Supplement Material
Figure S1. Regional waveform comparison of the foreshock, include, data, synthetics of
ref2, ref3 and 3D model
Regional waveform of the foreshock includes data (black), synthetics generated by 1D velocity
models (ref2/red and ref3/blue) and 3D velocity model (green). Stations are divided into two
groups according to paths as displayed in A (ref2, red) and B (ref3, blue). The true amplitude of
seismograms are plotted with scaling factors for different stations for better display purposes.
Seismograms are all filtered to 50s and longer.
Figure S2. Teleseismic P-wave inversion result for the Mw6.2 aftershock
Teleseismic P-wave fits for the Mw6.2 aftershock, both data and synthetics are filtered between
6s and 50s. The name of station is indicated at the beginning of each waveform pair, the number
about station name is epicenter distance in km and the number below is azimuth. The mechanism
of the inversion is shown on the top at the best depth (46km).
Figure S3. Regional waveform inversion result for the Mw6.2 aftershock
(A) Cut-And-Paste (CAP) inversion result for the Mw6.2 aftershock. Here the waveform fits for
the best depth (40km) are shown. Station names are indicated at the beginning of the record
along with distance in km (upper) and azimuth in degree (lower). The first number below each
waveform pair is the time shift needed to aligned synthetic and data, and the second number is
waveform cross-correlation coefficient in percentage. A positive time shift means the velocity
model is too fast. The red beach ball at the top shows the lower hemisphere projection of the
mechanism (13°/77°/-10°/6.26 for strike/dip/rake/Mw) and the dots indicate the station
1
projections on the lower hemisphere, each according to its first arrival P wave take-off angle. (B)
Depth resolution of the inversion.
Figure S4. Depth resolution of regional data
Sensitivity tests for source depth on three regional stations. The three component velocity
synthetics (red) are generated at different depth and compared with the data (black), both filtered
to 50s and longer. As an example, the Pnl wave and surface wave windows are indicated in the
radial component record section at station PSI. Note the different sensitivities between Pnl waves,
Rayleigh waves and Love waves.
Figure S5. Checker-board tests
Checker-board like slip models with different asperity dimensions (A,B) are used to generate
synthetic data and then inverted by the inversion setup we use for the real data. The inverted slip
models are presented in C and D, respectively.
Figure S6. Smoothness factor for joint inversion
Weighted errors are plotted against smoothness coefficients for the joint inversion of regional
and teleseismic data. There is not much change in error when smoothness coefficient is less than
0.03.
Figure S7. Decomposition of regional waveform fits
Decomposition of regional synthetics (gray) into the contribution of segment F1 (brown), F2
(red) and F3 (green).
2
Figure S8. Slip model on F1 and F2 for regional only inversion, teleseismic only inversion
and Joint inversion
Depth profiles of slip models for the regional only inversion (upper), teleseismic only inversion
(middle) and joint inversion (lower). Here we only display segment F1 and F2.
Figure S9. Representative teleseismic P-wave fits from regional data only model and
teleseismic only model
Selected teleseismic P-wave fits, (A) prediction from regional data only inversion, (B)
teleseismic only inversion. Data are displayed in black and synthetics are in red, station name is
indicated at the beginning of each trace with epicenter distance in degree (lower) and azimuth
(upper). Maximum amplitude of data in micro-meter is shown at the end of the seismogram.
Figure S10. Decomposition of teleseismic waveform fits
Decomposition of teleseismic synthetics (gray) into the contribution of fault segment F1
(brown), F2 (red) and F3 (green). The SH components with broader waveform are high-lighted
in the box.
Animation 1. Inversion process of the Mw8.6 mainshock
The slip distribution on fault segment F1 (top), F2 (middle) and F3 (bottom) are displayed from
the beginning of the iteration of simulated annealing inversion.
Animation 2. Inversion process of the Mw8.2 aftershock
Similar as Animation 1 for the Mw8.2 aftershock.
3