S13D-1076 1340h
Fault location and source process of the 2003 Boumerdes, Algeria, earthquake inferred from geodetic and strong motion data.
The Boumerdes earthquake occurred on a fault which precise location, offshore the algerian coast, was unknown. Geodetic data consist of GPS measurements, levelling points and coastal uplifts. They are first used to determine the absolute position of the fault. We performed a series of inversions assuming different positions and chose the model giving the smallest misfit. According to this analysis, the fault emerge at about 15 km offshore. Accelerograms are then used to infer the space-time history of rupture on the fault plane using a two-step inversion in the spectral domain. The observed strong motion records are in good agreement with the synthetics for the fault location inferred from geodetic data. The fault plane ruptured for about 16 seconds. The slip distribution on the fault indicates one asperity north-west of the hypocenter with a maximum slip amplitude larger than 2.5 m. Another asperity with slightly smaller slip amplitude is located south-east of the hypocenter. The rupture seems to stop its propagation westward when it encounters the Thenia fault, a structure almost perpendicular to the main fault. We computed the spatial distribution of ground motion predicted by this fault model and compared it with the observed damages.
S13D-1077 1340h
Low-Velocity Damaged Structure on the San Andreas Fault at Seismogenic Depths near the SAFOD Drilling Site, Parkfield, CA from Fault-Zone Trapped Waves
We deployed dense linear arrays of PASSCAL REFTEK130s across and along the San Andreas Fault near the SAFOD drilling site at Parkfield to record fault-zone trapped waves in the fall of 2003. We acquired the data for 114 local earthquakes with M0-2.3 including 2 SAFOD target events (located by Thurber and Roecker; Nadeau), five USGS explosions in a fan geometry, and 47 small shots (Hole; Catchings and Rymer). Prominent fault-zone trapped waves with large amplitudes and long duration at 2-5 Hz were observed for events within or close to the fault zone. The data from ~25 on-fault events with the raypath incidence angles smaller than 10 degrees from vertical show that the time duration of the dominant trapped wave energy increases from ~0.8 s to ~2.0 s as the event depth increases from ~2.5 km to ~10 km progressively. These observations show the existence of a low-velocity waveguide on the SAF that likely extends to seismogenic depths in this region, although the velocity reduction in the deeper part of fault zone becomes smaller due to the larger confined stress at greater depths. This is consistent with the low-velocity waveguide in the rupture zone of the 1992 M7.4 Landers earthquake, which has been mapped across seismogenic depths [Li et al., 1994; 2000]. 3-D finite-difference waveform simulations of the data delineate the internal damaged structure of the SAF near SAFOD drilling site using a depth-dependent model with the low-velocity waveguide, which is 150 m wide at the top and narrower at 10 km depth, and has shear velocities reduced by 30-45% from wall-rock velocities and Q of 10-50. The model parameters are consistent with those obtained on the SAF near Parkfield in previous studied using fault-zone trapped waves [Li et al., 1990; 1997; 2004], showing that the low-velocity zone extends along the SAF strike from the rupture of the 1966 M6 earthquake to the NW creeping segment. We interpret that the trapped wave inferred low-velocity waveguide on the SAF is a remnant of repeated damage due to M6 earthquakes and other large historical earthquakes on the principal slip plane at Parkfield. We also find some seismic energy trapped in the Buzzard Canyon fault that is approximately 1 km SW to the SAF main trace and dips toward the NE to connect to the SAF at a shallow depth. Among many other methods, fault-zone trapped waves can be used to document fine-scale damaged fault zones with high-resolution.
S13D-1078 1340h
Quasi-Spherical Approach: A Fast Method for Modeling of Seismic Wave Propagation in a 2D Slice of a Global Earth Model With Lateral Heterogeneity
We propose a new method for calculation of seismic wavefield propagating in a global earth model includes lateral heterogeneity. In the field of global waveform modeling, 2D or axisymmetric modeling has been used because full 3D modeling is too intensive for real application. Considering 2D modeling cannot correctly model 3D geometrical spreading effects, axisymmetric modeling is the most appropriate method for global seismology. However, axisymmetric modeling couldn't treat unaxisymmetrical structure with respect to the source axis. Furthermore the scattered and reflected waves from the symmetric continuation of the structure can be returned to the target zone as artificial numerical noise. In order to overcome these problems, we here propose an ultra efficient approach for modeling 3D elastic wavefield. We solve the elastodynamic equation for spherical coordinates not in the conventional spherical domain: $(0<r<\infty, \,0 \leq \theta \leq \pi, \,-\pi \leq \phi \leq \pi)$ but instead in the ``quasi-spherical domain'': $(0<r<\infty, \,-\pi \leq \theta \leq \pi, \,-\pi/2 \leq \phi \leq \pi/2)$ . This approach can simulate seismic wave propagation in a 2D slice of global earth model with an arbitrary lateral heterogeneity in similar computation time and storage as for 2D modeling. In addition it can correctly model 3D geometrical spreading effects and make it possible a direct comparison of real and synthetic waveform data. We applied this approach for a point source with an axisymmetric force system and developed a numerical code using a velocity-stress finite-difference method (FDM). The accuracy of this method was checked by comparing our results with those obtained by the Direct Solution Method (DSM) using spherically symmetric structure. By adopting an irregular grid configuration in the vertical direction and choosing the grid size to be small at the free surface, we succeeded in getting sufficient accuracy. In this presentation we show the comparisons of the seismograms and some numerical examples to demonstrate the validity and feasibility of our method.
S13D-1079 1340h
The Coupling of the Random Properties of the Source and the Ground Motion for the 1999 Chi Chi Earthquake
Two of the most fundamental principles in science are founded on the similitude in properties between a single event and a sum of these events. In seismology, the principle of superposition stipulates that during an earthquake, the waveform observed at a given distance of the fault is essentially the sum of waves emitted by point sources distributed over the fault surface. On the other hand, the Central Limit theorem postulates that the sum of Levy random variables is also a Levy random variable. In studies of source models for several earthquakes-the 1979 Imperial Valley, the 1989 Loma Prieta, the 1994 Northridge and 1995 Hyogo-ken Nanbu (Kobe), we have found that the spatial distributions of slip and pre-stress are characterized by a Levy law. This result can be used to deduce statistical properties of the radiated field. During an earthquake the rupture front propagates over the fault surface; as the rupture front reaches different points on the fault, each point source will emit a wave with an amplitude proportional to the stress released or the "stress drop." Because the magnitude of the stress drop is distributed according to a Levy law, so will the point source wave amplitude. The signal observed at a given distance from the source will be the sum of the signals emitted by the point sources. Because the point source wave amplitudes are distributed according to a Levy law, the sum of these signal amplitudes observed at a given distance from the sources will also be distributed according to a Levy law. We show that both the slip distribution and the peak ground acceleration (PGA) for the 1999 Chi Chi earthquake can be described by Levy laws. Furthermore, we found that the tails of the probability density functions (PDF) characterizing the slip and the PGA are governed by a parameter, the Levy index, with almost the same values as predicted by the Central Limit theorem. The PDF tail controls the frequency at which extreme large events can occur. These events are the large stress drops -or asperities- distributed over the fault surface and the large PGA observed in the ground motion. Our results suggest that the frequency of these events is coupled, and the PDF of the PGA is a direct consequence of the PDF of the asperities.
S13D-1080 1340h
High attenuation in the lower crust in Kinki region, Japan
Seismic attenuation is important for predictions of strong earthquake ground motions, and also it gives additional constraints on the physical properties of deep medium. In (Petukhin et al., 2003) the high-frequency {\it Q}-value, parameter of seismic attenuation, was studied in Kinki region. In that work effect of geometrical spreading, which is necessary to remove before inversion for {\it Q}-value, was calculated numerically using realistic 3D velocity model and ray approximation. Generally, estimated total {\it Q}-values agree well with the results of other studies and with common expectations based on the tectonic structure, except of one striking result: {\it Q}-value for the lower crust become extremely low, {\it Q$_{total}$} ~ 20{\it f}$^{0.9}$. To interpret this result we compiled phenomena, related to attenuation, that were observed in the studied region. They are: (1) seismogenic upper crust, depth 0-17km; (2) aseismic lower crust; (3) reflective lower crust (RLC) in depth range 17-35km revealed by the deep seismic exploration studies in the studied region; (4) belt-like zone of the deep low-frequency tremor generation (LFT), parallel to the slab, which was observed using the high-sensitivity borehole Hi-net stations (Obara, 2002); (5) a few deep low-frequency earthquakes (LFE) near Moho boundary were observed in the central part of Kinki region, far from volcanic centers. Generation of the LFT and LFE is usually explained by the presence of liquid phase in the crust near/above Moho boundary, which in turn can be explained by the dehydration and/or partial melting in the depth range 30-50km. Analysis shows, that anomalously low {\it Q}-value in lower crust can be explained by two processes: (1) high intrinsic attenuation (low {\it Q$_{in}$} value) due to presence of fluids in the lower crust, which are indicated by the LFE and LFT phenomena; (2) high scattering attenuation (low {\it Q$_{sc}$} value) due to scattering on the lower crust heterogeneities, indicated by the RLC. To separate these two cases we analyzed HF envelopes: in first case envelopes should have pulse-like shape with short duration and low coda, in second case envelopes should have intensive scattered part (coda) and long duration respectively. HF average envelopes of records of small earthquakes are analyzed for two cases: for shallow earthquakes with rays covering mostly upper crust and for deep earthquakes with rays covering mostly lower crust. The results indicate that the case of low {\it Q$_{in}$} value and/or low {\it Q$_{sc}$} value for the case of small-scale heterogeneities, is valid for the lower crust in Kinki region.
S13D-1081 1340h
The Physics of Earthquakes: In the Quest for a Unified Theory (or Model) That Quantitatively Describes the Entire Process of an Earthquake Rupture, From its Nucleation to the Dynamic Regime and to its Arrest
For the past four decades, great progress has been made in understanding earthquake source processes. In particular, recent progress in the field of the physics of earthquakes has contributed substantially to unraveling the earthquake generation process in quantitative terms. Yet, a fundamental problem remains unresolved in this field. The constitutive law that governs the behavior of earthquake ruptures is the basis of earthquake physics, and the governing law plays a fundamental role in accounting for the entire process of an earthquake rupture, from its nucleation to the dynamic propagation to its arrest, quantitatively in a unified and consistent manner. Therefore, without establishing the rational constitutive law, the physics of earthquakes cannot be a quantitative science in a true sense, and hence it is urgent to establish the rational constitutive law. However, it has been controversial over the past two decades, and it is still controversial, what the constitutive law for earthquake ruptures ought to be, and how it should be formulated. To resolve the controversy is a necessary step towards a more complete, unified theory of earthquake physics, and now the time is ripe to do so. Because of its fundamental importance, we have to discuss thoroughly and rigorously what the constitutive law ought to be from the standpoint of the physics of rock friction and fracture on the basis of solid evidence. There are prerequisites for the constitutive formulation. The brittle, seismogenic layer and individual faults therein are characterized by inhomogeneity, and fault inhomogeneity has profound implications for earthquake ruptures. In addition, rupture phenomena including earthquakes are inherently scale dependent; indeed, some of the physical quantities inherent in rupture exhibit scale dependence. To treat scale-dependent physical quantities inherent in the rupture over a broad scale range quantitatively in a unified and consistent manner, it is critical to formulate the governing law properly so as to incorporate the scaling property. Thus, the properties of fault inhomogeneity and physical scaling are indispensable prerequisites to be incorporated into the constitutive formulation. Thorough discussion in this context necessarily leads to the consistent conclusion that the constitutive law must be formulated in such a manner that the shear traction is a primary function of the slip displacement, with the secondary effect of slip rate or stationary contact time. This constitutive formulation makes it possible to account for the entire process of an earthquake rupture over a broad scale range quantitatively in a unified and consistent manner.
S13D-1082 1340h
High-frequency envelope inversion analysis of the 2003 Tokachi-Oki, JAPAN, earthquake (Mw8.0)
The 2003 Tokachi-Oki earthquake (Mw 8.0) took place on September 26, 2003 at the plate interface between the subducting Pacific plate and the Hokkaido island, northern Japan. The focal depth is around 30km and the focal mechanism is thrust type. This earthquake caused 2 missings, more than 100 injures, 2000 collapsed houses, and so on. Slip distribution on the fault plane was already estimated by inversion analyses of low-frequency seismograms. However, source characteristics for the earthquake in frequencies higher than 1 Hz is not so far clarified. In this study, we execute an envelope inversion analysis based on the method by Nakahara et al. (1998) and clarify the spatial distribution of high-frequency seismic energy radiation on the fault plane of this earthquake. We use three-component sum of mean squared velocity seismograms multiplied by a density of earth medium, which is called envelopes here, for the envelope inversion analysis. Three frequency bands of 1-2, 2-4, and 4-8 Hz are adopted. We use envelopes in the time window from the onset of S waves to the lapse time of 128 sec. Green functions of envelopes representing the energy propagation process through a scattering medium are calculated based on the radiative transfer theory, which are characterized by parameters of scattering attenuation and intrinsic absorption. We use the values obtained for eastern Hokkaido (Hoshiba, 1993). We assume the fault plane as follows: strike=$249^{\circ}$, dip=$15^{\circ}$, rake=$130^{\circ}$, length=150km, width=165km with reference to a waveform inversion analysis in low frequencies (e.g. Yagi, 2003). We divide this fault plane into 110 subfaults, each of which is a 15km x 15km square. Rupture velocity is assumed to be constant. Seismic energy is radiated from a point source as soon as the rupture front passes the center of each subfault. Time function of energy radiation is assumed as a box-car function. The amount of seismic energy from all the subfaults and site amplification factors for all the stations are estimated by the envelope inversion method. Rupture velocity and the duration time of a box-car function should be estimated by a grid search. Theoretical envelopes calculated with best-fit parameters generally fit to observed ones. The rupture velocity and duration time were estimated as 3.0 km/s and 6 sec, respectively. The high-frequency seismic energy was found to be radiated mainly from two spots on the fault plane: The first one is the deeper part beneath the initial rupture point and the second is the southern shallow part of the fault plane. Radiated energy was estimated to be $7.2 \times 10^{16}J$ in the 1-8Hz band. Acknowledgements: We used strong-motion seismograms recorded by the K-NET and KiK-net of NIED, JAPAN.
S13D-1083 1340h
A Kinematic Source Time Function Compatible With Earthquake Dynamics: Relations Between Kinematic and Dynamic Parameters
We show in this study that the adoption in kinematic source models of slip-velocity time functions which are not consistent with the dynamic propagation of an earthquake rupture can affect the inferred dynamic traction evolution. To overcome this limitation, we propose a new source time function to be used in kinematic modeling of ground motion time histories which is compatible with elastodynamics and it makes feasible the dynamic interpretation of slip models. This function is derived from a different one, first proposed by Yoffe (1951), which models a slip velocity pulse propagating during the earthquake rupture. In order to remove its singularity at the crack tip, we apply a convolution with a triangular function and we obtain a regularized source time function called regularized Yoffe. We provide an analytical expression of this slip velocity time function and propose its parameterization in terms of the final slip, its duration and the duration of the positive slip acceleration. We use this function to prescribe the slip velocity history and compute the dynamic traction evolution on the fault plane through a 3D finite difference algorithm. We estimate the relevant dynamic parameters from the inferred traction evolution. Therefore, we examine the relations between kinematic and dynamic parameters, such as peak slip velocity, slip weakening distance (D$_c$) and breakdown stress drop. The inferred scaling relations are consistent with those proposed by Ohnaka and Yamashita (1989) from laboratory experiments. Our numerical simulations reveal that the inferred values of D$_c$ depend on the adopted slip velocity time function: they can range between 30% and 80% of the total slip. The source time function proposed in this study yields reasonable values of D$_c$ and includes the healing of slip.
S13D-1084 1340h
The Influence of the Geometry of the San Andreas Fault System on Earthquakes in California
The San Andreas Fault is believed to be the main surface trace of the plate boundary between the North American and the Pacific plates. From 1800 to present, three large historical earthquakes (1857 M7.9, 1906 M8.25, and 1989 M7.1) ruptured the San Andreas Fault. At the same time, more than a dozen M$>$7.0 earthquakes occurred outside the main trace of the San Andreas Fault. Most of the off-main-trace large earthquakes were scattered in Southern California, whereas in northern and central California, earthquakes were clustered along the main trace of the San Andreas Fault. Such a seismic distribution may be related to the geometry of the San Andreas Fault, which is curved with a major bending in southern California. In this study, we constructed a finite element model to explore the influence of the geometry of the San Andreas Fault system on stress distribution and seismicity in California. In the model, the San Andreas Fault is simulated with a weak zone that obeys the Coulomb Friction Law. The model results show that along relative straight segments of the San Andreas Fault in northern and central California, fault slip on the main fault trace causes low level stresses in nearby regions. Along the bended San Andreas Fault in southern California, however, the relative plate motion causes significant off-main-trace stress buildup, consistent with the distribution of large historical earthquakes outside the San Andreas Fault.
S13D-1085 1340h
Dynamics of Stress-Field Proxies Associated With Large Earthquake Cycles in a Quasidynamic Fault Model
We use the recently developed quasidynamic model of a discrete strike-slip fault in an elastic solid (Z{\"o}ller et al., 2004) to study proxies of the stress field on the fault in relation to large earthquake cycles. The model, based on earlier work of Ben-Zion and Rice (1993) and Ben-Zion (1996), is governed by realistic boundary conditions, static-kinetic friction, aseismic creep, and stress-transfer for static dislocations in an elastic half-space (Chinnery, 1963). The dynamic rupture is approximated by stress propagation with a constant velocity. Model simulations over thousands of years show that important properties of observed seismicity, e.g. distributions of the earthquake sizes, times and hypocenters, can be reproduced. A key question for time-dependent seismic hazard assessment is the identification of the state of an earthquake cycle at a given time. This requires knowledge of the stress on the fault, which is known in the model, but not accessible for real fault systems. Therefore, we search for useful proxies for the stress field in the framework of the above model. In addition to various seismicity-based signals, a possible proxy is the surface displacement field, which can be determined by geodetic arrays. Using model realizations with different amount of heterogeneities, representing different evolutionary stages of a fault, we discuss the ability of such quantities to identify stages of a large earthquake cycle on a fault.
S13D-1086 1340h
Scaling Relationship Among Source Parameters of Microearthquake," From Near Source Observation in a Deep Mine
Scaling relationships among various source parameters are important clues to understand the source process. In particular the relationship between the corner frequency, $f_C$, and the seismic moment, $M_O$, has been investigated by many researchers. Aki(1967) investigated $f_C$ and $M_O$ using the spectra of seismic waves and reported that these parameters obeyed a relationship of $M_O \propto f_C^{-3}$. For small earthquakes, the breakdown of this relationship was often reported. On the other hand, no breakdown of the relationship for microearthquakes has been reported from high quality observation at deep boreholes and in a deep gold mine. We report here these scaling relationships using waveform of microearthquakes observed at the distance range of 15m to 1km. We installed nine tri-axial borehole accelerometers within 200 m along a haulage tunnel 2650m deep in Mponeng mine in South Africa from February to December in 1996. More than 25 thousand seismic events were recorded with a sampling frequency of 15 kHz and a dynamic range of 120 dB. The recording system has flat response up to 2 KHz. Among those events, we select 378 events with high S/N. We locate hypocenters assuming infinite medium with the P-wave velocity 5.5 km/s and the S-wave velocity 3.2 km/s. We calculate the green function using the discrete wavenumber integral method into account the effect of anelasticity by Takeo (1985) and determine the seismic moment and the mechanism using moment tensor inversion. We apply the omega square model by Brune (1970) to determine the corner frequency and the stress drop. Minimizing L2 norm between the observed spectra of P and S waves and the synthetic ones give the optimum source parameters. The earthquakes analyzed in this study show the constant stress drop of 0.4$\sim$8MPa in the ranges of 40$<f_C<$380 Hz and $10^{8}<M_O<10^{11}$Nm. This means that Mo is scaled as $f_C^{-3}$. This result is consistent with the previous work by Ogasawara et al. (2001). Compared with other studies, the scaling relationship$M_O \propto f_C^{-3}$ as is confirmed. Our data is very similar to the result of natural earthquakes observed at deep boreholes in the fault zone by Abercrombie (1995) and Hiramatsu et al. (2002), implying that there is no difference between mining earthquakes and natural earthquakes.
S13D-1087 1340h
Crustal Tomography Imaging Beneath The Gulf of Corinth Using a new Seismic Dataset Collected in 2002.
The Gulf of Corinth is the most active rifting zone of the Mediterranean region. Despite many studies, the crustal structure of this region is still a debating subject, especially for the western part of the Gulf where authors assume the presence of a low-angle detachment lying at 9-11 km depth. In this study, we construct improved mid-crust velocity models and accurate earthquake locations in order to provide new constraints on the tectonic and the geodynamic framework of the Corinth region, thanks to the new large dataset collected during the 3F-Corinth project around the Aigion area in 2002. The tomographic method we use is based on both an accurate finite-difference travel-time computation and a linearized iterative inversion scheme. P- and S- first-arrival times are simultaneously inverted for both earthquake locations and velocity distribution. Relocation of sources is also obtained from double difference procedures in order to highlight the active structures. A cross-correlation method is applied to the selected dataset for the improvement of the quality of the time picking catalog. The retrieved Vp and Vs models and the absolute location of events allow to image the mid-crust structure. The deduced Vp/Vs and Vp*Vs images give us additional information about the spatial distribution of porosity and fluid content. These results indicate the presence of two different zones with depth. The first 5 km corresponds to the shallower zone which is controlled by the NS extension and where we observe a lack of seismicity. The deeper zone (7-13 km) is characterized by faster and more heterogeneous anomalies. Moreover this zone matches the seismogenic zone. Our images confirm previous tomographic results obtained from data collected in 1991 in the same area with a less dense network and their interpretation can provide new insights on the presence and the nature of an active low-angle detachment in the western Gulf.
S13D-1088 1340h
3D Structural Model in Northwestern Kagoshima, Japan, for Strong Motion Simulation
Nowadays the dramatic development of computer resources and computational techniques have allowed us to calculate the seismic wavefields of three-dimensionally heterogeneous structure models (3D models). Using the 3D models, seismic-wavefield simulations have been widely performed in order to estimate strong motion or to calculate the Green functions for source inversions. Therefore, the construction of realistic 3D models is the key to success for strong motion predictions or source inversions. The purpose of this study is to construct a 3D model which has a adequate ability for strong motion simulation. We construct a 3D model in northwestern Kagoshima region of Kyushu island, Japan using all kinds of geophysical and geological data which were well investigated in the study area. The 3D model consists of six layers with width of 80 km, length of 80 km and depth of 40 km. When we construct the 3D model, we estimate a depth for a top of each layer and next assigned $P$-wave velocity to the top and its gradient for each layer. Using the constructed 3D model we then simulate the seismic waveforms of two small earthquakes (M4.6, M3.6) by the finite difference method to assess the ability of the 3D model for seismic motion simulation with frequency range between 0.1 to 1 Hz. These test events are aftershocks of the 1997 Northwestern Kagoshima, Japan, earthquakes, which were well recorded at K-NET (NIED) stations and our stations located on near-field hard rock sites. As the results, it can then be seen that the good agreement of $P$ and $S$ waves between the synthetic and the observed waveforms. It suggests that our 3D model is suitable to seismic wavefield simulations in the frequency range. We also compare the synthetic waveforms calculated by using the 3D structural model and those from the corresponding 1D model, and found that 1D synthetics can poorly explain the observed waveforms as compared to the 3D ones, which may suggest that the constructed 3D model can be applied to the strong motion simulations or source inversions using the 3D Green functions for large events occurring in this region, such as the mainshocks of the 1997 Northwestern Kagoshima earthquakes ($M$6.5, $M$6.3).
S13D-1089 1340h
Multi-resolution extension of Spectral Elements to non smooth dynamics: earthquake rupturing process
Spectral element method (SEM) already proved to be a powerful tool for the simulation of the seismic waves at regional and global scales. For smooth dynamics, the SEM actually combines the geometrical flexibility of the finite elements with the exponential convergence rate, proper to high-order spectral techniques. Extension to non smooth dynamics, such as faulting process and short wave radiation, requires specific solutions. We present here these main extensions in order to solve earthquake dynamic rupturing for realistic fault geometries and complex geological media. First, a new second-order time stepping allowing an implicit solution of the non-smooth contact and frictional conditions on faults of arbitrary shape is discussed and shown to be rather effective. Second, consistent high frequency dissipation is added to the SEM through a complex prolongation of the elastodynamics Green's functions. This leads to a selective absorption of the high-frequencies within a finite thickness fault zone. Finally, a multiresolution strategy has been developped by extending the non conforming Mortar method in space to a non conforming matching condition for domains with different space and time steps. Such a solution will allow to accurately resolve the high-frequencies generated during the rupturing process in the vicinity of the fault and to propagate the low-frequency field away from the fault itself to the receivers. The efficiency of these extensions is quantitatively addressed by several examples that will be discussed. This work is part of the SPICE European Project Festa, G., and Vilotte, J.-P., Efficient PML for Spectral element methods, Submitted to J. Geophys. Int. Festa, G., Vilotte, J.-P. and Ampuero, J.-P., Dynamic rupturing within Spectral elements, Submitted to Bull. Seism. Soc.Am.
S13D-1090 1340h
Low Velocity Oceanic Crusts at the uppermost part of the Subducting Plates Beneath Japan Arc Derived From Seismic Tomography
Aki and Lee (1976) and Aki $et$ $al$. (1977) are the pioneer studies of the body-wave seismic tomography for the local and regional scale. Aki and Lee (1976) use blocks and Thurber (1983) and Zhao $et$ $al$. (1992) use grid nodes for model parameterization. A weak point of the grid method is that the velocity must be continuous for any direction. Zhao $et$ $al$. (1992) proposed a new parameterization for 3-D velocity-discontinuity with complex shape. Assumption of the velocity discontinuity is available only if the depth of the discontinuity is correct, however, the wrong position of the discontinuity has some bad influence on the velocity structure. We place many grid nodes and introduce correlation among velocities at surrounding grid nodes into Zhao_fs method (Matsubara $et$ $al$., 2004) to obtain more fine and stable solution. Thus the obtained models might be more close to the real. We apply this modified method to 1,033,879 $P$- and 716,561 $S$-wave arrival times from 21,126 natural sources recorded by 698 stations of the NIED Hi-net with a grid spacing of $0.125\deg$ to obtain the crustal and upper mantle structure beneath the whole Japan. No velocity discontinuity such as Moho or the upper boundary of the Pacific (PAC) plate is assumed. The PAC plate beneath the northeastern Japan arc has about +5% velocity disturbance and high-velocity (high-$V$) zone at depths of 20-40 km beneath the southwestern Japan is considered as the Philippine Sea (PHS) plate. Low-velocity (low-$V$) zones considered as oceanic crust in the uppermost part of the PAC plate is obtained from south Tohoku to Kanto district at depths of 30-140 km. Low-$V$ zones considered as the oceanic crust in the uppermosts part of the PHS plate beneath from the Kanto to Kyushu district at depths of 30-60 km beneath the Kanto and southern Kinki district and at depths of 30-70 km beneath the western Shikoku district. We focus the underground structure around the Tokyo Metropolitan area, the Kanto region, central Japan, where the PAC and PHS plates subduct beneath the Eurasian plate. The same tomographic method is applied to 422,799 $P$- and 369,596 $S$-wave arrival times from 15,214 natural sources recorded by 176 stations with a grid spacing of $0.05\deg$. The high-$V$ PHS plate subducts to depths around 80 km. We clearly found low-$V$ oceanic layer in the uppermost part of the PAC and PHS plates. Low-$V$ oceanic layer of the PAC plate exists at depths of 60-120 km. The $V_P$/$V_S$ ratio of this layer is abount 1.85-1.90 and the possibility of the existence of the molten rock is small because of not so large $V_P$/$V_S$ ratio. There is garnet-granulite oceanic crust which is transferred from basalt or gabbro and not to eclogite with temperature about 400$\deg$C. Above the oceanic layer of the PHS plate, low-$V$ zones also exist in the mantle wedge at depths of 30-60 km. The low-$V$ zone in the western part of the mantle wedge about 60 km away from the Sagami Trough is composed of gabbro and that in the eastern part of the mantle wedge about 100 km away from the trough consists of about 30% serpentinized gabbro with consideration of $V_P$/$V_S$ ratio about 1.80-1.90. The PHS plate also has double seismic zone. The seismicity is high where the $V_P$/$V_S$ ratio is large.
S13D-1091 1340h
The Strain Energy, Seismic Moment and Magnitudes of Large Earthquakes
The strain energy $E_{st}$, as potential energy, released by an earthquake and the seismic moment $M_o$ are two fundamental physical earthquake parameters. The earthquake rupture process ``represents'' the release of the accumulated $E_{st}$. The moment $M_o$, first obtained in 1966 by Aki, revolutioned the quantification of earthquake size and led to the elimination of the limitations of the conventional magnitudes (originally ML, Richter, 1930) mb, Ms, m, MGR. Both $M_o$ and $E_{st}$, not in a 1-to-1 correspondence, are uniform measures of the size, although $E_{st}$ is presently less accurate than $M_o$. Est is partitioned in seismic- ($E_s$), fracture- ($E_g$) and frictional-energy $E_f$, and $E_f$ is lost as frictional heat energy. The available $E_{st} = E_s + E_g$ (Aki and Richards (1980), Kostrov and Das, (1988) for fundamentals on $M_o$ and $E_{st}$). Related to $M_o$, $E_{st}$ and $E_s$, several modern magnitudes were defined under various assumptions: the moment magnitude $M_w$ (Kanamori, 1977), strain energy magnitude $M_E$ (Purcaru and Berckhemer, 1978), tsunami magnitude $M_t$ (Abe, 1979), mantle magnitude $M_m$ (Okal and Talandier, 1987), seismic energy magnitude $M_e$ (Choy and Boatright, 1995, Yanovskaya et al, 1996), body-wave magnitude $M_{pw}$ (Tsuboi et al, 1998). The available $E_{st} = (1/2\mu)\Delta\sigma M_o$, $\, \Delta\sigma$~=~average stress drop, and M$_E$ is % \[M_E = 2/3(\log M_o + \log(\Delta\sigma/\mu)-12.1)\, ,\] % and $\log E_{st} = 11.8 + 1.5 M_E$. The estimation of $E_{st}$ was modified to include $M_o$, $\Delta$ and $\mu$ of predominant high slip zones (asperities) to account for multiple events (Purcaru, 1997): % \[E_{st} = \frac{1}{2} \sum_i {\frac{1}{\mu_i} M_{o,i} \Delta\sigma_i} \, , \hspace{0.8cm} \sum_i M_{o,i} = M_o \] % We derived the energy balance of $E_{st}$, $E_s$ and $E_g$ as: % \[ E_{st}/M_o = (1+e(g,s)) E_s/M_o \, , \hspace{0.8cm} e(g,s) = E_g/E_s \] % We analyzed a set of about 90 large earthquakes and found that, depending on the goal these magnitudes quantify differently the rupture process, thus providing complementary means of earthquake characterization. Results for some classes of large earthquakes are discussed.
S13D-1092 1340h
Whole Earth Tomographic Models: a Resolution Analysis
The evaluation of the resolution of global tomographic models of the Earth through synthetic tests has been proven to have intrinsic limits, strongly depending on the choice of the "input model". We have evaluated the resolution of whole-Earth structure achieved by compressional wave travel time data from the ISC bulletins, computing directly the model resolution matrix (e.g., Menke, 1989), i.e. the operator that relates output and input model in any synthetic test. The knowledge of R permits to detect what fictitious trade-offs (aliasing, leaking) occur, in the least-squares inversion, between any couple of model coefficients. While the robustness of mantle images is confirmed, the problem of mapping the Earth's CMB and core is inherently unstable: we found severe fictitious trade-offs within the CMB and outer core, or between the two, suggesting that a reliable modeling of those regions should take into account further complexities of the real Earth like the high-frequency structure in the D'' and the full mantle anisotropy. Our computational resources and the speed with which the inversions are carried out (few minutes) on a shared-memory, multiprocessor computer makes a refinement of our parameterization feasible.
S13D-1093 1340h
Temporal Changes of Seismic Structure Around Iwate Volcano as Inferred From Waveform Correlation Analysis of Multiplet Earthquakes
Temporal change of crustal structure around Iwate volcano, northeastern Japan, for the period from 1995 to 2002 is investigated by examining waveform similarity in P- and S-waves of multiplet earthquakes occurring at the subducting Pacific Plate boundary. We analyze seventy-three groups of multiplets consisting of 2 to 20 events whose epicentral distances range from 80 to 260 km and magnitudes from 2.1 to 4.5. These multiplets are recorded at 30 stations within approximately 80 km from the summit of Iwate volcano. We calculate cross-correlation coefficients of P- and S-waves for each pair of multiplets belonging to the same group. Totally, 1142 pairs are analyzed. We first band-pass filter each waveform from 4 to 8 Hz and align them at P-wave onset. The cross-correlation coefficients of P- and S-waves are calculated for 5-second time window starting at onset time of each phase. The spatial distribution of the coefficient is examined by averaging the cross-correlation coefficients at each station. At most stations far from Iwate volcano, the averaged coefficients of S-waves are estimated to be from 0.95 to 0.97. On the other hand, two stations, IKG and GNB, located close to the west of the volcano show smaller average of 0.85 and 0.91, respectively. These stations are located close to active volcanic pressure sources detected by geodetic measurements. We further examine temporal changes in S-wave by aligning the cross-correlation coefficients for multiplet pairs with an occurrence interval of 6 to 18 months. The results show that no significant temporal change is observed at most of stations located far from the volcano. However, it is found that the coefficients at IKG and GNB remarkably decreased from 1998 during which significant volcanic activity was observed and an M6.1 earthquake took place at the southwest of the volcano. As the volcanic activity slowly subsided from 2000, the coefficients gradually become large. These changes are well correlated in time with activities of Iwate volcano. Since changes in cross-correlation of P-waves are smaller compared with those of S-waves, the crustal structure changed its property properties sensitive to S-waves, such as rigidity, related to magmatic and/or seismic activity in and around the volcano.
S13D-1094 1340h
Numerical Investigation of the Effects of Anisotropy in a Fault Zone on Seismic Wavefield
The physical properties in and around the fault zone has came under intense study and observational data using dense seismic array network near the fault zone enables us to obtain the general view of spatial distribution of anisotropy associated with the fault zone such as fractures or cracks. Based on these anisotropic distributions, in this study we make and analyze the synthetic seismogram in order to estimate the influence of anisotropy in the fault zone on seismic wavefield. The propagator matrix method is used to calculate synthetic seismograms for a vertically layered medium. In the synthetic seismograms a later phase can be seen just after the main S-wave, and the separation between them are proportional to the propagation distance in the anisotropic fault zone and the degree of anisotropy. We suggest that observing and analyzing the later phase could be a clue to detect the fault zone. We also find that in the strike slip source with the strike parallel to the fault zone the synthetic seismograms are more affected by the velocity structure than the anisotropy. This means that it is difficult to estimate the shear wave splitting parameters describing the degree of anisotropy and the direction of symmetry axis from the waveform records for such source mechanism. However, the shear wave splitting parameters are available in the case of the normal-fault or reverse-fault source, and also in the case of the strike slip, normal fault and reverse fault striking 45 degrees against the strike of the fault zone. We suggest that the mechanism solution is needed for estimating the anisotropy in the fault zone.
S13D-1095 1340h
Moho Depth Variation Beneath Southwest Japan Revealed From Inverted Velocity Structure Based on Receiver Functions
We determine the depth variation of the Moho discontinuity beneath Chugoku-Shikoku region, southwest Japan. We apply the receiver function analysis to teleseismic waveforms from more than 250 earthquakes with magnitude 5.5 or larger recorded by the High Sensitivity Seismograph Network (Hi-net). Integrating estimated receiver functions into six groups according to the back azimuth of each station, we estimate the seismic velocity structure for every group of the receiver functions by using the improved linearized time-domain waveform inversion method. This improved method adopts a weighting function to determine the shallow structure well and estimate both S and P wave velocity, simultaneously. We detect a clear velocity discontinuity corresponding to the Moho across which the S wave velocity changes to 4.5 km/s from 3.7 km/s. The depth of the discontinuity is about 30 km beneath northern (the Japan Sea) and southern (the Pacific) coastlines and more than 40 km beneath central part of the study region. In the central part, a low velocity layer (LVL) with 10 km thickness exists under the Moho. The depth of the upper boundary of the LVL is 45 to 50 km. The Philippine Sea plate (PHS) is subducting toward the northwest from the Nankai Trough beneath the Chugoku-Shikoku region where both the continental and the oceanic Moho exist. The LVL corresponds to the subducting oceanic crust of the PHS and the oceanic Moho is the bottom of the oceanic crust. The continental Moho of the Eurasian plate lies above the low velocity oceanic crust. However, at stations in the northern and southern part of the study region, we find only one major velocity discontinuity. We read the depth of these clear discontinuities from the inverted velocity models and map the Moho depth at the conversion point. By interpolating the results, we separately draw the depth contour of the continental and the oceanic Moho beneath Chugoku-Shikoku region under the assumptions: (1) the Moho of the Pacific side and the Japan sea side are the oceanic and the continental Moho, respectively; and (2) each type of the Moho is continuous. We found that the oceanic Moho continues down to northwest and shows a complicated configuration with ridges and valleys. The corresponding depth is about 25 km beneath the Pacific seashore and 45 km beneath the central part of the study region. The oceanic Moho extends to the 34.5$^\circ$N at least. The depth of the continental Moho is relatively shallower at northern coastlines and shows a major depression with more than 40 km in depth at around 34.7$^\circ$N and 133.5$^\circ$E. In the south of 34$^\circ$N, we found only one discontinuity indicating the oceanic crust of the PHS is attached to the continental lower crust. The geometry of these Moho discontinuities has a high correlation with the seismicity in the crust and the Bouguer gravity anomaly.
S13D-1096 1340h
Three dimensional fine structure of seismic attenuation beneath Japan Islands derived with NIED Hi-net maximum amplitude data
The attenuation structure beneath the Japanese islands should be complicated three dimensionally because of the plate convergence area with subducting Pacific and Philippine Sea plates. We carried out tomographic inversions for the three-dimensional attenuation structure beneath the Japanese islands with the high-sensitivity seismograph network of Japan (Hi-net) data operated by the National Research Institute for Earth Science and Disaster Prevention (NIED). We use the maximum velocity amplitudes from vertical component of seismograms for earthquakes which occurred from April 2001 to May 2004. 205,501 amplitudes data from 3,410 earthquakes are used in this analysis. Location of hypocenter and focal mechanism of these earthquakes are reported by NIED Hi-net. The study area between latitude of 30-46N degree and longitudes of 130 - 145E, mainly covers the four main islands of Japan. Many grids are placed with an interval of 0.5 degree in the horizontal subsurface at depths of 10, 25, 40, 65, 90km and every 30km up to 360km. The shallowest two depths correspond to the upper and lower parts of the crust respectively, and the others are in the mantle. From the checkerboard test, the resolution is generally good for the layers shallower than 90km. There are three significant results. (1) High attenuation (low-Q) zones can be found beneath the volcanic front in the northeastern Japan clearly, and the distinct low attenuation (high-Q) zone is recovered in the east of the front. Low-Qp bodies appear only just below volcanoes in the upper and lower crust, while the low-Qs area extends continuously along the volcanic front. In the mantle wedge the low-Q zone are laid toward the west. (2) In the central Japan, a low-Qs area is found at a depth of 40 km in the Kanto region, east side of the volcanic front. In this area low-velocity materials with larger Poisson ratios was found, and considered the materials to be serpentine on the Philippine Sea slab (Matsubara et. al., 2004). Below 65km in the Chubu district, we found a distinct low-Q layer, west side of the volcanic front. (3) The high-Q area is found along the upper boundary of the Philippine Sea slab, which is determined from seismicity in the southwestern Japan. In the Shikoku region, the high-Q area does not extend beyond latitude of 34.2N, and the area looks falling down into a deeper part there. On the other hand, in the Kyushu region, the high-Q zone reaches a depth of 100km or deeper coincident with the intraslab seismicity.
S13D-1097 1340h
Rheological Models in the Time-Domain Modeling of Seismic Motion
The time-domain stress-strain relation in a viscoelastic medium has a form of the convolutory integral which is numerically intractable. This was the reason for the oversimplified models of attenuation in the time-domain seismic wave propagation and earthquake motion modeling. In their pioneering work, Day and Minster (1984) showed the way how to convert the integral into numerically tractable differential form in the case of a general viscoelastic modulus. In response to the work by Day and Minster, Emmerich and Korn (1987) suggested using the rheology of their generalized Maxwell body (GMB) while Carcione et al. (1988) suggested using the generalized Zener body (GZB). The viscoelastic moduli of both rheological models have a form of the rational function and thus the differential form of the stress-strain relation is rather easy to obtain. After the papers by Emmerich and Korn and Carcione et al. numerical modelers decided either for the GMB or GZB rheology and developed 'non-communicating' algorithms. In the many following papers the authors using the GMB never commented the GZB rheology and the corresponding algorithms, and the authors using the GZB never related their methods to the GMB rheology and algorithms. We analyze and compare both rheologies and the corresponding incorporations of the realistic attenuation into the time-domain computations. We then focus on the most recent staggered-grid finite-difference modeling, mainly on accounting for the material heterogeneity in the viscoelastic media, and the computational efficiency of the finite-difference algorithms.
S13D-1098 1340h
Simulation of Earthquake Strong Ground Motion Using the Specific Barrier Model
There are two key fault parameters that represent length scales and control the intermediate and high frequency content of near-fault ground motion: (1) the barrier interval; and, (2) the length of the cohesive end-zone at the crack tip. The barrier interval controls the rise time and therefore is related to the pulse duration of the near-fault pulses that carry considerable destructive potential for man-made structures. On the other hand, the length of the cohesive end-zone at the crack tip, which is a measure of the effective thickness of the fault gauge, is expected to control the intensity of the radiated near-fault pulses. In the present work we focus on the barrier interval, which is equal to the size of the representative sub-event of a main event. The barrier interval is equal to the diameter of the circular cracks of the equal-size sub-events that compose the main event in the Specific Barrier Model (SBM) proposed by Papageorgiou and Aki (1983). We have recently calibrated the model using the most up-to-date databases of earthquake events representing three tectonic regimes (Halldorsson and Papageorgiou, 2004). Using the parameters of the SBM that were obtained from the above calibration, we proceed to simulate time histories for a number of earthquake events that were well recorded, cover a wide magnitude range (Mw 5.9 - 7.9), and are representative samples of different source mechanisms. Objective measures are used to assess the quality of fit of the simulated time histories to the recorded motions. A key assumption of the SBM is the uniform distribution of the seismic moment released over the fault plane (i.e., all sub-events are of equal size). We assess the bias (if any) introduced to the overall simulation by the above assumption. The overall goal of the exercise is to assess the effectiveness of the SBM to provide time histories of earthquake ground motion that can be used with confidence by earthquake engineers in aseismic design.
S13D-1099 1340h
Recent Data from CSMIP Instrumented Downhole Arrays
The California Strong Motion Instrumentation Program (CSMIP) operates 19 downhole geotechnical arrays throughout California, with 8 arrays in Southern and 11 arrays in Northern California. 13 arrays were instrumented with the support and cooperation of the California Department of Transportation (Caltrans). Three more arrays are expected to be instrumented soon. More than 30 low amplitude recordings from earthquakes with 2.4<M<7.1 were recorded at the La Cienega array in Los Angeles area where the freeway collapsed during the Northridge earthquake. This deep soft soil array was instrumented in 1995 with sensors installed at the surface and at depths of 18 m, 100 m and 252 m. The strongest acceleration of 0.5g was recorded at this array during a relatively small earthquake of M4.2 at an epicentral distance of 2.7 km and depth of 7.9 km (almost vertical wave incidence) The downhole array at Treasure Island near San Francisco was installed in 1992 with instruments located at the surface, in fill, alluvium and rock. Transfer functions from the surface to rock (at 104 and 122 m depth) were computed. Both horizontal components demonstrate strong amplifications at: 0.8, 1.9, 3.4, 4.4, 5.8 and 6.8 Hz. The first frequency is associated with the depth to the rock (h = 88 m, average Vs=261 m/s, f = Vs/4h gives 0.74 Hz). During the last year, new low amplitude data were recorded at downhole arrays. An important set of records was obtained during the December 22, 2003 M6.5 San Simeon earthquake at the three arrays: Treasure Island, San Francisco - Bay Bridge, and Hayward - San Mateo Bridge at epicentral distances of 230-260 km. The records from the San Simeon earthquake obtained at different depths demonstrate a similar pattern to that shown during the Hector Mine earthquake: In contrast to accelerations (high frequency part of seismic signal), displacements (relatively low frequency part of seismic signal) demonstrate minimal near-surface site amplification. The one dimensional wave propagation program SHAKE91 was successfully used to model ground motion in the layers for the La Cienega (deep alluvium), Treasure Island (fill and alluvium over rock) and other arrays. It produces good results in most cases when modeling ground motion from low amplitude local events with M<5. Data recorded at the CSMIP instrumented downhole arrays are available through the CISN Engineering Data Center at http://www.quake.ca.gov/cisn-edc/.
S13D-1100 1340h
Seismic Tomography of the Mygdonia Basin (Northern Greece) Using Earthquakes Located With a Double-Difference Algorithm
The structure of the area of the Mygdonia Basin has been extensively studied during the last decade and its seismic potential is sufficient to produce an adequate set of seismic phase data. This makes the broader Mygdonia Basin area an appropriate "test-site" for developing and testing new techniques regarding regional-local scale seismic tomography. In the present work, the P- and S-wave velocity structure of the broader area of the Mygdonia Basin is studied by the inversion of traveltimes of earthquakes. The main objective is to produce a more detailed velocity model of the structure of the area, using a technique based on double differences. The quality of the results of seismic tomography depends, among other factors, on the errors in the locations of the earthquakes used. To reduce these errors, the double difference earthquake location algorithm can be used to relocate the earthquakes before introducing them in travel-time tomography. Using a local-scale 1-D and preliminary 3-D seismic wave velocity model of the area, we have implemented a 3-D ray tracing routine, which was incorporated in the original double-difference algorithm. To further improve the quality of the results, we have merged all available arrival time data from permanent and temporary networks, regional or local. These data were used in both earthquake location and tomography inversion. The applicability of the method is demonstrated through synthetic tests and an improvement of the obtained velocity images is clearly identified by both synthetic and real data.
S13D-1101 1340h
Tomographic Imaging of Active Hawaiian Volcanoes
In the past 25 years, seismic travel time tomography has become a widely, if not routinely, used tool taking advantage of earthquake arrival times, recorded on regional microearthquake monitoring networks, to image earth structure. When applied in volcanically active regions, tomographic studies have revealed the locations of magma chambers, pathways, and other features whose presence and properties will need to be incorporated into physically reasonable models of volcanic structure and process. We have expanded our calculations to include over 30 years of earthquake arrival time data cataloged at the Hawaiian Volcano Observatory. The large datasets extend from 1970 through 2003, and span Mauna Loa's 2 most recent eruptions in 1975 and 1984. Our results are consistent with those of earlier calculations and other investigators. The summit calderas and rift zones of the active volcanoes are associated with high velocity signatures, interpreted to suggest the presence of olivine cumulates developed via repeated eruptive and intrusive activity in these regions. In addition, fault zones beneath the southeastern flanks of both volcanoes are associated with velocity contrasts that extend quite deep into the volcanic edifice, possibly to the basal decollement. We also look to the identification of possible time-varying changes in volcanic structure. A distinct P-wave high-velocity anomaly under Mauna Loa deepens, from beneath its summit caldera and upper southwest rift zone into the southeast flank. Its associated P-wave speeds are comparable to those of features imaged in earlier calculations and which we have interpreted as magma bodies. At depths greater than 9 km below sea level, the lateral velocity heterogeneity weakens, but seismicity in the Kaoiki fault system beneath Mauna Loa's southeast flank does appear to localize along velocity contrasts. Prior to the 1975 eruption, the Mauna Loa high-velocity feature is principally centered beneath the summit caldera and is most apparent above depths of 3 km below sea level. Between 1975 and 1984, the anomaly is at its largest and extends to depths of 9 km. Since 1984, this feature has decreased in lateral extent but also exhibits greater velocity contrasts at shallower depths. Determining the possible relationships of these changes to eruptive processes will require additional volcanological monitoring data, as well as further testing and modeling using the tomographic results.
S13D-1102 1340h
Fine-Scale Structure of the Moho From Receiver Functions: Effects of a Deforming Crust
Andrija Mohorovicic, a Croatian seismologist, is credited with the first estimation in 1906 of crustal thickness using the critically refracted phase Pn. The crust-mantle boundary has become commonly known as the Moho and its depth, structure, formation, and evolution remains an important research topic in seismology, petrology, and tectonics. Other seismic phases sensitive to Moho depth and structure are the converted phases Ps and Sp, and the associated 2p1s and 1p2s reverberation phases that are isolated in receiver function waveforms. With sufficient station coverage, multiple receiver functions can be migrated and stacked into cross-sections of the crust. Crustal cross-sections from tectonically active regions reveal dramatic variations in amplitude and frequency content of Moho phases that we associate with fine-scale structure, and possibly anisotropy at the crust-mantle boundary. The Moho amplitude or "brightness" is a measure of the crust-mantle impedance contrast, thickness and structure within the crust-mantle boundary, and effects of scattering from 3D structure. Processes directly related to these Moho structures include crustal thickening, crustal extension, crustal flow, delamination or convective removal, and eclogitization. Therefore, the fine-scale seismological structure of the Moho is an important constraint in regional tectonic reconstructions. Examples of receiver function crustal images and their tectonic implications from the western US, South American Andes, and the Tibetan plateau will be reviewed.
S13D-1103 1340h
Crustal Structure and Micro-seismic Activity Along the Atotsugawa Fault System, Central Japan
Detailed seismic velocity structure along the Atotsugawa fault system, central Japan, was obtained to infer the structure effect on microseismic activity and fault creep movement. The Atotsugawa fault system consists of subparallel, right-lateral and strike-slip faults (e.g. the Atotsugawa fault, the Ushikubi fault) trending ENE-WSW. The baseline survey indicates a possibility of creep movement of 1.0-1.5mm/year in the middle part of the Atotsugawa fault, where the microseismic activity is very low comparing the other parts of the fault. Thus we clarify crustal structure in the whole fault system by seismic tomography, especially the structure effect on seismic activity along the fault or fault creep. We applied the Double-Difference Tomography method (Zhang and Thurber, 2003), which can make inter-hypocenter structure clear. We used data of local routine network and dozens of temporary stations that were deployed especially in the vicinity of the fault system. In order to avoid local concentration of ray paths, we adjusted the number of used hypocenter data so as to equalize seismic distribution, but except the fault system area. Consequently the number is 522, in particular 287 along the fault system. After tomography analysis, we checked resolution of the result and confirmed to be reliable down to 12km deep in the middle of the fault system. We summarize features of the obtained results as follows. In the cross section along the Atotsugawa fault, there is low velocity zone below 10km deep, which continues from middle lower part of the fault up to near-surface of the eastern fault edge. The micro-earthquakes correspondingly distributes along the upper boundary of the low velocity zone. The western part of the fault has a little higher Vp/Vs ratio than the eastern part, which is corresponding to the higher seismicity. The fault system does not have large-scale low velocity structure along the whole fault system. By several synthetic tests, it is sure that the width is less than at least several kilometers even if such a large-scale low velocity structure exists. In the plane view, low velocity area is localized in the shallow part (down to $\sim$2-4 km) between the Atotsugawa and the Ushikubi fault, which is in close vicinity to the creep area. From the obtained velocity structure, we cannot find the evidence of the Atotsugawa fault creep.
S13D-1104 1340h
Nearly Optimal Acoustic and Elastic Boundary Conditions
Seismic Modeling needs effective boundary conditions to model the free surface and to limit the size of the computational model. The typical three dimensional seismic model is large and would exceed the size of available computer memories without some effective boundary on the sides and bottom of the model. The reflection coefficient of the surface with the air water/land interface is considered strong given the vast differences in densities. The finite nature of the sides and bottom cause the real problem. Numerous numerical approaches have been proposed for creating these artificial boundaries. Among them are the one-way wave equations and the sponge damping zones. First order hyperbolic wave equations have waves moving in one direction. To truly model the propagation of acoustic waves and all the internal reflections the second order wave equation is needed. However, at the boundary edge what we need is a transmission of energy out of the model and no reflection of energy back into the model. Because the finite model is truncated at the boundary we will not generate any of the energy which ought to really be reflected back into the model region from the missing external regions. This is a source of error in our modeling process, an omission error. We have to expect that the effective region of interest is far enough inside the model that the recorded surface or well data is sufficiently accurate in representing reflection data. One-way wave equations result from factoring the full wave equation into two first order systems. This was the technique of Reynolds as documented in his 1976 paper. He factored the wave equation and computed results using two forms of his scheme. His second form was more successful and presented without comment. The sponge or damping zone was presented by Cerjan in 1982 and he showed a scheme that computed a gently tapered a set of weights that were applied to the edges of model wave fields. In this scheme the formula was presented without any rigorous development. By comparison of these two methods it is possible developed a mathematical relationship between the sponge numerics and the one way angle dependent wave equation as developed by Higdon. Further, by using an energy measure it can be shown that the two parameter space of the sponge method has one or more local minima. Using these local minima parameters the effectiveness of the sponge method is enhanced to the point where the boundary effects are reduced to the noise level of the data. Results for 2D and 3D acoustic and 2D elastic wave codes are used to demonstrate the benefits of this nearly optimal sponge boundary condition.