Subducting Slab Ultra-Slow Velocity Layer Coincident with ... · We meshed the 3-D slab geometry underneath Central Mexico (S1) and made a direct 2-D profile from the source to the

www.sciencemag.org/cgi/content/full/324/5926/502/DC1

Supporting Online Material for

Subducting Slab Ultra-Slow Velocity Layer Coincident with Silent Earthquakes in Southern Mexico

Teh-Ru Alex Song,* Don. V. Helmberger, Michael R. Brudzinski, Rob. W. Clayton, Paul Davis, Xyoli Pérez-Campos, Shri K. Singh

*To whom correspondence should be addressed. E-mail: [email protected]

Published 24 April 2009, Science 324, 502 (2009)

DOI: 10.1126/science.1167595

This PDF file includes:

Materials and Methods

SOM Text

Figs. S1 to S11

Table S1

References

Supplementary Online Information

Subducting Slab Ultra-Slow Velocity Layer Coincident with Silent

Earthquake in Southern Mexico

Teh-Ru Alex Song1, Don. V. Helmberger2, Michael R. Brudzinski3, Rob. W.

Clayton2, Paul Davis4, Xyoli Pérez-Campos5, Shri. K. Singh5

1. Department of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad Branch

Road, NW, Washington, DC 20015, USA

2. Seismological Laboratory, Division of Geological and Planetary Science, California Institute of

Technology, 1200 E. California Blvd, Pasadena, California 91125, USA

3. Geology Department, Miami University, 114 Shideler Hall, Oxford, OH 45056, USA

4. Department of Earth and Space Sciences, Center of Embedded Network Systems (CENS),

University of California, Los Angeles, 595 Charles Young Drive East, Los Ageles, CA 90095-

1567

5. Instituto de Geofísica, Universidad Nacional Autónoma de México, Circuito de la

Investigación Científica s/n, Ciudad Universitaria, 04510 México D. F

*To whom correspondence should be addressed. E-mail: [email protected]

Method Summary

We meshed the 3-D slab geometry underneath Central Mexico (S1) and made a direct 2-

D profile from the source to the receiver so that it directly mimics the slab geometry

between the source and receiver. We numerically propagate the wavefield using a 2D

finite-difference scheme (S2) from the source to the receiver. We set the background

velocity in the mantle at 8.0 km/s for P wave and 4.5 km/s for S wave, respectively (Fig.

S11A). We assume that the S wave and P wave velocity of the slab are 6% and 3% faster

than the background velocity, respectively. In addition, we also include a 1-D basin

structure directly below the station UNM located in the Mexico Valley (S3). Such a slow

structure underneath the receiver only changes the particle motion of the P wave from

linear into semi-elliptical (S4), but it does not introduce the anomalous pulse shown in the

data. The assumption that seismicity defined slab surface interface is not strictly valid

particularly if these events are within the slab. In our modelling, we slightly adjust the

depth and the angle of the slab surface, which can be constrained by the timing of the

converted SP wave (Fig. S11B). Typically, we have to decrease the dip of the slab near

the source by 6° to explain data from events near Guerrero, which is consistent with

recent receiver function analysis (S5). The source depth is checked against the teleseismic

pP and sP waves. Uncertainties in the earthquake focal mechanism, location and velocity

directly above the USL can slightly modify our estimate on the USL (Fig. S11C-E). The

USL with dlnVs/dlnVp of 2 is consistent with the data, but less well constrained due to

uncertainties in the velocity directly above the USL (Fig. S11F). Because the SP wave is

primarily sensitive to the S wave velocity anomaly, we summarize our model

emphasizing S velocity of the USL and its layer thickness.

Discussions on the permeability near the top of the slab

This interpretation relies on the presence of a low permeability layer that seals the

HPFP layer directly below it, as well as a permeability increase near the 450°C isotherm

where the HPFP layer disappears. While intrinsic permeability is grain-size dependent

(S6), the fluid flow also depends on the hydraulic gradient and the fluid generation

process. We speculate that fine-grained blueschist in the subducting crust is intrinsically

less permeable than the coarse-grained eclogite that forms near 450°C at depths near 40-

50 km. Reduction in grain size induced by shear along the plate interface (S7) could

effectively seal fluids to form the HPFP where blueschist is present. Dehydration reaction

can locally increase the hydraulic gradient near the phase transition and facilitate the flow

movement. Continental mantle is typically well serpentinized relative to the continental

crust near the wedge corner (S8) (see also velocity profile in Fig. S11A). It is possible

that changes from low to high permeability across the continental Moho also contributes

to the termination of the HPFP and the occurrence of the NVT.

A working model of for the spatial-temporal variation in seismicity and SSEs in

southern Mexico

The down-dip extent of the HPFP layer and its along-strike variation is probably a

key to explain the duration and recurrence interval of the SSEs. While long-term along-

strike segmentation in the occurrences of great earthquakes exists (S9, S10), our result

clearly demonstrates that the occurrences of the SSEs seem also persistent in time but

spatially varying. However, spatial sampling of the HPFP layer is dictated by the

locations of intra-slab events, which are not even along southern Mexico. We find almost

no intra-slab events directly below the transition zone in the west Oaxaca (98º-99ºW,

16.9º-17.4ºN, see also Fig. S10A) in the last 15 years or so when a few large thrust

earthquakes occur. Instead, most intra-slab events are located below the transition zone

where seismic gaps are present in the seismogenic zone (Fig. 1), similar to observations

in other coupled subduction zones (S11). Furthermore, it has been suggested that intra-

slab events may trigger the SSEs in the transition zone and subsequently megathrust

earthquake (S12, S13). We purpose a stress feedback system where megathrust

earthquakes, intra-slab events and SSEs are linked (Fig. S10B). In such a system,

temporal variation in the degree of coupling at the plate interface manifests in the

occurrences of intra-slab earthquakes and SSEs and it may provide a way to monitor mid-

term seismic potential in subduction zones. Recurrent SSEs near Guerrero are likely to

stress the seismic gap close to failure (S12). Our hypothesis suggests that SSEs do not

currently exist beneath western Oaxaca, or are much smaller than the detection limit of

current GPS instrumentation. New GPS instrumentation in western Oaxaca (S14) will

help test this hypothesis. Currently, there is no evidence showing significant transient

slow slip beneath western Oaxaca during the 2006-2007 slow slip event (S14).

2 1 0 1 2 3 4 5 6

Time (sec)

32_UNM

A_MIXC

31_MIXC

M_UNM

M_MIXC

2 1 0 1 2 3 4 5 6

Time (sec)

A_SAME

31_SAME

M_SAME

----

Song et al, Supp. Figure 1

B

A

104˚W 103˚W 102˚W 101˚W 100˚W 99˚W 98˚W 97˚W 96˚W 95˚W

16˚N

17˚N

18˚N

19˚N

20˚N

21˚N

0 50 100

km

E2AM

TMVB

20 km

40 km

60 km

80 km

100 km

UNMPTRPESTAMULUMIXCTEPETONI

TEMP

IXCATIBL

OCOL

SAME

PASU

TONNTECA

PSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQKM67

1

2

3

4

5

6

7

8

10

11

12

13

14

15

161719

20

212223

24

25

2627

28

29

303132

3334

3536

37

3839

40

M

Mexico

CO

TMVBRA

PA

NA

Orozco F.Z 5.8 cm/yr

A

18

9

13

-3 -2 -1 0 1 2 3 4 5 6Time [sec]

Event 22, mb = 5.1

0.01-4 Hz

0.01-2 Hz

0.01-1 Hz

0.01-0.6 Hz

-3 -2 -1 0 1 2 3 4 5 6Time [sec]

Event 16, mb = 4.6

0.01-4 Hz

0.01-2 Hz

0.01-1 Hz

0.01-0.6 Hz

-3 -2 -1 0 1 2 3 4 5 6Time [sec]

Event 30, mb = 4.6

0.01-4 Hz

0.01-2 Hz

0.01-1 Hz

0.01-0.6 Hz

Song et al, Supp. Figure 1 (cont)

C

A A AB B B

Fig. S1 Robustness of observations. (A) We compare P waveforms recorded by the

MASE station MIXC and the nearby permanent GEOSCOPE station UNM from four

different events, event M (060811, long: 101.061ºW, lat: 18.561ºN, depth: 60.1 km, mb =

5.9) and its aftershock (event no. 34, see also Table S1), event A (051214, long:

100.97ºW, lat: 18.661ºN, depth: 81.9 km, mb = 5.0) and event No. 31 (see also Table S1).

See caption of Fig, 1 for details. (B) The redundancy of these waveforms indicates that

our observations are robust and are not due to the complexity of the earthquakes or

instrumentation. (C) P waves from event 16, 22 and 30 recorded at station UNM (from

left to right). The data are displayed at different frequency band (0.01-4 Hz, 0.01-2 Hz,

0.01-1 Hz, 0.01-0.6 Hz). We choose to model data at frequency band 0.01-0.6 Hz to

enhance the coherency of the SP arrivals (pulse A and B) and to insure the uniformity of

the modelling.

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

dVs = -50%

dVs = -40%

dVs = -30%

dVs = -20%

dVs = -10%

HUSL = 3 km

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

HUSL = 12 km

HUSL = 10 km

HUSL = 8 km

HUSL = 6 km

HUSL = 3 km

dVs = -40%

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

DLVL = 7 km

dVp = -11%

dVp = -9%

dVp = -7%

dVp = -5%

dVp = -3%

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

dVp = -7%

HUSL = 11 km

HUSL = 9 km

HUSL = 7 km

HUSL = 5 km

HUSL = 3 km

A

B

USL

LVL

HUSL

DLVL

dVp

dVs dlnVs/dlnVp = 2

dlnVs/dlnVp = 2

----

----

Song et al., Supp. Figure 2

C BA

A B C

USL

LVLPS

BAC

C C

Fig. S2 Sensitivity tests on the thickness and S wave velocity reduction of the USL. The

amplitude of the pulse A and pulse B increases with the velocity reduction of the USL,

whereas the lag-time of the pulse B increases with the thickness of the USL, HUSL.The

parameter space is explored by searching for synthetic waveforms that can match those

presented in Fig. S1. In (A) and (B), we assume dlnVs/dlnVp = 2.

USL

LVL

intra-slab event (depth = 60 km)

DUSL

DLVL

HLVL = DUSL + DUSL

DLVL

DU

SL

d = 9 km d = 11 kmd = 7 kmd = 5 kmd = 3 km

d = 5 km

d = 7 km

d = 9 km

d = 11 km

d = 13 km

d = 1 km

HLVL = 12 km

HLVL = 14 km

HLVL = 16 km

HLVL = 18 km

HLVL = 20 km

HLVL = 22 km

HLVL = 10 km

HLVL = 8 km

HLVL = 6 km

8 sec

HLVL

dVs = -39%, dlnVs/dlnVp = 2

dVp = -7 %, dlnVs/dlnVp = 2


Fig. S3 We illustrate how the P wavetrain, including up-going SP converted arrival and

down-going turning P arrival, varies with the location of the earthquake inside the LVL

(top panel). Assuming HUSL= 3 km, we show how the waveforms vary with two model

parameters, DUSL and DLVL. DUSL is the separation between the source and the USL and

DLVL is the separation between the source and the bottom of the LVL, whereas HLVL is

the thickness of the LVL. For a fixed DUSL, the amplitude of pulse C increases with DLVL.

For a fixed DLVL, the timing and amplitude of pulse A and pulse B increase with DUSL.

For a fixed HLVL, we observe small but recognizable interferences between SP arrivals

and reflected P arrivals. Nearly all the blue waveforms displayed in Fig. 1 can be

modelled from this library of synthetics.

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

DATA

TONI

TEPE

MIXC

MULU

ESTA

PTRP

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

SYNX = 0.88

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

DATA

TEMP

TIBL

IXCA

OCOL

SAME

PASU

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

SYNX = 0.93

A B

30 35 40 45 50 55

Time (sec)

Syn_USL

Data_MIXC

Syn_noUSL

sPP SPUSLC

--------

Song et ai, Supp. Figure 4

Fig. S4 Comparison between data (black) and synthetics (red). In (A), data from event M

recorded by the MASE are modelled using NEIC CMT solution (strike=88º, dip=29º,

rake=-99º). Synthetics are computed with a USL of 3 km, velocity reduction dVs of -

39%. In addition, the P wave reduction dVp of the LVL is -7%, whereas the thickness of

the LVL is about 22 km. In (B), the synthetics are computed with a USL of 3 km,

velocity reduction dVs of -20%. The LVL is identical to (A). X is the mean coefficient of

cross-correlation between the data and synthetics. Although we model the first 6 secs P

waveforms, our model predicts the data reasonably well after 6 secs. The energy arriving

after 6 secs is generally weaker and it possibly bounces multiple times within the USL

before converting to P wave and leaking out of the USL. In addition, converted wave

from the continental Moho can also arrive at this later time window. In spite of this

complexity, our model predicts the data satisfactorily. In (C), we show that data from

event M (see also Fig. S1A) recorded at MIXC are well modelled including later arrivals,

which supports the validity of our model. Note the timing of sP wave is not well

predicted probably because of lateral heterogeneities in the shallow continental crust.

� 2� 10 1 2 3 4 5 6 7 8 9

Time (sec)

Event 10

� 2� 10 1 2 3 4 5 6 7 8 9

Time (sec)

Event 14

� 2� 10 1 2 3 4 5 6 7 8 9

Time (sec)

Event 29

� 2� 10 1 2 3 4 5 6 7 8 9

Time (sec)

Event 13

UNM

UNM

Syn

Syn

A B

DC

--- -

----


Fig. S5 Comparison between data and synthetics. (A) event 10 (50 km, strike=116º,

dip=83º, rake=-119º), (B) event 14 (52 km, strike=101º, dip=17º, rake=86º), (C) event 29

(51 km, strike=134º, dip=38º, rake=96º), (D) event 13 (55 km, strike=289º, dip=34º,

rake=-76º). The source depths of these events are relocated by the stacking teleseismic

depth phases from Yellowknife array. We can model these events by including a USL of

3 km with slightly different S velocity reduction dVs. In (A) and (C), dVs = -55%. In (B),

dVs = -45%, In (D), dVs = -30%. A LVL of 13 km with dVp = -5% is also included in

the modelling. In all cases, focal mechanisms are inverted from regional CMT as kindly

provided by K. Singh at UNAM. Note dlnVs/dlnVp is fixed at 2. In general, a focal

mechanism with predominantly 45º dip-slip component can reproduce the waveform,

which is consistent with most of the focal mechanisms determined by the global centroid

moment tensor.

� 2 � 1 0 1 2 3 4 5 6

Time (sec)

dVs = -42%

dVs = -33%

dVs = -24%

dVs = -15%

dVs = -6%

USL = 4 km

Event 2

� 2 � 1 0 1 2 3 4 5 6

Time (sec)

dVs = -42%

dVs = -33%

dVs = -24%

dVs = -15%

dVs = -6%

USL = 6 km

Event 4

----


Fig. S6 Modelling of P wavetrain of while circles for event 2 and event 4 (strike=129º,

dip=50º, rake=142º) (see also Fig. 1). Both events are located near Oaxaca. The USL

with a velocity reduction dVs of -40% (2.7 km/s) do not reproduce the data well. These

data are better explained by the USL with a velocity reduction dVs of -15% (3.8 km/s).

B

� 5 0 5 10 15 20 25

T� TP [sec]

Event 9, 99.3765oW,18.0166oN,72.7 km

P pP sPsUSLP sMP

YKB1

YKB2

YKB3

YKB4

YKR9

YKR8

YKR6

YKB6

YKR5

YKR4

YKR3

YKB7

YKR2

YKR1

YKB8

YKB9

YKB0

350.141

350.152

350.164

350.175

350.216

350.186

350.128

350.198

350.098

350.068

350.039

350.208

350.01

349.98

350.22

350.231

350.241

45.602

45.6231

45.6455

45.6665

45.6793

45.6874

45.7031

45.7093

45.711

45.7186

45.7267

45.7305

45.7345

45.7426

45.7516

45.7716

45.7935

� 5 0 5 10 15 20 25

T� TP [sec]

Event 9,STACK,55.2 km

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

P pP sPsUSLP sMP

A

----


� 5 0 5 10 15 20 25

T� TP [sec]

Event 4,98.152oW,17.856oN,67.1 km

P pP sPsUSLP(?) sMP

YKB1

YKB2

YKB3

YKB4

YKR9

YKR8

YKR7

YKR6

YKB6

YKR5

YKR4

YKR3

YKR2

YKR1

YKB8

YKB9

YKB0

349.427

349.439

349.452

349.463

349.505

349.475

349.446

349.418

349.487

349.389

349.359

349.331

349.302

349.273

349.511

349.523

349.534

45.966

45.9868

46.009

46.0298

46.0418

46.0505

46.059

46.0673

46.0721

46.0757

46.0839

46.0926

46.101

46.1096

46.114

46.1338

46.1554

� 5 0 5 10 15 20 25

T� TP [sec]

Event 4, STACK,57.7 km

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

P pP sPsUSLP(?) sMP

-

---

C

Song et al, Supp. Figure 7 cont.

� 5 0 5 10 15 20 25

T� TP [sec]

Event 27,101.314oW,17.912oN,60.4 km

P pP sPsUSLP sMP

YKB1

YKB2

YKB3

YKB4

YKR9

YKR8

YKR7

YKR5

YKB6

YKR4

YKR3

YKR2

YKB7

YKR1

YKB8

YKB9

YKB0

351.352

351.362

351.372

351.382

351.422

351.392

351.362

351.302

351.402

351.272

351.242

351.212

351.411

351.182

351.422

351.432

351.44

45.4081

45.4295

45.4522

45.4736

45.4877

45.4948

45.5018

45.5156

45.517

45.5222

45.5294

45.5363

45.5386

45.5434

45.5601

45.5804

45.6026

� 5 0 5 10 15 20 25

T� TP [sec]

Event 27,STACK,52.4 km

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

P pP sPsUSLP sMP

----

Fig. S7 Stacking of teleseismic short period data. Teleseismic short-period data recorded

by the Yellowknife array in Canada are stacked to increase the signal-to-noise ratio. The

stacking begins with the bottom trace YKB0 as trace 1 on the right. Trace 2 contains

YKB0+YKB1. Trace 3 contains YKB0+YKB1+YKB2, etc. Record sections from event

27 (west Guerrero), event 4 (Oaxaca) and event 9 (Guerrero) are shown in (A)-(C). The

left panel displays the data before stacking and the right panel shows the stacking

seismograms. All traces are aligned on the P wave and filtered between 0.5 Hz and 1 Hz.

Depth phases such as pP and sP can be clearly identified. In particular, the sP is strong

such that underside reflections from the USL are favourably excited for all three events.

In (A), we do not observe sUSLP wave in the stacking seismograms from event 4 near

Oaxaca, whereas in (B) and (C), the amplitude of sUSLP relative to sP increases from

event 9 to event 27. Note predicted arrival times of depth phases pP and sP from Supp.

Table1 are shown in green dashed line while red dashed lines indicate the timings

consistent with the observed depth phases.

� 30 � 20 � 10 0 10 20 30 40 50 60

Time [sec]

Event 31 (Mw=5.9)

990621 (Mw=6.3)

941210 (Mw=6.4)

SYN (USL, dVs=� 45%)


SsUSLS

sS

� 30� 20� 10 0 10 20 30 40 50 60 70

Time [sec]

970522

DATA_SDV



SsUSLS

sS

-

-

-

-

------

230˚ 240˚ 250˚ 260˚ 270˚ 280˚ 290˚ 300˚0˚

10˚

20˚

30˚

40˚

50˚

60˚

70˚

SDV

FFC

Yellowknife

FFC


Fig. S8 Modelling broadband teleseismic waveforms containing underside reflection

sUSLS wave at IRIS stations FFC and SDV displayed on the left. The energy arriving in

the time window between the S wave and the depth phase sS recorded at station FFC is

the underside reflections sUSLS from the USL. Data from event 31 and two larger events

990621 (Mw=6.3, long=101.62º, lat=18.09ºN), 941210 (Mw=6.4, long=101.39ºW,

lat=18.18ºN) are similar and they are consistent with model predictions as displayed. Its

polarity is opposite to the depth phase sS wave due to a velocity reversal across the top of

the USL. The USL model can also better explain the S wave data from a large event

970522 (Mw=6.5, long=101.73ºW, lat=18.76ºN) recorded at station SDV.

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

year

2526

35 36

15 16

24 34 14

27

17 37 23

3028 38

2939 19

20

1033

22

3231

12

11

A

Song et al., Supp Figure 9

Fig. S9: Temporal relationship between the USL and the slow-slip events (SSEs) (S15-

S18). Data between 99ºW and 102ºW near Guerrero are included to demonstrate that

strong SP waves from the USL are not only observed during the SSEs, but also are

observed before and after the SSEs. The grey zones indicate the occurrences of the SSEs.


104˚W 103˚W 102˚W 101˚W 100˚W 99˚W 98˚W 97˚W 96˚W 95˚W

16˚N

17˚N

18˚N

19˚N

20˚N

21˚N

0 50 100

km

E2

E1AM

TMVB

20 km

40 km

60 km

80 km

100 km

UNMPTRPESTAMULUMIXCTEPETONICIREARBO

COAC

TEMP

IXCATIBL

OCOL

SAME

PASU

TONNTECA

PSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQPSIQKM67

20

Mexico

CO

TMVBRA

PA

NA

Orozco F.Z 5.8 cm/yr

1960 - 19851985 - 19951995 - 2007

CMT Engdahlgreat EQ

Guerrero

Oaxaca

Michoacan

A

0 T < 15 years 15 < T < 25 years

T > 25 years

98oW - 99oW 102oW - 103oW 100oW - 101oW

Str

ess

on th

e fa

ult

slow

-slip

in th

e tr

ansi

tion

zone

The lapse of time after previous large megathrust earthquake, T

???

???

No.of intra-slab events

B

Fig. S10: (A) Spatial-temporal variations in seismicity along southern Mexico. An

enlarged map shows the mapped USL (HPFP layer) along with great earthquake slip

zones and intra-slab events from global centroid moment tensor solution (with normal

fault mechanism) and Engdahl catalog (depth > 35 km). The orange line depicts

approximate down-dip limit of the transition zone where SSEs take place or are expected.

Contours of slip patches for previous SSEs are shown in green lines. Note that no intra-

slab events beneath the transition zone have occurred in west Oaxaca (98°W-99ºW)

(outlined by the white dotted line, where recent megathrust earthquakes are located. More

frequent intra-slab events beneath the transition zone in Michoacan section (102ºW-

103ºW) have occurred in the period 10 years after previous megathrust earthquakes.

Currently, no SSEs are reported in these segments. Both intra-slab events and SSEs are

observed in Guerrero (100ºW-101ºW) where a seismic gap exists for more than 90 years.

(B) A schematic map showing the working hypothesis for spatial-temporal variations in

seismicity and SSEs along southern Mexico. We categorize southern Mexico based upon

the lapse of time (T) after previous megathrust earthquake. When T is less than

approximately 15 years, there are no intra-slab events beneath the transition zone

possibly caused by a temporal decrease in plate coupling on the subduction zone interface

and consequently a decrease in the extensional stress inside the slab (S11). We

hypothesize that afterslip in the transition zone likely prohibits the occurrences of the

SSEs temporarily. While the stress continues accumulating on the subduction zone

interface, the plate coupling increases so that the extensional stress inside the slab

increases as well. At this stage, more frequent intra-slab earthquakes occur beneath the

transition zone, but we hypothesize the episodic slow-slip on the transition zone is small

due to limited plate coupling in the transition zone. For T longer than 25 years, the

subduction zone interface is strongly coupled with some prominent coupling extending

into the transition zone. There are frequent intra-slab earthquakes and we observe

episodic slow slip in the transition zone. This hypothesis can be understood as a stress

feedback system where megathrust earthquakes, intra-slab events and SSEs are linked

(inset). Future GPS instrumentation will test its validity.

A

-2 -1 0 1 2 3 4 5 6

Time (sec)

PTRP

ESTA

MULU

TEPE

TONI

-2 -1 0 1 2 3 4 5 6

Time (sec)

PTRP

ESTA

MULU

TEPE

TONI

slab dip ~ 14O slab dip ~ 8OB

S

N


2 3 4 5 6 7 8

Moho

USL

LVL

km/s

Dep

th (

km)

0

25

50

75

100

125

1500 50 100 150 200 250

Distance (km)

LVL

Moho

PTRP

???

USL

� 2 � 1 0 1 2 3 4 5 6

Time (sec)

dlnVs/dlnVp = 2.15

dlnVs/dlnVp = 2

dlnVs/dlnVp = 1.75

dlnVs/dlnVp = 1.5

dlnVs/dlnVp = 1.25

dlnVs/dlnVp = 1

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

dVs = -10%

dVs = -8%

dVs = -6%

dVs = -4%

dVs = -2%

E F

� 2 � 1 0 1 2 3 4 5 6 7 8 9

Time (sec)

� 10 km

10 km

PDE

D

--

-- - -

Song et al. Supp. Figure 11 cont.

� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)

-5o+5o

� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)

� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)

-5o+5o

� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)� 2 � 1 0 1 2 3 4 5 6

Time (sec)

Strike Dip Rake

Strike Dip Rake

C

� 2 � 1 0 1 2 3 4 5 6

Time (sec)

model A1(usgs,str= 5)

model A (usgs,str= 5)

model A (usgs)

� 2 � 1 0 1 2 3 4 5 6

Time (sec)

model A1(usgs,str=� 5)

model A (usgs,str=� 5)

model A (usgs)

- - ----

------

- -

- -

Fig. S11 (A) Slab geometry from event M to station PTRP (see also Fig. S1A for event

locations). Distance is measured with respect to the source location. A depth-section of

velocity structure near the source (green dashed line) is shown on the right. Note P wave

velocity directly below the Moho is slow at 7.5 km/s and it is consistent with travel time

analysis and waveform shape observed at the MASE array. Using a typical mantle

velocity of 8.0 km/s produces a long period diffraction along the Moho, which is not

observed in the data. A 30% serpentinization can explain this low seismic velocity and it

is similar to the findings in Cascadia (S16). Note the slab geometry at deeper depth

(below the receiver) is not well defined.

(B) Sensitivity test on the dip angle of the slab. Left panel shows systematic time shift of

the SP wave between the data and synthetics with slab dipping at about 14º near the

source24. Synthetic SP wave arrives late relative to the observation at stations toward the

south and arrives early relative to the observation at stations toward the north. With a

shallower dip angle of about 8º, we can explain the timing of the SP wave very well

shown on the right panel.

(C) Sensitivity test on focal mechanism. We test sensitivity of P waveforms against

strike, dip and rake for station PTRP (upper left panel) and station SAME (lower left

panel). The synthetics are computed with a USL (3 km, dVs = -40%) and a LVL (22 km,

dVp = -7%). On the right panel, Model A is constructed based upon USGS mechanism.

Assuming uncertainty in the strike of 5º, we show that synthetics computed with such a

focal mechanism is slightly different from that computed with the USGS mechanism

(right panel). Using such a perturbed focal mechanism, model A1 is modified from model

A to explain the data. It is very similar to the model A except with a slightly thicker LVL

(24 km) and a slower USL (dVs = -44%). It suggests that uncertainties in the focal

mechanism do not change our model. We assume dlnVs/dlnVp = 2 in our calculation.

(D) Sensitivity test on earthquake mis-location. Moving the earthquake location ±10 km

does not produce noticeable waveform difference.

(E) Sensitivity test on dlnVs/dlnVp of the USL. Synthetics show that the converted SP

converted wave does not have great sensitivity on the dlnVs/dlnVp of the USL, except

for dlnVs/dlnVp = 1.

(F) Sensitivity test on the S wave velocity directly above the USL. Decreasing its

velocity reduces the velocity contrast across the top of the USL and the amplitude of the

converted SP wave from the top of the USL.

Table S1: Earthquake source parameters.

Event No. Time Long. (º) Lat. (º) Depth (km) Mb

1 1994/02/23 -97.1601 18.0463 70 5.6

2 1992/04/22 -96.5835 17.16 65.4 4.8

3 1994/05/06 -98.0373 18.3536 68.5 5

4 1999/12/27 -98.152 17.856 67.1 4.7

5 1999/09/08 -98.305 17.637 68.3 4.5

6 1999/12/14 -98.573 18.123 66.2 4.8

7 2007/10/02 -98.7 17.57 52 4.7

8 2000/07/21 -98.9699 18.29 66.2 5.4

9 2003/11/19 -99.3765 18.0166 72.7 4.9

10 1997/03/22 -99.526 17.302 76.3 4.7

11 1994/10/29 -99.5025 17.5405 89.2 4.5

12 1991/03/25 -99.8185 17.2076 53.4 4.6

13 2005/05/26 -99.593

(-99.97)

18.219

(17.94)

93.2

(58.0)

4.7

14 2007/04/13 -100.029

(-100.31)

17.453

(17.135)

52

(34.0)

5.4

15 1991/04/27 -100.207 17.2378 58.9 4.6

16 1997/07/19 -100.131 17.4711 71.9 4.6

17 1997/05/08 -100.251 17.4628 63.3 5

18 2004/10/28 -99.7908 18.4016 68.2 4.7

19 1999/11/08 -100.54 17.397 52.2 4.7

20 1993/07/29 -100.475 17.6242 66 5

21 1998/08/05 -100.202 17.9912 70.6 4.6

22 2002/12/10 -100.909 17.884 85.5 5.1

23 2007/07/28 -100.84

(-100.843)

18.05

(18.052)

48

(49)

5.1

24 2006/02/20 -100.754

(-100.754)

18.145

(18.145)

51.1

(51)

5.1

25 1993/07/19 -100.46 18.3715 70.4 4.9

26 2007/07/18 -101.14

(-101.664)

17.98

(17.766)

43

(65)

4.9

27 2006/12/17 -101.314

(-101.314)

17.912

(17.912)

60.4

(60)

4.9

28 1993/08/29 -100.597 18.421 87.9 4.8

29 1992/02/12 -101.527 17.8911 51.3 5.1

30 2002/05/12 -100.96 18.301 64.1 4.6

31 1999/12/29 -101.49 18.24 66.7 5.9

32 2003/05/16 -101.22 18.381 67.3 5

33 1991/09/24 -100.945 18.5718 71.7 4.8

34 2006/08/11 -101.175

(-101.245)

18.482

(18.391)

64.9

(64)

4.9

35 1995/12/20 -101.068 18.5982 76.6 5.2

36 2002/09/21 -101.259 18.523 60.9 5

37 2002/01/02 -101.491 18.766 82.6 4.7

38 2004/02/06 -102.526 18.506 83 5.0

39 1992/06/01 -102.885 18.5874 76.6 4.7

40 1991/08/23 -97.8121 16.6186 54.4 4.9

*Earthquake source parameters are requested from IRIS event catalogue preferred location between 1990 and 2007. Source location from SSN local seismicity catalogue is included in the parenthesis when available.

Reference:

1. M. Pardo, G. Suárez, J. Geophys. Res. 100, 12357 (1995).

2. D. Helmberger, J. Vidale, Bull. Seis. Soc. Am. 78, 109 (1988).

3. N. M. Shaprio, S. K. Singh, D. Almora, M. Ayala, Geophys. J. Int. 147, 517 (2001).

4. K. Kruger, Bull. Seis. Soc. Am. 84, 149 (1994).

5. X. Perez-Campos et al., Geophys. Res. Lett. 35, doi:10.1029/2008GL035127 (2008).

6. R. G. Shepherd, Ground Water 27, 633(1989).

7. L. G. Montesi, G. Hirth, Earth Planet. Sci. Lett. 211, 97 (2003).

8. M. G. Bostock, R. D. Hyndman, S. Rondenay, S. M. Peacock, Nature 417, 536 (2002).

9. T.-R. A. Song, M. Simons, Science 301, 630 (2003).

10. R. E. Wells, R. J. Blakely, Y. Sugiyama, D. W. Scholl, P. A. Dinterman, J. Geophys.

Res. 108, doi:10.1029/2002JB002072 (2003).

11. T. Lay, L. Astiz, H. Kanamori, D. H. Chirstensen, Phys. Earth Planet. Inter. 54, 258

(1989).

12. N. Kato, Earth Planet Space. 56, 553 (2004).

13. Y. Liu, J. R. Rice, K. M. Larson, Earth Planet. Sci. Lett. 262, 493 (2007).

14. F. Correa-Mora, C. DeMets, E. Cabral-Cano, O. Diaz-Molina, B. Marquez-Azua,

Geochem. Geophys. Geosys, in review.

15. V. Kostoglodov et al., Geophys. Res. Lett. 30, doi:10.1029/2003GL017219 (2003).

16. K. M. Larson et al., J. Geophys. Res. 109, doi:10.1029/2003JB002843 (2004).

17. A. R. Lowry, Nature 442, 802 (2006).

18. K. M. Larson, V. Kostoglodov, S. Miyazaki, J. A. Santiago, J. A., Geophys. Res.

Lett. 34, doi:10.1029/2007GL029912 (2007).

Subducting Slab Ultra-Slow Velocity Layer Coincident with ... · We meshed the 3-D slab geometry underneath Central Mexico (S1) and made a direct 2-D profile from the source to the

Documents