Seismicity shadows: observations and modelling D. Marsan 1 , G. Daniel 2 , and M. Bouchon 2

1. Seismicity shadows: observations and modelling D. Marsan 1, G. Daniel 2, and M. Bouchon 2 1Laboratoire de Géophysique Interne et Tectonophysique, Universite de Savoie, Le Bourget du Lac, France. 2Laboratoire de Géophysique Interne et Tectonophysique, Observatoire de Grenoble, France. Corresponding Author: David.Marsan @ univ-savoie.fr

2. Shadows Parsons (2002)

3. Shadows Marsan (2003) First 100 days after the Landers earthquake

4. Shadows Ogata et al. (2003)

5. Shadows Also: Toda and Stein (2002) Mallman and Zoback (2003) Felzer and Brodsky (2005)  causality? Marsan and Nalbant (2005) Ma et al. (2005) Instances of clear immediate quiescences: Dieterich et al. (2000) Toda and Stein (2003) Woessner et al. (2004)

6. Shadows • So: • Quiescences are rare • They can develop late after the mainshock • They could be triggered by something else than the mainshock… • Or simply happen by chance • Why? • They are difficult to detect (easier to see increases than decreases in seismicity rates) • They could be masked by dynamic triggering • They could be delayed by stress heterogeneity

7. Modelling 1 2 Rate and state friction (Dieterich 1979, Ruina 1983) with slowness law (Dieterich 1986) Total # of earthquakes triggered by a set of stress steps {ti} is proportional to the mean t = E{t} For a fault experiencing a Gaussian distribution N (t,st) of stress changes at time t=0 with t = E{t} < 0 and large st:

8. Modelling lfault(t) / m initial triggering rate increase 1 t rate decrease total number < 0

9. Modelling • 10 km x 10 km self-similar slip distribution down to 40 m • Hurst exponent H = 0.7 (Mai and Beroza 2002)

10. Modelling Change in shear stress for parallel strike-slip faults at various distances

11. Modelling

12. Modelling • Stress variability can be caused by: • slip heterogeneity • roughness of the main fault plane • crustal heterogeneity • roughness of the target fault • averaging stress over (large) volumes

13. Analyses We analyse two sequences: The 1999 Chi-Chi earthquake (Taiwan) The 1999 Izmit – Düzce earthquakes (Turkey)

14. Analyses • Non-stationary Poisson seismicity (cost function) • N triggers occurring at specified times {ti} • Each creating a normal N(t,st) distribution of stress steps • Numerically invert for the 2xN parameters {ti} and {sti}

15. Analyses synthetics inversion

16. Analyses 100 trials of forward model + inversion

17. Analyses Ma et al. (2005)

18. Analyses due to Chi-Chi?

19. Analyses

20. Analyses

21. Analyses

22. Analyses

23. Analyses

24. Analyses • Quiescence but: • late (starts > 1.5 years after Chi-Chi) • more likely due to a trigger (aseismic?) rather than stress varibility • also: strong dynamical triggering

25. Analyses

26. Analyses

27. Analyses

28. Analyses

29. Analyses

30. Analyses Daniel et al. (in revision) Düzce

31. Analyses

32. Analyses Quiescence Düzce

33. Conclusion • For Chi-Chi, there are 2 (+1?) cases of shadows, but: • Both develop after initial (dynamic) triggering that lasts several days • One starts >1.5 years after mainshock, possibly triggered by something else than Chi-Chi. • One develops late (>2.4 years) at depth • For Düzce, no clear observation of shadows, except at Yalova, but: • Activity was most likely dynamically triggered… • … and sustained itself for months. • Shadow develops late (> 1 month after Düzce, > 4 months after Izmit)

34. Conclusion • Off-fault shadows are more likely to occur • At shallow depths, • After several weeks / months of delay • And could be directly caused by (aseismic?) triggers rather than by co-seismic stress heterogeneities.