DI21A-1722
Global P-wave tomography of mantle plumes and subducting slabs
There are many volcanoes on the Earth which can be generally classified into 3 categories: island arc volcanoes, mid-ocean ridge volcanoes, and hotspot volcanoes. Hotspot volcanoes denote intraplate volcanoes like Hawaii, or anomalously large mid-ocean ridge volcanoes like Iceland. So far many researchers have studied the origin of hotspot volcanoes and have used mantle plume hypothesis to explain them. However, we still have little knowledge about mantle plumes yet. In this study, we determined a new model of whole mantle P-wave tomography to understand the origin of hotspot volcanoes. We used the global tomography method of Zhao (2001, 2004). A 3-D grid net was set up in the mantle, and velocity perturbations at every grid nodes were taken as unknown parameters. The iasp91 velocity model (Kennett and Engdahl, 1991) was taken as the 1-D initial model. We selected 9106 earthquakes from the events occurred in the last forty years from the ISC catalog. About 1.6 million arrival-time data of five-type P phases (P, pP, PP, PcP, and Pdiff) were used to conduct the tomographic inversion. In our previous model (Zhao, 2004), the grid interval in the E-W direction is too small in the polar regions. In this study, in order to remedy this problem, we use a flexible-grid approach to make the lateral grid intervals in the polar regions nearly the same as the other portions of the mantle. As a result, the tomographic images in the polar regions are remarkably improved. Our new tomographic model shows huge low-velocity (low-V) zones in the entire mantle under Tahiti and Lake Victoria, which reflect the Pacific and African superplumes, being consistent with the previous studies. A clear low-V zone is revealed under Mt. Erebus volcano in Antarctica. Other major hotspots also exhibit significant low-V zones in the mantle under their surface locations. Beneath Bering Sea, we found that the Pacific slab is subducting from the Aleutian trench and it is stagnant in the mantle transition zone. In Bering Sea, there are several intraplate volcanoes such as St. Paul island. Given the existence of the stagnant Pacific slab and very low-V mantle wedge above the slab, we think that the origin of the intraplate volcanoes in Bering Sea is most likely related to the deep subduction of the Pacific slab and its stagnancy in the mantle transition zone, similar to the Changbai and Wudalianchi volcanoes in Northeast Asia (Zhao, 2004). Zhao, D. (2001) Earth Planet. Sci. Lett. 192, 251-265. Zhao, D. (2004) Phys. Earth Planet. Inter. 146, 3-34.
DI21A-1723
Surface Wave Tomography for the Hawaiian PLUME Project and the Seismic Structure of the Hawaiian Swell
During the two-stage seismic component of the Hawaiian PLUME (Plume-Lithosphere Undersea Melt
Experiment) project from January 2005 through June 2007, we collected continuous seismic data at ten land
stations and nearly 70 ocean bottom sites which were occupied with broad-band seismometers. This provides
an ideal basis to analyze surface waves across a broad frequency band to image the crust and mantle of the
Hawaiian swell. In the first OBS deployment phase from January 2005 through January 2006, 35 sites were
occupied in an elongated array centered on the island of Hawaii, with a station spacing of roughly 75~km and
an aperture of 500~km. In the second phase from May 2006 through June 2007, 37 sites were occupied in
a larger array with a station spacing of roughly 200~km.
Our current analysis concentrates on long-period teleseismic Rayleigh waves. During the first phase we
collected records from upward of 95 suitable large, shallow earthquakes with scalar seismic moment
M0≥ 0.015 × 1020~Nm (MS≥ 5.6) or larger and source depths of 200~km or less. We
also identified 70 smaller events with signal levels suitable for analysis. For the second phase, our initial
analysis includes 163 larger earthquakes. We currently have over 5000 unique single-station phase
measurements for the first deployment stage and 2500 for the second. We use this primary phase database
to obtain two-station path-averaged phase velocity curves. These path-averaged dispersion curves are each
well constrained by many earthquakes and are internally consistent between 15 and 50~s, allowing us to
image the lithosphere and upper asthenosphere. Some larger events provide constraints beyond 100~s,
thereby illuminating the lower asthenosphere. Using these dispersion curves we determine path-averaged
depth profiles for nearly 300 two-station legs for the first deployment. The analysis of the second stage has
provided over 100 legs and is still ongoing. We combine these profiles in an inversion for 3-D structure. We
have also begun to determine group velocities, which provide additional constraints and improve depth
resolution of the crust and mantle.
Our analysis reveals a roughly 30~km thick low-velocity anomaly in the lower lithosphere beneath the islands
of Hawaii and Maui that indicates that the lithosphere has undergone some degree of rejuvenation. Deeper
imaged features include anomalously low velocities in the asthenosphere to the west of Hawaii. These results
are consistent with those from the 1997/98 SWELL pilot experiment that covered an area in the
southwestern corner of the PLUME array.
http://igppweb.ucsd.edu/~gabi/plume.html
DI21A-1724
Completion of three-year observation by BBOBS and OBEM arrays probing the stagnant slab
To investigate the stagnant slab beneath the northern Philippine Sea, we had conducted a three-year array observation from 2005 until 2008 by using broadband ocean bottom seismometers (BBOBSs) and ocean bottom electro-magnetometers (OBEMs). It is a key part of the "Stagnant Slab Project" started in 2004 for 5 years as a cross disciplinary project, because of the first direct dense observation to reveal the fine physical structure above the stagnant section of the Pacific slab lies in the mantle transition zone. And, the change of the slab morphology along the Izu-Ogasawara(Bonin)-Mariana arc shown by a global tomography is also an interest to be resolved with high resolution. The experiment was to be conducted in three phases, each consisting of one-year deployment, resulting in a three-year time series data available. The aim is to acquire improved images of the stagnant slab and surrounding mantle in terms of seismic parameters and electrical conductivity. Such images cannot be constructed from existing data. To complete three phases of this experiment, four installation/recovery cruises were required. BBOBSs and OBEMs were installed for two years at most of the sites and for three years at a number of selected sites. Total numbers of BBOBSs and OBEMs deployed were 39 and 37, respectively. Both of BBOBS and OBEM were originally developed by ocean bottom seismology and electromagnetic groups of the Earthquake Research Institute, University of Tokyo. During the phase 1 and 2, all instruments were recovered, indicating the high reliability of them. The first, second and fourth cruises were conducted by using a R/V KAIREI (JAMSTEC), and the third was conducted using a chartered ship, ASEAN-MARU (Dokai MS ltd.). Quality of the seismic data is good, especially in the southwestern part of the area due to the low background noise level. The electromagnetic data is almost perfect. Analyses of the data are still on the way, but the data accumulation for three years should give us fine images of the stagnant slab soon.
DI21A-1725
S-wave tomographic imaging of the mantle beneath the Hawaiian Islands from the PLUME deployments of ocean-bottom and land seismometers
The Hawaiian PLUME (Plume-Lithosphere Undersea Melt Experiment) project is a multidisciplinary program to study the deep mantle structure of the Hawaiian hotspot and address the debate over whether one end of the island chain is underlain by a classical plume from the deep mantle and how mantle upwelling interacts with the overlying lithosphere beneath the Hawaiian Swell. PLUME involved two consecutive ocean-bottom seismometer (OBS) deployments and a concurrent deployment of 10 land seismometers along the islands. The first deployment of 35 broadband OBSs in 2005-2006 was centered on the island of Hawaii with stations spaced about 75 km apart. A second deployment of 38 OBSs, in a pattern with a larger aperture and a station spacing of about 250 km, was carried out in 2006-2007, although the number of seafloor stations on the second deployment that yielded useful data was reduced by instrument malfunction or loss. We present S-wave tomographic images of the regional mantle structure beneath the Hawaiian Islands using the combined data from the PLUME experiment. The OBS horizontal components were oriented from particle motions of teleseismic P waves. High-quality relative arrival times of S waves were measured with the multi- channel cross-correlation method of VanDecar and Crosson. In the ~0.05-0.1 Hz frequency band, we obtained 1191 arrivals during the first deployment and 955 arrivals during the second deployment. Because of the high noise levels on the horizontal components of seafloor instruments at Hawaii, most measurements are from Mw ≥ 6.0 earthquakes, which are moderately well distributed in azimuth. The relative arrival time dataset has been inverted for S-wave velocity models beneath Hawaii using both ray theoretical and finite-frequency methods; results from both approaches display general similarity. The S-wave arrival time data from the second OBS deployment are useful in broadening the region of consideration and extending downward the depth resolution of imaging into the transition zone. Although resolution below the transition zone is limited, the arrival times of SKS phases recorded by the PLUME deployments display azimuthally varying patterns that may be compatible with a broad low velocity volume near ~1,000-1,500 km depth in the lower mantle beneath Hawaii.
DI21A-1726
Imaging Methods and Algorithms in Global Seismic Tomography Using Waveform Data
There remain significant differences between the various global tomographic models. It appears that details of these models depend to some extent upon choices made:
DI21A-1727
Mantle Anisotropy Beneath the Hawaiian Islands from Measurements of Shear-wave Splitting: Results from the PLUME Ocean-Bottom and Land Seismograph Deployments
The fieldwork component of the Hawaiian PLUME (Plume-Lithosphere Undersea Melt Experiment) project consisted of two consecutive one-year deployments of ocean-bottom seismometer (OBS) and land stations, respectively offshore and on the Hawaiian Islands. Thirty-five OBSs were deployed in the first year in a relatively dense array around the modern locus of the Hawaiian hotspot; in the second year, 38 OBSs were deployed over an area extending from west of Kauai to east of Hawaii. Ten portable land stations were operated for a period spanning both OBS deployments. We have analyzed SKS phases recorded by both OBS and land stations for anisotropy-induced shear-wave splitting. Splitting measurements were typically made in the frequency band 0.05-0.1 Hz in order to minimize tilt-generated noise at the low-frequency end and microseismic noise at the high end. Only events with Mw ≥ 6 yielded measurements with adequate precision. Data quality is such that there are about 5 events per station that yield good splitting measurements. Splitting parameters were measured using the stacking technique of Wolfe and Silver [1998]. The geographical distribution of fast-polarization azimuths does not show an obvious signature of a localized center of mantle upwelling and divergence. Fast azimuths are predominantly parallel to the fossil spreading direction (~75°), with a smaller number parallel to the present-day direction of absolute plate motion (-58°). Measured delay times are typically about 1 s or less, although some stations display larger splitting times of 1-2 s. The variability in the delay times across the different stations may indicate differences in either the degree of anisotropy or thickness of the anisotropic lithosphere. Some well- constrained null measurements may provide constraints on the amount of heating and deformation of the lithosphere due to interaction with upwelling mantle.
DI21A-1728
Imaging of Mantle Upwelling and Downwelling by Broadband Ocean Bottom Seismograms
Uniform coverage of the oceanic part of the globe by permanently operated broadband ocean bottom seismographs (BBOBS) is one of major observational goals in global seismology. Our effort in this direction is to deploy large-scale, long-term BBOBS arrays to fill spatial gaps of land-based broadband networks. Our main targets are the regions of mantle upwelling in the south Pacific (7 stations, 2003-2005) and downwelling in the northwestern Pacific (16 stations, 2005-2008). The background spectra based on one-year long records at 28 stations at depths ranging from 1200 to 5800 m represent the most comprehensive dataset of the Pacific seafloor noise. Noise in the tidal band synchronizes almost exactly with the tilt component of theoretical solid Earth tide. Vertical noise at 5 to 20 mHz is dominated by infragravity wave signal, with its peak shifting to the lower frequency side with increasing seafloor depth, a behavior explainable by the hydrodynamic filtering effect. Noise well below the NLNM model occurs at 30 to 100 mHz around the primary frequency of microseisms. Noise as high the NHNM model is ubiquitous in a band of the secondary frequency of microseisms. The BBOBS data have been incorporated into travel time tomography for mantle structure, receiver function analyses for topography of the 410- and 660-km discontinuities and surface wave dispersion studies for uppermost mantle structure. The large-scale Pacific superplume extends from the lowermost mantle to a depth of 1000 km, above which only narrow plumes appear to exist and continue up to the Earthfs surface near the hotspots. The Pacific plate subducted along the Izu-Bonin-Mariana trenches accompanies a low velocity zone at its top down to depths of more than 200 km under the Philippine Sea. The stagnant slab beneath Mariana is an isolated body, well confined in a depth range from 800 to 1000 km. We lastly touch on a future direction of BBOBS seismology.
DI21A-1729
Imaging of Stratified Upper Mantle Anisotropy
Global tomographic models have been improved over years not only by an increase in the number of data but more importantly by using more general parameterizations, now including anisotropy (radial anisotropy and then general slight anisotropy) and anelasticity. The imaging of seismic anisotropy renews our vision of upper mantle dynamics because different physical processes (cracks or fluid inclusions, lattice preferred orientation of crystals, fine layering) give rise to observable seismic anisotropy (S-wave splitting, surface wave radial and azimuthal anisotropy). Surface waves provide an almost uniform lateral and azimuthal coverages, particularly below oceanic areas and are used to image large scale (>1000km) lateral heterogeneities of velocity and anisotropy in the upper mantle (0-660km depth). The interpretation of anisotropy makes it possible to relate surface geology and plate tectonics to underlying mantle convection processes, and to map at depth the origin of geological objects such as continents, mountain ranges, slabs, ridges and plumes. Usually, several different processes create a complex stratification of anisotropy which can be unraveled by simultaneously taking account of effects of anisotropy on body waves and surface waves. The example of stratification of anisotropy beneath the Horn of Africa will be presented.
DI21A-1730
Imaging Earth's Interior Based Upon Adjoint Methods
Modern numerical methods in combination with rapid advances in parallel computing have enabled the simulation of seismic wave propagation in 3D Earth models at unpredcented resolution and accuracy. On a modest PC cluster one can now simulate global seismic wave propagation at periods of 20~s longer accounting for heterogeneity in the crust and mantle, topography, anisotropy, attenuation, fluid-solid interactions, self-gravitation, rotation, and the oceans. On the 'Ranger' system at the Texas Advanced Computing Center one can break the 2~s barrier. By drawing connections between seismic tomography, adjoint methods popular in climate and ocean dynamics, time-reversal imaging, and finite-frequency 'banana-doughnut' kernels, it has been demonstrated that Fréchet derivatives for tomographic and (finite) source inversions in complex 3D Earth models may be obtained based upon just two numerical simulations for each earthquake: one calculation for the current model and a second, 'adjoint', calculation that uses time-reversed signals at the receivers as simultaneous, fictitious sources. The adjoint wavefield is calculated while the regular wavefield is reconstructed on the fly by propagating the last frame of the wavefield saved by a previous forward simulation backward in time. This aproach has been used to calculate sensitivity kernels in regional and global Earth models for various body- and surface-wave arrivals. These kernels illustrate the sensitivity of the observations to the structural parameters and form the basis of 'adjoint tomography'. We use a non-linear conjugate gradient method in combination with a source subspace projection preconditioning technique to iterative minimize the misfit function. Using an automated time window selection algorithm, our emphasis is on matching targeted, frequency-dependent body-wave traveltimes and surface-wave phase anomalies, rather than entire waveforms. To avoid reaching a local minimum in the optimization procedure, we begin by using longer-period waveforms, which the starting model fits reasonably well, and work our way toward shorter periods. As we improve the model and increase the frequency content of the waveforms, we not only improve the fit to the current data used to constrain the model, but also steadily increase the number of time windows selected for inversion, i.e., more-and-more parts of the seismograms are incorporated in the iterative inversion process. We will present a preliminary southern California crustal model determined based upon this approach.
DI21A-1731
S40RTS a new degree-40 shear velocity model of the mantle
We present a new shear velocity model of the mantle based on a new set of teleseismic travel time, surface wave dispersion and normal mode frequencies. The travel time data set includes approximately 500,000 long-period (T>20 s) traveltimes of S and SKS, core-reflections (ScS, ScS2, ScS3, ScS4), and surface reflections along the minor (SS, SSS, SSSS) and major (SSm, SSSm) arcs. The traveltimes have been corrected for Earth's elliptical shape. Long period traveltime corrections are calculated using the a priori model Crust2.0. The surface wave data consists of almost 20 million dispersion measurements (38-300 s) of fundamental mode and overtone surface waves along the minor and major arcs. Crustal effects are corrected for using local eigenfrequencies calculated for Crust2.0. The normal mode dataset consists of a new set of splitting functions for frequencies lower than 3mHz. Variations of velocity in S40RTS are parameterised using 30 vertical splines and spherical harmonics up to degree and order 40. We illustrate S40RTS with maps and cross-sections and we discuss how the model and its resolution matrix can be made available to interested researchers.
DI21A-1732
Punctuality is the Thief of Time
Historically, our knowledge of the Earth is derived from reading onset times of seismic waves. Ray theory has been the main tool to derive 3D, topographic Earth models from such data. Ray theory is valid in global tomography to resolve structures with a size of several hundred km using short period teleseismic waves traveling minimum time paths. But it leads to systematic under-estimation of the predicted delays for long- period body waves. And even for short-period waves, important information is lost by observing only the onset. Heterogeneities in the Earth influence the seismogram, so that waveforms are expected to contain more information than the onsets alone. For late times in the seismogram, waves may be scattered multiple times, and once the waves have traveled a distance in excess of the 'transport mean free path', the concept of a localized sensitivity becomes degenerate, the inverse problem is fundamentally ill-posed, and interpretations of the data are limited to the estimation of statistical properties. For imaging purposes, our best opportunity therefore is to study the change in waveform of early arriving energy, in particular those of body waves. For the inversion we may attempt to linearize the problem using single scattering (Born theory). A scattered wave behaves as Lord Henry in Oscar Wilde's Picture of Dorian Gray: "He was always late on principle, his principle being that punctuality is the thief of time". We could also conclude that punctuality in our experimental set-up (when reading onsets only) robs us of important information. Some less patrician phases of minimax character like PP violate Lord Henry's rule, and do not even possess an onset time. Even for high frequencies, such phases can only be interpreted properly using finite-frequency methods. Centering a cross-correlation window on a body wave phase, or an identifiable surface wavetrain, has the advantage of concentrating the information around the geometrical ray, thus optimizing the measurement for a single-scattering interpretation and yielding a sparse matrix system. One may read the maximum of the correlation function, or even of the uncorrelated trace, or obtain a best fit to either in a quadratic sense. Each experimental method, except those of reading onsets, comes with its own Fréchet kernel. The phase and attenuation kernels are zero on the ray for first arriving body waves such as P or S, whereas the amplitude (focusing) kernels have a maximum on the ray. The width of the kernel can be manipulated by band-pass filtering before correlation. We shall discuss several such kernels and the improvement they provide with respect to ray theory.
DI21A-1733
Regional 3D tomography of the upper mantle using a summed source approach
Global and large scale regional waveform tomography has so far been based on approximate expressions of the wavefield in the framework of normal mode perturbation theory. As powerful numerical methods, such as the Spectral Element Method (SEM), are progressively adapted to waveform tomography and replace approximate computations, we gain significantly in the accuracy with which the effects of strong lateral heterogeneity of different scales can be represented. However, the main drawback is the large increase in computational time, which limits the frequency range as well as the number of iterations that can be performed. Capdeville et al. (2005, GJI) introduced an approach in which the computational time is significantly reduced, by simultaneously computing the wavefield for a large number of events, and comparing it to the correspondingly aligned and summed seismograms. This approach was tested on a global synthetic dataset with added noise, and showed considerable promise. We have implemented and tested this approach in the case of regional tomography, using a regional version of the SEM, RegSEM (Cupillard, 2008), developed for spherical geometry, which uses Perfectly Matched Layers (PML) at the border of the region and includes general 3D anisotropy, Moho and surface topography, ocean bathymetry, attenuation and ellipticity. We show the results of tests of this approach against the standard "single source single station" waveform inversion and discuss various challenges encountered. We present a model of upper mantle structure in southeast Asia, using a dataset of over 100 events observed at 6+ stations, with epicentral distances between 5-40 degrees, and in the period range 300-32 s.
DI21A-1734
Multiple-Frequency SH-Wave Tomography of the Western U.S. Upper Mantle
We estimate the SH-wave velocity and attenuation structures of the western U.S. upper mantel using the dense network of the USArray, and new techniques: we observe a multiple-frequency data set of both travel- time and amplitude anomalies, and interpret these with full 3-D finite-frequency sensitivity kernels. Amplitudes show stronger frequency dependence than travel times. We perform a joint inversion on the measured travel-time and amplitude anomalies, interpreting them in terms of velocity and attenuation heterogeneities. Aside from the expected clear division between the slow, tectonically active region in the west and the fast craton in the east, some smaller velocity anomalies are observed. The subduction along the Cascades at 100-300 km depth shows varying strength, weakest around 45°N. The Yellowstone plume seems to have an origin from 1000 km depth,and the plume seems to be broken by a fast anomaly, which might be a remnant of the Farallon slab. The S velocity model supports the heavy fragmentation of the Farallon slab proposed by Sigloch et al. (Nature Geo, 2008) based on the P model, by observing similar slab breaks as in the P model. We also investigate the benefits of the methodical improvements we use. Amplitude data help to sharpen the edges of narrow velocity heterogeneities in the shallow upper mantle. In general, velocity and attenuation heterogeneities correlate positively suggesting one major (but not the only) common cause for velocity and attenuation anomalies, mostly likely the temperature. The focusing effect from velocity heterogeneity dominates over that of attenuation and must be considered when interpreting amplitude anomalies.
DI21A-1735
Multi-Mode Surface-Wave Sensitivity Kernels and Mode-Ray Duality
I calculate the three-dimensional (3-D) sensitivities of multi-mode surface-wave phase and amplitude measurements to perturbations in seismic velocity and radial anisotropy. The 3-D phase-delay and amplitude kernels are formulated based upon Born approximation in the framework of surface-wave mode summation. The long-period, multi-mode, surface-wave sensitivity kernels show characteristics of multiple-reflected body waves; strictly speaking, they are often associated with families of "rays" instead of a single "ray" due to the finite length of the measurement window; and, the "ray" features do not necessary correspond to standard body-wave phases. I show that Jean's relation can be applied to quantify the "mode-ray duality" of multi-mode surface waves and provide guidelines for measurement-window determination. The 3-D multi-mode kernels can be computed very efficiently for large datasets of long-period measurements using a fast computation scheme. This multi-mode surface-wave approach opens the opportunity for imaging high-resolution structure of seismic velocity and radial anisotropy in the top 1500 km of the mantle.
DI21A-1736
Near-Field Surface-wave Sensitivity Kernels
In global seismic tomography, three-dimensional (3-D) surface-wave sensitivity kernels have been used to improve the resolution of lateral heterogeneities in the upper mantle. To date, Born sensitivity kernels formulated in the framework of surface-wave mode summation are based upon a far-field approximation. In the far-field approximation, calculations of Legendre functions are approximated with an asymptotic representation which may not be valid in regions close to the receiver (or source), especially for long period surface waves. In this work, we go beyond the far-field approximation and compute the exact 3-D surface- wave sensitivity kernels based upon calculations of the exact Legendre function of fractional orders. We show that (1) in regions more than a couple of hundred kilometers away from the receiver (or source), sensitivity kernels computed with a far field approximation agree very well with the exact sensitivity kernels; and therefore can be safely applied in global tomography; and (2) the differences are significant in regions close to the receiver (or source), and it becomes necessary to use the exact sensitivity kernels. This is especially important for regional/local studies where tomography studies often involve inter-station phase- delay measurements and the station spacing is less than several hundred kilometers. Our sensitivity kernels based upon exact computation of fractional-order Legendre functions are valid for near-field applications and can be computed very efficiently.
DI21A-1737
Multiscale Analysis of Mantle P-wave Velocity Models
An important data base for the analysis of the mantle structure and dynamic is the mantle tomographic velocity models. While it is great that many global mantle velocity models are now available based on different or similar datasets and methods, it is importance to compare and analyze these models quantitatively. We have often been puzzled by questions such as how much similarity or difference between the mantle tomographic models? What is behind their similarity and difference? How much of the similarity and difference is due to real signal or artifact? In this study we employed a multiscale model parameterization of spherical triangles to analyze a dozens of P-wave mantle velocity models published in the last 15 years. At each mantle depth range, the laterally-vary velocity variations are decomposed into components in spherical triangles of different scales. The zero-th scale contains the global average, the first scale consists of eight spherical triangles each is 90 degree long along its sides, and each spherical triangles of the k-th scale will be divided into four smaller triangles of the (k+1)-th scale. The hierarchical structure of the spherical triangles allows a wavelet-like localized decomposition of velocity heterogeneities with different scale lengths. This is advantageous for comparing the mantle models because it can avoid the holes in ray coverage that is problematic for spherical harmonics, and it offers a multiscale representation at each location which is lacked in block or grid model parameterizations. We decomposed each mantle P-wave model into the multiscale system of spherical triangles and compared them via multiscale power spectrum and correlation analysis. The analysis has been carried out for different depth ranges, global and selected regions, and weighted by ray coverage. The model patterns and potential artifacts were analyzed in terms of ray coverage and model parameterization. Despite a positive correlation between the overall patterns of the mantle models, the correlation level changes considerably on different scales and depths. While the power spectral and RMS analysis shows the existence of the 410-km and 670- km discontinuity, individual models on some scales show the discontinuities in different depths, indicating a multiscale character of mantle discontinuities. The comparative analysis also resulted in a composite mantle P wave based on the most similar components of the multiscale individual models weighted by ray coverage.
DI21A-1738
On mantle heterogeneity and anisotropy as mapped by inversion of global surface wave data
We jointly invert Love and Rayleigh wave dispersion curves for the Earth's mantle composition, thermal state, P and S wave anisotropy at different locations on the Earth, based on self-consistent thermodynamic calculations. The method consists of four parts: 1. The composition of the Earth is modeled by the chemical system CaO-FeO-MgO- Al2O3-SiO2. Given these parameters and a geotherm (also an unknown), we calculate stable mineral modes, elastic properties, bulk density at the prevailing physical conditions using Gibbs free energy minimisation. Voigt-Reuss-Hill averaging is subsequently emplouyed to compute radial isotropic P and S wave velocity profiles in the elastic limit. 2. Anisotropic P and S wave velocities are determined from the isotropic ones by employing the relations ξ=(Vsh/Vsv)2, φ = (Vpv/Vph)2, η=F/(2A-L), Vs=(2Vsv2+Vsh2)/3 and Vp=(Vpv2+4Vph2)/5. The former three parameters are the standard anisotropy parameters, that we also invert for. 4. From these radial profiles, i.e. of Vsv, Vsh, Vph, Vpv and ρ, sunthetic Love and Rayleigh wave dispersion curves are calculated. The dispersion curves, which comprise fundamental and overtones up to 5th (Love) and 6th (Rayleigh) order have been extracted from global surface wave velocity maps. Given the above scheme, the data are at each location are jointly inverted using a Markov Chain Monte Carlo algorithm, from which a range of compositions, temperatures and radial profiles of anisotropy parameters, fitting data within uncertainties, are obtained. Our method has several advantages over standard approaches, in that no scaling relationships between Vs and Vp and ρ and Vs have to be introduced, implying that the full sensitivity of Rayleigh and Love waves to the parameters Vs, Vp and ρ is accounted for. In this particular study we investigate 5 locations distributed across the globe and reveal mantle chemical and thermal differences at these locations.
DI21A-1739
S-Receiver Functions and the Lithosphere-Asthenosphere Boundary (LAB)
For seismologists the LAB is a very exotic boundary. It is originally not defined by seismic observations. There is broad agreement that the asthenosphere can be imaged as low velocity zone by surface waves. The boundary between asthenosphere and lithosphere, however, cannot be imaged laterally and vertically with this technique with the possible resolution. This is not seen as a problem by those, who define the LAB by an isotherm, since no sharp transition is expected in this case. Wide angle body wave data, from natural source or controlled source experiments have a higher resolution, but no sufficiently dense observations are available. Seismic techniques which use converted waves are now far enough developed to be successful in observing the LAB with high resolution and density. The principle of this technique (called receiver function technique) is that a strong mother phase (e.g. P or S) produces a small converted daughter phase at the LAB beneath a seismic station, which is interpreted. Due to the freely available data from many seismic stations it is now possible to obtain maps of the LAB topography with so far unprecedented resolution. The receiver function technique has the potential to gain the same significance for the lower lithosphere like steep angle seismics for the upper lithosphere. We are discussing global observations of the LAB in tectonically different region obtained from mainly S-receiver functions and compare it with results from other geophysical observational techniques and also with the multiplicity of geodynamic LAB definitions.
DI21A-1740
Imaging the Lithosphere-Asthenosphere Boundary with Scattered Waves
Advances in velocity and attenuation tomography have yielded increasingly clear models of three- dimensional Earth structure, including global constraints on the thickness of the lithosphere. Migration of scattered teleseismic phases such as Ps and Sp provides complementary and often sharper resolution of velocity gradients at the base of the lithosphere. These gradients contain important information on the physical and chemical properties that differentiate the lithosphere from the underlying mantle. For example, very rapid negative velocity gradients could require enhanced water or melt content in the asthenosphere, as has been proposed beneath regions of the northeastern United States and southeastern Canada. We have extended migration of Ps and Sp receiver functions to over 100 permanent broadband stations in North America and 20 stations in Australia. Beneath the Phanerozoic orogens of eastern Australia and eastern North America, Ps data at some stations contain negative phases that migrate to depths (70-110 km) that are consistent with the broad lithosphere-asthenosphere transition imaged in prior surface wave studies. A phase from similar depths is not as clearly imaged by Sp migration, likely due to lower signal-to-noise in Sp waveforms. In the western United States, negative Ps phases indicate a shallow lithosphere-asthenosphere boundary (< 70 km in some locations). Phase amplitudes at a few stations suggest a velocity drop at this discontinuity that is larger than has been observed elsewhere in North America. A phase that could represent a sharp lithosphere-asthenosphere boundary has not yet been unambiguously observed at stations above thick sub-cratonic lithosphere in either Australia or North America.
DI21A-1741
Improved Statistical Processing for Common Conversion Point Stacked Receiver Functions
The interpretation of teleseismic receiver functions is typically limited by poor constraints on the uncertainty of amplitudes of converted phases. In continental regions these problems are overcome by stacking large amounts of data. In oceanic regions, however, data quality is notoriously noisy and the number of events are limited by significantly shorter station deployment times. In order to obtain maximum value from a data set, it is necessary to have estimates of uncertainty. Here we combine a common-conversion point stacking technique with multiple-taper correlation RF estimates that allow frequency domain weighting. We then compute jackknife uncertainties to estimate local uncertainties in RF amplitude. We apply this technique to a continental station in Arabia (RAYN) as a benchmark, and also to the ocean island station at Raratonga, Cook Islands (RAR). The structure we recover matches previous crustal studies at both stations, and provides new interpretations of conversions in the upper mantle. At single stations, this technique works well to resolve crust and mantle structure up to a depth of 100 km. Geographical dispersion of raypaths at larger depths decreases the number of events per bin, and therefore increases the uncertainty in converted amplitude. We therefore propose that this method will be well suited to the analysis of data from seismic arrays.
DI21A-1742
Receiver function imaging in strongly laterally heterogeneous crust: Synthetic modeling of BOLIVAR data
We have generated 2D elastic wave synthetics and from them calculated receiver functions and CCP stacks for an extremely heterogeneous crust and upper mantle model based on an onshore-offshore active source seismic profile acquired as part of the BOLIVAR experiment in the southeastern Caribbean. The goal of the study was to interpret receiver functions from 20 BB stations deployed approximately along the same profile as part of BOLIVAR (Niu et al., 2006). The 500 km long active source model (Clark et al., 2008) acquired along the 64th meridian, has large lateral and vertical velocity variations at all levels, and particularly in the ~350 km wide Caribbean-South American plate boundary zone. The crust thickens by more than a factor of 2 as crossing from the Caribbean plate (~20 km) into the fold and thrust belt on the northern edge of the South American continent (~50km). The upper crustal structure along the profile is complicated by deep, >10 km, irregular sedimentary basins both onshore and offshore in the plate boundary zone. The active source Vp model was gridded at a fine scale (0.25 km) with a shear velocity model calculated using Brocher's (2005) Vp/Vs relationship, and a density model derived from a standard relationship in the same paper. Plane P waves with a center frequency of 0.4 Hz were input from both azimuths at incidence angles of ±10, ±20, and ±30 degrees. The vertical component of motion was deconvolved from the horizontal component using a water level method, the receiver functions were repositioned in space using the average 1D P and S wave velocity model, and were then binned for CCP stacking. Calculating receiver functions using the IASP91 crustal model (Kennett et al. 1991), which differs substantially from the active source average model, provides a poor representation of the crust, with Moho thickness estimates in error by >5 km. A number of interesting observations can be made comparing the field receiver functions with the synthetics. Notably, the density of the fd synthetics shows that several intracrustal and near Moho events in the sparsely sampled field data are basin multiple reflections. Where the Moho has the steepest dip, under the plate bounding strike-slip fault system, the CCP stacks fail to image the Moho well, regardless of the density of spatial sampling. A suitable spatial sampling criterion for imaging the lower crust and Moho is to overlap Fresnel zones by 50% at Moho depth, which in this case requires an instrument spacing of 10-15 km, about 3 times the actual data density.
DI21A-1743
Smooth Crustal Models Derived from Surface Wave Dispersion Data for Waveform Tomography based on the Spectral Element Method
As the frequency band typically considered in global S wave tomography of the earth's mantle does not allow the resolution of crustal structure, but the effect on seismic waves of strong heterogeneity in the crust cannot be neglected, a standard approach has long been to perform approximate crustal corrections, based on normal mode perturbation theory. As we move towards 3D numerical modeling, which can represent the effects of the crust very precisely, we are faced with several issues: 1) uncertainties in the crustal model; 2) heavy numerical costs associated with the thin layer meshing required to accurately compute the effects of the crust; 3) the need to connect the input crustal model, and its large Moho depth variations, with the domain of the tomographic inversion. Several approaches have been proposed to replace layered crustal models by smooth models that predict the same short period surface wave dispersion while significantly decreasing the computation time. Here we have chosen a somewhat different approach, in which, rather than smoothing existing crustal models, we directly develop a smooth 3D crustal model from surface wave dispersion data in the period range 10-100s. A constraint on the model is to properly account for the crustal contribution to the long period waveforms used in our SEM-based tomographic approach. We test our methodology on a dataset covering a large part of the Eurasian continent, which is characterized by large lateral variations in crustal structure.
DI21A-1744
Degree-2 in the Transition Zone and Near the CMB: Bottom up Tectonics?
The 2008 Cooperative Institute for Deep Earth Research (CIDER) program facilitated collaboration between researchers from seismology, geodynamics, mineral physics, and geochemistry to study, model and better understand the interior of the Earth. Through this multidisciplinary approach, we have developed a self- consistent paradigm of mantle structure and dynamics. Geochemical studies necessitate multiple mantle components, a requirement that can be met by a layered mantle structure with the 660-km discontinuity serving as a boundary between a depleted upper mantle and undepleted lower mantle. Seismological studies show strong evidence for reorganization of flow at the 660 km discontinuity, but some tomographic models also suggest a significant mass flux across this depth. We investigate the possibility that the large low-velocity seismic anomalies beneath south Africa and the central Pacific are thermochemical reservoirs that may serve as the undepleted, undegassed mantle end-member commonly seen in intraplate volcanics. These superplumes may represent an extension of the degree-2 heterogeneity dominating the deepest 500-1000 km of the lower mantle, and may comprise >20% of the mantle by volume. A comparison of S-velocity anomalies in the mantle with the slab reconstruction model of Lithgow-Bertelloni and Richards (1998) reveals a high degree-2 correlation between these models; in general, the slab model has much more power in higher harmonics. However, for degree-2, the slab density anomaly integrated over the upper mantle has the same pattern as the velocity anomalies at the bottom of the transition zone, but not at other depths in the upper mantle, suggesting that the transition zone acts as a low-pass filter, preferentially removing shorter wavelengths of mantle flow. The degree-2 velocity anomaly just above the core-mantle boundary (CMB) parallels the last 120 Ma of slab signal integrated over the entire mantle, indicating a long-lived origin of this boundary. The degree-2 anomaly consists of a ring of high velocities circumscribing both superplumes, leading us to speculate that the pattern of the anomalies near the CMB has imposed the primary control on the planform of mantle convection for at least the last 200 Ma. The two chemically and thermally distinct superplumes may be relatively immobile for long periods of time, anchoring mantle plumes and influencing the paths of Wilson cycles.
DI21A-1745
Using Synthetic Tomography to Investigate the use of Tomographically-derived Buoyancy as an Initial Condition to Compute Global Mantle Flow Models
Global mantle flow fields are widely used to understand upper mantle seismic anisotropy, to estimate tectonic driving forces, and for use in multi-scale geodynamical models. Currently, present day mantle flow is calculated from tomography-derived buoyancy fields, which are found by converting a seismic tomography model either to temperature or directly to buoyancy using a linear velocity-to-density scaling. The tomography-derived buoyancy fields are then used as initial conditions to a one-step Stokes Flow calculation which determines the present day mantle flow field. However, due to the heterogeneous nature of seismic ray path coverage in the mantle it is possible that these tomographically-derived buoyancy fields are damped and distorted versions of the true thermal structure, leading to an unevaluated amount of error when these buoyancy fields are used to generate global flow models. In previous work we developed a method to produce predicted tomographic models from 3D spherical isochemical mantle convection calculations. Numerical temperature fields were converted to shear-wave velocity using a linear scaling and the resultant shear-wave velocity fields were subsequently convolved with the resolution operator of the seismic model S20RTS to mimic the damping and distortion associated with heterogeneous sampling of the mantle. In this study we investigate the effect of the heterogeneous nature of tomographic resolution on global mantle flow fields by deriving a global flow field from this isochemical synthetic tomography model. We are interested in the manner that the original and tomographically-derived flow fields deviate from one another in both velocity magnitude and direction. We perform a detailed comparison and error analysis and find that to first order, the flow directions agree well between the models, with angular differences in the range of 0-20◦. There are significant differences in the magnitudes of the flow field between the two models suggesting that tomographically-derived mantle flow fields underestimate the velocity magnitude.
DI21A-1746
Seismic Constraints on Convective Patterns in the Earth's Mantle: a Geodynamicist Shopping List
Both laboratory experiments, numerical simulations, and theoretical work on mantle convection show that the morphology of convective patterns strongly depends on the mantle physical properties and the existence of density heterogeneities. For exemple, depending on the amplitude of the latter, mantle upwellings can be narrow or fat, continous or patchy. Moreover, their evolution and their signatures on the surface of the Earth in terms of volcanism, temperature or geochemistry, can be highly time-dependent. So, with improved resolution of geochemical and geophysical data, both in space and time, we can begin to invert the observations using a geodynamic framework to gather new insights into the structure and evolution of the Earth's mantle. In this quest, seismic data and models are critical since only they can image present-day heterogeneities (morphology, thermal or compositional nature) and present-day flow (anisotropy) deep in the mantle. As exemples, we shall try to decipher the histories of La Reunion and Iceland hot spots, combining new experiments on thermochemical plumes with geological, geochemical and seismological data.