Earth and Planetary Science Letters 388 (2014) 166–174

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

The influence of cooling, crystallisation and re-melting

on the interpretation of geodetic signals in volcanic systems
Luca Caricchi a,b,∗ , Juliet Biggs b , Catherine Annen b , Susanna Ebmeier b
a
Section of Earth and Environmental Sciences, University of Geneva, Rue des Maraîchers 13, CH-1205, Geneva, Switzerland
b
School of Earth Sciences, University of Bristol, Wills Memorial Building, Queen’s Road, Bristol, BS8 1RJ, UK

a r t i c l e i n f o a b s t r a c t

Article history: Deformation of volcanic edifices is typically attributed to the movement of magma within the volcanic
Received 4 May 2013 plumbing system, but a wide range of magmatic processes are capable of producing significant volume
Received in revised form 9 October 2013 variations and may also produce deformation. In order to understand the evolution of magmatic systems
Accepted 1 December 2013
prior to eruption and correctly interpret monitoring signals, it is necessary to quantify the patterns and
Available online xxxx
Editor: L. Stixrude
timescales of surface deformation that processes such as crystallisation, degassing and expansion of the
hydrothermal system can produce. We show how the combination of petrology and thermal modelling
Keywords: can be applied to geodetic observations to identify the processes occurring in a magmatic reservoir
geodetic signals during volcanic unrest. Thermal modelling and petrology were used to determine the timescales and
petrology volumetric variations associated with cooling, crystallisation and gas exsolution. These calculations can
thermal modelling be performed rapidly and highlight the most likely processes responsible for the variation of a set
of monitoring parameters. We then consider the magnitude and timescales of deformation produced
by other processes occurring within the vicinity of an active magma system. We apply these models
to a time series of geodetic data spanning the period between the 1997 and 2008 eruptions of
Okmok volcano, Aleutians, examining scenarios involving crystallisation, degassing and remelting of the
crystallising shallow magmatic body and including a viscoelastic shell or hydrothermal system. The
geodetic observations are consistent with the injection of a water-saturated basalt, followed by minor
crystallisation and degassing. Other scenarios are not compatible either with the magnitude or rate of
the deformation signals.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction
Driven by the increasing availability of satellite technology such
as InSAR and GPS, the number of volcanoes for which geodetic
measurements exist is growing rapidly, as is the spatial and tempo-
ral resolution of those measurements (e.g. Sparks et al., 2012). The
global catalogue of volcanoes known to be deforming now exceeds
one hundred (Fournier et al., 2010), but remarkably few of these fit
simple cycles of magma chamber recharge and discharge within an
elastic medium. A new paradigm is required. In this paper, we cou-
ple existing thermal models with petrologic constraints to estimate
the contributions of crystallisation and degassing on the magni-
tude and time scales of surface deformation. We then consider the
timescale and magnitude of other processes, such as perturbation
of the hydrothermal system and evolution of the physical prop-
*Corresponding author at: Section of Earth and Environmental Sciences, Univer-
sity of Geneva, Rue des Maraîchers 13, CH-1205, Geneva, Switzerland. Tel.: +41 22
37 96 630.
E-mail address: luca.caricchi@unige.ch (L. Caricchi).
erties of the crust, which are known to play a role in controlling
deformation in regions of active volcanism.
Of over a hundred volcanoes known to be deforming, roughly
twenty volcanoes are thought to be undergoing long-term sub-
sidence (Table 1). This has been attributed to the cooling and
crystallisation of magma bodies (e.g. Medicine Lake; Poland et al.,
2006) or associated hydrothermal activity (e.g. Kiska; Lu et al.,
2010; Lu and Dzurisin, 2010). Sudden pulses of uplift are typically
linked to magmatic intrusions (e.g. Three Sisters; Wicks, 2002),
but discrete pulses of subsidence (e.g. Suswa; Biggs et al., 2009)
are more challenging to explain. In several cases, exponentially
decaying subsidence is observed immediately following pulses of
inflation (e.g. Okmok; Fournier, 2008; Fournier et al., 2009; Alutu;
Biggs et al., 2011). The diversity of these observations illustrates
that we must consider not only the spatial pattern and magnitude
of the deformation, but also its temporal evolution if we are to
identify the processes driving volcanic unrest.
In this paper, we use Okmok volcano, Alaska (Fig. 1A) as a case
study because a 15-year geodetic record including InSAR, contin-
uous and campaign GPS measurements is available. The time se-
ries, shown in Fig. 1B, covers major eruptions in 1997 and 2008
and provides detailed time-series of inter-eruptive deformation

0012-821X/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.epsl.2013.12.002
 L. Caricchi et al. / Earth and Planetary Science Letters 388 (2014) 166–174 167

Table 1
Global compilation of subsiding volcanoes.

Volcano Method Rate Diameter Inferred References

and (cm/yr) (km) depth
coverage (km)
Aniakchak I: 92–02 1.3 (const.) 10 ∼4 Kwoun et al., 2006
Askja L: 83–98 5 (83–98), <25 3.2–3.8 Sturkell et al., 2006;
G: 93–98 2.5–3.0 (00–09), Pagli et al., 2006;
I: 92–00 decaying de Zeeuw-van Dalfsen et al., 2012
Campi G: 85–05 1.4–2.8, ∼12 2–4.2 Fernandez et al., 2001;
Flegrei I: 93–96 (decaying Lundgren et al., 2001;
1985–) Gottsmann et al., 2006.
Cerro Blanco I: 92–00 1.8–2.5 <30 <4.8 Pritchard and Simons, 2004
(decaying)
Copahue∗ I: 02–08 2 (const.) 2–3 ∼4 Euillades et al., 2008; Fournier et al., 2010;
Velez et al., 2011
Fisher Caldera G: 99–01 – >10 >1.6 Mann and Freymueller, 2003
Kiska∗ I: 95–01 1.6 3 <1 Lu et al., 2002
Krafla L: 89– 5 (89–92) ∼10 1.5–2.5 Sigmundsson et al., 1997;
G: 93–95 0.3 (00–05) de Zeeuw-van Dalfsen et al., 2006;
G: 00–05 Sturkell et al., 2008
I: 92–
Medicine G: 96–04 ∼0.9–1.2 ∼12 – Dzurisin et al., 2002;
Lake L: 54, 89, Poland et al., 2006
90, 99
I: 93–00
Sakurajima L ∼0.1 (79–) Oct-20 – Dvorak and Dzurisin, 1997;
Mogi, 1958;
Yokoyama, 1986.
Okmok I: 92–12 Variable ∼8 03-Apr Miyagi et al., 2004; Lu et al., 1998, 2000, 2005,
G: 00–12 (Fig. 1) 2010; Lu and Dzurisin, 2010; Mann et al., 2002;
Fournier et al., 2009; Biggs et al., 2010
The table lists the details of the observation method and temporal coverage, the rate and temporal behaviour of the subsidence, the diameter of the deformation pattern
and the estimated source depth. Okmok volcano is given for comparison. G—GPS, I—InSAR, L—Leveling. ∗ refers to volcanoes at which the deformation has been attributed to
hydrothermal processes.

(Lu et al., 2010; Lu and Dzurisin, 2010; Fournier et al., 2009;
Biggs et al., 2010). The petrology of this system is well charac-
terised, which is fundamental for quantifying the volumetric vari-
ations associated with crystallisation and eventual exsolution of
volatiles from the residual melt phase.
Following a magma intrusion, each part of a magma body will
experience a unique cooling history depending on the depth of
emplacement, thermal state of the crust, and magma composition,
which ultimately controls the variation of crystallinity and volume
with decreasing temperature. Cooling leads to magma crystallisa-
tion and because minerals are generally denser than the melt from
which they crystallise, to a decrease in volume (Lange, 1994). The
rate of crystallisation, and therefore of volumetric variation, is a
function of the shape, area and volume of the cooling magma in-
trusion, the thermal state of the surrounding rocks and the amount
of latent heat released during crystallisation. Hence thermal mod-
elling is necessary to calculate the appropriate rates of magma
contraction or expansion that could be the source of deformation
observed in volcanic systems. In volatile bearing magmas, crystalli-
sation is accompanied by a relative enrichment of volatiles in the
residual melt. Once the saturation limit is reached, volatiles exsolve
from the residual melt generating gas bubbles (second boiling) and
leading to a volumetric increase of the crystallising magma. The
variations of magma density during cooling control the rate and
magnitude of volumetric variation and thus surface deformation.
Any model of subsidence in a magmatic system must therefore
incorporate not only appropriate thermal models, but also petro-
logical information on the relationships between temperature and
melt/crystal/volatile fractions.
Here we focus on three scenarios that could be envisaged to
explain the sequence of inflation and deflation signals observed
at Okmok between 1997 and 2008: (1) injection of basalt in the
country rock followed by crystallisation; (2) injection of H2 O bear-
ing basalt followed by cooling and degassing; (3) interaction be-
tween the injected H2 O bearing basalt and a partially crystallised
andesitic magma that may be present in the system as a residuum
of previous eruptions (Finney et al., 2008). Here crystallisation,
melting and gas exsolution are considered to have only volumetric
effects:
(1) In the first scenario, cooling of a volatile-undersaturated
magma produces volumetric contraction with an associated
subsidence rate that can vary depending on the ratio between
the density of the crystallising phases and the density of the
residual melt along the crystallisation sequence, the shape and
dimensions of the magma intrusion, the magma composition
and the thermal state of the crust.
(2) If the magma present in a reservoir is volatile saturated, crys-
tallisation leads to the exsolution of gas bubbles, which, in
turn, within years, results in an inflation signal not related
to magma injection and eventual subsidence if the system
becomes permeable and gas can be expelled from the subvol-
canic reservoir.
(3) For the last scenario, the injection of a volatile-saturated
magma into a crystal mush can potentially cause partial re-
melting of the host. In this case, volatiles released from the
injected magma can potentially be resorbed by the volatile-
undersaturated melt produced by re-melting. The result is in-
flation (exsolution) followed by subsidence (resorption) but no
significant degassing observable at the surface.
In the following, we show how geodetic measurements, de-
tailed petrology, and thermal modelling can be combined to iden-
tify the most likely processes occurring in a magmatic reservoir
168 L. Caricchi et al. / Earth and Planetary Science Letters 388 (2014) 166–174
Fig. 1. (A) Map showing the location of Okmok volcano and the position of the GPS
stations (after Fournier et al., 2009). (B) Variation of volume in the Okmok mag-
matic system between 1997 and 2008 (after Biggs et al., 2010). The volumes were
obtained by inverting InSAR, continuous and campaign GPS signals using a fixed-
position Mogi model. The deflation observed between 2005 and 2006 corresponds
to about 10% of the total volume increase in the system and is the subject of this
study.
and eventually preceding an eruption. Finally, we consider the
thermal effects of the injection of magma on the surrounding
crust. In some cases, the increase in temperature is sufficient to
remelt the rocks surrounding the intrusion, but even smaller in-
creases in temperature will alter the rheology of cold country rock.
This has led to the concept of a viscoelastic shell surrounding
the magma reservoir (e.g. Dragoni and Magnanensi, 1989). Fur-
thermore, hot magma reservoirs are typically associated with hy-
drothermal systems and a change in temperature is likely to alter
the pattern of fluid circulation. Both these processes may produce
geodetic signals and/or affect the rate of inflation/deflation.
1.1. Case study: Okmok volcano
Okmok is a shield volcano located on Umnak Island in the Aleu-
tian Chain (Fig. 1), and it has erupted every 10–20 years for the
past 200 years (Miller et al., 1998). In 1997, the eruption originated
from Cone A (Fig. 1A) and produced a basaltic lava flow 5 km long
and tens of meters thick (Lu et al., 2003), while the 2008 eruption
produced new vents located near Cone D (Fig. 1A) and produced
mainly basaltic-andesite tephra, as a result of magma interaction
with the water table (Larsen et al., 2013). The 10-km diameter
summit caldera was formed by eruptions 12,000 and 2050 years
ago (Miller and Smith, 1975; Beget et al., 2005), which deposited
dominantly andesitic ash-flow units.
The petrological characterisation of the products places con-
straints on the interaction between magmas injected at depth and
those residing in the shallower part of the system and suggests
a chemically stratified magmatic system at depths of 2.5–5 km.
The most evolved products seem located in the shallowest part
of the systems and more basaltic magmas occupy its deepest por-
tions. The temperature in the plumbing system varies from 850 to
1125 ◦ C, with water contents ranging from 2 to less than 6 wt%
(Finney et al., 2008).
Geodetic observations began at Okmok volcano in 1992 and
since then a combination of campaign and continuous GPS and In-
SAR have been used to measure the surface deformation (Lu et al.,
1998, 2000, 2005, 2010; Miyagi et al., 2004; Fournier et al., 2009;
Lu and Dzurisin, 2010; Mann et al., 2002; Freymueller and Kauf-
man, 2010). The broad pattern is that of rapid, metre-scale sub-
sidence during the 1997 and 2008 eruptions and uplift during
inter-eruptive periods (pre 1997 and 1998–2008). All GPS and In-
SAR displacements can be reproduced well using a simple point
source model (Mogi, 1958) at a depth of 3.4 km (Biggs et al., 2010;
Lu et al., 2010; Lu and Dzurisin, 2010; Fournier et al., 2009) al-
though more complex rheologies and geometries have also been
proposed (Masterlark, 2007; Masterlark et al., 2010).
Biggs et al. (2010) combined campaign and continuous GPS
with InSAR to produce a time-series of volume flux, which spans
the entire geodetic record (1992–2007) and captures both the spa-
tial resolution of InSAR and the temporal resolution of the contin-
uous GPS. There are significant variations in the rate of uplift and
periods of minor subsidence not associated with eruption (Fig. 1B;
Fournier et al., 2009; Biggs et al., 2010). Uplift in 1992–1995 was
followed by subsidence in 1995–1996 preceding the 1997 erup-
tion. Uplift resumed from 1997 to 2002 followed by a major pulse
of uplift in 2002–2004. The 2002–2004 uplift can be further sub-
divided into two sub-periods of uplift (summer 2002 to late 2003,
and spring–summer 2004) interposed by a short repose or possibly
subsidence in early 2004. A further period of subsidence followed
in 2004–2005. The volcano then entered a 3-year period of quies-
cence that ended with a 6–7 month period of inflation leading to
the eruption in July 2008.
The deformation during the 1992–2005 period was associated
with a single, static source, but the pre-, co- and post-eruptive de-
formation associated with the 2008 eruption occurred in a new
source that changed in both location and geometry with time. De-
flation during the 2008 eruption is attributed to pressure decrease
in a source 2.1 km below sea level, notably shallower than that
associated with the 1997 eruption. Deflation of a deep source cou-
pled with re-inflation of a shallow change began a few weeks after
the eruption.
This study focuses on the inter-eruptive deformation that oc-
curred in 2002–2005. The peak uplift in 2003 corresponds to vol-
ume change of 0.03 km3 while the subsequent deflation corre-
sponds to a volume decrease of 0.003 km3 , or 10% of the original
input volume (Fig. 1B). Although the 2008 eruption will have sig-
nificantly altered the patterns of magma storage only 6 years after
the magma input, we run our models forward for 60 years to cap-
ture processes occurring over longer timescales.
A previous study constrained the variations of volume and pres-
sure responsible for the observed deformation using thermody-
namic models to describe magma properties and a mechanical
model that accounts for the properties of the crust surrounding
the magma chamber (Fournier, 2008). By performing thermal mod-
elling we can add temporal constraints and account for the ef-
fects of magma properties on the deformation signals observed at
Okmok between 1997 and 2008.
 L. Caricchi et al. / Earth and Planetary Science Letters 388 (2014) 166–174 169

2. Methodology

2.1. Thermal modelling

A range of simulations were performed to model the injec-

tion of basalt at 1075 ◦ C (near-liquidus temperature of an arc-
type water-saturated basalt similar to the basic magma erupted at
Okmok; Wagner et al., 1995) at 4 km depth either in a cold crust
(T crust = 150 ◦ C) or in a reservoir of partially crystallised andesite
(T = 850 ◦ C, crystal fraction = 0.5; Moore et al., 1998). The evolu-
tion of temperature for basalt, crust and andesitic magma was cal-
culated using an explicit finite difference scheme (see Appendix A).
The model includes heat transfer by conduction and latent heat
produced or absorbed by crystallisation or melting. The consump-
tion of latent heat related to gas exsolution is neglected because it
represents a relatively small contribution with respect to the la-
tent heat released by crystallisation (Sahagian and Proussevitch,
1996). The model uses cylindrical coordinates with temperatures
and melt fractions calculated on a 2D vertical slice of a 3D system
characterised by a rotational symmetry around its vertical axis. The
progressive injection of basalt is simulated with discrete emplace-
ment of successive magma pulses, which are wafer-shaped with a
diameter of 400 m. Each injection is emplaced in the middle of the
former one to simulate the thickening of a sill-like body (Annen,
2011). The effect of convection is not calculated in the thermal
model, however will be considered later in the discussion of the
timescales of volumetric variations.

2.2. Petrology

In order to calculate volume changes caused by cooling, crys-

tallisation, re-melting, it is necessary to have a complete descrip-
tion of the crystal and gas volume fraction, as function of tem-
perature for the different magmas present in system. Water is the
most abundant gaseous phase in volcanic systems and hence the
only one considered here. We use petrological data from Wagner
et al. (1995) and Moore et al. (1998) for arc basalts and an-
desites from experiments performed at 1 kbar, temperatures be-
tween 1100 and 800 ◦ C and 3.15–4.05 wt% H2 O (water saturated
conditions at 1 kbar for basalt and andesite respectively; Newman
and Lowenstern, 2002). The pressure condition was chosen be-
cause it is consistent with the geodetic estimate of 3–4 km for
the magma chamber depth (e.g. Biggs et al., 2010; Fournier et
al., 2009). To calculate the phase equilibria for the Okmok basalt
containing 2 wt% H2 O (such water content, and higher ones, are
measured in melt inclusions for this volcano; Finney et al., 2008)
and to extend the temperature range over which experimental
data are present, we used the thermodynamic software MELTS
(Ghiorso, 1985; Ghiorso and Sack, 1995). We first modelled the
water-saturated composition and compared the results with the
experimental data finding good agreement (Fig. 2A). Given the
good match, we used MELTS to compute the phase equilibria for
the Okmok magma containing 2 wt% H2 O (Finney et al., 2008;
Larsen et al., 2013). All the calculations were performed at a con-
stant confining pressure of 1 kbar. MELTS was also used to calcu-
Fig. 2. (A) Variation of crystal mass fraction with temperature for the different mag-
late the density of the crystallising minerals, residual melt, and ex- mas used in the calculations. The diamonds are experimental data and the lines
cess gas volume present in the system (Fig. 2B and C). We then pa- were calculated using the MELTS software (Ghiorso, 1985; Ghiorso and Sack, 1995).
rameterise the volume–temperature relationships and couple them (B) and (C) show the relative volumetric fraction of minerals and melt as function
of temperature calculated at a constant confining pressure of 1 kbar for a water sat-
with the thermal model to trace the evolution of temperature and
urated magma and a magma containing 2 wt% water, respectively. The variations
volume associated with a magmatic injection for comparison with of magma relative volume during cooling are calculated assuming that excess water
the geodetic observations. is continuously removed from the magma (thick dashed line), and considering that
all the excess water is maintained in the magma and no outgassing occurs (thick
3. Volume balance black line).

Analytic models of the geodetic observations suggest that the 0.03 km3 of magma at depths of 3–4 km while the 2005–2006
total uplift from 1997 could be explained with the injection of subsidence signal requires a volume decrease 0.003 km3 (10% of
170 L. Caricchi et al. / Earth and Planetary Science Letters 388 (2014) 166–174

the initial volume) in less than 1 year (Fig. 1; Biggs et al., 2010).
The main target of our study is to determine the process respon-
sible for this subsidence. All our calculations assume that the total
amount of magma injected in the system is the volume obtained
from the inversion of data recorded since 1997.
We start testing the hypothesis that the geodetic observa-
tions are an elastic response to change in volumes related to
magma injection, cooling and crystallisation, gas exsolution and
older magma remelting. We numerically simulated the injection
of magma at a rate that approximates the increase of volume cal-
culated from the inversion of the geodetic data. The results of the
thermal modelling were used to calculate the possible volumetric
variations associated with magma cooling. Fig. 3 shows the dif-
ferent effects of cooling, crystallisation, exsolution of gas from the
crystallising magma, and re-melting, on the volume of the subvol-
canic reservoir.
These calculations provide both the magnitude and the rate of
volumetric variations in time, which are fundamental to identify-
ing the possible processes operating in the magmatic reservoir and
responsible for uplift or subsidence.

3.1. Effect of crystallisation

The injection of magma is followed by cooling, which, in turn,

leads to crystallisation of minerals with higher density than the
initial melt and in absence of gas, this would result in contraction.
Petrological studies of Okmok, however, suggest that the magma
in the magmatic system contains volatiles (Finney et al., 2008;
Larsen et al., 2013). Therefore, for crystallisation to account for the
deflation, excess volatiles should be continuously removed from
the system during crystallisation. This is physically difficult be-
cause the presence of crystals tend to inhibit the buoyant rise of
bubbles in magma (Belien et al., 2010), even if deformation could
facilitate the escape of gas bubbles (Caricchi et al., 2011). Assuming
that volatile saturation results in continuous degassing, a volumet-
ric contraction of a volumetric contraction of 10% (i.e. a total vol-
ume fraction of magma of 0.9; Figs. 2B and 3) would be achieved
at a temperature of about 1000 ◦ C (Fig. 2B; in the model basaltic
magma is injected at its liquidus temperature of 1075 ◦ C). At this
temperature the crystal mass fraction is 0.44. However, crystalli-
sation cannot be responsible for the observed deflation because
the thermal model shows that even with injection in relatively
cold crust (150 ◦ C, unlikely for an active system such as Okmok)
more than 50 years are required for a volume of 0.03 km3 of
basalt to cool to 1000 ◦ C and account for the relative decrease of
volume of 10% (Fig. 3A). Depending on the magma body geome-
try and thickness, magma convection might accelerate cooling and
crystallisation. However, cooling of a convecting magma layer is
limited by the capacity of the country rock to conduct away the
magma heat. Analysis by Worster et al. (1990) of the solidification
of a convecting magma chamber cooled from above by conduction
through the country rock indicates that the solidification time for
a convecting magma layer is about half the time predicted by con-
ductive cooling. Thus, even if the magma is vigorously convecting,
crystallisation alone cannot explain a 10% volume contraction over
less than 1 year from the end of the injection episode as recorded
Fig. 3. Volume of crystal, melt and gas phases with time, normalised to the max-
at Okmok (Fig. 1B).
imum volume obtained from the inversion of geodetic data at the end of the
inflation period. The first 8 years of the thermal modelling are inspired by the
3.2. Crystallisation, second boiling and gas release geodetic records in the period 1997–2004. (A) H2 O-saturated basalt injected into
crust at 150 ◦ C. (B) Basalt containing 2 wt% H2 O injected into crust at 150 ◦ C.
The crystallisation-only model implies complete and continuous (C) H2 O-saturated basalt injected into crust at 850 ◦ C. Volume fraction–temperature
relationships are taken from Fig. 2 and temperature–time relationships from the
degassing, yet crystals tend to trap gas bubbles (Belien et al., 2010).
thermal model. The arrows in (A) show the necessary conditions for 10% deflation
Alternative models may account for the delay between the end of based on two assumptions: (1) If we assume all gas is lost instantly, the volume of
the magma injection episode and the onset of degassing: (1) gas melt+crystal falls to 90% of the peak value 52 years after the start of injection (in
is initially trapped in the magma until bubble coalescence leads to the year 1997). (2) If we assume the gas is stored in unconnected bubbles, the gas
accumulated accounts for 10% of the total volume by the time injection is complete.
a significant increase of permeability; (2) enough time is required
 L. Caricchi et al. / Earth and Planetary Science Letters 388 (2014) 166–174
for buoyant gas bubbles to rise and leave the system and the mag-
matic reservoir starts to release the gas and deflates. Considering
for simplicity that the excess gas at intrusion depth (3–4 km)
is principally water, that the intruding fully molten basalt has a
density of 2700 kg/m3 (Lange, 1994) at 1075 ◦ C and the molar
volume of H2 O at these conditions is 110 cc/mol (Ghiorso, 1985;
Ghiorso and Sack, 1995), approximately 0.6 wt% excess water has
to be released from the injected 0.03 km3 of basalt to account for a
volume change of 0.003 km3 (i.e. 10% of the total injected volume).
Fig. 3A shows that by the time the injection is completed, a rela-
tive volume of 10% excess gas has already been accumulated in the
system, the release of which would account for the deflation ob-
served between 2005 and 2006. Considering that convection would
have increased the cooling rate of the basalt and also taking into
account the uncertainty on the recalculated volumes, the calcu-
lated time is in good agreement with the time interval of almost
10 years between the onset of magma injection and the beginning
of deflation (Fig. 1).
Given the rate of deflation measured between 2005 and 2006,
we can also calculate an approximate gas flux that would have
been measured during this period. Considering a weight ratio
S/H2 O for the basalt of about 0.04 as measured in melt inclusions
in Okmok volcanic products (Zimmer et al., 2010), the mass of S
released daily would be a modest 100 tons. Clearly the reality lies
between these end-member models and the timescale over which
the gas would be released would depend on the mechanism of de-
gassing. Additionally, the gas fluxes measured at the surface could
be affected by the presence of a hydrothermal system and there-
fore for such relatively small quantities of gas it may be difficult to
link degassing and deflation rates.
So far we have considered the injection of a water-saturated
basalt. Fig. 2C shows that for a water undersaturated basalt con-
taining 2 wt% H2 O, no excess gas starts to accumulate in the sys-
tem until the magma reaches an average temperature of 1050 ◦ C.
This could suggest that the first modest subsidence episode
(2000–2002; Fig. 1B) following the first inflation event, could re-
sult from cooling and crystallisation and the second uplift event
could reflect the point at which the basalt reaches second boil-
ing and expands (Fig. 2C). However, this is not physically possible
because if all the inflation would be produced by the volumetric
expansion of magma at depth, this would imply an expansion of
about 200% (Fig. 1B), and the compressibility of gases at pressures
compatible with the depth of the magma chamber is not com-
patible with such extreme increase of volume (Pitzer and Sterner,
1994). On the other hand, the magnitude of the last inflation mea-
sured between before the 2008 eruption could be explained by the
exsolution of about 10 vol% volatiles from the magma. The thermal
model indicates that the rate of inflation observed before eruption
is too elevated to be accounted for by the second boiling of magma
(Figs. 1B and 3B).
In conclusion, (1) the deflation observed between 2005 and
2006 cannot be explained by crystallisation alone and can only
be explained by degassing if the injected magma was volatile
saturated. (2) The second inflation episodes cannot be related to
second boiling. (3) The last minor inflation episode preceding the
eruption occurred with higher rates than those that would be pro-
duced by second boiling. Therefore our preferred explanation for
this last event is the injection of another, smaller, pulse of magma
into the magmatic system.
3.3. Interactions with host andesite
At frequently erupting volcanoes, new magma is likely to be
injected into a system that contains residual magma from pre-
vious eruptions. At Okmok, a major caldera-forming eruption re-
leased andesitic magma (Finney et al., 2008), which suggests the
171
presence of chemically evolved bodies of magma in the plumb-
ing system. The magma erupted in 2008 had a more chemically
evolved composition with respect to previous basaltic eruptions
(Larsen et al., 2013), which could be related to interaction be-
tween basaltic magma and more chemically evolved magmas. We
use our thermal models to simulate the injection of basalt at
1075 ◦ C into an andesitic magma body at 850 ◦ C with a crystal
fraction of ∼0.5 (Moore et al., 1998; Fig. 2). We consider also
that the excess gas in the andesite magma left the system in the
time preceding the unrest that started in 1997. The volume of
0.03 km3 of basalt leads to the re-melting of about 0.005 km3 of
crystals from the andesite producing a water-under-saturated melt
(Fig. 3C). This figure is for conductive heat transfer between the
basalt and the andesite. We notice that convection of the basalt
layer and of the andesite would result in more basalt crystallis-
ing and more andesite remelting (Huppert and Sparks, 1988). The
re-melting has two main effects on the deformation signals: (i) Be-
cause the density of melt is lower than the density of crystals,
partial re-melting produces an additional volume increase which
contributes to the uplift signal (Fig. 3C). (ii) Excess volatiles re-
leased from a water-saturated basalt can be resorbed by the re-
molten andesite causing subsidence without any surface degassing.
The volume of re-molten andesite is in fact sufficient to resorb the
excess water in the basalt and account for the total subsidence
measured.
4. Thermal effects on the surrounding crust
The previous section discusses the subsurface volume changes
caused by the thermal evolution of a magma intrusion. In order
to link these processes to observations of surface deformation, we
have made the assumption of a purely elastic, homogeneous crust
implicit within the Mogi model (Mogi, 1958; Biggs et al., 2010). In
this section, we discuss the effects of intrusions on the surround-
ing crust and hydrothermal system. We use an energy balance to
discuss the thermal expansion associated with heat input into a
hydrothermal system and use an analytic solution for the effects
of pressure change within a viscoelastic shell (Segall, 2010). The
results allow us to estimate the direction and order of magnitude
of the deformation resulting from each mechanism independently.
Multiple mechanisms may act together and feedbacks between
them, for example, crystallisation of a magma intrusion within a
viscoelastic medium, would alter the spatial and temporal patterns
of surface deformation hence we do not try to fit the data with
any individual mechanism.
4.1. Rheology
The high-temperature environment around an active magmatic
system favours a viscoelastic rheology over a purely elastic one,
leading to a time-dependent response to volumetric and baro-
metric variations in the plumbing system of an active volcano.
Masterlark et al. (2010) constructed finite element models to simu-
late subsidence associated with the 1997 eruption of Okmok based
on a magma reservoir that is surrounded by a viscoelastic rind
of country rock embedded in an elastic domain, with the shal-
low caldera fills represented by lower density and elastic mod-
uli.
Here, we use a simpler analytic solution to consider the direc-
tion, timescale and magnitude of deformation following a pressure
change of a magma reservoir surrounded by a viscoelastic rind
within an elastic halfspace. The characteristic relaxation time, tr ,
is given by the equation (Segall, 2010)
3η(1 − υ ) R 32
tr =
3μ(1 + ν ) R 31
172 L. Caricchi et al. / Earth and Planetary Science Letters 388 (2014) 166–174
where η is the viscosity of the shell; μ is the shear modulus; ν
is Poisson’s ratio; R 1 is the radius of the chamber and R 2 is the
radius of the shell. The displacements have the same spatial distri-
bution as for a Mogi source, but following the instantaneous elastic
response, the displacements continue to increase at an exponen-
tially decreasing rate that scales with t R (Segall, 2010).
An andesite at T = 850 ◦ C with a crystal fraction of 0.5 would
be expected to have a Newtonian (i.e. maximum) viscosity of
the order of 107 –108 Pa s (considering a water saturated rhyolitic
melt; Giordano et al., 2008; Caricchi et al., 2007). Using this vis-
cosity, even for shells 10 times wider than the magma reservoir
(R 2 / R 1 = 10), the relaxation time is still less than 1 s. From a
monitoring perspective, any viscoelastic response from this part of
the magma chamber would be included within the initial intru-
sion. Further from the intrusion, the wall rock will remain solid,
but may still behave viscoelastically and viscosities on the order
of 1017 are required to produce relaxation times on the order of
a year. The dimensions of the thermal aureole (and hence R 2 / R 1 )
depends on the intrusion history: at low rates of emplacement,
the thermal profile will be the same as for an individual intrusion,
but if new magma is supplied before the previous batch has fully
cooled, the thermal aureole will grow (Annen, 2011).
For the case under consideration, that of a magmatic intrusion,
the post-intrusion viscoelastic response would be continued uplift
at an exponentially decreasing rate. The rate at which this occurs
will depend on the intrusion and eruption history of the magma
chamber.
4.2. Hydrothermal system
A new influx of magma would also be expected to perturb the
hydrothermal system. In a simple model, an increase in tempera-
ture at the base would increase the energy of hydrothermal system
above it. The total energy released can be calculated using:
E = ρb V b (cb  T b + L  X )
For a volume of basalt, V b of 0.03 km 3
at 1075 ◦ C cooling to
1030 ◦ C ( T b = 45 K) causing 20% crystallisation ( X = 0.2)
would release 1016 J. We assume the following constants: density
of basalt, ρb = 2700 kg m−3 ; specific heat capacity of basalt, cb =
1000 J kg−1 K−1 (Robertson, 1988), and latent heat, L = 4 × 105 J
(Bohrson and Spera, 2001).
Combining the equations for specific heat capacity with those
for thermal expansion, shows that the overall volume change,
 V w , is independent of the volume of water in the hydrothermal
system.
ρb (cb  T b + L X)
V w = α Vb
ρw c w
Thus for a basalt volume, V b = 0.03 km3 and a temperature
contrast,  T b of 45 ◦ C and thermal expansion coefficient of water,
α = 0.01 K−1 ; gives a volume increase of 0.024 km3 .
In the short term, the temperature change in a thin boundary
layer at the base could be considerably higher, driving a brief pe-
riod of enhanced convection or causing a phase transition to steam,
neither of which are addressed by these equations. More com-
plex hydrothermal responses, such as intrusion-triggered changes
in permeability leading to draining of the system are plausible, but
require a full model of the hydrothermal system, which is beyond
the scope of this study.
For both a viscoelastic and hydrothermal scenarios, the initial
response to intrusion would be additional uplift and hence these
mechanisms cannot explain the change from uplift to subsidence
observed at Okmok. In fact, since they act in the opposite direc-
tion, they are likely to reduce the magnitude of the subsidence
observed, making our previous estimates minimum values.
5. Discussion
This paper details thermal and petrological calculations for vol-
ume balances based on the geodetic record at Okmok volcano
but the concept may apply to other deforming volcanoes around
the world (Table 1). Like Okmok, geodetic measurements at the
well-studied calderas of Askja, Campi Flegrei and Sakurajima, cap-
ture pulses of uplift followed by subsidence that last for several
decades. Similarly, several volcanoes in the East African Rift show
complex patterns of uplift and subsidence; for example, Alutu vol-
cano shows two pulses of rapid uplift in 2004 and 2008 inter-
spersed with gradual subsidence (Biggs et al., 2011). However, for
this system, the lack of petrological characterisation means that
equivalent volume-balanced calculations cannot be performed.
At volcanoes such as Medicine Lake, Krafla and Cerro Blanco,
geodetic observations began during an extended period of subsi-
dence, and we infer an intrusion prior to the geodetic record. These
are caldera systems, where intrusions have been large enough to
result in long-lived subsidence signals and many show a slow de-
cay in rate. For example, the 10-km wide subsidence pattern at
Krafla caldera decreased in rate from 5 cm/yr in 1989–1992 to
0.3 cm/yr in 2000–2005 (Sturkell et al., 2008). Traditionally, these
subsidence signals have been attributed to drainage of magma (e.g.
Askja, de Zeeuw-van Dalfsen et al., 2012) but the calculations pre-
sented here suggest that volumetric changes associated with crys-
tallisation in a cooling magma body could produce deformation
patterns of the required magnitude and duration. This may also be
true for the smaller areas of persistent shallow subsidence com-
monly attributed to hydrothermal cooling e.g. Copahue (Velez et
al., 2011) or Kiska (Lu et al., 2002), although the presence of a
hydrothermal system may also effectively mask volume changes
associated with crystallisation and second boiling.
6. Conclusions
(1) This paper demonstrates that thermal models can be used to
link petrological processes to deformation patterns on geodetic
timescales. We examine candidate mechanisms for the 2005–2006
period of subsidence at Okmok volcano. Although the volume
change associated with crystallisation alone is of the right mag-
nitude, the timescale required is greater than the 9-year inter-
val between injection and onset of deflation (Fig. 3). However,
these timescales are compatible with a scenario where the defla-
tion signal is linked to the degassing of excess volatiles produced
by crystallisation of a water-saturated magma. The presence of
volatile-saturated magma at a depth corresponding to the mag-
matic reservoir of Okmok is confirmed by petrological evidences
(Finney et al., 2008). In this case, the timescale of subsidence
(2)
would be controlled by the mechanism of degassing. Alterna-
tively, re-melting of a partially crystalline andesite would generate
a volatile under-saturated reservoir capable of resorbing the ex-
cess volatiles without surface degassing. This study shows how
an appropriate chemical characterisation of eruptive products, in
combination with thermal modelling can be used to increase our
understanding of sub-volcanic processes from the interpretation of
geodetic signals. A similar approach would help understanding the
processes responsible for the subsidence observed in several vol-
canic systems.
Appendix A. computation of temperatures and melt fractions
Energy conservation implies that:
∂T ∂X
ρc + ρL = k∇ T (1)
∂t ∂t
where ρ is density, T is temperature, t is time, L is latent heat of
crystallisation, X is melt fraction, and k is conductivity, or
 L. Caricchi et al. / Earth and Planetary Science Letters 388 (2014) 166–174 173

∂H
= k∇ T (2) References
∂t
Annen, C., 2011. Implications of incremental emplacement of magma bodies for
with
magma differentiation, thermal aureole dimensions and plutonism–volcanism
relationships. Tectonophysics 500, 3–10.
H = ρ cT + ρ L X (3) Beard, J.S., Lofgren, G.E., 1989. Effect of water on the composition of partial melts of
greenstone and amphibolite. Science 244, 195–197.
The forward finite difference expression of (2) on a squared grid in Beget, J.E., Larsen, J.F., Neal, C.A., Nye, C.J., Schaefer, J.R., 2005. Preliminary volcano-
cylindrical coordinates is: hazard assessment for Okmok volcano, Umnak Island, Alaska. Alaska Division of
Geological & Geophysical Surveys Report of Investigation 2004-3, 32 pp.
p +dt p    Belien, I.B., Cashman, K.V., Rempel, A.W., 2010. Gas accumulation in particle-
H i, j − H i, j k  p p  dr rich suspensions and implications for bubble populations in crystal-rich
= T i −1 , j − T i, j r− magma. Earth Planet. Sci. Lett. 297, 133–140. http://dx.doi.org/10.1016/
dt r dr 2
  j.epsl.2010.06.014.
 p p  dr  p p  Biggs, J., Anthony, E.Y., Ebinger, C.J., 2009. Multiple inflation and deflation
+ T i +1 , j − T i , j r + + T i , j −1 − T i , j r events at Kenyan volcanoes. East African Rift. Geology 37 (11), 979–982.
2 http://dx.doi.org/10.1130/G30133A.1.

 p p 
Biggs, J., Mothes, P., Ruiz, M., Amelung, F., Dixon, T.H., Baker, S., Hong, S.-H.,
+ T i , j +1 − T i , j r (4) 2010. Stratovolcano growth by co-eruptive intrusion: The 2008 eruption of
Tungurahua Ecuador. Geophys. Res. Lett. 37, L21302. http://dx.doi.org/10.1029/
2010GL044942.
with subscripts i and j denoting the spatial position on the grid Biggs, J., Bastow, I.D., Keir, D., Lewi, E., 2011. Pulses of deformation reveal frequently
and superscript p denoting position in time. dt is the numerical recurring shallow magmatic activity beneath the Main Ethiopian Rift. Geochem.
timescale. r is the distance to symmetry axis and dr is the length Geophys. Geosyst. 12. http://dx.doi.org/10.1029/2011GC003662.
of a cell. To ensure numerical stability, dt < K /4 dr 2 , with the ther- Bohrson, W.A., Spera, F.J., 2001. Energy-constrained open-system magmatic pro-
cesses I: General model and energy-constrained assimilation and fractional crys-
mal diffusivity K = k/ρ c. tallization (EC-AFC) formulation. J. Petrol. 42 (5), 1019–1041.
To determine temperatures and melt fractions, we proceed as Caricchi, L., Burlini, L., Ulmer, P., Gerya, T., Vassalli, M., Papale, P., 2007. Non-
follows: Newtonian rheology of crystal-bearing magmas and implications for magma
With T s , the solidus temperature: ascent dynamics. Earth Planet. Sci. Lett. 264, 402–419. http://dx.doi.org/10.1016/
j.epsl.2007.09.032.
Caricchi, L., Pommier, A., Pistone, M., Castro, J., Burgisser, A., Perugini, D., 2011.
H S = ρ cT s (5) Strain-induced magma degassing: insights from simple-shear experiments on
H bubble bearing melts. Bull. Volcanol. 73 (9), 1245–1257. http://dx.doi.org/
T (H  H s) = (6) 10.1007/s00445-011-0471-2.
ρc de Zeeuw-van Dalfsen, E., Rymer, H., Williams-Jones, G., Sturkell, E., Sigmundsson,
F., 2006. Integration of micro-gravity and geodetic data to constrain shallow
and system mass changes at Krafla Volcano, N Iceland. Bull. Volcanol. 68, 420–431.
http://dx.doi.org/10.1007/s0445-005-0018-5.
X (H  H s) = 0 (7) de Zeeuw-van Dalfsen, E., Pedersen, R., Hooper, A., Sigmundsson, F., 2012.
Subsidence of Askja caldera 2000–2009: Modelling of deformation pro-
For H > H s , we use laboratory-determined temperatures and melt cesses at an extensional plate boundary, constrained by time series In-
fraction to calculate the corresponding H with Eq. (3) and we as- SAR analysis. J. Volcanol. Geotherm. Res. 213–214, 72–82. http://dx.doi.org/
10.1016/j.jvolgeores.2011.11.004.
sume that T – X–H relationship between laboratory data is linear,
Dragoni, M., Magnanensi, C., 1989. Displacement and stress produced by a pressur-
so that for any H calculated with Eq. (4), we can recover T and X ized, spherical magma chamber, surrounded by a viscoelastic shell. Phys. Earth
by linear interpolation. The Matlab code with H and the pressure Planet. Inter. 56 (3), 316–328.
P as input is as follows: Dvorak, J., Dzurisin, D., 1997. Volcano geodesy: The search for magma reservoirs and
the formation of eruptive vents. Rev. Geophys. 35, 343–384.
%“labTemp” are experimental temperatures (Wagner et al., 1995, Green,
Dzurisin, D., Poland, M.P., Bürgmann, R., 2002. Steady subsidence of Medicine Lake
%1973, Holloway and Burnham, 1972) volcano, northern California, revealed by repeated leveling surveys. J. Geophys.
%labTemp = [930 985 1000 1015 1035 1045 1050 1075]; Res. 107 (B12), 2372. http://dx.doi.org/10.1029/2001JB000893.
%“dT” is a correction for pressure of 75C per 2 kb (Beard and Lofgren, Euillades, P.A., Velez, M.L., Meljem, M., Cadelli, A., Martinez-Diaz, J.J., 2008. DInSAR
1989) SBAS data processing in the southern Andes: The Copahue-Agrio Caldera study
case. Paper presented at 4th Alexander von Humboldt International Conference.
dT = −(P−1)*75/2; Eur. Geosci. Union, Santiago.
labTemp = labTemp + dT; Fernandez, J., Tiampo, K.F., Rundle, J.B., 2001. Viscoelastic displacement and gravity
%“labMelt” is the set of experimental melt fractions corresponding to changes due to point magmatic intrusions in a gravitational layered solid earth.
labTemp %(Wagner et al., 1995; Green, 1973; Holloway and Burnham, Geophys. J. Int. 146, 155–170.
Finney, B., Turner, S., Hawkesworth, C., Larsen, J., Nye, C., George, R., Binde-
1972)
man, I., Eichelberger, J., 2008. Magmatic differentiation at an island-arc
labMelt = [0 0.53 0.56 0.63 0.83 0.89 0.96 1]; caldera: Okmok volcano, Aleutian Islands, Alaska. J. Petrol. 49 (5), 857–884.
%“ labH” is calculated for the experimental points http://dx.doi.org/10.1093/petrology/egn008.
labH = ro *c.* labTemp + ro * L. * labMelt; Fournier, T.J., 2008. Analysis and interpretation of volcano deformation in Alaska:
%Temperatures T and melt fractions are interpolated between experimen- studies from Okmok and Mt. Veniaminof volcanoes. PhD thesis. University of
Alaska, Faribanks.
tal %points for input H Fournier, T., Freymueller, J., Cervelli, P., 2009. Tracking magma volume recovery at
for i = 1:length(labH) − 1 Okmok volcano using GPS and an unscented Kalman filter. J. Geophys. Res. 114
H1 = labH(i); (B2). http://dx.doi.org/10.1029/2008JB005837.
H2 = labH(i + 1); Fournier, T.J., Pritchard, M.E., Riddick, S.N., 2010. Duration, magnitude, and fre-
quency of subaerial volcano deformation events: New results from Latin Amer-
T1 = labTemp(i);
ica using InSAR and a global synthesis. Geochem. Geophys. Geosyst. 11 (1).
T2 = labTemp(i + 1); http://dx.doi.org/10.1029/2009GC002558.
m1 = labMelt(i); Freymueller, J.T., Kaufman, A.M., 2010. Changes in the magma system during the
m2 = labMelt(i + 1); 2008 eruption of Okmok volcano, Alaska, based on GPS measurements. J. Geo-
index = logical(H > H1 & H <= H2); phys. Res. 115, B12415. http://dx.doi.org/10.1029/2010JB007716. 14 pp.
Ghiorso, M.S., 1985. Chemical mass-transfer in magmatic processes. 1. Thermody-
T(index) = T1 + ((T2 − T1)/(H2 − H1)).*(H(index) − H1); namic relations and numerical algorithms. Contrib. Mineral. Petrol. 90, 107–120.
melt(index) = m1 + ((m2 − m1)/(H2 − H1)).*(H(index) − H1); Ghiorso, M.S., Sack, R.O., 1995. Chemical mass-transfer in magmatic processes. 4.
end A revised and internally consistent thermodynamic model for the interpolation
174 L. Caricchi et al. / Earth and Planetary Science Letters 388 (2014) 166–174
and extrapolation of liquid–solid equilibria in magmatic systems at elevated-
temperatures and pressures. Contrib. Mineral. Petrol. 119, 197–212.
Giordano, D., Russell, J.K., Dingwell, D.B., 2008. Viscosity of magmatic liq-
uids: A model. Earth Planet. Sci. Lett. 271, 123–134. http://dx.doi.org/
10.1016/j.epsl.2008.03.038.
Gottsmann, J., Rymer, H., Berrino, G., 2006. Unrest at the Campi Flegrei caldera
(Italy): A critical evaluation of source parameters from geodetic data inversion.
J. Volcanol. Geotherm. Res. 150, 132–145.
Green, D.H., 1973. Experimental melting studies on a model upper mantle compo-
sition at high pressure under water-saturated and water-undersaturated condi-
tions. Tectonophysics 17, 285–297.
Holloway, J.R., Burnham, C.W., 1972. Melting relations of basalt with equilibrium
water pressure less than total pressure. J. Petrol. 13, 1–29.
Huppert, H.E., Sparks, R.S.J., 1988. The generation of granitic magmas by intrusion
of basalt into continental crust. J. Petrol. 29 (3), 599–624.
Kwoun, O.I., Lu, Z., Neal, C., Wicks Jr., C., 2006. Quiescent deformation of the Ani-
akchak Caldera, Alaska mapped by InSAR. Geology 34, 5–8. http://dx.doi.org/
10.1130/G22015.1.
Lange, R.A., 1994. The effect of H2 O, CO2 and F on the density and viscosity of
silicate melts. In: Volatiles in Magmas. In: Rev. Mineral., vol. 30, pp. 331–369.
Larsen, J.F., Śliwiński, M.G., Nye, C., Cameron, C., Schafer, J.R., 2013. The 2008 erup-
tion of Okmok volcano, Alaska: Petrological and geochemical constraints on the
subsurface magma plumbing system. J. Volcanol. Geotherm. Res. 264, 85–106.
Lu, Z., Dzurisin, D., 2010. Ground surface deformation patterns, magma sup-
ply, and magma storage at Okmok volcano, Alaska, from InSAR analysis: 2.
Coeruptive deflation, July–August 2008. J. Geophys. Res. 115. http://dx.doi.org/
10.1029/2009JB006970.
Lu, Z., Mann, D., Freymueller, J., 1998. Satellite radar interferometry measures
deformation at Okmok volcano. Trans. Am. Geophys. Union 79 (39), 461.
http://dx.doi.org/10.1029/98EO00348.
Lu, Z., Mann, D., Freymueller, J.T., Meyer, D.J., 2000. Synthetic aperture radar inter-
ferometry of Okmok volcano, Alaska: Radar observations. J. Geophys. Res. 105
(B5), 10791. http://dx.doi.org/10.1029/2000JB900034.
Lu, Z., Masterlark, Tim, Power, John, Dzurisin, Daniel, Wicks, C.J., 2002. Subsidence
at Kiska volcano, Western Aleutians, detected by satellite radar interferometry.
Geophys. Res. Lett. 29 (18), 2-1–2-4. http://dx.doi.org/10.1029/2002GL014948.
Lu, Z., Fielding, E., Patrick, M.R., Trautwein, C.M., 2003. Estimating lava volume by
precision combination of multiple baseline spaceborne and airborne interfero-
metric synthetic aperture radar: The 1997 eruption of Okmok volcano, Alaska.
IEEE Trans. Geosci. Remote Sens. 41 (6), 1428–1436.
Lu, Z., Masterlark, T., Dzurisin, D., 2005. Interferometric synthetic aperture
radar study of Okmok volcano, Alaska, 1992–2003: Magma supply dynam-
ics and postemplacement lava flow deformation. J. Geophys. Res. Solid Earth
(1978–2012) 110 (B2). http://dx.doi.org/10.1029/2004JB003148.
Lu, Z., Dzurisin, D., Biggs, J., Wicks Jr., C., McNutt, S., 2010. Ground surface de-
formation patterns, magma supply, and magma storage at Okmok volcano,
Alaska, from InSAR analysis: 1. Intereruption deformation, 1997–2008. J. Geo-
phys. Res. 115. http://dx.doi.org/10.1029/2009JB006969.
Lundgren, P., Usai, S., Sansosti, E., Lanari, R., Tesauro, M., Fornaro, G., Berardino, P.,
2001. Modeling surface deformation observed with synthetic aperture radar in-
terferometry at Campi Flegrei caldera. J. Geophys. Res. 106 (B9), 19355–19366.
Mann, D., Freymueller, J., 2003. Volcanic and tectonic deformation on Unimak Island
in the Aleutian Arc, Alaska. J. Geophys. Res. 108 (B2), 2108. http://dx.doi.org/
10.1029/2002JB001925.
Mann, D., Freymueller, J., Lu, Z., 2002. Deformation associated with the 1997 erup-
tion of Okmok volcano, Alaska. J. Geophys. Res. 107 (B4), 2072.
Masterlark, T., 2007. Magma intrusion and deformation predictions: Sensitivi-
ties to the Mogi assumptions. J. Geophys. Res. Solid Earth 112, B06419.
http://dx.doi.org/10.1029/2006JB004860.
Masterlark, T., Haney, M., Dickinson, H., Fournier, T., Searcy, C., 2010. Rheologic
and structural controls on the deformation of Okmok volcano, Alaska: FEMs,
InSAR, and ambient noise tomography. J. Geophys. Res. 115. http://dx.doi.org/
10.1029/2009JB006324.
Miller, T.P., Smith, R.L., 1975. Two caldera-forming eruptions on Umnak Island, east-
ern Aleutian Islands. U.S. Geol. Surv. Circ. 733, 45.
Miller, T.P., McGimsey, R.G., Richter, D.H., Riehle, J.R., Nye, C.J., Yount, M.E., Dumoulin,
J.A., 1998. Catalog of the historically active volcanoes of Alaska. U.S. Geol. Surv.
Open-File Rep. 98-582, 104 pp.
Miyagi, Y., Freymueller, J.T., Kimata, F., Sato, T., Mann, D., 2004. Surface deformation
caused by shallow magmatic activity at Okmok volcano, Alaska, detected by GPS
campaigns 2000–2002. Earth Planets Space 56 (12), E29–E32.
Mogi, K., 1958. Relations between the eruptions of various volcanoes and the de-
formations of the ground surfaces around them. Bull. Earthq. Res. Inst. Univ.
Tokyo 36, 99–134.
Moore, G., Vennemann, T., Carmichael, I., 1998. An empirical model for the solubility
of H2 O in magmas to 3 kilobars. Am. Mineral. 83, 36–42.
Newman, S., Lowenstern, J.B., 2002. VOLATILECALC: a silicate melt-H2 O–CO2 solu-
tion model written in Visual Basic for excel. Comput. Geosci. 28 (5), 597–604.
Pagli, C., Sigmundsson, F., Árnadóttir, T., Einarsson, P., Sturkell, E., 2006. Deforma-
tion of the Askja volcanic system: constraints on the deformation source from
combined inversion of satellite radar interferograms and GPS measurements.
J. Volcanol. Geotherm. Res. 152, 97–108.
Pitzer, K.S., Sterner, S.M., 1994. Equations of state valid continuously from zero
to extreme pressures for H2 O and CO2 . J. Chem. Phys. 101 (4), 3111–3116.
http://dx.doi.org/10.1063/1.467624.
Poland, M., Burgmann, R., Dzunisin, D., Lisowski, M., Masterlark, T., Owen, S.,
Fink, J., 2006. Constraints on the mechanism of long-term, steady subsi-
dence at Medicine Lake volcano, northern California, from GPS, leveling,
and InSAR. J. Volcanol. Geotherm. Res. 150, 55–78. http://dx.doi.org/10.1016/
j.jvolgeores.2005.07.007.
Pritchard, M.E., Simons, M., 2004. An InSAR-based survey of volcanic deforma-
tion in the southern Andes. Geophys. Res. Lett. 31, L15610. http://dx.doi.org/
10.1029/2004GL020545.
Robertson, E.C., 1988. Thermal properties of rocks. U.S. Geol. Surv. Open-File Rep.
88-441, 106 pp.
Sahagian, D.L., Proussevitch, A.A., 1996. Thermal effects of magma degassing. J. Vol-
canol. Geotherm. Res. 74 (1–2), 19–38.
Segall, P., 2010. Earthquake and Volcano Deformation. Princeton University Press,
ISBN 9780691133027.
Sigmundsson, F., Vadon, H., Massonet, D., 1997. Readjustment of the Krafla spreading
segment to crustal rifting measured by satellite radar interferometry. Geophys.
Res. Lett. 24, 1843–1846.
Sparks, R.S.J., Biggs, J., Neuberg, J.W., 2012. Monitoring Volcanoes. Science 335
(6074), 1310–1311. http://dx.doi.org/10.1126/science.1219485.
Sturkell, E., Sigmundsson, F., Slunga, R., 2006. 1983–2003 decaying rate of deflation
at Askja caldera: pressure decrease in an extensive magma plumbing system at
a spreading plate boundary. Bull. Volcanol. 68, 727–735. 99–134.
Sturkell, E., Sigmundsson, F., Geirsson, H., Ólafsson, H., Theodórsson, T., 2008. Multi-
ple volcano deformation sources in a post-rifting period: 1989–2005 behaviour
of Krafla, Iceland constrained by levelling, tilt and GPS observations. J. Volcanol.
Geotherm. Res. 177 (2), 405–417.
Velez, M.L., Euillades, P., Caselli, A., Blanco, M., Díaz, J.M., 2011. Deformation of Co-
pahue volcano: Inversion of InSAR data using a genetic algorithm. J. Volcanol.
Geotherm. Res. 202 (1), 117–126.
Wagner, T.P., DonnellyNolan, J.M., Grove, T.L., 1995. Evidence of hydrous differen-
tiation and crystal accumulation in the low-MgO, high-Al2 O3 lake basalt from
medicine lake volcano, California. Contrib. Mineral. Petrol. 121 (2), 201–216.
Wicks, C.W., 2002. Magmatic activity beneath the quiescent Three Sisters volcanic
center, central Oregon Cascade Range, USA. Geophys. Res. Lett. 29 (7), 1122.
http://dx.doi.org/10.1029/2001GL014205.
Worster, M.G., Huppert, H.E., Sparks, R.S.J., 1990. Convection and crystallization in
magma cooled from above. Earth Planet. Sci. Lett. 101, 78–89.
Yokoyama, I., 1986. Crustal deformation caused by the 1914 eruption of Sakura-
jima volcano, Japan and its secular changes. J. Volcanol. Geotherm. Res. 30,
283–302.
Zimmer, M.M., Plank, T., Hauri, E.H., Yogodzinski, G.M., Stelling, P., Larsen, J., Singer,
B., Jicha, B., Mandeville, C., Nye, C.J., 2010. The role of water in generating the
calc-alkaline trend: new volatile data for aleutian magmas and a new tholeiitic
index. J. Petrol. 51 (12), 2411–2444. http://dx.doi.org/10.1093/petrology/egq062.