S21C-0712
Influence of body-wave velocity characteristic on the seismic data interpretation in TI media with arbitrary spatial orientation
TI media with arbitrary spatial orientation (ATI) is the actual anisotropic model used to describe tilted PTL (period thin layers) and no vertical fractured rock. We devote our paper to study the Influence of body-wave velocity characteristic on the seismic data interpretation in the ATI media. Based on the method of coordinates transformation for the TI media with arbitrary strength of anisotropy and arbitrary spatial orientation, we present the characteristic of body-wave velocity with incident angle and azimuth variation in the ATI media. The result shows that patterns of body-wave velocity are fixed relative to the symmetry axis, and these fixed patterns are closely relative to Thomsen's anisotropic parameters. Body-wave velocity characteristic depends just on the angle between the propagation direction and the symmetry axis. Therefore, with variations of spatial orientation of the symmetry axis and the surveying line azimuth, patterns of body-wave velocity have a variety, and have some symmetries, gradual changes and repetitions. The result is useful to processing and explaining of seismic data.
S21C-0713
An Integrated Crustal Dynamics Simulator
Numerical modelling offers an outstanding opportunity to gain an understanding of the crustal dynamics and complex crustal system behaviour. This presentation provides our long-term and ongoing effort on finite element based computational model and software development to simulate the interacting fault system for earthquake forecasting. A R-minimum strategy based finite-element computational model and software tool, PANDAS, for modelling 3-dimensional nonlinear frictional contact behaviour between multiple deformable bodies with the arbitrarily-shaped contact element strategy has been developed by the authors, which builds up a virtual laboratory to simulate interacting fault systems including crustal boundary conditions and various nonlinearities (e.g. from frictional contact, materials, geometry and thermal coupling). It has been successfully applied to large scale computing of the complex nonlinear phenomena in the non-continuum media involving the nonlinear frictional instability, multiple material properties and complex geometries on supercomputers, such as the South Australia (SA) interacting fault system, South California fault model and Sumatra subduction model. It has been also extended and to simulate the hot fractured rock (HFR) geothermal reservoir system in collaboration of Geodynamics Ltd which is constructing the first geothermal reservoir system in Australia and to model the tsunami generation induced by earthquakes. Both are supported by Australian Research Council.
S21C-0714
Homogenization of the SH-Wave Equation in 2D Heterogeneous Media
In the Earth, seismic waves propagate through 3-dimensional heterogeneities characterized by a large variety of scales, much smaller than their minimum wavelength. Computing the wavefield in such media with the use of heavy numerical methods, leads to high calculation costs. To lower these latter, but also to obtain a better geodynamical interpretation of tomographic images, we aim at calculating convenient effective properties of heterogeneous media, by deriving appropriate upscaling rules for the wave equation. To progress towards this goal, we extend the successful work of Capdeville and Marigo (2007), from 1-D to 2-D; basically, we apply the so-called homogenization method -based on a two-scale asymptotic expansion of the field variables-, to model wave propagation in 2-D periodic media. These are characterized by short-scale variations of elastic properties, compared to the smallest wavelength of the wavefield. Seismograms are obtained using the 0th-order of this asymptotic expansion, at a lower computational cost. They are in good agreement with reference solutions calculated with spectral elements simulations, at least in the bulk of the medium. We finally suggest an extension of the homogenization of the wave equation, to 2-D nonperiodic, deterministic media.
S21C-0715
Seismic wave propagation in complex geological media: The hexahedral grid generation problem
Unstructured hexahedral mesh generation is a critical part of the modeling process in the Spectral-Element Method (SEM). We present some examples of seismic wave propagation in complex geological models, automatically meshed on a parallel machine based upon CUBIT (Sandia Laboratory, cubit.sandia.gov), an advanced 3D unstructured hexahedral mesh generator that offers new opportunities for seismologist to design, assess, and improve the quality of a mesh in terms of both geometrical and numerical accuracy. The main goal is to provide useful tools for understanding seismic phenomena due to surface topography and subsurface structures such as low wave-speed sedimentary basins. Our examples cover several typical geophysical problems: 1) "layer-cake" volumes with high-resolution topography and complex solid-solid interfaces, and 2) models with an embedded sedimentary basin. We discuss the impact of the design of the mesh on the simulation results.
S21C-0716
Propagation of the Near-Field Point Source Finite-Difference Seismograms to Teleseismic Distances Using Propagator Matrices
In this study we present an approach to model teleseismic P waves by interfacing near-field point source seismograms computed using a parallelized finite-difference (FD) algorithm for a localized three-dimensional earth model with the one-dimensional teleseismic seismograms calculated using the propagator-matrix technique. Our finite-difference algorithm includes the effect of topography and treats the free surface as a boundary between two media where the upper medium is represented by the Lame's constants set equal to zero and the lower medium by the established geophysical parameters. The free-surface formulation to handle topography is taken from Pitarka and Irikura (1996) in which the free surface perpendicular to the x axis is set just on the grid points and for the surfaces perpendicular to the z and y directions are set half way between two grid points in the z and y directions, respectively. To propagate the near-source wave field to a teleseismic distance, the point-source FD wavefield is stored on a 2D plane underneath the source and transformed to the ω-p domain using a 2D radon transformation. The transformed F(ω,p) response is filtered using a 2D exponential filter around the desired ray parameter and inverse transformed to the time domain. This filtered wavefield is now a down-going wavefield into the lower structure which is propagated through the remaining part of the source crust followed by a convolution with the t* of the medium and propagation through the receiver crust. The method has proved useful in understanding the effects of the topography and laterally varying crustal layers on the initial part of the P-wave seismograms.
S21C-0717
Assessment of a Multidomain Chebyshev Method for seismic wave modeling
Simulations of seismic wave propagation for large areas and realistic 3D Earth models require the use of multi- processor computers in order to obtain a relevant range of frequencies. Pseudo-spectral techniques, like other high order methods, provide highly accurate solutions while reducing the grid resolution. Due to the global character of the discrete wave operator, these techniques suffer from very low computational scalability on massive parallel computers and the wave field may suffer from random non-causal errors. The Overlapping Multidomain Chebyshev Method is particularly well suited for developing efficient and scalable parallel algorithms using a domain decomposition approach. The pseudospectral Chebyshev differential operator is applied to a large number of smaller subdomains and is local to each of them. Only a reduced amount of data needs to be exchanged in order to couple adjacent subdomains using a small overlap region. In the present work we use this method in order to implement a velocity-stress formulation of the elasto-dynamic wave equation in cartesian and spherical coordinates and we assess the obtained solutions in dependence of various parameters and implementation alternatives for timestepping scheme and boundaries.
S21C-0718
Purpose Built Computers for the Wave Equation
The construction of a purpose built computer for the wave equation is presented as a novel and powerful method to improve the rate at with synthetic seismograms are computed. The new hardware of Waves in Linear Motion Analyzer uses field programable gate array logic circuits to compute the finite difference solutions to elastic and acoustic wave equations in two and three dimensions. Examples of problems which this machine can solve are near surface Rayleigh waves, elastic/acoustic reflection profiles, and vertical seismic profiles. The kernel of several inverse problems is the modeling algorithm of the machine. Hence, large scale 3D seismic inversion methods are now feasible.
S21C-0719
Earthquake occurrence in regional-scale fault models with rate- and state-dependent friction
Long-term (~10,000 year) catalogs of simulated earthquakes can be used to address a host of questions related to both seismic hazard calculations and more fundamental issues of earthquake occurrence and interaction (e.g. Ward [1996], Ziv and Rubin [2000, 2003], Rundle et al. [2004]). With a goal of simulating earthquake occurrence in geometrically complex, regional-scale fault networks (such as the SCEC Community Fault Model) at a resolution on the order of 1 km2, we have extended the models of Dieterich [1995] and Ziv and Rubin [2000, 2003] for faults which obey rate- and state-dependent frictional laws to handle fault elements of arbitrary orientation and slip mode. These simulations are computationally very efficient as they use analytic expressions for the nucleation process that include the effects of time-varying normal stress. We will present work on our exploration of how both geometry (fractal roughness, offsets, bends, and other simple but non-planar) and material property heterogeneities affect aspects of earthquake occurrence such as a) frequency-magnitude distribution, b) spatial and temporal clustering, c) multi-segment ruptures, and d) recurrence statistics in our simulations. Interesting results so far include: the surprisingly small effect of random fractal roughness on the frequency-magnitude distribution for single-fault models driven by uniform backslip but a profound effect in such models driven by tapered backslip; contrasting preferential nucleation locations for small and large events; the occurrence of paired large events with time separations of seconds to years in models consisting of two parallel faults separated by an offset; and the transition of the recurrence time distribution for large events from nearly periodic (coefficient of variation of ~0.04) to much more random (coefficient of variation of ~0.9) with the addition of large-scale bends and additional sub-parallel fault strands in models loosely based on southern California.
S21C-0720
Wavelet modelling of broad-band receiver functions
We present a wavelet modelling approach to invert for S-wave velocities from broad-band receiver functions. Taking spline function as the basic wavelet, the broad-band receiver function is decomposed into five resolution scales by Mallat's pyramid algorithm. The linearized least-squares inversion procedure is applied to every resolution scale. The fifth-scale approximation of receiver function is first inverted to recover the slowly varying background velocity variations with respect to a reference model. This solution is then taken as the initial model for fitting the fourth-scale wavelet coefficients of receiver function to further tune the solution to resolve sharper variations. This procedure is iteratively carried out up to the first-scale wavelet coefficients of receiver function. In this manner, the model neighbourhood containing the global minimum is first searched from the coarsest-scale receiver function, and the search gradually focuses on the global minimum by introducing finer-scale information of receiver function. Noise-free synthetic receiver function tests show that wavelet modelling of receiver functions can guide a certain range of initial models to converge to the true velocity distribution. Tests on actual data indicate that wavelet modelling can provide results very similar to those inferred by joint inversion of receiver function and surface wave dispersion.