S33C-01 INVITED 13:40h
What controls slip heterogeneity- prestress, fracture energy, or sliding friction?
We are exploring the physics of earthquake ruptures to understand what controls slip and stress heterogeneity. We consider many variations in the fault friction model and seek a set of parameters that allows the system to evolve into a stable heterogeneous state. In other words, earthquake ruptures with heterogeneous slip are able to occur across a wide range of length scales. We test friction models with and without shear restrengthening (friction rises as slip rate drops) and with and without spatial variations in the fracture energy and dynamic sliding stress. Results from 3-D finite-element dynamic rupture simulations of earthquakes with magnitudes between 5.0 and 7.7 on a planar strike-slip fault show that: (1) {\em heterogeneous prestress} with {\em no shear restrengthening}, {\em uniform fracture energy} and {\em uniform dynamic sliding stresses} rapidly evolves into a nearly homogeneous stress state with earthquakes that only rupture the entire fault and produce distributions of {\em slip} that are much {\em too smooth}; (2) {\em heterogeneous prestress} with {\em shear restrengthening}, {\em uniform fracture energy} and {\em uniform dynamic sliding stresses} slowly evolves into a homogeneous stress state with earthquakes that rupture larger and larger portions of the fault and produce distributions of {\em slip} that are {\em too smooth}; (3) {\em heterogeneous prestress} with {\em no shear restrengthening}, {\em heterogeneous fracture energy} and {\em heterogeneous dynamic sliding stresses} evolves into a slightly heterogeneous stress state with too many large earthquakes that produce {\em reasonable} distributions of {\em slip}; (4) {\em heterogeneous prestress} with {\em shear restrengthening}, {\em heterogeneous fracture energy} and {\em heterogeneous dynamic sliding stresses} evolves into a heterogeneous stress state with a wide variety of earthquake sizes that produce {\em realistic} distributions of {\em slip}. As the stress state on the planar fault tends toward a more homogeneous distribution, the slip distributions become smoother and fail to exhibit the degree of heterogeneity seen in kinematic source inversions. The set of parameters that appear to produce the most realistic conditions (set 4) yields faults that are in a critical state and can fail across a wide range of length scales with slip distributions that have more appropriate levels of heterogeneity. This set of parameters might also be a proxy for nonplanar fault geometry; we suspect that nonplanar geometry coupled with relatively homogeneous friction parameters would yield similarly realistic behavior.
S33C-02 13:55h
Multi-cycle Dynamics of Branched Fault Systems
Recent earthquakes such as the 1992 Landers, 1999 Hector Mine, and 2002 Denali fault (Alaska) have shown the importance of understanding the dynamics of branched fault systems. Realistic dynamic models of earthquakes on such systems require a fault stress field that is consistent with both tectonic loading and the fault event history. Toward this goal, we perform multi-cycle dynamic simulations on 2D branched fault systems by modeling both the dynamic rupture process and the interseismic loading process. A 2D finite element method is used to simulate the co-seismic process with a slip-weakening friction law and full inertial dynamics, and a linear viscoelastic model is used to calculate stresses during the interseismic loading process. In our branched fault models, a secondary planar fault intersects the main planar fault in the center of the main fault and divides the main fault into two segments. We find that the branched fault geometry causes the fault stress to depart from the regional stress field significantly over multiple earthquake cycles. Locally increased normal stress can slow down or stop rupture propagation, while locally reduced normal stress facilitates jumping (discontinuous) rupture between branches. The fault behavior also includes examples of backwards branching, where rupture propagates around the acute angle between the primary and secondary segments. Various types of events occur on our branched fault systems over multiple earthquake cycles, including 1) events that only rupture the favorable segment(s), 2) events that rupture part of the main fault and the secondary fault, and 3) events that rupture the entire fault system. The results of our study may have implications for understanding branching behavior in the three earthquakes above, and also may help to delineate the range of expected behaviors for branched fault systems in the future.
S33C-03 14:10h
Short Fault Branches as Sources of Seismic Complexities
We analyze an earthquake rupture propagating along a straight "main" fault that is perturbed by a finite-length branch fault. Such intersections are often encountered in natural events. The predicted effects of the encounter with the branch that we report can be remarkable; it can strongly perturb the propagation velocity on the main fault and, in some cases, even arrest that propagation. From previous studies [Poliakov et al., {\it JGR}, 2002; Kame et al., {\it JGR}, 2003; Bhat et al., {\it BSSA} in press, 2004] it is understood what determines whether rupture begins along such a branch fault, and whether the branch fault captures the rupture path exclusively or if rupture continues on the main fault too. However, in the present case of rupture along a finite-length branch fault, the propagation must stop abruptly at the branch end (we neglect fresh fault generation there). Such a sudden stoppage will radiate significant stress changes [Freund and Fossum, {\it JGR}, 1975; Harris and Day, {\it JGR}, 1993]. What happens when those stress changes reach the main fault? Do finite branches thereby influence the rupture propagation characteristics on the main fault? Those questions are addressed by considering mode II rupture propagation along a planar main fault with a finite branch. We simulate the propagation using a 2D elastodynamic BIE formulation incorporating a slip-weakening Coulomb friction failure criterion. Different parameters used to describe the fault configuration and rupture propagation (inclination of the maximum principal compressional stress with the main fault, inclination of the branch, rupture propagation velocity at the branching point, length of the branch) have different influences on the rupture. While, in some cases, an infinite branch would have completely captured the rupture, thus stopping propagation on the main fault, a finite branch in the same configuration will sometimes allow the rupture to propagate. A finite branch also sometimes induces the stoppage of the rupture on the main fault in cases for which an infinite branch would have let it continue to propagate. In general the finite branch, if not completely ignored by the rupture process on the main fault, introduces sudden deceleration and then acceleration of that rupture propagation. Those will contribute to the high frequency content of radiated ground motions. The branch also introduces complexities in the slip pattern along the main fault, and in the residual distributions of normal and shear stress which remain on the fault after the rupture event. Hence, finite branches can be seen as sources of complexities on small spatial and temporal scales during the dynamic rupture, as possible arrestors of that rupture, and as generators of local stress field non uniformities that may affect nucleation and propagation of future events.
http://esag.harvard.edu/rice/OlivesBhRiDm_FinBr_27Aug04.pdf
S33C-04 14:25h
Dynamic Rupture Propagation on a Branched Fault System: The 1891 Nobi, Japan, Earthquake (M8.0)
The 1891 Nobi earthquake was the largest intraplate earthquake that occurred in recent 200 years in Japan. Although it occurred more than 100 years ago, there are several photographs with geological descriptions (Koto, 1893, J. Coll. Sci. Imp. Univ. Tokyo) as well as the fault offsets some of which still remain on the surface. We constructed the fault model based on Mikumo and Ando (1975, J. Phys. Earth), who used the field survey results of the fault traces by Matsuda (1974, Spec. Rep. Earthq. Res. Inst.). The fault model consists of 5 subfaults, three of which form the curved northwesten segment and the rest two are for the branched faults. We take into account the static stress distribution, referring to the results by Mikumo and Fukuyama (2004, AGU Fall Meeting). Concerning the hypocenter location, there are two possibilities: the northern end (Mikumo and Ando, 1975) and the southern edge of the northern subfaults (Muramatu et al., 2003). If we considered the arrival time differences of the initial P- and S- waves recorded at Gifu and Nagoya, the northern edge is more probable, and we used this locatrion as the hypocenter. We also assumed that the rupture started at a depth of 10 km, and that pure strike slip occurs down to a depth of 15 km. We computed a spontaneous rupture propagaton based on the boundary integral equation method with triangular elements (Fukuyama et al., 2002, AGU Fall Meeting). In this computation, a slip-weakening constitutive law is employed with exponential decay and the critical slip-weakening distance of 1m. We assumed a uniform tri-axial stress field around the fault system, and the yield and frictional stresses are computed from the normal stress multiplied by static and dynamic coefficients of friction, respectively. A series of computations tells us that since the rupture has to propagate a long way (55km) up to the junction before changing its direction, the rupture velocity becomes very sensitive to the stress field applied and it becomes rather easy to propagate with super-shear speed in order to make the rupture propagate along both of the branced subfaults. In this fault geometry, since the branch angle is large enough (44$^\circ$), there are no interactions between the branched faults, which are the same as shown by Aochi et al. (2000, GRL). Another critical parameter was the direction of the principal stress. We first used N105$^\circ$E (Mikumo and Fukuyama, 2004) but the dynamic simulation prefers to N90$^\circ$E in order for the rupture to propagate along both of the branched faults. These investigations provide us with additional information on the old earthquake whose rupture process is not well known. These approaches are found to be very useful especially for the case of insufficient information on the earthquake rupture.
S33C-05 14:40h
Resolution of Fault Processes in Near-source Records: Supershear Ruptures and the 2002 Denali Fault Earthquake
Ruptures propagate in one of two velocity regimes: either less than the Rayleigh wave speed (sub-Rayleigh) or between the S and P wave speeds (supershear). We present an overview of dynamic source representation, the basis of which is a consideration of how waves released by material failure processes in the fault zone transmit shearing forces ahead of the propagating rupture. Depending on the sign of the forces, they will either drive further material failure or act to lock the fault. The allowed velocity regimes follow naturally from this representation. Furthermore, any departure of rupture growth from steady-state conditions generates a set of elastic waves that diffract off of the moving crack edge. These transient diffractions provide the mechanism that allows ruptures to jump between the two propagation velocity regimes. This theory also predicts that the supershear transition should be accompanied by the release of a Rayleigh interface wave on the fault surface, which manifests as a secondary slip pulse trailing the supershear rupture. Until recently, seismic inversions have revealed most, if not all, earthquakes to be sub-Rayleigh. The 2002 Mw 7.9 Denali Fault earthquake offers evidence to the contrary (Ellsworth et al. 2004), in the form of strong ground motion pulses recorded at pump station 10 just 3km from the fault that differ completely from typical near-source motions from sub-Rayleigh ruptures. We present a simple dynamic rupture model of the event, in which an initially sub-Rayleigh rupture accelerates to supershear velocities 30km before the station. The ground motion is characterized by two sets of pulses, one set appearing when the supershear rupture passes the station, and the second when the theoretically predicted Rayleigh interface wave goes by. The interface wave, which appears naturally in our dynamic models, would be a difficult feature to reproduce in kinematic inversions. An exact match to the pulse widths is not possible using a homogeneous stress and strength distribution along with a slip-weakening friction law. Instead, matching the pulse widths requires either including healing on a 1.5s time scale within the friction law, or having stress heterogeneity with a 5km length scale. The ground motion from these two cases are virtually identical. We augment this analysis by calculating synthetic ground motions using 2D analytical solutions for steady-state ruptures, valid for any arbitrary traction function in the breakdown zone, in both the sub-Rayleigh and supershear regimes. These solutions place constraints on which friction law properties can be inferred from near-source records. At supershear rupture velocities, information remains unattenuated as it is transported away from the fault along the S wave Mach front, a situation that does not occur for sub-Rayleigh speeds. Consequently, ground motion from supershear ruptures offers unprecedented insight into such parameters as the rupture pulse length and the extent of the process zone.
S33C-06 14:55h
Shock S-Wave Characterization for Kinematic Fault Rupture Models With Constant Supershear Rupture Velocity
We present the specific amplitude and waveform characteristics of near-source S-shock-wave generated by a kinematic model of super-shear rupture at constant velocity v. Asymptotic analytical solutions are provided for the shock wave amplitudes, in relationship with the geometrical singularities carried by the S-wave isochrones on the fault plane. The solution is dominated by waves radiated near a critical point source A defined by $cos(\theta)=\beta / v $, where $\theta$ is the angle between the rupture ray and the S-wave ray normal to the rupture front at A, and b is the S-wave velocity. The far-field, dominant shock-wave velocity related to the mode II component of the slip is proportional to the slip velocity at A and to $cos(2\theta) / sin(\theta)$. Thus, the shock-wave front "carries" the motion on the fault plane at large distances, with little attenuation, within a "shock-wave beam" of rays characterized by their angle $\theta$. Numerical calculation of the complete field has been achieved up to 4 Hz, in a homogeneous elastic half-space, and for a vertical strike-slip fault 50 km long equivalent to a magnitude 7.1. It confirms these theoretical developments, and shows that the peak acceleration and velocities are at least twice that of a standard sub-Rayleigh rupture at 10 km, and up to 5 times its value at 30 km. Although the diffusion and diffraction of S-waves in the real crust is expected to reduce the coherence of the shock-wave front and hence its peak amplitude, specially at large distances and for high frequencies, our analytical and numerical developments demonstrates that supershear rupture can produce unusually large levels of ground motion at distances ranging from 10 to a few tens of kilometres, within the shock-wave beam.
S33C-07 15:10h
A High Frequency View of 1999 Chi-Chi, Taiwan, Source Rupture and Fault Mechanics
High-frequency band-pass filtering of broadband strong-motion seismograms recorded immediately adjacent to the fault plane of the 1999 Chi-Chi, Taiwan earthquake reveals a sequence of distinct bursts, each of which can be considered as a sub-event from an asperity source of the Chi-Chi mainshock. These bursts collectively make up the entire mainshock accelerogram. Each burst may have released a significant portion of the total energy release of the Chi-Chi mainshock. Many of these bursts contain quasi-periodic sub-bursts with periods on the order of a few tenths of a second. Most bursts occur well behind the propagating rupture front. Detailed pictures of these asperity sources do not appear in conventional slip-map studies, presumably because of the low-pass filtering used in these waveform inversions. We directly used the high-frequency data to determine the origin times, locations and magnitudes of these sub-events. The first asperities to rupture in a given location follow the Chelungpu rupture propagation history at a velocity of about 2.0 km/s. Later asperity events at a given location can be interpreted as aftershocks that begin before the Chi-Chi rupture has terminated. Spatially these asperity sources appear in groups, most of which are located at shallow depth along the Chelungpu surface rupture and are consistent with the large asperities presented in source inversion studies. Asperities located at great depth suggest a non-planer rupture surface with dip increasing to the east. The frequency-magnitude distribution of these sub-events has b-value equal to 1.0. In space, the larger sub-events are located at greater depth, while the small sub-events are only located at shallower depths.
S33C-08 15:25h
Modeling Physical Limits on Extreme Earthquake Ground Motion
When predicting ground motion at critical structures for very low probability of exceedance, Probabilistic Seismic Hazard Analysis can yield values much larger than any ground motion that has actually been observed in an earthquake. Are there physical limits that can constrain such predictions? Two physical principles can be applied: (1) maximum stress drop available at the source, and (2) strength of material through which waves propagate. Unfortunately, we do not know the state of stress on faults in the earth's crust. Shear stress ranging up to 100 MPa is allowed by laboratory values of friction. A stress drop of 100 MPa may occur in small patches, but such stress drop over a large area of a fault would produce ground motion that has never been observed. Whatever the stress, complete stress drop may be possible. Thermal pressurization of pore fluid from frictional heating can produce near-complete stress drop in large events in sufficiently impermeable material. The character of ground motion near the northern part of the rupture of the Chi-chi earthquake suggests that thermal pressurization may have occurred there. More understanding is needed of when near-complete stress drop may be expected. Strength of material provides a physical constraint on ground motion. Particle velocity propagated in an S wave is limited to shear strength divided by shear impedance, which effectively limits short-period motion, even though velocity can increase further as waves reverberate in a layer. Non-elastic response near the earth's surface constrains short-period motion. In addition, non-elastic response near a rupture front increases fracture energy and limits particle velocity at the source. To establish physical limits on earthquake ground motion, we need to use non-linear calculational methods.