Snmek 1

1
Problematic non-double-couple mechanism of the
2002 Amfilochia Mw5 earthquake, western Greece
Adamová P (1), Sokos E (2), Zahradník J (3) (1)
Academy of Sciences, Institute of Geophysics,
Prague, adamova_at_ig.cas.cz (2) University of
Patras, Seismological Laboratory, Patras (3)
Charles University, Faculty of Mathematics and
Physics, Prague
Abstract Two seismic agencies reported a very low
double-couple percentage (DC) of the Amfilochia
earthquake, of about 30 and 60, by SED and
MEDNET, respectively. Near-regional waveforms,
carefully analyzed in this paper for the DC,
suggested a higher DC, ranging from 75 to 100,
dependent on the uncertainty of the source
position (optimum value DC93). Using a
statistical F-test, forward modeling of the
near-regional data with a single-event low-DC
source yields a significantly worse waveform
match. The fit of near-regional data can be
further improved (although at the 90
significance level only) if considering a
speculative two-event model. The same model, when
viewed at the very-low frequency range, reaches
the very low DC values. However, two features
make the two-event model unlikely The two
subevents strongly differ in their focal
mechanism, and their mutual separation (3.5
seconds) is larger than the expected duration of
this earthquake. Therefore, the two-source model
appears to be nothing but an interesting
equivalent representation of the non-DC model,
providing some insight into possible origin of
the apparently low DC. Preference is given to
the simple interpretation, most clearly supported
by the near-regional data, that the Amfilochia
earthquake was an almost pure-shear event. See
also a paper on this topic to appear in J.
Seismol.
Amfilochia earthquake, December 31, 2002, Mw5
F test

• Data, location
• stations from the network of NOA (National
  Observatory of Athens, IG) Lennartz Le-3D/20 sec

Each model is characterized by its misfit
triangles BB stations used for waveform
modeling stars epicenter locations by different
agencies Epicentral distances 50 to 150 km
O observed data S synthetic data F ratio is
defined as it is
compared with F distribution with N-M degree of
freedom In our case M5 (deviatoric MT solution),
N84 (7 independent samples at each record 3
components and 4 stations)
DC 70 100
DC 31, 2 sec shift from the optimal solution
Stable strike, dip and rake Unstable DC
(differences between agencies and differences in
our solution) F-test prefers DC93 with respect
to DC31 at the 99 significance level F test
prefers our solution (DC93) with respect to
agency solutions (30 and 61) at the 90
significance level
Agencies
This study
Note drastic difference between the DC values.
Speculative 2-event model Using F-test we
find that adding second subevent represents an
improvement of the waveform match, but
significant at the 90 level only
Waveform fit
The difference between the single-source and
two-source synthetics is not large, but visible
on seismograms, basically as a small shift
(compare black and dashed line). Blue line
observed data Black line synthetic data for
subevent 1 Dashed line synthetic data for both
subevents Red line synthetic data for the SED
model
Synthetic very low frequency solution
(0.005-0.025 Hz) derived from the two-event
model

One dominant subevent with DC24 Agreement with
SED solution achieved with two subevents.

• Conclusion and discussion
• Single source at flt0.11 Hz it must have a
  large DC otherwise regional data are not fit
  well
• Regional data at flt 0.11 Hz do NOT support a
  SINGLE source with a SMALL DC.
• If a single-source,low DC model (e.g. SED) is
  correct at very low frequencies (flt 0.025 Hz),
  the DC is frequency dependent. This case can be
  modeled using a two-event source model at flt 0.11
  Hz with DC100 for each subevent.

Reference Adamova P., Sokos E., Zahradnik J
Problematic non-double-couple mechanism of the
2002 Amfilochia Mw5 earthquake, Western Greece,
Journal of Seismology, 10.1007/s10950-008-9112-4
(in press).
Acknowledgmets The authors thank NOA-IG and the
Univ. of Thessaloniki for waveforms. The GMT
software of Wessel and Smith (1991) was used.
This work was financially supported by following
projects in the Czech republic GAUK
279/2006/BGEO/MFF, GACR 205/07/0502 and MSM
0021620860