A study of the 2006 and 2007 earthquake sequence of Pisco, Peru, with InSAR and teleseismic data

A study of the 2006 and 2007 earthquake sequence of Pisco, Peru, with InSAR and teleseismic data
of 6
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A study of the 2006 and 2007 earthquake sequence of Pisco, Peru, with InSAR and teleseismic data M. E. Pritchard 1 and E. J. Fielding 2 Received 22 January 2008; revised 14 March 2008; accepted 20 March 2008; published 15 May 2008. [ 1 ] We combine interferometric synthetic aperture radar (InSAR) and teleseismic body waves to study the largest earthquake (  M  w  8.1) in a sequence of events on thesubduction megathrust near Pisco, Peru. Our analysisincludes some of the first InSAR data from the ALOSsatellite and wide swath data from the Envisat satellite. Theteleseismic data indicate the slip maximum occurred 60– 90 seconds after the mainshock started. The InSAR dataconstrain the main slip patch to be about 70 km from thehypocenter, suggesting an extremely low rupture velocity(<1.5 km/s) or long slip rise time. No large earthquake hasoccurred in the 2007 rupture area since at least 1746 and possibly 1687, suggesting significant aseismic deformationin the area. The slip deficit apparently cannot be filled withrapid after-slip. In addition, the area where the Nazca Ridgeis subducting appears to be either a seismic gap or a persistent area of aseismic slip.  Citation:  Pritchard, M. E., andE. J. Fielding (2008), A study of the 2006 and 2007 earthquakesequence of Pisco, Peru, with InSAR and teleseismic data, Geophys. Res. Lett. ,  35 , L09308, doi:10.1029/2008GL033374. 1. Introduction [ 2 ] Between 1992 and 2007, nine  M  w  7 or larger earth-quakes have rocked thesubductionzone offshore of southernPeru and northern Chile. To assess the location and size of earthquakes that could occur in areas that have not recentlyslipped, it is necessary to determine the areas of thesubduction megathrust that moved in these most recent earthquakes.[ 3 ] Traditionally, the rupture area of past earthquakes isapproximated by the region where numerous aftershocksoccur [e.g.,  Lay and Wallace , 1995] (Figure 1), but theadvent of global seismographic networks revealed that earthquake slip was not uniformly distributed within theaftershock area [e.g.,  Lay and Wallace , 1995]. Satellite- based interferometric synthetic aperture radar (InSAR) canreveal details of the earthquake slip distribution beyondconventional body wave seismology, particularly for largeevents (  M  w  > 7.5) [e.g.,  Pritchard et al. , 2007], withimplications for understanding stress triggering, postseis-mic deformational processes, and assessing future seismichazard.[ 4 ] AsequenceofearthquakesnearPisco,Peru,culminatedinthe  M  w 8.1earthquakeon15August2007(Figure1),whichis hereafter called the 2007 mainshock. We use InSAR andteleseismic body wave data both jointly and separately toinfer the slip distribution from this earthquake and comparethe results with nearby earthquakes. The sequence appar-ently began in October 2006, when two  M  w  > 6 earthquakesoccurred near the USGS-derived hypocenter of the 2007mainshock. We perform a teleseismic analysis of the larger of these events (October 20,  M  w  6.7, hereafter called the  M  w  6.7 2006 earthquake) to better understand the fault depth and dip in this location. 2. InSAR Data [ 5 ] Thanks to special efforts on the part of the EuropeanSpace Agency (ESA) and the Japanese Space Agency(JAXA), InSAR data for the 2007 mainshock are availablefrom at least ten different orbital tracks, more than for anyearthquake studied to date (Figure 2 and Table S1 in theauxiliary material). 1 The 2007 mainshock data include bothascending and descending orbits from ESA’s Envisat andERS-2 satellites (with C-band radars, 5.6 cm wavelength)and ascending data from JAXA’s ALOS satellite (L-band,23.6 cm). Because of the generally arid climate on the coast of Peru, the interferometric coherence is excellent. However,theL-bandradarclearlyhassuperiorcoherencetotheC-bandradars in sandy or agricultural areas, particularly near thecoast (Figure 2). Due to infrequent observations before the2007 mainshock, we were not able to form any interfero-grams that spanned just the 2006 earthquakes.[ 6 ] The InSAR data span different time intervals, andmost includes several weeks to months of postseismicdeformation, although we think the effect on the inferredslip distributions is small. Postseismic deformation follow-ing the 2007 mainshock appears similar to the 2001  M  w  8.5Peru [  Pritchard  , 2003] and 1995  M  w  8.1 Chile [  Pritchard and Simons , 2006] earthquakes in that the inferred peak  postseismic slip appears to be about an order of magnitudelower than the peak coseismic slip. For example, while thecoseismic interferogram from Envisat track 82 includesabout 14 fringes (39 cm), a postseismic pair spanning 11to 80 days after the earthquake includes no obvious tectonicdeformation. Of course, the most rapid deformation imme-diately following the earthquake might have been missed.All of the ERS-2 and some of the Envisat interferogramsspan the 2006 earthquakes (Table S1), but the predictedsurface displacement from these earthquakes is only a cm or less, or at most a few percent of the 2007 mainshock deformation. 1 Auxiliary materials are available in the HTML. doi:10.1029/ 2008GL033374. GEOPHYSICAL RESEARCH LETTERS, VOL. 35, L09308, doi:10.1029/2008GL033374, 2008 1 Department of Earth and Atmospheric Sciences, Cornell University,Ithaca, New York, USA. 2 Jet Propulsion Laboratory, California Institute of Technology,Pasadena, California, USA.Copyright 2008 by the American Geophysical Union.0094-8276/08/2008GL033374 L09308  1 of 6  [ 7 ] The Envisat ASAR data includes data from four orbital tracks in the standard strip-map or image mode beam 2 that has a radar viewing geometry the same asERS SAR, and also includes data from two tracks acquiredin ScanSAR or wide-swath (WS) mode. The Envisat WSmode scans back and forth among five subswaths to image awidth of 400 km instead of the roughly 100 km width of theimage mode, at the expense of coarser resolution. ScanSAR interferometry has not been used previously to study largeearthquakes, largely because the bursts must be aligned between the two acquisitions. ESA started a new process toimprove the WS burst alignment in the fall of 2006 [  Rosichet al. , 2007].[ 8 ] For further technical details on InSAR processing, seethe auxiliary material. 3. Seismic Data [ 9 ] We select azimuthally well-distributed digital P andSH teleseismic displacement records (between 30 and 90)from the digital global network [  Butler et al. , 2004]. See  Pritchard et al.  [2006] for more details about selection and processing of the teleseismic data. For the 2007 mainshock,we use 15 P and 15 SH waveforms extending 150 secondsafter the arrival of each phase, although we cut the recordshort at a few stations to avoid interference from PP(Figure 3). For the  M  w  6.7 2006 earthquake, we use 21 Pand 15 SH waveforms extending 60 seconds after the arrivalof each phase (Figure S1). 4. Modeling Methods [ 10 ] We use the method of   Ji et al.  [2002] to invert for fault slip using the seismic and InSAR data both separatelyand together. The seismic waveforms are transformed intothe wavelet domain to better utilize the temporal andfrequency content of the waveforms while solving for therupture properties. The forward model is generated by acombination of point-source dislocations to create subfaultswith a specified fault geometry (see Table S2). Calculationsare performed in an elastic layered media using the velocitymodel of   Husen et al.  [1999]. See  Pritchard et al.  [2007] for discussion of the validity of this model and its impact.[ 11 ] The detailed location of the fault interface betweenthe Nazca and South American plates is not well-knownwithin the area of interest [e.g.,  Spence et al. , 1999;  Pritchard et al. , 2007], and these details (particularly thevariations of dip as a function of depth) can impact slipinversions. Therefore, we test different configurations of thefault plane, and we list the parameters used to generate themodels in Table S2. There is evidence for along-strikevariations in fault dip between central Peru and northernChile based on relocated aftershocks and USGS and GlobalCMT locations of large earthquakes (Figure S2). Althoughthe 2007 earthquake occurred within the Peruvian flat-slabregion, most of the slip occurs updip of where the slabflattens [e.g.,  Cahill and Isacks , 1992].[ 12 ] We perform non-linear inversions using simulatedannealing, which searches through bounded regions of  parameter space to find models that best match the data ina least-squares sense. See Table S2 for the range of  parameters used. In addition to finding a model that wellfits the data, we also seek to minimize the roughness of theslip distribution using a Laplacian operator and for someinversions we also minimize the difference between thecalculated seismic moment and an a priori value. Some testsof the resolving power of the technique are in  Ji et al. [2002], and a discussion of combining seismic and InSAR  Figure 1.  Oblique Mercator projection of recent largeshallow interplate thrust earthquakes in central and southernPeru plotted over seafloor bathymetry (color) and shadedtopography (gray scale) with the trench shown by a red barbed line. The rupture areas of the earthquakes areapproximated by the labeled ellipses to the right of thetrench (defined by the aftershock locations from the USGScatalog, except for the 1974 earthquake where aftershocksare from  Langer and Spence  [1995]). We show contours of earthquake slip (from models that fit both seismic andgeodetic data) which are significantly smaller than theellipses for the 1996 and 2001 Peru earthquakes with 1 and4 m contours, respectively [  Pritchard et al. , 2007], and for the 2007 earthquake from this work (3 m contours). Veryapproximate rupture areas of historic earthquakes are shownwith their dates and magnitudes as dashed lines to the left of the trench [e.g.,  Dorbath et al. , 1990;  Spence et al. , 1999].The subducting Nazca Ridge is labeled and the direction of  Nazca Plate motion is shown by the arrow. L09308  PRITCHARD AND FIELDING: THE 2006 AND 2007 PISCO, PERU, EARTHQUAKES  L09308 2 of 6  Figure 2 L09308  PRITCHARD AND FIELDING: THE 2006 AND 2007 PISCO, PERU, EARTHQUAKES  L09308 3 of 6  data can be found in the auxiliary material and  Pritchard et al.  [2006, 2007]. 5. Results [ 13 ] The inferred slip during the  M  w  6.7 2006 earthquakeis shown in Figure 4a, and the predicted waveforms fromthis model are compared to the data in Figure S1. While theGlobal CMTsolution for this earthquake has a dip of 16, wefind that a larger dip can also match the data and is moreconsistent with the inferred dip of the mainshock whichruptured in this location. We find that the Global CMTlocation is shifted west of the hypocenter and this offset cannot be explained by the slip distribution (Figure 4a).Based on our work with  M  w  6-7 earthquakes in thissubduction zone [  Pritchard et al. , 2006, 2007], we suspect that the CMT locations are incorrect.[ 14 ] The slip distributions from the teleseismic-only,geodetic-only and joint inversions for the 2007 earthquakeare shown in Figure 4. Our model predicts the maximumsubaerial vertical deformation to be about 30 cm of subsi-dence, which agrees with the field observation of no verticaldeformation beyond the 40 cm tidal zone [  Audin et al. ,2007]. The fit to the seismic data is not significantlydifferent between the seismic-only and the joint inversion(Figure3).ThefittotheinterferogramsisshowninFigureS3and the root-mean-squared residual (RMS) for each inver-sion is in Table S3. In general, the RMS values calculatedfor full (not subsampled) interferograms are larger for the2007 mainshock than for the other large earthquakes wehave studied with InSAR in the region [  Pritchard and Simons , 2006;  Pritchard et al. , 2007]. Part of the reasonis that the InSAR coherence in the Pisco area is generally poorer and more heterogeneous, which complicates phase Figure 2.  (a) Ascending Envisat wide swath data, where the red box shows the regions of Figures 2b–2q). The trench (red barbed line), Global CMT location (red mechanism), and hypocenter (USGS, black star), are also shown. (b–i)Interferograms from 10 orbital tracks spanning the 2007 earthquake (see Table S1 for dates). Only a fraction of eachinterferogram used in the inversion is shown. All scenes have been unwrapped and re-wrapped at a 5 cm interval. Thesatellite to ground radar line-of-sight (LOS) is shown with an arrow. The descending Envisat/ERS interferograms containmore fringes than the ascending ones because the westward and downward coseismic ground motion add constructively for the descending orbits but destructively for the ascending orbits (as observed for the 2001 Peru and 1995 Chile earthquakes[  Pritchard and Simons , 2006;  Pritchard et al. , 2007]). The largest InSAR displacement observed is about 1.1 m from theascending ALOS track 111 which has a higher incidence angle (about 39) than the ERS-2/Envisat Image Mode 2interferograms (about 23). (j–q) Modeled interferograms. The residual between data and model is shown in Figure S3. Figure 3.  Teleseismic displacement data (P and SH) used in the slip inversions for the 2007 mainshock (black lines) andcalculated synthetics from the teleseismic-only (gray lines), and joint inversions (black dashed lines). To the left of eachtrace is the station name, epicentral distance (lower number) and azimuth (upper number, which increases from the bottomof the page to the top). The type of record (P or SH) is listed above each station name. Each amplitude has been normalized by the maximum displacement, shown in microns in the upper right of each trace. L09308  PRITCHARD AND FIELDING: THE 2006 AND 2007 PISCO, PERU, EARTHQUAKES  L09308 4 of 6  unwrapping. Some of the misfit is correlated with topogra- phy (particularly obvious in valleys), indicating likelychanges in tropospheric water vapor.[ 15 ] The 2007 earthquake has one large concentration of slip 60-90 seconds after the earthquake begins (Figure 4d).The distance between the main slip pulse and the hypocen-ter depends on the rupture properties, particularly therupture velocity and/or the slip rise time, which trade-off with each other. While our inversions allow us to explore arange of parameters, we are unable to use the  Ji et al.  [2002]method to test complex scenarios that involve more thanone peak of slip at a given fault location. Therefore, wediscuss an end-member set of models where we explain thedata with variations in rupture velocity and a single pulse of slip at each fault location (with variable rise time).[ 16 ] If we constrain the rupture velocity in the teleseismicmodel to conventional values for subduction zone earth-quakes (2-4 km/s) the dominant slip patch is >100 km fromthe hypocenter (Figure S4). This asperity location and theteleseismic fault slip map (not shown) generally matchesresults from two independent groups who use modifiedversions of the  Ji et al.  [2002] method (O. Konca andA. Sladen at Caltech, available at slip_history/2007_peru/peru.html; and C. Ji and Y. Zengat the USGS and UCSB, available at eqcenter/eqinthenews/2007/us2007gbcv/finite_fault.php).In detail, there are many differences between these modelsand ours, including we use two fault planes instead of one,we use a different velocity model and subfault size, theUSGS/UCSB solution includes surface waves, and we favor a lower rupture velocity.[ 17 ] The InSAR data require the slip to be significantlycloser to the hypocenter (Figure 4) than the previous tele-seismic only inversions, and implies an average rupturevelocity of 1.3 km/s (about 30% of the shear wave velocity).When we expand the allowed rupture velocities in theteleseismic inversion (0.1-3.5 km/s), the main slip pulsemoves closer to the hypocenter, but the predicted surfacedeformation still does not well fit the InSAR data (Table S3).(Teleseismic-only models of other large subduction earth-quakes also produce poor fits to InSAR data [e.g.,  Pritchard et al. , 2007].)  Sladen et al.  [2007] independently estimated arupture velocity of <2 km/s based on the aftershock distri- bution (130 km) and rupture time of 110 seconds. The verylow rupture velocity is comparable to earthquakes that generate abnormally large tsunamis for a given seismicmoment (called tsunami earthquakes). This earthquake hasnot been classified as such, although tsunami heightsreached up to 10 m [  Fritz et al. , 2008].[ 18 ] A time lag between the earthquake initiation and theslip maximum is observed at several other large subductionearthquakes in this area: the 1974  M  w  8.0 Peru, 1995  M  w  8.1Chile, 1996  M  w  7.7 Peru, and 2001  M  w  8.5 Peru earth-quakes [  Pritchard et al. , 2007]. The 2007 earthquake alsohas the same southward rupture direction as these earth-quakes [  Pritchard et al. , 2007]. However, the 2007 earth-quake is unlike the other southern Peru earthquakes in that it had two  M  w  > 6 foreshocks near the hypocenter 10 months before the mainshock. The 1995 Chile earthquake also had a  M  w  6.2 foreshock near the hypocenter 4 months before theearthquake [e.g.,  Delouis et al. , 1997].[ 19 ] Even though the Pisco area has been inhabited for centuries, no large earthquake had been reported in the 2007rupture area since at least 1746 and possibly 1687 [  Dorbathet al. , 1990]. The 10 m or more of slip in the 2007mainshock has only relieved a portion of the 20 m slipdeficit accumulated in the last   300 years. Furthermore, the2007 mainshock has only partly filled the 100 km seismicgap [ Swenson and Beck  , 1999] between the 1996 and 1974Peru earthquakes: a gap remains, particularly at the crest of the Nazca Ridge (Figure 1). While the slip deficit may befilled in future earthquakes, it is also possible that thesubduction of the Nazca Ridge is at least partly aseismic(e.g.,  Dorbath et al.  [1990], although see  Okal et al.  [2006]for an opposing view). As the intersection between the Nazca Ridge and the trench migrates to the southeast, Spence et al.  [1999] speculate that the leading edge of ridgeis more likely to host earthquakes than the trailing edge. Figure 4.  Contours of slip from the 2007 mainshock from inversions using (a) only seismic data, (b) only InSAR data,and (c) both data sets. The maximum slip is about 15 m in the joint model and the contour interval is 2 m. The contours of slip from the seismic inversion of the  M  w  6.7 2006 earthquake and the Global CMT locations are shown in Figure 4a. Thehypocenter location (USGS) of the  M  w  6.7 2006 earthquake and the  M  w  6.0 from 26 October 2006 are shown as a white star with a black outline in all plots. For the 2007 mainshock, the USGS location is shown as the black star and the focalmechanism is from the Global CMT catalog. L09308  PRITCHARD AND FIELDING: THE 2006 AND 2007 PISCO, PERU, EARTHQUAKES  L09308 5 of 6
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks