ArticlePDF Available

Abstract and Figures

Lying below Vatnajökull ice cap in Iceland, Bárðarbunga stratovolcano began experiencing wholesale caldera collapse in 2014 August 16, one of the largest such events recorded in the modern instrumental era. Simultaneous with this collapse is the initiation of a plate boundary rifting episode north of the caldera. Observations using the international constellation of radar satellites indicate rapid 50 cm d−1 subsidence of the glacier surface overlying the collapsing caldera and metre-scale crustal deformation in the active rift zone. Anomalous earthquakes around the rim of the caldera with highly nondouble-couple focal mechanisms provide a mechanical link to the dynamics of the collapsing magma chamber. A model of the collapse consistent with available geodetic and seismic observations suggests that the majority of the observed subsidence occurs aseismically via a deflating sill-like magma chamber.
Content may be subject to copyright.
Geophysical Journal International
Geophys. J. Int. (2015) 202, 446–453 doi: 10.1093/gji/ggv157
GJI Mineral physics, rheology, heat flow and volcanology
The collapse of B´
ararbunga caldera, Iceland
B. Riel,1P. Milillo,1,2M. Simons,1P. Lundgren,3H. Kanamori1and S. Samsonov4
1Seismological Laboratory, California Institute of Technology, 1200 E. California Blvd., Pasadena, CA 91125, USA. E-mail:
2Scuola di Ingegneria, Universit`
a degli Studi della Basilicata, Viale dell’Ateneo Lucano 10,I-85100 Potenza, Italy
3Jet Propulsion Laboratory, 4800 Oak Grove Dr., Pasadena, CA 91109,USA
4Canada Centre for Mapping and Earth Observation, Natural Resources, Ottawa, ON K1S5K2, Canada
Accepted 2015 April 8. Received 2015 April 2; in original form 2015 January 21
Lying below Vatnaj¨
okull ice cap in Iceland, B´
ararbunga stratovolcano began experiencing
wholesale caldera collapse in 2014 August 16, one of the largest such events recorded in the
modern instrumental era. Simultaneous with this collapse is the initiation of a plate boundary
rifting episode north of the caldera. Observations using the international constellation of radar
satellites indicate rapid 50 cm d1subsidence of the glacier surface overlying the collapsing
caldera and metre-scale crustal deformation in the active rift zone. Anomalous earthquakes
around the rim of the caldera with highly nondouble-couple focal mechanisms provide a
mechanical link to the dynamics of the collapsing magma chamber. A model of the collapse
consistent with available geodetic and seismic observations suggests that the majority of the
observed subsidence occurs aseismically via a deflating sill-like magma chamber.
Key words: Satellite geodesy; Radar interferometry; Magma chamber processes; Calderas;
Remote sensing of volcanoes.
In 2014 August 16, a swarm of earthquakes was detected under-
neath B´
ararbunga caldera in Iceland, a stratovolcano located within
the Eastern Volcanic Zone (EVZ) and completely covered by the
Va tn aj ¨
okull ice cap (NEIC 2014). These earthquakes signaled the
onset of subsurface magma movement. As magma propagated out
of the confines of the caldera, earthquake activity tracked its mo-
tions, revealing the emplacement of a large dyke along a northeast
oriented fissure swarm of the B´
ararbunga volcanic system, consis-
tent with a plate boundary rifting event (Gudmundsson et al. 2014;
Sigmundsson et al. 2014). The dyke intrusion triggered an effu-
sive eruption 40 km away from the caldera at the surface of the
Holuhraun lava field north of Vatnaj¨
okull beginning in 2014 Au-
gust 29. Prior to the surface eruption, geodetic observations revealed
that the ice over the caldera was subsiding rapidly, with measured
rates of approximately 50 cm d1(Fig. 1). The rapid subsidence
was accompanied by moderate earthquakes (Mw>5) with epicen-
tres concentrated along the caldera rim (Fig. 2). Nearly all of these
earthquakes exhibited anomalous behaviour with large deviations
from traditional double-couple sources (i.e. motion confined to a
shear fault plane; Aki & Richards 2002). These earthquakes ap-
pear to be the manifestations of simultaneous vertical compression
and outward horizontal expansion. Such motion is commonly in-
terpreted as a compensated linear vector dipole (CLVD; Knopoff &
Randall 1970). The close spatiotemporal association of the caldera
collapse, anomalous seismicity and large-scale rifting provides a
unique opportunity to study the mechanics of a caldera collapse
in a basaltic system. The large subsidence within the caldera rim,
which has never been previously observed at B´
ararbunga, provides
critical constraints on the collapse sequence within the caldera.
Since the start of the eruption, a suite of synthetic aperture radar
(SAR) images over northwest Vatnaj¨
okull and adjacent regions has
been acquired by the international constellation of radar satellites.
With these images, we can use interferometric SAR (InSAR) to
measure surface deformation between two successive SAR images
along a line-of-sight (LOS) direction (e.g. Simons & Rosen 2007).
We use images acquired by the COSMO-SkyMed (CSK) constella-
tion, which consists of four X-band radar satellites operated by the
Italian Space Agency (ASI), to image ground deformation within
the vicinity of the B´
ararbunga caldera (Fig. 1). One-day separa-
tion between CSK images over the collapsing ice-covered caldera
permits the formation of high-resolution interferograms with good
coherence, providing snapshots of daily subsidence of the overlying
ice. We complement the CSK data with 24- and 48-d-interval InSAR
observations from RADARSAT-2 (RS2), a C-band satellite operated
by the Canadian Space Agency, to measure ground deformation in
ice-free regions north of Vatnaj¨
okull. This deformation primarily
results from emplacement of an intrusive dyke, producing 1.5 m of
surface ground motion very close to the dyke and measurable defor-
mation as far as 60 km away from the surface trace. Near Herubrei
volcano northeast of Askja, a cluster of earthquakes is associated
with left-lateral fault motion where discrete centimetre-scale sur-
face rupture can be observed. The left-lateral motion agrees with
the previously inferred bookshelf faulting for that area and implies
that dyke emplacement enhanced the background stress field in the
446 C
The Authors 2015. Published by Oxford University Press on behalf of The Royal Astronomical Society.
by guest on May 11, 2015 from
Collapse of B´
ararbunga caldera, Iceland 447
Figure 1. Map of northwestern Vatnaj¨
okull, Iceland with the locations of known calderas/volcanos indicated by barbed lines: As, Askja; B´
a, B´
ararbunga; Gr,
otn, H, Herubrei;Kv,Kverkfj
oll; Tu, Tungnafellsj ¨
okull. Coloured fringes represent contours of the line-of-sight (LOS) component of ground motion
at 15 cm per colour cycle. Areas of high interferometric phase noise have been masked out. The interferogram over the ice was formed from COSMO-SkyMED
(CSK) images acquired from 2014 September 12–13 while the 24-d interval interferogram over the ground was formed from RADARSAT-2 (RS2) images
acquired on 2014 August 8 and September 1. A clear bullseye pattern over the B´
ararbunga caldera indicates subsidence of the glacier surface. The RS2
interferogram shows the effects of the rifting associated with spreading of the ground away from the active dyke. Black arrow indicates the satellite-to-ground
direction for both the CSK and RS2 interferograms. White dots indicate earthquakes that occurred between 2014 August 15 and 2015 February 1, as recorded
by the SIL network in Iceland (B¨
ovarsson et al. 1999). Lower left inset shows the location of the study area in Iceland with respect to the glaciers and
active volcanic zones (dark grey shading). Upper left inset shows surface rupture near the Herubreivolcano (denoted by white box) observed in the RS2
Askja rift segment (Green et al. 2014; Gudmundsson et al. 2014).
In the ice-free areas adjacent to and west of B´
ararbunga caldera,
inspection of the LOS displacements in the RS2 data reveals several
centimetres of motion consistent with a deflating magma chamber
beneath the caldera.
We formed five 1-d interferograms from CSK images acquired on
August 27–28, September 12–13, September 13–14, September
17–18 and October 19–20 (Fig. 2). All interferograms show strong
subsidence signals within the caldera boundary, presumably due to
subsidence of the caldera floor. While melting of the overlying ice at
its base could also result in the observed subsidence, there has been
no evidence of anomalous glacial outwash or changes in the ice flow
rates for the central B´
ararbunga caldera (Sigmundsson et al. 2014).
The first two and the last interferogram show similar, axisymmetric
bullseye patterns due to the subsidence of the ice over the caldera.
However, only a day after the September 12–13 interferogram, the
September 13–14 interferogram shows a distinctly different defor-
mation pattern with greater displacement near the northern rim of
the caldera. Because ground displacement in one satellite line-of-
sight direction is generally a combination of horizontal and vertical
motion, we use the method of Yun et al. (2006) to estimate the ratio
of horizontal to vertical motion on the ice by comparing interfero-
grams with different viewing geometries (Supporting Information
Fig. S1). We estimate that the ratio of horizontal to vertical motion
does not exceed 0.3 within the caldera, and any horizontal motion
is diminished close to the centre of the subsidence signal. Thus,
assuming purely vertical motion, we can extract profiles across the
caldera to estimate instantaneous subsidence rate for each of the
interferograms (Supporting Information Fig. S2). The northward
trending subsidence pattern in the September 13–14 interferogram
is clearly associated with a sharp increase in apparent instantaneous
subsidence rate over the background rate: 20 cm increase in the
centre of the caldera and 25 cm on the northern edge of the caldera.
Two earthquakes with Mw4.9 and Mw5.3 occurred on the north
rim of the caldera during the time spans of the September 13–14
by guest on May 11, 2015 from
448 B. Riel et al.
Figure 2. (a)–(e) Evolution of subsidence within the caldera observed from five 1-d interval CSK interferograms. Note the similarity in LOS directions for
the interferograms, with the exception of the September 17–18 interferogram. Coloured circles indicate earthquakes that occurred during the time span of the
interferogram and are coloured by hours elapsed from the acquisition time of the first image of a given pair. Earthquakes with Mw>4.9 are shown with focal
mechanisms derived from moment tensors obtained from the GFZ Potsdam catalog (Centre 1993). The location of the transects in Supporting Information
Fig. S2 are indicated in the first panel for the August 27–28 interferogram. (f) Distribution of all seismicity near the caldera between 2014 August 16 to October
21. The coloured focal mechanisms are coloured by days elapsed since 2014 August 23. Yellow dots denote the locations of the earthquakes presented in Fig. 1.
The black focal mechanisms with vertical tensional axes correspond to selected CLVD events at B´
ararbunga between 1976 and 1996 (Nettles & Ekstr¨
and September 17–18 interferograms, respectively (Fig. 2). These
two earthquakes were part of a persistent sequence of Mw>5
earthquakes along the rim of the caldera with highly nondouble-
couple characteristics, exhibiting dominant vertical compressional
axes and horizontal expansions consistent with CLVDs (Fig. 2)
(Centre 1993). The focal mechanisms for this sequence of events
are consistent between different earthquake catalogs (Supporting
Information Table S1). While previous studies of global CLVD
earthquakes in the vicinity of active volcanoes have shown that
such earthquakes tend to have longer-than-average source durations
for their magnitudes, estimates of the magnitudes themselves are
weakly dependent on the complexity of the source time functions
(Shuler et al. 2013a,b). Therefore, the small differences in mo-
ment magnitudes between different catalogues for the current event
(Supporting Information Table S1) suggests that the magnitudes are
correct to within approximately 0.1 magnitude unit.
The Mw4.9 event has a clear CLVD component in the moment
tensor while the Mw5.3 event appears to show more normal fault
motion. The location, size and focal mechanisms of these events,
and the absence of similar events during the time spans of the Au-
gust 27–28, September 12–13 and October 19–20 interferograms,
suggests that the asymmetry of the September 13–14 and Septem-
ber 17–18 interferograms is linked to the occurrence of the larger
earthquakes on the northern rim. A simple relationship between
earthquake size and ground displacement, such as assuming seis-
mic potency is proportional to a really integrated displacements of
the ice surface, does not appear to apply to these events since the
potency associated with the earthquake that occurred during the
September 13–14 interval was nearly five times less than the earth-
quake that occurred during the September 17–18 interval, yet the
estimated volume change within the caldera as estimated from the
integrated displacements is larger for the former (Fig. 3d). There-
fore, most of the subsidence occurs aseismically while the larger
Mw>5 events may produce localized additional displacements of
up to 25 cm along the caldera rim as measured at the ice surface.
Experimental and numerical studies of caldera collapse consistently
show that the size, shape and depth of subsurface magma chambers
strongly affect the final geometry of the collapsed caldera, as well as
the rate at which it will form Roche et al. (2000), Geyer et al. (2006),
Gudmundsson (2007,2008). In order to estimate the depth of a sub-
surface magma chamber while avoiding modelling errors due to
uncertain ice-rock coupling, we consider LOS displacements of the
subsidence signal from the ice-free areas in interferograms formed
from longer time interval RS2 images (24 and 48 d, Supporting In-
formation Fig. S8). We also include data from continuous global
positioning system (GPS) stations located north of Vatnaj¨
(Sigmundsson et al. 2014). As part of the analysis, we remove
ground displacements due to the rift zone by estimating an elastic
model for tensile opening along the dyke interface using a collection
of RS2 and CSK interferograms and the GPS data (see Supporting
by guest on May 11, 2015 from
Collapse of B´
ararbunga caldera, Iceland 449
Figure 3. (a) Difference between September 12–13 and September 13–14 interferograms for isolating the ground deformation associated with a Mw4.9
earthquake that occurred on September 14. Also shown are synthetic LOS displacements due to slip along a vertical ring fault (b) and a closing crack (c).For
the ring fault (thick black line), the fault is placed at a depth of 2 km with a width of 1 km and approximately 3 m of normal motion. The closing crack is
simulated using the model of Fialko et al. (2001) at a depth of 2 km, a radius of 1 km and a volume change of 0.003 km3. (d) Seismic potency for the northern
and southern halves of the caldera compared with estimated volume change rate (dV/dt) within the caldera inferred from the 1-d interferograms. The thick blue
and red lines correspond to the cumulative potency of the northern and southern regions, respectively, of the caldera. To calculate seismic potency, we divide
the seismic moment by a shear modulus of 24 GPa. Thin vertical lines mark the occurrence of earthquakes greater than Mw4.9. The shaded brown boxes show
the time spans of each 1-d interferogram and the estimated volume change during that time span (right ordinate). The volume change for each interferogram
was computed by integrating the projected vertical displacements within the caldera boundary.
Information Section S1.1; Fig. S4). While the geodetic data alone
cannot resolve opening of the dyke under the ice, the robust com-
ponents of the model include approximately 5 m of opening in the
upper 5 km of the resolvable dyke segment(equivalent to 200–300 yr
of cumulative plate motion) and peak geodetic potency occurring at
a shallower depth (2 km) than the peak earthquake density, which
occurs at around 6 km depth (Sigmundsson et al. 2014; Support-
ing Information Fig. S6). These results agree with estimates of the
dyke geometry from other studies (e.g. Gudmundsson et al. 2014;
Sigmundsson et al. 2014).
After removal of the rift zone signal, we model the chamber as a
collapsing horizontal circular crack in an elastic half-space (Fialko
et al. 2001; Supporting Information Section S1.2). The parame-
terization of the chamber parameters is such that there are strong
trade-offs between the chamber depth, radius and excess pressure
(difference between the magma chamber and lithostatic pressures).
However, we are able to resolve a consistent depth-to-radius ratio
of approximately 3.6 (Supporting Information Fig. S7). By adjust-
ing the excess pressure, we can explain the geodetic observations
equally well with a shallow, small chamber or a deep, larger cham-
ber. Therefore, determining the ‘true’ depth of the magma cham-
ber would require other independent observations, that is re-located
seismicity of earthquakes occurring within and around the caldera or
an upper bound on the allowable values for excess pressure. Never-
theless, the ability of our model to fit the displacements at distances
greater than three times the chamber radius validates the assumption
of a symmetric source with a uniform pressure difference since no
obvious asymmetries appear in the residuals. Steady deflation of the
chamber is thus the primary contributor to subsidence observations
both on and off the ice. While we cannot rule out the possibility
of additional deeper magma chambers (depth >10 km), our lack of
reliable ground measurements over the caldera limits our ability to
resolve multiple chambers.
One of the most interesting aspects of the collapse sequence has
been the occurrence of the moderate earthquakes along the caldera
rim with large CLVD components in their focal mechanisms. As-
suming that the background subsidence rate is nearly constant
by guest on May 11, 2015 from
450 B. Riel et al.
Figure 4. A conceptual model for the mechanics of the collapsing caldera. Magma migration out of the chamber into the dyke system causes depressurization of
the chamber. The deflating chamber results in subsidence of a coherent block above the chamber, as well as subsidence of the overlying ice. Sustained depressur-
ization of the chamber leads to failure of internal support structures (causing seismic events marked by inward vertical motion and outward horizontal expansion)
or rupturing of curved, inward-dipping ring faults. Cross-sections of focal mechanisms representative of these events are shown on the edge of the chamber.
between September 12–14, we can isolate the ground deformation
associated with one of the earthquakes (Mw4.9 on September 13)
by forming a residual interferogram of the difference between the
September 12–13 and September 13–14 CSK interferograms since
only the September 13-14 image spanned the time of the earthquake
(Fig. 3a). The highly localized deformation due to the earthquake
indicates that the source is most likely shallow. For earthquakes with
vertical CLVD focal mechanisms in the vicinity of volcanoes, the
two most likely physical processes are rupture on curved ring faults
or opening/closing of a crack under tension/compression (Acocella
2007; Shuler et al. 2013b). Static simulations for both processes
require source depths of approximately 2 km to roughly match the
InSAR observations (Figs 3b and c). For reference, a Mw4.9 event
with an effective shear modulus of 5 GPa (typical for shallow depths
in volcanic systems within an extensional regime (Smith et al.
1996)) has a seismic potency of approximately 0.004 km3.How-
ever, the simulated ring fault rupture requires a much larger potency
(0.01 km3), which implies a very low effective shear modulus
(1 GPa) consistent with very weak materials such as water-
saturated basaltic tuff. While the closing crack provides a more
realistic measure of potency (equivalent volume change of about
0.003 km3), the synthetic ground deformation lacks the asymmetry
observed in the residual interferogram. Therefore, for both cases, it
is likely that the ground deformation is a combination of seismic de-
formation (due to ring faulting or a closing crack) and aseismic slip
on the ring fault. The inconsistency between earthquake size and
ground deformation (Fig. 3d) suggests that the amount of aseismic
slip per event is highly variable.
Based on the seismic evidence and the inversion results for the
chamber geometry, we propose a model for the sequence of the
caldera collapse and dyke emplacement (Fig. 4). The initial seismic
activity on the southern edge of the caldera and subsequent propaga-
tion of an oblique dyke caused depressurization of a magma cham-
ber that can be approximated as a horizontal circular sill. Magma
migrated out of the chamber to the short, oblique dyke and eventu-
ally migrated to the larger, regional-scale dyke. The underpressure
in the magma chamber resulted in subsidence of the caldera surface
and overlying ice. Stress concentrations in the vicinity of the deflat-
ing magma chamber led to the initiation of Mw>5 seismic events
located along the caldera rim. The large CLVD components in the
earthquake focal mechanisms indicate a seismic process character-
ized either by downward vertical motions and horizontal expansions
or rupture on a curved fault. The former process has been observed
in mine collapses where rapid closing of a horizontally oriented
underground cavity (the mine) leads to CLVD components in the
seismic moment tensor (Dreger et al. 2008). A mine collapse mech-
anism could imply failure of brittle support structures within the
partially molten magma chamber due to large compressive stresses,
which would require a larger (and thus deeper) magma chamber
since the earthquake clusters on the northern and southern rim of
the caldera are separated by approximately 8–10 km. Alternatively,
the deflating chamber could impart large shear stresses on existing
ring faults, leading to seismogenic slip on limited portions of a ring
fault. Static stress changes associated with both mechanisms could
then trigger larger aseismic slip on the ring fault with slip most
likely confined to shallower depths.
Earthquakes with significant nondouble-couple components in
their focal mechanisms have occurred at B´
ararbunga in the
decades prior to the current event (Supporting Information Table S2;
Nettles & Ekstr¨
om 1998; Konstantinou et al. 2003;Tkal
cet al.
2009). However, the polarities of the focal mechanisms for the
current B´
ararbunga eruption are opposite to those observed for
by guest on May 11, 2015 from
Collapse of B´
ararbunga caldera, Iceland 451
the earlier earthquakes, that is vertical compressional axes for the
former and tensional axes for the latter (Fig. 2f). One interpre-
tation of these earlier events suggested that their faulting mecha-
nisms are primarily due to rupture on outward dipping ring faults
which are activated by inflation of a very shallow magma chamber
(Nettles & Ekstr¨
om 1998). However, this interpretation is highly
nonunique, since a deflating magma chamber below the ring faults
could also produce similar focal mechanisms (Ekstr¨
om 1994). Our
inversion results for the magma chamber predict a chamber radius
smaller than the caldera radius for chamber depths less than 15 km
(Supporting Information Figs S7 and S8), which implies inward
dipping ring faults. The dip of the ring faults controls the rupture
arc-length necessary to create the CLVD focal mechanisms. For
this geometry, the earlier events can be explained by inflation of the
magma chamber, leading to reverse motion on those ring faults.
In addition to mine collapses and ring fault rupture, earthquakes
with large CLVD components like the ones observed on the rim of
the B´
ararbunga caldera have been observed for tensile failure due
to high fluid pressure (Sipkin 1986) and magma injection in water-
saturated environments (Kanamori et al. 1993). An earthquake with
a strong CLVD component was observed near Tori Shima, Japan
in 1984. Since the earthquake was tsunamigenic, the preferred ex-
planation was horizontal injection of magma into water-saturated
sediment with a volumetric component resulting from the explosive
magma-water interaction (Kanamori et al. 1993). A global search
for earthquakes prior to 2013 with vertical-CLVD mechanisms re-
vealed 100 vertical-CLVD earthquakes located near active vol-
canoes over the past century (Shuler et al. 2013a). Thus, the close
temporal and spatial association of the current B´
ararbunga anoma-
lous events with the active fissuring lends credibility to a collapse
mechanism driven by a drop in magma pressure. A sudden ver-
tical collapse with no volume change would be characterized by
downward vertical and outward horizontal motions.
Our model of caldera collapse due to magma withdrawal has
been proposed for volcanic systems in Iceland neighbouring
ararbunga. North of the EVZ is the Northern Volcanic Zone
(NVZ), consisting of five en echelon volcanic systems aligned
with the boundary of the North American and European plates
(Einarsson 1991). Within the NVZ, the Krafla caldera system was
the site of a major rifting event between 1975 and 1984. Eruptive
activity initiated with inflation of the central caldera, which then
underwent rapid deflation, leading to lateral basaltic magma injec-
tion into a northward trending fissure swarm. A shallow magma
chamber at approximately 3 km depth has been inferred at Krafla
via inversions of geodetic observations (Ewart et al. 1991). Seismic
tremor amplitude associated with dyke emplacement was corre-
lated with subsidence at the caldera (Einarsson 1991), suggesting
a mechanistic link between chamber pressure and stress on the
dyke. Numerical models support the hypothesis that magma pres-
sure, dyke overpressure, and background tectonic stresses are the
primary factors controlling dyke propagation (Buck et al. 2006).
Recent studies of the current dyke emplacement at B´
also predict a feedback mechanism where dyke-induced stresses
trigger seismicity at the caldera itself (Gudmundsson et al. 2014).
Kinematically, however, the two volcanic systems are quite different
since the maximum subsidence for Krafla (2 m) was substantially
smaller than that of B´
ararbunga (about 60 m). Additionally, the
Krafla episode experienced periodic inflation/deflation over 10 yr
whereas B´
ararbunga has thus far only experienced rapid deflation
and over a much shorter time period.
On a global scale, there have only been a few observed erup-
tive events and subsequent caldera collapses in basaltic systems,
although none with volume changes as large as B´
ararbunga. The
1968 caldera collapse at Isla Fernandina in the Gal´
apagos Islands
occurred on a single shield volcano, and aerial observations indi-
cated that the caldera floor collapsed 300 m as a 3 km wide coherent
block with motion confined to an elliptical boundary fault formed
from a prior collapse event (Simkin & Howard 1970; Filson et al.
1973;Francis1974). In 2000, the Miyakejima stratovolcano in the
Izu-Bonin volcanic chain experienced 12 d of increased seismicity
due to magma intrusion at its northwest flanks, which was then
followed by a minor phreatic eruption and formation of a collapsed
caldera 1.6 km in diametre (Ukawa et al. 2000; Geshi et al. 2002).
Tiltmeters stationed around the summit indicated intermittent abrupt
uplift events superposed on the longer term subsidence. The uplift
events were accompanied by very long period (VLP) seismic events
detectable over a wide area (Ukawa et al. 2000; Kumagai et al.
2001). Similar observations were collected for the 2007 Piton de
la Fournaise caldera collapse during its largest historical eruption
(Michon et al. 2009). For Miyakejima, all of the VLP events ex-
hibited large CLVD components for their focal mechanisms, which
was explained to be slip on both inward- and outward-dipping faults
(Geshi et al. 2002; Shuler et al. 2013a). 60 km south of the Isla
Fernandina caldera, an inflation event occurred at the Sierra Ne-
gra caldera which was followed a few years later by a Mw5.5
CLVD earthquake with a vertical tensional axis (Yun et al. 2006;
Shuler et al. 2013a), consistent with the interpretation of the older
ararbunga events being caused by inflation of a central magma
While the unique geodetic observations of ground deformation
within and around the B´
ararbunga caldera during its collapse were
the first of its kind for the caldera, there are still large uncertainties
regarding the mechanics of the collapse process. The difficulties as-
sociated with the unknown interaction between the overlying glacier
and the bedrock limits the spatial extent of usable data for estimat-
ing the geometry of the underlying magma chamber and active ring
fault systems. We have shown that a majority of the ice-free ground
deformation can be attributed to steady deflation of a horizontal
circular sill. However, the physical process driving the anomalous
Mw>5 earthquakes along the caldera rim is still uncertain. The
geodetic signature of one of these events, as measured by the dif-
ferences in successive 1-d interferograms over the ice, suggests a
shallow seismic source caused by rupture on a ring fault or a rapidly
closing crack. Since the amount of ground deformation expected
for either seismic mechanism is significantly less than the observed
deformation, we believe that the total ground deformation is caused
by a combination of seismic processes and aseismic slip on ring
faults. However, if the shear modulus in the vicinity of the caldera
were much lower than expected (perhaps due to the presence of
water-saturated basaltic tuff), then the predicted deformation for
the seismic component would be much larger. In that case, the
earthquakes can be explained entirely by rupture on dipping ring
We thank Egill Hauksson and Hilary Martens for discussions in the
early phase of this study. We also thank Agust Gudmundsson, J¨
Schuler and three anonymous reviewers for their helpful comments
and for improving the quality of this manuscript. BR was supported
under a National Aeronautics and Space Administration (NASA)
by guest on May 11, 2015 from
452 B. Riel et al.
Earth and Space Science Fellowship. Part of this research was
carried out at the Jet Propulsion Laboratory and the California Insti-
tute of Technology under a contract with NASA and funded through
the President’s and Director’s Fund Program. This research was car-
ried out using COSMO-SkyMed (CSK R
) products delivered under
an Italian Space Agency (ASI) license and is made possible through
a collaboration between JPL/Caltech/CIDOT and NASA/ASI.
Acocella, V., 2007. Earth-Sci. Rev., 85(3), 125–160.
Adam, L. & Otheim, T., 2013. J. geophys. Res., 118(3), 840–851.
Aki, K. & Richards, P.G., 2002. Quantitative Seismology, Vol. 1, University
Science Books.
ovarsson, R., R¨
ognvaldsson, S.T., Slunga, R. & Kjartansson, E., 1999.
The SIL data acquisition system—at present and beyond year 2000, Phys.
Earth planet. Inter., 113(1), 89–101.
Buck, W.R., Einarsson, P. & Brandsd´
ottir, B., 2006. Tectonic stress and
magma chamber size as controls on dike propagation: constraints from
the 1975–1984 Krafla rifting episode, J.geophys. Res.: Solid Earth (1978–
2012), 111, B12404, doi:10.1029/2005JB003879.
Centre, G.D., 1993. Geofon seismic network. Available at: deutsches geo-
forschungszentrum gfz.other/seismic network (last accessed 2014 Octo-
ber 13).
Dreger, D.S.,Ford, S.R. & Walter, W.R., 2008. Source analysis of the crandall
canyon, Utah, mine collapse, Science, 321(5886), 217–217.
Einarsson, P., 1991. Earthquakes and present-day tectonism in Iceland,
Tectonophysics, 189(1), 261–279.
om, G., 1994. Anomalous earthquakes on volcano ring-fault structures,
Earth planet. Sci. Lett., 128(3), 707–712.
Ewart, J.A., Voight, B. & Bj¨
ornsson, A., 1991. Elastic deformation models of
krafla volcano, iceland, for the decade 1975 through 1985, Bull. Volcanol.,
53(6), 436–459.
Fialko, Y., Khazan, Y. & Simons, M., 2001. Deformation due to a pressurized
horizontal circular crack in an elastic half-space, with applications to
volcano geodesy, Geophys. J. Int., 146(1), 181–190.
Filson, J., Simkin, T. & Leu, L.-k., 1973. Seismicity of a caldera collapse:
Galapagos islands 1968, J. geophys. Res., 78(35), 8591–8622.
Francis, T., 1974. A new interpretation of the 1968 Fernandina caldera
collapse and its implications for the mid-oceanic ridges, Geophys. J. Int.,
39(2), 301–318.
Geshi, N., Shimano, T., Chiba, T. & Nakada, S., 2002. Caldera collapse
during the 2000 eruption of Miyakejima volcano, Japan, Bull. Volcanol.,
64(1), 55–68.
Geyer, A., Folch, A. & Mart´
ı, J., 2006. Relationship between caldera col-
lapse and magma chamber withdrawal: an experimental approach, J. Volc.
Geotherm. Res., 157(4), 375–386.
Green, R.G., White, R.S. & Greenfield, T., 2014. Motion in the north Iceland
volcanic rift zone accommodated by bookshelf faulting, Nat. Geosci.,
7(1), 29–33.
Gudmundsson, A., 2007. Conceptual and numerical models of ring-fault
formation, J. Volc. Geotherm. Res., 165, 142–160.
Gudmundsson, A., 2008. Magma-chamber geometry, fluid transport, lo-
cal stresses and rock behaviour during collapse caldera formation, Dev.
Volcanol., 10, 313–349.
Gudmundsson, A., Lecoeur, N., Mohajeri, N. & Thordarson, T., 2014. Dike
emplacement at B´
ararbunga, Iceland, induces unusual stress changes,
caldera deformation, and earthquakes, Bull. Volcanol., 76(10), 1–7.
Kanamori, H., Ekstr¨
om, G., Dziewonski, A., Barker, J.S. & Sipkin, S.A.,
1993. Seismic radiation by magma injection: an anomalous seismic event
near Tori Shima, Japan, J. geophys. Res.: Solid Earth (1978–2012),
98(B4), 6511–6522.
Knopoff, L. & Randall, M.J., 1970. The compensated linear-vector dipole:
A possible mechanism for deep earthquakes, J. geophys. Res., 75(26),
Konstantinou, K., Kao, H., Lin, C.H. & Liang, W.T., 2003. Analysis of
broad-band regional waveforms of the 1996 September 29 earthquake at
ardarbunga volcano, central Iceland: investigation of the magma injec-
tion hypothesis, Geophys. J. Int., 154(1), 134–145.
Kumagai, H., Ohminato, T., Nakano, M., Ooi, M., Kubo, A., Inoue, H. &
Oikawa, J., 2001. Very-long-period seismic signals and caldera formation
at Miyake Island, Japan, Science, 293(5530), 687–690.
Michon, L., Villeneuve, N., Catry, T. & Merle, O., 2009. How summit
calderas collapse on basaltic volcanoes: new insights from the april 2007
caldera collapse of Piton de la Fournaise volcano, J. Volc. Geotherm. Res.,
184(1), 138–151.
National Earthquake Information Center, United States Geological Survey,
2014. Retrieved from
Nettles, M. & Ekstr¨
om, G., 1998. Faulting mechanism of anomalous earth-
quakes near Bardarbunga volcano, Iceland, J. geophys. Res.: Solid Earth
(1978–2012), 103(B8), 17 973–17 983.
Roche, O., Druitt, T. & Merle, O., 2000. Experimental study of caldera
formation, J. geophys. Res.: Solid Earth (1978–2012), 105(B1), 395–416.
Shuler, A., Nettles, M. & Ekstr¨
om, G., 2013a. Global observation of vertical-
CLVD earthquakes at active volcanoes, J. geophys. Res., 118(B1), 138–
Shuler, A., Ekstr¨
om, G. & Nettles, M., 2013b. Physical mechanisms for
vertical-CLVD earthquakes at active volcanoes, J. geophys. Res., 118(B4),
Sigmundsson, F. et al., 2014. Segmented lateral dyke growth in a rifting
event at B ´
ardarbunga volcanic system, Iceland, Nature, 517, 191–195.
Simkin, T. & Howard, K.A., 1970. Caldera collapse in the Gal´
apagos islands,
1968 the largest known collapse since 1912 followed a flank eruption and
explosive volcanism within the caldera, Science, 169(3944), 429–437.
Simons, M. & Rosen, P., 2007. Interferometric synthetic aperture radar
geodesy, Treatise on Geophysics, Elsevier, pp. 391–446.
Sipkin, S.A., 1986. Interpretation of non-double-couple earthquake mech-
anisms derived from moment tensor inversion, J. geophys. Res.: Solid
Earth (1978–2012), 91(B1), 531–547.
Smith, R.P., Jackson, S.M. & Hackett, W.R., 1996. Paleoseismology and
seismic hazards evaluations in extensional volcanic terrains, J. geophys.
Res., 101(B3), 6277–6292.
c, H., Dreger, D.S., Foulger, G.R. & Julian, B.R., 2009. The puzzle of
the 1996 B´
ardarbunga, Iceland, earthquake: no volumetric component in
the source mechanism, Bull. seism. Soc. Am., 99(5), 3077–3085.
Ukawa, M., Fujita, E., Yamamoto, E., Okada, Y. & Kikuchi, M., 2000. The
2000 Miyakejima eruption: crustal deformation and earthquakes observed
by the NIED Miyakejima observation network,Earth Planet Space, 52(8),
Yun, S., Segall, P. & Zebker, H., 2006. Constraints on magma chamber
geometry at Sierra Negra volcano, Gal´
apagos islands, based on InSAR
observations, J. Volcanol. Geotherm. Res., 150(1), 232–243.
Additional Supporting Information may be found in the online ver-
sion of this paper:
Figure S1. Ascending and descending one-day CSK interferograms
with time spans of Sep. 13–14 and Sep. 17–18, respectively. The
two viewing geometries allow for separation of LOS displacements
into vertical and horizontal components using the method of Yun
et al. (2006), assuming that the two interferograms are measuring
roughly the same ground deformation. While the two interferograms
appear to measure similar levels of ice subsidence, the earthquake
perturbations are more pronounced for the Sep. 13–14 interfero-
gram (see Fig. 2). Nevertheless, the ratio of horizontal to vertical
displacements does not exceed 0.3 within the caldera.
Figure S2. Profiles of projected vertical subsidence observed in
the five one-day CSK interferograms. Profile locations are shown
in Fig. 2. Vertical displacements were computed by dividing the
line-of-sight range changes by the cosine of the incidence angle for
by guest on May 11, 2015 from
Collapse of B´
ararbunga caldera, Iceland 453
each interferogram. Solid lines are constructed from interferogram
data with good coherence while the dashed lines interpolate through
areas of low coherence. The gold line shows transects extracted from
a 2013 one-day interferogram (Fig. S3), confirming that the current
subsidence is directly related to the ongoing magmatic activity.
Figure S3. One day interferogram of B´
ararbunga caldera and sur-
rounding ice-covered regions formed from CSK images acquired
on June 26, 2013 and June 27, 2013. No significant subsidence over
the caldera is observed, although LOS displacements due to ice
flow out of the east side of the caldera are consistent with the 2014
Figure S4. Time spans of interferograms used for estimating the
distribution of tensile opening along the dike interface. All inter-
ferogram time spans are referenced to Aug. 16, 2014, indicated by
the vertical dashed line. The yellow lines correspond to RS2 inter-
ferograms while the red lines correspond to CSK interferograms.
The orbital directions of the images used for the interferograms
are indicated as either ascending (ASC) or descending (DESC).
The alternating gray and white columns correspond to the temporal
subdomains used in the inversion.
Figure S5. Estimated distribution of opening along the dike inter-
face for the time span corresponding to the RS2 interferogram in
Fig. 1. The synthetic interferogram (see Fig. S6) is shown above the
fault, and the white dots correspond to earthquakes located along
the dike using the SIL network.
Figure S6. Synthetic ground deformation for the elastic model of
tensile opening along the dike interface. (A) RS2 interferogram from
Fig. 1and GPS displacements with 90% confidence error ellipses
used in inversion. (B) Synthetic ground displacements from the es-
timated model. The modeled LOS displacements include a bilinear
ramp to account for orbital errors and long-wavelength deforma-
tion. Ellipses represent 90% confidence posterior uncertainties in
the modeled GPS displacements. (C) Residual interferogram and
GPS displacements after subtracting the synthetic ground displace-
ments. Larger residuals near the surface trace north of Vatnaj¨
may indicate potential asymmetries due to dike dip not captured in
our vertical dike model. (D) Cumulative geodetic potency from the
dike model in Fig. S5 for Aug. 16–Sep. 20 vs. a histogram of the
depth distribution of earthquakes located along the dike.
Figure S7. (A) Samples for magma chamber depth and radius ob-
tained from an adaptive Metropolis algorithm. Red and blue circles
correspond to samples where the prior distribution for the chamber
depth has a maximum at 3 and 6 km, respectively. The black dashed
lines indicate lines of constant depth-to-radius ratio, h. (B) Samples
for magma chamber pressure (in units of the shear modulus μ)and
Figure S8. Modelling of the source magma chamber using geode-
tic measurements on ice-free areas. The chamber is modeled as a
horizontal circular sill at depth. (A) Unwrapped RS2 interferogram
formed from images acquired in an ascending orbit on Aug. 1,
2014 and Sep. 18, 2014, and GPS displacements with 90% confi-
dence error ellipses for VONC and HNIF after removing the esti-
mated signal from the rift zone; (B) unwrapped RS2 interferogram
from images acquired in a descending orbit on Aug. 27, 2014 and
Sep. 20, 2014; (C) modelled GPS displacements and LOS displace-
ments for the ascending interferogram for a chamber at a depth
of 8 km and a radius of 2.3 km (location and size indicated by
red dashed circle) or a depth of 4 km and radius of 1.1 km (blue
dashed circle); (D) modelled LOS displacements for the descending
interferogram; (E) residual between (A) and (C); and (F) residual
between (B) and (D). The estimated volume changes were 0.43 km3
and 0.14 km3for the ascending and descending interferograms,
Figure S9. Seismicity rates for earthquakes occurring within the
dike (top) and caldera (bottom). The number of earthquakes are
binned in latitude and time where 1-day intervals are used for the
time bins. The count is saturated at 15 earthquakes/day for visu-
alization clarity. Bins with less than two earthquakes are coloured
white. High seismicity rates within the dike for the first 10 days
are associated with the initial dike emplacement. While seismic-
ity along the dike has decreased with time, seismicity along the
northern rim of the caldera has increased.
Tab l e S 1 . Comparison of select earthquakes from the GFZ Pots-
dam and Global CMT (GCMT; Dziewonski et al. 1981; Ekstr¨
et al. 2012) catalogs. Focal mechanisms and moment magnitudes
are shown for the earthquakes from each catalog. We compute seis-
mic moment as M0=1/2ij Mij1/2. Both the mechanisms and
moment magnitudes are consistent between the catalogs.
Tab l e S 2 . Focal mechanisms and moment magnitudes for CLVD
events occurring near B´
ararbunga from 1976–1996. The fo-
cal mechanisms were formed using the moment tensor ele-
ments estimated by Nettles & Ekstr¨
om (1998). These prior
events exhibit opposite polarities from the current events (verti-
cal tensional vs. compressional axes). (
Please note: Oxford University Press is not responsible for the con-
tent or functionality of any supporting materials supplied by the
authors. Any queries (other than missing material) should be di-
rected to the corresponding author for the paper.
by guest on May 11, 2015 from
... They also built a moment tensor catalog for the M W > 5.0 events and inferred smaller events' seismic sources through a waveform similarity analysis. Their principal findings for the seismicity are: (1) the seismogenic zone extends to 12 km depth below the caldera, (2) the collapse was asymmetric with outward and inward dipping faults at the north and south sides of the caldera, respectively, (3) the caldera collapse was mainly aseismicas proposed by Riel et al. (2015) -where the geodetic moment (4 × 10 20 Nm) exceeded the seismic moment (5 × 10 18 Nm), and (4) the normal focal mechanisms have a dominant near-vertical P-axis and vertical compensated linear vector dipole (V-CLVD) components. In general, the Gudmundsson et al. (2016) model predicts the onset of the collapse in response to the reduced pressure support from the magma chamber to the caldera floor due to magma outflow; while the subsequent gradual collapse, magma outflow, and fissure eruption were driven by feedback between the pressure of the caldera floor on the magma reservoir and the outflow. ...
... Other Bárðarbunga caldera collapse seismic studies also revealed moment tensors with normal focal mechanisms and important V-CLVD components Riel et al., 2015) in addition to the near-real-time inversions provided by the Global Centroid Moment Tensor Project (GCMT; Dziewonski et al., 1981;Ekström et al., 2012) and the GEOFON program (Saul et al., 2011). ...
... Proposed models for explaining the underlying mechanisms of non-DC focal mechanisms fall into two main categories (Riel et al., 2015): ...
Between August 2014 and February 2015, a caldera collapse process in Bárðarbunga volcano accompanied a fissure eruption at the Holuhraun lava field after magma had propagated laterally 45 km from the Bárðarbunga magma chamber to the lava field. The Icelandic Meteorological Office (IMO) reported around 30,000 earthquakes at the caldera, including more than 70 earthquakes with moment magnitude MW ≥ 5.0. We built a moment tensor catalog of around 230 seismic events at the caldera with MW between 3.7 and 5.5 using a waveform inversion method. We identified five event families based on focal mechanisms, double couple (DC) and compensated linear vector dipole (CLVD) components, epicentral distribution relative to the caldera center, and time of occurrence. The main characteristics of these families are (1) DC to CLVD normal faulting events with steep (50°–70°) to near-vertical (>70°) dip angles on the northern side of the caldera; (2) DC to CLVD normal events with predominant steep (50°–70°) dip angles on the southern side of the caldera; (3) intra-caldera oblique strike-slip events; and (4) thrust events at the eastern side of the caldera with a horizontal tension axis compensated linear vector dipole (T-CLVD) component. Each family provided new insights about the Bárðarbunga collapse mechanism such as (1) a systematic increase in the CLVD component with magnitude, corroborating that the curvature in the ring fault is the likely cause for the vertical CLVD moment tensors observed for the largest events; (2) seismic evidence of near-vertical dip angles in agreement with the vertical faults observed in many calderas worldwide, and numerical and analog models of caldera collapse structures; (3) intra-caldera oblique and strike-slip events on the western side of the caldera occurred by asymmetric collapse; (4) thrust horizontal T-CLVD events observed at the last stage of the eruption, possibly related to a viscoelastic response.
... For small events, the curvature of the rupture area is negligible and therefore can be explained by a single double couple (DC) (Ágústsdóttir et al., 2019), on the other hand, for bigger ruptures, the curvature comes into play and a more complex model is needed to explain the observations. Moreover, MT solutions reported for these events (Riel et al., 2015;Gudmundsson et al., 2016) include an important Compensated Linear Vector Dipole (CLVD) component, which Ekström (1994) attributed to outward dipping ring-fault ruptures. Thus, these MT solutions are a good indicator that curved-ruptures are applicable. ...
... Previous results on the 2014 Bárðarbunga caldera collapse support the idea of an aseismic collapse (Riel et al., 2015), which implies creeping slip at the caldera rims or a tremor-like superposition of events forming a slow slip event. These processes are very likely to occur, therefore, there is always a discrepancy when comparing seismic and geodetic moment. ...
... Previous studies of the seismicity at Bárðarbunga showed tensional, vertical CLVD focal mechanisms supporting the conceptual model at stage 2 of caldera formation (Nettles and Ekström, 1998;Tkalčić et al., 2009). Riel et al. (2015); Gudmundsson et al. (2016); Ágústsdóttir et al. (2019) studied seismicity during the 2014-2015 collapse which was concentrated at both the north-northwest and southern parts of the caldera. In summary, the seismicity was interpreted as normal DC solutions for small events (M w >4.5), whereas for bigger events, non-DC component (CLVD and ISO) become dominant (Rodriguez Cardozo et al., 2018). ...
Previous studies have found discrepancies concerning the seismic radiation between planar and curved faults; moment tensor (MT) interpretations, seismic moment estimation and waveforms change dramatically when the rupture is not planar. Therefore, assuming a point source on a planar fault for earthquakes in volcanic environments can be an oversimplification that needs to be addressed if we observe some seismological clues. We study MT inversions for the biggest earthquakes during the 2014–2015 collapse of the Bárðarbunga caldera, which show non-double couple solutions, with vertical compression axis. We calculate synthetic seismograms for partial-ring ruptures using an ideal seismic network, and one emulating the existing monitoring network at Bárðarbunga. Observations using distal stations can return a better-constrained seismic moment, but they fail to characterise the dynamics involved. On the other hand, using proximal stations we obtain a reliable representation of the forces involved; however, the seismic moment is systematically overestimated due to the proximity to the curved source and the corresponding focusing effects. Finally, we correct the area of rupture due to fault shape to estimate the real cumulative seismic moment during the caldera collapse. The result shows a closer relationship between seismic and geodetic moment.
... In contrast, Bárðarbunga has erupted 26 times in the last 1100 years (Gudmundsson and Högnadóttir 2007) with the most recent eruption occurring in 2014-2015. This eruption, also known as Bárðarbunga-Holuhraun eruption, started with the lateral injection of melt from the Bárðarbunga magma chamber first towards the SE and later along a NE-SW oriented dike, leading to the collapse of the magma chamber roof (Gudmundsson et al. 2014;Riel et al. 2015;Sigmundsson et al. 2015;Gudmundsson et al. 2016;Á gústsdóttir et al. 2016;Ruch et al. 2016). Within a few days the dike propagated outside the Vatnajökull glacier and reached the surface in the area of Holuhraun. ...
... One distinct difference between the seismicity observed during the 1996 Gjálp and the 2014-2015 Bárðarbunga-Holuhraun eruption was the opposite sense of motion for the events around the Bárðarbunga caldera rim. During the 2014-2015 eruption earthquakes clustered in the NE and SE parts of the caldera exhibited normal faulting with large CLVD components (Riel et al. 2015). On the contrary, during the 1996 eruption events clustered in the NW and SW parts of the caldera and their moment tensor solutions indicate reverse faulting (cf. ...
Full-text available
This study performs a reanalysis of the seismicity recorded during the 1996 Gjálp eruption that occurred at NW Vatnajökull, Iceland. The seismicity was recorded by the temporary HOTSPOT network consisting of 30 three-component broadband stations. In total 301 events were identified between 29 September and 12 October and their phases were manually picked. A velocity model was estimated from P-phase travel times by using VELEST. Events were first located using the algorithm NON-LINLOC in order to obtain absolute locations. Precise relative locations were obtained with HYPODD by utilizing catalog and cross-correlation differential travel times. Results show that events clustered first along the SW rim of the Bárðarbunga caldera and later along the Gjálp fissure, with most hypocentral depths located between 3 and 8 km. Waveforms of the 10 largest events that followed the Bárðarbunga earthquake were inverted in order to obtain moment tensors. For all events we found that the deviatoric moment tensor fits the data better than pure double-couple or full moment tensor solutions. Events along the Bárðarbunga caldera exhibited reverse focal mechanisms, while those at the Gjálp fissure exhibited mostly strike-slip faulting. Seismic velocity variations calculated using ambient noise interferometry, point to the possibility that a small subglacial eruption occurred at Bárðarbunga before the main earthquake. This removed melt from the magma chamber causing its roof to collapse, and also resulted in the lateral migration of magma towards the Gjálp fissure. The 2014-2015 Bárðarbunga-Holuhraun eruption shares common characteristics with the 1996 Gjálp eruption, although the size of the latter was much smaller.
... This is supported by the lack of thick tephra deposits on their rims. Geophysical observations on some recent examples of caldera formation, at Fernandina, Galapagos, in 1968 (Simkin andHoward, 1970), Myakejima, Japan, in 2000 (Geshi et al., 2002), Piton de la Fournaise, La Réunion Island, in 2007 (Michon et al., 2007Peltier et al., 2009;Staudacher et al., 2009), Bárdarbunga, Iceland (Sigmundsson et al., 2015;Riel et al., 2015;Gudmundsson et al., 2016) and Kilauea, Hawaii, in 2018(Anderson et al., 2019Neal et al., 2019), favor an incremental collapse of the rock column into the magma chamber . In many cases, collapse of a new caldera or within a caldera accompanies voluminous flank eruptions, lateral magma intrusion, or long-lived eruptive activity (Sigmundsson, Hot Spots 2019). ...
The geomorphology of hot spot volcanism and specifically oceanic hot spot volcanism is best explained and understood by examining volcanic edifices development above hot spots using nested scales. This includes starting from a large overview to progressively more and more detailed scales. The construction of hot spot volcanic chains and the identification of morphologies from the deep-sea floor to volcano summits includes considering the mechanisms of construction, evolution, and dismantling of volcanic edifices utilizing well-known examples and models. Limitations and imperfections within models for hot spot volcanism mainly relate to the specific context and scale, and to magmatology. At the oceanic scale, the models are influenced by the geodynamic context (plate motion and proximity of a subduction or an accretion zone). At the volcanic edifice scale, the models are constrained by the history of the volcano development, for example, cones superimpositions, proximity of fractures, zone of weakness, stress field, major mass wasting, and rift zones.
... Assuming a magnitude of 5.0 for a seismic shock related to the caldera collapse at a distance of 5 km, a maximum horizontal acceleration coefficient (PHAC) of 0.29 would be generated . Comparing these results with other observational data from some recent calderas (e.g., Abe, 1992;Alvizuri et al., 2020;Filson et al., 1973;Riel et al., 2015), a Mw magnitude of 5-7.2 is required for seismic shocks to have triggered the Icod landslide . ...
Full-text available
Extreme geohazards (volcanic eruptions, earthquakes, landslides, and tsunamis) have the potential to inflict cascading effects whose associated risks are difficult to predict and prepare for. Thus, these events are generally not taken into account in hazard assessment. Anticipating the occurrence of such extreme events is thus key if our life‐styles are to remain safe and sustainable. Volcanic islands are often the source of complex successions of disastrous events, as is evident from any examination, for instance, of the geological record of regions such as Hawaii, the Canary Islands, Reunion and Indonesia. The island of Tenerife in the Canary Archipelago is an excellent example of where cascading extreme hazards have occurred several times in the past and could occur again in the future. A cascading sequence involving a caldera‐forming eruption, high‐magnitude seismicity, mega‐landslides and tsunamis occurred at least twice during the construction of this island. In order to understand the possible consequences of such processes if they were to reoccur, we simulated the extent and potential impact of a multiple, extreme geohazard episode similar to the last recorded one that took place on the island of Tenerife around 180 ka. The implications of such a disastrous succession of events are analyzed at local, regional and global scales, and the results obtained are discussed within the framework of disaster risk‐reduction policies.
... The openings are formed locally if the meltwater finally reaches the glacier surface and continues flowing supraglacially [5,7]. The 2014-2015 rifting episode between Bárðarbunga and Askja in Iceland resulted in the formation of ice cauldrons (on Dyngjujökull glacier) along the path of a 45-km long subglacial dyke, and included the collapse of the Bárðarbunga caldera [15,[22][23][24]. These cauldrons lie within a sector of ice 430-670 m thick [25]. ...
Full-text available
Globally, about 250 Holocene volcanoes are either glacier-clad or have glaciers in close proximity. Interactions between volcanoes and glaciers are therefore common, and some of the most deadly (e.g., Nevado del Ruiz, 1985) and most costly (e.g., Eyjafjallajökull, 2010) eruptions of recent years were associated with glaciovolcanism. An improved understanding of volcano-glacier interactions is therefore of both global scientific and societal importance. This study investigates the potential of using optical satellite images to detect volcanic impacts on glaciers, with a view to utilise detected changes in glacier surface morphology to improve glacier-clad volcano monitoring and eruption forecasting. Roughly 1400 optical satellite images are investigated from key, well-documented eruptions around the globe during the satellite remote sensing era (i.e., 1972 to present). The most common observable volcanic impact on glacier morphology (for both thick and thin ice-masses) is the formation of ice cauldrons and openings, often associated with concentric crevassing. Other observable volcanic impacts include ice bulging and fracturing due to subglacial dome growth; localized crevassing adjacent to supraglacial lava flows; widespread glacier crevassing, presumably , due to meltwater-triggered glacier acceleration and advance. The main limitation of using optical satellite images to investigate changes in glacier morphology is the availability of cloud-and eruption-plume-free scenes of sufficient spatial-and temporal resolution. Therefore, for optimal monitoring and eruption prediction at glacier-clad volcanoes, optical satellite images are best used in combination with other sources, including SAR satellite data, aerial images, ground-based observations and satellite-derived products (e.g., DEMs).
... Vertical-CLVD earthquakes cannot be explained by shear rupture on a planar fault, indicating that their anomalous mechanisms are associated with complex source structures or magmatic processes. For vertical-CLVD earthquakes at volcanoes, several models have been proposed, including ring-faulting (e.g., Ekström, 1994), rapid water-magma interaction initiated by magma intrusion into shallow crust (Kanamori et al., 1993), and opening or closing of a horizontal crack (e.g., Fukao et al., 2018;Riel et al., 2015). ...
Full-text available
Large earthquakes (M_w > 5) with moment tensors (MTs) dominated by a vertical compensated-linear-vector-dipole (vertical-CLVD) component are often generated by dip slip along a curved ring-fault system at active volcanoes. However, relating their MTs to ring-fault parameters has proved difficult. The objective of this study is to find a robust way of estimating ring-fault parameters based on their MT solutions obtained from long-period seismic records. We first model the MTs of idealized ring-faulting and show that an MT component representing the vertical dip-slip mechanism is indeterminate from long-period seismic waves owing to a shallow source depth, whereas the other MT components representing the vertical-CLVD and vertical strike-slip mechanisms are resolvable. We then propose a new method for estimating the arc angle and orientation of ring-faulting using the two resolvable MT components. For validation, we study a vertical-CLVD earthquake that occurred during the 2005 volcanic activity at the Sierra Negra caldera, Galápagos Islands. The resolvable MT components are stably determined from long-period seismic waves, and our estimation of the ring-fault parameters is consistent with the ring-fault geometry identified by previous geodetic studies and field surveys. We also estimate ring-fault parameters of two earthquakes that took place during the 2018 activity at the caldera, revealing significant differences between the two earthquakes in terms of slip direction and location. These results show the usefulness of our method for estimating ring-fault parameters of vertical-CLVD earthquakes, enabling us to examine the kinematics and structures below active volcanoes with ring faults that are distributed globally.
... The caldera collapse and the formation of the Holuhraun lava field during the eruption were well monitored (e.g. Dumont et al., 2018;Gudmundsson et al., 2016;Pedersen et al., 2017;Riel et al., 2015;Sigmundsson et al., 2015). A continuous GNSS station installed in the center of the caldera on the ice surface, repeated radar altimeter survey, and maps from optical satellite images revealed the temporal evolution of subsidence within the caldera (Gudmundsson et al., 2016). ...
Full-text available
Improvement of our understanding of the role of ground deformation due to viscoelastic relaxation following eruptions is important, as the generated signals can resemble renewed magma inflow. We study post‐eruptive unrest at the subglacial Bárðarbunga volcano, Iceland, after a caldera collapse and major magma drainage in 2014–2015. Elevated seismicity began about 6 months after the eruption ended, including nine Mlw > 4.5 earthquakes. Global Navigation Satellite System and Sentinel‐1 Interferometric Synthetic Aperture Radar geodesy are applied to evaluate post‐eruptive ground deformation from 2015 to 2018. Horizontal velocities locally exceed 10 cm/year and rapidly decay with distance away from the caldera. We explore two end‐member models and their combination to explain the post‐eruptive deformation field: 1) viscoelastic relaxation caused by the co‐eruptive caldera collapse and magma withdrawal, and 2) renewed magma inflow. We find parameter combinations for each model that explain the observed ground deformation. The purely viscoelastic relaxation model, consisting of a half‐space composed of a 7‐km thick elastic layer on top of a viscoelastic layer with a viscosity of 3.0 × 10¹⁸ Pa s reproduces broadly the observations. A simple magma inflow model consisting of a single point source with an inflow rate of 1 × 10⁷ m³/year at 0.7 km depth broadly fits the observations, but may be unrealistic. A more elaborate model of magma inflow into a 10‐km deep sill combined with slip on the caldera ring fault explains the observations well. Our results suggest that the co‐eruptive deformation field is likely influenced by viscoelastic relaxation, renewed magma inflow, or a combination of both processes.
... Because of the difficulty in capturing this process as it unfolds, much of what is known about caldera collapse has been learned from a combination of geological investigation of older, sometimes exhumed examples (e.g., Lipman, 1976), and from laboratory studies of analog materials (e.g., Roche et al., 2000). Field observations of eruptions of Fernandina (1968), Pinatubo (1991), Miyakejima (2000), Piton de la Fournaise (2007), and Bardarbunga (2014) volcanoes have all provided important constraints on caldera collapse processes (Kumagai et al., 2001;Michon et al., 2009;Mori et al., 1996;Riel et al., 2015;Stix & Kobayashi, 2008), yet none of these eruptions were captured by continuously recorded near-field instrumentation approaching the quality available for the 2018 eruption of Kīlauea. ...
Full-text available
The 2018 Kīlauea eruption and caldera collapse generated intense cycles of seismicity tied to repeated large seismic (Mw ~ 5) collapse events associated with magma withdrawal from beneath the summit. To gain insight into the underlying dynamics and aid eruption response, we applied waveform‐based earthquake detection and double‐difference location as the eruption unfolded. Here, we augment these rapid results by grouping events based on patterns of correlation‐derived phase polarities across the network. From April 29 to August 6, bracketing the eruption, we used ~2,800 events cataloged by the Hawaiian Volcano Observatory to detect and precisely locate 44,000+ earthquakes. Resulting hypocentroids resolve complex, yet coherent structures, concentrated at shallow depths east of Halema'uma'u crater, beneath the eventual eastern perimeter of surface collapse. Based on a preponderance of dilatational P wave first motions and similarities with previously inferred dike structures, we hypothesize that failure was dominated by coupled shear and crack closure.
Full-text available
In this study, we propose a new method to determine full moment tensor solution for induced seismicity. This method generalizes the full waveform matching algorithm we have developed to determine the double-couple (DC) focal mechanism based on the neighbourhood algorithm. One major difference between the new method and the former one is that we adopt a new misfit function to constrain the candidate moment tensor solutions with respect to a reference DC solution in addition to other misfit terms characterizing the waveform matching. Through synthetic tests using a real passive seismic survey geometry, the results show the new constraint can help better recover the DC components of inverted moment tensors. We further investigate how errors in the velocity model and source location affect the moment tensor solution. The synthetic test results indicate that the constrained inversion is robust in recovering both the DC and non-DC components. We also. test the proposed method on several real induced events in an oil/gas field in Oman using the same observation system as synthetic tests. While it is found that the full moment tensor. solutions without using the DC constraints have much larger non-DC components than solutions with the DC constraints, both solutions are able to fit the observed waveforms at similar levels. The synthetic and real test results suggest the proposed DC constrained inversion method can reliably retrieve full moment tensor solutions for the induced seismicity.
Full-text available
A 45-km-long regional dike was emplaced over a period of 2 weeks in August 2014 at the boundary between the East and North Volcanic Zones in Iceland. This is the first regional dike emplacement in Iceland monitored with modern geophysical networks, the importance of which relates to regional dikes feeding most of the large fissure (e.g., Eldgja 934 and Laki 1783) and lava shield (e.g. early Holocene Skjaldbreidur and Trölladyngja) eruptions. During this time, the dike generated some 17,000 earthquakes, more than produced in Iceland as a whole over a normal year. The dike initiated close to the Bardarbunga Volcano but gradually extended to the northeast until it crossed the boundary between the East Volcanic Zone (EVZ) and the North Volcanic Zone (NVZ). We infer that the strike of the dike changes abruptly at a point, from about N45ºE (coinciding with the trend of the EVZ) to N15ºE (coinciding with the trend of the NVZ). This change in strike occurs at latitude 64.7º, exactly the same latitude at which about 10 Ma dikes in East Iceland change strike in a similar way. This suggests that the change in the regional stress field from the southern to the northern part of Iceland has been maintained at this latitude for 10 million years. Analytical and numerical models indicate that the dike-induced stress field results in stress concentration around faults and particularly shallow magma chambers and calderas in its vicinity, such as Tungnafellsjökull, Kverkfjöll, and Askja. In particular, the dike has induced high compressive, shear, and tensile stresses at the location of the Bardarbunga shallow chamber and (caldera) ring-fault where numerous earthquakes occurred during the dike emplacement, many of which have exceeded M5 (the largest M5.7). The first segment of the dike induced high tensile stresses in the nearby part of the Bardarbunga magma chamber/ring-fault resulting in radially outward injection of a dike from the chamber at a high angle to the strike of the regional dike. The location of maximum stress at Bardarbunga fluctuates along the chamber/ring-fault boundary in harmony with dike size and/or pressure changes and encourages ring-dike formation and associated magma flow within the chamber. Caldera collapse and/or eruption in some of these volcanoes is possible, most likely in Bardarbunga, but depends largely on the future development of the regional dike.
Full-text available
A volcanic earthquake with M w 5:6 occurred beneath the Bárdarbunga caldera in Iceland on 29 September 1996. This earthquake is one of a decade-long sequence of M 5‡ events at Bárdarbunga with non-double-couple mechanisms in the Global Centroid Moment Tensor catalog. Fortunately, it was recorded well by the regional-scale Iceland Hotspot Project seismic experiment. We investigated the event with a complete moment tensor inversion method using regional long-period seismic waveforms and a composite structural model. The moment tensor inversion using data from stations of the Iceland Hotspot Project yields a non-double-couple solution with a 67% vertically oriented compensated linear vector dipole component, a 32% double-couple component, and a statistically insignificant (2%) volumetric (isotropic) con-traction. This indicates the absence of a net volumetric component, which is puzzling in the case of a large volcanic earthquake that apparently is not explained by shear slip on a planar fault. A possible volcanic mechanism that can produce an earthquake without a volumetric component involves two offset sources with similar but opposite volume changes. We show that although such a model cannot be ruled out, the circumstances under which it could happen are rare.
Full-text available
Along mid-ocean ridges the extending crust is segmented on length scales of 10-1,000km. Where rift segments are offset from one another, motion between segments is accommodated by transform faults that are oriented orthogonally to the main rift axis. Where segments overlap, non-transform offsets with a variety of geometries accommodate shear motions. Here we use micro-seismic data to analyse the geometries of faults at two overlapping rift segments exposed on land in north Iceland. Between the rift segments, we identify a series of faults that are aligned sub-parallel to the orientation of the main rift. These faults slip through left-lateral strike-slip motion. Yet, movement between the overlapping rift segments is through right-lateral motion. Together, these motions induce a clockwise rotation of the faults and intervening crustal blocks in a motion that is consistent with a bookshelf-faulting mechanism, named after its resemblance to a tilting row of books on a shelf. The faults probably reactivated existing crustal weaknesses, such as dyke intrusions, that were originally oriented parallel to the main rift and have since rotated about 15° clockwise. Reactivation of pre-existing, rift-parallel weaknesses contrasts with typical mid-ocean ridge transform faults and is an important illustration of a non-transform offset accommodating shear motion between overlapping rift segments.
Full-text available
Understanding the elastic behavior of basalt is important to seismically monitor volcanoes, subsea basalts, and carbon sequestration in basalt. We estimate the elastic properties of basalt samples from the Snake River Plain, Idaho, at ultrasonic (0.8 MHz) and seismic (2-300 Hz) frequencies. To test the sensitivity of seismic waves to the fluid content in the pore structure, measurements are performed at three saturation conditions: saturated with liquid CO2, water, and dry. When CO2 replaces water, the P-wave velocity drops, on average, by 10%. Vesicles and cracks, observed in the rock microstructure, control the relaxation of pore-fluid pressures in the rock as a wave propagates. The bulk and shear moduli of basalts saturated with liquid CO2 are not frequency dependent, suggesting that fluid pore pressures are in equilibrium between 2 Hz and 0.8 MHz. However, when samples are water saturated, the bulk modulus of the rock is frequency dependent. Modeling with Gassmann's equations predicts the measured saturated rock bulk modulus for all fluids for frequencies below 20 Hz but underpredicts the water-saturated basalt bulk modulus for frequencies greater than 20 Hz. The most likely reason is that the pore-fluid pressures are unrelaxed. Instead, the ultrasonic frequency rock moduli are modeled with high-frequency elastic theories of squirt flow and Kuster-Toksöz (KT). Although KT's model is based on idealized pore shapes, a combination of spheres (vesicles) and penny-shaped cracks (fractures) interpreted and quantified from petrographical data predicts the ultrasonic dry and saturated rock moduli for the measured basalts.
Satellite-based interferometric synthetic aperture radar (InSAR) provides a synoptic high spatial resolution perspective of Earth's deforming surface, permitting one to view large areas quickly and efficiently. We review basic InSAR theory for geodetic applications and attempt to provide an overview of what processing and analysis schemes are currently used and a glimpse of what the future may hold. As part of this discussion, we present a biased view of what constitutes best practices for use of InSAR observations in geodetic modeling. Finally, we provide a basic primer on the ties between different mission design parameters and their relationship to the character of the resulting observations.
Crust at many divergent plate boundaries forms primarily by the injection of vertical sheet-like dykes, some tens of kilometres long1. Previous models of rifting events indicate either lateral dyke growth away from a feeding source, with propagation rates decreasing as the dyke lengthens2, 3, 4, or magma flowing vertically into dykes from an underlying source5, 6, with the role of topography on the evolution of lateral dykes not clear. Here we show how a recent segmented dyke intrusion in the Bárðarbunga volcanic system grew laterally for more than 45 kilometres at a variable rate, with topography influencing the direction of propagation. Barriers at the ends of each segment were overcome by the build-up of pressure in the dyke end; then a new segment formed and dyke lengthening temporarily peaked. The dyke evolution, which occurred primarily over 14 days, was revealed by propagating seismicity, ground deformation mapped by Global Positioning System (GPS), interferometric analysis of satellite radar images (InSAR), and graben formation. The strike of the dyke segments varies from an initially radial direction away from the Bárðarbunga caldera, towards alignment with that expected from regional stress at the distal end. A model minimizing the combined strain and gravitational potential energy explains the propagation path. Dyke opening and seismicity focused at the most distal segment at any given time, and were simultaneous with magma source deflation and slow collapse at the Bárðarbunga caldera, accompanied by a series of magnitude M > 5 earthquakes. Dyke growth was slowed down by an effusive fissure eruption near the end of the dyke. Lateral dyke growth with segment barrier breaking by pressure build-up in the dyke distal end explains how focused upwelling of magma under central volcanoes is effectively redistributed over long distances to create new upper crust at divergent plate boundaries.
Most ring faults of collapse calderas are primarily shear fractures the initiation and development of which depends on the state of stress in the host rock. The state of stress in a volcano is controlled by the loading conditions, such as magma-chamber geometry and pressure, but also by the mechanical properties of its rock units and structures (such as existing contacts, faults, and joints). Ring-fault formation is thus essentially a problem in rock physics. Field observations show that some ring faults are dip-slip, whereas others are partly faults (shear fractures) and partly ring dykes (extension fractures). Although slip on existing ring faults is much more common in basaltic edifices (shield volcanoes) than in true composite volcanoes, in both types of volcanoes most caldera unrest periods do not result in ring-fault slip. Here I present new conceptual and numerical models of caldera formation in volcanoes with shallow spherical (circular) or sill-like (oblate ellipsoidal) magma chambers. In the layered models, the host rock above the chamber is composed of 30 comparatively thin layers with stiffnesses (Young's moduli) alternating between 1 GPa to 100 GPa. The chamber itself is located in a single, thick layer. The crustal segment hosting the chamber is either 20 km or 40 km wide but has a constant thickness of 20 km. The loading conditions considered are: (1) a crustal segment subject to 5 MPa tension; (2) crustal segment subject to excess magmatic pressure of 10 MPa at the bottom (doming of the volcanic field containing the chamber); (3) a combination of tension and doming; and (4) chamber subject to underpressure (negative excess pressure) of 5 MPa. The main results are as follows: (1) Excess pressure and underpressure in a chamber normally favour dyke injection rather than ring-fault formation. (2) For doming or tension, a spherical magma chamber favours dyke injection except when the layer hosting the chamber is very soft (10 GPa) or one with recent dyke injections, in which case the surface stress field favours ring-fault formation. (3) For a sill-like chamber in a 20-km wide crustal segment, a ring-fault can be generated by either tension or tension and doming; for a 40-km wide segment, doming alone is sufficient to generate a ring fault. (4) Since individual layers in a volcano may develop different local stresses, stress-field homogenisation through all the layers between the chamber and the surface is a necessary condition for ring-fault formation. (5) Because the mechanical properties of the layers that constitute basaltic edifices are more uniform than those that constitute true composite volcanoes, it follows that stress-field homogenisation, and thus ring-fault formation or slip, is more commonly reached in basaltic edifices than in composite volcanoes. (6) Both for basaltic edifices and composite volcanoes, the stress fields most likely to initiate ring faults are those generated around sill-like chambers subject to tension, doming, or both.
Many volcanic earthquakes large enough to be detected globally have anomalous focal mechanisms and frequency content. In a previous study, we examined the relationship between active volcanism and the occurrence of a specific type of shallow, non-double-couple earthquake. We identified 101 earthquakes with vertical compensated-linear-vector-dipole (vertical-CLVD) focal mechanisms that took place near active volcanoes between 1976 and 2009. The majority of these earthquakes, which have magnitudes 4.3 ≤ MW ≤ 5.8, are associated with documented episodes of volcanic unrest. Here we further characterize vertical-CLVD earthquakes and explore possible physical mechanisms. Through teleseismic body-wave analysis and examination of the frequency content of vertical-CLVD earthquakes, we demonstrate that these events have longer source durations than tectonic earthquakes of similar magnitude. We examine the covariance matrix for one of the best-recorded earthquakes and confirm that the isotropic and pure vertical-CLVD components of the moment tensor cannot be independently resolved using our long-period seismic data set. Allowing for this trade-off, we evaluate several physical mechanisms that may produce earthquakes with deviatoric vertical-CLVD moment tensors. We find that physical mechanisms related to fluid flow and volumetric changes are incompatible with seismological, geological, and geodetic observations of vertical-CLVD earthquakes. However, ring-faulting mechanisms explain many characteristics of vertical-CLVD earthquakes, including their seismic radiation patterns, source durations, association with volcanoes in specific geodynamic environments, and the timing of the earthquakes relative to volcanic activity.