Takashi Furumura

Parallel 3-D pseudospectral simulation of seismic wave propagation

Three-dimensional pseudospectral modeling for a realistic scale problem is still computationally very intensive, even when using current powerful computers. To overcome this, we have developed a parallel pseudospectral code for calculating the 3-D wavefield by concurrent use of a number of processors. The parallel algorithm is based on a partition of the computational domain, where the field quantities are distributed over a number of processors and the calculation is concurrently done in each subdomain with interprocessor communications. Experimental performance tests using three different styles of parallel computers achieved a fairly good speed up compared with conventional computation on a single processor: maximum speed-up rate of 26 using 32 processors of a Thinking Machine CM-5 parallel computer, 1.6 using a Digital Equipment DEC-Alpha two-CPU workstation, and 4.6 using a cluster of eight Sun Microsystems SPARC-Station 10 (SPARC-10) workstations connected by an Ethernet. The result of this test agrees well with the performance theoretically predicted for each system. To demonstrate the feasibility of our parallel algorithm, we show three examples: 3-D acoustic and elastic modeling of fault-zone trapped waves and the calculation of elastic wave propagation in a 3-D syncline model.

Anomalous Propagation of Long-Period Ground Motions Recorded in Tokyo during the 23 October 2004 Mw 6.6 Niigata-ken Chuetsu, Japan, Earthquake

Unusually large (5 cm) and prolonged shaking associated with long- period ground motions at periods of about 7 sec were observed in central Tokyo during the Mw 6.6 Niigata-ken Chuetsu earthquake of 23 October 2004. The long- period ground motions caused significant resonance in high-rise buildings of about 70 floors in height. Thus, it is an urgent matter to understand the development and amplification properties of long-period ground motions in Tokyo associated with large earthquakes. In this study, we use numerous waveform records from 585 stations in a nationwide accelerometer network (K-NET, KiK-net) and 495 intensity meters in the area around Tokyo. The data reveal that the long-period ground motion is characterized in most part by a surface, Rayleigh wave generated at the northern edge of Kanto basin, and the surface wave is developed as propagating through a thick cover of sediments (3000-4000 m) that overlies rigid bedrock. To complement the observational data, we conducted a large-scale computer simu- lation of seismic-wave propagation by employing the Earth Simulator supercomputer with a detailed source-slip model and a high-resolution 3D sedimentary structural model of central Japan. The results of the computer simulation demonstrate that the anomalously prolonged ground shaking of the long-period signal recorded in the center of Tokyo occurred because of the stagnation of seismic energy resulting from the multipathing and focusing of Rayleigh waves toward the bottom of the Kanto basin from surrounding mountain regions with interaction to the 3D basin structure.

Specific distribution of ground motion during the 1995 Kobe Earthquake and its generation mechanism

The 1995 Kobe (Hyogo-ken Nanbu) earthquake (Mw 6.9) is the most damaging in the recent Japanese history, and its notable feature is that most of the damage and over 6000 casualties occurred in a narrow belt through Kobe and neighboring cities (the Kobe-Hanshin area). Usually a fault-rupture propagation generates strong ground motions along the direction of propagation, and this directivity effect should be significant above the source faults and beyond its leading end. However, the damage zone of the Kobe earthquake is migrated noticeably from the fault trace into the center of Kobe city. Three-dimensional (3-D) simulations of ground motion in Kobe show that this migration is due to the strong amplification and ray bending in the sedimentary basin below the city in cooperation with the multipathing effects at a basin/bedrock boundary.

FDM Simulation of Seismic Waves, Ocean Acoustic Waves, and Tsunamis Based on Tsunami-Coupled Equations of Motion

We have developed a new, unified modeling technique for the total simulation of seismic waves, ocean acoustic waves, and tsunamis resulting from earthquakes, based on a finite difference method simulation of the 3D equations of motion. Using the equilibrium between the pressure gradient and gravity in these equations, tsunami propagation is naturally incorporated in the simulation based on the equations of motion. The performance of the parallel computation for the newly developed tsunami-coupled equations using a domain partitioning procedure shows a high efficiency coefficient with a large number of CPU cores. The simulation results show how the near-field term associated with seismic waves produced by shallow earthquakes leads to a permanent coseismic deformation of the ground surface, which gives rise to the initial tsunami on the sea surface. Propagation of the tsunami along the sea surface as a gravity wave, and ocean acoustic waves in seawater with high-frequency multiple P-wave reflections between the free surface and sea bottom, are also clearly demonstrated by the present simulations. We find a good agreement in the tsunami waveform between our results and those obtained by other simulations based on an analytical model and the Navier–Stokes equations, demonstrating the effectiveness of the tsunami-coupling simulation model. Based on this simulation, we show that the ratio of the amplitude of ocean acoustic waves to the height of the tsunami, both of which are produced by the earthquake, strongly depends on the rise time of the earthquake rupture. This ratio can be used to obtain a more detailed understanding of the source rupture processes of subduction zone earthquakes, and for implementing an improved tsunami alert system for slow tsunami earthquakes.

Regional Wave Propagation from Mexican Subduction Zone Earthquakes: The Attenuation Functions for Interplate and Inslab Events

The seismic waves from subduction zone earthquakes are significantly affected by the presence of 3D variation in crust and upper-mantle structure around the source area. These heterogeneous structures also profoundly modify the character of seismic waves as they propagate from the source area to regional distances. This is illustrated by studying shallow, interplate earthquakes along the Mexican subduction zone, and deeper, inslab, normal-faulting earthquakes in the subducted Cocos plate beneath Mexican mainland. The strong-motion recordings of these earthquakes are used to evaluate the character of wave propagation along the path between the source region and Mexico City. We compare the wavefield from two large earthquakes of different source type. During the shallow ( H = 17 km), interplate, 1995 Guerrero earthquake ( M w 7.3), the Lg phase is the most prominent feature at regional distances of about 150 to a few hundred kilometers from the source. The presence of a lateral velocity gradient in the crust, caused by the subduction of the Cocos plate, enhances the Lg -wave amplitude, which is then amplified further in the Mexican volcanic belt by amplification in the low-velocity volcanic rocks. Both effects lead to very large ground motions along the path from the coast to the Mexican inland, in the frequency band from 0.2 to 4 Hz. However, for the deeper ( H = 40 km), inslab, normal-faulting, 1999 Oaxaca earthquake ( M w 7.5), the amplitude of the Lg phase is too small to produce the abnormal wave propagation, and the direct S wave and its multiple SmS reflections between the free-surface and Moho show a simple attenuation with increasing distance. We compare these observations with numerical simulations of seismic-wave propagation using the Fourier spectral method. The results provide a key to the understanding of seismic-wave field generated by shallow interplate and deeper inslab earthquakes in a realistic 3D heterogeneous structure.

Parallel simulation of strong ground motions during recent and historical damaging earthquakes in Tokyo, Japan

Abstract The development of high-performance computing facilities such as the Earth Simulator supercomputer and the deployment of dense networks of strong ground motion instruments in Japan (K-NET and KiK-net) have made it possible to directly visualize regional seismic wave propagation during large earthquakes. Our group has developed an efficient parallel finite difference method (FDM) code for modeling the seismic wavefield and three-dimensional visualization techniques, both of which are suitable for implementation on the Earth Simulator. We will show examples of current state of the large-scale FDM simulations of seismic wave propagation by using the Earth Simulator to recast strong ground motions during damaging earthquakes in Tokyo such as the 2000 Tottori-ken Seibu ( M J 7.3) earthquake, the 1923 great Kanto earthquake ( M 7.9), and the 1855 Ansei Edo ( M 7) earthquake. Significant speed-up is achieved using 64–1406 processors of the Earth Simulator with good vector performance of over 40–60% of the theoretical peak speed.

Chapter 7 A Scattering Waveguide in the Heterogeneous Subducting Plate

The subducting plate is an efficient waveguide for high‐frequency seismic waves. Such effects are often observed in Japan as anomalously large ground acceleration and distorted pattern of seismic intensity extending along the eastern seaboard of the Pacific Ocean from deep earthquakes in the Pacific plate, and Kyushu to Shikoku region from deep events in the Philippine Sea plate. Seismograms in these high intensity zones show low‐frequency (f 2 Hz) later arrivals with a very long coda. Such observations are not explained by a traditional plate model comprising just high wave speed and low attenuation material in the slab. A new plate model that can produce such guided high‐frequency waves is characterized by multiple forward scattering of seismic waves due to small‐scale heterogeneities within the plate. The preferred model requires anisotropic heterogeneity of elongated properties in the subduction slab with longer correlation distance (10 km) in the plate downdip direction and much shorter correlation distance (0.5 km) across the plate thickness. The standard deviation of P‐ and S‐wave velocities and density from average is 3%. Such a quasi‐laminated structure in the plate, which is equivalent to random distribution of anisotropic heterogeneities of elongated properties in parallel to the plate surface, can guide high‐frequency signals with wavelengths shorter than the correlation distances along the plate. In contrasts, low‐frequency signals with longer wavelength are not affected by the small scale heterogeneities and travel through the heterogeneous plate as a forerunner of the scattering signals. The high wave speed property of the plate and a strong velocity gradient from the center to the outer part of the plate due to the thermal regime allows low‐frequency (f = 0.3–0.5 Hz) seismic waves to escape into the surrounding, low wave‐speed mantle by refraction of seismic waves. The net result is a frequency‐dependent waveguide in the subducting plate with efficient guiding of high‐frequency (f > 2 Hz) signals by multiple forward scattering and loss of intermediate frequency (f = 0.3–0.5 Hz) signals due to internal velocity gradients. Very low frequency signals (f < 0.15 Hz) with wavelength larger than the plate thickness are not significantly affected by the presence of the plate. We demonstrate the presence of the frequency selective wave propagation effect from comparisons of observations from deep earthquakes that occurred recently in the Philippine Sea plate and in the Pacific plate. A good representation of the behavior of scattering waveguide is provided by 2D finite‐difference calculations for seismic waves using heterogeneous slab models. The results of the simulations demonstrate that the frequency dependency of the models is quite sensitive to the thickness of the plate, and also depends on the scale lengths of heterogeneity distribution in the plate.

Simulation of strong ground motions caused by the 2004 off the Kii peninsula earthquakes

Strong ground motions caused by the Mj 7.4 2004 earthquake that occurred in the Nankai Trough to the southeast of the Kii Peninsula, Japan are simulated by a three-dimensional (3D) finite-difference method (FDM) using a fault-rupture model obtained by inversion of teleseismic seismograms and a 3D subsurface structure model for central Japan. Through simulations of the foreshock (Mj 7.1), the structural model is refined by comparison with observations, and the modified model is used to simulate the mainshock. The simulation provides a reasonable reproduction of the ground motions caused by the mainshock, including site amplification effects in the sedimentary basins of Osaka and Noubi. However, the current simulation model has limitations in producing the large and extended ground motion due to long-period Love waves in the Kanto Plain, as the model does not account for the sharp frequency selectivity for Love waves in the surficial structure of the Bouso Peninsula. It therefore appears necessary to develop a better model for longer-period waves.

Parallel 3-D Simulation of Ground Motion for the 1995 Kobe Earthquake: The Component Decomposition Approach

A new approach to parallel pseudospectral simulation of 3-D seismic wave propagation is developed based on component decomposition of the wavefield. Field quantities and equations of motion are distributed over three processors according to their relationship to the x-, y- and y-coordinates, and computation is carried out concurrently on each processor with inter-processor communications. The efficiency of this approach is evaluated by a theoretical estimate and actual benchmark computations. We then conduct a 3-D simulation of strong ground motion for the 1995 Kobe (Hyogo-ken Nanbu) earthquake in order to show the feasibility of the parallel pseudospectral method. The results of the simulation demonstrate that the complex 3-D structures of the subsurface medium and source fault greatly affect the strong ground motion on the surface.

FDM Simulation of an Anomalous Later Phase from the Japan Trench Subduction Zone Earthquakes

We investigated the development of a distinct later phase observed at stations near the Japan Trench associated with shallow, outer-rise earthquakes off the coast of Sanriku, northern Japan based on the analysis of three-component broadband seismograms and FDM simulations of seismic wave propagation using a heterogeneous structural model of the Japan Trench subduction zone. Snapshots of seismic wave propagation obtained through these simulations clearly demonstrate the complicated seismic wavefield constructed by a coupling of the ocean acoustic waves and the Rayleigh waves propagating within seawater and below the sea bottom by multiple reflections associated with shallow subduction zone earthquakes. We demonstrated that the conversion to the Rayleigh wave from the coupled ocean acoustic waves and the Rayleigh wave as they propagate upward along the slope of seafloor near the coast is the primary cause of the arrival of the distinct later phase at the station near the coast. Through a sequence of simulations using different structural models of the Japan Trench subduction zone, we determined that the thick layer of seawater along the trench and the suddenly rising sea bottom onshore of the Japanese island are the major causes of the distinct later phase. The results of the present study indicate that for realistic modeling of seismic wave propagation from the subduction zone earthquakes, a high-resolution bathymetry model is very crucial, although most current simulations do not include a water column in their simulation models.