S31A-1017 0800h
A New Model for Earthquake Occurrence World-Wide and its Implications for Seismic Hazard Assessment
We have developed a new model for the time interval between earthquakes greater than a given magnitude. The model allows us to calculate quantities of interest in seismic hazard assessment including the probability distribution of times to the next (large) earthquake in a region given the time since the previous one. We have established a database of earthquakes world-wide that is effectively complete above magnitude 5. The database and a comprehensive model for aftershocks are described elsewhere at this meeting. We use an extension of the aftershock model to define a circular region in which earthquakes are assumed to be causally related. The size of this region scales with magnitude: log Area \propto$ M. From these regions we assemble the distribution of time intervals between earthquakes with magnitudes $>$ some reference magnitude M$_{r}$ e.g. M 7. If earthquakes were independent, the distribution would be a negative exponential (Poisson process) with time constant equal to the mean time between events $>$ M$_{r}$. The short time intervals are well described by Omori's Law 1/t$^{p}$, with p=1.0. However we find that the whole distribution is well modeled as a mixture: a 1/t power law modulated by a negative exponential plus a negative exponential. These two parts may be thought of as comprising the 'aftershocks' and the 'background'. The model has two parameters to be determined: the weights of the two parts (which must sum to 1) and the time constant for the negative exponentials, assumed the same for both parts. To fit the model a correction must be applied for the finite duration of the catalogue, as time intervals longer than this cannot be observed and the frequencies of intervals that are a large fraction of the catalogue duration are biased downwards. We find that the time constant is several years, larger than the mean time between events. An implication of this is that 'aftershocks' persist as a significant contribution to earthquake occurrence for much longer - many years - than is often considered. This study was (part) funded by the EQC Research Foundation.
S31A-1018 0800h
Assessment of Seismic Hazard for the State of Kuwait
Kuwait considered quiet seismically since it characterized by low level of seismic activities by the world standard. In the last three decades, several earthquakes with magnitude (ML) near and above 4 have occurred in Kuwait. In addition, the Kuwait National Seismic Network (KNSN) recorded several low magnitude earthquakes mostly clustered within the vicinity of the oil fields since March 1997. This clustering within the oil fields indicates a possible association of seismicity with oil production in Kuwait. The seismic hazard map for 10 % probabilities of exceedance in a 50 year period shows that Kuwait has maximum PGA value 10 gal, while the seismic hazard map for 2 % probabilities of exceedance in a 50 year shows that the maximum PGA value 20 gal. The potential sources of hazard in Kuwait is the southwest near by the Minagish, Umm Qudair fields and externally from Zagros Mountain.
S31A-1019 0800h
A new Proposal to Mexico Valley Zonification
The effects of the Michoacan earthquake (19th September, 1985, Mw 8.1) in Mexico City caused a significant change in the political, social and scientific history, as it was considered the worst seismic disaster ever lived in Mexico. Since then, numerous efforts have been made to understand and determine the parameters that caused the special features registered. One of these efforts had began on 1960 with the work by Marsal and Masari, who published the Mexico Valley seismological and geotechnical zonification (1969), based on gravimetric and shallow borehole data. In this work, we present a revision of the studies that proposed the zonification, a description of the valley geology, and basing on it we propose a new zonification for Mexico Valley.
S31A-1020 0800h
Evidence for Repeated Liquefaction-Induced Lateral Spreading along the Lower Pajaro River, Watsonville, California
This investigation, designed to evaluate whether lateral spreads occur repeatedly in the same location, documents evidence of recurring liquefaction-induced sand injection and lateral spreading along a stratigraphic unconformity within the Pajaro River floodplain near Watsonville, California. We excavated two trenches across a lateral spread formed by the 1989 Loma Prieta earthquake and located south of the Pajaro River in floodplain sediments at the Miller Farms site that was originally identified by the US Geological Survey during post-earthquake studies. In addition to liquefaction-related features produced in 1989, the trench walls revealed evidence for at least two to three prior lateral spread failures and associated liquefied sand bodies. The site likely records evidence for failure from the 1906 M 7.8 San Francisco earthquake and earlier events on the San Andreas fault. The spreading repeatedly occurred along a one-meter-wide zone that coincides with a buttress unconformity between middle to late Holocene floodplain deposits (south of the unconformity) and late Holocene to historic fluvial deposits of an aggraded inset river terrace (north of the unconformity) of the Pajaro River. Trench walls exposed a secondary zone of discontinuous normal faults with small ($<$ 1 cm) vertical displacements, located toward the river and several meters north of the primary lateral spread zone. Minor normal faults generally coincide with ground cracks caused by the 1989 earthquake, although it is permissible that an earlier lateral spread produced some of the faults. The minor displacements on faults in the secondary zone relative to the massive failure along the primary lateral spread zone indicates that the primary mode of deformation at the Miller Farms site has been repeated failure localized along a buttress unconformity. Detrital charcoal collected from within, and above, a structurally tilted sand layer suggests that the ante-penultimate event happened after A.D. 1400. A single trench excavated across the distal alluvial fan of Coyote Creek near Milpitas, California, showed no evidence for lateral spread failures, despite reports of widespread lateral spreading in the vicinity during the 1906 San Francisco and the 1868 Hayward earthquakes. Two narrow sand dikes exposed in the trench walls indicate that the natural levee deposits have liquefied during past events, supporting prior site-specific and regional interpretations that underlying deposits at the site have a high susceptibility for liquefaction.
S31A-1021 0800h
Newly Recognized Fault Structures and Earthquake Hazards in the Western Santa Clara Valley, San Francisco Bay Area, California
A combined reflection and refraction imaging investigation from the San Andreas Fault to the central Santa Clara Valley, California shows the complex structures and tectonics of the Santa Cruz Mountains and the western Santa Clara Valley. Our seismic images show numerous faults within the Santa Cruz Mountains, consistent with surface geologic mapping. Because of the high velocities (> 6.4 km/s) in the upper 5 km, we interpret steeply northeast-dipping reflectors within basement in the Santa Cruz Mountains to be part of the Coast Ranges Ophiolitic rocks. Within the Santa Clara Valley, offset reflectors and folded Pre-Pliocene strata suggest multiple faults beneath the alluvium of the western Santa Clara Valley, and seismicity recorded over the past 30 years show that these faults are active. Some of these faults extend to the shallowest reflectors imaged (~20 m), including the base of the ground-water aquifer system, suggesting at least post-Pliocene movement on faults at shallow depths. The seismic image suggests that the principal fault zone beneath the western Santa Clara Valley is about 6 km wide in the uppermost crust but tapers to about 4 km in width at 5 km depth, similar to known flower structures elsewhere. Basement velocities are low (4 to 5 km/s) beneath this western Santa Clara fault zone relative to those velocities (~6 km/s) along the remainder of the seismic profile, an observation seen in a nearby seismic profile across the San Andreas fault zone. The high-population density within the valley and the proximity of these buried faults makes the earthquake hazard in the western Santa Clara Valley potentially high. This potential hazard is compounded by the thick sequence (~1.6 km) of low-velocity (< 4 km/s) sediments within the Cupertino Basin that can amplify seismic waves generated by movement on regional and/or local faults.
S31A-1022 0800h
Seismic-Reflection Profiling Across the Urban Area of Santa Clara Valley, California: Images of the Northeastern Margin of the Cupertino Basin
An 8-km-long, P-wave, vibroseis seismic-reflection profile provides new information regarding the Miocene to Pliocene-Quaternary (P-Q) sedimentary structure in the Cupertino Basin on the southwestern side of the Santa Clara Valley, California. We acquired the profile using as many as 177 channels with 10-m receiver and source intervals. A prominent southward dipping reflection package, believed to be the Miocene Monterey Fm. (MF), unconformably underlies 500-600 m of generally flat, well-layered P-Q section. The unconformity, which is only clear along the southern 2 km of the profile, has as much as 40 m of relief over lateral distances of about 250 m. Within the MF a wedge of clear reflections thicken to the south of Creekside Park in Cupertino, with the wedge base approaching 1.2 km depth below Rainbow Dr. These relationships are a striking confirmation of the previously proposed buried Cupertino Basin filled largely with MF. Wedge reflections onlap to the north against a strong reflection that dips about 20 degrees south. This south-dipping reflector separates a highly reflective zone above from a relatively transparent zone below and to the north, although there are hints of south-dipping reflections at 2 km depth below Rainbow Dr. The profile provides no evidence for fault offset greater than 20 m within the 300-m depth range of water wells from which the Cascade and Santa Clara Faults were proposed 30 years ago. There is evidence, however, of a south-dipping, south-side-up fault near the north end of the profile. This fault terminates at a depth of about 550 m, above which strata are slightly warped up to a depth of about 300 m. Faults with smaller offsets may be present but are not detected because of the 8 to 10 m vertical resolution limit of the data. The seismic profile crossed five busy highways, severely reducing the subsurface coverage, thus possibly missing some faults. However, prominent reflectors on either side of these roads are at similar depths, which indicates that faults having vertical offsets greater than 20 m are not likely beneath the roads.
S31A-1023 0800h
San Francisco Bay Area Velocity Structure From Controlled-Source Seismic Refraction Imaging
To better understand the velocities and structures of the crust and upper mantle in the San Francisco Bay area, we developed 2-D tomographic velocity models along four seismic refraction profiles acquired along and across the bay area in the early 1990's. The four profiles extended from (1) Hollister to Inverness along the San Francisco and Marin Peninsulas (~200 km long), (2) Hollister to Santa Rosa along the East Bay (~220 km long), (3) the Pacific Ocean to Livermore crossing the bay (~100 km long), and (4) the Pacific Ocean to the western Santa Clara Valley (~25 km long), centered on the epicenter of the1989 M. 6.9 Loma Prieta earthquake. Velocity models were not previously developed for three of the seismic profiles, and the previously developed model for the fourth profile (Catchings and Kohler, 1996) did not include some of the currently available seismic data. The profiles along the bay image structures from the near surface to about 25 km depth, and they show velocity anomalies associated with the major faults (San Andreas, Hayward, Rodgers Creek, Calaveras) and basins along the profile. Velocities range from about 2 km/s in the basins to about 7 km/s at the Moho, which dips southward along both sides of the bay. The cross bay profile shows velocity anomalies associated with six fault zones between the Pacific Ocean and the Livermore Valley and higher upper-crustal velocities (~6.2 km/s) between the San Andreas and Hayward faults than to the southwest (~5 km/s) or northeast (~4 km/s) of those faults. The Loma Prieta profile shows velocities ranging from 2 km/s to 6 km/s in the upper 5 km, with the highest velocities in the epicentral region of the 1989 Loma Prieta earthquake. A pronounced, northeast-dipping, low-velocity zone is located beneath the surface expression of the San Andreas fault zone, but other fault zones along the profile show high-velocity anomalies beneath their surface expressions. Collectively, the velocity images show the complexity of the crustal structure and provide constraints for 3-D tomographic models developed for the San Francisco Bay area.
S31A-1024 0800h
Subsurface Study of Faults Near Point Arena and the San Andreas Fault Zone, Northern California, Using High-Resolution Reflection and Refraction Profiling
We acquired seismic reflection and refraction data along a 600-m-long profile located about 6 km southwest of the San Andreas fault (SAF), near Point Arena, California to determine possible relations between faulting and synclinal folding in the Miocene Point Arena Formation. Seismic sources were generated by a combination of 118 BETSY Seisgun and sledge-hammer shots. Seisgun shot points were spaced every 50 m, and sledge-hammer shot points were spaced every 5 m between the seisgun shots. Data were recorded in a linear array of two Geometrics Strataview RX 60 seismographs, with each seismograph having 60 active channels. Geophones were spaced at 5-m intervals and were laterally offset 1 m from the shot points. We recorded 2 s of data at a sampling rate of 0.5 ms. The seismic profile crossed two known faults and a prominent, asymmetric syncline that trends northwest-southeast, subparallel to the SAF. The two faults extend into unconformable, late Quaternary marine terrace deposits and may still be active. The syncline has gently dipping beds (13 to 22 degrees) on the southwest side, and strata dip on the northeast side 25 to 56 degrees. Because the seismic profile was oriented NE-SW, approximately normal to the syncline axis and fault trend, we observe similar features on the seismic images. Seismic velocities range from about 800 m/s near the surface to about 2200 m/s at a depth of 40 m. We observe two low-velocity gradients along the profile that we interpret to be the northwest-southeast-trending syncline and a thrust fault, both of which have linear trends visible in aerial photographs. The seismic reflection image also reveals sub-horizontal reflectors apparently offset by faulting and the prominent syncline. The seismic reflection image shows at least four faults, two more than previously mapped. These faults, and their recency of activity, represent a significant change in understanding of the Quaternary tectonic setting of the region.
S31A-1025 0800h
High-Resolution Seismic Reflection and Refraction Imaging of a Thrust Fault Near Point Arena, Northern California
We acquired high-resolution seismic reflection and refraction data along a 240-m-long profile located about 6 km southwest of the San Andreas fault (SAF), near Point Arena, California. The study was conducted to determine structural relations of a thrust fault, which has natural exposures in a sink hole and coastal bluffs, that moved strata of the Miocene Point Arena Formation over uplifted late Quaternary marine terrace deposits. Seismic sources were generated by 120 sledge-hammer `shots'. Shot points and geophones were spaced every 2 m, and geophones were laterally offset 1 m from the shot points. Data were recorded in a linear array of two Geometrics Strataview RX-60 seismographs; we recorded 2 s of data at a sampling rate of 0.5 ms. The seismic profile crossed the known thrust fault within 15 m of exposures in a sink hole and a bluff. Seismic P-wave velocities range from about 500 m/s near the surface to about 1500 m/s at 10 m depth. Below 10 m velocities are laterally highly variable. Near the northwest end of the profile, maximum velocities are about 2700 m/s at 25 m depth; at about 80 m to the southeast and 20 m depth, velocities are 1700 m/s, increasing to 1900 m/s at 15 m depth near the southeast end of the profile. A migrated seismic reflection image shows the source of the thrust fault to be in subhorizontal bedding, which matches relations seen in the natural bluff exposures. The reflection image also shows a prominent, steeply dipping fault near the middle of the profile, the location of which coincides with the low-velocity zone in the deeper section of the velocity profile. The steeply dipping fault near the middle of the profile does not appear to offset reflectors in the uppermost, late Quaternary terrace deposit section of the reflection image. Fault relations revealed in the seismic sections further our understanding of the three-dimensional fault structure west of the San Andreas fault.
S31A-1026 0800h
Structure of the San Bernardino Basin Between the San Jacinto and San Andreas Faults
We acquired high- and medium-resolution seismic reflection and refraction data along two profiles through the San Bernardino Basin. The high-resolution profile (5 m shot and geophone spacing) was acquired along the I-215 freeway from I-10 to State Road 30. The medium-resolution profile (1 km and 50 m shot and geophone spacing, respectively) was acquired northeastward from southern Colton to the San Andreas fault (SAF). Images from both profiles show that the San Bernardino Basin has a maximum depth of 1.2 km northeast of the I-215/I-10 interchange. From southwest to northeast, the top of basement steps down about 800 m between surface traces of the Rialto-Colton and San Jacinto fault zones (RCFZ and SJFZ, respectively); basement rises in steps northeastward along faults between the interstate interchange and the SAF. We interpret three zones of faulting between the RCFZ and the SAF, including, the SJFZ, the Tippecanoe fault zone (TFZ), and the East Highlands fault zone (EHFZ). Double-difference-relocated earthquakes (Hauksson et al., 2003) illuminate all of these fault zones. The RCFZ and the SJFZ form a flower structure that is ~8 km wide in the upper crust and is centered near the I-215/I-10 interchange. The TFZ is ~3 km wide, has dominantly near-vertical fault strands, and is centered near downtown San Bernardino. Over the past 20 years, the TFZ has been the most seismically active fault zone in this part of the San Bernardino Basin. The EHFZ appears to be centered on, and probably is structurally related to, a series of hills exposed above alluvium near the northeast edge of the basin. The EHFZ forms an 8-km-wide flower structure where it crosses our medium-resolution profile. Because our medium-resolution profile extended only a few hundred meters northeast of the SAF, we did not image major strands of the SAF, suggesting that the SAF may dip to the northeast. Sediment P-wave velocities within the San Bernardino Basin range from about 800 m/s at the surface to about 3700 m/s at a depth of 1.2 km. Basement velocities are highly variable, ranging from about 4000 m/s at the sediment-basement contact to about 6400 m/s at 4 km depth. The thickness of the San Bernardino Basin, the relatively low sediment velocities, and the presence of active faults within the basin all suggest that even moderate local earthquakes may result in appreciable losses.
S31A-1027 0800h
Preliminary Late Pleistocene Slip Rate of the Green Valley Fault at Lopes Ranch Creek, Cordelia, California
The Concord-Green Valley fault system is part of the eastern San Andreas fault system, and is composed of two major fault segments from south to north: Concord fault and Green Valley fault (GVF). The GVF is subdivided into a southern segment that extends from the northern shores of Suisun Bay to near Cordelia, and a northern segment that continues north to Wooden Valley, east of Napa, CA. At the Lopes Ranch Creek site, along the southern GVF, an ephemeral creek preserves the cumulative dextral separation of an abandoned north-trending paleochannel located east of the main fault. Trenches excavated at the site expose a sequence of latest Pleistocene to historic alluvial and fluvial deposits overlying weathered bedrock of the Upper Cretaceous to Upper Jurassic Great Valley Sequence. The main active trace of the southern GVF is well constrained based on tectonic geomorphology (e.g., NE-facing scarps, vegetation lineaments, and springs), and trenches that exhibit a 3-meter-wide fault zone containing vertical faults, west-dipping dextral reverse faults and creep-related fractures. Preliminary estimates of cumulative right-lateral displacement of a distinct paleochannel deposit range from 31 to 58 meters. Radiocarbon analyses of charcoal collected from a burn horizon directly above the paleochannel provide a minimum age of 14,080 to 15,380 cal yr B.P. Based on the estimated cumulative displacement and minimum age of the paleochannel deposit, a preliminary long-term slip rate for the southern GVF is 2 to 4 mm/yr. This preliminary slip rate is close to the historical slip and creep rates for the GVF. For instance, a previous slip rate study of an offset late Holocene paleochannel deposit about 0.4 km northwest of the Lopes Ranch Creek site determined a slip rate of 3.8 to 4.8 mm/yr (over the last 300 years), consistent, in part, with the latest Pleistocene slip rate yielded by this study. These geologic slip rates also are close to the 14.7-year average creep rate of 4.4 $\pm$ 0.1 mm/yr for the southern GVF, and the geologic slip rate of the Concord fault (3.4 $\pm$ 0.3 mm/yr). Because of the: (1) limited exposures of the paleochannel deposit directly west of the fault zone, (2) limited radiocarbon ages, and (3) preliminary nature of the investigation at Lopes Ranch Creek, additional studies are necessary to refine the long-term slip rate for the GVF.
S31A-1028 0800h
Simulating Ground Motions from Geodetic Data for ShakeMaps
Over the past several years, we have developed an automated finite-source analysis procedure making use of data recorded by regional distance broadband stations. The method determines the best fault plane by testing the two possible nodal planes of the regional distance moment tensor. Both line-source and plane-source inversions are performed, and the source parameters from these inversions are used to characterize rupture finiteness and directivity. Near-fault ground motions obtained by integrating the derived slip distribution with near-fault Green's functions can be used to augment ShakeMap. For example, source finiteness information significantly improved the initial ShakeMaps of the 2003 Mw6.5 San Simeon, California, earthquake. Our present work has two primary thrusts: 1) development of a method for the near-realtime inversion of GPS data to independently determine finite-fault geometry and orientation, and slip distribution, and 2) investigation of methods to simulate high-frequency ground motions from the geodetic slip models. In this study, we will present a method for converting slip models obtained from GPS data into kinematic models whose rupture process is governed by the rupture and slip velocities. Preliminary results show that simply assuming a rupture-to-shear velocity ratio of 0.8 and a slip velocity derived from a constant stress drop model performs well. We will demonstrate the approach for the 1994 Northridge earthquake by simulating motions using the Wald et al. (1996) kinematic model, a uniform slip model, and the geodetic slip model of Hudnut et al. (1996). The simulated motions for the geodetic model will be compared to both the kinematic model reference and the data in both the time domain and the spectral acceleration domain. We will also compare the simulations in terms of peak ground velocity ShakeMaps. Finally the results will be characterized in terms of the uncertainty due to the unknown rupture velocity and stress drop.
S31A-1029 0800h
A Hybrid Method for the Generation of Broadband Ground Motions
A hybrid method for simulating broadband ground motions is presented. The low frequency ground motions are generated deterministically using the 3D finite-element method whereas the high frequency motions are obtained stochastically. The technique for calculating the high frequency components involves the methodology by Irikura (1983) and Kamae et al. (1996) for superposition of stochastic Green's functions for small events over sub-faults. One major modification is made in order to include a more realistic frequency-dependent radiation pattern. We incorporated the theoretical double-couple full-space radiation pattern for intermediate frequencies and used the average radiation factor proposed by Boore for the higher frequencies. In order to be able to better represent the near field ground motions, the corrections for the amplification due to surface layers and site-dependent factors are also applied in the technique. Finally, the low and high frequency simulations are combined via a matched pair of filters eliminating the higher and lower frequency components from the motions obtained with the deterministic and stochastic methods, respectively. The proposed hybrid method is tested for the 1994 Northridge earthquake. The results are presented in the form of ground motion simulations in the Sherman Oaks region and compared with observed data.
S31A-1030 0800h
Computation of Strong Ground Motions using Modified Empirical Green's Functions
We simulated strong ground motions using empirical Green's functions (EGF) containing information of amplification and attenuation of a specific site. The EGF is the seismic data observed from small earthquakes, such as foreshocks or aftershocks. The main shock can be predicted with EGFs using the finite fault model. The conventional method based on EGF did not consider locations of subfaults. Same EGF has been applied to all subfaults on the fault plane. Green's functions computed for shallow sources and deep sources, however, are not similar to each other. So, it is need to modify the EGF considering the depth of subfaults and epicentral distances. For modifying the EGF, we computed the theoretical Green's function using the wavenumber integration method in layered medium, and deconvolved the EGF with the TGF for each subfault. In order to compute the strong ground motions, we used the finite fault model with 2-asperity based on Somerville et al. (1999). This model was produced considering some empirical relations for large earthquakes. The slip contrast between asperity and other area of the fault plane was fixed to 2. We used the velocity structure proposed by Chang and Baag (2004) in Korea. After computing the TGFs using the wavenumber integration method and constructing the TGF database around the source area in advance, we simulated strong ground motions using this TGF database. This method was applied to the large earthquakes of the world for the verification. The ground motions at seismic stations in southeastern Korea for Mw 7 were properly simulated using modified empirical Green's function method in layered medium. The proposed method and fault model for estimation of site dependent seismic parameters and ground motions could be efficiently used in the low and moderate seismicity regions.
S31A-1031 0800h
Some Primary Observation from Earthquake Scenario Simulation Using Numerical Green Functions
Deterministic earthquake scenario simulation methods are playing an increasingly important role in seismic hazard and risk estimation. The numerical calculation of the complete wavefield in the observed frequency band for a seismically active basin remains a computationally expensive task for some time. Our aim is to provide a tool with which we can calculate a large number of difference finite source scenarios for a particular fault or fault system. In order to avoid having to calculate an individual scenario for each kinematic source description we propose the concept of "numerical Greens functions" (NGFs). The basic idea is that: a large seismic fault is divided into subfaults of appropriate size which will be found out with a trial and fail method. For those subfaults synthetic seismograms for the whole surface of this seismic active area are calculated and stored for the regional velocity model. Within some limitations, arbitrary kinematic sources can be simulated for the whole fault or parts of it by superposition of this set of separate seismograms obtained by the stored subfault excitations. In this work we test the potential of the NGF approach to simulate the seismicity and the associated seismic hazard for a given region and earthquake statistics
S31A-1032 0800h
Application of the 'Recipe for Strong Ground Motion Evaluation' to the 2003 Tokachi-oki, Japan, Earthquake
The National Research Institute for Earth Science and Disaster Prevention has carried on the special research project 'National Seismic Hazard Mapping Project of Japan'. In this project, strong ground motion evaluations for scenario earthquakes have been carried out based on the 'Recipe for strong ground motion evaluation' proposed by Irikura et al. (2004). Here we simulate strong ground motions during the 2003 Tokachi-oki earthquake (Mw=8.0) by applying the 'Recipe', and examine its validity. A simplified source model consists of some rectangular-shaped asperities on a fault plane is constructed following the method in the 'Recipe'. A 3D velocity structure model in and around Hokkaido, which is the target area for the simulation, is also constructed based on boreholes, reflection and refraction surveys, and geological data. Waveforms are calculated by the hybrid Green's function method (Kamae et al., 1998). We use the 3D finite-difference method and the stochastic Green's function method for low ($<$ 0.2 Hz) and high frequency content, respectively. As an overall feature, the peak ground velocity and seismic intensity distributions of simulated strong ground motions agree well with the observed ones. However, some problems become clear. The first is the underestimations in the frequency range of 0.2-1 Hz at sites located on deep basins. This may be a lack in the evaluation on the effects of 3D structure in the calculations using the stochastic Green's function. The second is the overestimations in high frequency range ($>$ 2 Hz) at sites located on back-arc side. This may be caused by the existence of low-Q zone beneath the volcanic front, which is not considered in this simulation. In addition, the non-linear site effects also should be considered in strong motion evaluations, especially for soft soil sites near source region.
S31A-1033 0800h
PS Converted-Wave Reflection Survey in Southern Osaka Plain, Japan
Osaka plain is one of the most populated areas in Japan. The velocity structure models have been created for the simulation of strong ground motion generated by expected earthquakes. But in southern Osaka plain, the information for the basement depth by wells and reflection surveys is few, and the models are created using the surface geology and the gravity information. We conducted a 3-component survey in southern Osaka plain, Japan. The survey line is 3.6km long. The Ryoke granite (Cretaceous) crops out in the south end of the line. It deepens to the north and covered by Osaka group (late Pliocene to early Pleistocene). The purposes of this survey are the delineation of Ryoke granite top and the estimation of P and S-wave velocity structure of Osaka group in this area. The source was a vibrator Y2400. The attenuation was large in southern part of the line and refractions are visible only to the distance about 300m. In one quarter of the line, coherent noises probably caused by the water line were very strong and made refractions and reflections hard to be recognized. From the P-wave section, we can see the Osaka group strata in the northern part of the section. And the basement structure with steps can be seen clearly. The depth is about 700m at the north end of the line. But because of the strong coherent noises, the Osaka group strata are not imaged in the middle and southern part of the section. That makes the interpretation of the steps difficult. They can be faults related to Uemachi Fault and/or Izumi mountains, or they can be just eroded unconformities. The radial component data synthesized from H1 and H2 components were processed for the PS converted-wave section. The same structure and the problem can be observed in the PS converted-wave section. The comparison of the two sections suggests the Vp/Vs is almost 3 for the Osaka group strata in the northern part of the section.
S31A-1034 0800h
A 3-D velocity structure in and around the Miura peninsula, Japan, using a 3-component off-line seismographic array.
A deep seismic profiling around the Metropolitan Tokyo region, the Kanto district, started in 2002 under the project titled as the Regional Characterization of the Crust in Metropolitan Areas for Prediction of Strong Ground Motion. The deep seismic profiling, Tokyo Bay 2003, was performed along the major axis of the Tokyo Bay. Because the seismic line in the Miura peninsula passes through a densely populated area, we have a low signal-to-noise ratio data due to the cultural noise. Thus, in addition to the conventional reflection profiling, we deployed 51 off-line recorders with a 3-compornent geophone of 4.5 Hz at carefully selected, quiet receiver points. During 90 days, we had continuous records including many shot signals produced by vibrators on land and air-guns at the bay area. These data provided far-offset first arrival signals and wide angle reflections. We focus on the common receiver gather records of the Tokyo Bay 2003 off-line stations data to identify first arrival and wide angle phases. We applied the first arrival tomography method using a finite difference travel time solver (Hole, 1992) to those data to obtain a 3-D P-wave velocity structure of the uppermost crust along the profile. We obtained a velocity model in and around the Miura peninsula as follows: Across the Tokyo Bay, near surface is a layer with velocities of 2.0-2.5 km/s. A low velocity area corresponds to the fore-arc basin sediment (post Early Miocene) which extends to a depth of approximately 4 km. High velocity patches are located at a depth of approximately 6 km under the Miura peninsula, which we interpreted as Pre-Neogene basement rocks. Finally, the velocity structure obtained by the tomography analysis is used to improve the processing of the reflection profiling data to clarify the deeper structure in the peninsula, including a good velocity constraint for a pre-stack migration of the reflection profiling data.
S31A-1035 0800h
Seismic reflection profiling between the Tone canal and eastern Saitama city in the Kanto plain, central Japan
Thick sedimentary layers down to the basement top affect the strong ground motions generated by earthquakes. Underground surveys are necessary to obtain the fundamental knowledge of active structures and earthquake hazard mitigation especially in thick sedimentary plains. The AIST carried out seismic reflection surveys almost in the middle of the Kanto plain, the metropolitan area and the largest plain of Japan. The survey lines are totally about 20km long between the Tone canal and eastern Saitama city. We used a 4-ton mini-vibrator or a 17-ton vibrator in the eastern and central parts (Yoshikawa line, 14km) and two 17-ton vibrators in the western part (Saitama line, 6km) as seismic source. Intentionally, a 3.5km part is left for future survey between the central part and the western part because of densely population. The data quality is poorer in the westernmost part of the Saitama line than in other part. This is perhaps due to energy attenuation in the surface sediments. A conventional data processing was applied to the seismic data. In the Yoshikawa line, the basement top is 1.1s deep in two way time at the eastern edge and gradually deepens westward to 1.6s at the western edge. In the Saitama line, the basement top is not clearly defined due to multiple reflectors and poorer data quality, but it is considered deeper than 2.1s. The basement depths in the seismic sections are almost compatible with those of well data near the seismic lines. The difference of depth is at least 0.5s (about 500m) in the gap between the two lines. Layers upper than the basement are horizontal or dip westward more gently than the basement top in both lines. On the basis of above results, a fault or inclined top of the basement is inferred in the gap and it has been inactive since the formation of upper sediments.
S31A-1036 0800h
Long-Period Ground Motion: 3-D Finite Difference Simulation of the 2003 Tokachi-oki, Japan, Earthquake (Mw 8.0)
During the 2003 Tokachi-oki earthquake (2003/09/26,04:50, 41.7797N, 144.0785E, 42km; JMA, Mw 8.0), which is an interplate earthquake of Kuril Trench, the long-period ground motion with long duration was observed. At Tomakomai located more than 200 km from epicenter, large oil tanks were damaged by sloshing and heavy fires broke out. It was pointed out that the cause of these damages was the long-period ground motion due to the combination of the large earthquake and the deep sedimentary plain. We performed a simulation of the wave propagation of this earthquake with 3D finite-difference method (FDM) to examine this phenomenon. We constructed a crustal model for FD simulation based on seismic velocity profiles by travel time analysis (Iwasaki et al., 1991). On the top of the model, taking into consideration the result of the refraction surveys, reflection surveys, downhole measurements and the geological information, we assumed sedimentary layer model consisting of five layers. There are several plains that have deep sediments such as Yuhutsu, Tokachi, Ishikari and Konsen plains. The Yufutsu plain where Tomakomai is located has especially thick sediment whose maximum thickness is near 10 km. We used a rupture process model of this earthquake by Honda et al. (2003) for FD simulation. It was estimated by the multi-time window linear waveform inversion analysis. Our 3D FD simulation (0.04-0.3 Hz) successfully reproduced observed waveforms, spectra, PGV's, and duration-time of later phases in a wide area of Hokkaido. This implies the validity of our source model and underground structures. The incident waves are amplified by the soft sediment and successfully reproduced the long-duration waves in the Yufutsu plain. Waves are trapped in the soft sediment and continue for several hundred seconds by propagating back and forth in the plain. The mechanism of wave trapping is complicated. Velocity gradient of the sediment as well as the basin topography affect both amplitude and duration of the ground motions.
S31A-1037 0800h
Surface Waves Observed in the Western Coast Plain of the Taiwan Island During the 1999 Chi-Chi, Taiwan, Earthquake and its Aftershocks
During the 1999 Chi-Chi, Taiwan, earthquake and its consequent aftershocks, significant surface waves were observed in the western coast plain of the Taiwan island. Ground-motion records from the 1999 Chi-Chi mainshock and its five aftershocks with magnitude larger than $M_w$ 6.0 are analyzed in this study. The ground-motions in the western coast plain, which can be classified as NEHRP $E$ site, were dominated by the surface wave in the period range larger than 1.0 second. The distance between the sources and the instruments in the western coast plain are about 20--45 km. We study the attenuation of the surface waves with distance to source. The Rayleigh waves attenuate faster than the Love waves in the coast plain. The amplitude of horizontal Rayleigh wave is about 1.5 time of that of the corresponding Love wave on average in the near-source area. The strong-motions observed in the coast plain are a mixture of $P$, $S$, Love, and Rayleigh waves. $S$-wave was overlapped by the surface waves in a relative wide period range (0.5--1.5 sec). However, the body wave and the surface wave were separated quite well in some seismograms triggered by one aftershock (Sep. 20, 1999, 18:03:41.16, $M_w$ 6.2) of the Chi-Chi mainshock. Based on a basic structural model of the studied area, we use 3D finite-difference modeling to study the special mechanism separating the body wave and surface wave. We find it is really difficult to generate the near-source seismograms with the body wave and surface wave being separated perfectly. Depth of the source and the azimuth of instrument to source are the two key elements that control the separating of the body wave and surface wave.
S31A-1038 0800h
The Apparent Periodicity of Felt Reports in the Alaskan Earthquake Record
Felt reports for Alaskan earthquakes were found to be non-uniformly distributed throughout the year. With a predominantly tourist economy, the Alaskan population nearly triples in the summer months, possibly affecting the reporting of earthquakes in the historical record. Using published felt reports from the National Earthquake Information Center and the Alaska Earthquake Information Center, the percentage of events felt each month in central mainland Alaska were tabulated and compared between the summer and winter seasons. Earthquakes were selected from January 1, 1990 to October 31, 2002, from latitudes 58 to 70 degrees N and longitudes 140 to 160 degrees W, and depths 0 to 200 km. 408 events were felt out of a total of 695 that occurred. A number of parameters, including time of day, latitude, longitude, and magnitude, were additionally compared to specify possible limiting factors within each season. While a strong seasonal effect was not found in magnitude 4.0 ML events and greater, the months of May and June were consistently found to have the highest percentage of felt events with a steep drop occurring in the month of July. We ascribe this effect to the summer melting of the top layer of frozen ground to a depth of about 1.5 meters. Additionally, smaller events from magnitude 1.0 to 4.0 ML were also examined. 396 events were felt out of a total of 7,451 that occurred. We found that small earthquakes were felt, with a significant difference, more readily during summer months than in winter. This is likely an effect of the higher summer population of tourists and greater distribution of open businesses. Together these observations suggest that the historical Alaskan earthquake record is likely biased in favor of more frequent reporting of events occurring in summer months as opposed to winter.