T33A-1325 1340h
Finite-difference modeling of upper-mantle heterogeneities imaged by Lithoprobe's controlled-source seismic experiments beneath the Precambrian platform of Western Canada
Observations of upper mantle reflectivity at numerous locations around the world have been linked to the presence of a heterogeneous distribution of rock types within a broad layer of the upper mantle. This phenomenon is observed in wide-angle reflection data from Lithoprobe's Alberta Basement Transect (the SAREX and Deep Probe Seismic Experiment of 1995) and Trans-Hudson Orogen Transect (the THoRE experiment of 1993). SAREX and Deep Probe image the Archean lithosphere of the Hearne and Wyoming Provinces, whereas THoRE images the Archean/Proterozoic lithosphere of the Trans-Hudson Orogen and surrounding areas. Finite-difference modeling schemes are applied to constrain the position and physical properties of the reflective layer. SAREX / Deep Probe modeling uses a 2D visco-elastic finite-difference routine; THoRE modeling uses a pseudospectral algorithm. In both cases, the upper mantle is parameterized in terms of two media. One medium is the background matrix; the other is statistically distributed within the first as a series of elliptical bodies. Such a scheme is suitable for modeling: (1) variations in lithology (e.g., a peridotite matrix with eclogite lenses) or (2) variations in rheology (e.g., lenses of increased strain within a less strained background.) Results indicate that the thickness of the heterogeneous upper-mantle layer in western Canada varies. Beneath the Trans-Hudson Orogen in Saskatchewan, the layer is best modeled to lie at depths between 80 and 160 km. Based on observations from perpendicular profiles, anisotropy of the heterogeneities is inferred. Beneath Alberta, 400 km to the west, upper-mantle heterogeneities are modeled to occur between depths of 90 and 140 km. In both cases, the heterogeneous bodies within the model have cross-sectional lengths of tens of km and vertical thicknesses on the order of 1 km.
T33A-1326 1340h
Mantle Fabric Beneath Ontario: Results From the CNSN, FEDNOR and POLARIS Arrays
The basement of the province of Ontario is Precambrian, consisting of the Archean Superior Province in the north and west, which abuts on the Proterozoic Grenville Orogen. The Grenville, which forms the southeastern edge of the Canadian Shield, is the result of extensive crustal shortening and deformation during the interval 1.3-0.98 Ga. The degree to which this crustal deformation is reflected in the underlying mantle is uncertain, though LITHOPROBE detection of a preserved subduction zone (Calvert et al., 1995) in the Superior indicates that relict Precambrian features are preserved below the crust. A number of permanent CNSN stations (e.g., SADO, GAC, KGNO, etc.) are located on or near the Grenville, and have large archives of teleseismic data available; the deployment of the dense POLARIS network on the Grenville in southern Ontario provides higher-resolution constraints, and has more recently been extended by FEDNOR stations in the Superior province. We present preliminary analyses of the FEDNOR data for SKS splitting and crustal thickness, along with a more detailed examination of receiver functions in the Grenville for upper-mantle structure. The SKS results for the Superior Province show fast directions to be largely in-line with absolute plate motion, in contrast with previously-presented results for the Grenville; receiver functions on a transect across Southern Ontario indicate ubiquitous fine anisotropic layering which varies laterally over distance scales of less than 200 km.
T33A-1327 1340h
Finite strain and relative rheology from field exposures of mantle peridotite, Twin Sisters, Washington
We present estimates of finite strain and relative rheology of naturally deformed mantle materials based on field observations in the Twin Sisters Range of Washington state. The Twin Sisters ultramafic body is a 16 by 5.5 km body located 30 km east of Bellingham, Washington. The outcrops show virtually no serpentinization away from the metamorphic sole. We conducted detailed structural mapping in a 100 by 150 meter field area located east of the crest of the Twin Sisters range and approximately midway between the north and south ends. The foliation strikes ~155 and the lineation pitches 40 S. Folded orthopyroxenite dikes within the host dunite allow us to characterize the finite strain. Dikes trending NE-SE were folded, while dikes trending NW-SE were elongated or boudinaged. Using the method of Talbot (1970), the principal stretch directions in the horizontal plane were calculated using the deformed dikes. We calculated a maximum stretch of 1.596 oriented at 151 (similar to the trace of the foliation) and a minimum stretch of 0.286 in direction 061. Assuming that the lineation and foliation represent the orientation of S1 and the S1S2 plane, respectively, a finite strain ellipsoid was determined. The best fitting answer defines an oblate ellipsoid with S1=3.15, S2=1.11, and S3=0.286. Thus, on this outcrop, the Twin Sisters dunite has an oblate-shaped finite strain ellipsoid whose long axis plunges 40 to the SE. The same area provides constraints on relative rheology. Folded orthopyroxenite dikes show a linear relationship between fold wavelength and dike thickness, indicating that they initiated as buckle folds. Using dynamic instability analysis, the orthopyroxene within the dikes is calculated to have ~31 times the effective viscosity of olivine of the dunite matrix, assuming a power law exponenent of n=3 (dislocation creep) for both the dikes and the matrix. Although not investigated in detail, similar orientations of fabrics are observed throughout the Twin Sisters body. Further, no high temperature localization zones were observed. This data suggests that large parts of the mantle may have relatively homogeneous fabric.
T33A-1328 1340h
Magnetic anisotropy at the lithosphere-asthenosphere boundary: preliminary results from the Ronda peridotite, Spain
The Ronda massif (S. Spain) is a large (300 km2) orogenic peridotite massif located in the Betic Cordillera, the westernmost part of the peri-Mediterranean Alpine orogen. One of the unique features of this massif is the coarsening front which marks the transition between two petrostructural and geochemical domains: [D1] an intensely foliated porphyroclastic spinel peridotite domain with subordinate garnet pyroxenite layers and [D2] a coarse-granular spinel peridotite domain with minor spinel pyroxenite layers. The coarsening front, which extends for 20 km along strike, is marked by a strong grain size increase and annealed microstructures. Based on previous olivine LPO data acquired using EBSD, deformation in [D1] occurred at a significantly lower temperature <$1000\deg$C than in [D2] ($1250\deg$C). Tectonic fabrics between [D1] and [D2] are parallel. The macroscopic foliation dims gradually from [D1] to [D2] but compositional layering and spinel trails (lineation) persist in [D2]. Totally recrystallized melting residues of former garnet pyroxenite dykes (now spinel websterites) display folded structures inherited from the spinel tectonites. Geochemical and textural data show that the coarse-grain peridotite domain [D2] does not represent a "true" asthenospheric mantle but rather derived from the lithospheric spinel tectonite domain [D1], through re-heating, partial melting and pervasive interaction with basaltic melt at quasi-asthenospheric conditions (ca. $1250\deg$C at 1.5 GPa). The increase in fabric intensity, from the tectonites to the coarsening front, that is deduced from olivine LPO data is currently unexplained. Magnetic measurements were performed on 80 specimens (20 mm cubes) from 10 stations. The low field magnetic susceptibilities and degree of anisotropy indicate the dominance of serpentinization-related secondary MD magnetite. The low field AMS fabrics are oblique to the mineral foliation and lineation. Measurements conducted in high magnetic field, using a vibrating sample magnetometer, reflect the contribution of paramagnetic minerals (mostly olivine and orthopyroxene). Torsion magnetometry measurements conducted at fields up to 1.8 T show the lack of high coercivity phases such as hematite. Preliminary results show that the low field AMS and the high field AMS principal axes (K1 and K3) are statistically distinct. These preliminary results allow to constrain the origin of the increase in fabric observed at the lithosphere-asthenosphere boundary.
T33A-1329 1340h
DEVELOPMENT AND APPLICATION OF A NEW METHOD FOR CALCULATING THE STRENGTH OF LATTICE-PREFERRED ORIENTATION (LPO)
Quantifying the strength of lattice-preferred orientation (LPO) has many practical applications in the study of deformation microstructures. We show that a commonly used indicator of fabric strength, the J-index (Bunge, 1982), which is calculated from the distribution of orientation data in Euler-space, is highly sensitive to numerical procedures. Thus, we conclude that the J-index is not a suitable indicator of fabric strength. We have developed a new method for determining the strength of LPO using the distribution of uncorrelated misorientation angles. This misorientation index (M-index) is defined as the difference between the observed distribution of uncorrelated misorientation angles and the theoretically determined distribution of uncorrelated misorientation angles for a random fabric. This method has been tested with orientation data from experimentally deformed olivine and is found to (1) correlate well with shear-strain and seismic anisotropy, (2) give robust measures of fabric strength for as few as 150 grains, and (3) exhibit minimal artifacts of its numerical calculation. We have applied the M-index to the study of recrystallized orthopyroxene bands in sheared lherzolite xenoliths from the Jagersfontein kimberlite, South Africa. Using the M-index, we can observe and quantify the progressive randomization of LPO inherited during dynamic recrystallization. This fabric randomization is used to infer a transition to grain-size sensitive deformation, indicating that strain-weakening of orthopyroxene may play an important role in the rheology of the upper mantle.
T33A-1330 1340h
Constraints on the Upper Mantle Structure of the Slave Craton From Surface Wave Tomography
The Slave craton in the NW Canadian Shield, because of its high degree of preservation and ideal location relative to global seismicity, has been the focus of many seismological studies that have enhanced our understanding of the structure and evolution of Archean cratons. Recent body-wave tomography has imaged a low P-wave velocity structure (2.8% lower than the surrounding mantle) beneath the high-velocity cratonic lithosphere of the Slave, implying persistent processes in the upper mantle relevant to kimberlite magmatism. While body-wave tomography provides good constraints on lateral velocity variations, the vertical extent of anomalous structures in the lithosphere and upper mantle may be further constrained by surface wave tomography due to its frequency-dependent sensitivity to Earth's structure as a function of depth. In this study, we analyze teleseismic Rayleigh wave phase velocities to investigate upper mantle structure of the Slave craton. We measured phase velocities for periods between 28s-143s of 3833 crossing seismic rays from 44 events recorded by the POLARIS broadband seismic network and the Yellowknife array. To better determine the velocity structure under this region, we employ an array-analysis technique that takes into account the multipathing effect of surface wave propagation by modeling the incoming wavefield with a two plane-wave approximation. Our preliminary results show slightly higher phase velocities at shorter periods compared to those of other cratons. Conversely, at periods greater than 100s, the phase velocities are reduced by up to 5% relative to average cratonic values. 1-D inversion of phase velocities yields high S-wave velocities ($\sim$4.65$\pm$0.2 km/s) associated with cratonic lithosphere between 30-130 km depths, underlain by a negative velocity gradient that reaches average velocities of $\sim$4.0 km/s in the 220-350 km depth range. While such negative gradients have been observed between 120 km and $>$200 km depth beneath most other cratons, the S-wave velocities inferred here are lower than elsewhere by at least 7%. These anomalous values, in conjunction with the low P-wave velocity observed in the body-wave tomography, indicate that a heterogeneous structure which may be related to kimberlite genesis persists in the upper mantle beneath the central Slave craton. We investigate the source of this anomaly by considering the influence of water content, thermal and compositional variations, and partial melting.
T33A-1331 1340h
Seismic Velocity Structure of the Siberian Mantle.
During the years between 1970 and 1990 an extensive program of deep seismic soundings (DSS) was carried out in northern Eurasia. As part of this program, data were acquired with peaceful nuclear explosions (PNE) as sources along several profiles. Seismic signals were recorded out to very large offsets (more than 3000 km) from the PNEs. In order to constrain the mantle structure, we perform two-dimensional P-wave velocity modeling and inversion of refraction and reflection travel times along three high resolution Russian PNE seismic data profiles (one perpendicular to the other two) from the Eastern Siberian platform: Kimberlite, Kraton and Meteorite. The seismic sections clearly show reflections from the discontinuities at 410, 520 and 660 km depth. Furthermore, a travel time delay is evident around 8 degrees offset in the profiles. It identifies the top of a low-velocity zone below ca. 100 km depth. A similar low-velocity zone has been identified by non-linear inversion of surface wave dispersion curves. Both inversion of group and phase velocities of an epicenter-station path crossing the Siberian Craton and the P-wave velocity structure obtained from the PNE data, suggest the presence of the low-velocity zone in the depth range of 70-150 km with S-wave velocity between 4.35 and 4.45 km/s and a velocity reduction less than 2 % as compared to the surrounding structure. This low-velocity zone is detected in a mantle with very high average seismic velocities, as typical of cratonic regions. We are presently interpreting the PNE data for possible anisotropy in the low-velocity zone and modeling the physical parameters of seismic scatterers in this layer.
T33A-1332 1340h
Continental Upper Mantle: T, Vs, Qs
Global maps of the thermal structure of the continental upper mantle at depths 50-150 km are compared with global Qs(-1) and Vs maps as determined from Rayleigh waves at periods between 40 and 150 seconds. This allows us to evaluate the relative contributions of thermal and non-thermal effects into seismic anomalies. Vs and Qs anomalies usually inversely correlate with lithospheric temperatures. However, the available dataset does not indicate that, at a global scale, seismic anomalies in the upper mantle are controlled solely by temperature variations. Continental maps have correlation coefficients of $<$0.56 between Vs and T and $<$0.47 between Qs and T. At 100 km depth, where the resolution of seismic models is the highest, seismic Vs and Qs are compared with theoretical Vs and Qs values that are calculated solely from temperature anomalies and constrained by experimental data on temperature dependences of velocity and attenuation. This comparison shows that temperature variations alone are sufficient to explain seismic Vs and Qs in $\sim$ 50% of continental regions. We hypothesize that compositional anomalies due to Fe-depletion can explain the misfit between seismic and theoretical Vs in cratonic lithosphere; partial melts and/or fluids may affect seismic parameters in regions of active tectonics.
T33A-1333 1340h
Fresh Insights into Surface Wave Tomography - Applications to the Australian Region
Strong variations in shear wavespeed in the Australian region are apparent from both global and regional studies and pose particular difficulties for successful surface wave tomography in capturing both the larger and smaller scale variations. A single pass inversion with model-norm damping will result in an underestimate of the small-scale features if the knee of the trade-off curve is used. However, a multi-scale inversion scheme is very effective in the construction of a 3-D shear wavespeed model from the suite of path specific 1-D models, viewed as averages along the path. With a smooth B-spline representation of the velocity variation at each depth the knot spacing can be adapted to allow refinement of scales in the inversion. We have found that an initial pass to retrieve large scale structure with an 8 degree knot spacing, and then a second pass with 2-degree knot spacing provides the best recovery of amplitude anomalies. The multi-scale nature of the wavespeed variations means that care needs to be taken with assessment of resolution, since a conventional chequer-board approach can only focus on one scale at a time. This limitation can be overcome by using a superposition of scale elements in the models constructed for resolution tests with smaller scale systematic variations superimposed on larger block features to provide a direct test of the influence of path coverage in model recovery. Inversion for azimuthal anisotropy is best performed at a moderate scale, e.g., with a 4-degree knot spacing. The observed changes in the direction of anisotropy beneath the continent between 100 and 200 km are consistent with a change from frozen to dynamically induced anisotropy. Surface wave tomography has concentrated on the recovery of shear wavespeeds, with simplifying assumptions about the influence of density. We are employing three-dimensional simulation using the SpeCFEM code of Komatitsch & Tromp as a means of investigating the characteristics of the interaction of seismic surface waves with realistic structures, and the way in which different assumptions in the tomographic inversion will affect the resultant images.
T33A-1334 1340h
On Resolution of Surface Wave Tomography: Geometrical Rays and Finite-Width Rays
Recent developments of finite-frequency sensitivity kernels for surface waves enable us to perform forward and inverse modeling taking account of the effects of scattering and diffraction of surface waves. We have recently developed a three-stage inversion technique to obtain regional surface wave tomography incorporating finite frequency effects utilizing the "influence zone", which can be represented as approximately one-third of the width of the first Fresnel zone (Yoshizawa and Kennett, 2002 GJI). The method has been applied to the Australian region (Yoshizawa and Kennett, 2004 JGR) and to the Philippine Sea region (Isse et al., 2004, AGU Fall Meeting). In this study we perform synthetic experiments, using the influence zone kernels, to assess the intrinsic differences in the resolutions of tomography models with or without the effects of finite frequency. The results of checker board resolution tests indicate improved resolution of finite frequency models; i.e., we can recover both the amplitude and patterns of lateral heterogeneity at scales of a few hundred kilometers quite well. Although large-scale heterogeneity patterns with scale length over 500 km can be retrieved well by both ray theory and finite-frequency theory, some artificial patch-like features tend to contaminate the ray-theoretical models, especially in regions with low path density. The variance reductions achieved by finite-frequency models for the Australian region are nearly 10% higher than those achieved by the ray-theoretical models. The main differences arise from different spatial sensitivities between ray-theoretical and finite-frequency kernels. In the ray theory, the along-path sensitivity is always constant. On the contrary, the finite-frequency kernels have conspicuous peaks of sensitivity near the source and receiver locations, which results in better resolutions in regions with higher sensitivity. Such effects are apparent in the actual tomography models in the Australian region, i.e., clearer images of subducting plates can be seen in the New Hebrides and Tonga-Kermadec.
T33A-1335 1340h
Combining different data to constrain the anisotropic shear-wave velocity structure of the Earth's mantle.
Over the last 10 years, dozens of new stations of the Global Seismographic Network have been deployed in remote and previously uninstrumented regions. The seismograms recorded at these sites provide new constraints on the structure of the Earth's mantle. We investigate the effect of including new measurements of long-period body and mantle-wave waveforms in the tomographic inversion for the anisotropic shear-wave velocity structure of the mantle. To ensure good radial resolution throughout the mantle, we combine the waveform data with measurements of Rayleigh- and Love-wave phase velocities and a diverse data set of absolute and differential body-wave travel times. To further improve the ray-path coverage we test the consistency of the $SS-S$ and $ScS-S$ differential travel times ({\it Bolton}, 1996) with our travel-time data and include them in the inversion. All waveform, surface-wave and body-wave travel-time measurements are corrected according to the global crustal model {\it CRUST 2.0} ({\it Bassin et al.}, 2000). We parameterize the mantle model using 362 spherical splines and 14 radial B-splines and minimize both horizontal and vertical gradients to regularize the inverse problem. To alleviate the effect of an arbitrary choice between a continuous and split model across the 670-km discontinuity, we define a split parameterization and control the change across the discontinuity using norm damping. Finally, we focus our attention on the capability of the combined data set to resolve the anisotropic structure in the mantle parameterized in terms of the velocities of horizontally- and vertically-polarized shear waves.
T33A-1336 1340h
Global Analysis of Upper Mantle Anisotropy Using Automated SKS Splitting Measurements
We have developed an automated method to measure shear wave splitting and applied it to SKS phases archived by various data centres. Using an automated system we are able to process data in a quantity and with a consistency not possible with manual measurements. To date we have processed over 100,000 station/event combinations from over 250 stations distributed globally, although our stringent quality control measures mean that only around 2 percent produce reliable estimates of splitting. When a polarised shear wave passes through an anisotropic medium, it is split into two perpendicular S-waves that are separated in time. In measuring splitting, we seek a polarisation angle and time shift that removes the effect of the anisotropy, minimising either the energy on the transverse component or the smaller of the two eigenvalues calculated from the covariance matrix. We report results from both methods, to allow consistency in comparisons with the results of other researchers and because we find consistent differences in results at some stations. Error estimates are calculated using both the methods of Silver and Chan (1991) and Sandvol and Hearn (1994). Where it is possible to compare, we find that our automated results agree well with manual measurements. The quality of results at individual stations is variable, being influenced both by the volume of available data (determined by station deployment date and the distribution of natural seismicity), and the performance of the STA/LTA picker used to define the start of the SKS window. When we have reliable results from a range of back azimuths, we compare the distribution of results with those predicted from single and multiple layers of anisotropy. Where this is not possible we report the polarisation angle and lag time for the most convincing splitting measurement. Where even this is not possible and where a sufficient number of events have been processed, a station is interpreted as a null, which may mean that it is underlain by an isotropic or transversely isotropic mantle. For stations suggesting either single layer anisotropy or where an individual splitting measurement has been used, we find an average lag value of 1.15 seconds. There seems to be some relation between splitting parameters and tectonic environment. These results are available through the ISC website and provide an aid to studies of anisotropy in the Earth.
T33A-1337 1340h
A Method For Measuring Shear Wave Splitting With S Phases Independently Of Source Side Splitting, With Application To Southern California Seismic Network Stations
The shear wave splitting community generally prefers to measure vertically averaged horizontal anisotropy with core phases such as $SKS$ because their initial polarization is well known, and any source side splitting will have been removed by the conversion to a compressional wave at the core-mantle-boundary. However, particularly for temporary deployments, the number and distribution of well-recorded $SKS$ phases is often too small to unambiguously determine the anisotropy, particularly if it cannot be easily described by a one-layer model. It is thus desirable to be able to carry out measurements with $S$ phases. Whereas it is possible to relax the assumption of known initial polarisation by imposing linearity of the corrected waveform, the ambiguity between source side and receiver side splitting has been harder to tackle. Measurements from deep events below $\sim$300~km are generally considered valid because dislocation creep is not thought to operate at large depths but the recent indications of anisotropy within the transition zone near subduction zones make questionable even this assumption. We present a method for measuring the receiver side splitting of S phases unbiased by source-side splitting by comparing waveforms with reference stations, whose anisotropy is well characterised (e.g., an IRIS station near a temporary network). Synthetic and data examples demonstrate the applicability of the assumptions underlying this method, and we present measurements at Southern Californian stations.
T33A-1338 1340h
Seismic anisotropy in the Western US as a testbed for advancing combined models of upper mantle geodynamics and texturing
Observed seismic fast axes are popularly equated with directions of mantle flow. In order to substantiate this assumption, we present improved models of mantle flow, derived textural anisotropy, and predicted seismic anisotropy. Results include global models, where we compare surface wave based seismologic inversions with predictions from large-scale mantle circulation computations. For body wave anisotropy, we focus on a regional model of the western United States, a relatively simply-parameterized region where a wide range of seismic data is available. Our work addresses the three stages needed to connect mantle flow and anisotropy: flow, texture development, and wave propagation in a heterogeneous upper mantle and lithosphere. For the flow models, we explore lateral variations in viscosity, improved rheologic realism, and joint regional/global convection computations. In terms of mineral physics, we compare predictions from finite strain (FS), lower-bound (LB), and kinematic (KR) texture formation theories. The seismological modeling includes computing apparent splitting from spatially variable, not necessarily hexagonal anisotropic elastic tensors. We also evaluate the potential role of the crust in partly obfuscating the underlying dynamic processes. Preliminary results indicate that FS, LB, and KR models are similar except in regions of large spatial variations in flow, as expected. Mantle circulation models that take the inferred Farallon slab density anomalies at depth into account show a return flow roughly opposite to the surface motion of North America. Those models tend to fit the data better, indicating a possible avenue to better constrain tectonic processes in the study region over time.
T33A-1339 1340h
Waveform modeling of upper mantle shear wave triplications in the western United States
Previous study (Song at al. 2004, Nature) has shown rapid variation of S wave triplication associated with possible low velocity zone atop the 410 seismic discontinuity beneath the Basin and Range in the western United States. The results are based on anomalous differential time between the CD (bottoming below the 410 seismic discontinuity) and the shallower AB-(bottoming around 200-300 km) triplication branches, along with anomalous receiver function estimate. To take into account variations in shallow mantle structure, we make velocity crosssections from recent regional tomographic images and compute 2-D synthetics. Here, we report on waveform modeling (2D) of record sections sampling the western edge of this structure from events offshore Washington-Oregon. Broadband data recorded by TriNet (12 events), Berkeley network (6 events) and NARS array (3 events) are used to study the western edge of this LVZ structure. Typically, at distances of 14-17 degrees, the S wave triplication recorded by TriNet exhibits rapid changes in both the AB-branch and the CD-branch when raypaths travel across the California-Nevada border. In particular, the CD-branch recorded by stations near the California-Nevada border arrive late for about 3-4 secs with larger amplitude than that recorded by stations close to the coastal California. Rapid azimuthal variations in the delay of CD branches are observed from record sections in multiple events.Such anomalous delay, however, is not observed for paths travelling offshore California-Oregon. Our preliminary result indicates a low velocity zone (5% reduction in shear velocity) atop the 410 seismic discontinuity beneath the western Basin and Range, with dimension of about 300-500 km long and 25 km thick. This low velocity anomaly extends to shallower depth ($\sim$ 250 km) beneath the southern part of the Western Great Basin with reduced amplitude (2%). Given the rapid variation in the S-wave triplication data, we attribute the low velocity anomaly to silicate melts, which are probably produced by the dehydration melting across the 410 seismic discontinuity.It may have strong implications on the water content in the mantle transition zone, melting processes and geochemistry.
T33A-1340 1340h
High-resolution Surface Wave Tomography of the Western and Central Alps
Global resolution of surface wave phase velocity maps is limited by data coverage. However, on local or regional scales data coverage can be quite dense in areas with closely spaced seismometers. It is possible to utilize data from high-density arrays by overparameterizing the inversion in areas with denser ray-path coverage. Two Swiss broadband arrays create a dense seismic network in and around Switzerland. The Swiss Digital Seismic Network (SDSNet) is a dense permanent array including 29 STS-2 broadband seismometers with a station spacing of 40-50 km. To supplement data from this array, 7 temporary STS-2 broadband seismometers of the TomoCH (Tomography of Confederatio Helvetica) array have been installed peripherally to Switzerland to increase the array aperture to approximately 5 degrees by 2 degrees. Source/receiver dispersion curves have been calculated by fitting synthetic waveforms to frequency-filtered event traces containing fundamental Rayleigh and Love waves. These observations have been merged with data previously collected and measured in European and global studies. The resulting comprehensive database has ray-path coverage sufficiently dense to allow relatively stable, lightly damped, highly-parameterized inversion for phase velocity perturbations in central-western Europe. Subsequent to dispersion measurements, a linearized inversion algorithm has been used to obtain phase velocity maps by fitting the measured dispersion curves. Observations of phase velocity perturbations contribute to high-resolution studies of the lithosphere beneath the Alps with the ultimate goal of understanding the dynamic processes that formed and stabilize these mountains.
T33A-1341 1340h
Velocity and Attenuation Structure beneath the Southeastern Carpathians
One of the seismically most active regions in Europe is situated at the southeastern bend of the Carpathian Arc (Vrancea region). Seismologically it is intriguing that the seismicity pattern is dominated by an intermediate-depth, finger-shaped volume with recurring strong earthquakes. From the tectonic evolution of the Pannonian-Carpathian subduction system we know that continental collision (followed by slab detachment) started in the northernmost part of the Carpathians and migrated towards the southeastern bend where the slab seems to be still coupled to the overlying crust. For our studies we used teleseismic P-wave data recorded during the CALIXTO-Experiment. Simultaneous inversion of P-wave traveltimes from the CALIXTO data and a global dataset of ISC- and NEIC-data provides a large-scale, high-resolution image of the P-wave velocity distribution in the upper mantle. The tomographic imaging technique includes implicit correction of the residuals for crustal structure, a variable parameterization, 3-D Ray Tracing and a nonlinear, iterative inversion scheme. Our results show an extended, well-resolved volume of increased P-wave velocity outlining the seismic active volume but extending down to a depth of approximately 280 km beneath the Vrancea region. The hypocenters are distributed within this high-velocity body, which points to stress accumulation induced by gravitational sinking. A low velocity anomaly is mapped northwest and west of the Vrancea region down to a depth of 200 km coinciding with geological constraints of Neogene-Quarternary volcanism. The analysis of the attenuation structure was performed using multiple taper spectral analysis. The slope of linear regression to the fall-off of the spectra is direct proportional to the amplitude absorption (t$^\ast$-Operator). Calculation of t$^\ast$-Operators from teleseismic CALIXTO data indicates that the largest t$^\ast$-values are observed at the bend of the Carpathians, in the mountains as well as in the foreland. It is complicated to interpret the distribution of t$^\ast$ values since observed attenuation parameters include both intrinsic and scattering attenuation effects. Synthetic t$^\ast$-operators can be modeled by calculation of a Q-model based on P-wave velocity structure and thus allow to give a guess on the intrinsic fraction of the observed attenuation. For the Carpathians we find that approximately 75% of the observed attenuation of teleseismic P-waves is induced by scattering, predominantly from complex boundary topographies and heterogeneities in the crust.
T33A-1342 1340h
S-splitting in Eastern Nepal and Southern Tibet Across High Himalaya
The 28 station HIMNT broadband seismic deployment in eastern Nepal and Southern Tibet in September, 2001-November, 2002 recorded teleseisms that allow the determination of S-splitting using multiple events from different back azimuths. We use mainly SKS waves for this study and the waves selected for analaysis are visually highly coherent in the band between 0.02 to 0.2 Hz. A code by Menke (see reference) is used for the determination of split parameters. The resulting fast directions for both eastern Nepal and southern Tibet are in the NNE direction (N $10\deg$-$30\deg$ E) with delay times varying in the range of less than 0.1 second to 0.9 seconds with most of the values in the 0.2 to 0.4 second range. The variations from station to station do not appear to have a clear pattern. These measurements are distinctly different from splitting parameters found from stations to the north of our study area. In northern part of the Tibetan plateau some of the largest splitting delays, up to 2.7 seconds (McNamara et al., 1993), in the world are found and the fast directions are generally trending ENE. Toward southern Tibet, the delays decrease and the direction changes to more northeasterly. Earlier sparse measurements of splitting in the high Himalaya shows NW or NE fast directions but with small delay times (Hirn, 1995). In many of the young mountain ranges of the world the fast directions areparallel to the trends of the ranges, e.g., New Zealand (Klosko et al., 1999), Taiwan (Rau et al., 2000), etc., some of which are undergoing significant shear. It is not yet clear as to the source of the small magnitude splitting that we observe across the High Himalaya. Such relatively small splitting delays could conceivably be generated mainly in the crust, with possible mantle contributions. The lack of large splitting delays with fast direction parallel to the mountain range reflect differences of the orogenic processes in the upper mantle beneath the Himalaya relative to other young mountain ranges. Selected References Hirn, A. and 11 others, 1995. Seismic anistropy as an indicator of mantle flow beneath the Himalayas and Tibet, Nature, 375,571-574. McNamara, D.E.,Owens, T.J.,Silver, P.G.,Wu, F.T.,1994. Shear wave anisotropy beneath the Tibetan Plateau,JGRB, 99, 13,655-13,665. Menke, W. (http://www.ldeo.columbia.edu/menke/software.html) Klosko, E. R.; Wu, F. T.; Anderson, H. J., and others, 1999, Upper mantle anisotropy in the New Zealand region,Geophysical Research Letters,26, 1497-1500 . Rau, R.J.Liang, W.T,, Kao,H., and others, 2000. Shear wave anisotropy beneath the Taiwan Orogen,EPSL,177, 177-192 .
T33A-1343 1340h
Crustal and Upper Mantle Structures of the High Himalaya
The crustal and mantle structures in the areas around the high Himalayan ranges are still enigmatic. The broadband data recorded at 28 stations in eastern Nepal and southern Tibet in 2001-2002 provide the basis for resolving some of the details. We used two approaches in the present study. First, the two-station differential surface wave phase velocity dispersion technique of Tanimoto and Prindle-Sheldrake (2002) and Prindle-Sheldrake and Tanimoto (submitted) is used to develop phase velocity maps for different station pairs (only those with separations greater than 50 km are selected), within a narrow range of the back-azimuths - usually less than 3.0 degrees. The assumption is that the differences in such waveforms are attributed to the path effects between the two stations. Under such conditions 151 events were recovered with good signals within specific frequency bands. Phase velocities between station pairs are inverted to generate phase velocity maps for an area discretized into 0.20 by 0.25 degree blocks (approximately 20 km) in latitude and longitude. Using a technique that approximately takes into account the finite frequency effect (Tanimoto, 2003; Prindle-Sheldrake and Tanimoto, submitted) a set of maps have been obtained for frequencies 0.01- 0.03 Hz. Variance reductions for the phase velocity maps are above 90 % in all cases. Current results indicate that at higher frequencies the main feature of higher velocities in the southern part of the region and lower velocities in the north dominates. At the lowest frequency (0.01 Hz) a regional low velocity feature under the high Himalaya is found. This last feature seems may imply the existence of a low velocity feature at depth below the high Himalaya, thus coming into conflict with results of a receiver function analysis (Schulte-Pelkum, personal communication) indicating the existence of a transitional crust. Inversion of the phase velocity maps into shear velocity with removal of elevation effect will be necessary. Secondly we are attempting to provide additional constraints by using teleseismic/local earthquake joint inversion (Roecker, ) to obtain the crustal and upper mantle velocities. Preliminary results show the differences between what is most probably the Indian plate lithosphere and that of the Eurasian plate. To resolve this question the spatial resolution for the joint version will have to be successively improved.
T33A-1344 1340h
Black upper mantle beneath the Sierras Pampeanas, Argentina? Results from seismic tomography.
Regional P- and S-wave travel-time data were jointly inverted with teleseismic P-wave residuals to obtain 3-D seismic tomography models for Vp, Vs, and Vp/Vs above the subducting slab in central Chile and Argentina. Additional constraints on the velocity model are provided by Pn and Sn apparent phase velocity data. In this region, there is an abrupt change from a normal subduction geometry south of $33^{\circ}$S to a flat subduction geometry to the north. We find low Vp, low Vs, and high Vp/Vs ratios in the southern half of our study area directly beneath the modern active volcanic arc which we interpret as localized pockets of melt. In the northern half of our study area, above where the subducting Nazca plate flattens at 100 km depth, we find low Vp, high Vs, and low Vp/Vs ratios. These unusual results point to a lack of melt or hydrated mineralogies such as serpentine, both of which are characterized by high Vp/Vs values. The only mantle rocks that have low Vp/Vs and high Vs are Mg rich compositions, such as dehydrated serpentinite or orthopyroxenite. We suggest that large portions of the mantle overlying the flat slab consists of orthopyroxenite, formed by a transient fluxing of silica rich fluids. Such fluids may have come from sediments which were subducted during the initiation of flat subduction at this latitude $\sim$10 Ma. This would imply that the hydration of mantle material above a flat slab can be a transient phenomenon, which leaves little residual free water behind, but significantly alters the mantle chemistry.
T33A-1345 1340h
Waveform Modeling Of The Subducting Slab Beneath Japan
Deep earthquakes ($400-600$ km) occurring in the down-going Pacific plate produce detailed waveform patterns recorded by the Japanese Hi-net array ($>500$ stations), at distance from about $7^\circ$ to $12^\circ$, with ray paths densely and uniformly sampling the vicinity of the subducting slab. Most P-wave direct arrivals are well predicted by a 1-D model. However, observations along paths sampling the slab are advanced by up to 4 seconds and many show strong multipathing over patches of about 100 km in width. Observed P-wave seismograms from stations in these patches can be well aligned by a time-shift from a preferred combination involving epicentral distance and azimuth by which the multipathing becomes well-organized. The P-wave 3-D tomographic model proposed for this region by Zhao (1993) has a well-developed slab with velocities $3-6%$ higher than the normal mantle and a thickness of about 90 km, and low velocity zones with anomalies up to $-6%$ in the mantle wedge. Synthetics calculated based upon 2-D finite-difference and 3-D spectral-element methods are compared with Hi-net data (vertical component) and F-net data (3 components) for regional deep earthquakes. The comparisons show that synthetics generated from Zhao's P-wave model can predict some of the data in terms of both timing and waveform complexity as measured by cross-correlation, but discrepancies also exist. Waveforms in certain azimuth and epicentral distance ranges are very complex, suggesting stronger 3-D model variations.
T33A-1346 1340h
Anisotropic Seismic Structure along the trace of the Dead Sea Rift
How the motion on a transform plate boundary is accommodated in the upper mantle is not very well known. A notion of a narrow zone of deformed mantle rock associated with the transform is intuitive, however such specifics as the width of this zone, the intensity of deformation etc. have been probed at a few locations only. Deformation of upper mantle rocks should result in the formation of fabric that makes of seismic wave velocity anisotropic, and thus studies of seismic anisotropy indicators hold considerable promise in improving the understanding of continental transform fault behavior. The Dead Sea Rift is a left-lateral transform boundary between the Arabian and African plates, with over 100 km of offset accumulated since $\sim$15Ma. It cuts through the former shield, and thus is modifying a lithosphere that has experienced numerous episodes of rock-fabric formation prior to the inception of the transform. The challenge in understanding the accommodation of this transform at depth is thus in discriminating rock fabric imparted by the modern motion from other signatures that have likely been left behind by previous tectonic episodes. We investigate depth dependence of seismic anisotropy along the trace of the Dead Sea Rift. We use observations of birefringence in $SKS$ phases from permanent and temporary observatories. In preliminary analysis we found considerable variation in values of effective splitting parameters (fast directions and delays), both with the propagation direction, and between sites. Such variations are likely when anisotropic structures are either stratified or laterally variable. We devise depth-dependent models of anisotropy and evaluate their relative merit using group inversion methodology of Menke and Levin (2003). Sets of azimuthally distributed observations are inverted together for models of different classes (single layer of anisotropy, single layer with inclined axis, two layers of anisotropy etc.) and their respective fits are compared. Models with best performance are then interpreted in the context of both the modern plate motion and the past tectonic episodes in the region.
T33A-1347 1340h
Evidence of Melt-Induced Seismic Anisotropy and Magma Assisted Rifting in the North Ethiopian Rift
The complex process of the transition from continental to oceanic rifting remains poorly understood. The Northern Ethiopian Rift, being free from interference from other tectonic processes, is an ideal place to study the continental breakup process. With this in mind, the recent EAGLE experiment saw the deployment of 79 broadband seismometers over an area 250-350km centred on the Northern Ethiopian Rift. We investigate the signature of crustal and upper-mantle anisotropy in these data as it provides insights into rifting processes. Recent studies employing shear-wave splitting techniques provide strong and consistent evidence for a rift-parallel (NNE-SSW) fast anisotropic direction beneath the rift. The detailed characteristics of these observations imply a single anisotropic layer confined to the upper 100km. Surface-wave tomography shows that the fast rift-parallel directions persist to a depth of 400km beneath a broader area surrounding the rift. These observations eliminate a number of plausible causes of anisotropy including plate motion drag, radial mantle-flow in the Afar plume head, mantle flow perpendicular to the rift induced by the rifting process, or pre-existing frozen-in crystallographic fabric. The observed anisotropy is more likely to be caused by {\sl either} channeled horizontal mantle-flow along the rift axis, which would cause the lattice preferred orientation (LPO) of olivine with the fast $\alpha$-axes paralleling the rift, {\sl or} the presence of rift-aligned melt-filled pockets (MFP) in the mantle. % SKS-splitting results show that the distribution of the fast anisotropic orientation mimics closely the distribution of strain and magmatism in the rift, implying MFP-induced anisotropy. % However, the techniques employed by the studies to date do not provide a means of conclusively separating the two candidate causes of anisotropy. The speeds of horizontally propagating $S_{\rm v}$ and $S_{\rm h}$ waves vary in similar fashions with azimuth for LPO- and MFP-induced anisotropy ( $S_{\rm v} \propto \cos({\theta})$, $S_{\rm h} \propto \cos({2\theta})$, where $\theta$ is the azimuth measured from the rift axis). However, the the relative change in the two shear-wave velocities is distinctive for LPO- and MFP-induced anisotropy. This provides a powerful tool for distinguishing between the two candidate causes of anisotropy. % We present strong evidence for MFP-induced anisotropy beneath the rift in the depth range of 20--60~km, by showing that the azimuthal variation of the speeds of $S_{\rm v}$ and $S_{\rm h}$ waves propagating horizontally through the rift area is in good agreement with predicted anisotropy models for vertical rift-parallel melt-filled dykes.We obtain shear-velocity models by inverting the group and phase velocity dispersion experienced by a number of local and teleseismic Rayleigh and Love waves with total or inter-station propagation paths crossing the rift area with a variety of azimuths. By only using highly coherent waves for the phase-velocity measurements and by applying a phase-matching filtering technique in the group velocity extraction procedure we ensure that our measurements are free from bias introduced by scattering and noise and that our results are reliable and robust.
T33A-1348 1340h
Constraints on Crust and Upper Mantle Velocities Beneath the Gulf of California from the Analysis of Love and Rayleigh waves
The Gulf of California is one of only a few newly forming ocean basins where seafloor spreading is still not fully established. Improved knowledge of crustal composition and thickness, and the inferred thermal state of the upper mantle is needed for a more complete understanding of how this plate boundary has evolved. This in turn provides a basis for explaining what role if any past periods of subduction may have played and sheds light on the mechanism by which the apparently slow, but imminent transition to seafloor spreading is manifested in small segmented basins. Whether the crust is mainly oceanic, continental or transistional in composition has broader implications for studies on magmatism, structure and deformation. Furthermore, the constraints provided by reliable crustal velocities on earthquake locations and determinations of crustal thickness via receiver functions are invaluable. Here we present a surface wave dispersion study of regional earthquake data recorded by the NARS-Baja array, a set of 14 broadband seismic stations surrounding the Gulf of California. This study is conducted with both Love and Rayleigh waves, whose dispersion allows their velocities to vary with period and depth of penetration. They thus provide valuable information on crust and upper mantle velocities along their paths. We will use a number of earthquakes to create a 2D map of the phase velocities in the region, which can then be used to determine the 3D structure. Our initial results show noticeable differences in the group velocities of different paths. Due to the fact that the NARS-Baja array straddles the plate boundary, we are able to make additional first order comparisons of the crust on opposite sides of the conjugate margins.
T33A-1349 1340h
Rapid variation in the upper mantle P velocity structure beneath Gulf California
Recent studies of upper mantle triplications involving large number of stations reveal a myriad of small-scale structures, ranging from deep low velocity zone just above the 410km discontinuity to breaking this sharp discontinuity into a combination of first-order and second-order changes. Here, we report on an even larger dataset involving over a thousand records containing a cluster of Mexican events at several different depths as recorded by CISN, IRIS, USGS networks and PASSCAL arrays. Absolute travel times of the various record sections seem to be reasonably well explained by recent surface wave tomography models, but are not very effective in predicting triplication behaviors. \par One of the sharpest anomalies occurs for record section recorded by CISN at distances between 19 and 30 degrees. At this range, the CD branch (410km arrival) and the EF branch (660km arrival) cross near 23 degree. However, about a third of the western portion of CISN array shows this cross-over occurring about a degree earlier and this edge is sharp (within 50km). The mid-point locations of these samples are at the eastern edge of the Gulf of California. Several neighboring events at different depths display similar patterns, but the pattern moves laterally for more distant events. \par Our preferred interpretation is that it is caused by lateral variations of the transition zone, either by about 10kms of topography on the 660km discontinuity (EF-changing) or a low velocity zone just above the 410km discontinuity (CD-changing).
T33A-1350 1340h
Anisotropy in the upper mantle beneath Mexico
We have developed a 3D model of the seismic velocity structure for the upper mantle beneath Mexico including measurements of radial anisotropy with the ultimate goal of establishing its thermal structure and composition. In particular, we want to establish the variability of structure between different tectonic provinces. We obtained high resolution of the region by using data from earthquake sources in the south to dense collections of seismometers in the north, in particular the RISTRA seismic array and the California network. Initially, we defined paths between a single source and a collection of receivers and calculated 1D models for each path using a seismic waveform inversion procedure designed to match the body and surface waves of the radial, vertical and tangential components simultaneously. These 1D models were then combined to create the 3D tomogram. The results show that the mantle beneath Mexico is characterized by a radially anisotropic seismic lid, which thickens from west to east, overlying a low velocity zone. Most of the lateral variation in the models occurs in the top 200 km, generally related to tectonic regime. S anisotropy is highest (7%) in the thin lid beneath the tectonically active western region. The anisotropy beneath central and eastern Mexico is fairly uniform (4-5%) and extends deeper into the mantle (150 km depth). P anisotropy is not well resolved, but P structure mimics the SH velocity structure, suggesting that P is also anisotropic within the lid. The VP/VSH ratio (1.86) is nearly constant as a function of depth from the Moho through the transition zone.
T33A-1351 1340h
Southern California Plate Parallel Azimuthal Anisotropy from Surface Wave data
Azimuthal anisotropy was recovered in Southern California using data from CISN (California Integrated Seismic Network) broadband seismic network. Approximately 150 stations were utilized in this study with the number of paths reaching about 3000. Phase velocity maps including azimuthal anisotropy terms were generated for Rayleigh waves at frequencies 0.02 Hz through 0.055 Hz. The inversion for phase velocity maps included lateral heterogeneity and 2 theta terms for anisotropy. First, we have carefully tested the resolving power of our data by computing resolution kernels and by performing checker-board input tests for anisotropy. Resolution analyses indicate that the data are capable of resolving 2 theta anisotropy terms for approximately the size 150 km x 150 km. It is not possible to determine short wavelength scale anisotropy below the length scale 100 km. Derived anisotropy patterns seem to display interesting tectonic implications. Rayleigh wave azimuthal anisotropy results show orientations of the fast axis parallel to the motion of the Pacific plate on the west side of the San Andreas fault for frequencies between 0.02 Hz and 0.055 Hz. East of the San Andreas fault, Rayleigh wave azimuthal anisotropy is comparatively small from 0.02 Hz to 0.03 Hz, suggesting a substantial anisotropic contrast across the tectonic margin in the upper mantle. The Pacific plate parallel anisotropy orientations west of the San Andreas may indicate upper mantle flow under the Pacific plate. At low frequencies, between 0.02 Hz and 0.03 Hz, the Salton Trough region has fast axis oriented perpendicular to Pacific plate motion and overall strike of margin faults, such as the San Andreas. Above 0.03 Hz the orientation changes to parallel to Pacific plate motion. If spreading underneath the Salton Sea exists, this pattern may indicate orientation of anisotropy related to the spreading flow in the upper mantle. At higher frequencies, the amplitude of anisotropy increases on both sides of this plate boundary region, indicating strong crustal anisotropy.
T33A-1352 1340h
A Teleseismic Experiment to Investigate Crust and Mantle Beneath the Seychelles: Crustal Structure and Upper-Mantle Anisotropy Beneath a Microcontinent
It is well known that the Seychelles microcontinent resulted from episodic rifting between Africa and India. However, the mechanics of such rifting are poorly understood. The aim of the SEISM project (Seismic Experiment to Investigate the Seychelles Microcontinent) is a better understanding of microcontinent formation through a seismic study of the upper-mantle and crustal structure beneath the Seychelles. From Feb. 2003 and Jan. 2004, 8 broadband and 18 intermediate-period three-component seismometers were deployed on 18 islands in the Seychelles. During this time period 240 events of Mb $>$ 5.8 were recorded. Microseismic noise was a problem on most islands and resulted in low signal to noise ratios in seismic recordings. A polarisation filter developed by Du et al. (GJI, 2000) has been used to enhance teleseismic signals. The first stage of data analysis has been a study of upper-mantle anisotropy using shear-wave splitting in core phases such as SKS. In general the degree of splitting is similar to global averages (i.e., $\sim$1 second). Variability in the orientation of the anisotropy suggests a mechanism related to deformation associated with microcontinent formation during the breakup of the Madagascar/Seychelles/India land mass, rather than a mechanism associated with current plate motions. The granitic inner-islands of the Seychelles plateau show a coherent NNE-SSW trend in anisotropy, whilst stations near the Mascarene plateau show more E-W trends. Such trends may support ideas of an anticlockwise rotation of the Seychelles during rifting events, or influence from local plume upwelling. The boundaries between oceanic and continental parts of the plateau are poorly know. Crustal thicknesses can be used to determine oceanic and contenental parts. Receiver functions will be used to estimate crustal depths and deeper mantle discontinuities.
T33A-1353 1340h
Upper Mantle Structure Beneath the Eastern Pacific Ocean Ridges
We analyze vertical-component body and surface waves for ten Mw$>$5 earthquakes, recorded by ocean bottom seismometers at regional and teleseismic distances. Through waveform modeling we place new constraints on along-axis variation in temperature and partial melt beneath the Eastern Pacific Ridges. The resulting best-fit models show over 9% variation in average lithosphere shear velocities between different ridge segments. We demonstrate that the lid velocity correlates with the square root of plate age consistent with a conductive cooling process, but find a more rapid dependence on age close to the axial rifts. We map the average plate age into a mean lithospheric temperature for each of our models using a half-space cooling model, and the temperature derivatives (dVs/dT) determined from least squares fits are -1.1 m/s/deg and -0.26 m/s/deg, respectively, for temperatures above and below 1000C. The former estimate is more negative than values determined by earlier reports (-0.4 to -0.7 m/s/deg), using global or regional data from a much wider range of sea floor age but with less resolution at young ages. The high absolute dVs/dT value suggests the presence of partial melt at shallow mantle depths beneath young ocean crust out to an age of approximately 5 Myr. Our data also show a strong north-south difference in mantle structure. The surface waves that traverse through the southern EPR experience shear velocities that are as low as $\sim$3.75 km/s, more than 0.2 km/sec slower than the average mantle structure at comparable depths beneath the northern EPR and the Galapagos spreading center. This difference cannot be explained by simple conductive cooling or spreading rate variation between ridge segments. We hypothesize that more melt exists near the ridge axis of southern EPR, either due to higher melt production in the south or more efficient melt extraction in the north.
T33A-1354 1340h
3-D Numerical Simulations of Mantle Flow Beneath Mid-Ocean Ridges
Small-scale perturbations of the geoid lined up in the seafloor spreading direction have been observed in the Pacific Ocean. These anomalies may possibly originate in small-scale convection. We perform 3-D numerical simulations of the cooling of an oceanic lithosphere above a convective mantle to test this explanation. We use a Newtonian rheology with a viscosity depending strongly on temperature. The cartesian box has an aspect ratio of 6x3x1 (or 12x6x1). A constant horizontal velocity (half spreading velocity) of 2 (or 4) cm/yr is applied at the surface of the box to mimick plate motion. In all cases, due to the temperature dependence of the viscosity, a rigid conductive lithosphere develops beneath the two plates separated by the ridge. After 16 to 40~Myrs, cold downgoing instabilities develop at the base of the lithosphere. Small-scale convection can then be observed superimposed on the large-scale circulation. It produces, and then interacts with, a slowly evolving short wavelength isotherms topography at the base of the lithosphere. Three kinds of ridge geometry are modelled: (A) a ridge perpendicular to the plate motion, (B) a ridge strongly oblique to the plate motion, and (C) a ridge with transform faults along the spreading center with a mean ridge orientation strongly oblique to the plate motion. All simulations reveal complex interaction between the isotherm topography within the base of the lithosphere, the small-scale flow generated at or just below the base of the lithosphere, and the large-scale flow dominating in the mantle below. The large-scale flow appears always controlled by the mean ridge axis direction, and thus may be oblique to the imposed plate motion direction. In order to obtain a very crude quantification of the anisotropy orientation generated by the modelled mantle flow, we have evaluated the average horizontal shear direction. We thus conclude that surface wave tomography in the uppermost mantle should average small-scale, possibly complex, features associated to the geometry of the developping thermal boundary layer instabilities, plus a large-scale pattern that derives from a combination between shearing induced by plate motion and an internally driven flow.
T33A-1355 1340h
A Numerical Simulation of Longitudinal Rolls and Its Implications to Small Scale Convection in the Mantle
The possible existence of longitudinal rolls in the Earth's mantle was first proposed by Frank Richter over three decades ago. Subsequently, many investigators have presented various observations to suggest their existence with varying degrees of success. In recent years, however, this feature has received renewed interests. Based on seismic anisotropy studies, such a structure has been suggested to exist beneath the Pacific Plate as well as within the D" layer above the CMB. Although the onset of longitudinal rolls has been investigated and laboratory experiments have been conducted to study their structures, direct numerical simulations have received relatively little attention. In this paper, we shall present some of our simulation results to gain improved understanding of longitudinal rolls and their possible implications to the structure of the mantle. Our simulations are carried out using FLUENT 6, a commercial computational fluid dynamic package available at the NCSA of the University of Illinois. We have examined flow structures over a range of Rayleigh numbers varying from 5000 to 10$^{6}$, and Peclet numbers ranging from 10 to 10$^{3}$. Our results show that longitudinal rolls can exist only within a range of specific combinations of Rayleigh and Peclet numbers, consistent with the stability analysis result of Koranaga and Jordan (2003). Particle motions of the longitudinal rolls appear to follow a helical path. This may have implications to the seismic anisotropy investigations. In addition, we have also examined the effect of a stress-free bottom boundary to study if such structure in the D" layer can exist. Our results show that longitudinal rolls can exist with a stress free bottom boundary. However, horizontal scale of the rolls in the transverse direction becomes larger. Core temperature of the rolls is also higher than that of a no-slip bottom boundary. This is probably because heat can be brought into the system from below more easily when the boundary is stress-free.
T33A-1356 1340h
Imaging Crack Systems in The Geysers with Shear-Wave Splitting
Clear shear-wave splitting (SWS) is observed in 1757 high signal-to-noise ratio microearthquake seismograms recorded in The Geysers geothermal field, CA. The high quality observations of shear-wave splitting parameters (fast shear-wave polarizations and time delays) and the good data azimuthal coverage provide a unique opportunity to test the observability of shear-wave splitting and its usefulness for accurate subsurface fracture modeling. Shear-wave splitting parameters are highly sensitive to the anisotropic fabric of the medium through which shear-waves propagate and constitute the basic dataset to invert for 3D crack geometry and crack density in the subsurface. Fracture inversion results from simultaneous minimization of polarization and time delay residuals show that the most common patterns of fracture-induced anisotropy in The Geysers can be simulated by horizontally transversely isotropic (HTI) media or rocks with vertical to steeply-dipping systems of parallel cracks, which in general strike parallel to the N-to-NE direction of maximum compressive stress. The average crack density is about 4%. Deviations from horizontal transverse isotropy conditions in The Geysers are modeled by non-vertically dipping crack systems or intersecting crack systems. The SWS data also show strong evidence of frequency-dependence. Observed time delays, which are used to measure shear-wave velocity anisotropy, show a consistent decrease as the dominant shear-wave frequency increases. We attribute such frequency-dependence to fluid-flow effects in subsurface macro-fractures and describe a method that allows estimation of fracture size from detailed measurements of frequency-dependent double shear-wave splitting.
T33A-1357 1340h
Seismic Anisotropy at the Krafla Geothermal Field in Northern Iceland
We deployed an array twenty PASSCAL L-28 4.5-Hz sensors for forty days during the summer of 2004 at the Krafla Geothermal field. During this time, twenty GeoSpace 1-Hz instruments recorded the seismicity for fifteen days. The Krafla Geothermal field is located approximately 60 km East of Akureyri in northern Iceland. The arrays recorded approximately 5 micro-earthquakes per day at a sampling rate of 500 Hz. This high sampling rate is required to exploit newly developed theories on the frequency-dependence of shear-wave splitting (SWS). The array covered an area approximately 5 km North/South by 4 km East/West. During the experiment, the re-injection of water into the geothermal reservoir was halted for ten days during which seismic activity decreased. When injection resumed, seismic activity returned to its previous intensity. We use SWS measurements to image fracture locations, sizes, and orientations in the geothermal field. These fractures control the directions of fluid migration in the subsurface. This information will aid engineers in locating productive drill sites. This will decrease the cost of geothermal exploration, hopefully making this clean source of energy economically competitive with more traditional forms of energy.
T33A-1358 1340h
Near field and surface contributions to shear-wave splitting, waveform and traveltime sensitivity kernels
We calculate shear-wave splitting, waveform and traveltime Fr\'echet sensitivity kernels near the receiver in a homogeneous medium. Sensitivity kernels for anisotropic parameters are derived. Near-field and surface effects near the receiver are important when imaging the lithosphere as well as conducting global tomographic inversions. Sensitivity kernels are based upon finite-frequency seismology which accounts for the effects of perturbations off the ray path upon seismic observables at the receiver. In the far field the kernels are insensitive along the ray path, this is no longer true in the near-field region close to the receiver where the kernels are maximally sensitive. These sensitivity kernels account for near-field, far-field, surface waves and reflected waves off the free surface.
T33A-1359 1340h
A new global P wave model
Using data from different seismic phases and improved tomography codes we have constructed a new model of 3-D variations in P-wavespeed in Earth's mantle. First, we have improved data coverage by adding to the large volume of routinely processed ISC catalog (P, pP, PKP ) data, as reprocessed by Engdahl et al. (BSSA98), several sets of differential travel times that have been measured by waveform cross correlation. These data include PKP-Pdiff (Wysession) and various PKP branches (Creager and McSweeney), which help improve resolution of structure in the lowermost mantle, and PP-P (Bolton and Masters), which greatly improves the mapping of structure in the upper mantle away from belts of high seismic activity. Second, we follow Karason et al. (JGR01) and use 3-D sensitivity kernels to account for the fact that the data are measured at different frequency, and in different ways. The kernels allow low frequency data (e.g., PP-P, Pdiff) to constrain long wavelengths without preventing short period data (P, PKP) to resolve small scale structure. We use approximate kernels since our research has shown that for this application the precise shape of the kernels is less important than effects of parameterization and regularization. Third, the localization of sensitivity is further aided by the use of a grid that is adapted to sampling density, so that small scale structure can be resolved in regions of dense data coverage. Fourth, for the ray geometry part we use 3-D ray tracing to account for the non-linear effects of heterogeneity on ray geometry. Finally, we correct for 3-D variations in crust structure, using CRUST2.0, which helps reduce the artificial smearing of shallow structure along steep ray paths. We present the new model, along with resolution tests, and discuss improvements over previous efforts.
T33A-1360 1340h
Detailed Structure of the Upper Mantle Discontinuities Around the Japan Subduction Zone Imaged by Receiver Function Analyses
We applied receiver function (RF) analyses to P wave coda potions of teleseismic events observed at 138 broadband stations in Japan and Korea to procure the Ps phases converted at the upper mantle seismic velocity discontinuities. These broadband stations are consist of F-net, J-array, IRIS, Korean ones, which realize enough high-density spacing to reveal the detailed structure around the Japan subduction zone. After careful inspection, we retrieved 7,725 RFs with good SN ratios from 389 teleseismic earthquakes with the magnitudes of 5.5 and greater. The numbers of used RFs and stations are far larger than those in previous studies, which enables us to investigate the more details in the discontinuity structure. RFs are conventionally constructed through frequency domain division of radial components by vertical ones with a water level of 0.01. Gaussian filters of 1.0, 0.3, 0.17, 0.1 Hz low-pass filter are also applied, respectively. Assuming the phases in RFs are generated by Ps converted ones at depths, we transformed the time domain RFs to the depth domain ones referring to 1-D IASP91 velocity model. Further, in addition to this model, examining 3-D tomographic one and 1-D JMA one, we investigated whether lateral variation in velocity around the subduction zone shifts the discontinuities largely. In our obtained RF images, remarkable positive RF amplitudes corresponding to the Pacific plate (PAC), the 410 km and the 660 km discontinuities appear beneath the whole Japan Islands. The PAC can be traced from the Japan trench to a depth of about 200 km remarkably, and further down to a depth of about 400 km. Phase boundary of the 410 km discontinuity is locally elevated by 30 km inside the cold PAC where the PAC penetrates the 410 km. Positive RF amplitudes dipping westwards near a depth of 660 km describes the phase boundary affected by the stagnation of the subducted PAC on the 660 km discontinuity. These undulations corresponding to the 410 and 660 km discontinuity varies from 380 km to 410 km locally and from 650 km to 700 km gradually with 1-D velocity migration. To investigate the sharpness of the 410 km discontinuity, we examined the frequency dependence of RFs by changing the width of Gaussian filter. Our resultant RF images show that the phase boundary near a depth of 410 km varies in a relatively local scale, in other words, the thickness of the 410 km one beneath Korea spreads out broader than the one beneath Japan. In contrast, the 660 km discontinuity can be seen everywhere and its transition seems to be sharp. Comparison of the 1-D JMA velocity-migrated RFs with the ones referring to recent 3-D tomographic velocity models confirms these obtained undulations are not apparent ones caused by existing lateral seismic velocity variations. Indeed, RF amplitudes of dipping layer in the cross-section referring to 3-D velocity model would be traced well compared with the ones using 1-D JMA one. Also, the absolute depths of the 410 km and 660 km discontinuity with 3-D and 1-D JMA velocity models are imaged deeper than the one with 1-D IASP91 one. They were estimated from 390 km to 420 km and 680 km and 730 km. This is caused by that 1-D JMA and 3-D velocity models include the lower velocity portion than 1-D IASP91 one above a depth of 400 km.