GEOPHYSICAL RESEARCH LETTERS

Supporting Information for “Uppermost mantle structurebeneath eastern China and its surroundings from Pn and Sntomography”

Weijia Sun1 and B.L.N. Kennett2

1 Key Laboratory of Earth and Planetary Physics, Institute of Geology and

Geophysics, Chinese Academy of Sciences, Beijing 100029, China.

2 Research School of Earth Sciences, Australian National University, Canberra,

ACT 2601, Australia.

Contents of this file

1. Data selection

2. The FMTOMO procedure

3. Resolution tests

4. The influence of relocation

5. Figure S1 - resolution tests

6. Figure S2 - residual histograms

2 SUN AND KENNETT: MANTLE VP , VS IN EASTERN CHINA

S1. Data selection

As noted in the main paper we have applied strict selection criteria to the bulletin arrivals times for

both P and S.

In the region closest to the source the crustal phases Pg, Sg will be the first arrivals for the particular

wavetypes, but beyond 1.8◦ epicentral distance it is expected that Pn, Sn becomes the first arrival, since

the average Moho depth is about 33 km [Li et al., 2014]. The Pn or Sn diving waves can exit from

the base of thin lithosphere, leading to significant changes in the frequency spectrum for epicentral

distances greater than 15◦ even for thick lithosphere. The thin lithosphere in parts of eastern China

[Hearn et al., 2008; Zhao et al., 2013], means that Pn and Sn are only trapped in the lithosphere out to

12◦. We have therefore limited the epicentral distance for the use of the Pn and Sn travel times to the

range 1.8◦ – 12◦, for the inversions for P and S velocity structure in the uppermost mantle.

The relocation step is also used as a filter on the arrival times to be used in the inversions. After

relocation we only retain events for which the focal depths are shallower than 2 km less than the Moho

depth. We also discard events for which there is a very large epicentral shift on relocation (>1◦). To

avoid erroneous readings we exclude any paths for which Pn or Sn travel-time deviations are larger than

8 s relative to the ak135 model [Kennett et al., 1995] using the relocated hypocentres.

S2. The FMTOMO procedure

The FMTOMO package is designed to invert multiple classes of body wave datasets, including re-

fracted waves, reflected waves, local and regional and teleseismic events, and data from active sources

[Rawlinson and Urvoy, 2006]. The multi-stage fast marching method is employed to solve the forward

problem of travel-time prediction. Assuming local linearity, a subspace inversion scheme is used to

optimise the objective function of model parameters and observed data, which is a iterative non-linear

SUN AND KENNETT: MANTLE VP , VS IN EASTERN CHINA 3

scheme. The model parameters are defined on regular grids in spherical coordinates, which are quite

suitable for regional and global tomography, because of minimising travel-time prediction errors at

these scales.

The FMTOMO method is quite suitable to solve large tomographic problems, and is computationally

efficient and robust for our study of Pn and Sn across the eastern China. The Pn and Sn phases are

represented diving waves in the uppermost mantle returned by interaction with the velocity gradients.

The FMTOMO procedure takes account of the variations of the Moho in the inversion.

Our approach to determine the 3-D wavespeeds can be summarised as:

• Establish a 3-D initial model; in this case we use the CRUST1.0 model [Laske et al., 2013] for the

crust and Moho and the SL2013sv model [Schaeffer and Lebedev, 2013] in the mantle.

• Relocate all sources in the 3-D initial model, and extract Pn and Sn residuals.

• FMTOMO inversion for an inversion domain down to 150 km. The procedure was run for multiple

iterations, and the model with the first minimum in misfit selected.

S3. Resolution tests

The ability of the available data to identify small heterogeneities can be assessed with the aid of a

formal resolution test. A 2◦ × 2◦ pattern of 2% negative and positive velocity perturbations is imposed

on the initial P model and ±4% velocity anomalies are imposed on the S model. To be more realistic,

we set two individual flat layers at depths of 50 km and 60 km for both P and S. Figure S1 shows the

recovered checkerboard resolution test with relative model to the initial 3-D model at depths of 50 km

and 60 km.

For P waves, the checkerboard patterns are generally recovered quite well with the imposed pattern

at 2◦ × 2◦ at both depths of 50 km and 60 km in the continental regions. The exceptions are in eastern

4 SUN AND KENNETT: MANTLE VP , VS IN EASTERN CHINA

Mongolia, due to infrequent events and rather sparse stations, and a small part of Yangtze Craton centred

at (108◦E,26◦N) where only weaker amplitudes can be recovered.

The recovered checkerboard for S wave is quite good, and is similar to that for P at a depth of 50

km. The recovered S velocity anomalies at 60 km depth is generally good, but with a little weaker

amplitudes than those of S at 50 km depth and P waves. This difference is linked to differences in the

ray path for P and S associated with the varying wavespeed distribution, since the P and S velocities

are dominated by different factors. For example, temperature has much more prominent influence on S

velocities in the crust and mantle than P velocities.

As might be expected from the similar pattern of available paths (Figure 2) we are able to achieve

comparable resolution for P and S. Features at a scale of 2◦ × 2◦ or larger should be well represented

for both P and S.

Construction of a direct resolution test for the Vp/Vs ratio is difficult, because the Vp/Vs ratio is ob-

tained by division of Vp by Vs rather than inverted directly. Figure S1e and S1f show indirect resolution

results for the Vp/Vs ratio at depth of 50 km and 60 km for comparison with the direct results for P and

S velocities. The background Vp/Vs is obtained by division of the background Vp by the background

V s, while the perturbed Vp/Vs is given by division of the inverted Vp by Vs. The checkerboard pattern

of Vp/Vs is then obtained by subtracting the background from the perturbation. Since the variation of

Vp/Vs is not proportional to those of Vp and Vs, the procedure cannot completely reflect the resolution

ability of the available data, and so we have small amplitudes of the relative Vp/Vs ratio in Figure S1e

and S1f. For a given Vp = 8.0km/s and Vs = 4.0km/s with perturbations of +0.16km/s for both P

and S, the perturbed Vp/Vs would be −0.027, so that the polarities are reversed. This can be seen by

careful comparison of Figure S1e-f with Figure S1a-d.

SUN AND KENNETT: MANTLE VP , VS IN EASTERN CHINA 5

Although the indirect resolution of the Vp/Vs ratio has been recovered in a good pattern, we still

suggest it is necessary to evaluate the resolution of the Vp/Vs ratio by comparing those for Vp and Vs.

S4. The influence of relocation

The event information for the arrival times we have collected come form location with a global or

local 1-D model. To put all observations on a common basis and to correct for much of the influence of

lateral heterogeneities in the Earth, we carry out a relocation of all events using the initial 3-D model,

before inverting the Pn and Sn travel-times.

This relocation step helps to reduce biases in the travel time distribution and so provide a good

starting point for the FMTOMO inversion. A convenient way of illustrating the effect of the relocation

is to examine the histograms of travel-time residuals. In Figure 2 we illustrate the results for station

GYA (26.46◦N, 106.66◦E) in southern China, an open station at Incorporated Research Institutions for

Seismology (IRIS). This station has a large number of arrivals, with a considerable spread in residuals

associated with paths sampling rather different tectonic provinces. In Figure S2 we show the Pn and Sn

residuals before and after relocation. At first sight, the Pn residuals from the original bulletin (Figure

S2a) seem good, but we note that the residuals are dominantly positive residuals centred at 3 sec.

After relocation with the 3-D initial model, the Pn residual distribution (Figure S2b) is much more

concentrated with a smaller residuals centred at 1 sec. We have similar results for Sn with residuals

centred at 2 sec from the original bulletin (Figure S2c) to around 0 sec after relocation (Figure S2d),

though in this case the spread of residuals does not reduce much. Similar improvements are observed

at most stations, which indicates the value of relocation before travel-time inversion.

6 SUN AND KENNETT: MANTLE VP , VS IN EASTERN CHINA

References

Hearn, T. M., Wang, S. Y., Pei, S. P., Xu, Z. H., Ni, J. F., Yu, Y. X., 2008. Seismic amplitude tomography

for crustal attenuation beneath China. Geophys. J. Int. 174 (1), 223–234.

Kennett, B. L. N., Engdahl, E. R., Buland, R., 1995. Constraints on seismic velocities in the Earth from

travel-times. Geophys. J. Int. 122 (1), 108–124.

Li, Y. H., Gao, M. T., Wu, Q. J., 2014. Crustal thickness map of the Chinese mainland from teleseismic

receiver functions. Tectonophysics 611, 51–60.

Laske, G., Masters, G., Ma, Z., Pasyanos, M., 2013. Update on CRUST1.0 - a 1-degree global model

of Earths crust. In: Geophys. Res. Abstracts. Vol. 15. p. 2658.

Schaeffer, A. J., Lebedev, S., 2013. Global shear speed structure of the upper mantle and transition

zone. Geophys. J. Int., 194 (1), 417–449.

Rawlinson, N., and M. Urvoy (2006), Simultaneous inversion of active and passive source datasets for

3-D seismic structure with application to Tasmania, Geophys. Res. Lett., 33(24), L24313.

Zhao, L. F., Xie, X. B., Wang, W. M., Zhang, J. H., Yao, Z. X., 2013. Crustal Lg attenuation within the

north China craton and its surrounding regions. Geophys. J. Int. 195 (1), 513–531.

SUN AND KENNETT: MANTLE VP , VS IN EASTERN CHINA 7

c)

d)

a)

b)

-0.12 -0.06 0.00 0.06 0.12

P Velocity [km/s]

-0.12 -0.06 0.00 0.06 0.12

S Velocity [km/s]

60.0 km

100

100

110

110

120

120

130

130

140

140

20 20

25 25

30 30

35 35

40 40

45 45

50 50

55 55

60 60

50.0 km

100

100

110

110

120

120

130

130

140

140

20 20

25 25

30 30

35 35

40 40

45 45

50 50

55 55

60 60

60.0 km

100

100

110

110

120

120

130

130

140

140

20 20

25 25

30 30

35 35

40 40

45 45

50 50

55 55

60 60

50.0 km

100

100

110

110

120

120

130

130

140

140

20 20

25 25

30 30

35 35

40 40

45 45

50 50

55 55

60 60

-0.06 -0.03 0.00 0.03 0.06

Vp/Vs

e)

f)

60.0 km

100

100

110

110

120

120

130

130

140

140

20 20

25 25

30 30

35 35

40 40

45 45

50 50

55 55

60 60

50.0 km

100

100

110

110

120

120

130

130

140

140

20 20

25 25

30 30

35 35

40 40

45 45

50 50

55 55

60 60

Figure S1. Results of checkerboard resolution test for Pn (a-b) and Sn (c-d) tomography at depths

of 50 km (top panel) and 60 km (bottom panel). The 2◦ × 2◦ pattern of positive and negative velocity

anomalies is imposed on the initial models for both P and S. The indirect resolution test for Vp/Vs is

shown at depth of 50 km (e) and 60 km (f).

8 SUN AND KENNETT: MANTLE VP , VS IN EASTERN CHINA

a) b)

c) d)

-10 -8 -6 -4 -2 0 2 4 6 8 10

Residuals [s]

0

5

10

15

20

25

30

Co

un

t

-10 -8 -6 -4 -2 0 2 4 6 8 10

Residuals [s]

-10 -8 -6 -4 -2 0 2 4 6 8 10

Residuals [s]

0

5

10

15

20

25

30

Co

un

t

-10 -8 -6 -4 -2 0 2 4 6 8 10

Residuals [s]

P P

S S

Catalog Relocation

Figure S2. Travel-time residuals for the Pn and Sn arrivals at station GYA located at (26.46N ,

106.66E), with the theoretical travel-time predicted with the ak135 model: (a) Pn residuals from the

original CEDC catalog, and (b) Pn residual pattern after relocation with the 3D initial model. The

corresponding Sn residuals before and after the relocation procedure are shown in (c) and (d).