A DOUBLEEVENT INTERPRETATION OF THE 2003 LEFKADA EARTHQUAKE, GREECE

1
A DOUBLE-EVENT INTERPRETATION OF THE 2003 LEFKADA
EARTHQUAKE, GREECE

• J. Zahradník1), A. Serpetsidaki2),
• E. Sokos3), G-A. Tselentis2)
• 1) Charles Univ. Prague, 2) Univ. of Patras, 3)
  National Observatory of Athens

2
Data sources Le-3D/20sec stations of NOA,
CMG-3T a joint station of Charles and Patras
Univ.,location by Patras network
3
Questions to be answered

• What can we see from regional distances (about
  100 km) at periods 10 to 20 seconds ? Just a
  single point source, or Mpq(x,t) ?
• Any relation between the source process and the
  aftershock pattern ?
• Heterogeneous slip during a single event, or a
  multiple event ?

4
M6 Lefkada earthquake, Aug.14, 2003
CephaloniaTransfromFault(Louvari et al.,
1999)
5 BB stationsepic.distancelt140 km
5
Two clustersoperated simultaneously
Lefkada
Cephalonia
the first M4 event unambiguously located in the
southern cluster 9 minutes after the mainshock
6
Three 1D crustal model tested, one selected
Haslinger et al. (1999)
the same model used for HYPO location as well as
for the waveform modeling the location
depth (13 km) proved to be good also for the
waveforms (!)
7
Forward mainshock modeling, a single point
source, period 10-20 sec
Late arrivals unexplained !
8
Forward aftershock (M5) modeling, a single point
source
Now the fit is quite good !
For T gt 10 sec the crustal model is satisfactory,
but the mainshock cannot be treated as a point.
9
Multiple point-source inversion based on Kikuchi
Kanamori, a completely new code was written,
suitable for inversion of regional and local data.

• complete wavefield by DW method (M.Bouchon)
• synthetic subevents iteratively deconvolved from
  observed waveforms (Tgt10 sec)

10
Trial source positions along the Cephalonia
Transform Fault

• Each subevent seeks its position,
• rupture time, and Mpq.
• Some trial sources will rupture more than once,
  some of them not at all.

11
Moment tensor (deviatoric)at each trial
space-time position by minimization of the L2
waveform misfit (least squares) Optimum
space-time positionof subevents by maximization
of the waveform correlation (grid search)
12
Correlation as a fctn. of the space-time
position(foc mech superimposed)
an interactive graphic tool possibility to
impose various constraints e.g. polarities
13
trying to find the first subevent ...
Remove it, and proceed with the second one ...
14
... and now the second subevent
The best-DC solution is very stable, but the
non-DC part varies a lot. Thus we continue with
the DC-constrained solution.
15
Stable best-DC focal mechanism and much less
stable DC
16
Two sources explain the Lefkadaearthquakebetter
than one
SERG
black one subevent red two subevents
17
Final solution 3-4 major subevents,with
basically no slipbetween Lefkada and Cephalonia
explaining the two aftershock clusters
18
An independent validation from teleseismic data

• Benetatos et al. (2004) resolved
• 3 subevents, and the two major ones are very
  similar to sub 1 and 2 in this study.

Another supporting argument from Harvard CMT the
13.5 sec difference between the centroid time
and the hypocentral time, too large for a
single M6.
19
How many subevents (for Tgt 10 sec) ? Two,
three, or more ?

• The answer is tightly related to
• the uncertainty and stability problems.

20
Matching data by synthetics (4 subevents) and
checking stability
stability check repeatedly removing 1 station
21
(No Transcript)
22
Checking stability of subevents (repeatedly
removing one station)
sub 4 is less certain
mom1mom20.5e18 Nm
mom3mom40.2e18 Nm
23
DC or non-DC ?
Note that the two largest subevents have a
slightly different focal mechanism. Their
tensorial sum can mimic an apparently non-shear
event. We have no indication about a true
departure from 100 DC.
24
DC or non-DC ?
The tensorial sum sub 1 sub 2, (both 100
DC) imitates the HRV non-DC solution ! (J.
Sileny)
25
Space-timeevolution
sub dist (km) delay(s) 1 7.5
2 3 15.0 5 .... note this
separation .... 4 45.0 16
rupture velocity 2.5 km/s could explain this
pattern
26
A single M6 earthquake with several asperities
? Recall that sub 2 has epicentral distance of
45 km !

• Empirical relations (e.g. Papazachos, Papazachou,
  1997) indicate that such a large fault length is
  not compatible with M6.

27
Thus double-event in the title does not only
indicate two main patches of a single earthquake,
but rather two earthquakes.Sub 2 probably
triggered by sub 1.
An alternative terminology is a segmented
earthquake, but it might create confusion with
geological segmentation of the Cephalonia fault.
28
Seismic moment sub 1,3,4 0.9e18 Nmsub 2
0.5e18 Nmtotal 1.4e18 Nm(Mw6.1)
29
Seismic momentin this study is about 1/2 of that
reported by major agencies. A possible
interpretation is that we resolved only major
asperities.
30

• Somerville et al.(1999) found empirical relations
  between the fault and asperities on it.
• Aarea, Ddisplacement, Momoment
• Aa, Da, Moa ... asperity,
• A, D, M ... the entire fault
• Aa/A0.22, Da/D2, implying Moa/Mo0.44

31
How to estimate entire moment Mo from asperity
moment Moa ?

• Based on Somerville et al.(1999), Moa/Mo0.44 and
  assuming that sub 134 represent asperities,
  Moa0.9e18 Nm, we arrive at the whole moment
  Mo2.0e18 on Lefkada
• and, analogously, the asperitysub2 yields
  Mo1.1e18 Nm on Cephalonia

then Mw 6.3, same as from major agencies
32
Speculation about the fault area

• Based on the same relations (Somerville et al.,
  1999), we estimate the entire fault area A from
  the area of asperities Aa, using Aa/A0.22
• A164 km2 on Lefkada,
• A110 km2 on Cephalonia
• while an average area of one asperity is 16 km2

33
The double-event geometry is quite essential for
the Coulomb stress
(S. Nalbant)
Is it to expect a rupture of this 20 km
segment between Lefkada and Cephalonia ?
34
Conclusions

• What can we see at 10 lt T lt 20 sec ? We
  recognize 3-4 asperities with a space-time
  varying moment tensor.
• Any relation between the source process and the
  aftershock pattern ? Two clusters are explained
  by the main slip patches.
• Heterogeneous slip during a single event, or a
  multiple event ? Rather two earthquakes.

http//seis30.karlov.mff.cuni.cz
35
Asking new questions

• Was the double-event nature of this earthquake
  caused by the dynamic stress transfer ?
• What is the implication for the future activity
  of the Cephalonia Transform Fault ?
• Is the source complexity like this anomalous or
  typical for M6 earthquakes in Western Greece ?

http//seis30.karlov.mff.cuni.cz
36
http//seis30.karlov.mff.cuni.cz

• The paper accepted for publication in Bull.
  Seism. Soc. Am.

37
(No Transcript)
38
Let me add one more event, not planned for
this talk, showing multiplicity of the source
process, but, opposed to Lefkada, the subevents
will be tightly clustered in space.
For details, see Thursday 1530, C2
39
Vartholomio (near Zakynthos)Dec. 2,
2002ETH-SED Mw5.7DC55 !(also HRVand
Mednet)
40
We found three subevents, delayed by 2 2 sec,
within 2 km from each other
rupture propagation with an arrest ? or 3
separate events

2 2 sec
41
Six data subsets prove a very robust
result Sub 1,2,3 are not separated more than
2 km from each other, although the time delay
is 22 seconds. Multiple event, no rupture
propagation !
vertical plane 8 x 2 km
42
... their time separation (2 2 sec) and focal
mechanism is stable, incl. sub 3
2 2 sec
43
(No Transcript)
44
NOA network (N. Melis)