S41A-1813
Effects of Initial Conditions on Inversions of Geodetic Data Using Rate-State Friction Models of Afterslip
Current afterslip modeling efforts in the geodetic community are moving away from a purely kinematic approach, in which slip is estimated using standard inversion methods, towards dynamic models that incorporate stress boundary conditions and fault rheology. Previous studies have modeled geodetic postseismic time series data as a response of a fault to instantaneous stress and velocity changes due to an earthquake using a single-degree-of-freedom spring-slider model or a 2D fault in an elastic half-space with a steady-state velocity strengthening or rate- and state-dependent friction laws. These models assume that initial conditions on the fault immediately after the earthquake are determined from steady-state sliding before the earthquake and an instantaneous stress step corresponding to the earthquake. In this study, we first show that the spring-slider assumption and the assumption on initial conditions employed in previous studies may lead to misestimates of friction parameters through simulations of earthquake cycles on a fault obeying rate-state friction law. This indicates that it is important to estimate appropriate initial conditions from postseismic data. We thus develop an inversion method to estimate initial conditions on a fault from postseismic time series data and test the method with synthetic postseismic time series data. We also investigate effects of initial conditions and the model assumptions on estimates of friction parameters.
S41A-1814
Numerical simulations of afterslip following the 2003 Tokachi-oki earthquake
We conduct numerical simulations of afterslip following the 2003 Tokachi-oki earthquake and compare
results with a kinematic GPS inversion for spatial and temporal distribution of slip in order to investigate the
frictional properties of the subduction interface. Previous geodetic studies have clarified that afterslip
following this earthquake is distributed around the rupture region. On the contrary, a Tsunami inversion
result for the previous even, the 1952 Tokachi-oki earthquake, shows the eastern end of the 2003 afterslip
region ruptured at the time of the mainshockt (Hirata et al., 2003). These two observations raise the question
of whether afterslip occurred in a velocity strengthening or weakening region. Miyazaki et al. (2004) have
addressed this question by running a single spring-slider simulation. They compared the phase plots (i.e.
log(velocity) - shear stress change) with the 30-day GPS data and concluded the data favors the velocity
strengthening case. Fukuda et al. (2007) also obtained the same conclusion. However both studies used
oversimplified models to take the stress change into account.
In this paper we extend the GPS data by one year and use a two-dimensional fault embedded in a three
dimensional elastic media. Following Miyazaki et al. (2004) we consider two cases: (1) the earthquake
occurred in a velocity weakening region with k>kc and afterslip occurs in a velocity strengthening region,
(2) the frictional properties for the rupture region remains same and the rest of the fault is also velocity
weakening but with k<kc (conditionally stable). Although the phase plots for the two simulations differ
significantly, in both cases we see significant decaying afterslip following the earthquake, similar to the GPS
inversion result. It therefore may be difficult to distinguish between two models using GPS data.
S41A-1815
Afterslip as a part of earthquake rupture process
Afterslip is a postseismic slow slip, which lasts from several days to a few years adjacent to the coseismic rupture zone. Therefore, its slip mechanism has been considered to be different from that of earthquake in terms of stress release process. Consequently, in the simulation based on the rate- and state-dependent friction (RSF) law, coseismic rupture zone is modeled as velocity weakening, in contrast, afterslip as velocity strengthening. In 2003 Tokachi-oki earthquake, the afterslip between the mainshock and the largest aftershock was inferred between the two source region (Miyazaki and Larson, 2008). This suggests that the mainshock and the afterslip, and the aftershock occur continuously in space and time. Thus, both earthquake and the afterslip are in the process of the rupture propagation on the plate boundary and it is natural to interpret they have the same generation mechanisms. To verify the idea, we perform numerical simulation of earthquake generation cycle based on the RSF law using sloping plain fault as plate boundary. Parameter of the dependency of friction to slip velocity, A-B, varies with only depth. Characteristic slip distance, L, is large in the whole model area, and small in two circular ones (asperities) located in the central part. The heterogeneity of L assumes that of fracture energy, Gc, and low stress level. As a result, rapid slip (earthquake) occurs in asperity, and after that, aseismic long-term slip (afterslip) occur in the adjacent part of the area of large L. Earthquake and the afterslip reproduce as a chain of rupture process due to stress perturbation. The slip, however, does not so accelerate in the area of large L because of large Gc. The smaller the slip velocity change is, the smaller the surrounding stress change is. Thus the propagation speed from the coseismic area to the afterslip one gets smaller radically and the slip becomes not to be detected with seismic wave. In another earthquake cycle, rapid slip can occur at large Gc area when the stress level is high. Therefore investigation of the afterslip generation mechanism is important to physical understanding of earthquake cycle and considering earthquake forecast.
S41A-1816
Granular Friction: Constitutive Laws and Particle Dynamics in a Wide Range of Slip Rate
In a "microscopic" view, natural faults generally consist of gouge layers, the frictional properties of which are much richer than the celebrated rate-state friction law. One of such examples is intermediate-to-high slip velocity (mm/sec-m/sec) regime, where anomalous weakening and, at the same time, strengthening are reported; the results differ from experiments to experiments. In order to understand such a complicated phenomenon, one must carefully control the physical processes that potentially affect the frictional properties. In this paper, the standard numerical model for granular materials is investigated in a wide range of slip rate, focusing on two situations: (a) stationary sliding with velocity control, (b) unstable slip with a spring-block system on gouge layers. In the stationary sliding system, the parameter-independent "master curve" is found, in which the friction coefficient increases as the power of the slip rate with a nontrivial exponent. This is mainly due to the increasing random motion of gouge particles. Furthermore, quantities that describe the random motion of particles and the dilatation also obey power-law master curves with different exponents. In unstable slip regime, the system exhibits slip-weakening behavior. The critical slip distance, over which the frictional strength decreases, is investigated as a function of the system parameters. It is found that the critical slip distance strongly depends on the maximum slip rate and can increase by several orders of magnitudes than that for stationary sliding. This means that the critical slip distance cannot be determined solely from the surface roughness and is essentially coupled with the slip dynamics. As a closing remark, an ongoing experiment on granular friction is introduced, in which the temperature and the (internal) shear rate is strictly controlled. It is found that the power-law dependence of the friction coefficient on slip velocity overwhelms the logarithmic rate dependence which originates from the rate-state friction law.
S41A-1817
Mechanics of Post Seismic Deformation of the 1999 Chi-Chi, Taiwan, Earthquake
GPS measurements of 450 days after the 1999 Chi-Chi, Taiwan, earthquake revealed large postseismic displacement, from dozens to hundreds of millimeters, especially on the hanging wall of the fault. Two possible mechanisms responsible for the postseismic deformation of the Chi-Chi earthquakes are the viscoelastic relaxation of the lower crust and the afterslip in the regions with coseismic slip deficit on the fault plane, respectively. In this study, the postseismic deformation process of the 1999 Chi-Chi, Taiwan, earthquake is modeled dynamically by the revised LDDA method with the GPS data as constraints and the possible postseismic deformation mechanisms, viscoelastic relaxation and afterslip, are discussed. The results indicate that the postseismic deformation observed by GPS can be well explained by the viscoelastic relaxation mechanism. The viscosity of the lower crust below the Taiwan area is estimated as 1017 Pa· s by applying the trial and error method. The slips on the fault focus mainly on the part near the ground surface, and the maximum and minimum values are on the ground surface and at the fault tip, respectively. The slips decrease with depth, postseismic slips on the fault increase with time, but the speed of increase diminishes. Observable slip of the fault on the ground surface due to postseismic stress relaxation will last about 13 years. The westward and upward components of the total slips accumulated during this period are about 25.5% and 14.7% of the coseismic slip, respectively. Shear stress drops on the fault are heterogeneous. The effect of afterslip on the postseismic deformation of the Chi-Chi earthquake can be investigated by the LDDA method with rate and state dependent frictional law. Supposing that the fault is a rate-strengthening material, the results obtained suggest that the GPS data cannot be explained by the afterslip mechanism alone, which can only affect the postseismic deformation in the early 50 days after the mainshock.
S41A-1818
Episodic Tremor and Slip: an Experimental Approach.
We have devised a laboratory experiment to investigate the frictional and acoustic patterns of a salt slider over a large number of deformation cycles. We observe a continuous change of the frictional behavior of the slider under constant experimental conditions of stiffness, temperature and loading velocity. The stick-slip regime is progressively vanishing, eventually reaching the stable sliding regime. Concomitantly, the contact interface, observed under the microscope, develops a striated morphology with contact asperities increase in length and width, arguing for an increase in the critical slip distance dc. Complementary experiments including velocity jumps show that the frictional parameters of the rate and state friction law, a and b, progressively vanish with accumulated slip. The ultimate stage of friction is therefore rate and state independent under our experimental conditions. The Acoustic Emission evolves with cumulative displacement and interface ageing, following a trend from strong impulsive events to a collection of smaller amplitude and longer duration signals. We tentatively extend these results to natural subduction zones: shallow loud earthquakes, medium depth slow, deeper silent quakes and deepest steady-state creep are reproduced by the ageing of contact interface with cumulative displacement. In the meantime, the seismic energy release is evolving from seismic-like signals to NVT-like signals. NVT would emerge as the local recollection of the unstable behavior of the contact interface globally evolving towards the stable sliding regime.
S41A-1819
Analog earthquakes: Friction experiments with bulk solids and implications for fault dynamics
We report on friction experiments on dry bulk solids used to simulate slow ("aseismic") and fast ("seismic") slip in analog laboratory experiments. Ring shear tests have been performed at typical experimental conditions, i.e. in the very low stress regime (normal loads < 0.02 MPa, shear velocities 0.05 - 25 mm/min, T = 23°), and compared to similar tests at higher stresses (Mair and Marone 1999; Mair et al., 2002). Friction tests have been complemented with scanning electron microscope analysis to explore the control of particle characteristics (size, shape, sorting, roughness) on the strength and stability of frictional deformation. By varying the ratio between particle size and displacement we aim at simulating different brittle deformation mechanisms: Fine-grained material (sand, glass beads, sugar, salt) exposed to relatively fast shear undergoes distributed granular flow (DGF) including particle rotation, dilation and sliding. Rather slow shear of coarse material (rice) is accommodated mainly by localized particle boundary sliding (LBS). Both mechanisms are capable to produce frictional instabilities resulting in stick-slip motion. We found that frictional strength during DGF is controlled by particle shape: the frictional coefficient increases with the amount of angular particles consistent with previous findings by Mair et al. (2002) in the high stress regime (> 5 MPa). Also consistent with their work, unstable slip during DGF occurs if well-sorted and isometric particles dominate presumably controlled by the breakdown of force bridges. Frictional strength and stability during LBS seems to be controlled by surface roughness consistent with the concept of asperities. For particle surfaces with a scale-invariant roughness similar to natural faults, the critical distance over which slip has to accelerate to become unstable is not constant but increases with "interseismic" sliding velocity both during LBS (this study) and DGF (Mair and Marone, 1999). We found that the product of the critical distance and the critical time during unstable LBS is constant. If we interpret this observation in terms of fault dynamics, slip on a creeping fault has to accelerate more to nucleate an earthquake than slip on a locked fault. This may result in a positive feedback damping earthquake nucleation on creeping fault. Moreover, since acceleration during earthquake nucleation controls the spectrum of radiated seismic energy this may result in a spectral shift towards higher corner frequencies for earthquakes nucleating on creeping faults. This implies a direct correlation between the seismic coupling coefficient and the seismic source spectrum.
S41A-1820
Slow Earthquakes Induced by Fluid Flow and Inelastic Pore Creation
We study the generation mechanism of gslow earthquakesh, which radiate negligible elastic wave energy compared with ordinary earthquakes; as the examples, we can mention slow-slip events, low-frequency earthquakes and very-low-frequency earthquakes. Typical features of these slow earthquakes will be that the fault growth speed and stress drop are markedly lower than those of ordinary earthquakes. The high-pressure fluid seems to be involved in the generation of slow earthquakes according to seismological observations. The diffusivity of fault zone materials suggests that the fluid can flow at least about a few centimeters during high-speed fault growth [Mase and Smith, 1987]; this flow distance is almost comparable to the thickness of ultracataclastic fault core [Chester et al., 1993]. Hence, the fluid flow will play more important role if the fault growth speed is much lower than that observed for ordinary earthquakes. We numerically simulate features of slow earthquakes to understand their generation mechanism in the framework of dynamic modeling. We assume a fault in a thermoporoelastic medium taking account of fluid flow and inelastic creation of pores on the fault. In our previous study Suzuki and Yamashita [2008], on which the present study bases, a nondimensional parameter Su was shown to play a critical role in the dynamic fault growth. The parameter Su represents the dominance of effect of inelastic pore creation over that of frictional heating under the condition of no fluid flow. However, it can be shown that Su plays an important role even if the fluid flow is considered. We successfully simulate slow fault growth and low stress drop, which characterize the slow earthquakes. Critical ingredients of our modeling are (1) to assume Su considerably larger than assumed for the simulation of ordinary earthquakes, (2) to assume the fluid flow into the inelastically created pores and (3) to assume the initial shear stress significantly smaller than assumed for the simulation of ordinary earthquakes.
S41A-1821
Modeling the Effect of Quasi-Drained Fault Zone Dilatancy on Friction Constitutive Properties
Dilatancy may have a significant effect on the strength and stability of fault zones through increases in
effective stress resulting from decompression of pore fluids. As fault accelerates from a slow steady
background value to earthquake like velocities the porosity of the gouge layer may change [e.g. Δφ = -ε ln(v0θ/Dc); Segall and Rice, 1995]. If the rate
of fluid diffusion into the gouge layer is small compared to the rate of pore volume increase, the effective
stress on the fault zone will increase, which may inhibit the nucleation of unstable slip. In addition to
strengthening through an increase in effective stresses, the evolution of frictional strength may also change
via changes in the rate and state friction parameters a, b, and Dc. Such changes may result
from changes in contact junction properties, as a function of normal stress or porosity, or via granular
effects. We explore such effects using a simple, 1-D elastic model for a gouge layer of known frictional
strength (μ), permeability (k), initial porosity (φ0), dilatancy coefficient (ε), and rate and state friction parameters (a, b, Dc). The gouge layer is saturated
with water of bulk modulus (Kw), initial pore fluid pressure (Pp), and viscosity (η).
We conduct velocity stepping experiments and study evolution of friction, porosity, and effective stress using
rate and state friction laws. Elastic interaction between the slider and the layer is controlled by the stiffness of
the spring pulling the slider (Ks) at effective normal stress (σn-Pp) via the
equation
dμ/dt = Ks(vlp-v)/(σn-Pp) + μ/(σn-Pp)dPp/dt.
This complex coupling can lead to changes in the
effective magnitude of all of the rate and state parameters, a, b, and Dc, but at steady-state
leave the combined friction rate parameter, a-b, unchanged. We explore the unsteady regime where
the parameters are evolving, representing the case where the time of drainage in the gouge layer is long
compared to the time necessary for the new steady-state porosity to evolve (θ). We investigate
these changes by modeling a range of drainage parameters (k,η,ε,φ0) paying particular attention to ε, which we determined experimentally for fine grained
quartz sand (4.7e-5 at 0.8 MPa to 3.8e-4 at 30 MPa), Westerly granite gouge (2.0e-4 at 5 MPa to 1.2e-4 at
30 MPa), and clay rich ODP gouge (1.5e-4 at 5 MPa to 1.1e-4 at 30 MPa). We invert our model output
results to obtain effective values of the friction parameters a, b, and Dc. Early results
indicate (using ε = 5e-4, φ0 = 0.15, and a 10-30 μm/s velocity step) that
Dc can be effectively doubled as a result of undrained fault zone depressurization.
S41A-1822
Determining Pore Pressures Along a Slip Surface Within a Saturated Elastic-Plastic Porous Medium
Here we consider shear rupture along a slip surface in a fluid-saturated elastic-plastic porous medium, like in landslide and earthquake modeling, and assume that there are different poro-elasto-plastic response properties on the two sides of the slip surface. This different response may be because the fault bordering materials are dissimilar, or just because one side is actively yielding and the other is not, or is yielding but in a different mode. In effect, we are representing a core about a slip surface that divides two similar or contrasting materials. This representation is especially relevant in earthquake rupture dynamics. Studies of mature fault zones have noted a trend of fractured host rock extending 10–-100m from the fault, with an ultracataclastic core ~100mm about or to one side of the principal slip surface (e.g., Chester and Chester, Tectonophys, 1998; Chester et al., Columbia Univ Pr, 2004). Furthermore, there is likely to exist a material contrast that may come from accumulating km of slip and a bias in accumulated damage. The local pore pressure at the slip surface influences the rupture dynamics because, through the effective stress concept, it controls the local shear strength along the fault, a feature neglected as a simplification in our preliminary poro-elasto-plastic modeling of dynamic rupture (Viesca et al., JGR, 2008). To determine pore pressures at the slip surface under locally elastic-plastic response, we must consider pore pressure discontinuities about that surface that arise in an undrained treatment of off-fault material and their amelioration within resulting thin diffusive boundary layers, such that pore pressure and fluid mass flux in the normal direction are continuous at the slip surface. Our approach builds on previous work considering the effect of contrasts in poroelastic properties on rupture propagation (Rudnicki and Rice, JGR, 2006; Dunham and Rice, JGR, 2008). Here we find expressions for the undrained pore pressure changes in the thin damage zone adjacent to the slip surface. This expression is dependent on the local fault-normal stresses and fault- parallel strains coupled to the undrained response in the adjacent bulk. Additionally, in the process of determining the pore pressure at the slip surface, we find a diffusivity under elastic-plastic response that is different from the diffusivity of linear poroelasticity theory (e.g., Rice and Cleary, Rev Geophys, 1976). We show how the full inclusion of pore pressure changes by these considerations affects rupture dynamics in explicit dynamic numerical simulations, first checking for validation against the poroelastic results of Dunham and Rice (2008), and then examining the effects relative to the fully undrained elastic-plastic simulations of Viesca et al. (2008) which neglected pore pressure changes on the fault itself.
S41A-1823
Healing of Experimentally Simulated Fractures: Contact Neck Growth, and Strength Evolution of the Interface
To investigate the physical processes operating in active fault zones, we conduct analogue laboratory experiments where we simulated a rough fracture that undergoes healing/shear cycles under dry condition or in the presence of a reactive fluid. This set-up is a surrogate for the healing/sealing of fractures and faults rocks, where fluid-rock interactions are operative at the millennium time scale. A rough slider of sodium chloride was left in contact with a flat glass plate under a constant normal load. The whole set-up was mounted on a microscope and left in a temperature-controlled box. The closure of the interface through several days was measured using high resolution displacement sensors and the contact surface was continuously imaged with a CCD camera. Under dry conditions, a small transient creep displacement perpendicularly to the fracture plane was measured. This deformation last for several minutes and finally stopped. Under fluid saturated conditions, a slow closure of the rough interface was measured over several days. This closure was concomitant with the growth of contact points, driven by surface tension forces. After 50 hours, up to 10% of the fracture surface was healed by this process. The force necessary to break the adhesion forces, allowing the sample to slide, was also measured after several periods of increasing holding times and showed a power law dependence with time. After each experiment, the fracture interface roughness was measured to nanometer resolution using white light interferometry. The morphology of the contacts was characterized by scaling relationships. Finally, we propose a macroscopic constitutive model of fracture closure and strength recovery, related to the dynamics of the contact asperities, which are seen to flatten and expand through time.
S41A-1824
Seismic and Aseismic Processes on the Psathopyrgos Normal Fault, Western Rift of Corinth, Greece.
The western rift of Corinth (Greece) displays a fast opening rate (1.5 cm/year from GPS) and a very high, strongly fluctuating microseismic activity, monitored since 2000 by a local seismometer array. Several moderate to large earthquakes (M=6 to 6.7) have occurred historically on its segmented normal fault system, and one earthquake in this magnitude range is expected to occur in the coming decades. We focus here on the activity of the Psathopyrgos fault, the westernmost one in the rift ( 15 km away from the city of Patras), which did not rupture in historical time (more than 300 years), despite its clear morphological slip activity. This raises the question of possible creep on this fault, as a future coseismic release of more than 3 to 4 meters of slip on a 15 km long segment seems unlikely. A large seismic swarm occurred on this fault, in November-December 2002, with several events above magnitude 3, mostly located between 4 and 10 km in depth, and one reaching magnitude 3.5, at the exceptionally shallow depth of 3 km: this swarm suggests a large-scale activation of the whole fault, by creep or fluid migration. About 30 minutes before this M3.5 event, a strong compression signal is recorded oh the borehole dilatometer installed 15 km away in the Trizonia island, reaching its maximum at the time of the event, then gradually decaying. This signal can be modelled by a 10 cm aseismic creep involving the shallowest, eastern part of the fault. These observations raise the question of the possibility for mature normal faults to slip intermittently, both seismically and by creep.
S41A-1825
Triggering effect of M>4 earthquakes on the earthquake cycle of repeating events at Parkfield
How the stress perturbation by nearby earthquakes influences earthquake recurrence is of fundamental importance to understanding of the earthquake cycle and to determining earthquake hazard. The large population of characteristically repeating earthquakes at Parkfield provides a unique opportunity to study the degree to which stress interactions between earthquakes may influence earthquake recurrence intervals. Here we examine the response of the repeating events to the occurrence of M > 4 earthquakes. Using 187 M -0.4 ~ 1.7 repeating earthquake sequences (1987-1998) from the High Resolution Seismic Network, we find that the time to recurrence of repeating events subsequent to the M4-5 earthquakes is unusually short, suggesting triggering by major events. The triggering effect is found to be most evident within a distance of 5 km and decays with distance. We also find coherently reduced recurrence intervals from 1993 to 1998. This enduring recurrence acceleration over several years reflects the accelerated slip associated with several M >4 events and aseismic transient during the early 1990s. Using 25 M -0.51 ~ 2.16 updated HRSN repeating sequences (1987-2006), we also evaluate the effect of the 2004 M6 Parkfield event sequence (i.e., M 6 mainshock and three M > 4 aftershocks). The recurrences of the 25 updated HRSN repeating sequences exhibit a strong acceleration pattern associated with 2004 M 6 Parkfield event. The accelerated recurrence patterns of repeating sequences appear to be similar when they are close in space. The characteristically decaying afterslip pattern is not obvious for some of the repeaters adjacent to the largest co-seismic slip area, suggesting either that the stress changes are very heterogeneous, or that the rupture erased or shut off some of the sequence source areas.
S41A-1826
Can rate and state friction laws explain frictional behavior at high slip rates?
With increasing number of data reported from high velocity frictional experiments (Tsutsumi & Shimamoto, 1997; Goldsby & Tullis, 2002; Mizoguchi et al., 2007; Han et al., 2007), it seems plausible to believe that most crustal material undergo dramatic weakening as slip rates exceeding decimeters per second. Sone and Shimamoto (2006, Eos Trans. AGU) reported results from high velocity frictional experiments using natural fault gouge samples that mimicked an accelerating and decelerating fault motion during an earthquake in contrast to constant velocity tests. Their results further imply that gouge friction not only weaken with slip, but is also sensitive to the continuously changing slip velocity (velocity weakening). In order to explain their observations, they used an empirical equation that describes an exponential slip-weakening from a peak to steady-state friction, but with a continuously renewed steady-state friction due to velocity weakening. Another notable result was that it required several decimeters of slip before friction reached its peak value (about 0.8) from the initial value (about 0.6). On the other hand, studies on dynamic rupture of numerical faults obeying rate and state friction (Perrin et al., 1995; Bizzarri & Cocco, 2003) show that the combination of the direct effect and state evolution yields fault traction curves that are qualitatively similar to above. The direct effect creates a strengthening phase in the early stage of slip, and the state evolution creates a slip- weakening phase which is followed by a slight or significant healing phase depending on the specific form of evolution laws and cut-off velocities used. Whether the qualitative similarity of the friction/traction curves, between these laboratory data and numerical studies, imply that the empirical equation by Sone and Shimamoto (2006) is simply a manifestation of rate and state friction does not seem to be a trivial problem. Attempts to describe high velocity friction data using rate and state friction laws yield a-b (=dfss/dln(V)) values and characteristic displacement, L, that are about an order or more higher than typical values reported in the literature. However, the choice of these parameters satisfying high velocity friction data are highly dependent on the selection of lower cut-off velocities and the recognition that slip-weakening distance from high velocity friction experiments become shorter at greater normal stress (Mizoguchi et al., 2007). Careful discussion is needed to understand whether rate and state friction laws are applicable at co-seismic slip rates.
S41A-1827
Dynamic Rupture on Rough Faults and Production of High-Frequency Radiation
We have developed a highly accurate 2D finite difference method to solve dynamic rupture problems in irregular geometries. Our objective is to connect properties of high frequency radiation produced during slip on rough faults to statistical measures of fault roughness (namely, the amplitude-to-wavelength ratio, γ, of self-similar fractal faults). We study differences between antiplane and in-plane propagation; while faults are rougher in the direction perpendicular to slip (γ = 10-2 vs. 10-3 in the slip direction), only in the in-plane case does slip on rough faults alter the normal stress. Antiplane propagation is only mildly perturbed at γ = 10-2, suggesting that it might be possible to construct approximate broadband seismograms as the sum of the wavefield from slip on a flat fault plus a first-order correction term to account for roughness. The changes in normal stress in the in-plane case can dramatically influence the rupture process; in some cases, conditions that permit propagation on flat faults are insufficient to host ruptures on rough faults. To handle irregular geometries, we transform the governing equations from a non-Cartesian coordinate system that conforms to the irregular boundaries of the physical domain to a Cartesian coordinate system in a rectangular computational domain, and solve the equations in the computational domain. To accurately capture the high frequency wavefield, we use a numerical method that produces far smaller oscillations than those plaguing conventional finite difference/element methods. The governing equations (momentum conservation and Hooke's law) are written as a system of first-order equations for velocity and stress, which are defined at a common set of grid points and time steps (i.e., there is no staggering in space or time). Time stepping is done using an explicit third-order Runge-Kutta method. The equations are hyperbolic and the fields can be decomposed into a set of waves (with associated wave speeds). Spatial derivatives are computed with fifth-order WENO (weighted essentially non-oscillatory) finite differences in the upwind direction associated with each wave [Jiang and Shu, J. Comp. Phys., 126(1), 202-228, 1996]. Rather than using data from a single stencil (i.e., set of grid points) to calculate the derivative, a weighted combination of data from several candidate stencils is used. The weights are assigned based on solution smoothness within each stencil, and stencils in which the solution exhibits excessive variations are given minimal weight. Consequently, numerical oscillations are suppressed, even in the vicinity of the rupture front and at wavefronts.
S41A-1828
Investigating the Creeping Segment of the San Andreas fault using Persistent Scatterer Interferometry
We analyze the temporal characteristics of the creeping section of the San Andreas fault in Central California, using persistent scatterer interferometry (PS-InSAR) time series methods. In PS-InSAR, we identify a network of pixels whose scattering properties vary little between multiple SAR acquisitions spanning a period of time and use phase measurements at these points as a function of time to derive deformations. Applying PS-InSAR to natural terrains where conventional interferograms tend to suffer decorrelation is difficult, yet several PS-InSAR methods have been proposed and have been shown to work reliably in urban environments. The Stanford Method for PS (StaMPS) was the first method developed to extend the scope of PS-InSAR to work effectively in vegetated regions. We applied a maximum likelihood approach to PS selection and find it to be effective in identifying PS points in vegetated areas of the San Francisco Bay Area and Imperial Valley in California, USA. A key advantage of both StaMPS and the maximum likelihood method are that they do not require an a priori temporal model for the deformation pattern. Here we present results from applying these methods to the creeping section of the San Andreas fault. This segment of the fault creeps at rates in excess of 20 mm per year. Geodetic measurements in this area from creepmeters, alignment arrays and GPS typically have poor spatial and/or temporal resolution. Conventional stacking of ERS interferograms covering this segment of the fault provides good surface deformation information in parts of this region, but is not viable in areas that are heavily decorrelated due to vegetation and topography. The PS methods generate time series of surface displacement, even in steep, vegetated areas, and readily reproduce the creep rate of about 26 mm/yr along the fault and the spatial distribution of deformation away from the fault. The results are consistent, but more detailed than, the observations from GPS networks. We use elastic dislocation maps to invert for the time-series of the fault creep.
S41A-1829
Slow-Speed Weakening of Serpentinite Sheared Against Quartzofeldspathic Rocks: Implications for Fault Creep in the San Andreas System
Serpentinized ultramafic rocks are closely associated with creeping faults of the San Andreas system in central and northern California. Although serpentinite is commonly invoked as the cause of the creep, serpentine minerals can exhibit unstable (seismogenic) slip under certain conditions. However, these serpentinites are often juxtaposed against quartzofeldspathic rocks at depth along the creeping faults of the San Andreas system, and the compositional contrast potentially could influence the mechanical behavior of the faults. To investigate this possibility, we are conducting triaxial experiments under hydrothermal conditions to determine the effect on the frictional properties of serpentinite gouge caused by shearing between forcing blocks of granite or quartzite. All water-saturated experiments were conducted at a fluid pressure of 50 MPa and effective normal stress of 100 MPa, at temperatures between 200 and 300C (corresponding to depths of 6-10 km) and shearing rates of 3.6-360 cm/yr. In this temperature range and under water-saturated conditions, the coefficient of friction of pure granite or quartzite is 0.7-0.8 and that of pure lizardite or antigorite serpentinite is 0.5-0.6, and the strength of each material is greater at 300C than at 200C. However, when either the lizardite or antigorite serpentinite is sheared against granite or quartzite at 200-300C, its strength is reduced by as much as 40 percent to a coefficient of friction of about 0.3, with the greatest strength reductions at the highest temperatures (temperature weakening). During velocity-stepping experiments, strength decreases substantially with decreasing shearing rate (velocity strengthening), and the strength reductions are reversible when the rates are increased. Initial tests substituting dunite for serpentinite show similarly marked weakening behavior. In contrast, dry serpentinite gouge sheared slowly between granite blocks at 250C has a coefficient of friction of about 0.8. Preliminary SEM and XRD analyses of the run products yielded little evidence of the growth of new, weak minerals on the slip surfaces over the duration of the experiments (2-9 days). Although the exact mechanism is still being investigated, the cause of the weakening is presumed to be a solution-transfer or other fluid-assisted process involving the dissolution of serpentine minerals at grain-to-grain contacts. In quartzofeldspathic rocks, solution-transfer creep becomes important at T>350C, corresponding to depths below the base of the seismogenic zone. The chemical contrast between serpentinite and quartzofeldspathic rocks juxtaposed across an active fault may therefore contribute to aseismic slip (creep) at T<350C.
S41A-1830
Generating low frequency radiation from stick slip rupture
Low frequency seismic radiation has been observed in a variety of diverse environments from volcanoes to subduction zones. The dominant low frequency content from such earthquakes has been attributed to a variety of causes from slow slipping faults in subduction zones [Shelly, 2006] to volumetric pressure changes in magma vents [Chouet, 1996]. Using 3D finite difference rupture modelling we investigate what conditions are required to generate low frequency radiation on a brittle stick-slip fault. Using a simple asperity model, we investigate how different rupture parameters and environmental conditions influence the frequency content of seismic radiation recorded at a given station. Preliminary results show that constraining the slip rate provides the most effective means of decreasing the frequency content of the recorded ground motion. Increasing Poisson's ratio in the media around the fault also leads to a decrease in the frequency content in recorded ground motion and a slower rupture velocity. The frequency content of seismic radiation does not appear to be as sensitive to different stress drops and rupture velocities. Our current focus is on investigating mechanisms that control the slip rate and hence low frequency radiation.
S41A-1831
Modeling Activity of Short-term Slow Slip Events in the Deeper Parts of the Nankai Subduction Zone Considering the Segmentation of the Slip Zone
We have developed a 3D model of the short-term slow slip events (SSEs) on the subduction interface beneath Shikoku, southwest Japan, using a rate- and state-dependent friction law (aging law) with a small cut-off velocity to an evolution effect, following the study conducted by Shibazaki and Shimamoto (2007). We assume low effective normal stress and small critical displacement at the SSE zones. On the basis of the hypocentral distribution of low-frequency tremors (LFTs) determined by Obara (2008), we set three SSE- generation zones: a large segment beneath western Shikoku and two smaller segments beneath central and eastern Shikoku. The numerical results show that the events reproduced by the model beneath western Shikoku have longer lengths (around 100 km) in the horizontal direction and longer recurrence times (0.25-- 0.5 years) than those beneath central and eastern Shikoku (20--40 km and 0.2--0.3 years). The numerical results are consistent with the observations of Obara (2008) in that events beneath western Shikoku have longer lengths (around 100 km) and longer recurrence intervals (0.25--0.5 years) than those beneath central and eastern Shikoku. The activity of SSEs is determined by a segmentation structure in the frictional properties in the transition zone. We also report the results of numerical modeling on the SSE activity beneath the Kii Peninsula and Tokai region by considering a segmentation structure of the SSE generation zone. Next, we attempted to model the very-low-frequency (VLF) earthquakes accompanied by short-term SSEs, on a 2D thrust fault. We consider a local patch in which the friction parameters are varied. There exist two plausible models for generating VLF earthquakes. In the case that the critical displacement is very small at the patch, fast multiple slips occur at the patch. In the case that the effective normal stress is high at the patch, the patch acts as a barrier to SSEs; when it ruptures, however, rapid slip occurs. Because the source time functions of these cases are somewhat different, it will be possible to assess which model is more appropriate for VLF earthquakes.
S41A-1832
Fast Aseismic Growth of a Fault-related Fold Associated with the 2007 Chuetsu-Oki Earthquake (M6.8) in Japan: Space Geodetic Observation and Modeling
Niigata basin is located at a diffuse plate boundary between Eurasian plate and North American plate. Although there are no clear plate boundaries, GPS data indicate that a broad area from Niigata to Kobe is undergoing significant strain concentrations (Sagiya et al. 2000, PAGEOPH). Over the past 50 years, there have been a couple of large inland earthquakes, and the 2007 Chuetsu-Oki earthquake (Mj6.8) on July 16 is one of such events. Geologically, this area is known for active folding and thick sedimentary layer (Sato and Kato, 2005, EPS; Okamura et al. 2007, JGR). Using an L-band ALOS/PALSAR InSAR data, we detected crustal deformation signals associated the earthquake. Besides broad signals near the hypocenter, we found that significant deformation took place around Nishiyama Hill, which was ~15 km away from the epicenter and known as a local anticline axis, extending ~30 km. The aftershock data revealed very few events around the Nishiyama Hill, and thus the detected deformation indicates an aseismic growth of fault-related fold. Neither InSAR data nor nationwide GPS data shows significant postseismic deformation signals; the earliest post-earthquake SAR data was acquired 3 days after the earthquake. It turns out, therefore, that the aseismic slip generating the fault-related fold was rather fast, although the slip velocity was not fast enough to generate seismic waves. We constructed a fault source model, consisting one main shock fault and two other inland faults. Our modeling result indicates that the moment magnitude of the aseismic fault for the fold growth is Mw 6.0, and that the slipped area is almost comparable to that of main shock fault (Mw 6.6). The estimated size is overwhelmingly larger than the earlier study by Nishimura et al (2008, GRL), which we think is due to the difference in the derived area of aseismic deformation.
S41A-1833
A Brownian Walk Model for Slow Earthquakes
Along some subduction plate boundaries, slow deformation is observable as seismically detected deep low- frequency tremor and geodetically detected slow slip events. These phenomena are considered as different manifestations of slow earthquakes characterized by fairly constant seismic moment rate. This paper presents a simple model of slow earthquakes that can explain wide variety of observed features including the steady moment rate and scaled energy, characteristics of tremor signals both in time and frequency domains, and the migration of the source location. In this model, slow earthquakes are represented as shear slip on circular faults whose radius is a random variable that is governed by a Langevin equation and three parameters, a diffusion coefficient, a damping coefficient, and a slip rate coefficient. This model expands on a a previous scaling law for the slow earthquakes by providing a specific image of kinematics. Allowing for spatial variations of the parameters could potentially explain differences in behavior of slow slip events worldwide.
S41A-1834
Particle Dynamics Simulations of Fault Slip From the Micro- to the Macro-scale
Numerical simulations of granular shear using particle dynamics techniques, e.g., the discrete element method (DEM), have captured some aspects of rate-dependent variations in fault slip in the presence of fault gouge. Low slip velocities favor stick-slip behavior associated with sharp drops in shear stress along the fault, whereas higher slip velocities produce oscillatory variations in stress. Other observed phenomena include changes in strength and dilatancy correlated with steps in slip velocity, as well as time-dependent hardening of the gouge during episodes of no slip. Through such models, we can isolate the effects of granular interactions, and in particular, changes in grain packing, which strongly influence the rate- dependence of granular shear. Such microscale models, however, are governed by controlled boundary displacements, and thus have limited ability to capture emergent slip modes due to changing stress states and physical properties. Rate-dependence is also demonstrated in large-scale simulations, for example, of convergence and accretionary wedge formation above a frictional decollement. Here, temporal and spatial variations in basal strength can develop and dissipate, influencing slip modes along the decollement. Heterogeneous properties within cohesive materials above and below the decollement control the build-up and release of strain energy, and the partitioning of slip along multiple faults within the system. Within this complex system, slip rates on a given fault can range from slow creep to stick-slip, with the latter accompanied by rapid rupture, often reaching the surface of the wedge. Although DEM simulations at all scales still have quantitative limitations, they can provide valuable insights into the physical processes that may influence slip behavior in discontinuous and heterogeneous media that make up the shallow crust, and how these processes are manifested (e.g., seismically, structurally, and geodetically) at the surface of the Earth, where most natural systems are studied.
S41A-1835
Influence of heterogeneous coupling on the spatial pattern and recurrence of seismic ruptures
Observations of co-seismic, interseismic and post-seismic deformation suggest that friction properties of faults are heterogeneous with interfingering of patches with velocity-weakening (mostly undergoing stick-slip motion) or velocity-strengthening (mostly creeping) properties. For example, the pattern of locking of the Sunda megathrust in the interseismic period is highly heterogeneous and show both downdip and along strike variations (Hsu et al., 2006; Chlieh et al, 2008). The large earthquakes there appear to have ruptured patches within areas that had remained locked in the interseismic period, but some areas produced repeating similar events, while others produced quite dissimilar successive events. The behavior seems to depend on areas of low interseismic coupling, which can act as systematic or non-systematic barriers during the seismic rupture. Motivated by these observations, we carried out numerical simulations to explore how lateral variations in friction properties affect earthquake rupture patterns over many earthquake cycles. The fault is governed by rate and state friction with the aging form of state variable evolution. On the fault, two velocity-weakening segments are separated and surrounded by velocity-strengthening patches. Velocity-strengthening patches inhibit seismic slip to varying degree based on their size and friction parameters. The model can qualitatively explain the relation of interseismic coupling to co-seismic asperities on the Sunda Megathrust. The width and friction properties of the central velocity-weakening patch can be chosen so that some events are stopped by the patch and hence rupture only one locked segment, while others propagate through the patch, rupturing both locked segments and resulting in larger earthquakes. In the latter case, the patch causes a decrease in moment rate. By varying the strength and width of the velocity-strengthening patch, we identify parameter regimes in which such a patch acts either as a "permanent" or as a "weak" barrier. It is argued that depending on the characteristics of intervening velocity-strengthening areas, seismic asperities can either rupture in isolation or simultaneously, leading to the kind of complexity observed on the Sunda Megathrust. The modeling suggests that variations in the seismic moment rate of an earthquake, combined with the distribution of afterslip, can be used to infer friction properties along a subduction interface.
S41A-1836
Fault Slip Velocities Inferred from the Spectra of Ground Motions
There is much practical need in obtaining independent information about earthquake-source dynamic properties directly from observable data. One such dynamic parameter, the peak slip velocity during earthquake rupture, can be calculated from the corner frequencies of the source spectra, in the assumption of the validity of the ω2 source model. To obtain source spectra, observed Fourier spectra should be corrected for site and path effects. Small-to-moderate earthquakes in Japan recorded on multiple rock sites are suitable for the application of this methodology. The results indicate that the maximum slip velocity of the selected earthquakes ranged from approximately 0.2 to 0.6 m/s. Direct observation-based determinations of this type provide valuable physical information about the in-situ faulting processes that can be used for constraining dynamics theories of faulting or in ground-motion prediction.
S41A-1837
Recovery of active faults from surface displacement fields.
The goal of our project is to process measurements of surface displacements in such a way to use them as data for the inverse problem consisting of locating faults and portraying their geometry. Our research is also aiming at determining whether a measured displacement field on the surface is indicative of the onset of a fault destabilization phase. We have already entirely solved a two dimensional problem for the strike slip model, which essentially reduces displacement fields to two dimensional scalar fields. Deriving the inversion method involved a rigorous mathematical eigenvalue asymptotic analysis, leading to closed form inversion formulas. Those formulas were then tested for robustness in numerical simulations. As the strike slip model is limited in scope, we have worked on extending our results to fully three dimensional fault problems. In this much more difficult case, we have already obtained very promising closed form formulas (valid for the dominant part of the asymptotic behavior), and we have tested their use on numerical data. This is joint work with I. R. Ionescu, with the support of NSF grant DMS 0707421.