Jökull - 01.01.2011, Qupperneq 46

J. Tarasewicz et al.
tificially shallow levels. This would imply unusually
high velocities above the source region, distributed
such that travel times were reduced significantly to the
proximal temporary stations but not to the more dis-
tant permanent stations. We view this interpretation
as unnecessarily convoluted and unlikely compared to
the simpler interpretation that the deep locations in-
ferred by using only the more distant stations are the
result of sub-optimal network geometry.
Our analysis also disproves the notion that the true
depth of seismicity is ∼10 km, and that artificially
shallow hypocentres at ∼5 km might simply be the re-
sult of switching between VM1 and VM2. Rather, the
choice of velocity model is shown to be only a sec-
ondary factor in determining the hypocentral depth in
this situation.
Instead, our synthetic tests and inversions of the
earthquake data when the temporary stations are in-
cluded indicate that the true depth of our test events is
shallower than 6 km. Inverting arrival times from our
real sample earthquakes using our preferred param-
eters (velocity model VM1 and including data from
all available stations) returns hypocentral depths of
5–6 km (Figure 5a). However, Figure 8a shows that
even if VM1 is a good approximation of the true ve-
locity structure, hypocentral depths of 5±1 km would
be found for all true source depths in the range ∼2–6
km. This is because the geometry of the network lim-
its resolution at shallow depths, even when data from
all stations are included in the inversion.
On the other hand, inverting the arrival times
from our real sample earthquakes using VM2 returns
hypocentral depths of 2–4 km (Figure 5b). However,
we have shown that hypocentral depths shallower than
∼4 km are not well constrained by the network in this
epicentral region, even when data from all stations are
included. The only scenario in which our synthetic
tests produce best-fit hypocentres at such shallow
depths is when a velocity model that deviates from
the true velocity structure is used to invert the data. In
that scenario, the best-fit hypocentral locations are in
fact too shallow compared to the true source depths.
For the specific case in which synthetics generated
using a true velocity model VM1 are inverted using
the ’wrong’ VM2, the shallow hypocentres found for
the real data (using VM2, Figure 5b) match the ar-
tificially shallow depths of synthetic hypocentres for
true source depths of ∼2–6 km (Figure 8c). There-
fore, the behaviour of the hypocentral solutions for
the real test data when inverted with both VM1 and
VM2 is consistent with the true earthquake hypocen-
tres lying within the 2–6 km depth range. Whilst not
conclusive, the synthetic tests suggest that the more
scattered hypocentral solutions in the ∼2–4 km depth
range for the real test events are likely to be poorly
constrained and may hint at material deviations from
the true velocity structure in velocity model VM2.
Therefore, whilst the synthetic and real data
clearly indicate that the true depths of our test events
are <6 km, it is harder to distinguish between well re-
solved hypocentres of true sources at ∼4–6 km depth
and apparent hypocentral locations in the same 4–
6 km depth range that derive from earthquakes with
true source depths of 2–4 km. The observed pat-
tern of hypocentral depths obtained by inverting the
earthquake data with VM1 (5–6 km) and VM2 (2–4
km) could be obtained either for well resolved true
sources at 5–6 km depth, or for incorrectly located
true sources at 2–3 km depth (Figures 8a and 8c). Our
synthetic tests suggest that true sources shallower than
∼2 km would have apparent hypocentral locations at
depths >6 km and therefore should be distinguishable
from true sources in the 2–6 km depth range.
On the basis of seismic data alone, we cannot con-
clusively say whether our test events really are well
resolved at ∼5 km depth, as suggested by inversion
using our preferred parameters (Figure 5a). However,
this remains our preferred interpretation (compared to
the alternative interpretation that the true sources are
at 2–3 km depth) because it is in agreement with the
position of a sill inferred from surface deformation
data to be inflating at 4–6 km depth (Sigmundsson et
al., 2010). These authors also model an inflating dyke
(in combination with the sill) to fit the surface defor-
mation observations. This modelled dyke extends to
very shallow levels (10s to 100s of metres from the
surface), and hence may be consistent with the ob-
served seismicity being towards the shallower end of
the 2–6 km depth range. However, the modelled dyke
is located towards the eastern end of the seismically
46 JÖKULL No. 61, 2011