S21B-1807
Fault Coupling Between Seismicity Streaks on Strike-Slip Faults: Using the Separation Distance Between Streaks as an Indicator of Coupling
Many strike-slip faults that are partially coupled exhibit lineations of microseismicity that are oriented nearly parallel to the slip direction of fault movement, referred to as seismicity streaks. The observed streaks are up to a few hundred meters wide in the direction perpendicular to fault movement and anywhere from one to ten kilometers long in the direction parallel to fault movement. Parallel streaks are typically separated by a quiescent zone that is devoid of microseismicity. It has been proposed that streaks represent a linear trend of strong asperities embedded on an otherwise creeping fault, or they delineate the discrete boundary between locked and creeping zones. Many of these zones are too narrow (<2 km) to be resolved from geodetic inversions. Thus direct imaging of their kinematic state is often not possible. We assess the coupling of these quiescent zones by considering the stress accumulation on a narrow fault patch. If a quiescent zone is coupled, then this patch would eventually fail in a time frame dependent on the loading rate and the dimensions of the patch. We focus our analysis on a quiescent zone on the San Jaun Bautista segment of the San Andreas fault. This patch has a down-dip width of 2 km, based on the double-differenced relocated seismicity. Assuming a 20 MPa failure threshold, we calculate that the fault patch should fail within 70 years if it is fully coupled. Historical seismicity records extending back to 1932 suggest that this patch has not failed in a major earthquake. Therefore, for this particular patch on the San Juan Bautista segment of the SAF, seismicity streaks are separated by an uncoupled patch that is creeping. The opposite scenario is exemplified on the Parkfield segment of the San Andreas fault where a large quiescent zone, separated by a pair of seismicity streaks, ruptured in the 2004 Parkfield earthquake. Thus we find evidence for both proposed models of coupling between seismicity streaks. Our calculations demonstrate that the down-dip width of the quiescent zone may be an important diagnostic as to whether it is coupled or uncoupled.
S21B-1808
Did a Stress Change due to a Long-Term Slow Slip Event in the Tokai Region Cause Distant Seismic Quiescence in the Tamba Region, Japan?
Seismic quiescence is useful information for the earthquake prediction. Relationships between seismicity rate change and stressing rate change have been reported by theoretical and observational studies (Dieterich, 1994; Toda et al., 2002). Recently, Ogata (2007) showed that a silent slip event might occur within the source region of an intraplate earthquake preceding the rupture from seismicity rate changes and GPS anomalies. The Tamba region in southwest Japan is located to the northeast of the rupture zone of the 1995 Hyogo-ken Nanbu Earthquake (Mjma 7.3). In the region, the seismicity was activated by a coseismic static stress change (+20kPa; Hiramatsu et al., 2000) due to the event. A distinct decrease in seismicity rate of microearthquakes was recognized in 2003 (Katao, 2005). Such a seismic quiescence had continued for two and a half years before the event (DPRI, 1999). It has, therefore, been controversial whether a major earthquake follows the quiescence or not (e.g., Umeda et al., 2005). We showed that the Tamba region was located in a region where Δ CFS decreased (-0.5kPa/yr) due to the long-term slow slip event (SSE) in the Tokai region and indicated that the beginning of the quiescence seemed to be associated with that of the event (Sugaya et al., 2007IUGG). Our purpose in this study is to investigate whether the quiescence in the Tamba region is caused by the stress change due to the long-term SSE or not based on the rate- and state- friction law (Dieterich, 1994). We use the hypocentral catalog of the DPRI from 1987 to 2001 and that relocated in this study from 2002 to 2006. We use declustered earthquakes (Reasenberg, 1985) greater than or equal to M 2.5 for following analyses. We find that the seismicity in the Tamba region after the 1995 Hyogo-ken Nanbu earthquake is explained by the Omorifs law (p=1) than the ETAS model (Ogata, 1986). The seismicity is, thus, interpreted as the aftershock-type activity of the earthquake. We estimate Aσ (A is the fault constitute parameter and σ is the effective normal stress) (Dieterich, 1994) using the seismicity rates in the Tamba region during a half year before and after the Hyogo-ken Nanbu earthquake and the stress change (+20kPa; Hiramatsu et al., 2000) by the earthquake. The obtained value of Aσ is 9.7 kPa and is smaller than 35kPa reported by Toda et al. (1998). We estimate the stressing rate as 0.5kPa/yr using the observed cumulative curve of earthquakes before the long-term SSE. This value is close to the stressing rate (0.7 kPa/yr) estimated by Toda et al. (1998). The observed curve, however, deviates downward from the theoretical curve calculated by above parameters from the mid 2000. This deviation means that the stressing rate changes in the mid 2000. The long-term SSE provides a decrease in the stressing rate of -0.5kPa/yr in the Tamba region. However, to fit the theoretical curve to the observed one, a decrease in the stressing rate of -2.5 kPa/yr is needed. This indicates that the long-term SSE contributes to the seismic quiescence partly but it is not a main cause of the quiescence.
S21B-1809
Dynamic Triggering Stress Modeling
It has been well established that static (permanent) stress changes can trigger nearby earthquakes, within a few fault lengths from the causative event, whereas triggering by dynamic (transient) stresses carried by seismic waves both nearby and at remote distances has not been as well documented nor understood. An analysis of the change in the local stress caused by the passing of surfaces waves is important for the understanding of this phenomenon. In this study, we modeled the change in the stress that the passing of Rayleigh and Loves waves causes on a fault plane of arbitrary orientation, and applied a Coulomb failure criteria to calculate the potential of these stress changes to trigger reverse, normal or strike-slip failure. We preliminarily test these model results with data from dynamically triggering earthquakes in the Australian Bowen Basin. In the Bowen region, the modeling predicts a maximum triggering potential for Rayleigh waves arriving perpendicularly to the strike of the reverse faults present in the region. The modeled potentials agree with our observations, and give us an understanding of the dynamic stress orientation needed to trigger different type of earthquakes.
S21B-1810
Tidal Triggering of Earthquakes in the Northeast Pacific Ocean
There have been many searches for evidence of tidal triggering in earthquake catalogs. With the exception of volcanically active regions, the more rigorous studies in continental settings tend to find no correlation or at best only a very weak correlation. In the oceans, the effect of loading by the ocean tides can increase tidal stresses by about an order of magnitude over continental settings. In recent years, several studies have reported evidence of tidal triggering in oceanic regions and such observations can represent a useful constraint on models of fault failure. In this study, I systematically search for a link between ocean tide height and the incidence of earthquakes in the Northeast Pacific Ocean, a region of high-amplitude open ocean tides. The focal mechanisms of most of the earthquakes in these catalogs are unknown but it can be shown that tidal stresses will in most instances promote failure at low tides. I investigate three declustered data sets comprising (1) earthquakes from 1980-2007 on the Juan de Fuca plate and in the Queen Charlotte Fault region from land based catalogs; (2) earthquakes from 1992-2001 on the Juan de Fuca plate located with the US Navy's SOSUS hydrophone array and (3) earthquakes from 1980-2001 south of Alaska and the Aleutians located with land based networks. Each data set comprises between 5000 and 6000 earthquakes. I look at the distributions of earthquakes with ocean tide phase, height, and tidal range and apply Schuster and binomial tests and Monte Carlo simulations to determine if they deviate significantly from random. The results show no evidence of triggering during intervals of increased tidal range but all three data sets show a significant increase in earthquake incidence at low tides. The signal is particularly strong in the land-based catalog for the Juan de Fuca Plate and Queen Charlotte Fault regions where there is a 15% increase in the rate of seismicity within 15° of the lowest tides. The signal is weakest in the SOSUS data set which may reflect the lower average tidal range at epicenters in this data set or an analysis that is influenced by gaps in this catalog. The triggering signal in the Alaska/Aleutian data set may be partially obscured by earthquakes in the Aleutians where the total tidal stresses can be significantly out of phase with the ocean tide height. The increases in the rates of seismicity I observe at low tides are reasonably compatible with predictions based on an earlier analysis of global thrust earthquakes and on laboratory simulations of fault failure under tidal loading.
S21B-1811
Triggering Effect of the Static Stress Transfer in Mining-Induced Seismicity from Rudna Mine in the Legnica-Gogow Copper District, in Poland
The static stress transfer is often considered as a possible source of earthquake interaction. Some examples from natural seismicity show that even small stress perturbations resulting from the coseismic slip can enhance or prevent future occurrences. In case of the mining-induced seismicity the coseismic stress changes expressed in terms of the Coulomb failure function (CFF) are at least one order smaller than those for earthquakes. Furthermore, they are only a small component of the stress field variations in mines. In order to recognize whether the static stress transfer can influence also the generation process of the mining induced seismicity we analyze seismic data from Rudna Mine in Poland. We consider events of ML>=2.0 occurred in Rudna Mine from 1993-2006. We examine the possible triggering checking correlation between event locations and the stress-increased zones, expressed by the proportion of events consistent with increased CFF areas. We find that more than 50 per-cent of the analyzed events occurred in areas where stress was enhanced due to the occurrence of previous events. In order to recognize the significance of the effect we test the null hypothesis stating that there is no influence of CFF changes due to previous events on the subsequent event. To attain the significance of this null hypothesis we estimate the distribution of the proportion of events located inside positive CFF areas from 2000 results for random permutations of the original series of events. For CFF changes >= 0.02 bar the null hypothesis was rejected at the 95% confidence level. This result indicates that the static CFF triggering in Rudna Mine exists and this effect is statistically significant. This work was prepared within the framework of the research project No. PBS- Grecja/10/2007, financed by the Ministry of Education and Science of Poland during the period 2007 to 2009.
S21B-1812
Rate State Seismicity Equations Applied to 3D Spatially Heterogeneous Stress to Study Aftershock Sequences: Implications for Crustal Stress
3D models of stress heterogeneity [Smith and Heaton, submitted 2008] are combined with a formulation for seismicity based on rate-state friction [Dieterich, 1994] to examine aspects of aftershock seismicity: 1) Spatial patterns as a function of time, 2) seismicity rates, and 3) changes of focal mechanism statistics in aftershock sequences. Some key observed features for the simulated aftershock period are: 1) Approximately 1/t, Omori Law behavior, 2) initial clustering of aftershocks in regions of greatest Coulomb stress then transitioning to a more even spatial distribution with time, and 3) some seismicity in stress shadow areas due to the pre-existing heterogeneous stress. With regard to focal mechanism statistics, we find that stress inversions of aftershock focal mechanisms with the spatially heterogeneous stress yield large "apparent" rotations of the inferred maximum horizontal compressive stress, SH, that are far in excess of the "true" SH rotations. Also, focal mechanism inversions for simulations with spatially variable stress perturbations due to slip on finite geometrically complex faults can produce transient increases in the stress inversion mean misfit angle, β, (generally thought to correlate with stress heterogeneity). Both the apparent changes in stress orientation and increase in misfit angles decay during the aftershock period. These results generally agree with previous conclusions [Smith 2006; Smith and Heaton, submitted 2008] that failures in a heterogeneous stress field are biased toward the stressing rate or stress perturbation. Biasing arises because failure surfaces that are well oriented with respect to stress perturbations experience larger transient seismicity rate changes than surfaces that are misaligned. Perhaps the most important conclusion is that one can generate this "apparent" rotation of SH, much larger than the "true" rotation, for a moderately strong crust (mean stress 50 MPa) if stress is heterogeneous. Consequently, estimates of crustal strength from aftershock studies may significantly underestimate the mean crustal stress state for the region; i.e., one cannot directly use rotations of, SH, from stress inversions, to conclusively demonstrate low crustal strength (< 10 MPa). Understanding to what degree biasing effects create rotations of the inferred stress orientations and increase the stress inversion parameter, beta, is important if one wishes to better parameterize stress in the Earth's crust. Indeed, we explore the potential extent of the biasing and its implications for crustal stress, by comparing our models to a few aftershock sequences.
S21B-1813
Fault Interaction and -Propagation in the South Iceland Seismic Zone in the Current Major Earthquake Sequence, On-Going since 1998.
Plate spreading of approximately 1.9 cm/yr and the eastward shift of the rift zones across Iceland builds up stress in the two transform zones connecting the displaced rift zones. Every century through historical time the 80-km-long E-W trending southern shear zone, or South Iceland Seismic Zone (SISZ), has experienced episodes of major earthquakes; each sequence often lasting a few years. The current one started in 1998 and is still on-going. The previous sequence lasted from 1896 to 1912. Even though the underlying motion is E-W oriented left-lateral slip, the earthquakes occur on N-S oriented right-lateral strike-slip faults, distributed along the shear zone. The present episode started with two M>5 earthquakes five months apart in 1998 at the western margin of the SISZ. In 2000, two ~M6.5 earthquakes occurred 3 days and 17 km apart in the center of the SISZ. Four additional events of M>5 followed the first event, of which two were dynamically triggered by the event's shear waves. On May 29 2008 the most recent earthquake of the sequence occurred in the western SISZ. It was generated by slip on two parallel N-S oriented vertical strike-slip faults, 4 km apart, with a combined magnitude of M6.3. The earthquake started on the eastern, 10-km-long Ingolfsfjall fault and waves from this event triggered slip on the western, 19-km-long Kross fault, presumably at its center. Slip on an E-W oriented fault, west of the Kross fault and active in 1998, may have contributed to the event, but so far aftershocks on this fault have not been found in the data until 53 minutes after the main event, suggesting it was not. The event was felt over most of SW Iceland and strongly felt in the near-by towns Hveragerdi and Selfoss. No human fatalities occurred, but over 30 houses were terminally damaged and live-stock killed. The closest seismic stations were saturated, but 3 cm/s peak ground velocity was recorded at 46 km distance. The event was well recorded on the continuous GPS network, with maximum horizontal displacement of 19 cm measured 2.5 km SE off the Ingolfsfjall fault. Details of the fault structures are mapped through relative location of aftershocks, showing two parallel 8-km-wide, vertical faults. Seismic data from the first few seconds are analyzed to search for the onset of slip on the Kross fault. An overview of the evolution of faulting and fault patterns of the previous earthquakes in the sequence is also presented.
S21B-1814
The Effects of Acoustic Waves on Stick Slip Behavior in Sheared Granular Media and Their Implications for Earthquake Recurrence and Dynamic Triggering
To better understand the physics of dynamic triggering and the influence of dynamic stressing on earthquake recurrence, we are conducting laboratory studies of stick slip in granular media with and without applied acoustic vibrations. In our 3-D experiments, glass beads are used to simulate granular fault zone wear material, sheared in a double-direct configuration under constant normal stress, while subject to transient or continuous perturbations by acoustic waves. We observe both instantaneous and delayed triggering when vibration is applied. Vibrations also cause significant disruption in the recurrence rate. The effects of vibration are observed for many major-event cycles after vibrations cease, indicating a strain memory in the granular material. Vibration-induced disruption of periodic stick slip is linked to failure of granular force chains. In 2-D experiments we are applying photoelastic discs in stick slip measurements in order to visualize the evolution of the force chain network. Photoelastic measurements provide insight into failure, and in particular small adjustments in the force chains network that presage failure. Our results should lead to a new understanding of the importance of seismic energy on earthquake physics and more generally, we anticipate that it will have broad impact on unexpected material failure induced by moderate-amplitude elastic waves, including avalanches, landslide and failure of incipient damage in solids.
S21B-1815
Relating Seismicity Rate Changes to Stress Shadows in Order to Quantify Local Stress Field Changes Caused by Earthquake Events
One of the current methods used to examine the seismicity rate and what information it can provide is the Pattern Informatics (PI) method (Tiampo et al., 2002). It is a phase dynamical analysis of the historic seismicity that provides insight into time-dependent earthquake probabilities. One particular area where it can be used is in the examination of the local stress field. In Tiampo et al. (2006), it was shown that the PI method could detect areas of seismic quiescence around where a large earthquake event had occurred. These areas can be related to 'stress shadows' regions where there has been a decrease in the local stress field due to a large earthquake. Using the results from the PI method as a reference model, the local stress field can be optimized by manipulating the stress field orientation in a stress model, as the field orientation is a key parameter used to calculate stress changes (King et al., 1994). For this study, the program Coulomb 3 (Lin and Stein, 2004; Toda et al., 2005) has been employed to calculate the coulomb stress changes from slip models of several Californian earthquakes. The stress model results are then quantitatively and qualitatively compared to those from the PI method in an effort to find the field orientation that gives the best correlation. Although the field orientation is the most important parameter, calculation depth and coefficient of friction also can influence the degree of correlation, as well introducing heterogeneities into the stress field. Further improvements on the PI method also allow for more ease of comparison, and more accurate determination of the size, location, and duration of the stress shadows. By understanding the changes in the local stress field associated with large magnitude earthquakes, we can further the ideas and concepts involved with earthquake and fault mechanics and improve earthquake modeling.
S21B-1816
Mapping deformation field prior to an Earthquake:InSAR Observations 1993-2004 before the Mw 6.0 Parkfield Event.
We propose to use Synthetic Aperture Radar Interferometry (InSAR) for detecting potentially seismogenic locked zones or structural barriers on an active major fault. We run a test on the well intrumented Parkfield segment of the San Andreas Fault, California, before the 2004 Parkfield earthquake (Mw 6.3). We obtain the interseismic deformation pattern before the 2004 Parkfield earthquake by stacking multiple selected InSAR interferograms. The interseismic interferogram clearly indicates strain building up over a narrow 10-15 km length along the San Andreas Fault Zone (SAFZ) adjacent to the coming epicentral area, Gold Hill. We then construct the coseismic InSAR interferogram. The coseismic interferogram shows how the 2004 earthquake rupture extends from this narrow stressed region to the North, consistent with the seismological and geodetic analyses. These two interferograms reveal that the forthcoming epicenter is located in the area that most deformed before the earthquake. This information cannot be easily detected by the Global Positioning System (GPS) network and thus provides the direct evidence of the preseismic deformation process inferred from the other geodetic inversions. We show the InSAR signal when a locked segment on a major fault acts as stress concentrator and thus it is potentially seismogenic. This approach could be applied in remote areas where major faults are not instrumented and seismic hazard is often underestimated.
S21B-1817
New Constraints on Fault-Zone Structure from Seismic Guided Waves
We propose two new methods that analyze dispersion from fault-zone guided waves (GWs) to constrain the fault structure at seismogenic depths. A two-station differential group velocity technique, such as that commonly used for surface waves, was adapted to solve for the local fault-zone structure between two earthquakes. This method was tested with synthetic data and shallow events recorded in the SAFOD borehole in the San Andreas Fault. A pair of deep earthquakes recorded in the SAFOD borehole indicate a about 150 meter wide San Andreas Fault waveguide with >20% velocity contrast at 10-12 km depth. With additional earthquakes, spatially varying fault-zone structure at seismogenic depth could be imaged. Our second method uses the fact that dispersive fault-zone guided waves are refracted, similar to body waves, due to the increase in velocity with depth. Within fault zones, an increase in seismic velocity with depth causes deeper parts of the fault to efficiently trap GWs at higher frequencies than GWs trapped near the surface. Higher-frequency signals are thus required to constrain deep fault structure. Finite-difference synthetic seismograms were generated for an explosive source and a receiver array along and within a low- velocity fault zone, analogous to a body-wave refraction survey. Subsurface fault structure was successfully derived from the surface guided-wave refraction survey. Using multiple sources and receivers, this method can be used to derive 3-D fault-zone structure to seismogenic depths, mapping variations both along strike and in depth.
S21B-1818
Tectonics of Coastal Central California: Deep Geophysics and Structural Interpretation of Active Faults
We have compiled a number of geological cross sections for west-central California in order to evaluate the regional tectonics. The study area extends from Monterey Bay, CA, southward to the western Transverse Ranges, and from the base of the continental slope eastward to the San Andreas Fault (SAF). The cross sections were compiled from the published and unpublished geological and geophysical literature. These sections allow us to compare mapped faults and their extension at depth, fault seismicity, fault dip and focal mechanisms, depths of high- and low-velocity reflective zones, and interpretations of crustal composition and structure at depth. Relocated seismicity and focal mechanisms were plotted on the cross sections in order to examine evidence for the geometry of faulting. The cross sections were aligned along the San Andreas Fault and have a common scale. At depth, major geological features were revealed by near-vertical seismic reflection profiles. Two cross sections through the San Andreas Fault Observatory at Depth (SAFOD), near Parkfield, CA, reveal reflective zones at 8- and 12-km depth that most likely correspond to the top of Salinian granitic rocks. Further to the south, seismic reflection profile SJ-6 provides clear evidence for blind reverse faulting along the northern La Panza fault and also reverse faulting along the San Juan/Red Hills fault, an apparent flower structure related to the SAF. Microseismicity and focal mechanism were plotted over the seismic profiles. Many earthquake hypocenters align along the San Andreas and Nacimiento faults. Most focal mechanisms are consistent with right-lateral strike-slip faulting, although clear thrust mechanisms are also found. These thrust mechanisms located west of the SAF provide evidence for thin-skinned tectonics in this region.
S21B-1819
The method to estimate a slip distribution of a large earthquake based on the spatial distribution of its aftershocks and rate- and state friction law: Further development
We had developed a method to estimate a spatial slip distribution of a large earthquake based on its on-fault aftershock activity and rate- and state-dependent constitutive friction law (Dieterich, 1994). The outline of this method is as follows: we divide the source area of the large earthquake into many subfaults and optimize slips of each fault for which the expected spatial distribution of aftershocks fits well to the observed distribution using the Dieterich's friction law and point-process modeling. Since we have a huge number of parameters in this optimization, to make the optimization stable, we construct a Bayesian model incorporating a constraint as a smoothness prior of spatial slip distribution. The weight of the smoothness constraint is determined by the maximization of the marginal likelihood or the minimization of the Akaike's Bayesian Information Criterion (ABIC). In the estimation of the marginal likelihood or ABIC, the Laplace approximation (e.g., Tierney and Kadane, 1986) is typically introduced to reduce the computation time, but this approximation cannot be applied to our method. Therefore, instead of the approximation, a computational technique called Markov chain Monte Carlo is used, but this algorithm requires much computation time compared with the Laplace approximation if the number of subfaults is huge. This means that practically we cannot analyze a fault model in which the subfault size is small. To overcome this problem, in this study we have modified our method. Namely, we also assume a prior distribution with respect to the weight of the constraint, and estimate the weight and slip distribution simultaneously. As a result of this modification, we succeed in reducing the computation time and can analyze a model in which the subfault size is smaller than before this modification.
S21B-1820
Distribution of S-wave reflectors in the crust and uppermost mantle beneath the source regions of large inland earthquakes in southwest and central Japan
It is important to estimate detailed heterogeneous structures beneath active fault zones, which will give us clues to understand the generating processes of large inland earthquakes. We estimated detailed seismic reflection properties in the crust and uppermost mantle in the Chugoku region, southwest Japan, and also in the Hokuriku region, central Japan, where several large (M>6) crustal earthquakes occurred within a few hundreds of years. In the Chugoku Region, we first analyzed the source region of the 2000 Western Tottori Earthquake with Mj 6.6, using dense aftershock observation data. Then, we extended the analysis to the whole Chugoku region, using routine network data. From the analyses of these two steps, we delineated the Conrad and the Moho discontinuities and also the top surface of the Philippine Sea Plate beneath the Chugoku region, at depths of 15-25 km, 30-45 km, and 50-70 km, respectively. We found that the strengths of seismic reflections have remarkable contrast between the two sides of the fault plane of the 2000 Western Tottori Earthquake. This suggests that the mainshock fault plane is located at the boundary of medium properties and that the fault plane has near-vertical extension down to the lower crust. In the Hokuriku Region, we conducted the same reflection analysis using waveforms from local earthquakes, including the source region of the 1948 Fukui Earthquake (Mj 7.1). We found two reflection zones at depths of 20-25 km and 40-50 km. Comparing with past studies of receiver function analyses, theses two reflection zones are thought to correspond to the Conrad and the Moho discontinuities, respectively. Reflection strength in the upper crust is higher on the ENE side of the fault plane of the 1948 Fukui Earthquake, while lower on the WSW side. In the source region of the 2000 Western Tottori Earthquake, crustal thickness changed along the fault strike just beneath the source. We will further look into such a relation between heterogeneous structures and earthquake generation in the Hokuriku region.
S21B-1821
The Influence of Fault Kinematic Variability in Ground-Motion Estimates: Comparison of Different Simulation Approaches
Ground motion simulations based on a kinematic representation of the rupture over a finite fault are an important tool in engineering seismology. Even though they lack in the physical aspects of the rupture process with respect to dynamic modeling, they are preferable for several engineering purposes due to the broader frequency content of seismic radiation and to their ease of application. In general simulations should be also preferred to empirical Ground Motion Prediction Equations (GMPEs) to assess the level of shaking generated by an earthquake at some sites as they are able to reproduce specific phenomena (particularly in the near-fault region) not accounted for by GMPEs. Anyway, the use of GMPEs is extremely easy requiring, in most cases, only the earthquake magnitude, the source-to-site distance and an approximate site condition coefficient to provide an average ground-motion estimate. On the contrary, the ground-motion simulation requires setting of a larger number of input parameters, whose values are generally characterized by large uncertainties. Different numerical approaches might require the definition of different input parameters and each simulation technique might be differently sensitive to each parameter. We investigate the ground motion variability related to uncertainties in the large-scale kinematic source parameters for three numerical approaches: two stochastic methods (EXSIM, Motazedian and Atkinson, 2005 and DSM, Pacor et al., 2005) and a hybrid deterministic-stochastic method (HIC, Gallovic and Brokesová, 2007). The capability of each method to reproduce earthquake ground motion has already been tested in several works. All the techniques are applied to simulate the largest instrumental earthquake occurred in Italy (1980, M6.9 Irpinia earthquake, southern Italy) at a grid of bedrock virtual receivers. A set of possible rupture scenarios is built up, for each simulation technique, varying the considered kinematic parameters (hypocenter position, value of rupture velocity and final slip distribution on the fault) within plausible ranges. The ground-motion variability obtained from each simulation method is discussed in terms of peak values and response spectra as function of the azimuth and distance from the source. We show that each technique generates ground-motion distributions characterized by different variabilities, related to the different modeling of the fault rupture. Synthetics have been compared with GMPEs from NGA project and with strong-motion data recorded during the Irpinia earthquake in order to discuss the validity of theoretical and empirical predictions. Finally, we computed residuals between median values from GMPEs and synthetic ground- motion values and analyzed their intra-event and inter-event distributions.
S21B-1822
Full-3D Waveform Tomography for the Seismic Velocity and Attenuation Structure on the San Andreas Fault Zone in Parkfield Area
We are conducting time-lapse full-3D waveform tomography for the seismic velocity and attenuation structure on the San Andreas Fault zone in the Parkfield area. The seismic waveform data we are using come from several Parkfield seismic networks, including the borehole seismographs installed in the SAFOD mainhole and pilot hole, the HRSN stations, an array of 45 portable seismographs deployed across and along the surface trace of the San Andreas Fault near SAFOD site in fall 2003 as part of a site characterization program and an array of 45 portable seismographs deployed in 2002 and 2004 near the town of Parkfield. The tomography technique we are employing for this study is the full-3D waveform tomography (F3DT) method based on the scattering-integral (SI) formulation. In our F3DT algorithm, both the reference structural model and the derived model perturbations are all 3D in space and the full-wave sensitivity kernels are calculated from the full physics of 3D wave propagation, which is implemented using the finite-difference method. In this study, we are conducting time-lapse, joint inversions for both seismic velocity and attenuation structures of the San Andreas Fault Zone using waveform recordings from repeated explosions and earthquakes. Intrinsic attenuation estimates are usually derived from frequency-dependent amplitude measurements, which are also strongly dependent on elastic effects such as focusing/defocusing and 3D scattering. By jointly inverting for seismic velocity along with Q in a fully 3D setting, we can remove much of the uncertainty associated with focusing/defocusing and 3D scattering. About 60 percent of our recorded earthquakes are members of nearly 200 event clusters of repeating, virtually identical earthquakes exhibiting very small source dimensions. These widely distributed repeating micro-earthquakes provide us the opportunity to detect transient or systematic temporal changes of physical properties within the fault-zone through time-lapse tomography.
S21B-1823
Relationship between earthquake source faults and 3D density structures derived by gravity anomaly inversion in Japan
In Japan, prediction of strong ground motion has important role for prevention of earthquake hazards. Information of subsurface and crust structure including source fault are require for the strong ground motion estimation. Various geological and geophysical information are required to define earthquake fault and underground models. Active fault trace is the major information to construct earthquake source faults. On the contrary, recent earthquakes occurred without obvious earthquake faults in Japan. Thus, the other information is required to model earthquake source faults. Seismic surveys reveals detailed fault configuration and subsurface structure. Three dimensional velocity distribution were estimated various seismic tomographic studies. Relationship between earthquake source fault and velocity distribution has been discussed by many papers. The detailed gravity anomaly distribution were derived from several gravity data base published in Japan. We have investigated the validity of gravity data for earthquake source faults. As a result, several earthquake source faults occurred in Japan were surrounded by steep gradient zones of local gravity anomaly (short wavelength component) derived frequency analysis of the gravity anomaly. The local (short wavelength component) or regional (long wavelength component) gravity anomaly correspond to shallow and deep density structure, respectively. In this study, the three dimensional density structure was estimated around the earthquake source fault applying linear inversion to cell model (dividing underground structure to small density bodies). We will discuss the relationship between the three dimensional density structure and the source fault in the poster.
S21B-1824
Relationship between Bouguer anomaly and active fault ( source fault) - For the purpose of estimate of source fault -
Concepts of setting source fault parameters from active faults are discussed for predictions of strong ground motion. Estimation of ground motion plays important rule for the prevention of earthquake hazards. From recent developments in waveform inversion analysis of source fault rupture processes through large earthquakes, it is found that strong ground motion is strongly affected by fault geometry and slip heterogeneity. At the prediction of strong ground motions for scenario earthquakes by active faults, the initial parameters of source faults, such as fault length, direction and dip are thus necessary to be determined. However, the study about relation between source fault and surface rupture, source fault length are certainly longer than surface ruptures (Kitada et al, 2004) and it is difficult to estimate exactly source fault length from the information around surface fault. In case of estimate the source fault, it might be better using spatial underground structure and its information such as gravity. Inoue et al (2007) discussed the relation between source fault and gravity data and suggest the short wave length component of bouguer anomaly have the possibility to show the distribution of density around seismogenic zone. In this study, we consider the length of source fault for scenario earthquake by distribution of the short wave length component of bouguer anomaly. The short wave length component of bouguer anomaly shows the relative high density area and low density area. These areas indicate the same tectonic block. Source fault are distributed not only on the boundary of these block but also in the same tectonic block. Therefore, the source fault seems to be difficult to continue over the other block.
S21B-1825
Simulation of Supershear Ruptures on Interfaces With Strong Velocity-Weakening Friction
We present finite-difference simulations of mode II ruptures transitioning to supershear speeds on rate-and- state faults. Frictional resistance is controlled by a slip evolution law with strong velocity-weakening behavior at high slip rates (Rice, 2006), representing thermal weakening due to flash heating of microscopic asperity contacts. In this study, initial fault configuration is mainly characterized in terms of two parameters: the initial background stress (τBACK), and a generalization of Andrew's (1985) seismic S ratio between strength excess and stress drop, to rate-and-state friction laws (SRS). The latter is a function of the initial state of the interface, which we parameterize in terms of the initial slip velocity, V0. Rupture modes evolve from cracks to pulses via reduction of τBACK, as predicted by the understressing theory of Zheng and Rice (1998), and supershear propagation is observed after SRS drops below a certain threshold that depends on τBACK. SRS decreases as the initial slip rate V0 increases, since this decreases the peak strength of the fault.
S21B-1826
Comprehensive Study on the Effects of Parameters on Slip Motion of the Fault
In recent years, the study on the numerical simulation of the fault slip motion has become the main method in research of the earthquake prediction and prevention. However in the past, most researchers paid attentions to the only one aspect which affects the slip motion of the fault, such as the normal stress of fault surface, depth or width dependence of friction influence the crustal faulting etc. There is now no study on slip motion of the fault affected by these parameters together. This paper, basing on the rate- and state-dependent frictional law and improved one-dimension mass-spring model, we will research these parameters including normal or tangential tectonic stress and depth or width dependence of frictional influence the fault slip comprehensively. Through establishing the mechanism of slip motion of the fault and investigating the observation evolution of the precursory slip, post-seismic slip, co-seismic slip, creep events and post-seismic fault position, we will seek the relations between these parameters. By numerical calculating, we will find that the influence of these parameters on the fault is much different from that of only one parameter. Finally, applying the historical earthquake data of strike-slip fault zone in Tibetan plateau and surrounding areas, we will prove theory and model and give the numerical simulation result of future slip history of the zone.
S21B-1827
A Case Study of Stress Proxies on the Hayward Fault
Currently, there are a number of seismicity analysis techniques that have been used to infer the state of stress within the lithosphere. While many of these "stress proxies" are based on theoretically plausible foundations and could potentially be incorporated in earthquake forecasting, rigorous comparisons between stress inversion results and other data that also reflect stress conditions at depth (e.g., fault creep rates) have not been performed. The Hayward fault presents an ideal locale in which to test the accuracy of these stress proxies in that we can qualitatively infer the stress along the fault by assuming that fault creep rates are inversely related to stress and thus that stress is high along asperities locked since the 1868 earthquake and low along sections that are accommodating a substantial amount of their slip budget through creep. In this study, we performed three stress analyses of Hayward fault seismicity by evaluating spatial variations of the b-value, accelerating moment release (AMR), and Load/Unload Response Ratio (LURR) methodologies. We then qualitatively compare the results of these analyses with best-fit Hayward fault slip inversions derived using GPS, InSAR, and seismicity patterns to determine if stress levels inferred from the inversions and independent data agree.
S21B-1828
Asymmetric Velocity across the San Andreas Fault System: The Effects of Fault Dip
Most models of the interseismic velocity across the San Andreas Fault (SAF) system use block motion at depth and backslip on vertical strike-slip faults for the locked zone. Studies have found an asymmetry in the velocity across straight segments of the SAF and have explained them by an across-fault contrast in shear modulus [Fay and Humphreys, 2005; Schmalzle et al., 2006]. However, as some research showed, heterogeneity alone can't explain the geodetic very well because the contrast ratio of rock rigidity is too high to be true [Fialko, 2006]. In addition, recent relocated seismic data shows the microseimic data location is several kilometers off from the surface fault trace, which suggests a 60-degree dipping of the Southern SAF (SSAF) near Salton Sea and an opposite dipping of the Southern SAF north of the San Gabriel Mountains, including Carrizo [Lin et al., 2007]. Here we investigate the effects of fault dip to explain the asymmetry in velocity and gravity anomaly across the SSAF near Carrizo, and SSAF and San Jacinto Faults (SJF) near the Salton Sea. 1D modeling of the GPS data of the SSAF near the Salton Sea show a 50 percent improvement of the misfit by introducing a 60-degree dipping SAF fault, while SJF remains vertical. The best fitting model suggests a SAF slip rate of 25 mm/yr and a SJF slip rate of 21 mm/yr slip rate with associated locking depths of 17 km and 12 km, respectively. 1D modeling of the gravity data in the same area shows a 30 percent improvement when dipping is included. The modeling of the GPS data of SAF near Carrizo shows a 50 percent improvement when the dip is included. The best fitting model here requires a 37 mm/yr slip rate and a 17 km locking depth. The dipping angle is 80 degrees from horizontal to the west. These preliminary results suggest that fault dip has an important effect on the interseismic velocity field and should be included in future plate boundary models.
S21B-1829
Constant Stress Drop Scaling from Small to Great Earthquakes
Earthquakes span a tremendous range of scales, more than 5 orders of magnitude in length. Are earthquakes fundamentally the same across this huge range of scales, or are the great earthquakes somehow different from the small ones? We show that a robust scaling law seen in small earthquakes, with stress drops being independent of earthquake size, indeed holds for great earthquakes as well. The simplest hypothesis, that earthquake stress drops are constant from the smallest to the largest events, combined with a more thorough treatment of the geometrical effects of the finite seismogenic layer depth, gives a new magnitude area scaling which matches the data well, and better over the whole magnitude range than the currently used scaling laws which have non-constant stress drop scaling. This has significant implications for earthquake physics and for seismic hazard estimates. Finally, we propose a physical origin for this scaling based in rock mechanics. We note that the stress drops seen in earthquakes correspond with the tensile strength of rocks. We suggest it is the cohesive breakdown of chemically recemented faults which thus sets this stress drop scale.
S21B-1830
Kinematic inversion of physically plausible earthquake source models obtained from Dynamic Rupture Simulations
A common approach to investigate earthquake physics consists of producing kinematic source models from the inversion of seismic records jointly with geodetic data. The regularization of the inversion requires some assumptions to restrict the range of possible models. Here, we evaluate to what extent physically plausible models are reliably restituted in spite these restrictions. More precisely we study which characteristics of ruptures, such as rupture velocity, slip distribution, and rise time can be reliably determined from the inversion of near-field data. We use a standard inversion scheme which assumes a rupture front propagating away from the hypocenter with a simple cosine slip-time function, and searches for solutions with minimum roughness [Ji et al, 2002]. To provide inversions with physically plausible sources, we generate several earthquake scenarios using 3D spectral-element simulations of dynamic rupture (Kaneko et al., 2008). The assumed model contains a planar fault in an elastic half-space. The fault is governed by rate and state friction, with a velocity-weakening region surrounded by slip-inhibiting velocity-strengthening regions. The fault properties are varied to obtain scenarios with different slip distributions and local slip durations, leading to pulse and crack-like ruptures. For the inversion, strike, dip, average rake, velocity model and the hypocenter are given, and we search for slip evolution that best fits strong-motion and GPS data at simulated stations, without a priori knowledge of moment, smoothness, rupture velocity, or slip distribution. The comparison with the input model is done only after the best-fit model is chosen among various constraint inversions. Our preliminary results show that, overall, rupture velocity and slip distribution are well- determined. Since we assume a single cosine for the slip-time function, both crack-like and pulse-like ruptures appear as pulses in the inverted models, but crack-like ruptures have larger spatial extent at each moment. The difference between the two kinds of ruptures is thus still observable. However, the slip history at a specific point on the fault cannot be obtained accurately due to the assumed shape of the slip time function. This is probably the major drawback of these inversion procedures. Our current work is therefore directed towards implementing different slip-time functions to allow a wider range of possible behavior without adding complexity to the inversion.
S21B-1831
Distribution of Seismicity Across Strike-Slip Faults in California
The decay of seismicity with distance from strike-slip faults in California is well described by a power-law of the form R~(x2+d2)-γ/2, where x is distance from a fault, γ is the decay rate of seismicity, and d is a near-fault inner scale. We establish this relationship by analyzing high-resolution regional and relocated catalogs in fault-referenced coordinate systems for four classes of faults (large faults in northern CA, large faults in southern CA, small faults in southern CA, and aftershocks of large earthquakes). Results from a multi-catalog analysis of hypocentral variance are used to estimate parameter bias due to event mislocation. Scaling parameters, d and γ, vary regionally: seismicity is more localized on faults in northern California (d=0.04± 0.01 km; γ=1.54±0.15) than in southern California (d=0.21±0.04 km; γ=0.95±0.05). An investigation of individual fault segments shows that the localization of small earthquakes correlates with cumulative offset, on-fault earthquake density, and aseismic slip rate and we interpret this evolution of seismicity in the context of rate-and-state and damage-zone models of faults. Scaling parameters for aftershocks of large earthquakes (γ=1.45±0.10) and declustered catalogs of southern California large faults (γ=1.22±0.05) also show increased localization of seismicity, suggesting how epidemic-type earthquake-triggereing models might be modified to improve earthquake forecasts.
S21B-1832
Depth Localization of Seismicity on Strike-Slip Faults in California
We investigate the distribution of earthquake ruptures in three separate dimensions along California strike- slip faults. Previous work by Powers and Jordan (in prep.) shows that the average rate of small earthquakes along California strike-slip faults obeys a power-law of the form R~(x2+d2)-γ/2, where the rate R is in events/km2, x is the distance from a fault, γ is the decay rate of seismicity, and d is the near-fault inner scale. However, they do not consider the depth variability of earthquake hypocenters. We therefore perform a reconnaissance of their fault-referenced data set to determine if there is significant on-fault versus off-fault variability in earthquake depths. For each fault segment, we compute the depth variance in 4d km wide fault-normal bins, centered on the fault. For particularly long fault segments, we take the average variance over several shorter fault-parallel sub-segments. Results show interesting regional variations. In southern California, on-fault earthquake hypocenters are strongly localized in depth, but become more distributed with distance from a fault. In contrast, variance of hypocenter depths in northern California is similar both on and off of faults. Similar regional variations are observed for γ and d, so depth variance likely correlates with fault properties such as seismic productivity, creep rate, and cumulative offset. These results have important implications for fault-based models of seismicity, which can be used to improve current earthquake forecasting methods such as ETAS.