S12B-01 INVITED
Spectral Matrix Analysis Method for the Detection of Polarized Wave Arrivals Using Confidence Levels
A new method for detecting the arrivals of linearly or elliptically polarized waves by using a spectral matrix was developed and applied to reflection imaging of the earthfs crust and upper mantle. In this method, parameter values representing the confidence levels of the detection of the arrivals of linearly or elliptically polarized waves are used to image the reflectors. In this paper, the theory for the detection of polarized waves is first described and their detectability is discussed in relation to a synthetic three-component signal. The dependence of detectability on time-window length and the center frequency of the analysis were quantitatively evaluated for this signal. Then, earthquake data for events around Sendai, Miyagi, Japan, with magnitudes (Md) ranging from 2.0 to 5.6 were analyzed and the reflectors imaged by a migration technique based on the confidence levels of polarized wave arrival detection. By assuming P-P and S-S reflections, reflectors due to characteristic changes in the mantle were identified at depths of around 350-700 km, demonstrating the feasibility of using a spectral matrix to detect polarized waves and to image the earth crustfs and upper mantle.
S12B-02
Dual-Frequency Coherence as Signature of Fluid Flow Noise Observed in Costa Rica Subduction Zone
In the CRSEIZE experiment of 1999-2000 (DeShon and others, 2006), an array of 14 "ONR" OBSs were deployed in an array across the forearc of the subduction zone west of Costa Rica. The station spacing was about 15 km and the array was about 35km by 60km. Half the sensors were PMD broadband electrochemical seismometers and half were 1-Hz inertial sensors. At three sites near an out-of-sequence thrust, correlations were observed between signals on collocated fluid flow sensors and seismic noise in the 1-10Hz range (Brown and others, 2005). It appears that this correlated noise is statistically different from the usual noise. In particular, there exists a pattern revealed by dual-frequency coherence calculations. This indicates a harmonic relationship between frequency multiples of about 0.4 Hz. A possible cause for this pattern is the type of fluid flow regime proposed by Julian, 1994. This proposed cause of this fluid flow is creep-induced strain change.
S12B-03 INVITED
Site Effect Assessment of Earthquake Ground Motion Based on Advanced Data Processing of Microtremor Array Measurements
High-resolution near-surface geologic information is essential for earthquake ground motion prediction. The near-surface geology forms the critical constituent to influence seismic wave propagation, which is known as the local site effects. We have collected microtremor data over 1000 sites in Beijing area for extracting the much needed earthquake engineering parameters (primarily sediment thickness, with the shear wave velocity profiling at a few important control points) in this heavily populated urban area. Advanced data processing algorithms are employed in various stages in assessing the local site effect on earthquake ground motion. First, we used the empirical mode decomposition (EMD), also known as the Hilbert-Huang transform (HHT), to enhance the microtremor data analysis by excluding the local transients and continuous monochromic industrial noises. With this enhancement we have significantly increased the number of data points to be useful in delineating sediment thickness in this area. Second, we have used the cross-correlation of microtremor data acquired for the pairs of two adjacent sites to generate a 'pseudo-reflection' record, which can be treated as the Green function of the 1D layered earth model at the site. The sediment thickness information obtained this way is also consistent with the results obtained by the horizontal to vertical spectral ratio method (HVSR). For most sites in this area, we can achieve 'self consistent' results among different processing skechems regarding to the sediment thickness – the fundamental information to be used in assessing the local site effect. Finally, the pseudo-spectral time domain method was used to simulate the seismic wave propagation caused by a scenario earthquake in this area – the 1679 M8 Sanhe-pinggu earthquake. The characteristics of the simulated earthquake ground motion have found a general correlation with the thickness of the sediments in this area. And more importantly, it is also in agreement with the seismic intensity anomalous areas revealed by intensity investigations for the more recently 1976 Tangshan earthquake.
S12B-04
Panel Deconvolution of Receiver-Function Gathers: Improved Images via Cross-Trace Constraints
A method is presented for deconvolving collections of receiver functions at individual seismic stations. Conventional deconvolution trades off signal enhancement (through stacking) with preservation of directional variation. I replace stacking of traces with minimization of directional derivatives, thus communicating information between traces without stacking, and recovering the minimum directional variation required by the data. Directional derivatives are defined by triangulation of the incident horizontal slownesses, computing derivatives at the center of each triangle. Deconvolution is as a simultaneous inversion for all traces, incorporating directional derivative minimization as cross-trace constraints. The constraints share frequency content between traces, while preserving directional variation. Resulting receiver function gathers have greater inter-trace coherence, lower noise levels and higher usable frequencies than standard methods. A decade of teleseismic data at station GAC (Quebec, Canada) are deconvolved using conventional and panel deconvolution; the new technique produces a coherent receiver-function image and reveals complex crust and mantle responses.
S12B-05
n Approximation to True Amplitude Limited-Aperture Migration: Semblance-Wei ghted Diffraction Stack
Prestack depth migration has become a regularly applied processing step within the flow and a quickly expanding method for reliable velocity model building. It has become an important tool for studying the nature of reflecting boundaries in the Earth's crust. In this contribution we propose a modification of the Kirchoff depth migration (diffraction stack) algorithm to include a weight fun ction on the amplitude part of the diffraction stack algorithm. For each receiver this weight is exclusively a function of the energy and the direction from which this reflected energy reaches the receiver. The weight function reduces the diffraction stack to a weighted stack of the amplitudes at a given travel time to every point along a corresponding isochron. In comparison with other published weight functions that disregard migration aperture, the work presented here develops a weighting function (taper) which is a function of the recorder data. This weight function is derived from the semblance of the slant stack data. The semblance of the slant stack for a particular offset (the receiver offset) represents the total amount of energy that reaches a particular receiver with specific ray parameter (ie. direction of propagation of the seismic energy). These scheme is data dependent and therefor is a true amplitude weight function. Most of the published weight functions are model dependent, obtaining the weight, the obliquity factors, by estimating the incident, reflection, refraction angles from the ray path through the model. The availability of very fast supercomputers has turned PSDM to a conventional processing step used routinely in an iterative loop for velocity model building and interpretation objectives prior to drilling. For example, this is the case for the determination of the geometry of the flank of salt domes or other reservoirs limited by nearly vertical structures. The scheme proposed here as it is data dependent decreases the number of iterations for the PSDM, as the direction of the imaging reflection is determined from the semblance of the tau-p converted shot gathers. This scheme, which is presented for the 2-D case can be easily extended to 3-D.
S12B-06 INVITED
Pg-pPg Time Delays from Sparse Networks Using the Time-Frequency Correlation Method
The arrival times of primary phases generally produce poor estimates of focal depth, particularly for shallow events. While depth phases may be detected at teleseismic distances, no reliable methods currently exist for identifying regional depth phase time delays, such as Pg-pPg. Scattering from crustal heterogeneities obscures the Pg-pPg time delay by producing multiplicative noise within the Pg time window. The theory of Time Reversed Acoustics (TRA) states that the autocorrelation of Pg time windows produces a clear sidelobe at the Pg-pPg time delay. Using large Pg time windows (i.e. including more scattering) and stacking the autocorrelations from an array of receivers improves the reconstruction of the sidelobe at the Pg-pPg time delay. In this paper, we expand the TRA concept to develop the Time-Frequency Correlation (TFC) method for measuring Pg-pPg time delays, which incorporates signal-processing techniques used in Sonar and Radar applications. The TFC method applies a 2D correlation function in time delay and frequency delay to the analytic representation of each Pg time window. Stacking the 2D correlation functions better identifies the sidelobe at the Pg-pPg time delay. Tests of the TFC method on synthetic Pg time windows provide guidance in detecting the Pg-pPg time delays for events with different source time functions, focal depths, and scattering distributions. We apply the TFC method to measure Pg-pPg time delays for 33 local earthquakes from the Southern California Earthquake Data Center (SCEDC) catalog. We include only catalog events with the most accurate locations ('A'), catalog depths between 2 and 16 km (±2 km), and magnitudes between 3.0 and 6.0. For each event, the TFC method uses a station array that is sparse (N < 20), narrow aperture (Δθ < 90°), and located greater than 300 km from the catalog epicenter. The Pg-pPg time delays are converted to focal depths assuming vertical propagation within the SCEDC 3D model at each event's epicenter. In general, the focal depths from the TFC method are strongly correlated with the catalog depths (R2 ~ 0.9).
S12B-07 INVITED
Complexities of the Wavefield in Subduction Zones
Models of subduction zones have a potential of impacting thinking about many of the central problems in the structure, dynamics, chemistry, and history of the solid earth, cummulative effects of which would reach across the earth sciences. The long-term goal of this work is to develop adaptive seismic imaging concepts for modeling and understanding of the subduction zone processes, taking into account the modification of the propagating seismic field due to reflection, refraction, and scattering from complex geological structures such as slabs, low velocity layers , scatterers, and topography. Different from the relatively small-scale approaches employed so far in reflection seismology, our approach provides the resultant field from deep geological structures, and surface topography spanning hundreds of kilometers, incorporating the signatures of the large-scale features typical in subduction zones. A part of the objective consists of obtaining a fundamental understanding of the 3D structure such as the high velocity slab, through the development of efficient physically based propagation and scattering models. This includes features such as aspect-dependent scattering objects, geological structures, and topography. Moreover, the analysis addresses the resultant physical mechanisms such as waveguide multipath effects produced by wave-trapping slabs. Additionally, the effort aims to analyze the existing experimental data and develop novel signal processing techniques for spatial, temporal, and spectral identification of different classes of waves, including early slab arrivals. The initial phase consists of the development and testing of the modeling and simulation work to explore the underlying physical mechanisms of seismic propagation through slabs beneath the ocean floor, scattering from complex structure and closed boundary objects, and the prediction of the resultant field to the surface receivers. For the representation of range-dependent propagation, we employ a hybrid coupled wavenumber integration approach to range-dependent seismic modeling based on the OASES environmental modeling framework. The generic model is a suite of modules covering a variety of range-dependent waveguide, topography, and source/receiver representations. Furthermore, the representation and the analysis of scattering from arbitrarily shaped structure as well as from closed boundary scatterers is performed by developing a wave theory based computational model combining a virtual source approach to open and closed boundary scattering with an established range- dependent wavenumber integration model for propagation. The current subduction zone modeling has shown that for specific receivers, early arrivals are a manifestation of the slab-trapped traveling waves reaching the surface first, while the rest of the response follows significantly later - a concept which will be further explored in the actual data. The spaciotemporal and frequency characteristics of data recorded by the Japanese seismic network is analyzed and compared to the numerical model. Time-frequency, array processing, and event classification methods for extracting properties of slab structure and scatterer signatures are developed and implemented on the data as well as on the synthetics. The model and experimental data time-frequency analysis exhibits frequency dependence and intricate time of arrival features. The understanding of these, and other manifestations of the underlying physical processes, are further explored and analyzed using signal processing techniques.
S12B-08 INVITED
Eikonal tomography for earthquake data: surface wave azimuthal anisotropy in the western US
In principle, seismic anisotropy provides powerful constraints on the deformation of crust and upper mantle. In recent years, numerous models of azimuthal anisotropy from surface wave studies have been interpreted to explain lithospheric and asthenospheric dynamics. The frequency dependence of the azimuthal anisotropy of surface waves yields information on both the lateral and vertical distributions of anisotropy. However, there is a trade-off between azimuthal anisotropy and isotropic wave speeds in traditional surface wave tomography based on large matrix inversions, which typically involve regulation, damping and smoothing. The trade-off potentially biases the direction and amplitude of azimuthal anisotropy. To overcome this problem, we use a new method of surface wave tomography based on the Eikonal equation to obtain surface wave azimuthal anisotropy. The Eikonal equation states that the gradient of a phase travel time surface constrains both the local phase speed and the direction of wave propagation when the amplitudes of seismic waves vary smoothly. In the western US, we have collected surface waves from more than 200 regional and teleseismic earthquakes with magnitudes larger than 5.0, which are recorded on the Earthscope/USArray Transportable Array (TA). We construct phase travel time surfaces for Rayleigh waves following each earthquake. We find that the variations of surface wave amplitudes are smooth compared to those of surface wave phases, which justifies applying the Eikonal equation in surface wave tomography. For each geographic location, we measure azimuthally dependent phase speed based on the phase travel time surface from each earthquake. Assembling results from all earthquakes, we statistically estimate isotropic phase speeds, azimuthal anisotropy, and their uncertainties at periods from 25 to 100 sec across the entire western US. Surface waves at these periods are mainly sensitive to a depth range from the crust to ~150 km. The resulting azimuthal anisotropy maps are comparable to those from ambient noise Eikonal tomography at overlapped periods between 25 and 40 sec.