Jökull - 01.01.2011, Page 36

J. Tarasewicz et al.
ond, we consider the effect of using different velocity
models to invert arrival time data.
Variation in network geometry
It is well established that, in order to place good con-
straints on earthquake hypocentres, a network ide-
ally needs to have both good azimuthal coverage of
the source and one or more stations close to the epi-
centre (e.g., Bondár et al., 2004; Bai et al., 2006;
Martens et al., 2010). Proximal stations are espe-
cially important for constraining the depth of earth-
quake sources. For ice-capped volcanoes such as
Eyjafjallajökull, seismicity occurring directly beneath
the glacier is rarely recorded by such an ideal network
configuration, given the difficulties of seismic moni-
toring on glaciers.
Seismicity in Iceland is routinely monitored by the
Icelandic Meteorological Office (IMO), which oper-
ates a permanent national network of seismometers.
In early March 2010, the IMO network included three
stations within 15 km of Eyjafjallajökull and another
four within 50 km. Two weeks prior to the initial fis-
sure eruption, the Institute of Earth Sciences, Univer-
sity of Iceland deployed an additional six seismome-
ter stations around the base of Eyjafjallajökull (Figure
1). These temporary stations were all within 4–15 km
of the epicentres of the main seismic activity during
early–mid March 2010. Data from the more proximal
temporary stations were combined with data from the
IMO stations to obtain the hypocentre locations re-
ported by Tarasewicz et al. (in press).
Variation in assumed velocity model
The velocity structure beneath Eyjafjallajökull has not
been constrained directly by tomographic studies or
refraction profiles that transect the edifice. Based on
data from other similar volcanoes, Eyjafjallajökull
is likely to have marked lateral variation in velocity
structure, at least at shallow levels. There is there-
fore uncertainty regarding what velocity structure it
is most appropriate to assume in calculating hypocen-
tral locations. Tarasewicz et al. (in press) use a 1-
D velocity model (VM1, see Figure 2e) based on the
eastern end of the SIST refraction profile (Bjarnason
et al., 1993) and the northern part of the Katla refrac-
tion profile (Gudmundsson et al., 1994). The SIST
profile crossed the South Iceland Seismic Zone, with
its easternmost shotpoint in the Gígjökull lake, at the
central-northern margin of Eyjafjallajökull. The Katla
profile lay NNW-SSE across Mýrdalsjökull, crossing
the western Katla caldera. Both of these profiles re-
port similar velocity gradients and increased upper
crustal thickness in the vicinity of Eyjafjallajökull. A
marked increase in upper crustal thickness is observed
going east across the South-Iceland Seismic Zone into
the Eastern Volcanic Zone (EVZ) (Bjarnason et al.,
1993; Pálmason, 1971). The depth below sea level
to the 5.0 km/s isovelocity surface, which is a proxy
for depth to the base of extrusive volcanics, is ∼2 km
within the southern propagating tip of the EVZ. The
depth to the 6.5 km/s isovelocity surface ranges from
5–10 km around Eyjafjallajökull (Brandsdóttir and
Menke, 2008). The IMO use an alternative velocity
model (VM2, Figure 2e) with slightly thinner upper-
most crust and thus higher velocities in the 3–6 km
depth range (Vogfjörð et al., 2002). Both velocity
models use a Vp/Vs ratio of 1.77 in the upper crust.
EARTHQUAKE DATA
We have used five earthquakes with local moment
magnitudes (Mlw) greater than 2.5 (Table 1) as a sam-
ple to test the effects of varying the hypocentral lo-
cation inversion parameters. All occurred under the
northeastern flank of Eyjafjallajökull during March
2010 and all have hypocentres within the region of in-
tense seismicity that occurred in the two weeks lead-
ing up to the Fimmvörðuháls eruption (Figure 2). The
test events were some of the largest-magnitude earth-
quakes to occur around Eyjafjallajökull during March
2010 and all display clear P- and S-wave arrivals at all
stations in the network that were operating at the time
(Figure 3). This means that arrival time picks of the
P-wave and S-waves are unambiguous. As such, these
test events are both some of the best-constrained ex-
amples and are also representative of the several thou-
sand other earthquakes that are found to be co-located
within a consistent depth range, when a consistent set
of location inversion parameters is employed.
We first use a Coalescence Microseismic Mapping
(CMM) technique (Drew, 2010; Brandsdóttir et al.,
2010; Tarasewicz et al., in press) to find hypocen-
36 JÖKULL No. 61, 2011