Jökull - 01.01.2010, Qupperneq 29

The 2008 South Iceland Seismic Zone aftershock sequence
The LUT was calculated for a one-dimensional
velocity model, being a simplification of the real
three-dimensional crustal structure. Localised near
surface velocity anomalies beneath each station could
potentially have a considerable effect on their over-
all travel time. No station corrections were applied.
The resultant location estimates and calculated uncer-
tainties are impacted by the uncorrected delays. In
addition to applying station correction terms, location
errors could be further reduced by introducing source
specific station terms (Richards-Dinger and Shearer,
2000 ) or using the double-difference relative location
technique of Waldhauser and Ellsworth (2000).
EARTHQUAKE LOCATIONS
Three sets of data were run through the CMM pro-
grams, initially using only data from the 11 tempoary
LOKI stations, followed by a rerun with additional
three SIL stations and finally data from nine SIL sta-
tions. A total of ∼19450 events were located using
data from 14 stations. Using the 11 LOKI stations
CMM detected 11057 events. Smaller, poorly con-
strained events were omitted resulting in 13868 events
with signal-to-noise (SN) ratio >2.5 and 7846 events
with SN ratio >3 and epicentral and depth location
errors within 1 km and 2 km, respectively. The signif-
icant jump in the number of detected events with 14
stations is due to more events being located during the
first two days while the LOKI network was being in-
stalled as well as improved event detection. Running
CMM with both the LOKI and SIL station data pro-
duced less scatter of events and better defined faults
within the epicentral zone. Running CMM on data
from nine SIL stations resulted in 18504 events being
detected, of which 6824 passed filtering.
Most events have latitude and longitude errors
0.7±0.3 km and depth errors 1±0.5 km, as estimated
by the CMM algorithm. These errors are within one
standard deviation measure as determined by the com-
puted probability density function (pdf). Each numer-
ically computed event location pdf incorporates inher-
ent assumptions regarding uncertainty, including the
measurement uncertainty as governed by the choice of
parameters for the STA/LTA function. As with other
location algorithms that have their own inherent as-
sumption, the uncertainty estimates provide a relative
measure of confidence in the location estimates. In
addition we assessed variations in crustal structure by
running CMM on three different crustal models, see
discussion below. No observed change in epicentral
distribution was observed showing the CMM result to
be quite robust against variations in velocity.
The epicentral map is dominated by seismicity
along two parallel N–S trending faults spaced 4–5 km
apart and diffuse activity stretching mainly westward
along an E-W aligned zone (Figures 1 and 5), into an
E-W zone mapped by Vogfjörd et al. (2005) from the
aftershock activity of the 13 November 1998 earth-
quake. Considerably smaller aftershock activity is ob-
served on the eastern N-S fault (Ingólfsfjall), possibly
suggesting that the mainshock rupture left little resid-
ual stress on the fault to cause aftershocks. Given the
extent of the aftershock clusters, they mark the fault
as being about 13 km long. The second main fault
(named Reykjafjall or Kross fault) lies about 5 km
to the west and is slightly longer (∼17 km). Most
of the aftershock activity originated along this fault.
Several smaller N-S faults are also active within the
main E-W zone. Events outside the seismic network,
at the western and eastern ends of the E-W zone, are
not well constrained. Clustering of event locations
suggests the main faults to be made up of numer-
ous smaller segments branching into conjugate faults
at each end (see discussion below). The two cross
sections highlight the depth distribution of events.
Most aftershocks lie between 1 and 9 km depth along
the two main faults with the eastern fault having a
slightly smaller depth distribution than the western
fault, whereas events along the E-W zone west of
the main faults are concentrated below 5 km depth.
Hypocentral depths are markedly shallower along the
northern portion of both major fault where they extend
down to 4–6 km, deepening southwards to 9 km. Of
note is the relative lack of events in the middle of the
fault, where the two main events originated (Hreins-
dóttir et al., 2009).
A more direct comparison of the CMM and SIL
locations along the two major N-S faults underlines
how a dense local network around the epicentral zone
is required in order to constrain the depth distribution
JÖKULL No. 60 29