S11A-0992 0800h
Simulations of Earthquake Nucleation, Its Static Perturbation, and Aftershock Rates on Faults with Rate and State Friction
Large earthquakes are usually followed by an increased seismic activity that decays to the background earthquake rate over time. The decay is well-described by the empirical Omori's law. Dieterich (JGR, 1994) proposed that Omori's law could result from statically perturbing a population of nucleation sites that are governed by laboratory-derived rate and state friction laws. Dieterich considered a spring-slider setting for each nucleation site with the assumption that the acceleration during nucleation is such that the state variable is significantly behind its steady-state value. This allowed him to derive analytical formulae for the nucleation process as well as for the resulting earthquake rate increase. This model has been used to analyze aftershock sequences and inspired further theoretical studies. In particular, Gomberg et al. (JGR, 2000) considered numerically aftershock rates in the spring-slider block model with the full rate and state friction relations. Our goal is to verify Dieterich's conclusions in a 2-D continuum setting, where the nucleation process can be more complicated than assumed in Dieterich's model (Lapusta, SCEC meeting, 2003). First, we simulate a sequence of earthquakes using the methodology of Lapusta et al. (JGR, 2000) and select the nucleation process of a typical event for further study. This ensures that the nucleation process proceeds in conditions naturally occurring in the model and not arbitrarily imposed. Next, we perturb the model by a static stress change at some time during the nucleation process and simulate the resulting behavior, which lets us to construct numerically the dependence F of the new time to instability on the timing of the static perturbation. Finally, following Dieterich (JGR, 1994), we imagine a population of nucleation sites each of which following the behavior observed in our model but being at a different point in the nucleation process such that the whole population would result in a uniform background rate. Then we use the dependence F to find how the earthquake rate generated by this population is changed when the static stress perturbation is imposed. We find that if the nucleation process proceeds in relatively uniform conditions then the results, in terms of the changed earthquake rate, are well-described by the Dieterich's model. This is despite the fact that the nucleation process is non-uniform spatially (the slowly sliding nucleation region is spreading) and hence is not well-represented by a single degree of freedom spring-slider model. If, on the other hand, the nucleation process occurs in more complicated conditions, such as stress concentration caused by the nearby creeping region, the resulting aftershock behavior can be somewhat different, with additional peaks in the aftershock activity. The important distinguishing feature seems to be the validity of the Dieterich's assumption about the behavior of the state variable; it is almost always valid in the uniform case but fails for some parts of the nucleation process in more complex scenarios. We will report on our current efforts to study the parameter space and understand the significance, if any, of the observed differences.
S11A-0993 0800h
Relation between Seismicity and Strain Rate in Japan
We examine the relationship between occurrence rates of small earthquakes and the crustal strain rates for the Japan Islands region. Recent GPS data from GEONET, operated by the Geographical Survey Institute, provide good measurements of the crustal deformation rates and the rates of small earthquakes are well recorded by the regional seismic networks. Therefore, there is good regional coverage of strain and seismicity rates across most of Japan. We want to look at the microearthquake occurrences associated with the secular rates of strain, so we eliminate locations of large earthquakes and high seismicity associated with volcanic events. For the crustal deformation rates, we calculated strain rate in a grid of even intervals of 0.1 degree. We used the method of Shen et al. (1996) which estimates horizontal displacement rate, strain rate and rotation rate at each grid point from observed displacement rates of the GPS stations. Then, we calculated dilatation rate and maximum shear strain rate using the estimated horizontal strain rate. For the seismicity rates, we used the JMA (Japan Meteorological Agency) hypocenter catalogue and estimated the number of earthquakes (M$>$2) that occurred within a 15km radius of each grid point at depths shallower than 20km. There is a large scatter in the plot of seismicity rate as a function of strain rate, but if values of seismicity rates are averaged over a range of strain rate values, some clear trends can be seen in the results. Comparing the numbers of earthquakes with the strain rates, we obtain the following results. Maximum shear strain rate and the number of earthquakes show a positive correlation for rates of 0 to 90 nanostrain/year. However, above 90 nanostrain/year, the number of earthquakes decreases with increase of the strain rate. Similarly, there is an increase in the number of earthquakes with increase of dilatation rate from 0 to -110 (negative dilatation rate indicates compression), and earthquake numbers decrease for values of compressive strain rate greater than -110. Thus, for both shear strain rate and dilatation rate, the number of earthquakes increased with increasing strain rate, but above some threshold strain rate, the numbers of earthquakes then decrease. The threshold above which seismicity rates decline, may be an important factor in the regional stress conditions that control the rates of small earthquakes.
S11A-0994 0800h
Low Stress Drop Swarm Events in the Yilgarn Craton, Western Australia
Since September 2001, the small rural community of Burakin, southwest Western Australia, has been at the focus of seismic activity in Australia. In the six month period following commencement of seismicity, some 18,000 events had occurred, the largest of which having a moment magnitude of M 4.6. At the onset of activity, Geoscience Australia made a concerted effort to deploy a temporary seismic network in the region. The primary objective of this network was to collect high-quality strong-motion data for use in attenuation studies. Levels of seismicity near Burakin have decreased significantly since the 2001-02 swarm, however the region continues to experience a few small earthquakes per month. Earthquake source and path parameters are evaluated for a subset of 67 earthquakes. The dataset comprises some 375 seismograph and accelerograph records for events of magnitude M 2.3-4.6, including strong-motion data for seven earthquakes of M 4.0 and greater recorded at hypocentral distances less than 10 km. Source parameters are evaluated from far-field displacement spectra. Average corner frequencies are typically quite low, chiefly ranging between 2-3 Hz for events M 3.0 and above. Given the small variability in corner frequency, stress drop is observed to increase with magnitude, from very low values of 0.04 MPa to 18 MPa for the largest events in the catalogue. The stress drops for lower magnitude events (M $<$ 4.0) are typically lower than those obtained for southeastern Australian earthquakes of similar seismic moment. Since corner frequency is not observed to vary significantly with seismic moment, it is thought that the spectral content of shallow, small swarm events and consequently, the stress drop, is characteristically different to that of isolated intraplate earthquakes. We suggest that the larger events may be faulting previously unfractured rock or healed fault asperities, while the smaller events are adjustment events or aftershocks and occur on recently faulted surfaces. The work described has provided a useful framework for the development of regional ground-motion relations for Western Australia and will enable a better understanding of the mechanisms driving intraplate seismicity.
S11A-0995 0800h
Earthquake Source Scaling of Moderate to Large Earthquakes in Taiwan: Study of 2003 Mw $>$6 Taiwan Earthquakes
We inverted velocity data from Broadband Array in Taiwan for Seismology (BATS) to study earthquake source parameters of three Mw$>$6 earthquakes in eastern Taiwan. These earthquakes occurred on June 9th, June 10th and Dec 10th, 2003, respectively. We use the least-square method to invert for the distribution of slip and other source parameters. The slip distribution and aftershock data were used to estimate the fault geometry. For each event, we derived a preferred model by testing different focal mechanisms. In order to have better azimuthal coverage to the earthquake, one Japanese station, YNG, from F-net was also used. Our results show that these earthquakes all have thrust fault mechanisms. The hypocenter of the June 9th earthquake is 24.4$\deg$N, 121.99$\deg$E, and has a depth of 21.3 km. The focal mechanism has a strike, dip, and rake of 225$\deg$, 26$\deg$, and 121$\deg$, respectively. The moment is 8.65$\times$1024 dyne-cm, which yields the Mw of 5.89. The main rupture is around the hypocenter, and propagates towards the northeast. The June 10th earthquake occurred at 23.52$\deg$N, 121.67$\deg$E, at a depth of 32.3 km. The focal mechanism has a strike, dip, and rake of 217$\deg$, 39 $\deg$, and 110$\deg$, respectively. The moment is 2.03$\times$1025 dyne-cm, which yields the Mw of 6.13. It ruptured toward the northeast. The September 10th earthquake occurred at 23.07$\deg$N, 121.40$\deg$E at a focal depth of 17.7 km. The focal mechanism has a strike, dip, and rake of 3$\deg$, 42$\deg$, and 104$\deg$, respectively. The moment is 2.68$\times$1025 dyne-cm, which yields the Mw of 6.22, with the main slip concentration to the north. According to the slip and aftershock distributions, this earthquake is believed to be associated with the Chihshang fault. We have investigated the relationship of earthquake magnitude (Mw) to fault length (L$_{2}$), fault area (A$_{2}$) and average slip (D$_{2}$). The relations among the source parameters are as follows: Mw = 1.4log(L$_{2}$)+4.5, Mw = 0.8log(A$_{2}$)+4.5, Mw = (1.3$\pm$0.45)log(D$_{2}$)+3.7 These scaling relationships provided important information for seismic hazard analysis and for better understanding of earthquake source characteristics in Taiwan.
S11A-0996 0800h
Deep Seismicity as a Proxy for Crustal Strength, a Case Study of the 1992 Landers Aftershock Zone
Strength is the stress required to cause permanent deformation in a rock. Strength may not be constant, but may vary with time. For example, because crustal strength depends in part on strain rate, the rock strength of a region may be temporarily altered in response to a large earthquake. Are such changes observable? I modeled crustal strength of the 1992 Landers aftershock zone considering the thermal state and changes in strain rate. These models, one for the northern and southern end of the aftershock zone, indicate that a small increase in the strain rate of a region has the potential to lower the depth of the brittle-ductile transition of a region, temporarily allowing earthquakes to occur deeper than before the Landers earthquake. I compared the pre-mainshock seismicity and the aftershock seismicity using the beta statistic of Matthews and Reasonberg (1988). The Southern California Seismic Network (SCSN) "A" quality events were divided into a fine grid for several time periods before and after the 1992 Landers earthquake. Comparing the pre and post-mainshock time periods revealed a small, but significant, increase in the depth of seismicity. Comparing distributions of the beta statistic between several post-shock time windows shows that the seismicity of the Landers region appears to consistently decrease in rate and to shallow over the course of 210 days following the earthquake. However, the northern and southern ends of the Landers aftershock zone behaved differently. The northern end had a faster decrease in the rate of deep aftershocks than the southern end, possibly as a function of an increase heat flow or possibly related to the directivity of the rupture itself.
S11A-0997 0800h
A 3D Mimetic Finite Difference Method for Rupture Dynamics
We are developing a method for solving earthquake rupture dynamics problems on structured curvilinear meshes. The advantage of a curvilinear mesh over a rectangular mesh is that it can accommodate free-surface topography as well as non-planar fault geometry. The advantages of using a structured mesh over an unstructured mesh (as used in many finite element methods) is simplicity and computational efficiency. Structured meshes also make a number of computational tasks easier, such as parallelization, or coupling with other codes that use similar structured meshes. To build the discretized equations of motion on a structured, yet non Cartesian mesh, we use a mimetic method, so named because it takes special care to mimic the important conservation properties of the original equations of motion. We begin by writing the equations of motion in terms of gradient and divergence operators. We then derive a discrete grad (or div) operator by differentiating an interpolation function of the discrete variable. Next, that grad (or div) operator is plugged into a discrete analog of Gauss' Identity and manipulated to find the adjoint div (or grad) operator. We use a computer algebra system to handle the manipulations, which is practically essential for the 3D case because of the extreme lengths of the expressions to be coded. The code is currently implemented as a "rapid prototype" in MATLAB and undergoing validation prior to conversion to a high performance language. We compare results for simple types of rupture that have analytical solutions.
S11A-0998 0800h
Modeling Finite Faults Using the Adjoint Wave Field
Time-reversal acoustics, a technique in which an acoustic signal is recorded by an array of transducers, time-reversed, and retransmitted, is used, e.g., in medical therapy to locate and destroy gallstones (for a review see \textit{Fink},~1997). As discussed by \textit{Tromp et al.}~(2004), time-reversal techniques for locating sources are closely linked to so-called `adjoint methods' (\textit{Talagrand and Courtier}, 1987), which may be used to evaluate the gradient of a misfit function. \textit{Tromp et al.}~(2004) illustrate how a (finite) source inversion may be implemented based upon the adjoint wave field by writing the change in the misfit function, $\delta\chi$, due to a change in the moment-density tensor, $\delta \mathbf{m}$, as an integral of the adjoint strain field $\epsilon^\dagger({\mathbf x},t)$ over the fault plane $\Sigma$: $\delta \chi = \int_0^T\!\!\!\int_\Sigma \epsilon^\dagger({\mathbf x},T-t) \!:\!\delta {\mathbf m}({\mathbf x},t)\:d^2{\mathbf x}dt$. We find that if the real fault plane is located at a distance $\delta h$ in the direction of the fault normal $\hat{{\mathbf n}}$, then to first order an additional factor of $\int_0^T\!\!\!\int_\Sigma \delta h ({\mathbf x}) \partial_n \epsilon^\dagger({\mathbf x},T-t)\!:\!{\mathbf m}({\mathbf x},t) \:d^2{\mathbf x}dt$ is added to the change in the misfit function. The adjoint strain is computed by using the time-reversed difference between data and synthetics recorded at all receivers as simultaneous sources and recording the resulting strain on the fault plane. In accordance with time-reversal acoustics, all the resulting waves will constructively interfere at the position of the original source in space and time. The level of convergence will be deterimined by factors such as the source-receiver geometry, the frequency of the recorded data and synthetics, and the accuracy of the velocity structure used when back propagating the wave field. The terms $\epsilon^\dagger({\mathbf x},T-t)$ and $\partial_n \epsilon^\dagger({\mathbf x},T-t)\!\!:\!\!{\mathbf m}({\mathbf x},t)$ can be viewed as sensitivity kernels for the moment density and the faultplane location respectively. By looking at these quantities we can make an educated choice of fault parametrization given the data in hand. The process can then be repeated to invert for the best source model, as demonstrated by \textit{Tromp et al.}~(2004) for the magnitude of a point force. In this presentation we explore the applicability of adjoint methods to estimating finite source parameters. \begin{thebibliography}{3} \bibitem{Fink1997} Fink, M. (1997), Time reversed acoustics, \textit{Physics {T}oday}, \textit{50}(3), 34--40. \bibitem{TalagrandCourtier1987} Talagrand, O., and P.~Courtier (1987), Variational assimilation of meteorological observations with the adjoint vorticity equatuation. {I}: {T}heory, \textit{Q. J. R. Meteorol. Soc.}, \textit{113}, 1311--1328. \bibitem{TrompTapeLiu2004} Tromp, J., C.~Tape, and Q.~Liu (2004), Waveform tomography, adjoint methods, time reversal, and banana-doughnut kernels, \textit{Geophys. Jour. Int., in press} \end{thebibliography}
S11A-0999 0800h
Imaging the M7.9 Denali Fault Earthquake 2002 rupture at the Delta River using LiDAR, RADAR, and SASW Surface Wave Geophysics
The Mw 7.9 Denali fault earthquake of 3 November 2002 resulted in approximately 5.5 meters of right-lateral offset and sub-meter (0.6m average) up-to-the north vertical displacement of alluvial deposits of the Delta River. We characterize the surface rupture and shallow fault structure of the Denali fault zone at the Delta River in order to better understand these most recent displacements and to estimate the total vertical offset of alluvium above glacially scoured bedrock. To analyze deformations along the fault-trace, we performed tripod-mounted ground-based LiDAR surveys, and Spectral analysis of Surface Wave (SASW) and Ground Penetrating RADAR (GPR) geophysical investigations. These studies were performed between the Trans-Alaska Pipeline (TAPS) corridor on the terrace deposits of the eastern flanks of the Delta River valley and the steeply sloping bedrock surface on the western side of the river. To produce digital terrain models (DTM) of the surface break we used a Riegl Z210i Laser-scanner to image eight independent LiDAR scans, and ISite3D modeling software to merge these scans into three DTM surfaces. We find that using a rotating scanning-laser allows us to produce ultra-high resolution quantitative DTMs for geomorphic analysis that can be used to resolve features and detect topographic changes on a fine-scale (0.9-2.5cm). Local geo-referencing control points are established using fixed auto reflectors. The near subsurface alluvium was imaged using reflection-based (GPR). A suite of parallel and orthogonal GPR reflection lines were measured to develop block models of the surface rupture at two locations. Radar imagery clearly delineates a plane of chaotic reflectors across the rupture zone. To characterize the depth of alluvium over bedrock on either side of the fault, we used the spectral analysis of surface waves (SASW) approach to invert the near-surface shear wave velocity profile. An Alyeska Co. Catepillar D9N track-mounted dozer was used as a high-energy random-wave source for the SASW test. This source allowed us to profile to depths in excess of 200 meters on either side of the fault. We found the combination of LiDAR and GPR allows us to analyze the surface and near-surface characteristics of a complex oblique rupture across the braid bars of the Delta River. SASW-based shear wave velocity profiles on either side of the fault indicate total up-to-the north uplift on the Denali fault of between 60-90 meters since Pleistocene (?) deglaciation. This investigation is the product of a collaborative research and development agreement between the Alyeska Pipeline Services Company, Pacific Gas and Electric Company and the U.S. Geological Survey.
http://walrus.wr.usgs.gov/geotech/
S11A-1000 0800h
2003 Bam Earthquake in Southeastern Iran: an Unexpected Event Associated With a Segmented Fault
The Bam fault is an active Quaternary fault that runs over 50 km. The segments north and south of the Bam scarp are high-angle faults with small east-side-up component. There are possible offset streams that indicate predominant right-lateral strike-slip offset. The Bam scarp represents an east dipping flexure or monocline on the east wing of an asymmetric anticline probably above a west dipping reverse fault. During the December 26, 2003 Bam earthquake, continuous ruptures with consistent right-lateral strike-slip occurred in the north of Posht-e-Rud. West strand extends about 5 km with small amount of extension. The west strand coincides with a trace of a Quaternary fault, namely, northern Bam fault. East strand extend about 2 km with small amount of compression. There is no evidence of precedent faulting along this east strand. These ruptures were probably triggered by the intense ground shaking of the 2003 earthquake. There is no systematic change in the amount of offset. The central part of the Bam scarp might have grown during the earthquake. The swarm of tension cracks extends mostly 165 degree with opening of 1 to 10 cm. There is a E-W trending normal fault away from the scarp. The N-S extension may be correlated to the E-W compression. In south of the Bam scarp, there was no systematic surface effect. The topography and geology suggest there is a southern extension of the Bam fault over 30 km. Geological and geomorphological evidence clearly indicate that the Bam fault has repeatedly ruptured over a 50 km long strand. The repeated faulting has created the Bam scarp and other surface features. However, the 2003 Bam earthquake did not create any significant fault topography on the surface. The absence of tectonic displacement at the surface is concordant with the small magnitude. This means the 2003 earthquake was not the largest earthquake expected from geologic evidence. Only a small portion of the Bam fault plane was ruptured at depth, or another adjacent blind fault plane was ruptured in 2003. There is no way to forecast such small blind earthquake either by seismological research or by geological research. The most devastating event for Bam has not occurred for past 700 years or more and did not occur in 2003, but will occur in the future judging from the characteristics of the Bam fault. It is now possible to foresee the large, probably M 7.5 or larger, event but there is no historic and geologic data to quantify the risks. In order to reasonably reconstruct the city of Bam, we need much information about an earthquake cycle of the Bam fault.
S11A-1001 0800h
3D Geometric Structure of the Blind Fault for the 2003 Bam Earthquake_CSoutheast Iran_CInferred From the Aftershock Distribution: Existence of the Arg-e-Bam Fault Proposed
The 2003 Bam earthquake (Mw 6.5) occurred in the southeastern part of Iran on December 26, 2003. The mainshock caused severe damage, more than 26000 people died and tens of thousands of people injured, in the epicentral area. The relation of the mainshock with the heavily damaged area has not been well understood since nobody could find the source fault and even the hypocenter of the mainshok was not determined with a high accuracy. In order to find the source fault and study the relation to the damaged area, we installed 9 temporal seismic stations in and around the Bam city and conducted an aftershock observation during 1 month. We successfully recorded more than 18000 aftershock data and scanned onsets of P- and S-waves manually after several automatic data processings. The determined epicenters from these onset data are distributed linearly over about 20 km in parallel with a line 3.5 km west of the geological Bam fault and extend from the southern part of Bam city to the heavily damaged area in the eastern part of the city including the historical mud brick citadel "Arg-e-Bam". The hypocenters are distributed in a depth range from 0 km to 15 km and a seismic gap can be distinguished at a depth of less than 8 km. We named this hypocenter distribution "Arg-e-Bam fault" (Nakamura et al., 2004). We modeled 3D structure of the Arg-e-Bam fault based on the hypocenter distribution map. The estimated model resolves 6 fault segments that are consistent with a map of the damaged area. The branches of the fault with 3 segments are found in the NNW and NNE directions with different depth ranges in the northern part. The dip angle of segments is nearly 90 degrees from the surface, which does not extend to the Bam fault completely. Our estimated fault model shows a complicated deformation field especially in the northern parts and also indicates that the Arg-e-Bam fault is distinguished from the Bam fault.
http://www.gaea.kyushu-u.ac.jp/research/iran2004/iran2004.html
S11A-1002 0800h
The Observation of Fault Finiteness and Rapid Velocity Variation in P$_n$$_l$ Waveforms for the Mw 6.5, San Simeon, California Earthquake
We observed the effect of the fault finiteness in the P$_n$$_l$ waveforms from regional distances ($4\deg$ to $12\deg$) for the M$_w$6.5 San Simeon Earthquake on 22 December 2003. We aimed to include more of the high frequencies (2 seconds and longer periods) than the studies that use regional data for focal solutions (5 to 8 seconds and longer periods). We calculated 1-D synthetic seismograms for the $P_n_l$ portion for both a point source, and a finite fault solution. The comparison of the point source and finite fault waveforms with data show that the first several seconds of the point source synthetics have considerably higher amplitude than the data, while finite fault does not have a similar problem. This can be explained by reversely polarized depth phases overlapping with the P waves from the later portion of the fault, and causing smaller amplitudes for the beginning portion of the seismogram. This is clearly a finite fault phenomenon; therefore, can not be explained by point source calculations. Moreover, the point source synthetics, which are calculated with a focal solution from a long period regional inversion, are overestimating the amplitude by three to four times relative to the data amplitude, while finite fault waveforms have the similar amplitudes to the data. Hence, a moment estimation based only on the point source solution of the regional data could have been wrong by half of magnitude. We have also calculated the shifts of synthetics relative to data to fit the seismograms. Our results reveal that the paths from Central California to the south are faster than to the paths to the east and north. The P wave arrival to the TUC station in Arizona is 4 seconds earlier than the predicted Southern California model, while most stations to the east are delayed around 1 second. The observed higher uppermost mantle velocities to the south are consistent with some recent tomographic models. Synthetics generated with these models significantly improves the fits and the timing at most stations. This means that regional waveform data can be used to help locate and establish source complexities for future events.
S11A-1003 0800h
Source Characterization of the 15 November 1994, Ms 7.1 Mindoro, Philippines Earthquake
The 14 November 1994, Mindoro, Philippines earthquake (Ms = 7.1) was a major right lateral strike slip event associated to the movement along zones of weakness transecting the Philippine archipelago along the Aglubang River Fault. We examined this event in order to characterize its tectonic origin using the inversion of teleseismic body waveform data and the method developed by Kikuchi and Kanamori (1990). The earthquake was particularly devastating, generating tsunami that has killed 78 persons, and considered as one of the major events since 1992 around the Pacific Rim. The origin of the tsunami has remained an enigma to date in the scientific community. The digitally recorded broadband teleseismic data was downloaded from the archives of IRIS-DMC and was deconvolved to check the direction of P wave initial motion. Retrieved polarity data was then utilized for the estimation of the fault plane solution using the program FOCMEC_@(Snoke). The result suggests an almost purely strike slip faulting in the NE-SW trending direction which correlates well with the mapped trace of ground rupture. The observed waveforms were complicated and show some aspect of multiple shocks. The focal mechanism solutions from the preliminary inversion of teleseismic body waveforms reveal that the event was dominated by major shear faulting, however, the dislocation source based on the parent earthquake failed to model exceptional run-up amplitude. We performed a careful inspection of the body waveforms for multiple event analysis in search for its spatio-temporal distribution. The result suggests that the 1994 Mindoro earthquake consists of two sub-events and that these events are separated by 9 seconds and are located about 30 km apart from each other. We also infer from the inspection that some later pulses observed in the waveforms could be related to depth phase rather than complexity of the source. Effect of wave propagation in the source region was tested by searching the existence of similar observations in the waveforms convolved for the main shock using a smaller event that has occurred in the vicinity of the source region under study. Finally, we again performed the inversion of all the available teleseismic body waveforms, using the information gathered from the above analysis. We separately analyzed the azimuthal amplitude and phase variations of long-period surface waves for determining the radiation pattern that constrain the source characterization at long periods.
S11A-1004 0800h
Insights on the 1990 Bohol Tsunamigenic Earthquake, Bohol Island, Philippines
The February 8, 1990 earthquake at Bohol area is one of the few strong earthquakes that have affected central Philippines since the early 1900's. This M6.0 1990 Bohol event nonetheless wrought havoc to at least 16 municipalities, caused numerous casualties, injured about three hundred people, rendered several thousand homeless and evacuated from the coastal areas, and damaged at least P154 million worth of properties. The epicenter of this earthquake was initially placed onshore at 17km east of Tagbilaran City and was attributed to the movement along the Alicia Thrust Fault- a fault trending northeast-southwest. Noticeably, there was no surface rupture and the succeeding aftershocks clustered along a northeast-southwest trend off the eastern shore of Bohol island. In addition, the southeastern part of Bohol island experienced tsunami inundation particularly the municipalities of Jagna, Duero, Guindulman, Garcia Hernandez, and Valencia. In this study, several issues were resolved regarding this seismic event. First, the 1990 Bohol earthquake was generated along an offshore thrust fault based on the reviews of seismicity data from the NEIC. -Post-determined plots of the mainshock and aftershocks indicate offshore event with focal mechanism solutions that imply thrust fault activity. Intensity data likewise indicates that intense ground shaking was mainly felt in the southeastern part of the island. Second, recent field investigations undertaken clearly indicated a widespread tsunami inundation wherein the southeastern shorelines of Bohol likewise experienced a regional retreat in sea level several minutes after the strong ground shaking. Lastly, such tsunamigenic structure could somehow explain the anomalously large waves that impacted Camiguin island, an island more than 50km southeast of Bohol. A reconstruction of true tsunami heights and runup distances was also undertaken based from eyewitness accounts. Future works would involve relocation of aftershocks and numerical modeling of tsunami.
S11A-1005 0800h
Measurement of Differential Rupture Durations as Constraints on Source Finiteness of Deep-Focus Tonga Earthquakes
One of the basic issues regarding the physical mechanism for deep earthquakes is the extent to which they occur on pre-existing faults: Do deep earthquakes reactivate faults that formed in the oceanic plate prior to subduction? A previous study of the Tonga subduction zone (Jiao et al., 2000) compared nodal plane distributions of outer-rise events to events as deep as 450 km depth, and found them to be very similar. While this comparison provides some support for the pre-existing fault hypothesis, a more stringent test of the hypothesis is available by identifying the actual rupture planes of the deep-focus events. We are developing the methodology to perform such a test. Between 1990 and 2002, the IRIS FARM database contains approximately 200~earthquakes with $m_b \ge$~5.5 in the region studied by Jiao et al. The Harvard CMT focal mechanism identifies two possible fault planes, and our goal is to distinguish the true fault plane using observations of source finiteness. Source finiteness is observable on seismograms at different azimuths and distances, for unilateral ruptures, as variations in the apparent rupture duration and, for complex ruptures, as differences in the travel-time delay between subevents. For each earthquake, the rupture duration (or travel-time delay) will be shortest in the direction of rupture propagation and longest in the opposite direction. Rather than measuring the actual rupture duration at each station, we use a cross-correlation technique that includes a stretching factor to measure the differential rupture duration between each pair of stations. These differential measurements then allow us to identify the rupture direction and fault plane for each earthquake. A comparison of the distribution of rupture-plane orientations of deep-focus and outer-rise seismicity allows us to determine whether or not deep seismicity occurs on reactivated faults.
S11A-1006 0800h
Analysis of the M=7.2 1992 Cape Mendocino Earthquake with Kinematic Source Models and Empirical Green's Functions
The M=7.2 Cape Mendocino earthquake occurred on April 25, 1992 and is potentially the first subduction zone earthquake recorded in the Mendocino Triple Junction region. We utilize recordings of weak motion events as empirical Green's functions to constrain the propagation path effects of the earthquakes in forward kinematic modeling. Using constraints based on the 1992 event, we create kinematic source models based upon physical parameters of fault rupture and generate a suite of synthetic accelerograms to identify potential fault rupture characteristics of the actual earthquake. The precise fault for the 1992 earthquake is not known, so we confine the calculations to a small source volume where the earthquake likely occurred and run 100 scenario earthquakes. This also provides a test of a physically-based methodology to predict the range of ground motion that may occur from a particular magnitude earthquake along a specific fault or within a specific source volume. We test whether the actual ground motion recordings fall within the range predicted. We also test the hypothesis that some of the synthesized records will match those observed, and we make the assumption that the source models for these records are close to what actually occurred. This work was performed under the auspices of the U.S. Department of Energy by the University of California, Lawrence Livermore National Laboratory under contract No. W-7405-Eng-48.
S11A-1007 0800h
Paritioned inversion of seismic and geodetic data for high-resolution source inversions
Due to the availability of earthquake data spanning a very broad range of frequencies and spatial scales (e.g. geodetic, strong-motion, teleseismic data) we now have the potential of obtaining highly detailed views of the earthquake rupture process. The sheer quantity of the data, and the heterogeneity, do pose problems for simultaneous inversion procedures both in terms of scale of the problem as well as how much influence different datasets have over the final result. A simple solution to the problem outlined above, taking advantage of the aforementioned richness in data density and variability, is to divide the inversion procedure into two steps. The first step involves inverting for the slip distribution, which would then be used as a constraint in the second step where we invert for the kinematic parameters. The advantage of this approach is that we now have a-priori knowledge of the slip model before the second inversion that can help reduce the number of free parameters significantly. Practically speaking, we can remove the grid points that have negligible slip, and we can tailor the maximum rise time for every subfault to the total amount of slip on that subfault. It would also be feasible to reduce the length of the timesteps, which gives us an improved resolution of the slip history and allows us to search for a larger range of rupture velocities. Such a method makes more optimal use of the data than a simultaneous inversion. We will present results for some large recent earthquakes to demonstrate the feasibility of this approach, and discuss the applicability of this method to more dynamically oriented inversions.
S11A-1008 0800h
3-D Numerical Investigation of the Tsaoling Landslide Induced by Chi-Chi Earthquake, Taiwan.
Large landslides occurred in the mountainous area near the epicenter of the Sept. 21st, 1999, Chi-Chi earthquake in central Taiwan. These landslides were triggered by the Mw = 7.6 earthquake, which resulted in more than 2,400 human casualties and widespread damage. The 1999 Chi-Chi earthquake triggered a catastrophic Tsaloing landslide, which mobilized about 0.125 km3 of rock and soil that slid across the Chingshui River and created a 5 km long natural dam. One fifth of the landslide mass dropped into the Chingshui River, the rest jumped over the river. At least five large landslides occurred in Tsaoling area are induced by big earthquake and heavy rainfalls since 1862 to 1999. Geological investigation shows that the prevailing attitude of sedimentary formation is about N45W with a dipping angle of 12S. First we used Remark Method to calculate the stability of slope. The bottom of slope has been eroded by Chingshui stream, and the PGA (Peak Ground Acceleration) in Chi-Chi earthquake was exceeded the yield acceleration along the sliding surface. The landslide mechanism may be including flowing, rolling, bouncing and sliding. The rock on the fault plane during faulting can generate pseudotachylyte resulted from melted rock by frictional heat energy along the sliding surface. The frictional melted rocks were found out in the Chiu-Fen-Erh-Shan collapses. However, we didn't found out the frictional melted rock in Tsaoling area. If we calculated the kinetic energy which was converted to heat energy, the increase of temperature was enough to melt the rocks on sliding surface. When the rocks on the sliding surface had been melted, the friction on the sliding surface must be decrease. Therefore, the 0.125 km3 debris had sufficient kinetic energy to across Chingshui River to the other side of the river. Using 3D distinct-element modeling (PFC3d code), we try to simulate kinematic process of Tsaoling landslide. Our numerical model was compose of about 10,000 spherical elastic particles that were bonded together to create an initial solid numerical rock mass, We assumed the frictional coefficient near to zero, thus the 0.125 km3 debris can move over the Chingshui stream. If the particles didn't bond together, the debris would dissipate in the Chingshui stream valley. The frictional coefficient of the ball were set over 3 degree, the debris wouldn't across the Chingshui stream valley. We conclude that the characteristics of Tsaoling landslide process are: (1) the rock were bond together on sliding, and (2) the frictional coefficient was near to zero.
S11A-1009 0800h
Atmospheric Pressure Oscillations Forced by Surface Waves From the 2003 Tokachi-Oki Earthquake
Clear atmospheric pressure changes associated with the 2003 Tokachi-Oki Earthquake with M 8.3 were recorded by 8 microbarographs along Japan. The maximum oscillatory pressure change is about 2 Pascal with dominant period is about 15-20 second, and lasted for more than 30 minutes. Comparing the pressure change with broadband seismic records observed near or at the microbarogram, the pressure change starts at the arrival of seismic waves and reaches its maximum amplitude at the arrival of Rayleigh waves. Four microbarographs, co-located with STS-1 broadband seismographs and suffering less atmospheric wind noise, show that peaks in vertical ground velocity records correspond to the peaks of atmospheric pressure records. Similar pressure changes were observed during the largest aftershock (M 7.4). All ground motion analyzed in this paper were recorded by STS-1 broadband sensors. Spectrum analysis in the frequency domain supports that the vertical ground velocity and the pressure change has the same phase and the amplitude ratio is constant up to a period of about 50 second. The constant amplitude ratio is about (atmospheric density) times (sound velocity in the atmosphere), indicating that the surface ground in vertical ground motion compresses or inflates the air above the ground locally and low-frequency sound waves are generated. Pressure change recorded after the passage of Rayleigh waves does not well correlate with the ground velocity. Through the precise atmospheric pressure and ground motion measurement at the same sites, we witnessed the process of low-frequency sound generation by the vertical ground surface motion acted as a vibrating plate of a speaker. The radiated low-frequency sound waves propagates upward and reaches to the ionosphere with large amplitude because of the energy conservation. The ionospheric turbulence reported in the past researches were originated from this low-frequency sound at the ground surface.
S11A-1010 0800h
Large earthquakes along the Nankai trough, the 1944 Tonankai earthquake (Mw 8.0) and two 2004 Tonankai-oki earthquakes (Mw 7.2 and 7.4)
Great interplate earthquakes have repeatedly occurred along the Nankai trough with an interval of about 120 years. The most recent events were the 1944 Tonankai and 1946 Nankai earthquakes. Recently, two large compressional outer-rise earthquakes occurred off the source region of the 1944 Tonankai earthquake near the Nankai trough on Sept. 5, 2004 (GMT). First, the detailed slip distribution of the 1944 Tonankai earthquake is estimated using the tsunami waveform inversion in this paper. Second, the tsunamis from two recent outer-rise earthquakes are analyzed to discuss the relation between the great interplate event and the large outer-rise events. The slip distribution of the 1944 Tonankai earthquake has been already estimated using the tsunami waveform inversion by several researchers. The slip estimation in this paper is improved by using smaller size of subfaults, 25 km x 25 km, and larger number of subfaults, 72 subfaults. We also use smoothness constraints for the discrete Laplacian to stabilize the inversion. Appropriate weights for the smoothness constraints are determined using the Akaike Bayesian information criterion. The results show that the largest slip of 3.7m is occured at the subfault off Shima peninsula. The large slip region, $>$1m, are elongated to the southwest-northeast direction from the subfault where the largest slip is estimated. The total seismic moment is 1.3 x 10$^{20}$ Nm (Mw 8.0) by assuming that the rigidity is 5 x 10$^{10}$ N/m$^{2}$. The two 2004 Tokachi-oki outer-rise earthquakes generated the moderate tsunamis. Those tsunamis were observed at several tide gauge stations and ocean bottom pressure gauges. The tsunami heights of the second event are about twice as large as those of the first event. This is consistent with the magnitudes of two events because the Harvard CMT moment magnitudes of the first and second events are 7.2 and 7.4, respectively.
S11A-1011 0800h
Seismic Signals Associated with Landslides and with Tsunami at Stromboli Volcano, Italy
The seismic signals produced by two landslides that occurred at Stromboli volcano on December 30, 2002, recorded by broadband seismic stations, have been analyzed. For both landslides, the characteristics of the low-frequency seismograms indicate a complex time-history in the release of seismic energy. The first landslide occurred over the submerged part of the NW sector of the volcano, and had associated a large-amplitude, low-frequency pulse representative of the abrupt detachment of a large mass. The highest amplitude low frequency signals are well described by a single-force source model. The second mass-failure episode is characterized by a complex source and can be interpreted as a multiple event, with less abrupt onset and at least four detachments occurring during 4-5 minutes. Synthetic seismograms generated by a shallow single force located in the submerged area of Sciara del Fuoco and directed upslope, fit well the first low frequency seismic pulse recorded at Stromboli and Panarea. From this simulation we estimated the force exerted by the first mass failure. The estimate of the volume through two different procedures, gives values in the range 1.0 - 1.5 million cubic meters and about 14 million cubic meters respectively. The landslides produced a tsunami that struck the coast of Stromboli island and reached in a few minutes the other islands of the Aeolian archipelago. Three broadband seismic stations installed on land about hundred meters from the coastline at Panarea island, located 20 km SW of Stromboli, recorded very long period seismic signals produced by the tsunami waves. Synthetic tsunami waves, obtained by a landslide source model and taking into account the bathymetry of the sea surrounding Stromboli and Panarea islands, fit well the observed phenomena and the experimental data. The ground tilt estimated by the broad band recordings is in good agreement with the expected value, given the distance of the station from the sea and the tsunami wave height.
S11A-1012 0800h
How Heterogeneous Stress Biases the Orientations of Focal Mechanisms
We explore the effect of spatially heterogeneous stress on standard focal mechanism inversions. We ask the following questions: Does the presence of spatially heterogeneous stress, coupled with a stress rate of a particular orientation, bias which points fail as earthquakes? If there is a bias to which points fail and are consequently included in focal mechanism inversions, then will inversions, as they are currently being interpreted, primarily yield the orientation of the stress rate tensor (due to far-field plate motion) regardless of the other components in the stress field? We numerically address these questions by generating synthetic stress fields, bringing points to failure as a function of time under a plastic yield constraint, using the failed points to produce focal mechanisms/earthquake catalogs, and inverting the synthetic catalogs with standard inversion techniques [ Michael, 1984; Michael, 1987]. The total deviatoric stress field is composed of three terms, a spatially and temporally uniform background stress, a spatially uniform secular stress that increases linearly as a function of time and has an orientation different from that of the background stress, and a temporally uniform but spatially heterogeneous stress that is the summed effect of all past earthquakes, modeled statistically. If the heterogeneous stress has no preferred mean orientation and the added secular stress is much smaller than the background stress then one might assume under the current paradigms that the average orientation of the synthetic focal mechanisms to closely align with the orientation of the background stress. Interestingly, this is not what we find. We what we do find is the following: Standard focal mechanism inversions [Michael, 1984; Michael, 1987] accurately reproduce the average orientation of the stress tensors associated with the points which fail. However, in the presence of significant heterogeneous stress, which points fail and produce focal mechanisms are biased toward the orientation of the secular stress rate. We found that the ratio of the standard deviation of the heterogeneous stress to the amplitude of the background stress determines the orientation of the failure focal mechanisms. When the ratio is very small, i.e. the heterogeneity is much smaller than the spatially uniform background stress, the focal mechanisms on average (with very little orientation heterogeneity) are aligned with the background stress. When the ratio is very large, i.e. the heterogeneity is much larger than the spatially uniform background component, the focal mechanisms on average (which significant orientation heterogeneity) are aligned with the secular stress rate. Real earthquake data often displays significant focal mechanism heterogeneity suggesting that the Earth may have a large component of heterogeneous stress that biases the orientations of the focal mechanisms. This suggests that studies based on focal mechanism inversions may need to be reinterpreted. By comparing our simulation results to real data we have begun developing tools to reinterpret focal mechanism data in the context of heterogeneous stress. It is our hypothesis that in the end we may be able to estimate both the spatially uniform background stress component and the degree of stress heterogeneity.