S11D-01 INVITED 08:15h
Precision Seismology --- Applications and Comments
Since {\em Quantitative Seismology --- Theory and Methods} by Aki and Richards appeared in 1980, its co-authors made major career changes that emphasized {\em applications} rather than theory. The senior author promoted integration of scientific information about earthquakes and its public transfer as the founding science director of the Southern California Earthquake Center. The junior author researched practical issues of nuclear explosion monitoring and subsequently addressed two issues using earthquake data, entailing orders of magnitude increase in precision from previous practice. Both issues have entailed the use of cross-correlation methods to find and analyze similar waveforms from extensive archives. Both are influenced by fine-scale heterogeneities in the deforming Earth. Both have involved contributions from many colleagues. The first issue, concerns observations that a systematic change in travel times has occurred for seismic waves traversing some paths through the Earth's inner core. Though claimed in 1996, and offered as evidence that the inner core has super-rotation, extraordinary claims demand extraordinary evidence. As reported in more detail by J. Zhang et al and Y. Li et al at this meeting, earthquake doublets in the South Sandwich Islands have recently been discovered to provide extraordinary evidence. The second issue, concerns the classic problem of earthquake location. It now appears that relative locations for a significant fraction of all earthquakes (in some regions, a majority) can be obtained with precision of a few hundred meters. Examples will be presented from China, the Charlevoix region of eastern Canada, the New Madrid region of the central U.S., and northern California. These two issues indicate that the successes of precision seismology largely depend upon easily searchable archives, of digital data from stations that have operated at the same site for decades.
S11D-02 08:45h
High Resolution 3-D Travel-Time Tomography Using Controlled Sources and Earthquakes: Application to the Cascadia Subduction Zone
High quality portable and permanent networks in regions where both earthquake and controlled source observations are made offer perhaps the best opportunity for high-resolution tomographic imaging of earth structure at lithosphere scales. This opportunity exists in the seismically active regions of the Cascadia subduction zone as a result of over 30 years of regional network operation, and controlled source experiments over the past 10-15 years such as the "SHIPS" experiment. Aspects of current methodology for non-linear P wave travel-time tomography will be described, including the use of a single fine grid spacing (in the 1-2 km range at regional scale) for both travel-time and inversion computations, a feedback method that regularizes both the final model as well as iterative changes in the model, and problems associated with the use of current 3-D finite-difference (FD) algorithms for travel-time computation in complex structure. Although the methodology may be reaching the limits of current FD travel-time accuracy for complex structure, extraordinary detail is revealed in parts of the model that are well constrained by observational data. Earthquake relocation with 3-D models has not made major changes in our view of seismicity obtained previously with 1-D models; however, there is now a clearer association between crustal earthquakes and the tectonically "strong" structural units of the forearc crust. At the highest structural resolution, clear details of the complex basin and fault structure of the forearc of Washington are apparent. At a slightly lower resolution regional scale (280x303x85 km grid at 1.45 km grid spacing), structure including the subducted Juan de Fuca slab to depths of approximately 70 km is imaged beneath Puget Sound, showing evidence of complex interaction between the forearc crust, the mantle wedge region, and the subducted slab. In this same region, no evidence of a continental Moho exists, consistent with the idea of a serpentinized mantle wedge in this part of the forearc.
S11D-03 09:00h
Imaging Kilauea Volcano Magmatic System Using an Efficient Algorithm for Double-Difference Tomography and Cross-Spectral Time-Delays
We have designed a method based upon double-difference relocation and tomography to image, as accurately as possible, a heterogeneous medium containing seismogenic objects. Our approach has been not only to introduce double-difference in tomography, but rather to partly revisit tomographic schemes for chosing accurate and adapted numerical strategies. We used a finite-difference solution to the Eikonal equation (Podvin and Lecomte, 1991) for travel-time computation and a Tarantola-Valette approach for both the double-difference and the 3D tomographic inversion to find accurate earthquake locations and seismic velocity estimates. Though an efficient estimation of the squared root of the inverse model covariance matrix in the case of the gaussian correlation function, we use correlation length and a priori model variance criteria to determine the optimal solution. Double-difference relocation of similar earthquakes is performed in the optimal velocity model, making absolute and relative locations less biased by the velocity model. Double difference tomography is achieved by using high-accuracy time delay measurements. These algorithms have been applied to earthquake data recorded in the vicinity of Kilauea and Mauna Loa volcanoes for imaging the volcanic structures. They allow to find stable and detailed velocity models : regional tomography allows to retrieve unambiguously the structure of Hawaii island, and double difference tomography shows a detailed image of the southern Kilauea Caldera - Upper East Rift Zone magmatic complex.
S11D-04 09:15h
High-Resolution Inverse-Based Determination of Seismic-Velocity Structure in Basins
Starting with the pioneering work of Aki, Christoffersson, and Husebye in 1976, there has been an increasing interest in developing inversion techniques for determining the three-dimensional crustal velocity structure in seismic regions. In this paper we describe a methodology that capitalizes on recent advances in optimization methods to adapt, extend, and refine powerful nonlinear Newton-Krylov adjoint-based inverse wave propagation algorithms to two- and three-dimensional velocity structure and kinematic source inversion problems. We present results of high resolution models for two-dimensional sedimentary valleys undergoing antiplane motion, and three dimensional acoustic approximations of models of the San Fernando Valley using parallel scalable inversion algorithms that overcome many of the difficulties particular to inverse heterogeneous wave propagation problems.
S11D-05 09:30h
Guided Waves in Subduction Zones and the Heterogeneity Structure of the Subducted Plate
Sharp shocks of high intensity are frequently observed on the eastern seaboard of Japan from deep earthquakes in the Pacific plate. The recordings of such events at stations in the region of high intensity show a low-frequency (< 0.5 Hz) onset for both P and S waves followed by large, high-frequency (> 2 Hz) later arrivals with a long coda. This behavior is not explained by a subduction zone model comprising a just high velocity plate with low attenuation. From the analysis of observed broadband waveforms and numerical simulation of seismic wave propagation in 2-D and 3-D models we demonstrate that the high-frequency guided waves traveling in the subducting plate arise from the scattering of seismic waves by heterogeneity within the plate. Our preferred model of the heterogeneity has elongated scatterers parallel to the plate margin described by a von Karmann function with a down-dip correlation length of about 10 km and much shorter correlation length of about 0.5 km in thickness. The standard deviation of wavespeed fluctuations from the averaged background model is about 2 percent. This new plate model produces the required frequency-selective propagation characteristics through scattering of the high-frequency components. Three-dimensional simulations of the expected propagation characteristics for a deep earthquakes recorded in northern Japan for a realistic plate model display anomalously large intensities and distorted intensity patterns comparable to observed patterns. The simulations employ parallel computation on the Earth Simulator supercomputer with a high-resolution structural model including stochastic heterogeneity in the subducted plate. The simulations demonstrate clearly the scattering waveguide effects for high-frequency waves.
S11D-06 09:45h
A Quantitative Evaluation of 3D Velocity Models in Southern California
We present a systematic methodology for evaluating and improving 3D seismic velocity models using waveform data from regional earthquakes. The operator that maps a synthetic waveform into an observed waveform is expressed in the Rytov form $D(\omega) = \rm{exp}[\rm{i} \omega \delta \emph{t}_{\rm{p}}(\omega) - \omega \delta \emph{t}_{\rm{q}}(\omega)]$. We measure the phase delay time $\delta t_p(\omega)$ and the amplitude reduction time $\delta t_q(\omega)$ as a function of frequency $\omega$ using Gee & Jordan's [1992] isolation-filter technique. We have applied this procedure to CISN recordings of 25 small earthquakes ($3.0 \le \rm{M_L} \le 4.8$) in the Los Angeles region. Synthetic seismograms were calculated from four types of velocity models: the 3D SCEC Community Velocity Model Version 3.0 (CVM3.0) [Magistrale et al., 2000],the Harvard 3D model (HAR3D) [S\"{u}ss and Shaw, 2003], the 1D Standard Southern California Crustal Model (SoCaL) [Dreger & Helmberger, 1993], and a set of path-averaged 1D models (A1D), which were extracted from CVM3.0 by horizontally averaging wave slownesses along source-receiver paths.(The 3D synthetics were computed using K. Olsen's finite difference code.) We measured 165 P waves, 110 SV waves, 171 SH waves and 48 other phases from 284 source-station paths, at five frequencies ranging from 0.2 Hz to 1 Hz, yielding a data set comprising 4940 $\delta t_{pq}(\omega)$ observations. Overall, the two 3D models provided a substantially better fit to the waveform data than either laterally homogeneous or path-averaged 1D models. Relative to SoCaL, CVM3.0 reduced the variance in $\delta t_p$ by 61% to 0.401 $\rm{s}^2$, and reduced the variance in $\delta t_q$ by 64% to 0.122 $\rm{s}^2$. The variance reductions of CVM3.0 relative to A1D were 55% and 53%, respectively. The variance of HAR3D is 0.362 $\rm{s}^2$ in $\delta t_p$ (65% less than Socal and 60% less than A1D) and 0.118 $\rm{s}^2$ in $\delta t_q$ (65% less than SoCaL and 55% less than A1D). The correlation coefficient between CVM3.0 and HAR3D is 0.6 in $\delta t_p$ and 0.8 in $\delta t_q$, suggesting high similarity between the two 3D models. Our measurements indicate that both CVM3.0 and HAR3D wave speeds might be too low in Los Angeles Basin. The variance reductions for S wave measurements are as large as for P wave measurements suggesting the scaling relations relating S wave velocity to P wave velocity in both 3D models are good approximations. Our analysis on 90 SV-SH pairs shows slight polarization anisotropy of about 4 $\rm{ms/km}$ in the upper crust.