File name: Supplementary Information
Description: Supplementary Figures and Supplementary Tables
File name: Supplementary Movie 1
Description: Rotating animation showing the ray coverage and the subducting slab. Initially we show the P-wave ray paths for shots (gray) and local earthquakes (blue), with the location of the recording stations (black pyramids). The topography of the islands and of the interpreted slab surface are shown for reference. The second part of the animation shows the P-wave velocity anomaly draped on the slab surface. The location of the slab Moho is drawn as a semi-transparent surface assuming a constant 7 km crustal thickness. The outline of the overriding plate Moho is drawn in blue. The contact between the slab surface and the Moho is drawn as a green curve. The location of the backstop is drawn in red. Notice the spatial relationship between the location of seismicity and the local VP minimum at 50 km depth. The volume of the model is the same as shown in Fig. 7. The vertical extent is 162 km. The horizontal extent is 250 km in the arc-parallel direction and 280 km in the arc-perpendicular direction. The animation was prepared using Paraview and encoded using Ffmpeg.
File name: Supplementary Movie 2
Description: Rotating animation for 3D glasses. Same as Supplementary Movie 1, but encoded as a 3D red-cyan stereo anaglyph video to be viewed with common red-cyan 3D glasses.
File name: Peer Review File Description:
1
Supplementary Figure 1. Selection of starting VP/VS. a) Distribution of ratio of S-wave traveltimes (TS) to P-wave traveltimes (TP) for local earthquakes. The blue line marks the mean of the distribution. b) Plot of TS vs. TP (Wadati diagram). The blue line represents the linear regression with slope of 1.76. We chose this value to build our starting VP/VS model.
Supplementary Figure 2. Traveltime residuals. a) Histogram of P-wave traveltime residuals for starting model (gray), 40x50 km model (blue), 20x20 km model (red), and 15x15 km model (black). b) Same as a) for S-P traveltime residuals.
0 500 1000
N
1.7 1.8 1.9 2.0
TS/TP
a
0 10 20 30 40 50
TS(s)
0 10 20 30
TP(s)
b
0 1000 2000 3000 4000 5000
N
a b
−1.0 −0.5 0.0 0.5 1.0
P Residual (s)
0 100 200 300 400 500
−1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 S−P Residual (s)
Supplementary Figure 3. VP resolution. a) Spread function along four vertical cross-sections of the model.
The profiles are the same as those shown in Fig. 4. b) Diagonal element of resolution matrix (colors) and 70% contours of resolution kernel (contours). c) Derivative weight sum. See Methods for explanation. The dashed lines mark the location of the top and Moho of the slab and the Moho of the overriding plate.
a
3 0 20 40 60 80 100 120 140 160
b c
y =60 km
0 20 40 60 80 100 120 140 160
y =30 km
0 20 40 60 80 100 120 140 160
y =−40 km
0 20 40 60 80 100 120 140 160
−100
−50 0 50
100 100 50 0 −50 −100
y =−110 km
−100
−50 0 50 100
0.0 0.1 1.0
Vp diag. resol. element
0 1 2 3 4 5 6
Vp spread function
1 10 100 1000 10000
Vp derivative weight sum
3
Supplementary Figure 4. VP/VS resolution. Same as supplementary Fig. 3, for VP/VS model.
a b c
0.0 0.1 1.0
Vp/Vs diag. resol. element
0 1 2 3 4 5 6
Vp/Vs spread function
1 10 100 1000 10000
Vp/Vs derivative weight sum
3 0 20 40 60 80 100 120 140 160
y =60 km
0 20 40 60 80 100 120 140 160
y =30 km
0 20 40 60 80 100 120 140 160
y =−40 km
0 20 40 60 80 100 120 140 160
−100
−50 0 50
100 100 50 0 −50 −100
y =−110 km
−100
−50 0 50 100
Supplementary Figure 5. VP checkerboard tests. (a, e) Input and (b, c, d, f, g, h) recovered VP anomaly for several checkerboard tests. We show the recovery of 15x15 km and 30x30 km anomalies along horizontal sections at depths of 20, 50 and 80 km. The anomalies extend ~30 km in the vertical direction. The dashed black line marks the intersection of the slab surface with the horizontal plane of the section.
0 0
0 0
0
0 0
0
0
0
z = 20 km anomaly(%)
z = 9 km z = 40 km
−4 V−2P 0 2 4
50 0 50 100
z = 20 km z = 50 km z = 80 km
Input anomaly
Input anomaly
0 0
0 0
0
0
0
0 0 0
0
0
0
−150
−100
−50 0 50 100
a b
0 0 0 0
0
0 0
0 0
0
0
0
0
c
0 0 0
0 0 0
d
0 0
0
0
0
0
0
−150
−100
−50 0 50 100
− e
0
0
h
−50 0 50 100
0
0 0
0
g
−50 0 50 100
0 0
0 0
0 0
0
f
−50 0 50 100
a b c d
e f g h
−4 −2 0 2 4
VP/VSanomaly (%)
−150
−100
−50 0 50 100
z = 9 km z = 20 km z = 40 km
−100
−50 0 50 100
z = 20 km z = 50 km z = 80 km
Input anomaly
Input anomaly
0 0
0
0 0
0
0 0 0
00 0
0
0 0
0 0 0
0 0 0
0
0
0
0 0
0 0 0
0
0 0
0
0 0
0
0
0
0 0
0 0
0 0
0
0 0 0 0
0
0
0 0
0
5
Supplementary Figure 7. Slab anomaly recovery test 1. Vertical sections at different locations in the model showing input and recovered VP model, VP anomaly and VP/VS. In this test the slab crust low-VP
anomaly terminates at 90 km depth.
Input anomalyRecovered anomalyInput anomalyRecovered anomaly
3 4 5 6 7 8
Vp(km/s)
−8 −4 0 4 8
Vpanomaly (%)
1.68 1.72 1.76 1.80 1.84 1.88 Vp/Vs
7.5
8
8.5 0
50
100
150
−4
0 0
0 0
1.76
y =10 km
7
7.5 8
8.5 0
50
100
150
−400 1.76
1.8
y =−110 km
7.5 8 0
50
100
150
−4
00
1.76
y =10 km
8 0
50
100
150
1.76
y =−110 km
−100
−50 0 50
100 100 50 0 −50 −100 100 50 0 −50 −100
a b c
d e f
g h i
j k l
Supplementary Figure 8. Slab anomaly recovery test 2. Vertical sections at different locations in the model showing input and recovered VP model, VP anomaly and VP/VS. In this test the slab crust low-VP
anomaly terminates at 120 km depth.
3 4 5 6 7 8
Vp(km/s)
−8 −4 0 4 8
Vpanomaly (%)
1.68 1.72 1.76 1.80 1.84 1.88 Vp/Vs
7.5 8
8.5 0
50
100
150
−4
−4 0
0
0
0 1.76
y =10 km
7
7.5
8 0
50
100
150
−4
00 1.76
1.8
y =−110 km
7.5
8 0
50
100
150
−4 0
0
0 0
1.76
y =10 km
7
8 0
50
100
150
00 1.76
1.761.8
y =−110 km
−100
−50 0 50
100 100 50 0 −50 −100 100 50 0 −50 −100
Input anomalyRecovered anomalyInput anomalyRecovered anomaly
a b c
d e f
g h i
j k l
7
Supplementary Figure 9. Slab VP checkerboard tests. Input and recovered VP anomaly for several slab checkerboard tests. We show the anomaly both along the slab top surface and on a vertical section. The islands are colored in black for reference.
−4 −2 0 2 4
Slab surface Vp anomaly (%)
−150
−100
−50 0 50
100 Input Recovered
0 50 100 150
Input Recovered
a
Input Recovered
Input Recovered
b
−150
−100
−50 0 50 100
0 50 100 150
Input Recovered
Input Recovered
c
Input Recovered
Input Recovered
d
−150
−100
−50 0 50 100
−100
−50 0 50 100 0
50 100
150
−100
−50 0 50 100
Input Recovered
Input Recovered
e
−100
−50 0 50 100
−100
−50 0 50 100
Input Recovered
Input Recovered
f
Supplementary Figure 10. Slab VP/VS checkerboard tests. Input and recovered VP/VS anomaly for several slab checkerboard tests. We show the anomaly both along the slab top surface and on a vertical section. The islands are colored in black for reference.
−4 −2 0 2 4
−150
−100
−50 0 50 100
−150
−100
−50 0 50 100 0 50 100 150
−100
−50 0 50 100
−100
−50 0 50 100
−100
−50 0 50 100 0
50 100
150
−100
−50 0 50 100
Input Recovered
a
Input Recovered
b
c d
Input Recovered Input Recovered
Input Recovered Input Recovered
Input Recovered Input Recovered
Slab surface Vp/Vs anomaly (%)
9
Supplementary table 1: Summary of inversion parameters and residual statistics.
Inversion step
Minimum horizontal spacing
Minimum vertical spacing
Damping parameters
RMS residual (s) P Data variance
S-P data variance
tot P S-P
σ2 % σ2 %
1d model n. a. n. a. n. a. 0.73 0.71 0.90
Starting model 40 km 5 km n. a. 0.55 0.47 0.88 0.218 100 0.220 100 40x50 km
model 40 km 5 km VS: 2000
VP/VS: 2000 0.37 0.35 0.49 0.095 44 0.130 59 20x20 km
model 20 km 5 km VP: 500
VP/Vs: 500 0.28 0.26 0.40 0.048 22 0.071 32 15x15 km
final model 15 km 3 km VP: 250
VP/VS: 125 0.26 0.24 0.39 0.038 17 0.067 30