T53C-01 INVITED
High-Resolution Imaging of San Andreas Fault at Parkfield, California, Using Seismic Velocity and Anisotropy Tomography and Seismic Interferometry
We characterized the detailed structure of the San Andreas fault zone at multiple scales using an extensive dataset collected around the SAFOD site from our long-term deployments of PASSCAL and USArray seismic instruments, and the USGS Northern California and UC Berkeley HRSN networks, SAFOD borehole logs, borehole seismometers, and several active-source projects. A suite of techniques are employed to better constrain the internal structure of the fault zone, including seismic travel-time tomography, shear-wave splitting tomography and seismic interferometry. Adaptive-mesh double-difference tomography is used to derive high-resolution Vp and Vs models around the fault zone with the waveform cross-correlation derived differential times. Knowing three-dimensional (3-D) Vp/Vs variations is helpful to have a more complete characterization of the mechanical properties and geological identity of fault zone materials. Vp/Vs variations are reliably determined by the inversion of S-P time differences constructed only from similar P and S ray paths. Our velocity models show the high-velocity granitic rocks on the southwest side of the fault, a complex low-velocity zone beneath and southwest of the surface fault trace, and an extensive low-velocity zone overlying deeper bedrock on the northeast side. We systematically analyzed shear wave splitting for seismic data observed at PASO and UC Berkeley HRSN networks. Although polarization direction of the fast shear wave and the delay time show substantial scatter for different events observed at a common station, there are spatially consistent patterns when projecting them to various depths along corresponding ray paths, derived from a 3-D shear velocity model. We developed a 3-D shear-wave splitting tomography method to image the spatial anisotropy distribution by back projecting shear wave splitting delay times along ray paths. The anisotropy percentage model shows strong heterogeneities, consistent with the strong spatial variations in both measured delay times and fast polarization directions. The fault zone is highly anisotropic down to a depth of ~4 km and then becomes less anisotropic at greater depths. Seismic interferometry means the process of generating new seismic responses by cross-correlating seismic observations at different receiver locations. Based on the principle of seismic interferometry and reciprocity, we created pseudo shot profiles as if a shot was excited at one earthquake location and observed by a set of receivers at other earthquake locations. Preliminary results show that several reflectors may exist inside the fault zone.
T53C-02
Small-Scale Variations in SKS Splitting Near Parkfield, California
Understanding the nature of deformation near plate boundaries is a key element of understanding the nature of the dynamics of plate tectonics. One particularly interesting question is how seismic anisotropy measured in teleseismic body waves can provide constraints on the nature of small-scale (i.e., <10 km) deformation. The San Andreas Fault (SAF) near Parkfield, California provides an ideal zone to study small-scale variations in seismic anisotropy. In this study, we utilized data from the PASO-DOS broadband seismic array. The array has an approximate aperture of 15 km and is diagonally bisected northwest to southeast by the surface trace of the SAF. We performed shear wave splitting analyses to determine fast polarization directions and delay times from 5 high-quality events recorded at the array. Results from the shear wave splitting analysis reveal clear variations in apparent seismic anisotropy across the array. Data for two events from NW backazimuths exhibited first-order variations in seismic anisotropy across the SAF, with a clear, smooth rotation in fast direction from ENE-WSW on the east side of the SAF to WNW-ESE on the west side. Splitting times range from 1.33 to 2.35 sec, average 1.90 sec, but do not show a clear regional trend. Data for three other events with either W or N backazimuths exhibit null measurements. While these variations may be due to strong isotropic lateral heterogeneity, the existence of clear variations in crustal anisotropy as imaged from both receiver function analysis and local S splitting suggests that the variations we have documented are most likely due to effects of seismic anisotropy in the crust. The null splitting results from events with W and N backazimuths are consistent with previous interpretations of E-W fast directions in the mantle across the region. Our results suggest the presence of broad-scale asthenospheric deformation, which generates a pervasive E-W fast direction across the region. The rapid sweep in fast directions near the SAF is consistent with a first-order change in uppermost crustal anisotropy domains across the fault. The implications of our results are threefold: 1) shear wave splitting from teleseismic body waves has the potential to provide unique constraints on deformation at plate boundaries; 2) crustal anisotropy have a much larger influence on shear wave splitting measurements than previously assumed; and 3) backazimuthal variations in shear wave splitting documented at single stations may not necessarily be due solely to variations in anisotropic structure with depth.
T53C-03
Small-Scale Structures Derived From Microearthquake Locations Using SAFOD and HRSN Data
In order to improve the understanding of the small-scale fault within the San Andreas Fault Zone at Parkfield, California, we analyze high frequency seismic data (4000 Hz) recorded at the San Andreas Fault Observatory at Depth (SAFOD). Deep borehole seismic data from the Pilot Hole (PH) and Main Hole (MH) collected from August to mid September 2006 have been processed together with additional data (250 Hz) from 13 stations of the shallow borehole High Resolution Seismic Network (HRSN). The MIMO system (Oye and Roth, 2003) is used for automatic processing of the data. A single channel STA/LTA based detection algorithm provides primary detections. Detections are associated to form a single event with a generalized beam forming algorithm, similar to the approach used in array seismology. A detection on the MH instruments was required, whereas any further detection from PH and HRSN channels was associated with the event when appropriate. Over this 6 week period about 8900 events were identified, of which about 1000 events were only detected with one of the MH geophones. Another 5100 events were detected on one MH geophone and one PH or HRSN station and 2100 events were detected with both MH instruments. Most of these events were associated with tool-slip in the borehole. However, we observed a series of about 20 ultramicroaftershocks (M < 0) following the August 11th M 1.8 "Hawaii" target event. These events were only observed on the MH. About 720 events were detected on both MH geophones and at least one HRSN station. Subsequently, P- and S-wave onsets as well as azimuth angles are determined automatically using the MIMO monitoring software. To locate the events, we conduct a grid search using look-up tables created by ray tracing in a 3D velocity model that was provided by S. Roecker. We have also established semiautomatic processing routines for shear wave splitting analysis for application to events that locate close to the MH stations. Thus we hope to image shear wave splitting along the San Andreas Fault and map the anisotropy related to the fracture density distribution within the Parkfield section of the San Andreas Fault Zone.
T53C-04
High-Resolution Imaging of the San Andreas Fault from Fault-Zone Trapped Waves Recorded at the SAFOD Borehole Seismograph and Surface Array
Highly damaged rocks along the San Andreas Fault at Parkfield create a low-velocity waveguide to trap seismic waves. We recorded prominent fault-zone trapped waves (FZTWs) at the SAFOD mainhole seismograph at 3km depth and the surface seismic array across the fault for microearthquakes. A systematic waveform analysis of borehole and surface seismograms from several hundreds of local earthquakes and aftershocks of the 2004 M6 Parkfield earthquake allowed us to evaluate the variations in rock damage magnitude and extent on the SAF with high-resolution. The FZTWs are characterized by relatively large amplitudes and dispersive wavetrains at 3-12 Hz following S-waves. The duration time of dominant FZTWs after S-arrivals increase with the travel distance between the source and receiver either along the fault strike or with the depth, showing the continuous low- velocity waveguide existing on the SAF at Parkfield. We measured duration time of FZTWs after S-arrivals for the events at different depths with raypath incidence angles from these events to the seismic station smaller than 30o from vertical. The wavetrain lengths of FZTWs measured at surface stations within the fault zone show a progressively increasing trend from ~1.2 s to ~2.2 s as the event depths increase from 2.6 km to 11.7 km. These measurements are confirmed by the data recorded at the SAFOD mainhole seismograph. In contrast, the seismograms recorded at seismographs installed in the SAFOD pilot borehole ~1.8 km away from the SAF and the surface stations deployed out of the fault zone show much brief wavetrains after body waves. These observations indicate that the low-velocity waveguide formed by the damaged rock on the SAF likely extends across seismogenic depths with prominent seismic velocity reduction at depths above ~7-8 km. The smaller velocity reduction on the deeper portion of the fault damage zone is probably due to the larger confined stress at greater depths. The surface array data show that the damage zone on the SAF is not laterally symmetric but extends farther on the southwest side of the main fault trace. This could be due to rocks already weakened from previous faulting. It could also be due to greater damage in the extensional quadrant near the propagating crack tip of Parkfield earthquakes. We modeled these FZTWs using 3-D finite-difference methods. The models suggest that, on average, the fault zone cross section consist of a composite of two nearly vertical layers, one a 30-40-m- wide fault core, the other a surrounding ~150-250-m wide damage zone depending on the depth. The damage zone velocities range between 70-80% of the fault zone wall rocks, while those of the core are even less, going as low as 40-50% of the intact rock. The widths and velocity reductions of the fault core and damage zone at ~3-km depth in our model are consistent with the direct measurements of fault-zone properties in the SAFOD mainhole [Hickman et al., 2005] as well as the borehole observations of fault guided waves following P-waves [Ellsworth and Malin, 2006]. We interpret the distinct low-velocity waveguide on the Parkfield SAF as being a zone of accumulated damage from recurrent major earthquakes, including the 2004 M6 earthquake. This type of damage varies with depth and also along the strike, and may relate to the on- and near-fault variations in stress and slip distribution during earthquake rupture.
T53C-05
Dynamic Weakening of the San Andreas Fault by the 2004 Sumatra-Andaman Earthquake
The time-varying properties of seismic scatterers have recently been used to probe stress-induced changes at seismogenic depth within the San Andreas Fault zone near Parkfield, CA. Temporal properties are measured by using the decorrelation index for a waveform pair from repeating earthquakes. We have been monitoring the behavior of a group of time-dependent scatterers for a 20-year period (1987-2007), using well-recorded repeating earthquake sequences. These scatterers are located within the fault zone at seismogenic (about 3 km) depth and are interpreted as fluid-filled fractures whose properties change due to the stress-induced migration of fluids. We have found three dramatic responses of the time-dependent scatterers during this 20-year period: to the 1993 Parkfield Aseismic Transient, to the 2004 Mw 6.0 Parkfield earthquake, and a third excitation about 3 months after the Parkfield earthquake. For this third excitation, the magnitude of change in decorrelation index is comparable to those observed for the 1993 Parkfield Aseismic Transient and for the 2004 Parkfield earthquake. We were able to limit the onset time of the third excitation to a 3.5-day time window between 22 and 26 December 2004, by combining the decorrelation indexes obtained from a suite of repeating earthquake sequences. This excitation is probably of tectonic origin for two reasons. First, it is unlikely to be environmentally induced, because the scatterers are deep, and there was no precipitation (the likeliest environmental influence) within the time interval of interest. Second, the excitation was also observed as a reduction in the recurrence interval and seismic moment of repeating earthquakes. This reduction, however, did not significantly change the corresponding seismic slip rate (proportional to seismic moment divided by recurrence interval), a result that is consistent with the absence of a velocity change at nearby GPS stations. This behavior suggests that there was a temporary reduction in the strength of the fault, rather than an increase in slip rate. The most dramatic event to occur during this 3.5-day time interval is the 26 December 2004 Mw 9.1 Sumatra- Andaman earthquake, which took place 7 hours before the end of the interval. This timing is strongly suggestive that a form of remotely-triggered fault weakening occurred, due to the dynamic stresses from this seismic event. We hypothesize that fault strength was temporarily reduced by a stress-induced increase in fault-zone pore pressure.
T53C-06
Detection of Subsurface Stress/Strain Changes with Active Source Monitoring at the Parkfield SAFOD Drill Site
The time-varying stress field at seismogenic depth is arguably the single most important parameter for understanding the earthquake triggering process. Measuring stress changes within seismically active fault zones has consequently been a long-sought goal of seismology. It is well known from laboratory experiments that seismic velocities vary with the level of the applied stress. In principle, this dependence constitutes a stress meter, provided the induced velocity changes can be measured precisely and continuously. We have conducted several continuous active-source cross-well experiments to measure in situ seismic velocity changes along fixed baselines at the Earth's surface and near-seismogenic depth. At the Parkfield SAFOD drill site in 2005-2006, we conducted a continuous active-source cross-well experiment to measure in-situ seismic velocity changes across a ~10 m baseline at ~1 km depth. Over a two-month period, we found a 0.3% change in the average S-wave velocity, which shows a good negative correlation with barometric pressure, corresponding to a stress sensitivity of 2.4x10{-7}Pa-1. We also observed two large excursions in the delay-time measurement, corresponding to 0.55% and 0.15% decreases of seismic velocity, that are coincident with two earthquakes that are among those predicted to produce the largest coseismic stress changes at SAFOD. Interestingly, the two excursions started approximately 10 and 2 hours before the events, respectively, suggesting that they may be related to pre-rupture dilatancy observed in the early laboratory studies, and that cross-well monitoring might provide an effective tool for understanding the stress changes that accompany and perhaps precede seismic activity.
T53C-07 INVITED
Signatures for Structural Asymmetry in Fault Zone Rocks from the San Andreas and the North Anatolian Fault Systems: Implications for Preferred Direction of Earthquake Ruptures
Theoretical considerations indicate that earthquake ruptures on faults separating different solids have a preferred propagation direction (e.g., Weertman, 1980; Ampuero and Ben-Zion, 2007). Repeated occurrence of such ruptures is expected to lead to more shallow damage on the side of the fault with faster seismic velocities at depth (e.g. Ben-Zion and Shi, 2005). To test this hypothesis, expressions for structural asymmetry with respect to the slip zone of major faults in the southern San Andreas and the North Anatolian fault systems were observed and mapped at various scales in several sites along each fault. Geological mapping of shear fabric on a cm to meter fault core scale, subsidiary faults and fault rocks on a 10's of meters fault zone scale and pulverized rocks on a 100's of meters damage zone scale show systematically damage asymmetry. For the San Andreas, San Jacinto and Punchbowl faults the northeast side is more damaged. For of the 1943 and 1944 rupture zones on the North Anatolian fault, the south and north sides are more damaged, respectively. Similar sense of asymmetry for the San Jacinto fault was observed seismically for a ~100 m wide low velocity layer south of the our mapping sites (Lewis et al., 2005) and in a morphometric analysis of erosion patterns in the vicinity of our mapping sites (Wechsler et al., 2007). Asymmetric erosion patterns along the North Anatolian fault including the presence of almost all the river valleys length south of the fault along the 1943 rupture and vice-versa along 1944 rupture, strong asymmetry in drainage density and other related morphometric parameters between two correlative rock bodies on the two sides of the 1944 rupture, and significant difference in the density of gully networks in bad-land terrains across the 1943 rupture. All these signals are consistent with higher rock damage south and north of the 1943 and 1944 ruptures, respectively. The asymmetric damage patterns are compatible with preferred rupture directions northwestward on the examined fault sections of the southern San Andreas system, and eastward and westward on the 1943-1944 rupture sections of the North Anatolian fault, respectively (as occurred in these two recent North Anatolian fault earthquakes). Regional and local tomographic studies (e.g. Fuis et al., 2003; Scott et al., 1994) show that the more damaged northeast sides of the San Andreas and San Jacinto faults are on the blocks with faster seismic velocities at depth. Significant damage content in sedimentary rocks of the Juniper Hills formation on the southwest side of the San Andreas Fault in the central Mojave section indicates that dynamic generation of damage can occur very close to the surface of the Earth, in agreement with other indications for minimal exhumation of pulverized and other damaged fault zone rocks. These asymmetric damage patterns correlated to the velocity structure (where known) with shallow inferred generation depth are compatible with predicted outcomes for rupture along a bimaterial interface (Ben-Zion and Shi, 2005).
T53C-08
Variations of the Velocity Contrast and Rupture Properties of M6 Earthquakes Along the Parkfield Section of the San Andreas Fault
We perform a comprehensive high-resolution imaging of bimaterial interfaces along the Parkfield section of the San Andreas Fault (SAF), based on analysis of fault zone head waves (FZHW) that refract along bimaterial fault interfaces. The employed seismic data are generated by 8993 earthquakes since 1984, and recorded by the NCSN and HRSN permanent seismic networks along with two temporary PASSCAL deployments: the 2001-2002 PASO and the 2004 Parkfield Guided Waves experiment. We stack waveforms of events in repeating earthquake clusters to increase the signal-to-noise ratio and confidence level of FZHW identification. Next we align the peak or trough of the direct P waves assuming right-lateral strike-slip focal mechanisms, pick the FZHW arrivals, and plot the waveforms against the along-fault-interface distances. Clear FZHW are observed for many stations on the NE (slow) side of the fault in the creeping section of the SAF north of Middle Mountain (MM), indicating a presence of a sharp bimaterial interface in that region with variable values of velocity contrasts. The obtained velocity contrasts are 5-10% north of MM, and systematically decrease towards Gold Hill (GH). No clear head waves are observed at stations on the NE side of the SAF for ray paths sampling the fault around GH, indicating an absence or reversal of the velocity contrast in this region. The obtained along-strike variations of velocity contrasts are consistent with geological observations of a sliver of high-velocity rock immediately to the NE of the SAF associated with the GH fault and 3D seismic tomography results. The existence of a local reversal of velocity contrast near GH offers a simple explanation for the opposite propagation directions of the M6 1966 and 2004 Parkfield earthquakes, and could also partially explain the apparent segmentation of the M6 events. The 1966 earthquake nucleated near MM and propagated to the SE, as expected for rupture on the bimaterial fault interface in that region. The local reversal of the velocity contrast near GH may have prevented the rupture from propagating further to the SE. On the other hand, the 2004 earthquake nucleated near GH and propagated to the NE, again as expected for rupture on a bimaterial interface. The rupture stopped at MM where the preferred rupture direction is to the SE.