved the text and its conclusions.98
Magmatic evolution of the Villarrica Volcano
Rosemary Hickey-Vargas Department of Earth Sciences, Florida International Universi, Miami, USA
hickey @fiv ed
Leopoldo LopezEscobar Universidas'de Concepcion, Concepeion. Chile
lopez @ udec el
Hugo MorenoR, Servicio Nacional de Geologia y Minera, P.O. Box 28-0, Temuco.
moreno sernageomin.c!
Jorge ClaveroR Servicio Nacional de Geologia y Mineria, Av. Santa Maria 0104. Santiago. Chile
Inaranjo@semageorin c
Luis Lara. Servicio Nacional de Geologia y Mineria, P.O. Box 23-0, Temuco
timoreno @semageomin.e
Murong Sun Florida international University, Miami, USA
ABSTRACT
Voleanic rocks from Villarrica volcano have been studied using a broader array of pettologic and
‘geochemical tools than any other SVZ volcano, Results to date show that magmas erupted at this centre evolve
through polybaric crystallization of olivine and plagioclase from hydrous basaltic magma, yielding basaltic
andesite and low-SiO, andesite differentiates. Geochemical evidence for assimilation of crust is minimal
at Villarrica volcano, however, assimilation of chemically related volcanic and plutonic wallrocks and mixing
with silicic differentiates trom within the same magmatic system are suggested by U-series data for Villarrica
basaltic andesite, which are closer to isotopic equilibrium than basalt. Magma generation occurred by tlux-
‘melting of the mantle wedge with a significant contribution of elements transported in fluids derived from
the subducted crust. The subduction input is indicated by high "Be/*Be, high B/Be, °U2"Th and *Ra/™"Th
These values are higher in Villarrica basalts than in those from any other SVZ volcanic centre.
INTRODUCTION
Villarrica voleano is one of the most well-characterised volanoes of the Central Southern
Volcanic Zone with respect to its magmatic evolution. Over the past 15 years, a number of papers
has described the compositions of magmas erupted at Villarrica volcano, and the relationship
between these magma compositions and those erupted at nearby small eruptive centres and the
Villarrica-Lanin cross-chain (Hickey-Vargas efa/, 1989: Lopez-Escobar and Moreno, 1994: Moreno
etal, 1994a; Lopez-Escobar efa/, 1995; Clavero, 1996; Sun, 2001). Villarrica volcanois also one of the
most frequently studied SVZ volcanoes for the purpose of elucidating subduction zone processes.
Four papers have used Villarrica as a type’ composition to characterise the SVZ subduction system
and subduction zone magmatism generally (Sigmarsson ef a/, 1990: Morris e/.a/, 1990: Hickey-
Vargas eta/., 2002; Sigmarsson eta/, 2002). in this report the authors review the magmatic evolution
of Villarrica volcano, first from the ‘standpoint of magma chamber and crustal processes and, second,
from the point of view of magma formation by melting and manile wedge processes.0
MAJOR ELEMENT VARIATION AND VOLCANIC CHRONOLOGY
The volcanic stratigraphy of the Villatrica volcano was describedin detail inLopez-Escobar etal,
1984) and subsequently refined by Moreno (1993). Moreno ef a/ (1994a), Clavero and Moreno
(1994) and work in progress by the coauthors. Three volcanic units are identified: Villarrica 3 (from
3,700 yrBP, including historic eruptions of 1948, 1964, 1971 and 1984), Villarrica 2 (14,000-3,700
ytBP), which encompasses postglacial lavas and pyrociastics that culminated with the development
of a2 km wide caldera that forms the base of the currently active cone. and Villarrica 1 (>14,000
yBP), which includes interglacial lavas and pyroclastics that precede the development of a 6 km
diameter caldera that forms the E-SE flank of the volcano. Ages for the units are defined by "C
dates for charcoal found in deposits of the Lican and Pucon Ignimbrites (Moreno ef a/, 1994a;
Clavero and Moreno, 1994)
Petrologic and geochemical data for the volcanic products of Villarrica have been reported by
Hickey-Vargas ef a/ (1989); Lopez-Escobar and Moreno (1994). Clavero (1996) and Sun (2001).
Major element geochemical data forlavas and pyroclastics compiled from these sourcesare shown
in figure 1. Overall, the three units exhibit a similar chemical and petrologic variation, with a few
significant differences. All units include products in the range basalt to basaltic andesite (50-54%
SiO,). Basaltic andesite and low-SiO, andesite with >54% SiO, is rare among Villarrica 1 and 3
products, whereas Villarrica 2 has more abundant basaltic andesite and andesite, and rhyolite is
associated with the late caldera-forming stage of Villarrica 1. Phenocryst minerals are plagioclase
andolivinein basalts and basaltic andesites. and plagioclase and clinopyroxene in andesite. together
with accessory magnetite. On variation diagrams. all three units show decreasing MgO, Al,O, and
CaO withincreasing SiO, , and increasing Na,O and K,O. These changesate broadly consistentwith
crystallization of the observed phenocryst phases. For example, Hickey-Vargas ef a/. (1989)
modeled crystallization of a Villarrica 2 basaltic andesite (55% SiO,) from basalt (52% SiO,) by 33%
crystallization of an assemblage consisting of 17% olivine, 55% plagioclase, 24% clinopyroxene,
and 4% magnetite. Newer data (Fig. 1) show that many basaltic andesites and andesites from
Villarrica 2 have higher Fe,O, (>11% Fe,0,) than the majority of basalts and basalticandesites from
this and other units, This type of andesite requires a somewhat different origin, including lower /O,
conditions to suppress magnetite crystallization.
MAGMA CHAMBER PROCESSES
Villarrica basalts and basaltic andesites are plagioclase and olivine phyric, with plagioclase
phenocryst contents ranging from 5% to more than 30%. Based on major element trends. Hickey-
Vargas ef al. (1989) interpreted that basalts with the highest Al,O, contents (>18%) contained
‘excess’ plagioclase crystals that accumulated by either floating, or by crystal entrainment during
magma ascentand eruption. Lopez-Escobar and Moreno (1994) observed that high Al,O,/CaO ratios
were associated with products of Villarrica’s most explosive pyroclastic eruptions. in general, the
origin of very high Al basalts a significant puzzle for petrologists what focused on subduction zone
magmatism (e.g.. Plank and Langmuir, 1988). Alternative explanations are that: 1) the abundant
feldspars are accumulative, even xenocrystals. so, the whole rock analyses do not reflect a liquid
composition, or 2) plagioclase crystallization is suppressed by moderate/high pressure, high PH o
crystallization, increasing the Al,O, contents of the liquid, followed by plagioclase crystallization
when the magma ascends to shallow depths and degasses."
The best approach to resolving this issue is through examination of mineral chemistry. and
ideally, the changing volatile contents of the magma. Sun (2001) analysed phenocrysts from 3
Villarrica basalts/basaltic andesites with SiO, <53% and MgO contents >6% (Fig. 2). Basalt with
relatively sparse phenocrysts have a bimodal distribution of plagioclase core compositions centred
at An... and An,,.,,. A highly phyric (30% phenocrysts) basalt analyzed for oxygen isotopes by
Hickey-Vargas ef a/ (1989) (Fig. 2) also has some An-rich cores, but the largest proportion of
plagioclase crystals have cores with An,,, .,. Bimodal phenocryst populations may indicate mixing
of mafic and evolved magma, however, in this case, mixing is not consistent with the exclusively
basaltic composition of the rocks or their high phenocryst content. It is more likely that the two.
generations of feldspars reflect polybaric crystallization, one stage at depth, under high Pro
conditions, and a second stage near the surface. where loss of water causes feldspar saturation
and reduced An contents. Since An-rich cores ate common, most erupted basaltic magmas
apparently undergo the first stage of crystallization, whereas the second stage can be by-passed
in some cases or extended in others. During the second stage, impediments to degassing could
result in the explosive behavior noted by Lépez-Escobar and Moreno (1994) as well as the
correlation with high Al,O, contents.
oT
20081
(Scar eine
Jac21 vance
sors
20,
Fa moisine
FIG. 2. Plagioclase and olivine composi
tions in three Villartica basaltic
andesites with dffenng phenocryst
S802 SiO: contents: 210281-1 (1971 flow)
FIG. 1. Variation of major element oxides versusSiO, for the volcanic 8% phenocrysts, 280976-1 (L21}
Units ofthe Vilarrica volcano. Data from Hickey-Vargas ef 14% phenocrysts, 190281-1 (V2-
(1989) and unpublished data trom Moreno (1993) 10-30% phenoerysts,2
ROLE OF THE CRUST
One of the most widely debated questions about SVZ magmatism is the importance of the crust
inchanging the composition of mantle-derived magma (¢.g., Hildreth and Moorbath, 1988). Aprimary
issue is whether the absence of typical indicators of crustal contamination in a volcanic suite
indicates a low extent of crustal involvement, or that the assimilated crust was chemically similar
to magma. In the central SVZ, chemically similar crust could be the numerous Miocene granitoid
plutons underlying active volcanic belt, voleanicand plutonic products of Quaternary volcanism, and
even partially solidified portions of the active sub-volcanic magmatic system. Older basement rocks
for Villarrica include Mesozoic and Paleozoic granitoids exposed at its base, and Paleozoic
subduction-related metamorphic rocks that hosted these intrusions.
Most geochemical indicators of assimilation of old sialic crust are absent from Villarrica volcanic
rocks. For example, *Sri®Sr ratios are low and uncorrelated with bulk composition, even for the
unusual rhyolite composition. 5°*O ratios are 6.1-6.2% in plagioclase in basalts and 5.8% for the
rhyolite whole rock (Hickey- Vargas efa/, 1989). These values are mantle like. andthe relatively high
value for the rhyolite mitigates against partial melting of plutonic or volcanic rocks altered by meteoric
water asa potential origin (e.g., Grunder, 1987). Perhaps the only indication of contamination by pre-
Cenozoic crustis "”Os/""Os measured in one Villarrica basalt that are slightly elevated compared
with MORB values (Chesley, personal communication). These values couldindicate assimilation of
ancient crust at depth, or alternatively, subduction mobilization of Os within the mantle wedge.
Trace element discriminants have been proposed to test for assimilation by young, chemically
similar crust. For example, McMillan efa/, (1989) and Dungan eff (2002) propose it that changes
in ratios of highly incompatible elements, like K/Rb, Rb/Ba and Rb/La in a differentiated sequence
may be a good indicator of crustal assimilation. For the Villarrica suite, Hickey-Vargas e¢a/. (1989)
argued that changes in these ratios from basalt to rhyolite could be caused by changing and
increasing bulk partition coefficients with increasing SiO, content. Although this hypothesisis borne
outby crystal chemical arguments, ithas never been tested directly. Variations between basaltand
andesite were modeled by mixing of basaltic andesite and rhyolite. The current, larger dataset (Fig
3) exhibits a larger range of K/Rb ratios than the smaller set modeled by Hickey-Vargas ef@/ (1989).
However, like the earlier data set, the widest range of K/Rb ratios is found at the matic end of the
series, with less variation among intermediate to silicic rocks. In general, therefore. such trace
element tests can be useful when changes are correlated with bulk composition. but variations
0 ao ~
o ‘Villarica S
BVillrrica 2
avitaricat |
Flank eruptions.
K/Rb
| FIG. 9, K/Rb versus SiO, for the
volcanic units of the Villarrica
Ca 6 voleano, Data sources as in
SiOz figure 18
among basaltic end-members can complicate their interpretation.
Current views of the nature of magma storage in the crust encompass a variety of assimilation
processes that have little impact on long-lived isotopes, such as assimilation of petrogenetically-
related wallrock and mixing of newly intruded basaltic magma with highly differentiated liquids
produced over time in the same magmatic system. in these cases, an effective indicator of
contamination can be the activity ratios of short-lived radioisotopes, like 2 Th/*Th or °Ra/2"Th,
which tend toward isotopic equilibrium with time, Sun (2001) measured U-series isotopes for one
Villarrica matic basaltic andesite (52.6% SiO,), and two higher silica basaltic andesites (55-56%
SiO, from Villarrica volcano. The more silicic samples plot closer to equilibrium values (the equiline.
Fig. 4) than the mafic sample, which is consistent with mixing of basaltic magma with silicic
differentiates from older parts of the system, or with magmatic assimilation of chemically-related
silicic voleanic or plutonic wallrocks. This pattem of decreasing activity with increasing SiO, is widely
observed among basaltic andintermediatesilicic rocks in subduction settings (Reagan efa/, 2003)
which suggests that assimilation of the older products of the same magmatic system is a common
process in both continental and oceanic arcs.
ea
4
to sobae ‘
&
E we
§ orten woo FO Sve
P| Sees
Pree 5 as
06 08 10 12
=The=Th
FIG. 4. (a) Th*Th versus**U/*Th (a) and (b) "Ber Be versus B/Be for matic and high-silica basaltic andesites from
the Villarrica volcano, and other SVZ volcanic centres. Villarrica samples are indicaled by the unit and SiO.
Content. Data from Sigmarsson e/ a/ (1990), Mortis er a/ (1990); Sun (2001), Hickey-Vargas ef a/ (2002)
land Sigmarsson ef af, (2002).
INFERENCES ABOUT MANTLE MELTING AND SUBDUCTION INPUTS.
Like most subduction related volcanic rocks, the products of the Villarrica volcano are depleted in
Nb and Ta relative to REE (rare earth elements) and enriched in LILE (large ion \ithophile elements)
relative to LREE (light REE) (Fig. 5). In the continuum of magma compositions observed in arcs.
Villarrica basalts are geochemically indistinguishable from those of oceanic island ares (e.g.
Hawkesworth ef a/, 1993), As discussed by Sigmarsson e¢a/ (1990); Morris ef.a/ (1990): Hickey-
Vargas eta/.(2002) and Sigmarsson e/a/ (2002), Villarrica basalts have detectible ""Be contents (Fig
4), which indicates that Be derived from subducted youngpelagic Nazca plate sedimentis incorporated
into their mantle wedge sources, and the primary magmas derived by melting of those sources. °U/
*=Thand**Ra/*"Th ratiosin Villarrica basalts are high. which indicates that excess™°U and"Rahave
recently been added to the sources of the magmas, in the case of Ra, within a few thousand years4“
prior to magma eruption. Current models of magma generation in ares are that basalt and sediment of
the subducted crust devolatilize and melt (sediment), and that hydrous fluids induce flux-melting of the
peridotite mantle wedge (¢.g., Plankand Langmuir, 1998). For Villarrica, mass balance estimates based
on the isotopic composition of Nd and Th in Villarrica basalt, indicate that only 36% of the Nd and 10%
of the Th is derived from the presubduction mantle wedge. and that 38% of the Nd and 68% of the Th
is derived from subducted sediment (Sun, 2001; Hickey-Vargas ef af. 2002)
Based on correlated B/Be, ""Be/"Be and 2"U/"Th ratios in volcanic rocks from several SVZ
volcanoes, including Villarrica, Morris efa/ (1990) and Sigmarsson e¢a/ (1990) suggested that the
composition of slab-derived fluids added to the mantle wedge was remarkably uniform along the
length of the SVZ, but variedin its amount at different volcanic centres. This correlation was shown
toextendto***Ra/*"Th by Sigmarsson efa/ (2002). Basalts from Villarrica, together with basalt from
Osorno, define the high B/Be. high '°Be/Be, high “°U/Th, high “*Ra/Th end of the correlation
line. According to Sigmarsson efa/. (2002), this observation indicates thatthe amount of sediment-
derived fluid added to the sources of these centres is greater than at others. An altemative
interpretation proposed by Hickey-Vargas efa/. (2002) is that the subduction-derived components
at Villarrica and similar centres have the shortest mantle and crustal residence times. Prolonged
contact of magma with Band "Be poor mantle and crust, combined with decay of U-series isotopes
overtime, resuitin the lower values foundat other SVZ centres, particularly those with more abundant
intermediate to silicic volcanic products. More thorough study of these chemical parameters at well-
characterised volcanic centres is needed to resolve these competing ideas.
100
Rock/Primitive mantle
GBR BRUT KL GR RF BN Inz HE GD OU Y be YW
FIG, 5. Primitive mantie normauisea trace element aounaance patierns for > voicanic rocks trom tne vuiarnica volcano,
Normalizing values are trom the GERM website: hitp:/earthref.org/GERM.
FUTURE DIRECTIONS
Work accumulated over the past 15 years shows that the Villarrica volcano exhibits qualities that
make it particularly useful for further study, using newly emerging approaches and geochemical
tools. Studies of phenocryst texture and chemistry, together with magmatic volatile contents
determined from melt inclusions, can help to clarify the origin of high Al basalts. the origin of basalticandesites and andesites, and the reasons for explosive volcanism. Study of phenocryst trace
element contents can help to distinguish the effects of cryptic crustal contamination and crystallisation
with changing partition coefficients. The strong signal of subduction indicators in Villarrica volcano
basalts, coupled with a small crustal signal and frequent recent eruption, make itone of the best SVZ
centres for examining and quantifying subduction zone fluxes.
ACKNOWLEDGEMENTS
R. Hickey-Vargas and M. Sun gratefully acknowledge support from NSF grant EAR 9725366.
Chilean researchers are indebted to Fondecyt grants No. 1911247, 1930992, 1831019.46
Magma degassing at Villarrica Volcano
Jeffrey B. Witter Deparment of Earth and Space Sciences.
University ol Washington, Seattle, WA 98195 USA
‘mil_98103yah00.com
Eliza. Calder
Volcano Dynamics Group. Department o! Earth Sciences,
The Opon University Milton Keynes, MK 2UN, U.K.
ES Calder@open ac.uk
ABSTRACT
Villartica voleano has been characterised by continuous, open-vent, passive degassing from a summit
lava lake, with negligible eruption of magmatic material, from the end of the last eruption in 1985 to the time.
of writing, In 2000, the emission rate of SO, at Villarrica volcano was measured as 2431155 ton/day SO,
In 2001, the emission rates of SO,, HCI, and HF at Villarrica voleano were measured as: 4644259 ton/day
SO,, 102442 ton/day HCI, and 13#3 tons/day HF. Diffuse degassing of CO, on the flanks of the volcano was
negligible. The authors used a variety of methods to constrain volatile emissions including: correlation
spectrometry (COSPEC and FLYSPEC), filter packs, anda soil gas flux survey. By combining the measured
gas emission rates with data on the pre-eruptive volatile content of the Villarrica magma, we found that, on
average. ~2.2 m! of magma is degassed per second (or ~5800 kg/sec). The authors interpret their volatile
flux measurements to represent the current baseline level of volcanic gas emission at Villarrica volcano.
Future measurements of gas flux at Villarrica volcano can be compared with this baseline level to assess
changes in the level of activity at the volcano.
INTRODUCTION
Villarrica volcano is notable in that it has persistently degassed from an active lava lake in its
‘summit crater from the end of the last eruption in 1985, and continues to do so now (Nov. 2003), a
period of 18 years. During this time, eruption of magmatic material has been negligible. In this
chapter, the authors describe the methods used to quantify gas emissions at Villarrica volcano, give
results for the current lux of acid gases (SO,, HCI. and HF) and present calculations of the degassing
rate of Villarrica magma. As far as the authors can determine, these are the first successful
measurements of gas emissions at Villarrica volcano to be published. Indeed, results of gas flux
measurements have been published for only a handful of Chilean volcanoes (@.g., Lonquimay and
Lascar, Andres e//, 1991). At present, gas emission measurements at Villarrica volcano are not
made on regular or ongoing basis. A continuous volcanic gas study program at Villarrica volcano
would be a fruitful avenue of research and a valuable too! for volcano monitoringa
GAS FLUX MEASUREMENT METHODS,
CORRELATION SPECTROMETRY (COSPEC)
Sulfur dioxide emission rates were measured using a correlation spectrometer (COSPEC; Stoiber
and Jepsen, 1973; Stoiber ef a/, 1983). The COSPEC contains an ultraviolet spectrometer that
measures the concentration-path length of SO, (in units of ppm-m) ina vertical column of atmosphere
above the instrument upwardly-pointing telescope. The COSPEC is driven or flown beneath, and
foughly perpendicularto, the downwind direction of the gasplume. The concentration-path length of
SO, is integrated across the width of the gas plume, and multiplied by the wind velocity and a
conversion factor to put the gas flux in units of ton/day SO,. A trigonometric correction is made for
portions of the transect that are not perpendicular to the gas plume
The largest source of errorin COSPEC measurements is in estimating wind velocity. The authors.
used three methods to estimate wind velocity. 1. The same day that ground-based COSPEC
measurements were made, they measured wind velocity at the summit of Villarrica volcano using
ahand-held anemometer. 2. During airbome COSPEC traverses, wind data were obtained from in-
flight GPS as wells by differencing the upwind and downwind groundspeed and dividing by two. This
method had the advantage of being a direct measurement of wind velocity at the same altitude as
the gas plume. 3. We supplemented their wind velocity measurements with winds aloft forecasts for
9,000 foot elevation obtained from the meteorological division of the Temuco airport. Errors in the
SO, emission rate associated with COSPEC data collection and reduction are estimatedto be ~20%
relative
In the year 2000, COSPEC measurements were made between January 29 and June 1. They
performed a total of sixty-three COSPEC traverses of the gas plume over the course of 14 survey
days using both airborne (17 traverses) and ground-based (46 traverses) methods. Inthe year 2001,
COSPEC measurements were made between January 15 and February 2. They made twenty-five
COSPEC traverses of the gas plume over the course of 4 days during the study period using only
ground-basedmethods,
CORRELATION SPECTROMETRY (FLYSPEC)
‘SO, emissions were also measured at Villarrica using FLYSPEC; anew-generation, miniature UV
spectrometer developed for field use at the University of Hawali (Horton ef a/. in press). The
instrument comprises a correlation spectrometer (USB 2000) with a small telescope that measures
the concentration-path length of SO, (in units of ppm-m) in the atmosphere in a similar fashion to
COSPEC. Power for the spectrometer is supplied through a mini-notebook computer additionally
equipped with a GPS. The retrieval method (described in Horton e/ /.. in press) benefits from real-
time comparison of reference tobackground spectrum to measured spectra, The software allows for
real-time display of the detected UV spectrum, calculation of the spectral absorbance and gas
concentration-pathlength values in ppm-m, and corresponding GPS position and time
High-resolution time series SO, concentration measurements were made at Villarrica crater on
December 16 and 17, 2002. Data were acquired for a total of 6, approximately one-hour scans
undertaken between 1330.and 1630LT (00:00 UTM) on December 16 and 17, 2002 and between 1230
and 1530LT (00:00 UTM) on December 17.2002, During this period, the height of the lava lake within
the crater was low (> 200 m lower than crater | rim) and beyond the field of view. However, continuous.8
gas puffing and audible explosions occurred with occasional events sending incandescent spatter
into view. FLYSPEC was deployed in the static mode, mounted on a tripod located 50 m NW of the
crater rim (2862 ma.s.|.). The instrument was set ata low angle targeting the plume center-point ~
200 mabovetthe ventas volcanic gas buoyed out of the crater and was blown eastwards (away from
the instrument). This method differs from the COSPEC surveys in that it does not acquire data for
a complete transect of the plume, and thus SO, flux data (ton/day) cannot be obtained (although
FLYSPEC can be used in a manor to achieve this).
Instead, high-resolution time series concentrations are acquired for a stationary point through
which the centerline of the plume passes. This allows a detailed quantification of how the gas plume
concentration varies over short time scales, and indeed can record the ‘puffing’thatis often visible
characteristic of the Villarrica volcano gas plume. Constraining, short term variations in this manner
isimportant in understanding the significance of the variation in the sparse COSPEC data and also.
when used in combination with other geophysical parameters, can provide insightful information on
conduit dynamics (Calder ef af, in prep.)
DIFFUSE DEGASSING OF CO,
The authors measured CO, gas flux through soil to test for diffuse degassing on the flanks of
Villartica using an instrument manufactured by West Systems of Pisa. Italy on loan from J. Rogie
(USGS. Menlo Park). The instrument consists of an accumulation chamber connected througha pump
and tubingtoa Drager infrared gas analyser. The instrumentmeasures CO, fluxin.units of ppmis, which
is then multiplied by the molecular weight of CO, and corrected for ambient pressure and temperature
conditions. Accounting for the volume and footprint (area) of the accumulation chamber, the CO, flux
is converted to units of g/m*day. Details of this technique are outlined in Evans ef af (2001).
FILTER PACKS,
The authors used filter packs in 2001 to measure the relative concentrations of acid gases (S,
Cl, and F) in Villarrica voleano gas plume. The filter pack method they employed (developed by P
Deimelie. Département. des Sciences du Sol, Université Catholique de Louvain, Belgium) collects
trace gases in the atmosphere using impregnated filters. It is analogous to gas filtration methods
used at Masaya (Stoiber e/a/, 1986); Erebus (Zreda-Gostynska e/a/, 1993), and Etna (Pennisi and
LeCloarec, 1998). The authors collected samples of gases using filter packs while standing in
Villarrica volcano gas plume on the rim of the actively degassing summit crater. The filter pack
instrument consists of a small vacuum pump which sucks the ambient gases (a mixture of
atmosphericair and volcanic gases) througha series of three paper filters (Whatman 2.2cm diameter
ashless Grade 42 filters) each impregnated witha saturated sodium bicarbonate solution. The gases
are pre-filtered with a 10 um Teflon mesh to minimize intake of airborne ash particles. Total filtration
time varied from 3 to 46. minutes. Filtration time was chosen so that nearly all of the acid gases
were absorbed by the first two filters in the filter stack. Relative concentrations of the acid gas
species (SO,, Cl, and F) absorbed onto the filter paper were obtained by ion chromatography at
Département des Sciences du Sol, Universite Catholique de Louvain, Belgium (P, Deimelle, analyst)
Twenty-three gas measurements were made using filter packs over the course of eight summit days.
They authors also sampled acid gas ratios from the Villarrica gas plume using volatile traps (small
buckets containing a KOH solution placed on the rim of the degassing crater) but we found the filter
pack technique was much less labor intensive and gave superior results.®
RESULTS
COsPEC
Inthe year2000, COSPEC measurements yielded an overall average (#16) SO, flux of 243+155
tons/day SO, (Witter e/4/, 2004). In2001, COSPEC measurements yielded an overall average (10)
SO, flux of 464259 tons/day SO, (Witter ef@/, 2004). individual measurements and daily averages
of COSPEC traverses are shown in table 1. The overall average SO, flux in 2001 is greater than, but
within, the 1 variation of the overall average SO, flux measured in 2000, which may suggest SO,
flux at Villarrica does not change significantly on an annual basis. The highest values obtained for
each year are also relatively consistent (Fig. 1), indicating that the variation in the yearly averages.
to some extent may be an artifact of survey frequency. Furthermore. considering the ~20% relative
errorinherentto the COSPEC method, the daily to weekly variationsin SO, flux measured at Villarrica
volcano during 2000 and 2001 (Fig. 1; Table 1) are real. Variations of this magnitude are not unusual
fora volcano of this type. For comparison, Mather ef. (2004) measured an SO, flux of ~350 tons/
day at Villarrica volcano in February 2003 using Differential Optical Abosorption Spectrometry
{DOAS). In addition, two COSPEC surveys from neighboring Liaima volcano on April 14 and 12, 2000
gave average SO, fluxes of 642 and 597 ton/day. respectively (E.S. Calder, unpublished data).
1200
1000
800
600
400
200
0
Jani Mar Mayi—Jun/30Aug/30_—Ocv29.—«(Deci29_—«Feb/28
Fe000—> 2001 >
Date
SO, Flux (bday)
FIG. 1. SO, flux data for each of the COSPEC traverses conducted at Villarica during the period 29, January 2000 to
February 2 . 2001. Individual runs (open diamonds) and daily averages (filed diamonds) for each survey day
are plotted
FLYSPEC
High resolution time series SO, concentration-pathlength results (ppm-m)and corresponding UTM
time are shown in figure 2. The complete scans for each day (Fig. 2a and c) show SO, concentration
variations between a few 10's to over 2000 ppm-m. Two periods with expanded time axes (Fig. 2b
and d), show gas plume ‘puffing’, which is considered to be related to the ascent of large gas slugs
in the conduit (Calder e¢2/, in prep.). From this data, itis evident that puffing episodes can vary in
time-scales (from 40 to 72 second intervals between puffs) and magnitude (from an average of 500
to 1500 ppm-m). Periods where degassing seems to occur more continuously also exist.