S33A-1071 1340h
Wide-Angle Controlled-Source Receiver Functions
Receiver function (RF) methodology has become an accepted standard for imaging the Moho, upper mantle, and sometimes crustal discontinuities. However, its present applications to teleseismic waves and deep structures contain virtually no means for independent confirmation of RF images by methods of comparable accuracy, and the resulting images are typically validated on the basis of their geodynamical plausibility. Such lack of corroboration could be partly resolved by an application of RF techniques in controlled-source seismology. Recently, we constructed RFs from the recordings from an ultra-long-range refraction profile QUARTZ in Northern Eurasia and showed that the converted P/S waves could be utilized to map the thickness of the sediments along the profile and to estimate their Poisson's ratio. Here, I present a further application of RF techniques to a high-quality wide-angle crustal dataset acquired in 1994 as a part of ACCRETE project in SE Alaska and Coastal British Columbia. The dataset included 1700 km of marine multichannel (MCS) profiling in BC fjords that were also recorded by ~40 three-component Reftek seismographs deployed at 3-5 km spacing on land. From the perspective of converted-wave studies, this dataset provides thousands of RFs recorded from a highly consistent airgun source fired at 50-m spacings. Three-component particle motion analysis and RF deconvolution reveal consistent P/S mode conversions in the first arrivals in nearly all ACCRETE receiver gathers. Particle motion changes from linear (primary P-wave) to circularly-polarized transverse (P/S converted mode) within 100-200 ms milliseconds of the first arrivals. Notably, and similarly to what had been observed in QUARTZ records, the amplitudes of P/S converted modes exceed those of the first arrivals. Although the RFs for the same receiver are generally consistent, notable variations occur, apparently due to the interplay of scattering modes and the effects of crustal velocity heterogeneities. The locations of P/S converting boundaries (or scatterers) were estimated from the time delays and polarizations of the converted waves. These results suggest that, also similarly to QUARTZ, ACCRETE mode conversions take place on the sediment-basement contact which was mapped in the MCS profiling. Thus, identification of mode conversions in the coda of wide-angle first arrivals helps constraining the shallow near-receiver structure. In addition, such analysis provides valuable experimental data on the factors controlling RF consistency.
S33A-1072 1340h
A Wave Equation Migration Method for Receiver Function Imaging
A wave equation based poststack depth migration method is proposed to image the crustal and upper mantle structures using teleseismic receiver functions. By utilizing a frequency-wavenumber domain one-way phase-screen propagator for wavefield extrapolation in the migration scheme, the common conversion point (CCP) stacked receiver functions are backward propagated to construct the subsurface structural images. Synthetic experiments demonstrate the validity of the migration method for a variety of laterally heterogeneous models. The migrated images show considerable improvement over the CCP images in recovering the structural features. The phase-screen propagator migration method proves to be particularly useful for imaging complex structures and deep discontinuities overlain by strong shallow anomalies, because of its capability of handling lateral velocity variations. Influences of several factors on the image quality of the poststack migration are further investigated, including inter-station spacing, noise level of the data, velocity model used in migration, and earthquake distribution (incident direction of source fields). Theoretical derivation and numerical results suggest that both the CCP stacking and the poststack migration of receiver functions need to be designed in a target-oriented way for reliable and efficient imaging, and special consideration on earthquake distribution is particularly required in designing seismic experiments if structures of large dips are present. The proposed wav equation migration scheme is applied to image the Earth's internal structures using a number of dense field data sets collected at many seismic arrays in Asia. The constructed images reveal several interesting subsurface structures of the study regions and synthetic tests indicate that those subsurface features are well resolved by the seismic data. Significant improvements of the image quality demonstrate the great potential and flexibility of the proposed migration method for imaging the Earth's internal structures using seismic data collected in the future dense passive seismic arrays, such as the USArray.
S33A-1073 1340h
Mantle Discontinuities Beneath Japan Determined from CCP Stacking of Receiver Functions
Many seismic velocity discontinuities have been observed in Earth's mantle over the past several decades. Of these discontinuities the 410-km and 660-km discontinuities that mark the upper and lower boundaries of the transition zone exist globally, while others appear to be associated with specific tectonic regimes. For example, the Hales discontinuity at $\sim$80km is found only beneath the continents, and the mid-mantle discontinuity at $\sim$1000km has been reported specifically in the western Pacific. These discontinuities usually reflect the boundaries of layers that are distinct from each other in either composition or structure. With the unprecedented high quality and volume of seismic data recorded by the Japanese borehole seismic network (Hinet) deployed since 2000, we are in an ideal position to characterize the discontinuity structure beneath the Japanese islands, which is a ``type locale'' of a subduction region. In this study we used P-to-S converted waves to image to 1000 km depth beneath Japan. Our dataset includes 7903 three-component seismograms in the distance range of $35\deg$ to $85\deg$ recorded at more than 500 borehole seismometers that cover the Japanese islands. We used 20 earthquakes with Mw$>$6.0 to form the CCP images. Receiver functions were formed by deconvolution of the vertical components from the rotated radial components. To ensure a stable deconvolution from the short-period seismograms, the short-period instrument response was removed to recover the longer-period signals($\sim$6s) before the deconvolution. A revised common-conversion-point stacking technique was employed to enhance the signal-to-noise ratio. We varied the bin size and fixed the number {\it N} of conversion points in each bin to improve the horizontal resolution in densely sampled regions. The value for {\it N} in a given bin, 30 to 40 depending on the signal-to-noise ratio of seismograms, was chosen so that conversions for the 410-km and 660-km discontinuities were clearly visible. The bin size varies between $0.01\deg$ and $1\deg$ with an average of $\sim$$0.5\deg$. For a conversion depth {\it d}, we first calculated the ray path of converted phase {\it Pds} and its arrival time relative to {\it P} by ray tracing the 1D {\it iasp91} velocity model. We then summed the {\it N} seismograms using a 4$^{th}$-root stacking method, which provides a more efficient way to reduce the uncorrelated noise relative to the usual linear stack ({\it n=1}). The 410-km and 660-km discontinuities are clear on the CCP images. The 660-km discontinuity is found to be depressed as much as 40 km beneath southwest Japan and the Izu-Bonin arc, while the 410-km is uplifted by 20 km within the subducted Pacific slab and some of the fast regions in tomographic images. On the other hand, a thinner transition zone with a depressed 410-km and an uplifted 660-km is observed below the subducting slab, suggesting the possible existence of plume-like upwelling behind slabs, usually a slow region in tomographic images. A more quantitative comparison of the undulations of the two discontinuities with 3D velocity models will be presented in order to examine the origins of both the velocity anomalies and lateral variations of the two discontinuities. We observed some conversion at $\sim$560km depth only in the higher velocity regions, suggesting that the conversion may occur at top of the stagnant slab instead of the 520-km discontinuity. We also see some conversion energy below the 660-km discontinuity. These conversions, however, exhibit an intermittent feature rather than a continuous discontinuity.
S33A-1074 1340h
Seismic Imaging of the Corinth Rift Structure (Greece) From Converted Waves Migration of Local Earthquake Recordings
In the last 20 years, different models have been proposed to explain extensional mechanisms of the Corinth Rift (Greece) based on various data information. In these models, seismic information plays a central role both in terms of seismicity and velocity model imaging. In order to improve our knowledge of the seismic velocity structure of the Corinth Rift, we have analyzed converted waves using micro-earthquake recordings of a two-months seismological experiment performed in 1991. We have applied a prestack depth migration, based on a 3D Kirchhoff summation technique. Kinematic ingredients of the Kirchhoff approach are obtained from P and S smooth velocity structures reconstructed from hypocenter and velocity tomography of first-arrival times. Traveltimes locate recorded amplitude on seismograms which are summed with weights. These weights include geometrical spreading along rays and Snell coefficients at the diffraction point. We have analyzed 637 micro-earthquakes recorded at a network of 17 three-component stations deployed in the western Gulf of Corinth. The obtained PS forward-scattered image, which has an higher frequency content than tomographic models, indicates the presence of a strong reflector at 7-9 km depth beneath the rift. This interface may be correlated with the presence at depth of the Phyllades Nappes which is an inherited structure of the Hellenic orogenic belt. In this case, the 7-9 km reflector would be a strong constraint for any geodynamic modeling of the Corinth Rift.
S33A-1075 1340h
P to S scattered-wave imaging of the subducting slabs beneath Japan
Recent developments in seismic observations make it possible to apply scattered wave imaging techniques developed for petroleum exploration, such as the pre-stack depth migration, to investigate the crustal and upper mantle structures with unprecedented detail. In this study, we used P to S scattered waves to image the top 1000 km beneath Japan islands. We are particularly interested in imaging the fine structures related to the descent of the Pacific slab, for example, whether the seismic velocity between the slab and the surrounding mantle changes rapidly or gradually, and how the boundary changes with depth. Our dataset includes 7903 three-component seismograms at the distance range of $35^{\circ}$ to $85^{\circ}$ recorded at more than 500 borehole short-period seismometers from a total of 20 earthquakes with Mw $>$6. We generate receiver functions from the radial components of the rotated seismograms. Ham {\it et al.} (this session) show the CCP stacked images from the same dataset. While the 410-km and 660-km discontinuities are very clearly shown in the CCP images, it is difficult to identify the subducting Pacific slab. The absence of the descending Pacific slab in the image is probably due to the nature of CCP stacking process, which assumes a horizontally layered structure in calculating time corrections that degrades the P to S scattered wave signals associated with the dipping slab. Analysis of the data using diffraction migration, yields clear dipping events. Strong scattering associated with the subducting slabs are found in 2D profiles across southwest Japan, northeast Japan, and Hokkaido. However, the preliminary images show that scattering locations are $\sim$20-50 km below the Wadati-Benioff zone. For simplicity we have used the 1D iasp91 velocity model in back projecting the P to S scattered waves. The 1D reference model appears to misposition the events in the subduction region, which has been reported to be very heterogeneous. It is well known that the absolute depths of migrated events in the images depend strongly on the migration velocity model. For example, our synthetic test shows that if the reference velocity model differs from the true structure by 3%, the subducting slab can be mislocated by about 30 km. As the seismic velocity in the mantle wedge is usually several percent slower than the global average, the observed discrepancy between the scattering locations and the Wadati-Benioff zone could be much smaller, if an accurate 2D migration velocity model is used. A 2D Kirchhoff depth migration of the data is ongoing. The migration is a scalar form of the Rayleigh-Sommerfeld diffraction integral, a far field approximation of the Kirchhoff integral and has be successfully applied to the waveform data collected in PASSCAL experiments such as CD-ROM and the Kaapvaal Seismic Array.
S33A-1076 1340h
Multi-Band Images from CD-ROM: Paleo-Subduction and Modern Basalt Extraction Structures in the Southern Rocky Mountains
The CD-ROM (Continental Dynamics of the Rocky Mountains) seismic experiments targeted two Paleoproterozoic suture zones in the western U.S. along a north-south study corridor that extends from central New Mexico to central Wyoming. Seismic reflection, refraction, and teleseismic measurements were made across the Cheyenne Belt in southern Wyoming, and across the Jemez Lineament in northern New Mexico. The Cheyenne Belt is a profound geologic boundary separating the Archean Wyoming craton from island arc terranes accreted to the proto-continent in the Paleoproterozoic. The Jemez Lineament is a linear trend of modern volcanics extending SW from southern Colorado to Arizona, and also coincides with the southern edge of the suture between Yavapai and Mazatzal Paleoproterozoic island arc terranes. Karlstrom and Humphreys (1998) have speculated that the ancient accretionary boundaries influence Cenozoic tectonism in the western U.S., noting the correlation of NE-SW low velocity upper mantle tomography anomalies with geochemical boundaries and mapped suture zones in the Southern Rocky Mountains. At the Cheyenne Belt, the reflection, refraction, P and S tomograms, and pre-stack depth-migrated receiver function images show crust and upper mantle subduction-collision structures that are inferred to have formed during Paleoproterozic island arc collision with the southern margin of the Wyoming craton. Of particular note are a north dipping, high velocity (+3$%$ in P, +5$%$ in S), slab-like structure and a fragment of imbricated crust imaged with both the P and S tomography and the depth-migrated receiver functions. At the Jemez Lineament the reflection data image a bi-vergent orogen marking the Yavapai-Mazatzal suture in the crust. Refraction data show that under the suture zone the crust thins; upper mantle velocity ($\sim$7.7-7.8 km/s) determined from P{_${n}$ suggests that the upper mantle contains 1$%$ partial melt. In the same upper mantle region P and S tomograms show large magnitude low velocity anomalies (-2.5$%$, -5$%$) that correspond to a series of moderately bright but complicated upper mantle events that extend over a distance of $\sim$200 km and to depths of $\sim$100km in the receiver function images. We have modeled this complex series of events using a stochastic description of the velocity field, and generated finite-difference seismograms that match the observations. We interpret the upper mantle low velocity zone as the source region for the recently erupted basaltic magmas found at the Jemez Lineament along the CD-ROM corridor. We speculate that the paleo-suture zone left from continental accretion acts as a crustal conduit for basaltic magmas to pass through the crust, form sills, and erupt.
S33A-1077 1340h
Using geologic information for constraining near surface velocity estimation
Tomography problems often have a large null space. A standard approach is to apply Tikonov regularization. The ideal roughening operator is the inverse model covariance, but isn't known {\it a priori}. Some initial guess of the geologic structure often is known. From this initial model we construct a non-stationary filter, a {\it steering filter} that tends to create velocity trends similar to our structural model. We apply this technique to the problem of estimating near surface velocity from earthquake arrivals. On two examples we show that the updated velocity model shows more consistency with our geologic model than more standard techniques. We also show how the steering filters can be used to assess model validity.
S33A-1078 1340h
Imaging Lower Mantle Structure by Waveform Migration in 3D Tomographic Models
Lower mantle migration applications have generally used a 1D reference model for computing travel times. Resulting images of reflectors or scatterers are intrinsically biased by inaccuracies of the background model. Given that differential time variations for shear wave phases can range over 5 s over 1000 km dimension regions, failure to account for volumetric velocity heterogeneity can produce substantial apparent reflector topography or mislocation of structures by hundreds of kilometers. Alignment of observations on a reference phase or application of travel time corrections for aspherical structure affecting a reference phase have been used as a first-order approach to correction for volumetric heterogeneity, but these approaches cannot account for significant heterogeneity within the migration volume. We present the first lower mantle migration in a 3D heterogeneous reference structure applied to shear wave paths traversing the deep mantle below the Cocos Plate. We calculate travel time perturbations throughout the scattering medium relative to ScS, our aligned reference phase, which we know must reflect from the core-mantle boundary. Tomographic mantle shear velocity models from Grand (2002) and Hung et al. (2004) are used to compute travel time anomalies on all path segments, computed as perturbations to ray paths for a 1D reference structure (full 3D ray tracing is not attempted in this application because of the mammoth number of travel time calculations involved). Models obtained by this 3D migration are compared with corresponding regional structures obtained by 1D migration (Thomas et al., 2004), stochastic Kirchhoff migration (Hutko and Revenaugh, 2004), and forward modeling of local seismogram stacks (Lay et al., 2004). Given the small differential travel times typically involved in deep mantle, wide-angle migration applications, our results demonstrate that accounting for volumetric velocity heterogeneity is imperative to avoid biased images. Iterative tomography/migration approaches will be required to attain model accuracy, as is the case for exploration seismology.
S33A-1079 1340h
Delay Times From Clustered Multi-Channel Cross Correlation and Simulated Annealing
Several techniques exist to estimate relative delay times of seismic phases based on the assumption that the waveforms observed at several stations can be expressed as a common waveform that has been time shifted and distorted by random uncorrelated noise. We explore the more general problem of estimating the relative delay times for regional or even global distributions of seismometers in cases where waveforms vary systematically across the array. The estimation of relative delay times is formulated as a global optimization of the weighted sum of squares of cross correlations of each seismogram pair evaluated at the corresponding difference in their relative delay times. As there are many local minima in this penalty function, a simulated annealing algorithm is used to obtain a solution. The weights depend strongly on the separation distance among seismogram pairs as well as a measure of the similarity of waveforms. Thus, seismograph pairs that are physically close to each other and have similar waveforms are expected to be well aligned while those with dissimilar waveforms or large separation distances are severely down-weighted and thus need not be well aligned. As a result noisy seismograms, which are not similar to other seismograms, are down-weighted so they do not adversely effect the relative delay times of other seismograms. Finally, natural clusters of seismograms are determined from the weight matrix. Examples of aligning a few hundred P and PKP waveforms from a broadband global array and from a mixed broadband and short-period continental-scale array will be shown. While this method has applications in many situations, it may be especially useful for arrays such as the EarthScope Bigfoot Array.
S33A-1080 1340h
Spectral Reconstruction of Teleseismic Green's Functions and Source Signatures
In both exploration seismic and global teleseismic studies, the signature of the source excitation must be removed from seismograms to interpret the scattered wavefield. Earthquake sources are frequently characterized by unknown and protracted rupture histories which hamper recovery of the teleseismic Green's functions in global seismology. The {\it S}-wave components of the teleseismic {\it P} Green's functions may be approximated by the so-called receiver function, the deconvolution of the {\it P} component of a given seismogram from the corresponding {\it SV} and {\it SH} components. While the receiver function approach has been very successful in constraining the {\it S}-wave velocity structure beneath stations, this technique effectively assumes that there are no discontinuities in the corresponding {\it P}-wave velocity profile. By casting the multichannel convolution problem in terms of logarithms of power spectra, the convolution of several sources on several impulse responses can be represented by a linear system that is sub-rank by one equation. In previous work, we have attempted to resolve this under-determinedness though a statistical constraint that the source signatures are independent as a final equation. However, the practical limitations in the number events we can record over a temporary array deployment mean that this constraint may not be adequately satisfied. In the present work, we demonstrate that an individual source signature can be constrained by considering cross-spectra of two seismogram components recording this event: using the phase of this cross-spectrum, the cross-spectrum of the two Green's functions can be estimated, and the power spectrum of the source signature to be estimated. These additional constraints on source signatures eliminate the necessity of appealing to the statistical independence of sources, and render the system massively over-determined. Using the well-founded assumption that the {\it P}-wave Green's function is minimum-phase, not only can we reconstruct the {\it P}-wave components of the impulse response, but highly accurate {\it SV}- and {\it SH}-wave impulse responses are also retrieved. We will present examples on real and synthetic data demonstrating the efficacy of this reconstruction technique.
S33A-1081 1340h
Waveform tomography at a ground water contamination site: comparison with depth migration
We have previously applied 2D acoustic waveform tomography to surface and VSP seismic data from a groundwater contamination site, where the results were compared with images from a 3D reflection dataset. In this study, 2-D waveform tomograms from the reflection dataset acquired at Hill Air Force Base (HAFB) are extensively compared with depth migrated images from the same dataset. Comparisons of tomograms and depth migrated images from a small subset of the data show good agreement in terms of the structural features identified in both types of images. In terms of shallow seismic imaging, the advantages of waveform tomography over depth migration are that the former can be applied using both direct and refracted waves, eliminating a number of processing steps while achieving resolution scales similar to depth migration. Further more, the images provide quantitative estimates of material property perturbations. The disadvantage of waveform tomography is its computational expense when compared with migration. This restricts waveform tomography to 2D applications at present. Thus far we have applied acoustic waveform tomography to first arrival waveforms from land data as an approximation to the elastic case. To improve the applicability of waveform inversion, we present initial efforts to develop a form of elastic waveform tomography in the frequency-space domain. The frequency domain approach has distinct computational advantages over time domain in 2D applications. The technique is being developed in Cartesian as well as cylindrical coordinates to take the curvature of the Earth into account, in order to allow application to teleseismic data.
S33A-1082 1340h
Inverse Multiple Scattering
In reflection seismology at both the exploration and global scale, a common assumption is that waves recorded at the surface have scattered only once in the subsurface (primaries). We have developed a theory, through the combination of the Lipmann-Schwinger and Bremmer series,that goes beyond this assumption to allow for multiple reflections in the subsurface (internal multiples). Using this theory we construct two sets of common image gathers (CIGs), one containing information from both primaries and multiples and another that contains information only from multiples. The two sets of CIGs can then be differenced to remove contributions to a subsurface image from multiples, or each set of CIGs can be examined separately. Since internal multiples have traveled a larger distance in the subsurface they are more sensitive to the velocity model and the position of reflectors than primaries. I will describe the theory and give synthetic data examples of its application.
S33A-1083 1340h
Depth Migration Comparison to Waveform Tomography from a high Resolution 3D Seismic Dataset
The accuracy of shallow environmental seismic images as input for quantitative groundwater and engineering studies strongly relies on the accuracy and detail of the velocity model used in processing, to map the structural features and predict depth and layer thickness. However, conventional NMO velocity analysis can be non-trivial in shallow reflection profiles, due to coherent noise often present in the data and poor signal resolution, associated with a limited range of offsets. Furthermore, the inherent assumption of horizontal reflectors may be erroneous or break down because of large velocity contrasts and lateral heterogeneities, leading to an incorrect depth profile based on Dix's equation. Different methods exist to improve interval velocity estimates; here we present a comparison of migrated images from a 3D seismic reflection dataset at a contaminated site obtained using conventional NMO velocity analysis, stack, inversion of the NMO velocity field and migration, versus depth migrated images obtained using travel time and waveform tomography velocity models. Inverse methods, in fact, provide a powerful tool to build a relatively complex velocity model in depth that is very close to the migration velocity model, and has the advantages of eliminating a number of processing steps and avoiding the depth conversion approximation. Another advantage over more processing oriented methods is that the velocity field provided is often sufficiently accurate without further fine-tuning through iterative profile migrations, which can be difficult to perform in poor signal resolution datasets. Although pre-stack depth migration is the most accurate and detailed imaging method, post-stack depth migration can often produce an acceptable image if an accurate and detailed depth-velocity model is used and the structures are not extremely complex.
S33A-1084 1340h
Imaging Offsets in the Moho: Synthetic Tests using Gaussian Beams with Teleseismic Waves
Over the past two decades, results from several different approaches indicate that rapid changes in crustal thickness, of the order of 10 km or more, may be common in both modern and relic contraction orogens. Offsets in the Moho, if true, have major geodynamic implications in how differences in crustal thickness are compensated, how stresses are transmitted over a wide region, and rheology of the continental lithosphere. We carry out a sequence of tests in order to understand conditions under which rapid changes in crustal thickness can be reliably imaged by teleseismic body-waves. Using the finite difference method over a 2-D grid, we calculate synthetic seismograms that result from a planar incidence of P-wavefield below the crust. We then image the Moho using a migration scheme based on the Gaussian beam representation of the wavefield. The use of the Gaussian beam approach is particularly advantageous in certain geologically critical cases such as the thrusting of crustal material under mantle rocks. The superposition of high-velocity mantle material over crustal rocks requires special treatment because of the possibility of triplications of the incident and scattered wavefields. However, such complexities can be accounted for in the Gaussian beam imaging. Our preliminary results suggest that Moho offsets, of the order of 10 km in height, can be detected beneath thickened crust of about 50 km assuming adequate station spacing and signal-to-noise ratios. Such an experimental configuration seems attainable in a number of field experiments that are either on-going (such as Project Hi-CLIMB) or being planned in the Himalayan-Tibetan region
S33A-1085 1340h
Illumination Assessment for the Seismic-Array, Applications to the USArray Project
The deployment of Earthscope USArray Project, including both its transportable and flexible components, will greatly expand the opportunity for obtaining the properties of the Earth's interior. However, its system detecting ability is highly dependent on the factors including seismic sources used to illuminate the subsurface, the geometry of the array and the subsurface structures. To optimize the capability of the USArray, directional illumination analysis based on the wave propagation in the known velocity model is a solution. In this research, we developed a quantitative method which can assess the illumination of a specific acquisition system to the targeted subsurface structures. The method is originally built and applied in the exploration seismology where the sources and receivers are usually on the surface and the data are regularly and densely sampled. The illumination study (the acquisition system response) for the USArray project is much more complicated. Several modifications have been taken into consideration. The current method can easily calculate illuminations for different array configurations. The acquisition geometries which can be handled are highly flexible. Both active and passive sources including local and teleseismic sources can be included into calculation. The source can illuminate the target from above, below or from inside the model. The best illumination can be estimated by testing different configurations. The results can be used to assist the design of the flex array, the quality control of data acquisition, or to correct the imaging/inversion results. To demonstrate the method, we calculated the illumination for different array configurations under different sources and velocity models.
S33A-1086 1340h
Separation of forward and free surface back-scattered teleseismic wavefields and source signature estimation using the reciprocity theorem and inverse scattering series
We separate the forward and free surface (FS) back-scattered teleseismic wavefields using the one-way wavefield reciprocity theorem and a FS effect removal algorithm. With the correlation type reciprocity relation among flux-normalized one-way wavefields, the reflection responses can be generated from the flux-normalized transmission responses. We apply the FS multiple removal algorithm to the reconstructed reflection responses, and derived a variation of the algorithm to recover the transmission responses with FS back-scattered waves removed. The FS back-scattered waves can be obtained by taking the difference of the transmission responses before and after the FS effect removal. The method depends on accurate reconstruction of the reflection response from the transmission response. Consistency of the reconstructed reflection response and separated FS back-scattered waves may prove useful in validation of results. Theoretical and numerical tests for 1D crustal structure with a normal incident plane P wave validate the theory. We are working on using this as a method of source signature estimation using a criterion of best performance of the FS effect removal (e.g. minimum-energy criterion). This holds promise as an alternative to conventional receiver function deconvolution as it could simultaneously provide source signature estimation and separation of forward and FS back-scattered wavefields.
S33A-1087 1340h
A Raindrop Model for Passive Imaging
In passive imaging one reconstructs the Green's function that accounts for the wave propagation between two receivers by correlating the waveforms recorded at these receivers. This technique is important because it makes it possible to carry out seismic imaging without using coherent sources. With the deployment of US-Array as part of Earthscope, this opens up the possibility for dense intra-continental imaging without relying on intra-plate earthquakes. Passive imaging has been explained using normal mode theory, or by assuming 1D Earth models. I present an alternative derivation of passive imaging that is based on the simple analogue of raindrops that hit the water surface of a pond. A stationary phase analysis of the random scattering integral shows in a simple way that the correlation of the waveforms recorded at two receivers indeed leads to the correct Green's function. The model gives physical insight into the working of passive imaging. This insight helps to asses the use and limitations of passive imaging. \noindent Snieder, R., Extracting the Green's function from the correlation of coda waves: A derivation based on stationary phase, Phys. Rev. E, 69, 046610, 2004
S33A-1088 1340h
Green's Functions, Source Signatures, and Teleseismic Body Wavefields
The separation of Earth's impulse response or Green's function from earthquake source signature is a canonical problem in seismology and an important prerequisite to the exploitation of scattered waves in structural studies. Receiver functions represent a leading order estimate of the {\it S} contribution to the teleseismic {\it P} Green's function and have proven exceptionally useful in characterizing discontinuous shear structure in studies of the lithosphere and upper mantle. They provide no information, however, on {\it P}-to-{\it P} scattering or, accordingly, compressional properties of the Earth's subsurface. An improved estimate of the teleseismic {\it P} Green's function can be achieved through consideration of its spectral characteristics. Under conditions typical of the real Earth, the {\it P} component can be shown to be minimum phase; thus it may be recovered soley from knowledge of its power spectrum. The minimum phase property can be exploited using both auto- and cross-spectra of multichannel measurements to accomplish a comprehensive source-impulse response deconvolution that affords an improved estimate of the teleseismic {\it P} Green's function and the potential for imaging of short wavelength heterogeneity in compressional properties.
S33A-1089 1340h
Beamlet imaging using local cosine basis and aperture correction for true-reflection imaging
Beamlet imaging using local cosine basis (LCB) is implemented and acquisition aperture corrections are introduced in the local angle-domain (beamlet domain). In beamlet migration, both the source field and the receiver array field are decomposed into LCB beamlets and sunk down to image levels using beamlet propagators. Imaging condition can be applied either in beamlet domain or space domain. The beamlet propagators have localization in both the spatial location and the direction. In addition, beamlet propagators have wide-angle capacity, high-resolution, high-fidelity to amplitude information, and no numerical dispersion and numerical anisotropy, Using the beamlet decomposition (local angle-domain) of Green's function, acquisition aperture corrections for image amplitudes in local angle-domain are formulated and applied to the local image matrix, which is obtained from the imaging condition in local angle-domain (beamlet domain). The resulted images are true-reflection images, whose amplitudes are proportional to the total scattering coefficients of local heterogeneities. The local image matrices can be reduced to common refection-angle image (CRAI) gathers and common dip-angle image (CDAI) gathers. CRAI gathers can be used for local AVA analysis. Numerical examples for a four layers model with curved interfaces and for the SEG-EAGE salt model are shown to validate the approach.
S33A-1090 1340h
Receiver Function Seismic imaging of the Crust and Upper Mantle Using Frequency-Wavenumber Filtering and Regularized Multimode Kirchhoff Migration
Receiver functions provide an indispensable tool for producing discontinuity images of the crust and upper mantle from teleseismic earthquake arrivals. However, image quality can be significantly compromised by instabilities inherent in the deconvolution process and by data coverage irregularities driven by source and/or receiver geometries. We present receiver function estimation and prestack migration techniques which both reduce receiver function deconvolution instability and produce regularized, multimode receiver function images. The receiver function estimation technique exploits frequency-pseudowavenumber domain filtering of gathers to enhance features with consistent and physically allowed moveout characteristics. Synthetic tests confirm excellent recovery of key phases in the presence of high noise levels. To process regularized receiver functions into subsurface seismic scattering potential images, we employ a regularized Kirchhoff migration inversion methodology that migrates both direct and reverberated P-to-S converted modes to their correct subsurface locations. Synthetic studies and associated resolution tests demonstrate that the methodology is well suited to the irregularly spaced station and uneven data coverage common in teleseismic imaging experiments. We demonstrate these techniques using data from the "LA RISTRA" (Colorado PLAteau--Rio Grande RIft--Great Plains Seismic TRAnsect) IRIS PASSCAL experiment to image the seismic structure of southwestern United States lithosphere. Results show that crustal thinning beneath the rift is broadly symmetric about the rift axis, with the thinnest crust (35 km) located directly beneath the rift axis, suggesting a lithosphere that has been extended by pure shear. We observe an upper mantle discontinuity at 250-300 km depth that may correlate with similar discontinuities observed beneath eastern North America. We also observe relatively flat discontinuities at 410 and 660 km depth, indicating the lack of a large scale thermal anomaly beneath the rift at transition zone depths, with vigorous mantle convection confined to less than 400 km.
http://www.ees.nmt.edu/Geop/Ristra/
S33A-1091 1340h
Teleseismic Imaging with Wave Equation Migration
We demonstrate the utility of wave equation migration for lithospheric imaging with teleseismic phases by showing results using from both example synthetic and real datasets. Employing the shot-profile formulation of wave equation migration allows us to independently propagate the source and receiver wavefields using a Fourier domain downward continuation operator. Before propagation, we separate the P-Sv-Sh components of the receiver wavefield results by applying the free-surface transfer matrix. Following separation, we model the source wavefield in one of two ways: (1) the P component is assumed to be a direct measure of the source wavefield as in more traditional receiver function ananlysis, or (2) the source wavefield is modeled as a band limited plane wave with moveout and relative delays similar to that of the direct P arrival. After the choise of source representation and independednt propagation of both wavefields we compare them at each image point by calculating the zero-lag of the cross-corrrelation or deconvolution of the two wavefields. Where the two wavefields are coincident (e.g. at a locus of scattering), the wavefields are similar and produce an image point with non-zero amplitude. Fundamentally, we perform the same operations on the recorded data as in more traditional receiver function and common conversion point procedures (e.g. deconvolution followed by depth migration) except we reverse the order of operations. This allows us to recapture portions of the wavefield not following easily predicted plane wave moveout such as triplicated or diffracted phases leading to a more accurate image.
S33A-1092 1340h
Seismic Anisotropy From the Crust to the Core: Measurements and Interpretations with Large Datasets.
Seismic anisotropy is caused by the alignment of crystals, shapes and layers, and results from depositional and deformational processes. As such, observations of anisotropy offer insights into the dynamical nature of the Earth on a range of length scales. Perhaps the most unambiguous indicator of anisotropy is shear-wave splitting. There are though other techniques for estimating anisotropy, many of which have been developed for oil-industry reflection data. These include the analysis of variations in reflected amplitudes as a function of offset and azimuth, non-hyperbolic travel-time moveout, and converted-wave amplitude ratios. The onset of large semi-permanent seismic arrays (e.g., USArray) will mean that many techniques for estimating anisotropy in industry datasets can be adapted to global seismic datasets. An example involves the passive seismic monitoring of microseismicity, which is used to monitor stress changes in oil-fields. The resulting datasets can be very large. For example, a recent 18-day survey in a North Sea field produced nearly 10,000 earthquake records. Shear-wave splitting analysis on such a dataset cannot be done manually. Instead we have developed automated techniques for measuring shear-wave splitting and applied it to microseismic datasets. More recently, in a global survey of upper-mantle anisotropy we have applied the same automated methodology to teleseismic datasets archived by various data centres (e.g., IRIS). It can be difficult to determine the cause of anisotropy using individual measurements and one method of analysis. In the upper-mantle, anisotropy is primarily attributed to the lattice-preferred-orientation (LPO) of crystals, but there is increasing evidence that other factors, such as melt-alignment can cause anisotropy. Petrofabric analyses using techniques such as electron back scattered diffraction can provide insight into LPO anisotropy. This has been applied to mantle xenoliths and shows that the degree of olivine alignment can be used as a strain indicator. In contrast, such analysis has shown that LPO-anisotropy in the crust is primarily controlled by mica content. However, in shallow crustal rock, preferred fracture or crack alignment also provides an effective means of generating anisotropy. Using a range of methods it is increasingly possible to unravel LPO-induced anisotropy from fracture-induced anisotropy. Similar approaches can be applied to studying the deep Earth. For example, using more than one technique and large datasets allow anisotropy due to melt alignment to be distinguished from that due to crystal alignment. We will show examples of such analyses for both a sedimentary basin setting and the upper-mantle.
S33A-1093 1340h
Adapting Industry Multiple Attenuation Techniques to Crustal-Scale Marine Seismic Surveys
Academic marine seismic surveys often focus on crustal targets situated in areas with deep water and rough topography. Thinly-sedimented seafloor creates strong and late-arriving water-column reverberations, often termed multiples, that can completely obscure deeper primary reflections. In a 2002 survey of the Mariana back-arc, arc, and fore-arc regions, large topographic variations produced strong multiples which were not significantly attenuated by stacking or migration. Using swath bathymetry, collected by the onboard multi-beam sonar system, we adapt industry multiple attenuation tools to extract useable data from below the water-bottom multiple. Standard approaches to multiple removal either take advantage of differences in move-out velocities between primary and multiple arrivals in order to filter out multiples or attempt to model multiples so that they can be adaptively subtracted from the data. Until recently, most modeling tools were restricted to 2D but still performed effectively against the well-behaved multiples often encountered in commercially important areas. But these algorithms have limited effectiveness against multiples generated from 3D structures such as salt domes, so the petroleum industry has recently made a strong push for 3D algorithms. In academic surveys, out-of-plane effects are all-too-often too large for successful application of 2D models, but due to the large regions of interest and budget constraints, 3D surveys are typically out of reach. Surface-Related Multiple Elimination (SRME) is a powerful approach that predicts multiples that reflect at least once off the free surface and can model any multiples that bounce off the surface along the source-receiver line. Developed from 1D theory laid down in the late `70s at Stanford University and extended to 2D in the `80s by Delft University, it has become widely used in commercial hydrocarbon exploration. The version of SRME we adapt to the Marianas survey convolves field shot gathers with synthetics constructed from picked horizons. For us, its advantage over pure SRME is that there is more control over which multiples should be predicted and how they are modeled. In particular, the approach allows us extra flexibility to model out-of-plane multiples. To help adapt this process to the rough topography of the Mariana subduction zone, we use swath bathymetry to allow the hybrid SRME to construct synthetic shot records with out-of-plane as well as in-plane reflections. We also model multiples of the seafloor reflection itself directly using Kirchhoff summation over all possible ray-paths. This latter tool only predicts first-order multiples from the seafloor, but unlike SRME, is not inherently limited by the source-receiver data recording geometry. Finally, we use the relationship between the predicted and actual multiple arrivals to extract information about cross-dips, i.e. out-of-plane, effects.