1. Evidence on non-self-similarity source scaling in cluster earthquakes Yen-Yu Lin 1, Kuo -Fong Ma 1, Hiroo Kanamori 2, Teh-Ru Song 3, Nadia Lapusta 2, Victor Tsai 2 1 Department of Earth Sciences, National Central University, Taiwan 2 Seismological Laboratory, California Institute of Technology, USA 3 Department of Earth Sciences, University College London, UK IES 2014/3/19

2. Earthquake self-similarity???? 2 Smaller earthquake Larger earthquake r1 r2 Dimension (Circular) Source time function Constant rupture speed (Aki, 1967) Seismic waveforms shorter P-dur longer P-dur Similar Q structure

3. Cluster with constant P- durations Cluster A P-phase durations Pdur = 0.07s (~15Hz) Magnitude variation Mwe 0.27 to 1.97 Filter: Notch filter 60Hz 0.27 1.97

4. TCDP Borehole seismometers Instrument response Natural frequency: 4.5Hz Damping: 0.29 Gain: 100 Sampling rate for analysis 1000 p/s before 2008 (Nyquist frequency 500Hz) 200 p/s after 2008 (Nyquist frequency 100Hz) Corrections before analysis - Instrument response - Galperin angle - Orientation (Lin et al., 2012)

5. Earthquake Clusters -2006/11~2007/12 (14 months) - Correlation coefficient (3 comp.) > 0.8 -130 clusters (2~11 events) - Mwe 0.0~2.0 (S-wave max. amp.) - Located 7 clusters (A-G) (> 4 events) by stacked waveforms 5

6. Magnitudes estimation 6

7. Magnitude variation for the located clusters

8. Seismic clusters- finite P duration 8 0.27 1.97 0.49 1.62 0.35 1.32

9. 9 Instrument problem? NO!

10. 10 Analysis 1 : Empirical Green’s function Analysis 2 : Futterman Q correction (Futterman, 1962) within a cluster, B ak (f)=B bk (f) For event a and b with receiver k Observed velocity spectrum: (Wang et al., 2010; 2012)

11. 11 Analysis 1 : Empirical Green’s function T w =0.010~0.024 s Source dimension=20~50m Power of 0.04(1/20)~0.10(1/10)

12. 12 Analysis 2 : Futterman Q correction Q=202 & Q=101 T w =0.020~0.054 s (a constant for each cluster) Source dimension=40~110m Q=202 Q=101

13. 13 Comparison to earthquake self-similarity empirical relation (9.2>M>6.5) Source scaling for the clusters (Duputel et al., 2013) Clusters with the constant P- wave durations observations break down the earthquake self-similarity behavior. Power of 0.04(1/20)~0.10(1/10)

14. Constant rupture speed “a characteristic length “

16. Rupture speed and source dimension It can be constrained by source time function estimations. If α=1/3, earthquake self-similarity! But, selection of α can be arbitrary! -Any specified α could also be another possible model to explain the unique observations.

17. Summary - We discovered 3 clusters with constant P-phase durations (Pdur) for events Mw 0.5 to 2.0 in TCDPBHS records, which had been shown to be natural events from deformed zone of decollement. -The constant P-phase durations observations in the clusters break down the earthquake self-similarity behavior of Tw ∝ Mo (1/3) as examined by both the empirical Green’s function and Futterman Q correction analyses based on assumption of a constant rupture speed. -A potential model to explain the constant Pdur phenomenon is an existence of a characteristic length on the fault, limiting the duration of the source time function. - The difference in high frequency component between smaller and larger events might be due to the heterogeneity of fault. -If not characteristic length related, another possibility is the significant difference in rupture speed among the cluster events.

18. Thank you very much!

20. Analysis 1 : Empirical Green’s function