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
Song et al., Supp. Figure 3
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
--- -
----
Song et al, Supp. Figure 5
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
----
Song et al, Supp. Figure 6
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
----
Song et al, Supp. Figure 7
� 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%)
SYN (USL, dVs=� 20%)
SsUSLS
sS
� 30� 20� 10 0 10 20 30 40 50 60 70
Time [sec]
970522
DATA_SDV
SYN (USL, dVs=� 45%)
SYN (USL, dVs=� 20%)
SsUSLS
sS
-
-
-
-
------
230˚ 240˚ 250˚ 260˚ 270˚ 280˚ 290˚ 300˚0˚
10˚
20˚
30˚
40˚
50˚
60˚
70˚
SDV
FFC
Yellowknife
FFC
Song et al., Supp. Figure 8
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.
Song et al., Supp. Figure 10
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
Song et al, Supp. Figure 11
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).