1. Complex earthquake directivity during the 2009 L’ Aquila mainshock Tinti E., Scognamiglio L., Cirella A., Cocco M., and A. Piatanesi Istituto Nazionale di Geofisica e Vulcanologia, Rome, Italy

2. What we knew just after the 2009 L’Aquila earthquake… 1)The 2009 L’ Aquila earthquake (Mw=6.1) occurred in the Central Appenines on April 6 th at 01:32 UTC. The focal mechanism identifies a normal fault having a strike of ≈133° and a dip ≈50°. INTRODUCTION:

3. What we knew just after the 2009 L’Aquila earthquake… 1)The 2009 L’ Aquila earthquake (Mw=6.1) occurred in the Central Appenines on April 6 th at 01:32 UTC. The focal mechanism identifies a normal fault having a strike of ≈133° and a dip ≈50°. INTRODUCTION: 2) Accelerometers of RAN Network show high values of PGV and PGA along south-east direction (see Akinci et al 2010).

4. What we knew just after the 2009 L’Aquila earthquake… 1)The 2009 L’ Aquila earthquake (Mw=6.1) occurred in the Central Appenines on April 6 th at 01:32 UTC. The focal mechanism identifies a normal fault having a strike of ≈133° and a dip ≈50°. INTRODUCTION: 3) Moment rate functions of velocimeters highlight a clear south-east directivity (Pino and Di Luccio 2009). 2) Accelerometers of RAN Network show high values of PGV and PGA along south-east direction (see Akinci et al 2010).

5. Accelerometers close to the fault show the existence of two P phases: EP and IP (Di Stefano et al 2011). EP represents the rupture onset, while IP is an impulsive phase occuring only 0.9 seconds after. IPEP What we knew just after the 2009 L’Aquila earthquake… INTRODUCTION:

6. Main features of kinematic models inferred from non-linear joint inversion (Cirella et al. 2009, 2012) are: 1)Two main patches: larger along SE, smaller along UP DIP. 2)Two separated rupture modes (Mode II and Mode III); 3)High rupture velocity in UP-DIP direction; slower rupture propagation along strike. What we knew just after the 2009 L’Aquila earthquake… INTRODUCTION: Poster XL 337; A. Cirella et al., this afternoon 17.30-19.00 Rupture modes computed according to Pulido and Dalguer 2009

7. The 2009 L’ Aquila earthquake provided the collection of an excellent data set of seismograms and High Frequency GPS data (CGPS). We study in details the complexity of rupture process: To unravel the directivity of this moderate magnitude earthquake; To study the details of near source accelerograms; To discuss the implications for ground motion prediction (engineering seismology) MOTIVATIONS:

8. In this study we have selected the following near source data: 1)Eight accelerograms of RAN Network and one of Mednet Network (distance < 35 km); 2)Two continuos (High-frequency, 10Hz) GPS data: CADO and ROIO RECORDED DATA: SITE EFFECTS: Because of complex geological structure, we are forced to use only frequencies lower than 0.4 Hz to exclude site effects at each station.

9. RECORDED DATA: Filter: 0.02 – 0.4 Hz FP fault parallel FN fault normal Z vertical seconds cm/s FP FNFN

10. RECORDED DATA: Filter: 0.02 – 0.4 Hz FP fault parallel FN fault normal Z vertical seconds cm/s

11. RECORDED DATA Seismograms in velocity, filtered between 0.02 – 0.4 Hz. Comparing AQU (black) and AQK (green) stations and AQV (blue) and AQG (red) stations. Fault parallel Fault normal Vertical AQU-AQK distance ≈500m

12. RECORDED DATA: POLARIGRAM We show the “polarigram” rather than the “particle motion” representation, because it gives at each time sample the precise amplitude and polarization of the velocity vector and makes easier a visual analysis of time series. (Bouin and Bernard,1994) Horizontal Components Particle Motion Polarigram

13. RECORDED DATA Polarigrams of filtered velocities for FN and FP components FP FN FP FN time

14. RECORDED DATA Arias FP fault parallel FN fault normal Z vertical

15. RECORDED DATA Arias FP fault parallel FN fault normal Z vertical high % of total energy in few seconds

16. With the aim to reproduce the details of the rupture process, in particular of the complex directivity, we compute a new kinematic model by using only stations with hypocentral distance within 35 km and all the stations available on the fault plane. NEW KINEMATIC INVERSION: Ingredients: 1)PROCEDURE: Non-linear inversion technique (Piatanesi et al. 2007, Cirella et al 2008). 2)VELOCITY STRUCTURE: Defined by Herrmann and Malagnini (2011). 3)We use a L1 + L2 hybrid norm to evaluate the fit. Inversion on Finite Fault

17. NEW KINEMATIC INVERSION: synthetics data Velocity (cm/s)

18. Inversion on Finite Fault synthetics data NEW KINEMATIC INVERSION: Velocity (cm/s)

19. Polarigrams synthetics data NEW KINEMATIC INVERSION:

20. Polarigrams synthetics data NEW KINEMATIC INVERSION:

21. Polarigrams synthetics data NEW KINEMATIC INVERSION:

22. How the different patches contribute to the seismograms 020 0 0 NEW KINEMATIC INVERSION: synthetics data

23. KINEMATIC MODELING: Finite Fault Forward Modeling How the different patches contribute to the seismograms 020 0 0 synthetics data

24. KINEMATIC MODELING: Finite Fault Forward Modeling How the different patches contribute to the seismograms GREEN: synthetics obtained only with the along up-dip slip patch. 020 0 0 synthetics data updip patch

25. KINEMATIC MODELING: Finite Fault Forward Modeling How the different patches contribute to the seismograms GREEN: synthetics obtained only with the along up-dip slip patch. 020 0 0 synthetics data updip patch

26. KINEMATIC MODELING: BLUE: synthetics obtained only with the along-strike slip patch. NWSE How the different patches contribute to the seismograms Finite Fault Forward Modeling m 020 0 0 synthetics data along strike patch

27. KINEMATIC MODELING: BLUE: synthetics obtained only with the along-strike slip patch. NWSE How the different patches contribute to the seismograms Finite Fault Forward Modeling m 020 0 0 synthetics data along strike patch

28. KINEMATIC MODELING: Three forward models with: 1) homogeneous slip UPDIP (20 % of total moment) 2) constant rupture velocity (3.5-4- 4.5km/s) 3) Constant rake (100°)

29. KINEMATIC MODELING: Three forward models with: 1) homogeneous slip UPDIP (20 % of total moment) 2) constant rupture velocity (3.5-4- 4.5km/s) 3) Constant rake (100°) data cm/s

30. KINEMATIC MODELING: Three forward models with: 1) homogeneous slip UPDIP (20 % of total moment) 2) constant rupture velocity (3.5-4- 4.5km/s) 3) Constant rake (100°) data cm/s

31. KINEMATIC MODELING: Three forward models with: 1) homogeneous slip UPDIP (20 % of total moment) 2) constant rupture velocity (3.5-4- 4.5km/s) 3) Constant rake (100°) data cm/s UP DIP Directivity is important but more complex (GSA is overestimated with a simple model)

32. CONCLUSIONS: A heterogeneous rupture process for a moderate-magnitude earthquake; A complex directivity not revealed by initial interpretations of ground motions; Near-fault waveforms (see AQU, AQK, AQV and AQG) mainly controlled by the UPDIP directivity, explaining the short duration of strong shaking (Arias Intensity) Very near-source recording sites (e.g., ROIO) clearly measure the separate contributions of the two rupture propagation phases; A detailed modeling of recorded waveforms (strong motion & HF-CGPS data) allows us to understand and demonstrate the effects of the initial fast up-dip rupture propagation, but our targeted inversion attempts do not allow us to match all the details of these waveforms revealed by polarigrams The rupture initially propagates at high speed (super-shear not yet excluded) These results have relevant implications for near-source ground motion prediction and GMPE

33. CONCLUSIONS: A heterogeneous rupture process for a moderate-magnitude earthquake; A complex directivity not revealed by initial interpretations of ground motions; Near-fault waveforms (see AQU, AQK, AQV and AQG) mainly controlled by the UPDIP directivity, explaining the short duration of strong shaking (Arias Intensity) Very near-source recording sites (e.g., ROIO) clearly measure the separate contributions of the two rupture propagation phases; A detailed modeling of recorded waveforms (strong motion & HF-CGPS data) allows us to understand and demonstrate the effects of the initial fast up-dip rupture propagation, but our targeted inversion attempts do not allow us to match all the details of these waveforms revealed by polarigrams The rupture initially propagates at high speed (super-shear not yet excluded) These results have relevant implications for near-source ground motion prediction and GMPE

34. THANKS

36. KINEMATIC MODELING:

37. EVENTUALE EFFETTO DEL DIP???

38. Distribution of local rupture velocity reveals rupture front of Vr≈ 4km/s for the UPDIP rupture propagation

39. This kinematic model is obtained by inverting DinSar, GPS and strong motion data. This model doesn’t include AQK and AQV and ROI and CADO (CGPS). In these stations we compute the forward synthetics. KINEMATIC MODEL: Fitting Polarigrams using Cirella et al (2012) joint inversion model synthetics data

40. KINEMATIC MODELING: Arias (in velocity)

41. FP FN RECORDED DATA Arias of filtered accelerograms (cumsquared acc) for near field stations: 1) Few seconds 80% of total energy 2) Energy partitioned in different way FN FP Z

42. Arias. Inversion on Finite Fault NEW KINEMATIC INVERSION: FP fault parallel FN fault normal Z vertical Data Synthetics

43. KINEMATIC MODELING: Point Source Modeling

44. KINEMATIC MODELING: Point Source Modeling

45. KINEMATIC MODELING: Point Source Modeling