Journal of Volcanology and Geothermal Research 113 (2002) 379^389

[Link]/locate/jvolgeores

Detecting volcanic eruption precursors: a new method using

gravity and deformation measurements
Glyn Williams-Jones
, Hazel Rymer
Volcano Dynamics Group, Department of Earth Sciences, The Open University, Walton Hall, Milton Keynes, UK
Received 21 January 2000; accepted 12 June 2000

Abstract

One of the fundamental questions in modern volcanology is the manner in which a volcanic eruption is triggered;
the intrusion of fresh magma into a reservoir is thought to be a key component. The amount by which previously
ponded reservoir magma interacts with a newly intruded magma will determine the nature and rate of eruption as well
as the chemistry of erupted lavas and shallow dykes. The physics of this interaction can be investigated through a
conventional monitoring procedure that incorporates the simple and much used Mogi model relating ground
deformation (most simply represented by vh) to changes in volume of a magma reservoir. Gravity changes (vg)
combined with ground deformation provide information on magma reservoir mass changes. Our models predict how,
during inflation, the observed vg/vh gradient will evolve as a volcano develops from a state of dormancy through
unrest into a state of explosive activity. Calderas in a state of unrest and large composite volcanoes are the targets for
the methods proposed here and are exemplified by Campi Flegrei, Rabaul, Krafla, and Long Valley. We show here
how the simultaneous measurement of deformation and gravity at only a few key stations can identify important
precursory processes within a magma reservoir prior to the onset of more conventional eruption precursors. 3 2002
Elsevier Science B.V. All rights reserved.

Keywords: volcano; magma physics; micro-gravity; deformation; caldera; eruption triggers

1. Introduction as the chemistry of erupted lavas and shallow

dykes. To understand the process therefore re-
Explosive volcanic eruptions are often initiated quires information on the physical and chemical
by magma intrusion into a pre-existing magma processes occurring at depth within the magma
reservoir (cf. Sparks et al., 1977; Pallister et al., reservoir. The majority of geophysical techniques,
1992). The degree of interaction between the in- however, focus on shallow processes within the
truding magma and the pre-existing reservoir will volcanic edi¢ce. Micro-gravity monitoring, for ex-
determine the nature and rate of eruption as well ample, has been used at numerous active volca-
noes in order to identify the relatively shallow
processes occurring within the feeder conduit (cf.
* Corresponding author. Fax: +44-1908-655-151.
Eggers and Chavez, 1979; Johnson et al., 1980;
E-mail addresses: [Link]-jones@[Link] Yokoyama, 1989; Rymer and Brown, 1989; Ber-
(G. Williams-Jones), [Link]@[Link] (H. Rymer). rino et al., 1992; Rymer et al., 1998b). The con-

0377-0273 / 02 / $ ^ see front matter 3 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 7 - 0 2 7 3 ( 0 1 ) 0 0 2 7 2 - 4

VOLGEO 2381 6-5-02

380 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389

ventional micro-gravity survey involves repeated

measurements of gravity change (vg) and defor-
mation (vh) at a network of stations in and
around the active crater or caldera (Rymer,
1994; Rymer et al., 1998b). The recent deaths of
some well-known volcanologists (cf. Baxter and
Gresham, 1997; Fujii and Nakada, 1999) empha-
sise the need to monitor volcanoes at a safe dis-
tance from the active centre. This, however, can
often lead to reduced data quality. Here we dis-
cuss a new approach, ¢rst proposed by Rymer
and Williams-Jones (2000), which actually bene¢ts Fig. 1. A Mogi-type magma reservoir modelled in an elastic
from gravity and deformation measurements half space. The largest ground deformation changes (vh) are
observed in zone A. Measurable e¡ects are also seen at some
made up to 6 km from the centre of activity. distance (x) from the centre of activity, such as at B. Using
Physical processes deep within the magma reser- the measured or estimated values of the elastic properties of
voir, which can occur months to years before con- the country rock and the amount of ground deformation al-
ventional precursors, can now be interpreted from lows for the estimation of the change in volume of the sub-
vg/vh gradients measured at a safe distance from surface magma reservoir (vVr ). Small changes in gravity (vg)
may also be observed at these points (A and B) and when
the central volcanic edi¢ce. This is exempli¢ed by combined with ground deformation data, one can quantify
important examples from the calderas of Campi the change in sub-surface magma mass (vMm ). Modi¢ed
Flegrei, Rabaul, Kra£a, and Long Valley, exam- after Rymer and Williams-Jones (2000).
ined below.

2. Gravity and height correlations over the area of deformation. Measured or esti-
mated values for the elastic properties of the
Conventional gravity and deformation data are country rock along with the amount of ground
usually modelled in terms of either a sphere or a deformation can then be used to estimate the
Bouguer slab. Both are mathematically simple to change in volume of the sub-surface magma res-
model. To a ¢rst approximation, a magma reser- ervoir (vVr ) (McKee et al., 1989; Vasco et al.,
voir can usually be most realistically modelled as 1990; Dvorak and Mastrolorenzo, 1991; Wicks
a spherical body. If the depth of the magma res- et al., 1998). However, the change in volume
ervoir is large compared to its radius, the gravita- alone provides little information on the actual
tional e¡ect of the reservoir can be considered to processes taking place within the magma reser-
be that of a point source (Rymer, 1994). The voir. If, in addition, small variations in the accel-
Mogi model (Mogi, 1958) quanti¢es the deforma- eration due to gravity are monitored, the changes
tion at the surface caused by dilation of a point in sub-surface magma mass (vMm ) can be quan-
source in a homogeneous elastic half space. The ti¢ed (Berrino et al., 1992). Simultaneous gravity
deformation will depend on the amount of dila- and deformation measurements can therefore pro-
tion and the elastic properties of the medium (Fig. vide an estimate of vMm and vVr in order that
1). Using ground deformation measurement tech- changes in the average density of the magma res-
niques such as the Global Positioning System ervoir may be deduced.
(GPS), Synthetic Aperture Radar di¡erential in- Changes in gravity and elevation are normally
terferometry (DInSAR), altimetry or levelling (cf. inversely correlated. The amount by which gravity
Smith et al., 1989; Murray et al., 1995; Dixon et varies with elevation is given by the free air gra-
al., 1997; Avallone et al., 1999), the change in dient (FAG). If there is no change in the density
edi¢ce volume (vVe ) of the volcano can be esti- (b) of the magma reservoir, the amount by which
mated by integrating the observed height changes gravity varies with elevation can be described by

VOLGEO 2381 6-5-02

 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389 381

the Bouguer-corrected free air gradient (BCFAG) therein). The actual gradient of the BCFAG,
of a spherical body. The BCFAGspherical can be which will vary depending on the density assumed
calculated using the following relationship in for the surrounding rock, ranges from 3253 to
which the result is in units of WGal m31 : 3230 WGal m31 for magma densities of 2000 to
2800 kg m33 , respectively, assuming the theoret-
4ZGU108 ical FAG.
BCFAGspherical ¼ FAG þ b ð1Þ
3 If at all possible, the actual FAG should always
be measured at each station during a micro-grav-
where G is the universal gravitational constant ity/deformation survey. This can easily be accom-
(6.67U10311 N m2 kg32 ) and b is in kg m33 . plished in the ¢eld by making measurements ¢rst
The theoretical value for the FAG is 3308.6 at the surface and then some distance o¡ the
WGal m31 . Terrain e¡ects and Bouguer anomalies ground (e.g. V1 m using a levelling tripod), and
can cause this value to di¡er by up to 40% from then dividing the di¡erence of the two sets of
its theoretical value (Rymer, 1994 and references measurements (vg) by the di¡erence in elevation

Fig. 2. Changes in gravity (vg) and elevation (vh) can be plotted in terms of vg/vh gradients. When the gradient di¡ers from the
FAG, the variations are interpreted in terms of mass changes. Changes in density are shown by deviations from the BCFAG
and FAG (3FAG+BCFAG). The theoretical FAG is commonly taken as 3308.6 WGal m31 but may vary by V40% depending
on the local terrain and Bouguer anomaly. Assuming a Mogi point-source model and a theoretical FAG, the BCFAG may vary
between 3253 and 3230 WGal m31 for densities between 2000 and 2800 kg m31 . Modi¢ed after Brown and Rymer (1991).

VOLGEO 2381 6-5-02

382 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389

(vh) (Berrino et al., 1984; Yokoyama, 1989; Ry- Fig. 2, where during in£ation there are mass in-
mer, 1994). creases and yet density decreases. Recognition of
Gravity/elevation gradients are most easily vi- this is fundamental to understanding the physics
sualised using a vg/vh diagram (Fig. 2). vg/vh of magma reservoir processes and the detection of
gradients that deviate from the FAG are inter- eruption precursors (Fig. 3).
preted in terms of sub-surface mass changes. In order to investigate relatively shallow sub-
Data plotting above the FAG re£ect mass in- surface processes, micro-gravity and ground de-
creases while data plotting below it re£ect mass formation surveys are commonly carried out in
decreases. Similarly, deviations from the BCFAG the region of maximum uplift, central to the vol-
are interpreted as sub-surface density changes canically active region (zone A, Fig. 1). However,
(Brown and Rymer, 1991; Berrino et al., 1992; sensitive instrumentation and techniques (e.g. La-
Rymer et al., 1995). During periods of in£ation, coste and Romberg or Scintrex gravity meters
data plotting above the BCFAG re£ect density coupled with precise levelling or di¡erential
increases, whereas data plotting below the GPS) can detect small o¡ axis variations
BCFAG re£ect density decreases. There is an im- (vg s 20 WGal, vh s 1 cm) at signi¢cant lateral
portant and intriguing region between the FAG distances (a few kilometres) from the active region
and the BCFAG, in the lower right quadrant of (zone B, Fig. 1). Shallow volcanic events within
the edi¢ce (such as vesiculation and fracturing ;
(Rymer et al., 1998b), may be responsible for ele-
vated ‘noise’ in the gravity data in region A.
However, by making measurements away from
the active centre, the signal-to-noise ratio of the
data is signi¢cantly increased. Another potential
source of uncertainty, the variation in ground-
water levels, can be minimised in several ways;
the seasonal variation may be reduced by making
measurements at approximately the same time
every year (Arnet et al., 1997) while the e¡ects
of water table £uctuations can be reduced by
measuring at stations located on crystalline bed-
rock (Jachens and Roberts, 1985). The depth and
distance at which it is possible to detect subtle
changes within a Mogi-type magma reservoir
can be calculated by solving for the total observed
gravity change, vg (WGal), in the following rela-
tionship (Dzurisin et al., 1980; Johnson, 1987;
Eggers, 1987):
Fig. 3. During periods of in£ation, increasing elevation (vh)  
is accompanied by decreasing gravity (negative vg), de¢ned vM m WGWz
vg ¼ W108 ð2Þ
by the BCFAG. Region 1 represents anomalously large grav- ðx2 þ z2 Þ3=2
ity increases and may be interpreted in terms of magma in-
trusion into a magma reservoir resulting in an increase in the
average density of the reservoir. Region 2 re£ects an overall where vMm is the change in sub-surface magma
density decrease and mass increase that may be interpreted mass, G is the universal gravitational constant
in terms of gas build up within the magma reservoir, an im- (6.672U1011 N m2 kg32 ), x is the surface distance
portant trigger mechanism for explosive eruptions. Data fall- (m) out from the centre of the Mogi source, and z
ing below the FAG and close to the vg line may re£ect shal-
low processes such as magma and gas £uctuations within the
is the depth (m) to the Mogi point source (Fig. 1).
feeder conduit. Modi¢ed after Rymer and Williams-Jones Thus, for sub-surface magma mass changes of
(2000). 1011 kg (e.g. Rabaul, Campi Flegrei; McKee et

VOLGEO 2381 6-5-02

 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389 383

There will, however, be an overall mass increase

in the system, measurable at the surface as a grav-
ity increase along with limited ground deforma-
tion. This will result in a vg/vh gradient that will
plot between the BCFAG and the vh axis (region
1) or along the BCFAG (Fig. 3) implying an over-
all density increase (region 1) or no density
change (BCFAG) within the reservoir. Increasing
density within a magma reservoir is often inter-
preted in terms of magma devesiculation and/or
void ¢lling (cf. Brown et al., 1991; Rymer, 1994).
The dense intruded magma may be unable to rise
and, in a closed system, the reservoir would even-
tually stagnate. However, if the magma pressure
Fig. 4. The depth and distance at which it is possible to exceeds the local lithostatic pressure, dykes may
measure gravity changes (vg) within a Mogi source can be be injected into the country rock and can in some
determined for given changes in sub-surface magma mass cases eventually feed lava eruptions. The essential
(Eq. 2). For mass changes on the order of 1011 kg (similar point here is that there is very limited interaction
to those of Campi Flegrei and Rabaul calderas), small grav-
ity changes ( s 20 WGal) can be measured at horizontal dis-
between the new intruding magma and the mag-
tances of up to V4 km. For larger changes, e.g., 1012 kg, ma already in the reservoir. The mass increase is
variations in gravity could be measured at up to 10 km from associated either with an overall density increase
the centre of activity. Contours are 20 WGal. Depth (x) and in the reservoir or in no density change. Either
distance (z) from the Mogi source are in metres (see Fig. 1). way, an eruption, and especially a large one, is
unlikely.
al., 1989; Berrino et al., 1992, 1997) at a depth At the other end of the spectrum, model 2 as-
between 2 and 6 km, gravity variations ( s 20 sumes a high Reynolds number (i.e. turbulent
WGal) can be measured at a horizontal distance £ow; Sparks et al., 1980) which will interact vig-
of up to V4 km. For a mass change of the order orously with the reservoir magma, heating it,
of 1012 kg in a reservoir between 2 and 7 km deep, causing convection, and therefore cooling the in-
gravity changes are detectable at up to V10 km truding magma (Fig. 5b; Eichelberger, 1980). This
distance (Fig. 4). cooling can then lead to oversaturation and vesic-
ulation (Huppert et al., 1982; Tait et al., 1989;
Pallister et al., 1992), while superheating and de-
3. Modelling a magma reservoir compression of the convecting reservoir magma
may also cause volatile supersaturation and exso-
Although the Mogi model represents a very lution (Sparks et al., 1977). Consequent gravity
simple magma reservoir, many published geo- and deformation variations would plot as a vg/
physical data are consistent with this type of vh gradient between the FAG and BCFAG (re-
source (Berrino et al., 1992; Arnet et al., 1997; gion 2, Fig. 3) due to a net sub-surface mass in-
Avallone et al., 1999). Here we present two end crease and density decrease. Signi¢cantly, the in-
member models that illustrate how, by observing teraction between the intruding magma and the
density changes within the magma reservoir, the older ponded magma is much more pervasive
hazard potential of a volcanic system can be de- than in model 1. Magma mixing and mingling
termined. will cause signi¢cant bubble formation and the
An intruding magma with a relatively low Rey- resulting increase in gas pressure within the reser-
nolds number (i.e. laminar £ow ; Sparks et al., voir is an essential pre-requisite to explosive erup-
1980) will have little or no interaction with the tive activity (Eichelberger, 1980; Woods and
surrounding magma reservoir (model 1, Fig. 5a). Koyaguchi, 1994).

VOLGEO 2381 6-5-02

384 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389

Fig. 5. (a) Model 1 depicts the intrusion of a low Reynolds number magma (i.e., laminar £ow) into a magma reservoir. Impor-
tantly, there is little or no interaction with the surrounding magma. The observed vg/vh gradient will fall into region 1 of Fig. 3.
(b) Model 2 represents the other end member. Assuming an identical magma mass in£ux as model 1 but higher Reynolds number
(i.e., turbulent £ow), the intruding magma will interact vigorously with the surrounding magma, resulting in heating, convection,
vesiculation and expansion of the reservoir. The observed gradient will fall into region 2 of Fig. 3. Modi¢ed after Rymer and
Williams-Jones (2000).

These models suggest that the observed vg/vh only limited heating, magma convection within
gradient will evolve as a volcano develops from a the reservoir will follow; this has recently been
state of dormancy through unrest into explosive linked with pulses of activity at the surface (Ka-
eruptive activity. The application of these models zahaya et al., 1994; Stevenson and Blake, 1998).
to hazard warning is as follows. As the gradient However, with excess heating, as gas pressure in-
migrates from region 1 towards the BCFAG, the creases, the likelihood of an explosive eruption
density of the magma reservoir decreases (Fig. 3) will increase (Woods and Koyaguchi, 1994).
as a result of reduced crystallisation, increased
buoyant melt and vesicle content. On the
BCFAG, the magma reservoir volume increases 4. Application of the models
but its density remains constant. In either scenar-
io, eruption hazard is minimal. Once the mea- Calderas in a state of unrest and large compo-
sured vg/vh gradient crosses the BCFAG, the site volcanoes are the most appropriate targets for
density of the magma reservoir has decreased be- the models proposed here. In an open volcanic
low the previous average value of the reservoir system (i.e. one that displays persistent surface
and the system becomes unstable. It is at this manifestations such as a stable lava lake, frequent
stage that the magma will be able to rise and explosive eruptions, etc.) gravity and deformation
presents an increased eruption hazard. If there is measurements made in region A (Fig. 1) will typ-

VOLGEO 2381 6-5-02

 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389 385

ically be dominated by shallow processes. Thus, in er 4400 yr BP Agnano-Monte Spina eruption

order to study processes within such a Mogi-type (Barberi and Carapezza, 1996). Due to the exten-
source, measurements must be made in region B, sive history of caldera unrest at Campi Flegrei
o¡ the central axis of activity. A closed volcanic and the increasing population of the Naples re-
system on the other hand, generally has little or gion, the need to distinguish between magma
no surface activity allowing for vg/vh gradient chamber processes (models 1 and 2, Fig. 5) is of
measurements to be made safely in both regions critical importance to hazard mitigation.
A and B. Some examples of these systems are
presented below. 4.2. Rabaul, Papua New Guinea

4.1. Campi Flegrei, Italy Another excellent example of a closed system is

the Rabaul caldera (New Britain Island, Papua
The Campi Flegrei caldera (Naples, Italy) is a New Guinea) which has been in a state of unrest
prime example of a persistently active closed sys- since at least the late 1800s. The caldera (14U9
tem and a caldera in a state of unrest. Formed km) lies within a shallow ignimbrite volcano and
V35 000 years ago with the explosion of the was modi¢ed by an earlier explosive episode 1400
Campanian Ignimbrite, the 12-km-diameter cal- years ago (Walker et al., 1981). In a more recent
dera last erupted in 1538. More recently, however, crisis (1973^1985), there was an overall uplift of
in the 1970s and early 1980s, the caldera was V100 mm yr31 with a maximum in£ation of
characterised by a bradyseismic crisis (an ex- 1.8 m and gravity change of up to 3410 WGal
tended period of very slow vertical instability of (McKee et al., 1989). Although no FAG was mea-
the crust (Jackson, 1997) and a maximum uplift sured, given the amplitude and wavelength of a
and gravity change of 1.616 m and 3331 WGal, static Bouguer anomaly, the FAG has been esti-
respectively. The measured FAG was 3290 P 5 mated at 3300 WGal (Berrino et al., 1992). As-
WGal m31 and assuming a point source density suming a point source density of 2500 kg m33 ,
of 2500 km m33 , the BCFAG was calculated to the BCFAG was then calculated to be 3239
be 3220 WGal m31 (Berrino et al., 1992). The vg/ WGal m31 (Berrino et al., 1992). This episode
vh gradient during in£ation was 3213 P 6 WGal was believed to have been caused by an intrusion
m31 and fell into region 1 (Fig. 6a) and as our of basic magma into a pre-existing reservoir that
model predicts, there was no eruption. In fact, resulted in an overall sub-surface mass increase of
between February 1981 and March 1983, the gra- 108 kg (McKee et al., 1989; Rymer, 1994). As at
dient actually evolved from the BCFAG towards Campi Flegrei, the vg/vh gradient (3216 P 4 WGal
the horizontal axis, well into region 1 (Fig. 3; m31 ) measured at Rabaul fell into region 1 (Fig.
Berrino, 1994). This episode was interpreted in 6b) ; there was no eruption observed during this
terms of intrusion into a magma reservoir with period. If the gradient were to have progressed
a resulting overall sub-surface mass increase of below the BCFAG of 3239 WGal m31 (into re-
V2U1011 kg (Berrino et al., 1984). Mobilisation gion 2), our model predicts that an explosive
of the hydrothermal system has also been pro- eruption would have followed.
posed as a possible cause of the observed uplift The most recent period of activity (1992^1999)
(Bonafede and Mazzanti, 1998). which has been characterised by in£ation and seis-
Our model predicts that during in£ation, once mic swarms in fact developed into an extensive
the vg/vh gradient steepens beyond the BCFAG series of explosive and e¡usive eruptions. Petro-
(3220 WGal m31 in this case) into region 2, an logical analyses of material erupted between 1997
explosive eruption is likely to follow. A modern and 1998 suggest that, as with the 1973^1985 cri-
explosive eruption of the same magnitude as the sis, current activity was initiated by the intrusion
1538 Pozzuoli eruption would a¡ect an estimated of new basaltic magma into the shallow dacitic
80 000, while 200 000 people would be at risk if reservoir (Bulletin of the Global Volcanism Net-
the eruption was of similar magnitude to the larg- work, 1992^1999). Unfortunately, no gravity data

VOLGEO 2381 6-5-02

386 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389

Fig. 6. The vg/vh gradients observed during periods of in£ation at the calderas of Campi Flegrei (A), Rabaul (B), Kra£a (C),
and Long Valley (D). The FAG and BCFAG gradients are measured or estimated/calculated values. vh is in m and vg in WGal.
Grey shading represents estimated error for the measured gradients. Modi¢ed after Berrino et al. (1992).

are available for this period and thus it is not 4.3. Kra£a, Iceland
known whether the vg/vh gradient had actually
progressed into region 2 prior to the onset of The Kra£a caldera (northeast Iceland) is part
eruptive activity (Fig. 6b). Although the town of of the large Kra£a central volcano that straddles
Rabaul is not as densely populated as Naples, a 100-km-long N^S-trending ¢ssure swarm. The
the activity at this centre poses a threat to the 8U10-km caldera is believed to have formed
livelihood of tens of thousands of people. The V0.1 Myr ago during the last interglacial period
population is well informed about volcanic activ- with the explosive eruption of rhyolitic and dacitic
ity and more than 50 000 were successfully and rocks (Saemundsson, 1978). Kra£a was last active
safely evacuated prior to the eruption in 1994 during the 1975^1984 rifting episode that was
(Bulletin of the Global Volcanism Network, characterised by lava extrusion, dyke emplace-
1992^1999). ment and steady in£ation interrupted by rapid

VOLGEO 2381 6-5-02

 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389 387

subsidence within the caldera. The in£ationary (Dzurisin et al., 1990; Battaglia et al., 1999).
periods are interpreted in terms of pressure in- This episode has most recently been interpreted
crease within a shallow magma reservoir followed in terms of a mass increase (3.8U1011 to
by rupture of the reservoir walls with rifting and 17.8U1011 kg) due to the intrusion of basaltic
dyke formation leading to rapid subsidence magma into a rhyolitic reservoir beneath the re-
(Tryggvason, 1984). The January^June 1978 in£a- surgent dome (Battaglia et al., 1999). Assuming
tionary period was characterised by gravity and the theoretical FAG and a density of 2700 kg
height variations which followed a gradient of m33 , the calculated BCFAG is 3233 WGal m31
3250 P 20 WGal m31 (Johnson et al., 1980). A and as for some of the previous examples, the
FAG and a BCFAG for the caldera were calcu- measured vg/vh gradient falls above the BCFAG
lated at 3308.6 and 3200 WGal m31 , respectively into region 1. Our model predicts that the net
(Rymer et al., 1998a). This is the only published density of the reservoir will increase, thus making
example, where during in£ation, the observed an eruption unlikely (Fig. 6d).
vg/vh gradient fell into region 2 (Fig. 6c). Our
model suggests that magma reservoir mass in-
crease but density decrease is likely to lead to 5. Conclusions
an eruption. In this basaltic extensional tectonic
regime, ¢re fountaining occurred, feeding lava Large composite volcanoes and calderas in a
£ows and there was also considerable dyke injec- state of unrest are appropriate targets at which
tion. Post-eruptive activity has been characterised to consider the models presented here, where
by continued de£ation (30 mm yr31 ; Sigmunds- they are characterised by a spherical Mogi-type
son et al., 1997) and a net gravity decrease caused source. Conventional micro-gravity surveys in-
mainly by the drainage of 4U1010 kg of magma volve the measurement of a large network of sta-
(Rymer et al., 1998a). The hazard to the popula- tions, often in proximity to centres of volcanic
tion is extremely small at this site. However Lake activity, and inevitably focus on relatively shallow
Myvatn, just a few kilometres away, is a very processes. Here, we require simultaneous mea-
popular tourist centre and Kra£a supports an in- surements of deformation and gravity at only a
dustrial geothermal power plant so that eruptive few key stations o¡ axis of the centre of activity.
activity a¡ects thousands of people directly. Clearly the number of stations required would
depend on the individual circumstances and avail-
4.4. Long Valley, USA able instrumentation, but at 1 to 5, will be con-
siderably less than required for conventional mi-
Long Valley caldera (California, USA), which cro-gravity surveys. The advent of continuous
is located on the eastern edge of the Sierra Neva- gravity and deformation monitoring (cf. Berrino
da tectonic block, formed 730 kyr ago during the et al., 1997; Dixon et al., 1997) will allow for a
Bishop Tu¡ eruption. This was followed shortly dramatically increased temporal resolution by this
by the formation of a resurgent rhyolite dome method. The spatial resolution is not compro-
(Bailey et al., 1976). A negative Bouguer anomaly, mised badly, since the sources of interest in the
with a steep gradient s 20 mGal in some places, proposed method are deeper than for convention-
is concentric with the caldera and re£ects the lat- al micro-gravity surveys. Very small changes in
eral transition from low-density caldera material mass or density of a magma reservoir at depths
to the high-density crystalline rocks of the Sierra of 2^7 km can be made at a safe distance (up to
Nevadas (Kane et al., 1976; Jachens and Roberts, 10 km) from the active centre, greatly reducing
1985). During the 1982^1998 in£ationary period, the risk to scientists. By quantifying the magni-
centred on the resurgent dome, a gravity decrease tude and rate of these changes, eruption precur-
of up to 107 P 6 WGal and maximum residual sors within the reservoir may be detected well
in£ation of 0.42 P 0.05 m occurred, giving an before the magma begins to rise towards the sur-
average vg/vh gradient of 3215 P 11 WGal m31 face, o¡ering signi¢cantly more time for hazard

VOLGEO 2381 6-5-02

388 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389

mitigation and evacuation. Determining the na- Berrino, G., Rymer, H., Brown, G.C., Corado, C., 1992. Grav-
ity-height correlations for unrest at calderas. J. Volcanol.
ture of these processes is also critical for evaluat-
Geotherm. Res. 53, 11^26.
ing the potential magnitude of the hazard. Some Berrino, G., Corrado, G., Magliulo, R., Umberto, R., 1997.
138 calderas have shown a degree of unrest in Continuous record of the gravity changes at Mt. Vesuvius.
historical times (Newhall and Dzurisin, 1988) In: Boschi, E. (Ed.), Special issue dedicated to Professor
and about half the well-documented cases resulted Michele Caputo on his seventieth birthday. Ann. Geo¢s.
in eruption. At silicic calderas, only 15% of the 40, 1019^1028.
Bonafede, M., Mazzanti, M., 1998. Modelling gravity varia-
cases of unrest resulted in eruption, but increasing tions consistent with ground deformation in the Campi Fle-
population and vulnerability to explosive erup- grei caldera (Italy). J. Volcanol. Geotherm. Res. 81, 137^
tions means that a better understanding of caldera 157.
unrest is an essential goal for the new millennium. Brown, G.C., Rymer, H., 1991. Microgravity monitoring at
active volcanoes: a review of theory and practice. Cahier
Cent. Eur. Ge¤odyn. Se¤ismol. 4, 279^304.
Brown, G.C., Rymer, H., Stevenson, D., 1991. Volcano mon-
Acknowledgements itoring by microgravity and energy budget analysis. J. Geol.
Soc. London 148, 585^593.
This research was supported by the Royal Bulletin of the Global Volcanism Network, 1992^1999. Ra-
baul. Smithsonian Institution, 17:05^24:10 ([Link]
Society, The Open University Research Develop-
[Link]/gvp/volcano/region05/newbrit/rabaul/[Link]).
ment Fund and NERC. We thank J. B. Murray, Dixon, T.H., Mao, A., Bursik, M., He£in, M., Langbein, J.,
C.A. Locke, J. Cassidy and D. Rothery for many Stein, R., Webb, F., 1997. Continuous monitoring of surface
useful discussions about this work. This work was deformation at Long Valley Caldera, California, with GPS.
greatly improved by the comments of an anon- J. Geophys. Res. 102, 12017^12034.
ymous reviewer. Dvorak, J., Mastrolorenzo, G., 1991. The mechanisms of re-
cent vertical crustal movements in Campi Flegrei caldera.
Spec. Pap. Geol. Soc. Am. 263, 47.
Dzurisin, D., Anderson, L.A., Eaton, G.P., Koyanagi, R.Y.,
References Lipman, P.W., Lockwood, J.P., Okamura, R.T., Puniwai,
G.S., Sako, M.K., Yamashita, K.M., 1980. Geophysical ob-
Arnet, F., Kahle, H.-G., Klingele¤, E., Smith, R.B., Meertens, servations of Kilauea volcano, Hawaii, 2. Constraints on the
C.M., Dzurisin, D., 1997. Temporal gravity and height magma supply during November 1975^September 1977.
changes of the Yellowstone caldera, 1977^1994. Geophys. J. Volcanol. Geotherm. Res. 7, 241^269.
Res. Lett. 24, 2741^2744. Dzurisin, D., Savage, J.C., Fournier, R.O., 1990. Recent crus-
Avallone, A., Zollo, A., Briole, C., Delacourt, C., Beauducel, tal subsidence at Yellowstone caldera, Wyoming. Bull. Vol-
F., 1999. Subsidence of Campi Flegrei (Italy) detected by canol. 52, 247^270.
SAR interferometry. Geophys. Res. Lett. 26, 2303^2306. Eggers, A.A., 1987. Residual gravity changes and eruption
Bailey, R.A., Dalrymple, G.B., Lanphere, M.A., 1976. Volcan- magnitudes. In: Williams, S.N., Carr, M.J. (Eds.), Richard
ism, structure and geochronology of Long Valley Caldera, E. Stoiber 75th Birthday Volume. J. Volcanol. Geotherm.
Mono County, California. J. Geophys. Res. 81, 725^744. Res. 33, 201^216.
Barberi, F., Carapezza, M.L., 1996. The problem of volcanic Eggers, A.A., Chavez, D., 1979. Temporal gravity variations at
unrest: the Campi Flegrei case history. In: Scarpa, R., Till- Pacaya volcano, Guatemala. J. Volcanol. Geotherm. Res. 6,
ing, R.I. (Eds.), Monitoring and Mitigation of Volcano Haz- 391^402.
ards. Springer, Berlin, pp. 771^786. Eichelberger, J.C., 1980. Vesiculation of ma¢c magma during
Battaglia, M., Roberts, C., Segall, P., 1999. Magma intrusion replenishment of silic magma reservoirs. Nature 288, 446^
beneath Long Valley Caldera con¢rmed by temporal 450.
changes in gravity. Science 285, 2119^2122. Fujii, T., Nakada, S., 1999. The 15 September 1991 pyroclastic
Baxter, P.J., Gresham, A., 1997. Deaths and injuries in the £ows at Unzen Volcano (Japan): a £ow model for associ-
eruption of Galeras Volcano, Colombia, January 14, 1993. ated ash-cloud surges. J. Volcanol. Geotherm. Res. 89, 159^
J. Volcanol. Geotherm. Res. 77, 325^338. 172.
Berrino, G., 1994. Gravity changes induced by height-mass Huppert, H.E., Sparks, R.S.J., Turner, J.S., 1982. E¡ects of
variations at the Campi Flegrei caldera. J. Volcanol. Geo- volatiles on mixing in calc-alkaline magma systems. Nature
therm. Res. 61, 293^309. 297, 554^557.
Berrino, G., Corrado, G., Luongo, G., Toro, B., 1984. Ground Jachens, R.C., Roberts, C.W., 1985. Temporal and areal grav-
deformation and gravity changes accompanying the 1982 ity investigations at Long Valley Caldera, California.
Pozzuoli uplift. Bull. Volcanol. 47, 187^200. J. Geophys. Res. 90, 11210^11218.

VOLGEO 2381 6-5-02

 G. Williams-Jones, H. Rymer / Journal of Volcanology and Geothermal Research 113 (2002) 379^389 389

Jackson, J.A. (Ed.), 1997. Glossary of Geology, 4th edn., Rymer, H., Williams-Jones, G., 2000. Volcanic eruption pre-
American Geological Institute, Alexandria, VA, 769 pp. diction: magma chamber physics from gravity and deforma-
Johnson, D.J., 1987. Elastic and inelastic magma storage at tion measurements. Geophys. Res. Lett. 27, 2389^2392.
Kilauea Volcano. In: Decker, R.W., Wright, T.L., Stau¡er, Saemundsson, K., 1978. Fissure swarms and central volcanoes
P.H. (Eds.), Volcanism in Hawaii. U.S. Geol. Surv. Prof. of the neo-volcanic zones of Iceland. Geol. J. Spec. Issue 10,
Pap. 1350, pp. 1297^1306. 415^432.
Johnson, G.V., Bjo«rnsson, A., Sigurdsson, S., 1980. Gravity Sigmundsson, F., Vadon, H., Massonnet, D., 1997. Readjust-
and elevation changes caused by magma movement beneath ment of the Kra£a spreading segment to crustal rifting mea-
the Kra£a caldera, Northeast Iceland. J. Geophys. Res. 47, sured by satellite radar interferometry. Geophys. Res. Lett.
132^140. 24, 1843^1846.
Kane, M.F., Mabey, D.R., Brace, R.L., 1976. A gravity and Smith, R.B., Reilinger, R.E., Meertens, C.M., Hollis, J.R.,
magnetic investigation of the Long Valley caldera, Mono Holdahl, S.R., Dzurisin, D., Gross, W.K., Klingele, E.E.,
County, California. J. Geophys. Res. 81, 754^762. 1989. What’s moving at Yellowstone? The 1987 crustal de-
Kazahaya, K., Shinohara, H., Saito, G., 1994. Excessive de- formation survey from GPS, levelling, precision gravity, and
gassing of Izu-Oshima volcano: magma convection in a con- trilateration. EOS Trans. Am. Geophys. Union 70, 113^
duit. Bull. Volcanol. 56, 207^216. 125.
McKee, C., Mori, J., Talai, B., 1989. Microgravity changes Sparks, R.S.J., Meyer, P., Sigurdsson, H., 1980. Density var-
and ground deformation at Rabaul caldera, 1973^1985. iation amongst Mid-Ocean Ridge basalts: implications for
In: Latter, J.H. (Ed.), Volcanic Hazards: Assessment and magma mixing and the scarcity of primitive lavas. Earth
Monitoring. IAVCEI Proceedings in Volcanology 1. Spring- Planet. Sci. Lett. 46, 419^430.
er, Berlin, pp. 399^428. Sparks, S.R.J., Sigurdsson, H., Wilson, L., 1977. Magma mix-
Mogi, K., 1958. Relations between the eruptions of various ing: a mechanism for triggering acid explosive eruptions.
volcanoes and the deformations of the ground surfaces Nature 267, 315^318.
around them. Bull. Earthquake Res. Inst. 36, 99^134. Stevenson, D.S., Blake, S., 1998. Modelling the dynamics and
Murray, J.B., Pullen, A.D., Saunders, S., 1995. Ground defor- thermodynamics of volcanic degassing. Bull. Volcanol. 60,
mation surveying of active volcanoes. In: McGuire, B., Kil- 307^317.
burn, C.R.J., Murray, J. (Eds.), Monitoring Active Volca- Tait, S., Jaupart, C., Vergniolle, S., 1989. Pressure, gas content
noes. UCL Press, London, pp. 113^150. and eruption periodicity of a shallow, crystallising magma
Newhall, C.G., Dzurisin, D., 1988. Historical unrest at large chamber. Earth Planet. Sci. Lett. 92, 107^123.
calderas of the world. U.S. Geol. Surv. Bull. 1855, 1108. Tryggvason, E., 1984. Widening of the Kra£a ¢ssure swarm
Pallister, J.S., Hoblitt, R.P., Reyes, A.G., 1992. A basalt trig- during the 1975^1981 volcano-tectonic episodes. Bull. Vol-
ger for the 1991 eruptions of Pinatubo volcano. Nature 356, canol. 47, 47^69.
426^428. Vasco, D.W., Smith, R.B., Taylor, C.L., 1990. Inversion for
Rymer, H., 1994. Microgravity change as a precursor to vol- sources of crustal deformation and gravity change at
canic activity. J. Volcanol. Geotherm. Res. 61, 311^328. the Yellowstone caldera. J. Geophys. Res. 95, 19839^
Rymer, H., Brown, G.C., 1989. Gravity changes as a precursor 19856.
to volcanic eruptions at Poa¤s volcano, Costa Rica. Nature Walker, G.P.L., Heming, R.F., Sprod, T.J., Walker, H.R.,
342, 902^905. 1981. Latest major eruptions of Rabaul volcano. In: John-
Rymer, H., Cassidy, J., Locke, C.A., Murray, J.B., 1995. Mag- son, R.W. (Ed.), Cooke-Ravian Volume of Volcanological
ma movements in Etna volcano associated with the major Papers. Geol. Surv. Papua New Guinea 10, 81^193.
1991^1993 lava eruption: evidence from gravity and defor- Wicks, C., Thatcher, W., Dzurisin, D., 1998. Migration of
mation. Bull. Volcanol. 57, 451^461. £uids beneath Yellowstone caldera inferred from satellite
Rymer, H., Cassidy, J., Locke, C.A., Sigmundsson, F., 1998. radar interferometry. Science 282, 458^462.
Post-eruptive gravity changes from 1990 to 1996 at Kra£a, Woods, A.W., Koyaguchi, T., 1994. Transitions between ex-
Iceland. J. Volcanol. Geotherm. Res. 87, 141^149. plosive and e¡usive eruptions of silicic magmas. Nature 370,
Rymer, H., van Wyk de Vries, B., Stix, J., Williams-Jones, G., 641^644.
1998. Pit crater structure and processes governing persistent Yokoyama, I., 1989. Microgravity and height changes caused
activity at Masaya Volcano, Nicaragua. Bull. Volcanol. 59, by volcanic activity: four Japanese examples. Bull. Volcanol.
345^355. 51, 333^345.

VOLGEO 2381 6-5-02