Jökull - 01.01.2010, Blaðsíða 51

Upptyppingar seismic swarms
brittle-ductile boundary of 15–18 km, along a pre-
sumed dyke that dipped at 50◦.
Deep seismic activity attributed to melt movement
is also observed beneath active volcanic centres in Ice-
land, as well as elsewhere in the world such as Mt. Ki-
lauea (Wright and Klein, 2006) and Japan (Hasegawa
et al., 1991), but the hypocentres typically assume
conduit-like as opposed to planar distributions.
Near Mt. Upptyppingar, hundreds of lower crustal
earthquakes have been observed beneath Askja vol-
cano since the ASN began monitoring the area in 2005
(Soosalu et al., 2010). The seismicity outlines three
separate vertical conduit structures that extend from c.
10 km down to the crust-mantle boundary. The most
recent eruption at Askja volcano occurred in 1961.
Mt. Eyjafjallajökull in southwest Iceland also ex-
hibited deep seismicity in the mid-1990s that ex-
tended from the crust-mantle boundary through to
the surface (Hjaltadóttir et al., 2009). Seismic activ-
ity persisted at shallower depths with sporadic deep
events for over a decade, eventually leading to an
eruption in 2010 that halted European air traffic for
several days (Sigmundsson et al., 2010; Hjaltadóttir
et al., 2009; Pedersen and Sigmundsson, 2004, 2006).
GPS and InSAR modelling of surface deformation
at Mt. Upptyppingar in 2007–2008 constrain the vol-
ume of injected melt to be ∼0.040–0.047 km3, corre-
sponding to the inflation of a ∼0.1–1 m thick, south-
ward dipping dyke at depths of 10–18 km (Hooper
et al., 2009). For comparison, the inferred volume
of the pre-eruptive melt intrusion beneath Eyjafjalla-
jökull was of the same magnitude (Sigmundsson et
al., 2010).
Seismicity that clearly defines a planar structure,
as observed beneath Mt. Upptyppingar, presents an
ideal opportunity to evaluate the effects of different
processing techniques, network size and geometries,
and phase picking accuracy by comparing hypocen-
tral location precision. This form of analysis has been
applied successfully in previous studies that, for ex-
ample, demonstrate the benefits of relative relocation
techniques (Waldhauser and Ellsworth, 2000; Slunga
et al., 1995).
The precision of seismic hypocentre locations af-
fects how the seismicity is interpreted. Outstanding
questions in crustal formation, such as the extent of
host rock deformation caused by an active dyke in-
trusion in visco-elastic crust, require extremely high
hypocentral precision.
ASN PROCESSING TECHNIQUES
Data selection is primarily limited by the deploy-
ment dates of the ASN (i.e., 6 July–22 August 2007).
Within this date range, we further restrict our study
period to 6–24 July, during which the most inten-
sive and dynamic bursts of seismic activity occurred,
including several earthquakes that exceeded Ml 2.0
and seismic propagation rates that reached as high
as 0.05 m s−1. Moreover, signal to noise ratios
were higher on average during this period than in late
July and August. The study period comprises 547
events that are drawn from a SIL catalogue of over
9000 earthquakes observed beneath Mt. Upptypping-
ar during 2007–2008. We then manually filtered the
547 events based on signal-to-noise ratio by inspect-
ing waveforms for clarity of phase onsets. The final
dataset consists of 288 high-quality events.
Processing of seismic data from the ASN is per-
formed in multiple steps. Firstly, events are lo-
cated using the Coalescence Microseismic Mapping
(CMM) software developed by Drew (2010). The SIL
event catalogue is provided as input to CMM, which
searches the continuously recorded data for phase on-
sets near each catalogue event time through a Short
Term Average to Long Term Average ratio (STA/LTA)
(Drew et al., 2005). For a given search volume of dis-
crete grid spacing and a specified velocity model,
a look-up table is produced by forward-modelling
travel times from each grid node to each receiver. The
look-up table is then used to migrate seismic energy
from both P-wave (vertical component) and S-wave
(horizontal components) onsets at each station into
the subsurface. Finally, a coalescence function is
used to determine the subsurface location at which
the seismic energy is focussed, yielding spatial and
temporal information about the imaged seismic event.
Any mis-identified onsets (e.g., from noise bursts) are
smeared out over the migrated volume and thus do not
contribute to the final CMM locations. Here we have
used a grid spacing of 300 m; however, by virtue of
JÖKULL No. 60 51