A Rayleigh wave back-projection method applied to the 2011 Tohoku earthquake Daniel Roten, Hiroe Miyake, and Kazuki Koketsu (2012), GRL Earthquake of the Week - 27 January 2012 Roten, D., H. Miyake, and K. Koketsu (2012), A Rayleigh wave back-projection method applied to the 2011 Tohoku earthquake, Geophys. Res. Lett., 39, L02302, doi:10.1029/2011gl050183.
Motivation and Summary
Motivation and Summary Region of maximum slip in recent source inversions of 11 March, 2011 Mw 9.1 EQ strongly data-dependent: Near trench [e.g. Lay et al., 2011; Pollitz et al., 2011] Near JMA hypocenter [e.g. Ammon et al., 2011; Simons et al., 2011] Down-dip of JMA hypocenter beneath coastline [e.g. Meng et al., 2011] Up-dip extent of rupture overall difficult to assess [e.g. Lay et al., 2011]
Motivation and Summary Region of maximum slip in recent source inversions of 11 March, 2011 Mw 9.1 EQ strongly data-dependent: Near trench [e.g. Lay et al., 2011; Pollitz et al., 2011] Near JMA hypocenter [e.g. Ammon et al., 2011; Simons et al., 2011] Down-dip of JMA hypocenter beneath coastline [e.g. Meng et al., 2011] Up-dip extent of rupture overall difficult to assess [e.g. Lay et al., 2011] Authors suggest frequency-dependent radiation as potential explanation [e.g. Koper et al., 2011]
Motivation and Summary Region of maximum slip in recent source inversions of 11 March, 2011 Mw 9.1 EQ strongly data-dependent: Near trench [e.g. Lay et al., 2011; Pollitz et al., 2011] Near JMA hypocenter [e.g. Ammon et al., 2011; Simons et al., 2011] Down-dip of JMA hypocenter beneath coastline [e.g. Meng et al., 2011] Up-dip extent of rupture overall difficult to assess [e.g. Lay et al., 2011] Authors suggest frequency-dependent radiation as potential explanation [e.g. Koper et al., 2011] Employ multi-frequency (T=13-100 s) surface-wave back projection in order to assess this effect
Methods: Back projection
Methods: Back projection u 1 u2 u 3 X u k ( k ) k Given grid of potential source locations, identification by waveform stacking and migration to predicted travel time of chosen phase Want: highest amplitude phase, well-separated
Methods: Back projection u 1 u2 u 3 X u k ( k ) k Given grid of potential source locations, identification by waveform stacking and migration to predicted travel time of chosen phase P arrival used successfully for teleseismic body-wave BP, but local to regional distances difficult: Want: highest amplitude phase, well-separated Local distance: little separation between P and stronger S Regional distance: (>150km) Sn before direct Sg, followed by strong Lg
Methods: Back projection u 1 u2 u 3 X u k ( k ) k Given grid of potential source locations, identification by waveform stacking and migration to predicted travel time of chosen phase P arrival used successfully for teleseismic body-wave BP, but local to regional distances difficult: Want: highest amplitude phase, well-separated Local distance: little separation between P and stronger S Regional distance: (>150km) Sn before direct Sg, followed by strong Lg Authors use Rayleigh wave BP from K-NET and KiK-net recordings, as these phases are both strong and dispersive
Methods: Waveform processing
Methods: Waveform processing Authors use a continuous wavelet transform in order to isolate energy at the period of interest: wt k (s, ) = 1 p s Z +1 1 u k (t) t s dt
Methods: Waveform processing Authors use a continuous wavelet transform in order to isolate energy at the period of interest: wt k (s, ) = 1 p s Z +1 1 u k (t) t s dt where B(t) = p f b applesinc fb t p p exp {2 if c t} f b f c bandwidth center Chose 1 Hz center in mother wavelet - scale becomes period
Methods: Waveform processing Authors use a continuous wavelet transform in order to isolate energy at the period of interest: Synthetic (FD) example for vertical component waveforms of Mw 7.3 foreshock (T=13s) Rayleigh waves clearly visible as strongest phases in CWT envelopes Roten et al., 2012
Methods: Back projection (again)
Methods: Back projection (again) NX 2 S m (s, ) = wt k (s, + m,k ) w m,k p m,k k=1 Authors take square modulus of migrated, weighted, stacked waveforms s m k period source time source index station index
Methods: Back projection (again) S m (s, ) = NX k=1 wt k (s, + m,k ) w m,k p m,k 2 Authors take square modulus of migrated, weighted, stacked waveforms s m k period source time source index station index w m,k 1 = w m,k source-station azimuth weighting NX i=1 1 m,i m,k
Methods: Back projection (again) S m (s, ) = NX k=1 wt k (s, + m,k ) w m,k p m,k 2 Authors take square modulus of migrated, weighted, stacked waveforms s m k period source time source index station index w m,k 1 = w m,k p m,k source-station azimuth weighting NX i=1 1 m,i m,k source-station distance weighting zero for distances less than 100 km
Methods: Back projection (again) S m (s, ) = NX k=1 wt k (s, + m,k ) w m,k p m,k 2 Authors take square modulus of migrated, weighted, stacked waveforms s m k period source time source index station index w m,k 1 = w m,k p m,k source-station azimuth weighting NX i=1 1 m,i m,k source-station distance weighting zero for distances less than 100 km m,k Predicted travel time from group-velocity dispersion map
141 Methods: Back projection (again) S m (s, ) = NX 44 wt k (s, + m,k GV ) (km/s) w m,k p m,k 43 1 2 3 4 2 k=1 42 Authors take square modulus of migrated, weighted, stacked waveforms s m k period source time source index station index w m,k 1 = w m,k Right: Fundamental-mode Rayleigh wave group velocities derived from the p m,k Japan Integrated Velocity Structure Model (JIVSM) [Koketsu et al., 2008] - here shown for T=20s. 41 source-station azimuth weighting 40 39 38 NX i=1 37 1 m,i m,k source-station distance weighting 36 zero for distances less than 100 km 35 138 139 143 m,k Predicted travel time from group-velocity dispersion map
Methods: Synthetic validation L02302 ROTEN ET AL.: RAYLEIGH B FD simulation of Mw 7.3 foreshock as a point source in the Japan Integrated Velocity Structure Model [Koketsu et al., 2008] Stack maximum at zero time and true location Ghost peaks migrate from ESE to WNW At right: BP of Mw 7.3 foreshock radial-component synthetics (T=20s) using same 384 stations as in later study of Mw 9.1 mainshock
Methods: Synthetic validation FD simulation of Mw 7.3 foreshock as a point source in the Japan Integrated Velocity Structure Model [Koketsu et al., 2008] 44 42 0.0 0.5 1.0 S m (s, ) Stack maximum at zero time and true location 40 Ghost peaks migrate from ESE to WNW 38 At right: T=20s array response function estimate for the 384 K-NET / KiK-net stations used in mainshock BP. 36 138
141 141 141 141 Methods: Data validation 143 145 143 145 BP of Mw 6.5 aftershock of 28 March using 284 K-NET / KiK-net stations. Same behavior as in synthetic example. 41 40 39 (a) 0.0 0.5 1.0 S m (s, ) (b) 41 40 39 38 38 Stack maximum at zero time and true location 37 36 20 s 1.41e+07 0 s 1.82e+07 37 36 Ghost peaks migrate from ESE to WNW 41 40 39 (c) (d) 41 40 39 At right: T=13s BP result. JMA epicenter is yellow star. 38 37 36 143 20 s 9.23e+06 145 143 40 s 3.21e+06 145 38 37 36
Application to the Mw 9.1 mainshock Authors performed Rayleigh-wave BP at 13, 20, 30, 50, and 100 s ROTEN ET AL.: Prefer RAYLEIGH a three-stage BACK-PROJECTIONprogressive TOHOKU rupture model L02302 Stage 1: Local maximum in stack amplitude at 10 s just SW of JMA epicenter At left: 20 s mainshock BP
Application to the Mw 9.1 mainshock Authors performed Rayleigh-wave BP at 13, 20, 30, 50, and 100 s ROTEN ET AL.: Prefer RAYLEIGH a three-stage BACK-PROJECTIONprogressive TOHOKU rupture model L02302 Stage 1: Local maximum in stack amplitude at 10 s just SW of JMA epicenter Stage 2: Dual peaks to SE (33 s) and NE (39 s) At left: 20 s mainshock BP
Application to the Mw 9.1 mainshock Authors performed Rayleigh-wave BP at 13, 20, 30, 50, and 100 s ROTEN ET AL.: Prefer RAYLEIGH a three-stage BACK-PROJECTIONprogressive TOHOKU rupture model L02302 Stage 1: Local maximum in stack amplitude at 10 s just SW of JMA epicenter Stage 2: Dual peaks to SE (33 s) and NE (39 s) Stage 3: Northern segment peaks off Iwate at 61 s; Southern segment peaks at 124 s off near Fukushimaoki At left: 20 s mainshock BP
139 138 139 138 145 143 141 145 143 141 Application to the Mw 9.1 mainshock (a) Location and rupture times of 9 subfaults used for verification against seis-50, Authors performed Rayleigh-wave BP at 13,synthetic 20, 30, and 100 s an extended source. (b) Location and time of emitters identified from back-projection of Prefer a three-stage progressive rupture model mograms. 42 (a) 0.0 0.5 1.0 42 (b) Sm(s,τ) 38 38 36 36 1.36e+15 3.70e+14 86 s 138 50 s 40 138 40 Stage 1: Local maximum in stack amplitude at 10 s just SW of JMA epicenter Note difference in timing and distribution left: 50 s mainshock BP ame as Figure 3, but showing BP results for the Tohoku earthquake at T = 50 s.at Snapshots the peak time of the two identified Rayleigh wave emitters.
139 138 139 138 145 143 141 145 143 141 Application to the Mw 9.1 mainshock (a) Location and rupture times of 9 subfaults used for verification against seis-50, Authors performed Rayleigh-wave BP at 13,synthetic 20, 30, and 100 s an extended source. (b) Location and time of emitters identified from back-projection of Prefer a three-stage progressive rupture model mograms. 42 (a) 0.0 0.5 1.0 42 (b) Sm(s,τ) Stage 1: Local maximum in stack amplitude at 10 s just SW of JMA epicenter 40 40 36 36 1.36e+15 3.70e+14 86 s 138 50 s 38 138 38 Stage 2: Dual peaks to SE (33 s) and NE (39 s) Note difference in timing and distribution left: 50 s mainshock BP ame as Figure 3, but showing BP results for the Tohoku earthquake at T = 50 s.at Snapshots the peak time of the two identified Rayleigh wave emitters.
139 138 139 138 145 143 141 145 143 141 Application to the Mw 9.1 mainshock (a) Location and rupture times of 9 subfaults used for verification against seis-50, Authors performed Rayleigh-wave BP at 13,synthetic 20, 30, and 100 s an extended source. (b) Location and time of emitters identified from back-projection of Prefer a three-stage progressive rupture model mograms. 42 (a) 0.0 0.5 1.0 42 (b) Sm(s,τ) Stage 1: Local maximum in stack amplitude at 10 s just SW of JMA epicenter 40 40 36 36 1.36e+15 3.70e+14 86 s 138 50 s 38 138 38 Stage 2: Dual peaks to SE (33 s) and NE (39 s) Stage 3: Northern segment peaks off Iwate at 61 s; Southern segment peaks at 124 s off near Fukushimaoki Note difference in timing and distribution left: 50 s mainshock BP ame as Figure 3, but showing BP results for the Tohoku earthquake at T = 50 s.at Snapshots the peak time of the two identified Rayleigh wave emitters.
02302 ROTEN ET AL.: RAYLEIGH BACK-PROJECTION TOHOKU Application to the Mw 9.1 mainshock Authors select maxima of coherent ESE-WNW migrating peaks Stage 2 represents single broad peak at T=50,100s; at T=30s and shorter, two peaks are observed Stage 3 northern segment visible only at T=30s and shorter Stage 3 southern segment visible at all periods; single emitter at T=30-100s while second down-dip emitter observed at T=13,20s
Application to the Mw 9.1 mainshock Result: coarse, frequency-dependent view of rupture process Stage 1: rupture down-dip at 2-3.5 km/s; potential up-dip component leading to Stage 2 at ~35 s; Figure 4. (a) Rayleigh wave emitters identified from BP of radial-component seismograms of the Tohoku earthquake using periods of 100 s (triangles), 50 s (diamonds), 30 s (squares), 20 s (hexagons), and 13 s (spheres). Symbol colors indicate source time relative to the JMA hypocentral time, and symbol sizes (defined as the diameter of circumscribed circle) reflect stack amplitude. The yellow star shows the JMA epicenter. The location of the Japan
Application to the Mw 9.1 mainshock Result: coarse, frequency-dependent view of rupture process Stage 1: rupture down-dip at 2-3.5 km/s; potential up-dip component leading to Stage 2 at ~35 s; Stage 2: long-period surface waves show onset within 50-100 km of trench near Miyagi-oki (dual-peaked); propagates down-dip, peaking at 40-60 s (strongest Figure maxima 4. over (a) Rayleigh entire rupture); wave emitters identified from BP of radial-component seismograms of the Tohoku earthquake using periods of 100 s (triangles), 50 s (diamonds), 30 s (squares), 20 s (hexagons), and 13 s (spheres). Symbol colors indicate source time relative to the JMA hypocentral time, and symbol sizes (defined as the diameter of circumscribed circle) reflect stack amplitude. The yellow star shows the JMA epicenter. The location of the Japan
Application to the Mw 9.1 mainshock Result: coarse, frequency-dependent view of rupture process Stage 1: rupture down-dip at 2-3.5 km/s; potential up-dip component leading to Stage 2 at ~35 s; Stage 2: long-period surface waves show onset within 50-100 km of trench near Miyagi-oki (dual-peaked); propagates down-dip, peaking at 40-60 s (strongest Figure maxima 4. over (a) Rayleigh entire rupture); wave emitters identified from BP of radial-component seismograms of the Tohoku earthquake to north usingoff periods Iwate ofat 100 ~60 s (triangles), s (short period: 50 s (diamonds), T<=30s) Stage 3: bilateral - propagation and south off Fukushima-oki at 3085-100 s (squares), s. Inferred 20 s (hexagons), rupture velocity and 13 of s (spheres). 3-3.5 km/s; Symbol colors indicate source time relative to the JMA hypocentral time, and symbol sizes (defined as the diameter down-dip component at short periods (T<=20s) below Fukushima-oki at ~120 s of circumscribed circle) reflect stack amplitude. The yellow star shows the JMA epicenter. The location of the Japan
Application to the Mw 9.1 mainshock Result: coarse, frequency-dependent view of rupture process Stage 1: rupture down-dip at 2-3.5 km/s; potential up-dip component leading to Stage 2 at ~35 s; Stage 2: long-period surface waves show onset within 50-100 km of trench near Miyagi-oki (dual-peaked); propagates down-dip, peaking at 40-60 s (strongest Figure maxima 4. over (a) Rayleigh entire rupture); wave emitters identified from BP of radial-component seismograms of the Tohoku earthquake to north usingoff periods Iwate ofat 100 ~60 s (triangles), s (short period: 50 s (diamonds), T<=30s) Stage 3: bilateral - propagation and south off Fukushima-oki at 3085-100 s (squares), s. Inferred 20 s (hexagons), rupture velocity and 13 of s (spheres). 3-3.5 km/s; Symbol colors indicate source time relative to the JMA hypocentral time, and symbol sizes (defined as the diameter down-dip component at short periods (T<=20s) below Fukushima-oki at ~120 s of circumscribed circle) reflect stack amplitude. The yellow Total rupture star shows along the trench JMA epicenter. > 400km The location of the Japan
Comparison with other studies Distribution of rupture broadly consistent with finite source models in other studies - particularly maximum slip near epicenter [e.g. Simons et al., 2011] Progression (three-stage: 1. weak; 2. strong; 3. bilateral rupture) also broadly consistent with others [e.g. Koketsu et al., 2011] Peak moment release in most studies at 60-80 s, while maximum BP amplitudes at 33-55 s (freq. dependent) Stage 3 bilateral rupture velocities faster than reported elsewhere: 3-3.5 km/s vs. 2.5 km/s [e.g. Koketsu et al., 2011] Stage 3 southern segment stronger than implied by long-period finite fault inversions - source directivity effect on BP? (synthetic experiment suggest this is possible) Majority of long-period emitters within 100 km of trench, unlike shorter-period BP studies - highlights frequency-dependent radiation effect [e.g. Koper et al., 2011]
Conclusions Provides novel method for studying megathrust earthquake rupture Inferred distribution of rupture consistent with other studies using a variety of methods (GPS, waveform, tsunami) Highlights effect of frequency-dependent seismic wave radiation Slip inferred near trench more consistent with studies including longer-period seismic and tsunami waveforms Method is low-resolution and limited by the azimuthal coverage of the array, but potentially valuable as it is inexpensive