50
ESC2010 6-10 September 2010, Montpellier, France - Keynotes
Spatio-temporal evolution of volcano seismicity: laboratory simulations
The measurement and analysis of seismicity in active volcanic areas is the main tool used in volcanic hazard mitiga-
tion. However, seismic monitoring technologies have met with only mixed success in forecasting eruptive episodes.
This arises because, although our understanding of the spatio-temporal processes generating the observed seismic
signals has increased substantially with the wide scale adoption of new and improving technologies, such as modern
broadband seismology and GPS, there is still no universally accepted quantitative physical model for determining
whether or not a sequence of precursory phenomena will end in an eruption, or for forecasting the time or the type
of eruption. Hence, most forecasting strategies rely on long-term observation and monitoring of the edifi ce, rather
than directly assessing the fundamental micromechanics of the rock/fl uid interaction processes involved.
It is therefore our view that, in order to move forward, we need to study these coupled rock-fl uid-mechanical proces-
ses in detail under the controlled conditions available in laboratory-scale simulations. Specifi cally, we aim to investi-
gate the micromechanical original of the three key types of reported volcano seismicity: (1) high frequency volcano
tectonic (VT) earthquakes, generated by fracture and faulting; (2) low frequency (LF) earthquakes, hypothesised to
be generated by rapid fl uid fl ow though fracture-damage zones; and (3) seismicity that exhibits features of both VT
seismicity and LF tremor, known as hybrid (HY) events.
Here, we report results from a suite of laboratory experiments in which dry and water-saturated samples of basalt
from Mount Etna volcano (50 mm in diameter by 125 mm in length, with a centrally pre-drilled 3mm diameter con-
duit) were deformed and fractured under an effective confi ning pressure representative of conditions at depth under
a volcanic edifi ce (40 MPa). Particular attention was paid to the formation of a fracture and damage zone that would
allow us to stimulate the types of coupled fl uid-mechanical interactions thought to be responsible for generating the
different seismic signals recorded from deforming volcanic edifi ces preceding eruption events. Experiments were
conducted in two phases. In phase 1, samples were deformed at a constant strain rate of 4 x 10
-6
s
-1
until brittle failure
occurred. This resulted in the creation of a localized shear fault and an associated crack damage zone. In phase 2,
the fl uid stored in the sample (including both pre-existing microcracks and the new fault damage zone) was rapidly
decompressed from the top of the sample, stimulating rapid fl uid fl ow out of the sample via the damage zone and
the pre-drilled conduit. The output of microseismicity (acoustic emission) was recorded continuously throughout both
phases using an array of up to 16 transducers. In addition to recording waveforms, the array was used to locate indi-
vidual acoustic emission (AE) events using a downhill simplex routine and a triaxial velocity model. We also computed
source characteristics of the located events via relative amplitude moment tensor analysis.
Following an initially low rate of AE as pre-existing cracks were closed in phase 1, we observed an exponentially
increasing rate as differential stress was increased and new, dilatant cracks nucleated and propagated. Of the lo-
cated events, the majority were located within the damage zone of the shear fault, or its conjugate (Fig. 1A). They
exhibited dominantly double-couple (DC) source characteristics, similar to VT events, and entirely as expected for
the faulting phase of the experiment. We verifi ed that the presence of the central conduit did not signifi cantly affect
the mechanics of deformation and failure by conducting identical experiments without a conduit, and obtaining es-
sentially identical results.
In phase 2, we rapidly decompressed the pressurized pore fl uid by means of a valve at the top of the apparatus. The
decompression was accompanied by a swarm of AE events that were again located primarily within or near the fault
damage zone generated during stage 1 (Fig. 1B). However, In contrast to those from phase 1 of the experiment, the
source characteristics of the decompression-related events exhibited low components of shear, but a high volumetric
component (Fig. 1B). We therefore postulate that the rapid fl ow of the pore fl uid through the tortuous fracture da-
mage zone created conditions conducive to the generation of a swarm of AE events analogous to LF events recorded
at active volcanoes. Source mechanisms involving high levels of volumetric change have been widely reported in
volcanic regions, and in areas of tectonic subduction and fault overpressure; all of which have been linked to fl uid
movement.
In order to investigate the microstructural origin of located AE events, we analyzed backscattered scanning electron
microscopy images of the deformed and decompressed sample. These showed a complex damage zone formed in the
lower half of the sample, dominated by two major, conjugate faults (as seen in Fig. 1). During decompression, events
were located within this damage zone. The detailed locations of AE clusters imply that the fl uids producing these
events were following highly tortuous pathways, with many pinch-outs and undulating features. Such geometries have
long been postulated to be responsible for tremor-type events. SEM observations showed that many of the cracks
were fi lled with broken and comminuted rock, as also reported from fi eld observations of fractured magma. Taken
together, these observations suggest that LF events in volcanic areas are generated when hydrothermal fl uids (water,
Docsity.com
