SEARCH

SEARCH BY CITATION

Keywords:

  • Pisco earthquake;
  • back-projection;
  • rupture segmentation

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The 2007 Mw 8.0 Pisco, Peru Earthquake
  5. 3. Back-Projection of Short-Period Teleseismic P-Waves
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] The short-period (0.5–2 s) seismic radiation properties of the August 15 (23:40:57 UTC) 2007 Mw8.0 Pisco, Peru earthquake are imaged by back-projecting P waves recorded at 374 elements of USArray deployed in western North America at distances of 54°–74° from the source region. The coherent short-period seismic energy release has two main intervals similar to moment-rate functions determined by inversion of longer-period teleseismic body waves; however, the spatial locations of the coherent bursts of short-period energy release are located north and down-dip of the region of major slip. The contrast between short- and long-period seismic radiation properties of the Pisco earthquake is more subtle than for the 2011 Mw9.0 Tohoku earthquake, but provides further support for the idea of depth-dependent changes in sliding behavior during megathrust ruptures.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The 2007 Mw 8.0 Pisco, Peru Earthquake
  5. 3. Back-Projection of Short-Period Teleseismic P-Waves
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] Great megathrust earthquakes have rupture zones that span thousands of square kilometers, so it is reasonable to expect properties such as slip, stress drop, rupture velocity, rise time, etc., to have significant spatial variation during the rupture. Along-strike variations in megathrust rupture properties have been extensively investigated and attributed to factors such as gravity anomalies, subducted sediment thickness, slab contortions, and ridge or seamount subduction [Loveless et al., 2010 and references therein]. The Mw9.0 Tohoku-oki earthquake in 2011 definitively established that variations in megathrust rupture properties also occur across the width of the megathrust, along-dip [e.g.,Ide et al., 2011; Koper et al., 2011a, 2011b; Meng et al., 2011; Simons et al., 2011; Yao et al., 2011]. For the 2011 event, the down-dip portion of the rupture produced high frequency enriched seismic radiation, possibly indicating abrupt fluctuations in slip velocity and an overall rougher rupture process compared to the relatively smooth, tsunamigenic sliding that occurred up-dip. Importantly, it appears that the local strong ground motions in Japan generated by this earthquake (with observed accelerations as high as 2.7 g) originated from the deeper portions of the megathrust [Kurahashi and Irikura, 2011].

[3] The Mw9.0 Tohoku-oki earthquake was unusual in that it ruptured the entire width of the megathrust extending from the Japan Trench, approximately 200 km offshore, to the down-dip termination of interplate seismicity 45–55 km beneath the Honshu coastline [e.g.,Ammon et al., 2011; Ide et al., 2011; Simons et al., 2011]. Combined with the extensive seismic monitoring network in Japan, this allowed for unprecedented imaging of spatial variations in the rupture, especially along-dip. Recent re-evaluations of teleseismic data from other giant megathrust events suggest similar along-dip variations in the spectrum of seismic radiation, with high-frequency enrichment in seismic radiation from deeper portions of the rupture. Back-projections of P, PP, and PKIKP waves from the Mw8.8 Chile earthquake of 2010 show pulses of short-period energy release located 25–45 km down-dip of the patches with greatest slip [Koper et al., 2012], while back-projections of P waves from the Mw9.15 Sumatra earthquake of 2004 show bursts of short-period energy shifted noticeably down-dip of high-slip regions [Lay et al., 2012]. Together, these results suggest a new framework for interpreting megathrust rupture variability in terms of depth-dependent changes in frictional sliding behavior [Lay et al., 2012].

[4] Here we analyze teleseismic data from the Mw8.0 Pisco, Peru earthquake of 2007 that ruptured the megathrust boundary between the Nazca and South America plates. We search for a similar along-dip segmentation of the rupture properties by comparing locations of coherent short-period energy release imaged by new back-projections of teleseismic P waves with previously published finite-fault models of slip determined from inversion of longer-period seismic, geodetic, and tsunami data. The new back-projection images are created with data from the dense, uniformly instrumented Transportable Array (TA) that was deployed in the western United States at the time of the Pisco earthquake. Owing to the high coherence of teleseismic P-waves recorded by the TA, back-projections of the Pisco rupture can be extended to higher frequencies (∼2 Hz) than was possible with previous back-projections that used a sparse, global geometry of broadband seismometers [Lay et al., 2010a], helping to highlight any frequency-dependent variations in the rupture.

2. The 2007 Mw 8.0 Pisco, Peru Earthquake

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The 2007 Mw 8.0 Pisco, Peru Earthquake
  5. 3. Back-Projection of Short-Period Teleseismic P-Waves
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[5] The 2007 Mw 8.0 Pisco earthquake (Figure 1) ruptured a portion of the Nazca/South America megathrust just to the northwest of the subducting Nazca ridge (∼15°S). Based on the first few days of aftershocks, the Pisco rupture zone was approximately 200 km long (along strike) and 100 km wide (along dip), separating a region of flat-slab subduction to the northwest from more traditional dipping subduction to the southeast. Over the last hundred years, several great earthquakes have occurred along the Peru megathrust (M8 in 1940, M8 in 1942, M8.1 in 1974, M8.4 in 2001) and in-depth discussions of the tectonic setting and historical seismicity in Peru are given byBeck and Ruff [1989] and Dorbath et al. [1990].

image

Figure 1. Tectonic setting of the 2007 Mw8.0 Pisco earthquake. In the main panel the black circle represents the USGS epicenter and the smaller red dots represent USGS-located aftershocks that occurred within a few days of the main shock. The inset shows a close-up of the epicentral region, and in particular the epicenters from several global agencies: ISC- International Seismic Center, IDC - International Data Center, NEIC - National Earthquake Information Center (USGS). The global centroid moment tensor (gCMT) centroid location and best double couple solution is also shown.

Download figure to PowerPoint

[6] The final United States Geological Survey (USGS) epicenter of 13.39°S 76.60°W is in between two epicenters provided by the International Seismic Center (ISC, see www.isc.ac.uk), and slightly east of the epicenter determined by the International Data Center (IDC, see www.ctbto.org). Together the four locations span a linear distribution about 20 km long oriented roughly perpendicular to the coastline (Figure 1). This trend is indicative of the relative uncertainty in longitude caused by the lack of seismometers to the west, in the Pacific Ocean. The hypocentral depths among the four locations are clustered between 39 km and 44 km, and are relatively well constrained by the observation of the pP phase at several stations. The global Centroid Moment Tensor (gCMT, see www.globalcmt.org) centroid location is 75 km to the southwest of the USGS hypocenter, mostly in the up-dip direction. The less well-constrained gCMT centroid depth is 34 km; the centroid location corresponds to the 10 km contour of slab depth as defined by previous seismicity [Hayes et al., 2012].

[7] Several researchers have inverted teleseismic data for space-time models of the coseismic slip distribution for the 2007 Pisco earthquake (C. Ji and Y. Zeng, Preliminary result of the Aug. 15, 2007 Mw 8.0 coast of central Peru earthquake, 2007, available at http://earthquake.usgs.gov/earthquakes/eqinthenews/2007/us2007gbcv/finite_fault.php) [Pritchard and Fielding, 2008; Biggs et al., 2009; Hébert et al., 2009; Lay et al., 2010a; Sladen et al., 2010] and generally find good agreement in the moment rate function, with an initial burst of moment release just after the origin time, and a second larger burst peaking approximately 60–80 s later. However, the spatial pattern of slip, and in particular, the estimated locations of the second, larger slip pulse, differ significantly (Figure 2, left). For instance, the location of the dominant slip patch in the model of Ji and Zeng (http://earthquake.usgs.gov/earthquakes/eqinthenews/2007/us2007gbcv/finite_fault.php) is centered approximately 90 km south-southeast of the main slip patch in the teleseismic-only model ofBiggs et al. [2009]. The discrepancies are caused by different assumptions in the parameterization of the rupture process (particularly the rupture velocity), different types of regularization, and the fact that teleseismic body waves alone provide a relatively weak constraint on the spatial distribution of slip for this relatively compact rupture [Lay et al., 2010a].

image

Figure 2. Comparison of (left) teleseismic-only and (right) geodetic models of slip for the 2007 Mw 8.0 Pisco earthquake. The colored slip contours enclose the major slip zones associated with the second, large pulse of seismic energy release that occurred approximately 60–80 s after the origin time. The trench is indicated by the serrated line, contours of slab depth [Hayes et al., 2012] in 10 km increments are shown by the dashed lines, the USGS epicenter is shown by the white star, and the gCMT solution is shown by the focal mechanism.

Download figure to PowerPoint

[8] Models of the time-integrated spatial variation of slip for the Pisco earthquake based on geodetic data show greater similarity with one another (Figure 2, right), with the second large pulse located offshore of the Paracas peninsula, about 60–70 km due south and 10–15 km shallower than the epicenter [Pritchard and Fielding, 2008; Biggs et al., 2009; Motagh et al. 2008; Sladen et al. 2010]. The two studies that jointly inverted seismic and geodetic data obtained similar results [Pritchard and Fielding, 2008; Sladen et al. 2010], with the latter authors validating their model against tsunami records obtained by DART buoys deployed in the open ocean. Significant slip offshore and up-dip of the Paracas peninsula, as obtained in the joint inversions, is also consistent with local observations of tsunami run-up [Fritz et al., 2008] and direct inversion of the DART tsunami records [Hébert et al., 2009]. This consistency is important because geodetic data can include contributions from aseismic afterslip, while the tsunami observations are sensitive only to coseismic slip.

3. Back-Projection of Short-Period Teleseismic P-Waves

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The 2007 Mw 8.0 Pisco, Peru Earthquake
  5. 3. Back-Projection of Short-Period Teleseismic P-Waves
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[9] Previous P wave back-projection images of the Pisco 2007 rupture were derived using teleseismic signals recorded by a global configuration of 92 seismometers with wide azimuthal (easterly) coverage of the source region [Lay et al., 2010a]. The data were band-pass filtered between 0.05 Hz and 0.4 Hz and back-projected to the source region using an approach similar toXu et al. [2009]. Although back-projection images do not directly estimate slip, the inferred distribution of seismic radiation often corresponds well with finite fault models of slip [Koper et al., 2012]. The Pisco earthquake back-projections ofLay et al. [2010a]are similar in basic characteristics to some of the finite-fault models discussed above, showing an initial burst of seismic energy just down-dip of the epicenter and a larger energy release 60–80 s after the origin time. This later patch of energy was located in the vicinity of the Paracas peninsula, north of the slip concentrations. (For comparison with our TA back-projections, we reproduce the global results ofLay et al. [2010a] in the auxiliary material using the same methodology and control parameters described below.)

[10] Here we back-project teleseismic P waves recorded across the Transportable Array (TA), which was deployed in the western United States at the time of the Pisco earthquake (Figure 3). The data are exceptionally coherent because (1) the stations span a distance range of 54°–74° in which the P waves turn in the relatively homogeneous lower mantle, (2) span an azimuthal range of only 321°–335°N leading to similar sampling of the focal sphere, and (3) were installed with nearly identical instrumentation and site construction. This configuration potentially has lower spatial resolution then the global configuration of Lay et al. [2010a]because of the more limited sampling of vector slowness; however, the TA configuration much more accurately satisfies the back-projection assumption of identical Green functions except for a time shift (H. Yao et al., Subevent location and rupture imaging using iterative back-projection for the 2011 Tohoku Mw 9.0 earthquake, submitted toGeophysical Journal International, 2012). While global arrays do sometimes provide back-projection results consistent with those determined from regional arrays [e.g.,Xu et al., 2009], there are also examples in which the variation in Green function across a global array generates severe back-projection artifacts, as in the case of the 2009 Samoa-Tonga great earthquake doublet [Lay et al., 2010b]. Essentially, regional arrays have signal coherence to significantly higher frequencies than global arrays, extending the bandwidth for which short-period rupture properties of great earthquakes can be probed teleseismically.

image

Figure 3. Locations of 374 seismometers used to image the rupture of the Pisco earthquake. Most of these seismometers were included in the Transportable Array (TA) component of USArray. Seismograms were selected if the initial 10 s of the P wave had a mean correlation coefficient of at least 0.8 with respect to the other seismograms. The inset panel shows the unfiltered, aligned and normalized P waves ground velocity recordings. The high coherence of these P waves allows back-projections to be stable for frequencies up to ∼2 Hz.

Download figure to PowerPoint

[11] Results of the TA back-projection are presented inFigure 4for a lower frequency band defined by a 4-pole Butterworth band-pass with corners at 0.1 Hz and 0.5 Hz, and a higher frequency band defined by the same filter with corners at 0.5 Hz and 2.0 Hz. Otherwise the processing consisted of the same steps: alignment and normalization using a multi-channel cross-correlation (MCCC) algorithm on the first 10 s of unfiltered P waves, calculation of travel times using AK135 [Kennett et al., 1995] and the final USGS hypocenter, gridding of the source region at 0.05° in latitude and longitude at a constant depth of 39 km, beam generation by fourth-root stacking, beam power averaging over a 10-s long sliding window shifted in 1-s increments, and a post-processing step in which spatial frames were time-averaged over a 20-s window. We plot local maxima of beam power with respect to the three dimensions of time, latitude, and longitude to mitigate the influence of spurious spatial maxima (saddle points) created by the swimming or streaking artifact common to conventional back-projection (Yao et al., submitted manuscript, 2012).

image

Figure 4. Comparison of back-projection results with slip models derived jointly from seismic and geodetic data. The colored circles are local beam power maxima as a function of space and time, with circle size proportional to beam power. The colored slip contours were selected to enclose the major portion of slip associated with the second, large pulse of moment release that occurred approximately 60–80 s after the origin time. The trench is shown by the serrated line, contours of slab depth [Hayes et al., 2012] in 10 km increments are shown by the dashed lines, the USGS epicenter is shown by the white star, and the gCMT solution is shown by the focal mechanism.

Download figure to PowerPoint

[12] For both frequency bands shown in Figure 4, the locations of short-period radiation at lag times of 60–80 s are offset along-strike (NNW) by about 50 km, and horizontally in the down-dip direction (ENE) by about 30 km from the regions of major slip as defined by the two joint inversions of seismic and geodetic data. Assuming the rupture occurred at the plate interface, this is equivalent to an increase in depth of 10–15 km, based on the slab contours ofHayes et al. [2012]. The higher frequency locations are about 5–10 km down-dip (ENE) of the lower frequency locations. We conducted several tests to determine the robustness of these results, varying parameters such as the length of the sliding window used to compute beam power (1.5 s–24 s), the style of beam formation (linear vs. fourth root), the length of the frame averaging (10 s–20 s), and the frequency band used to filter the traces prior to back-projection, in particular experimenting with a series of narrow overlapping pass-bands centered at 0.125 Hz, 0.25 Hz, 0.5 Hz, 1.0 Hz, and 2.0 Hz. Back-projections at lower frequencies and with higher levels of averaging showed, as expected, fewer local maxima and broader maps of time-integrated beam power (i.e., lower spatial resolution), however there were only minor changes in the locations of the short-period radiators. In general, the locations of significant maxima varied over a region similar to those shown inFigure 4, bracketed by slab-depth contours of 30–50 km, the epicenter to the north, and the Paracas peninsula to the south. We note that an automated back-projection of North American data dominated by TA stations for this earthquake performed with different software, and tuning parameters shows results similar to ours (seewww.iris.edu/spud/backprojection/112167, last accessed Jan. 20, 2012).

[13] We checked for any bias in the TA-derived back-projection locations by comparing USGS epicenters to locations of peak time-integrated beam power for 13 earthquakes that occurred in the source region (Figure 5). Eleven of these events were aftershocks that occurred within the first month after the mainshock and two were earthquakes that occurred several months before the mainshock. The events varied between Mw5.6 to 6.7 and can be treated essentially as point sources. Back-projections were carried out in a pass-band of 0.5–2.0 Hz using the same processing described above, including the same static time corrections derived from the MCCC analysis of the main shock. The mean mis-location vector is in the ENE direction but is only 8 km long and cannot explain the much larger offset between the locations of the peak slip and peak beam power for the mainshock.

image

Figure 5. Mislocation vectors for 11 aftershocks and two previous earthquakes. Arrows point from locations of peak time-integrated beam power toward USGS epicenters. The mean mislocation vector is shown outlined in red at the peak beam power location for the mainshock. The slight tendency for back-projection derived locations to be northeast of the corresponding USGS epicenters is not large enough to account for the difference between the locations of peak beam power and peak slip in the mainshock.

Download figure to PowerPoint

4. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The 2007 Mw 8.0 Pisco, Peru Earthquake
  5. 3. Back-Projection of Short-Period Teleseismic P-Waves
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[14] During the 2007 Mw8.0 Pisco earthquake, bursts of coherent short-period seismic energy release, as imaged by back-projection of teleseismic P waves recorded in North America, originated in different locations than the main coseismic slip, as defined by joint inversions of long-period seismic and geodetic data. The ∼50 km along-strike (NNW) offset of the short-period radiation and the ∼30 km strike-perpendicular (ENE) offset of the short-period radiators are significant, even when considering the uncertainties in the approaches. While along-strike variations in megathrust rupture properties have been repeatedly observed in the past, the observation of significant down-dip offset of the short-period seismic radiation is more unusual and is similar to those made for the 2011 Mw 9.0 Tohoku earthquake, the 2010 Mw 8.8 Chile earthquake, and the Mw 9.15 Sumatra earthquakes (though the magnitude of the offsets differ).

[15] Taken together, these observations support a new framework for interpreting megathrust earthquakes in terms of depth dependent zones of frictional sliding behavior [Lay et al, 2012]. In particular, the spectra of seismic radiation from deeper portions of the megathrust ruptures are relatively enriched in short period energy compared to the shallower portions. Physical factors that may play a role in creating the spectral enrichment include changes in material contrast across the fault zone, variation in pore fluid type and/or content, and changes in fault roughness (perhaps created by changing pressure-temperature conditions). Future work on determining the relative importance of these factors will require numerical simulation of earthquake ruptures in realistic Earth models and/or laboratory-based rock mechanics experiments. For example, recent dynamic rupture simulations of M7 normal faulting earthquakes showed that using a realistic, depth-dependent distribution of initial stresses led to a decoupling of slip and slip-rate, with peak slip-rates located in the deeper portions of the fault plane, down-dip of the peak slip [Roten et al., 2011]. It remains to be seen if such a phenomenon holds for reverse faulting M8-9 events, but if so this may provide an explanation for our observations.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The 2007 Mw 8.0 Pisco, Peru Earthquake
  5. 3. Back-Projection of Short-Period Teleseismic P-Waves
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

[16] This work made use of GMT and SAC software. The IRIS DMS data center was used to access the seismic data. This work was supported by NSF grants EAR0635570 (T.L.), EAR098660 (T.L.), and EAR0951558 (K.D.K.). We thank Alex Hutko for advice and discussion related to this work. We thank Gavin Hayes and an anonymous referee for helpful reviews.

[17] The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The 2007 Mw 8.0 Pisco, Peru Earthquake
  5. 3. Back-Projection of Short-Period Teleseismic P-Waves
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The 2007 Mw 8.0 Pisco, Peru Earthquake
  5. 3. Back-Projection of Short-Period Teleseismic P-Waves
  6. 4. Conclusions
  7. Acknowledgments
  8. References
  9. Supporting Information

Auxiliary material for this article contains results from a global back-projection of P waves.

Auxiliary material files may require downloading to a local drive depending on platform, browser, configuration, and size. To open auxiliary materials in a browser, click on the label. To download, Right-click and select “Save Target As…” (PC) or CTRL-click and select “Download Link to Disk” (Mac).

Additional file information is provided in the readme.txt.

FilenameFormatSizeDescription
grl29101-sup-0001-readme.txtplain text document5Kreadme.txt
grl29101-sup-0002-fs01.epsPS document2903KFigure S1. P waves from 55 globally distributed seismometers.
grl29101-sup-0003-fs02.epsPS document467KFigure S2. Results of back-projecting data from the seismometers shown in Figure S1.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.