[1] Subduction zone plate boundary megathrust faults accommodate relative plate motions with spatially varying sliding behavior. The 2004 Sumatra-Andaman (M w 9.2), 2010 Chile (M w 8.8), and 2011 Tohoku (M w 9.0) great earthquakes had similar depth variations in seismic wave radiation across their wide rupture zones -coherent teleseismic short-period radiation preferentially emanated from the deeper portion of the megathrusts whereas the largest fault displacements occurred at shallower depths but produced relatively little coherent short-period radiation. We represent these and other depth-varying seismic characteristics with four distinct failure domains extending along the megathrust from the trench to the downdip edge of the seismogenic zone. We designate the portion of the megathrust less than 15 km below the ocean surface as domain A, the region of tsunami earthquakes. From 15 to $35 km deep, large earthquake displacements occur over large-scale regions with only modest coherent short-period radiation, in what we designate as domain B. Rupture of smaller isolated megathrust patches dominate in domain C, which extends from $35 to 55 km deep. These isolated patches produce bursts of coherent short-period energy both in great ruptures and in smaller, sometimes repeating, moderate-size events. For the 2011 Tohoku earthquake, the sites of coherent teleseismic short-period radiation are close to areas where local strong ground motions originated. Domain D, found at depths of 30-45 km in subduction zones where relatively young oceanic lithosphere is being underthrust with shallow plate dip, is represented by the occurrence of low-frequency earthquakes, seismic tremor, and slow slip events in a transition zone to stable sliding or ductile flow below the seismogenic zone.Citation: Lay, T

Supplemental Information on Rupture Analyses P and SH Wave Finite Fault InversionThe body wave inversions yield predominantly bilateral slip, although there may be a slight asymmetry in the slip distribution with greater extent toward the northwest. The P-and SH-waveform matches for our preferred finite source model are shown in Fig. S3. The slip model accounts for about 78% of the weighted signal power in the 120-s-long interval used in the inversion, with the source time function representing slip during the first 60 s of rupture. The seismic moment was estimated as 2.5 x 10 21 Nm (M w 8.2), but experience indicates that this may be less reliable than a determination made at longer periods and we adopt a final moment of 1.8 x 10 21 Nm (M w 8.1) based on composite source modeling of 714-1000 s period Rayleigh waves. There are clearly some P wave motions about 100 s after the first arrival in the waveforms of stations in Asia (TATO, ENH, BJT) that are not matched by the source model.Using very long fault models extending toward the northwest allows the normal fault model to match this late energy, but as discussed below, the late signal appears to originate from a secondary event with different fault mechanism. The depth extent of rupture is not well resolved, but we find that localized regions of significant slip are present in all models extending to depths of 30-36 km (>24 km below the 6 km thick ocean layer). The likelihood that rupture during the mainshock extended this deep is strongly supported by the depth of the largest aftershock in the trench-slope region with a location and mechanism consistent with being on the mainshock rupture surface. This is the event of October 19, 2009 (22:49:37 UTC, 15.3°S, 172.2°W, M w 5.9). The W-phase and GCMT solutions (Fig. 1) for this event have compatible normal fault geometries, with the W-phase centroid depth being 32.5 km. We inverted 40 azimuthally welldistributed teleseismic P waves for a finite source model for this event, assuming a hypocentral depth of 30 km. The resulting slip model and examples of fits to the data are shown in Fig. S4.The rupture centroid is confirmed to be about 32 km deep.The rupture velocity of predominantly bilateral ruptures is usually not well resolved by teleseismic body wave data because their high apparent velocities provide limited resolution.Essentially, the slip distribution scales spatially directly with the assumed rupture velocity.However, surface waves have lower apparent velocities and are thus more sensitive to directivity effects. We compared observed and predicted of short-arc Rayleigh wave (R1) source time functions (discussed below) for many body wave models to bound the rupture velocity and spatial extent of the slip model. We compared predicted R1 STFs for the model in Fig. S2 with azimuthally binned and stacked STF observations, finding little evidence for azimuthal variation of STF duration, consistent with a largely bilateral rupture. Preferred rupture velocities based on fitting the STF characteristics are less tha...

1] The M w 7.9 Wenchuan earthquake of 12 May 2008 was the most destructive Chinese earthquake since the 1976 Tangshan event. Tens of thousands of people were killed, hundreds of thousands were injured, and millions were left homeless. Here we infer the detailed rupture process of the Wenchuan earthquake by back-projecting teleseismic P energy from several arrays of seismometers. This technique has only recently become feasible and is potentially faster than traditional finite-fault inversion of teleseismic body waves; therefore, it may reduce the notification time to emergency response agencies. Using the IRIS DMC, we collected 255 vertical component broadband P waves at 30-95°from the epicenter. We found that at periods of 5 s and greater, nearly all of these P waves were coherent enough to be used in a global array. We applied a simple down-sampling heuristic to define a global subarray of 70 stations that reduced the asymmetry and sidelobes of the array response function (ARF). We also considered three regional subarrays of seismometers in Alaska, Australia, and Europe that had apertures less than 30°and P waves that were coherent to periods as short as 1 s. Individual ARFs for these subarrays were skewed toward the subarrays; however, the linear sum of the regional subarray beams at 1 s produced a symmetric ARF, similar to that of the groomed global subarray at 5 s. For both configurations we obtained the same rupture direction, rupture length, and rupture time. We found that the Wenchuan earthquake had three distinct pulses of high beam power at 0, 23, and 57 s after the origin time, with the pulse at 23 s being highest, and that it ruptured unilaterally to the northeast for about 300 km and 110 s, with an average speed of 2.8 km/s. It is possible that similar results can be determined for future large dip-slip earthquakes within 20-30 min of the origin time using relatively sparse global networks of seismometers such as those the USGS uses to locate earthquakes in near-real time.

The frequency-dependent rupture process of the 11 March 2011 M w 9.0 off the Pacific coast of Tohoku Earthquake is examined using backprojection (BP) imaging with teleseismic short-period (∼1 s) P waves, and finite faulting models (FFMs) of the seismic moment and slip distributions inverted from broadband (>3 s) teleseismic P waves, Rayleigh waves and regional continuous GPS ground motions. Robust features of the BPs are initial down-dip propagation of the short-period energy source with a slow rupture speed (∼1 km/s), followed by faster (2-3 km/s) rupture that progresses southwestward beneath the Honshu coastline. The FFMs indicate initial slow down-dip expansion of the rupture followed by concentrated long-period radiation up-dip of the hypocenter, then southwestward expansion of the rupture. We explore whether these differences correspond to real variations in energy release over the fault plane or represent uncertainties in the respective approaches. Tests of the BP results involve (1) comparisons with backprojection of synthetic P waves generated for the FFMs, and (2) comparisons of backprojection locations for aftershocks with corresponding NEIC and JMA locations. The data indicate that the down-dip environment radiates higher relative levels of short-period radiation than the up-dip regime for this great earthquake, consistent with large-scale segmentation of the frictional properties of the megathrust.