Jökull - 01.01.2015, Blaðsíða 56

J. Schuler et al.,
We also used the probabilistic location algorithm
NonLinLoc (Lomax et al., 2009) to estimate the lo-
cation error of some events introduced by assigned
picking uncertainties. The location probability den-
sity function, or location misfit function, is sampled
5000 times (scatter points in Figure 3b) during the lo-
cation process of each event. For some hypocenters,
the highest location likelihood are displayed in Fig-
ure 3. It is common that location constraints are poor
and the scatter points do not have an elliptical shape,
in which case we cannot estimate location errors us-
ing a standard Gaussian error approximation. Regions
with higher sample densities (scatter points) represent
more likely hypocenter locations. For the selected
earthquakes we display the 68% contour of the scat-
ter points, together with the most likely location. We
arbitrarily chose the 68% contour, which would repre-
sent one standard deviation of a normally distributed
dataset. If instead of our preferred VELEST veloc-
ity model the SIL model is employed in the (NonLin-
Loc) event relocations, we can estimate the difference
of their highest likelihood location. For the same se-
lected events as before, their location differences are
plotted as error bars in Figure 3b.
Polarity information for the P- and S-waves were
used to estimate earthquake focal mechanisms using a
probabilistic inversion algorithm (Pugh, 2015). Only
three maximum probability moment tensor solutions
could be calculated from the deep (>8 km depth b.s.l.)
events (Figure 1a). More solutions were computed
for the shallow events, but we show only a few well-
constrained fault plane solutions of earthquakes lo-
cated in the center of the cluster beneath Bæjarfjall
(Figure 1b).
DISCUSSION
The 1D velocity model obtained for Þeistareykir re-
sembles the regional SIL velocity profile within the
uppermost 2 km, but deviates from it, with higher ve-
locities below 2 km. Similar higher velocities are ob-
tained beneath Krafla from seismic tomography (e.g.
Schuler et al., 2015). We calculated the event location
errors for both profiles, because the manually picked
earthquakes are clustered in space and represent, to-
gether with our receiver array, a sub-optimal prereq-
uisite for robust velocity estimation that is based on
a trial and error process. Nevertheless, it is the first
local seismic velocity model for Þeistareykir.
Most of the deep (8–20 km depth b.s.l.) earth-
quakes are located southeast of Þeistareykir. Both P-
and S-wave first arrival amplitudes of these events are
small, some emergent, compared to the majority of
earthquakes beneath Bæjarfjall. Hence, we assigned
large picking uncertainties that translate to large loca-
tion error estimates (Figure 3b). Despite their large
uncertainties, their hypocenters lie below the base of
the shallow seismogenic zone of the crust, at 6–8 km
depth b.s.l., based on seismic information gathered
around Askja (Martens and White, 2013; Greenfield
and White, 2015). We surmise that these deep earth-
quakes, caused by brittle (tectonic) rock failure in a
plastic regime, occur due to either high strain rates or
reduced normal friction caused by melt movement in
the lower crust. Some of their hypocenters lie beneath
the Borgarhraun lava flow (Figure 1). The likely depth
of crystallization of Borgarhraun and other primitive
basalts (Maclennan, 2008) indicate that melt move-
ment at the base of the crust occurs and possibly
causes brittle rock failures at these depths. Þeista-
reykir is also located at the southeastern end of the
onland-offshore Húsavík-Flatey system (HFFS, Fig-
ure 2c). The HFFS merges with the Þeistareykir fis-
sure swarm just north of Bæjarfjall (Hjartardóttir et
al., 2015) and may play an important part in the
crustal strength of this region. Spatial clustering of
events may point to a preferred interpretation of such
data, but the wide scatter of the deep events may be a
result of our sparse station distribution and we there-
fore do not prefer one interpretation over another.
The seismograms in Figure 2 show the arrivals
from a 16–17 km (b.s.l.) deep earthquake. Both the P-
and S-wave first arrivals can be followed to epicentral
distances of over 90 km towards the Askja region. An
interesting arrival in this record section is a seismic
energy package, indicated as an orange line, arriving
after the S-wave first breaks. The energy package can
be identified well on a few traces, especially towards
the north, but is rather difficult to follow with increas-
ing station offsets as well as towards the south. The
arriving energy could be interpreted as being the S-
56 JÖKULL No. 65, 2015