skip to main content


Title: Accelerating full-waveform inversion using source stacking: synthetic experiments at the global scale in a realistic 3-D earth model
SUMMARY

The spectral element method is currently the method of choice for computing accurate synthetic seismic wavefields in realistic 3-D earth models at the global scale. However, it requires significantly more computational time, compared to normal mode-based approximate methods. Source stacking, whereby multiple earthquake sources are aligned on their origin time and simultaneously triggered, can reduce the computational costs by several orders of magnitude. We present the results of synthetic tests performed on a realistic radially anisotropic 3-D model, slightly modified from model SEMUCB-WM1 with three component synthetic waveform ‘data’ for a duration of 10 000 s, and filtered at periods longer than 60 s, for a set of 273 events and 515 stations. We consider two definitions of the misfit function, one based on the stacked records at individual stations and another based on station-pair cross-correlations of the stacked records. The inverse step is performed using a Gauss–Newton approach where the gradient and Hessian are computed using normal mode perturbation theory. We investigate the retrieval of radially anisotropic long wavelength structure in the upper mantle in the depth range 100–800 km, after fixing the crust and uppermost mantle structure constrained by fundamental mode Love and Rayleigh wave dispersion data. The results show good performance using both definitions of the misfit function, even in the presence of realistic noise, with degraded amplitudes of lateral variations in the anisotropic parameter ξ. Interestingly, we show that we can retrieve the long wavelength structure in the upper mantle, when considering one or the other of three portions of the cross-correlation time series, corresponding to where we expect the energy from surface wave overtone, fundamental mode or a mixture of the two to be dominant, respectively. We also considered the issue of missing data, by randomly removing a successively larger proportion of the available synthetic data. We replace the missing data by synthetics computed in the current 3-D model using normal mode perturbation theory. The inversion results degrade with the proportion of missing data, especially for ξ, and we find that a data availability of 45 per cent or more leads to acceptable results. We also present a strategy for grouping events and stations to minimize the number of missing data in each group. This leads to an increased number of computations but can be significantly more efficient than conventional single-event-at-a-time inversion. We apply the grouping strategy to a real picking scenario, and show promising resolution capability despite the use of fewer waveforms and uneven ray path distribution. Source stacking approach can be used to rapidly obtain a starting 3-D model for more conventional full-waveform inversion at higher resolution, and to investigate assumptions made in the inversion, such as trade-offs between isotropic, anisotropic or anelastic structure, different model parametrizations or how crustal structure is accounted for.

 
more » « less
NSF-PAR ID:
10476594
Author(s) / Creator(s):
;
Publisher / Repository:
Oxford University Press
Date Published:
Journal Name:
Geophysical Journal International
Volume:
236
Issue:
1
ISSN:
0956-540X
Format(s):
Medium: X Size: p. 644-658
Size(s):
["p. 644-658"]
Sponsoring Org:
National Science Foundation
More Like this
  1. SUMMARY

    Accurate synthetic seismic wavefields can now be computed in 3-D earth models using the spectral element method (SEM), which helps improve resolution in full waveform global tomography. However, computational costs are still a challenge. These costs can be reduced by implementing a source stacking method, in which multiple earthquake sources are simultaneously triggered in only one teleseismic SEM simulation. One drawback of this approach is the perceived loss of resolution at depth, in particular because high-amplitude fundamental mode surface waves dominate the summed waveforms, without the possibility of windowing and weighting as in conventional waveform tomography.

    This can be addressed by redefining the cost-function and computing the cross-correlation wavefield between pairs of stations before each inversion iteration. While the Green’s function between the two stations is not reconstructed as well as in the case of ambient noise tomography, where sources are distributed more uniformly around the globe, this is not a drawback, since the same processing is applied to the 3-D synthetics and to the data, and the source parameters are known to a good approximation. By doing so, we can separate time windows with large energy arrivals corresponding to fundamental mode surface waves. This opens the possibility of designing a weighting scheme to bring out the contribution of overtones and body waves. It also makes it possible to balance the contributions of frequently sampled paths versus rarely sampled ones, as in more conventional tomography.

    Here we present the results of proof of concept testing of such an approach for a synthetic 3-component long period waveform data set (periods longer than 60 s), computed for 273 globally distributed events in a simple toy 3-D radially anisotropic upper mantle model which contains shear wave anomalies at different scales. We compare the results of inversion of 10 000 s long stacked time-series, starting from a 1-D model, using source stacked waveforms and station-pair cross-correlations of these stacked waveforms in the definition of the cost function. We compute the gradient and the Hessian using normal mode perturbation theory, which avoids the problem of cross-talk encountered when forming the gradient using an adjoint approach. We perform inversions with and without realistic noise added and show that the model can be recovered equally well using one or the other cost function.

    The proposed approach is computationally very efficient. While application to more realistic synthetic data sets is beyond the scope of this paper, as well as to real data, since that requires additional steps to account for such issues as missing data, we illustrate how this methodology can help inform first order questions such as model resolution in the presence of noise, and trade-offs between different physical parameters (anisotropy, attenuation, crustal structure, etc.) that would be computationally very costly to address adequately, when using conventional full waveform tomography based on single-event wavefield computations.

     
    more » « less
  2. Abstract

    A new azimuthal anisotropy model for the North American and Caribbean Plates, namely,, is constructed based on full waveform inversion and records from the USArray and other temporary/permanent networks deployed in the study region. A total of 180 earthquakes and 4,516 seismographic stations are employed in the inversion to simultaneously constrain radially and azimuthally anisotropic model parameters:,,, and, within the crust and mantle. Thirty‐two preconditioned conjugate gradient iterations have been utilized to minimize frequency‐dependent phase discrepancies between observed and predicted seismograms for three‐component short‐period (15–40 s) body waves and long‐period (25–100 s) surface waves. Modelexhibits complicated variations in anisotropic fabrics underneath the western and eastern United States, especially at depths shallower than 100 km. For instance, the fast axis orientations in modelsuggest the presence of trench‐perpendicular mantle flows underneath the Cascadia Subduction Zone and also follow the strikes of the Snake River Plain, the Ouachita Orogenic Front, and the Grenville and Appalachian Orogenic Belts. The amplitudes of azimuthal anisotropy reduce to around 1% at depths greater than 200 km, and the orientations are subparallel to the global plate motion directions to the east of the Rocky Mountain, except for large discrepancies in central and eastern Canada. At a depth of 700 km, the fast axes change along the trajectory of the Farallon slab underneath the Great Lakes region and Gulf of Mexico, which might indicate the development of 2‐D poloidal‐mode mantle flows perpendicular to the strike of the sinking slab within the uppermost lower mantle. Comparisons between modelwith a western U.S. model from ambient noise tomography and SKS splitting measurements demonstrate a relatively good agreement for the fast axis orientations, considering the usage of different data sets and imaging techniques. However, the absolute magnitude of azimuthal anisotropy in modelmight be underestimated, especially at greater depths, given the poor agreement on the amplitudes of predicted and observed SKS splitting times. At the current stage, the agreement among different azimuthal anisotropy models at global and continental scales is still poor even for the United States with a dense station coverage.

     
    more » « less
  3. Abstract

    Over the past decade, the seismicity rate in the state of Oklahoma has increased significantly, which has been linked to industrial operations, such as saltwater injection and hydraulic fracturing. Taking advantage of induced earthquakes and recently deployed seismometers, we construct a 3‐D radially anisotropic seismic velocity model for the crust of Oklahoma by using full waveform inversion. To mitigate the well‐known cycle‐skipping problem, we use misfit functions based on phase and waveform differences in several frequency bands. Relative velocity perturbations in the inverted model allow us to delineate major geological provinces in Oklahoma, such as the Anadarko Basin and the Cherokee Platform/Shelf. In addition, radial anisotropy in the inverted model reflects deformation within the crust of Oklahoma, which might correlate with sedimentary layering, microcracks/fractures, as well as dominant orientation of anisotropic minerals. The crystalline basement beneath Oklahoma can be inferred from the new velocity model, which enables us to better classify induced seismicity in current earthquake catalogs. Furthermore, synthetic experiments suggest that the new velocity model enables us to better constrain earthquake locations in Oklahoma, especially for determining their depths, which are important for investigating induced seismicity.

     
    more » « less
  4. Abstract

    We present the first continental‐scale seismic model of the lithosphere and underlying mantle beneath Southeast Asia obtained from adjoint waveform tomography (often referred to as full‐waveform inversion or FWI), using seismic data filtered at periods from 20 to 150 s. Based on >3,000 hr of analyzed waveform data gathered from ∼13,000 unique source‐receiver pairs, we image isotropicP‐wave velocity, radially anisotropicS‐wave velocity and density via an iterative non‐linear inversion that begins from a 1‐D reference model. At each iteration, the full 3‐D wavefield is determined through an anelastic Earth, accommodating effects of topography, bathymetry and ocean load. Our data selection aims to maximize sensitivity to deep structure by accounting for body wave arrivals separately.SASSY21, our final model after 87 iterations across seven period bands, is able to explain true‐amplitude data from events and receivers not included in the inversion. The trade‐off between inversion parameters is estimated through an analysis of the Hessian‐vector product.SASSY21reveals detailed anomalies down to the mantle transition zone, including multiple subduction zones. The most prominent feature is the (Indo‐)Australian plate descending beneath Indonesia, which is imaged as one continuous slab along the 180° curvature of the Banda Arc. The tomography confirms the existence of a hole in the slab beneath Mount Tambora and locates a highS‐wave velocity zone beneath northern Borneo that may be associated with subduction termination in the mid‐late Miocene. A previously undiscovered feature beneath the east coast of Borneo is also revealed, which may be a signature of post‐subduction processes, delamination or underthrusting from the formation of Sulawesi.

     
    more » « less
  5. SUMMARY

    We use source-encoded waveform inversion to image Earth’s Northern Hemisphere. The encoding method is based on measurements of Laplace coefficients of stationary wavefields. By assigning to each event a unique frequency, we compute Fréchet derivatives for all events simultaneously based on one ‘super’ forward and one ‘super’ adjoint simulation for a small fraction of the computational cost of classical waveform inversion with the same data set. No cross-talk noise is introduced in the process, and the method does not require all events to be recorded by all stations. Starting from global model GLAD_M25, we performed 100 conjugate gradient iterations using a data set consisting of 786 earthquakes recorded by 9846 stations. Synthetic inversion tests show that we achieve good convergence based on this data set, and we see a consistent misfit reduction during the inversion. The new model, named SE100, has much higher spatial resolution than GLAD_M25, revealing details of the Yellowstone and Iceland hotspots, subduction beneath the Western United States and the upper mantle structure beneath the Arctic Ocean.

     
    more » « less