Docstoc

The End of a Seismic Cycle on the Gofar Transform Fault_ East

Document Sample
The End of a Seismic Cycle on the Gofar Transform Fault_ East Powered By Docstoc
					Capturing the End of a Seismic Cycle on the Gofar Transform Fault,
                         East Pacific Rise
       Jeffrey J. McGuire1, John A. Collins1, Pierre Gouedard2, Emily Roland3, Dan
    Lizarralde1, Margaret S. Boettcher4, Mark D. Behn1, and Robert D. van der Hilst2

1Department of Geology and Geophysics, Woods Hole Oceanographic Institution,
Woods Hole MA 02543

2 Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute
of Technology, Cambridge MA 02155

3   MIT-WHOI Joint Program

4 University   of New Hampshire, Dept. of Earth Sciences, Durham, NH 03824

It has long been known that Mid-Ocean Ridge Transform Faults (RTFs) are deficient
in producing large earthquakes relative to subduction zone and continental
transform faults1. The largest RTF earthquakes rupture only a small fraction of the
available fault area2 and most plate motion occurs aseismically2,3. Despite this
overall inefficiency in earthquake generation, RTFs produce an order of magnitude
more foreshocks than continental faults4,5. Here we show that these defining
features of RTFs result from along-strike variations in fault-zone material
properties: rupture barrier regions that prevent large earthquakes from covering an
entire fault are also the regions that generate spectacular foreshock sequences. With
an ocean bottom experiment specifically designed to use the intermediate-term
predictability of RTFs5,6, we captured the September 18th, 2008 magnitude 6.0 Gofar
earthquake and the preceding swarm of over 20,000 foreshocks. The week-long
foreshock sequence was confined to a 10-km long rupture barrier and coincided
both in time and space with a ~3% decrease in the average shear-wave velocity in
the fault-zone. Both the velocity change and the foreshock swarm suggest the
occurrence of a larger, aseismic, slow-earthquake in the barrier region. Our
observations suggest that material properties of fault segments capable of rupturing
in large earthquakes and those of rupture barrier regions differ, possibly as a result
of enhanced fluid circulation within the latter. Such heterogeneity may explain why,
on a global scale, purely temperature dependent rheologies7 can explain the
maximum depth extent of seismicity8 but not the overall deficiency of large
earthquake generation on RTFs1-3.

       With its short length (~90 km) and fast slip-rate (~14 cm/yr), the Gofar
transform fault9 on the equatorial East Pacific Rise (EPR; Figure 1) has a warm
thermal structure, which limits its largest earthquakes to 6.0Mw6.2. These
magnitude 6 earthquakes repeat quasi-periodically every ~5–6 year5 (Figure S1).
For instance, the westernmost segment of the Gofar transform fault has two
asperities that have repeatedly ruptured in M ~6.0 earthquakes (Figures 1, S1). The
eastern asperity (centered on the blue circle in Figure 1) ruptured in 1997, 2002,

                                           1
and 2007, while the western asperity (centered on the orange circle in Figure 1)
ruptured in 1992, 1997, 2003, and 20085,6. These fully coupled fault patches6 are
anomalous relative to the dominant behavior of RTFs where 80% of fault-motion
occurs without earthquakes2,3. The repeating large ruptures present an opportunity
to understand the physical processes that limit the size of RTF ruptures to be much
smaller than the full fault length as well as the high rates of RTF foreshock
activity4,10.
        Knowing the regularity of EPR seismic cycles we installed an OBS array to
capture the 2008 Mw 6.0 Gofar earthquake as well as its foreshocks and aftershocks.
Seven OBSs, each with a strong-motion accelerometer and a broadband
seismometer, were deployed on the western asperity (Figure 1), which ruptured on
18 September in an Mw 6.0 earthquake. The entire earthquake sequence was
recorded on-scale by the strong-motion network, providing an unprecedented
dataset covering the end of the seismic cycle on an RTF.

Earthquake Catalog and the Space-Time Evolution of Seismicity
        During our deployment the Gofar fault failed in series of ruptures that
propagated from east to west. Figure 1 displays epicenters of 21,919 earthquakes
between August 1st and December 30th 2008. The use of waveform-derived
differential arrival times yields a high-resolution image of the space-time evolution
of seismicity (see Methods). Throughout 2008, the 2007 M6 rupture zone (~105.7
to 105.9 W) experienced a low rate of seismicity, but the area immediately west
(~105.9 to 106.0W) experienced a high rate of seismicity between January 1 and
September 1st (Figures 2 and S2). The sequence of foreshocks began on September
1st (day 245) with a Mw 4.5 event located in the eventual mainshock rupture zone
(Figure 2). The western portion of the foreshock zone (~105.99W to 105.96W),
approximately 10km east of the M4.5 earthquake, responded by increasing its
seismicity rate by about a factor of two. Then, from September 10th (day 254) to
17th (day 261) the seismicity rate in the entire foreshock region increased by about
an order of magnitude (yellow line, Figure 2) until this earthquake swarm was
effectively terminated on September 18th (day 262) by the rupture of the M6.0
mainshock on the segment directly to the west (~106.0 to 106.2 W). The
mainshock’s epicenter occurred at about 106.09W and its aftershocks delineate a
~15–20 km long rupture zone (red circles in Figure 1). The westernmost ~10 km of
the plate boundary (106.2 to 106.3 W) failed in a swarm of ~20,000 earthquakes
on December 10-17 (cyan in Figures 1 and 2 and S3) to complete the seismic cycle.
Following the Sept. 18th M6 mainshock, the seismicity rate in the foreshock region
was one (two) order of magnitude smaller than before (during) the swarm (Figure
2).

Physical Properties of the Rupture Barrier
       Two first-order observations suggest that the physical properties in the
swarm/foreshock regions differ from that of the M6 rupture zones. The first
concerns the behavior of this fault segment during previous large earthquakes.
Since 1992, 7 M6 ruptures have occurred on the western Gofar fault, all of which

                                         2
have centroids that are indistinguishable from either the eastern or western
asperity5,6 (Figures 1 and S1). The 2008 aftershocks suggest rupture lengths of
approximately 15–20 km for these Mw 6.0–6.2 earthquakes. Thus, as the sharp
eastern boundary between the 2008 foreshocks and aftershocks indicates (Figure
S3), the foreshock region separates the M6 ruptures of the eastern and western
asperities (Figure 1). Each of the last 7 large ruptures failed to propagate through
the foreshock region and rupture both asperities, despite the near synchronization
of their seismic cycles (Figure S1) and despite the fact that the fault appears
continuous in this region (with offset less than 2km based on earthquake locations
and bathymetry). The ability of the foreshock region to repeatedly stop M6 rupture
fronts, despite being only ~10 km long, demonstrates that this segment is
comprised of a strongly velocity strengthening frictional material11 that is not
present in the adjacent M6 rupture zones.
        Additionally, microearthquakes in the barrier region extend a few km into
the uppermost mantle compared to being confined within the crust in the
mainshock rupture zone. Our array has limited control on focal depths of shallow
earthquakes (<~5km) owing to its station spacing, but we resolve a difference in the
range of the maximum earthquake depth between the M6 rupture zones compared
to the foreshock and December swarm fault segments.                  Figure 3 shows
seismograms aligned on the S-wave arrival (vertical red line at 3.0 s) for clusters of
earthquakes directly beneath stations located within ~1 km of the fault. The red
curve near 1.8 to 2.3 s marks the predicted P-wave arrival time from a local velocity
model12. S-P times for deep foreshocks and December swarm events (beneath
stations G08 and G04 respectively) are ~1.25 s while for events in the M6 rupture
zones (G06 and G10) they are ~0.6-0.8 s. With a constant Vp/Vs ratio of 1.73, as
inferred from our data (Figure S6) and consistent with normal oceanic crust, many
earthquakes in the foreshock and December swarm fault segments appear to be
located in the uppermost mantle (~8–10 km depth) whereas seismicity in the 2008
and 2007 rupture zones is confined to the lower oceanic crust (~3–6km). We note
that the ~1.25 s S-P times for the near-vertical rays would be compatible with
hypocenters within the lowermost crust if the entire fault-zone in the foreshock
region had an average Vp/Vs ratio larger than 2.0 (see Figure S5). This is unlikely
because the Vp/Vs ratio for rays propagating through the lower crust is ~1.73
(Figure S6). While in the swarm zones high values of Vp/Vs may occur at shallow
depth, sub-moho depths would still be required to explain the large S-P times shown
for stations G08 and G04 in Figure 3.
        Subtle changes in waveforms from earthquakes directly underneath station
G08 indicate that the elastic properties of the rupture barrier region changed during
the foreshock swarm. We inferred such medium changes, represented as changes
(dv/v) in S-wave velocity, measured from time-dependent stretching of the S-wave
coda (Figure 4) measured with a doublet method10 (see “Methods”). In order to
produce a reference signal relative to which waveform changes are measured, we
stacked waveforms from selected small earthquakes that occurred in the foreshock
region before day 240. We then determined, for each selected earthquake, a
stretching coefficient by comparison against this reference signal (Figure 4a,b). The
dense foreshock sequence allows for a continuous monitoring of S-wave velocity
                                          3
variations through day 262 when the main shock occurred.
        We find a ~3% decrease in S-wave velocity in the fault-zone during the
foreshock swarm (Figure 4). The coda is probably most sensitive to the shallow
(~0-3km) part of the damage zone, which has a greatly reduced P-wave speed
(~30%)12 compared to nearby crust and is likely fluid saturated. If this change
represents the response of fluid filled cracks to a change in stress, the minimum
stress change would be about 0.03 MPa (stress sensitivity of 10-6), but it could be 1-
2 orders of magnitude larger13. Generating such a stress change over a significant
portion of the fault zone likely requires a deformation event larger than the biggest
foreshock (Mw 4.1). Such a slip event would have been primarily aseismic similar to
other earthquake swarms triggered by creep events10,14,15. However, linear elastic
estimates of the stress-change implied by the velocity reduction may be inaccurate
because the velocity reduction was transient (Figure 4) (unlike the expected static
stress change from a creep event) and correlated with the seismicity-rate variation
which is expected to be a function of stressing-rate not stress during creep
events15,16. A similar correlation between stressing-rate and velocity change has
been observed for a large slow-slip event in the Guerrero subduction zone17. We
conclude that both the foreshock swarm and the seismic velocity changes resulted
from the elevated stressing rates caused by a large aseismic creep event in the week
before the mainshock that occurred within the rupture barrier region.

Implications for Transform Fault Mechanics
        Oceanic earthquakes are generally confined to depths shallower than the
600C isotherm7,8,18. For a half-space cooling model, this depth is ~4 km in the
center of the Gofar transform, and it would increase to only 5–6 km for thermal
models19 that include hydrothermal cooling. The depth extent of seismicity in the
mainshock regions is consistent with these thermal/rheological models, but the
deeper seismicity in the foreshock zone (~7-10 km, Figure 3) is not.
        The anomalous properties of the foreshock zone (compared to the
mainshock regions) likely result from significant fluid circulation within the
transform-fault domain. Two candidate mechanisms for explaining the velocity-
strengthening frictional behavior and the large S-P times of the swarm regions are:
(1) alteration of the fault-zone to serpentine, talc or other hydrous phases, and (2)
unusually high porosity and/or fluid pressures in this region. A ~10km long
serpentine body along the Garrett transform fault20 further south on the EPR
provides a potential example of along strike compositional variation, and serpentine
has been proposed as an explanation for the lack of large ruptures in the creeping
section of the San Andreas Fault21. Serpentine is velocity strengthening in standard
laboratory friction experiments22 but becomes velocity weakening at seismic-slip
velocities23. If present in large quantities, serpentine would produce a Vp/Vs ratio of
~2.0, explaining the S-P times24 and much of the apparent along-strike variations in
the depth extent of seismicity. Alternatively, high porosity and/or high fluid
pressure zones have been proposed as an explanation for the low P velocities within
the Clipperton fault-zone25,26. High porosity could substantially reduce the P
velocity in the lower crust27, and pore pressures in excess of hydrostatic could
increase the Vp/Vs ratio to the level required24 (~2.0) to explain the S-P times. In
                                          4
addition, near lithostatic pore pressures enable high-temperature (300–500C)
faults in gabbro to fail aseismically in slow earthquakes28. Hence a high pore-
pressure region could potentially explain both the rupture barrier and earthquake
microseismicity properties of the foreshock region. Only a narrow region of high
pore-pressure and/or serpentine around the rupture surface might be required to
effect rupture propagation, and this localized heterogeneity in physical properties
might not be detected in our Vp/Vs estimates (Figure S6).

       While the exact combination of enhanced cooling, serpentinization, and high
fluid pressure that exists in the foreshock zone is not resolved by our dataset, all
three mechanisms likely contribute to both the apparently greater depth extent of
seismicity and velocity-strengthening behavior of this region. Moreover, all three
mechanisms would result from enhanced fluid circulation in the foreshock region
relative to the portions of the fault that fail in large earthquakes. Given that
velocity-strengthening regions like the foreshock zone predominate on the 10,000+
km of faults in the global transform system1,2, it is possible that transform faults
represent a significant contribution to the global budget of seawater/rock
interactions as has been suggested by geochemical studies29. Future detailed
seismic imaging and submersible studies of the aseismic regions of transform faults
will help clarify the relationships between fluid circulation, fault-zone composition,
and rupture barriers.




                                          5
Methods Summary
        We constructed an initial earthquake catalog covering calendar year 2008
using standard STA/LTA based detection algorithms in the Antelope software
package for P-waves and wavelet based detections30 for S-wave arrival times.
However owing to the very low amplitude P-wave arrivals on OBSs and the short
temporal separations between P and S waves, these algorithms missed many
arrivals. To overcome this limitation, we used the best located events from the
STA/LTA catalog to provide waveform templates for a matched filter based
detection algorithm similar to that of Peng31. This computationally intensive
procedure detected over 200,000 earthquakes on the Gofar fault between August to
December 2008. For each day of the waveform-derived catalog, we used all
STA/LTA events with local magnitudes between 1.8 and 2.1 and within 20 days of
the target day as S-wave templates. Both horizontal components of the S-wave
records for the template event were filtered between 5 and 15 Hz. Accelerometer
horizontal components were used for the 7 stations in the center of the array while
the broadband horizontals were used for the other 7 stations (Figure 1). The
templates were then cross-correlated with the entire day’s seismogram. Events
were defined at times when the average cross-correlation coefficient (cc) across all
available (~28) horizontal components was cc=0.25 and at least two stations had
at least one S-wave component with a cc0.5. Only components with cc0.5 were
saved as detected arrivals. These criteria were set by visual inspection of trial runs
and produce very limited false identifications (i.e. events lacking clear P+S arrivals
on at least two stations). If more than one template event registered a detection
within a 5 s window, only the origin time and location corresponding to the
template with the highest average cc was saved in the catalog.
        A subset of 21,919 earthquakes between August and December 2008 from
the waveform-derived catalog were utilized in the relative relocations. These events
met several criteria. They had detections on at least 6 stations, and they were
retained by the hypoDD relocation program’s32 clustering algorithm, requiring at
least 77 differential time observations per pair with a cutoff of cc0.75. Waveform
derived differential arrival times were calculated using a time-domain interpolation
algorithm33 and resulted in a significant (~0.2 s) improvement over first-arrival
picks (Figure S7). A 2.56 s window around each arrival was extracted from the
waveform database, tapered and filtered. S-waves were bandpass filtered between
5 and 12 Hz while P-waves were filtered between 5 and 15 Hz. In general, the
broadband channels were utilized for the correlation except for a period of 3-9 days
following the M6.0 earthquake when they were not functioning due to clipping/re-
leveling problems. On these days, the accelerometer components were utilized.
Only arrival times with cc0.75 were used in the relocations. The relocations were
done in three subsets by longitude, 106.40W to 106.00W, 106.04W to 105.90W,
and 105.94W to 105.50W. For the eastern and western sections, we required 6
observations per pair, while for the middle section we required 8 observations per
pair due to denser instrumentation and higher seismicity rates in this region. For
the three groups, we used 1.9M, 0.7M, 0.6M P and 3.7M, 1.9M, 3.7M S to relocate
10779, 9462, 6258 earthquakes respectively. For earthquakes in the regions of

                                          6
overlap, we used the location estimate from the middle relocation with stricter
criteria. We did not use catalog times owing to their significantly higher percentage
of mis-identified phases. We utilized a 1-D version of the P-wave velocity model of
Roland et al.12 that accounts for the significant (~20-30%) reductions in P-wave
speeds in the Gofar fault-zone relative to ordinary EPR crust. We assumed a Vp/Vs
ratio of 1.73.
        Owing to the station spacing (~10km) the the earthquake relocations have
only limited constraint on the earthquake depths (~3-10 km).) We performed
several tests including starting all quakes at the same depth, starting with the
depths from the STA/LTA catalog, and starting with the catalog depths shifted
deeper by 1, 2, and 3 km. The clearest result is that the greater range of depths (~6
km) in the foreshock region compared to the aftershock region (~2 km) is required
by the differential times. However, the absolute depths are not well constrained, as
shifting the starting depths effects the centroid of the relocated depths (i.e. deeper
starting depths lead to deeper relocated depths overall) and has almost no effect on
the post-relocation residuals (i.e. data fit). We feel the best way to evaluate the
absolute depths given our sparse station coverage is from the S-P times of the
vertically propagating rays as show in Figure 3. Our preferred catalog (which used
initial catalog depths shifted 2 km deeper) matches the S-P times of earthquakes
underneath the four near fault stations G04, G06, G08, G10 (see Figure 3). We have
made similar plots for numerous clusters of earthquakes under each station and
these absolute depths appear robust given our velocity models.
        The relative variation dv/v of the S-wave velocity during the foreshock
sequence is measured using the doublet method34. This method consists in looking
for time shifts dt between a waveform and a reference trace, as a function of the
time t in the record. Here we selected foreshocks with waveforms as recorded at
station G08 similar to a master event, and used a stack of those occurring before day
240 as a reference trace. For each of the 670 selected foreshocks, and each of the 3
components, the waveform, filtered in the 5-12 Hz frequency band, is compared to
the reference trace using a tapered, one second long, moving time window centered
on time t to measure any shift dt as a function of t. While this method works well for
small dv/v, when velocity variations are as high as 10% the same time windows for
the reference and the current event doesn’t corresponds to the same cycles in the
waveform (especially for large times t), and the time shift dt cannot be measured. To
overcome this limitation, we first determine a coarse stretching coefficient ε (which
is an approximate value for dt/t) using the stretching method35, and use it to define
the center of the time window for the selected event in the doublet method as (1+ε)t
instead of t. A stretching coefficient dt/t is then inferred using a linear regression to
fit dt as a function of t (Figure 4b). Assuming a homogeneous velocity variation dv,
dt/t is equal to the opposite of the relative change in S-wave velocity -dv/v between
the (arbitrary) reference and the event date. The resulting dv/v estimates are of
similar amplitude on the 3 components (within 10%), which gives confidence that
they are not biased by any coupling resonance issues. Measurements on the 3
components are then averaged to increase the accuracy on dv/v. Repeating this
process for each event of the dense foreshock sequence allows for a continuous


                                           7
monitoring of S-wave velocity variations from day 215 to day 262 when the main
shock occurred (Figure 4c).
       Estimates of the centroid locations for the 2007 and 2008 M w 6 earthquakes
were derived using relative Rayleigh wave locations5. Both events were located
relative to a Mw 5.1 aftershock of the 2008 event (Figure S8). To determine the
absolute position of the M6 earthquakes, we assumed that the centroid location of
the M5.1 aftershock was equal to its epicenter location (-106.184, -4.512)
estimated from first arrival times. This yielded estimates of the 2008 (-106.047, -
4.559) and 2007 (-105.789, -4.618) centroids.


References

1      Brune, J. N. Seismic Moment Seismicity and Rate of Slip Along Major Fault
       Zones. J Geophys Res 73, 777-& (1968).
2      Boettcher, M. S. & Jordan, T. H. Earthquake scaling relations for mid-ocean
       ridge transform faults. J. Geophys. Res.-Solid Earth 109, 21, doi:B12302
10.1029/2004jb003110 (2004).
3      Bird, P., Y., K. Y. & D., J. D. in Plate Boundary Zones Vol. 30 Geodyn. Ser. eds
       Stein S. & Freymueller J. T.) 203–218 (AGU, 2002).
4      McGuire, J. J., Boettcher, M. S. & Jordan, T. H. Foreshock sequences and short-
       term earthquake predictability on East Pacific Rise transform faults. Nature
       434, 457-461, doi:Doi 10.1038/Nature03377 (2005).
5      McGuire, J. J. Seismic cycles and earthquake predictability on east pacific rise
       transform faults. Bull. Seismol. Soc. Amer. 98, 1067-1084, doi:Doi
       10.1785/0120070154 (2008).
6      Boettcher, M. S. & McGuire, J. J. Scaling relations for seismic cycles on mid-
       ocean ridge transform faults. Geophys Res Lett 36, -, doi:Artn L21301
Doi 10.1029/2009gl040115 (2009).
7      Boettcher, M. S., Hirth, G. & Evans, B. Olivine friction at the base of oceanic
       seismogenic zones. J. Geophys. Res.-Solid Earth 112, 13, doi:B01205
10.1029/2006jb004301 (2007).
8      Abercrombie, R. E. & Ekstrom, G. Earthquake slip on oceanic transform faults.
       Nature 410, 74-77 (2001).
9      Searle, R. C. Multiple, Closely Spaced Transform Faults in Fast-Slipping
       Fracture Zones Geology 11, 607-610 (1983).
10     Roland, E. & McGuire, J. J. Earthquake swarms on transform faults. Geophys J
       Int 178, 1677-1690 (2009).
11     Kaneko, Y., Avouac, J. P. & Lapusta, N. Towards inferring earthquake patterns
       from geodetic observations of interseismic coupling. Nat. Geosci. 3, 363-
       U324, doi:10.1038/ngeo843 (2010).
12     Roland, E. C., McGuire, J. J., Lizarralde, D. & Collins, J. A. in Am. Geophys. Un.
       Fall Mtg. (AGU).




                                           8
13    Silver, P. G., Daley, T. M., Niu, F. & Majer, E. L. Active Source Monitoring of
      Cross-Well Seismic Travel Time for Stress-Induced Changes. Bull. Seism. Soc.
      Am. 97, 281-293, doi:10.1785/0120060120 (2007).
14    Lohman, R. B. & McGuire, J. J. Earthquake swarms driven by aseismic creep in
      the Salton Trough, California. J. Geophys. Res.-Solid Earth 112, 10, doi:B04405
10.1029/2006jb004596 (2007).
15    Llenos, A. L., McGuire, J. J. & Ogata, Y. Modeling seismic swarms triggered by
      aseismic transients. Earth Planet. Sci. Lett. 281, 59-69, doi:Doi
      10.1016/J.Epsl.2009.02.011 (2009).
16    Dieterich, J. A CONSTITUTIVE LAW FOR RATE OF EARTHQUAKE
      PRODUCTION AND ITS APPLICATION TO EARTHQUAKE CLUSTERING. J.
      Geophys. Res.-Solid Earth 99, 2601-2618 (1994).
17    Rivet, D. et al. Seismic evidence of nonlinear crustal deformation during a
      large slow slip event in Mexico. Geophys Res Lett 38, doi:L08308
10.1029/2011gl047151 (2011).
18    Wiens, D. A. & Stein, S. Age Dependence of Oceanic Intraplate Seismicity and
      Implications for Lithospheric Evolution. J Geophys Res 88, 6455-6468 (1983).
19    Roland, E., Behn, M. D. & Hirth, G. Thermal-mechanical behavior of oceanic
      transform faults: Implications for the spatial distribution of seismicity.
      Geochem. Geophys. Geosyst. 11, 15, doi:Q07001
10.1029/2010gc003034 (2010).
20    Hekinian, R., Bideau, D., Cannat, M., Francheteau, J. & Hebert, R. Volcanic
      Activity and Crust Mantle Exposure in the Ultrafast Garrett Transform-Fault
      Near 13-Degrees-28S in the Pacific. Earth Planet. Sci. Lett. 108, 259-275
      (1992).
21    Moore, D. E. & Rymer, M. J. Talc-bearing serpentinite and the creeping section
      of the San Andreas fault. Nature 448, 795-797, doi:10.1038/nature06064
      (2007).
22    Reinen, L. A., Tullis, T. E. & Weeks, J. D. Two-mechanism model for fricitonal
      sliding of sepentinite. Geophys. Res. Lett. 19, 1535-1538 (1992).
23    Kohli, A. H., Goldsby, D. L., Hirth, G. & Tullis, T. Flash weakening of
      serpentinite at near-seismic slip rates. J. Geophys. Res.-Solid Earth 116, 18,
      doi:B03202
10.1029/2010jb007833 (2011).
24    Christensen, N. I. Pore Pressure and Oceanic Crustal Seismic Structure.
      Geophysical Journal of the Royal Astronomical Society 79, 411-423 (1984).
25    Van Avendonk, H. J. A., Harding, A. J., Orcutt, J. A. & McClain, J. S. A two-
      dimensional tomographic study of the Clipperton transform fault. J. Geophys.
      Res.-Solid Earth 103, 17885-17899 (1998).
26    Van Avendonk, H. J. A., Harding, A. J., Orcutt, J. A. & McClain, J. S. Contrast in
      crustal structure across the Clipperton transform fault from travel time
      tomography. J. Geophys. Res.-Solid Earth 106, 10961-10981 (2001).
27    Korenaga, J., Kelemen, P. B. & Holbrook, W. S. Methods for resolving the
      origin of large igneous provinces from crustal seismology. J. Geophys. Res.-
      Solid Earth 107, 27, doi:2178
10.1029/2001jb001030 (2002).
                                           9
28    Liu, Y. J. & Rice, J. R. Slow slip predictions based on granite and gabbro
      friction data compared to GPS measurements in northern Cascadia. J.
      Geophys. Res.-Solid Earth 114, 19, doi:B09407
10.1029/2008jb006142 (2009).
29    Manac'h, G., Lecuver, C. & Juteau, T. A fluid inclusion and stable isotope study
      of hydrothermal circulation in a transform zone: Western Blanco Depression,
      northeast Pacific. J. Geophys. Res., 104, 12941-12969 (1999).
30    Simons, F. J., Dando, B. D. E. & Allen, R. M. Automatic detection and rapid
      determination of earthquake magnitude by wavelet multiscale analysis of the
      primary       arrival.    Earth     Planet.    Sci.   Lett.    250,     214-223,
      doi:10.1016/j.epsl.2006.07.039 (2006).
31    Peng, Z. G. & Zhao, P. Migration of early aftershocks following the 2004
      Parkfield earthquake. Nat. Geosci. 2, 877-881, doi:10.1038/ngeo697 (2009).
32    Waldhauser, F. & Ellsworth, W. L. A double-difference earthquake location
      algorithm: Method and application to the northern Hayward fault, California.
      Bull. Seismol. Soc. Amer. 90, 1353-1368 (2000).
33    Schaff, D. P. et al. Optimizing correlation techniques for improved earthquake
      location. Bull. Seismol. Soc. Amer. 94, 705-721 (2004).
34    Poupinet, G., Ellsworth, W. L. & Frechet, J. Monitoring Velocity Variations in
      the Crust Using Earthquake Doublets - an Application to the Calaveras Fault,
      California. J Geophys Res 89, 5719-5731 (1984).
35    Sens-Schonfelder, C. & Wegler, U. Passive image interferometry and seasonal
      variations of seismic velocities at Merapi Volcano, Indonesia. Geophys Res
      Lett 33, 5, doi:L21302
10.1029/2006gl027797 (2006).

Supplementary Information is linked to the online version of the paper at
www.nature.com/nature.

Acknowledgements. We would like to thank the W. M. Keck Foundation for the
funding to build the OBSs that carried strong motion accelerometers and made this
study possible. We thank the crews of the R/V Thomas Thompson, Marcus
Langseth, and Atlantis for three nearly flawless cruises needed to collect this
dataset. We thank the WHOI OBSIP group, Ken Peal, Rob Handy, Alan Gardner, Dave
Dubois, Dan Kot, Peter Lemmond, and Jimmy Ryder, for collecting such an amazing
dataset. This work was done with the support of NSF grant OCE-0242117.


author contributions
J.M and J.C. conceived the OBS experiment. J.M, J.C., M.B., and E.R. participated in the
three research cruises and associated data collection. J.M. and J.C. derived the
earthquake catalog. P.G. and R. vdH. performed the velocity change analysis. E.R.
and D.L. determined the seismic velocity model used for relocations. E.R. and M. B.
carried out the thermal modeling. All authors discussed results and contributed to
the manuscript.

                                          10
author information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of this article at
www.nature.com/nature. Correspondence and requests for materials should be
addressed to J.M. (jmcguire@whoi.edu).

figure legends

Figure 1. Epicenters of 23,877 earthquakes from August through December 2008
(black dots), determined with a double difference relocation scheme (see Methods).
The inset map shows the Gofar fault’s location (star) on the equatorial East Pacific
Rise. Earthquakes occurring on September 10th, 11th, and 12th (foreshocks) are
shown in yellow. Earthquakes on September 18th, 19th, and 20th (aftershocks) are
shown in red. Earthquakes on December 7th and 8th (swarm) are shown in cyan.
Ocean bottom seismographs (OBS) with only a broadband sensor are shown as
white triangles, while those with both broadband and strong-motion sensors are
shown as white stars. The epicenter of the largest (M5.2) aftershock, the centroid of
the 2008 Mw 6.0, and the centroid of the 2007 M6.2 earthquakes are shown as large
brown, orange, and blue circles respectively. OBS G04, G06, G08 and G10 are
labeled below their symbols. The background colors denote seafloor depth (See
Figure S3).

Figure 2. Left) Cumulative number of earthquakes in the waveform-detection
earthquake catalog (black curve). The yellow, red, and cyan curves show the
cumulative curve for earthquakes in the foreshock zone (105.9 to 105.98 W),
mainshock rupture zone (106.04 to 106.18 W), and the December swarm zone
(106.2 to 106.3 W) (see Figure 1). The M6.0 mainshock occurred on Julian day
262 or Sept. 18 (vertical black line). Right) zoom into the time period of the
foreshock sequence. The solid yellow, red, and black curves are the same as the
righthand panel. The vertical lines denote the M4.5 earthquake on day 245 and the
M6.0 earthquake on day 262. The dashed and dash-dot black lines denote
seismicity located in the eastern foreshock zone (between 105.91 and 105.945W)
and the western foreshock zone (-105.98 and 105.96 W) respectively. These are
subsets of the foreshock zone shown with the yellow curve.

Figure 3. Examples of S-P times for vertically propagating rays in the Gofar fault-zone. The
top panels show vertical-component P-waveforms for clusters of earthquakes directly
beneath each station (G04, G06, G08, and G10) within the Gofar fault-zone (Figure 1). The
vertical-component waveforms are aligned using cross-correlation of the horizontal-
component waveforms at the same station (see Figure S4). Thus, the S-wave first arrivals
are precisely aligned at ~3.0 seconds in each panel. These clusters are located within 1–3
km epicentral distance from the station and hence are dominantly vertically propagating
rays. The red lines show the observed S-wave first-arrival time (vertical line at ~2.8 s) and
the P-wave arrival time (earlier red line) predicted using a velocity model derived from a


                                             11
refraction experiment that crosses the foreshock zone12 and assumes a Vp/Vs ratio of 1.73.
There is a significant difference in S-P time between stations in the rupture zones of M6
earthquakes (G06 ~0.8 s, G10 ~0.7s) and those in the rupture barriers on either side of the
2008 aftershock zone (G08 ~1.25 s, G04 ~1.1 s). The location of the clusters is shown on
the bottom panel, which depicts the relocated earthquake catalog as a function of distance
along the fault (west is negative on the x axis, the origin is at -106.00 longitude). The cyan,
red, yellow, and blue circles are the locations of the clusters beneath stations G04, G06, G08,
G10 (blue triangles from left to right) respectively.

Figure 4.




                                              12
Figures




Figure 1. Epicenters of 23,877 earthquakes from August through December 2008
(black dots), determined with a double difference relocation scheme (see Methods).
The inset map shows the Gofar fault’s location (star) on the equatorial East Pacific
Rise. Earthquakes occurring on September 10th, 11th, and 12th (foreshocks) are
shown in yellow. Earthquakes on September 18th, 19th, and 20th (aftershocks) are
shown in red. Earthquakes on December 7th and 8th (swarm) are shown in cyan.
Ocean bottom seismographs (OBS) with only a broadband sensor are shown as
white triangles, while those with both broadband and strong-motion sensors are
shown as white stars. The epicenter of the largest (M5.2) aftershock, the centroid of
the 2008 Mw 6.0, and the centroid of the 2007 M6.2 earthquakes are shown as large
brown, orange, and blue circles respectively. OBS G04, G06, G08 and G10 are
labeled below their symbols. The background colors denote seafloor depth (See
Figure S3).




                                         13
Figure 2. Left) Cumulative number of earthquakes in the waveform-detection
earthquake catalog (black curve). The yellow, red, and cyan curves show the
cumulative curve for earthquakes in the foreshock zone (105.9 to 105.98 W),
mainshock rupture zone (106.04 to 106.18 W), and the December swarm zone
(106.2 to 106.3 W) (see Figure 1). The M6.0 mainshock occurred on Julian day
262 or Sept. 18 (vertical black line). Right) zoom into the time period of the
foreshock sequence. The solid yellow, red, and black curves are the same as the
righthand panel. The vertical lines denote the M4.5 earthquake on day 245 and the
M6.0 earthquake on day 262. The dashed and dash-dot black lines denote
seismicity located in the eastern foreshock zone (between 105.91 and 105.945W)
and the western foreshock zone (-105.98 and 105.96 W) respectively. These are
subsets of the foreshock zone shown with the yellow curve.




                                       14
Figure 3. Examples of S-P times for vertically propagating rays in the Gofar fault-zone. The
top panels show vertical-component P-waveforms for clusters of earthquakes directly
beneath each station (G04, G06, G08, and G10) within the Gofar fault-zone (Figure 1). The
vertical-component waveforms are aligned using cross-correlation of the horizontal-
component waveforms at the same station (see Figure S4). Thus, the S-wave first arrivals
are precisely aligned at ~3.0 seconds in each panel. These clusters are located within 1–3
km epicentral distance from the station and hence are dominantly vertically propagating
rays. The red lines show the observed S-wave first-arrival time (vertical line at ~2.8 s) and
the P-wave arrival time (earlier red line) predicted using a velocity model derived from a
refraction experiment that crosses the foreshock zone12 and assumes a Vp/Vs ratio of 1.73.
There is a significant difference in S-P time between stations in the rupture zones of M6
earthquakes (G06 ~0.8 s, G10 ~0.7s) and those in the rupture barriers on either side of the
2008 aftershock zone (G08 ~1.25 s, G04 ~1.1 s). The location of the clusters is shown on
the bottom panel, which depicts the relocated earthquake catalog as a function of distance
along the fault (west is negative on the x axis, the origin is at -106.00 longitude). The cyan,
red, yellow, and blue circles are the locations of the clusters beneath stations G04, G06, G08,
G10 (blue triangles from left to right) respectively.

                                              15
Figure 4. Top panel shows a comparison of the S-waveforms between the reference
trace (black line, a stack of aligned waveforms for events with similar waveforms
occurring prior to day 240) and a single event occurring on day 255 (red line), in the
5-12 Hz frequency band.T he middle panel shows the measured time shifts dt as a
function of time t from S-wave arrival (t = 0) for the day 255 earthquake waveform
in the top panel relative to the reference trace. The red line indicates the best fit,
which is a measurement of dt/t = -dv/v.The bottom panel shows the measurements
of dv/v for individual earthquakes in the foreshock zone as a function of the time of
the earthquake (black dots). 670 measurements are shown for earthquakes located
within 3 km of station G08. Insets show close ups of the velocity changes on days
245 and 254. The blue curve shows the number of earthquakes detected in the
waveform catalog in 12 hour bins in a 3km long portion of the fault that coincides
with the earthquakes utilized to measure the velocity changes.


                                         16
Supplementary Figs:




Figure S1. Map and space-time evolution of large earthquakes on the Quebrada,
Discovery, and Gofar transform faults, after McGuire 2008. Earthquakes with Mw >
5.0 since 1990 are shown as circles. Events with overlapping rupture areas (defined
as relative centroid locations < 5 km, see McGuire 2008) are shown in a constant
color. The 2008 Gofar Mw 6.0 event is the most recent of the red circles. The
vertical gray lines show the locations of segment boundaries defined by intra-
transform spreading centers.



                                        17
Figure S2. Cumulative earthquake counts for the STA/LTA derived catalog (black
line). The yellow, red, and cyan curves show the cumulative curve for earthquakes
in the foreshock zone (105.9 to -105.98 W), mainshock rupture zone (106.04 to
106.18 W), and the December swarm zone (106.2 to 106.3 W) (see Figure 1).
The M6.0 mainshock occurred on Julian day 262 or Sept. 18 (vertical black line).




                                       18
Figure S3. (A) Bathymetry of the western portion of the Gofar tranform fault. Stars and triangles
denote the locations of ocean bottom seismographs equipped with both a strong-motion
accelerometer and a broadband seismometer or just a broadband seismometer respectively. (B) The
relocated earthquake catalog is shown as black circles. Earthquakes occurring on September 10–12
2008 (Julian days 254–256) are shown in yellow, while earthquakes occurring on December 7 th and
8th 2010 (Julian days 342 and 343) during the December swarm are shown in cyan. The epicenter of
the largest aftershock, the centroid of the 2008 Mw 6.0, and the centroid of the 2007 M6.2
earthquakes are shown as brown, orange, and blue circles respectively. (C) The relocated
earthquake catalog. Earthquakes occurring on September 18–20 2010 (Julian days 262–264) during
the early part of the aftershock sequence are shown in red.



                                              19
Figure S4. Examples of S-P times for nearly vertically-propagating rays in the
foreshock and aftershock areas. The first and third panels show horizontal-
component S-waves aligned by cross-correlation for a cluster of earthquakes
beneath station G06 (aftershock zone, Figure 1) and G08 (foreshock zone)
respectively as a function of epicentral depth in the relocated catalog. These
clusters are located within 2–3 km epicentral distance and hence are dominantly
vertically-propagating rays. The second and fourth panels show the vertical-
component waveforms for the same earthquakes with the same (S-wave derived)
temporal alignment. The red lines show the observed S-wave first-arrival time
(vertical line at ~3 seconds) and the predicted P-wave arrival time using a velocity
model derived from a refraction experiment that crosses the foreshock zone [Roland
AGU 2010] and assumes a Vp/Vs ratio of 1.73. There is a significant difference in S-
P time between G08 (~1.25 seconds) and G06 (~0.85 seconds) for the two clusters
of earthquakes. The location of the two clusters is shown on the bottom panel,
which depicts the relocated earthquake catalog as a function of distance along the
fault (west is negative on the x axis, the origin is at 106.00 W). The red circles are
the locations of the cluster beneath G06 (left-hand blue triangle) and the yellow
circles are the locations of the cluster beneath G08 (right-hand blue triangle).


                                          20
Figure S5. Plots of S-P times for different epicentral distances assuming a Vp/Vs
ratio of 1.73 (top) and for different Vp/Vs ratios assuming an epicentral distance of
3 km (bottom) for the P-wave model of Roland12. The waveforms shown in Figure 3
are typically for epicentral distances of 2.0 to 2.5 km. End-member models with
earthquake depths of 8–10 km and a Vp/Vs ratio of 1.73 or depths of ~6 km and a
Vp/Vs ratio of >2.0 could explained the observed S-P times in the foreshock zone.
Similarly, intermediate models between these end members are also consistent with
the data.




                                         21
Figure S6. Comparison of the difference between P-wave first arrival times at two
stations with S-wave first arrival times at those stations for ~10,000 earthquakes.
The left-hand panel shows stations G04 and G08, which are 20 km apart so the
measured Vp/Vs ratio (1.73) is predominately sensitive to the lower crust of the M6
rupture zone. The right-hand panel shows stations G06 and G10, which are 20 km
apart so the measured Vp/Vs ratio (1.67) is predominately sensitive to the lower
crust of the foreshock zone. Red lines are least-squares fits to the data points and
their slope gives an estimate of the Vp/Vs ratio.




                                        22
Figure S7. Examples of S-wave differential time measurements for 229 earthquakes
recorded at station G06. The left-hand panel shows the S-waveforms filtered
between 5 and 12 Hz aligned using the STA/LTA catalog arrival times. The right-
hand panel shows the same seismograms aligned by cross-correlation. The
earthquakes are predominately in the foreshock zone.




                                      23
Figure S8. (Left) Relative Rayleigh wave based locations for the 2007 M w 6.2 and
2008 Mw 6.1 Gofar transform earthquakes. The 2008 event is located 32 km from
the 2007 event at an azimuth of 293. The top panel compares the aligned
waveforms at global seismic stations. The bottom panel shows the differential time
measurements (blue circles) and the fit to them (black line and red circles). (Right)
Similar relative locations for the Mw 5.2 aftershock on 9/18/2008 and the 2008 M w
6.0 mainshock. The aftershock’s centroid is located 16 km west of the mainshock’s
centroid.




                                         24
25

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:27
posted:11/14/2011
language:English
pages:25