T21B-0515 INVITED 0800h
Natural Examples of Olivine Lattice Preferred Orientation Patterns With a Flow-Normal a-Axis Maximum
Olivine lattice preferred orientation (LPO) due to ductile deformation is one of the main causes of mechanical anisotropy in the upper mantle and the patterns are useful to infer the direction of mantle flow from the seismic anisotropy in various settings. In subduction zones the mantle anisotropy near subduction boundaries suggests that olivine a-axes are arranged roughly perpendicular to plate motion. This anisotropy has been attributed to localized subduction-normal flow, applying a common type of olivine LPO with a `flow-parallel' a-axis maximum to the mantle. However, a recent deformational experiment provides an alternative interpretation that the B-type LPO with a `flow-normal' a-axis maximum can be developed in water-rich mantle above subducting slab. We document the widespread occurrence of B-type LPO in the Higashi-akaishi peridotite body, SW Japan, and examine the physical conditions in which it was formed. Our structural studies define four deformational phases in the Higashi-akaishi body (D1-D4) that are related to the tectonic evolution in the Cretaceous subduction zone at the Eurasian margin. The main deformational stage, D2, is associated with dynamic recrystallization of olivine to form porphyroclastic microstructure consisting of clear olivine neoblasts and porphyroclasts with abundant micro-inclusions. Parallel alginment of olivine neoblasts defines a stretching lineation (L2) and tectonic foliation (S2) and the D2 olivine LPO is identified as the B-type fabric with a-axes normal to L2, b-axes normal to S2 and c-axes parallel to L2. Micro-Raman spectroscopic analyses reveal that the syn-D2 micro-inclusions include hydrous minerals such as serpentine, indicating water-rich conditions for the D2 deformation. Garnet-orthopyroxene geothermobarometry applied to the D2 garnet peridotite reveals that the D2 stage was associated with the almost isothermal burial (700-800C, 2-3GPa). These D2 physical conditions in which the B-type LPO was formed are expected for mantle wedge near subduction boundaries. These results confirm the suggestion from the experimental work and imply that mantle flow above subduction zones is parallel to plate motion.
T21B-0516 0800h
Slip, Stress, and Lattice Preferred Orientation in Ultramafic Xenoliths From Cerro Mercedes, Costa Rica
Ultramafic xenoliths are abundant in Quaternary lavas from Cerro Mercedes, Costa Rica, a small back-arc cone roughly 70 km behind the Central American volcanic front. These rare subduction zone xenoliths offer a glimpse into the deformation conditions in the mantle wedge behind an active subduction zone. Examination of more than 75 xenoliths reveals a lack of garnet, suggesting an origin depth of approximately 60-70 km. These samples are mainly lherzolites and harzburgites with some dunites and pyroxenites. The xenoliths examined thus far are coarse grained and have either a protogranular or porphyroclastic texture. The presence of both of these textures implies varying amounts of deformation between samples. In addition, many xenoliths have deformation textures such as undulose and kinked olivine grains and recrystallization around grain boundaries. By comparing the orientation of olivine crystallographic axes and kink-band boundaries on a Universal Stage Microscope, we have determined that slip is dominantly along the [100] crystallographic axis in the samples, with both [100](010) and [100](001) slip systems present. These systems match slip observed in other upper mantle samples and samples experimentally deformed at mantle conditions, indicating that the deformation features seen in the xenoliths were created at depth. Stress estimates taken from the recrystallized grain size and the size of subboundaries created by dislocations match well with published values for other xenoliths, reinforcing our interpretation that the deformation seen in these xenoliths occurred in the upper mantle. Currently, we are analyzing the lattice-preferred orientations of these xenoliths. Many xenoliths have strongly aligned kink band boundaries, and the alignment of these lineations implies that the xenoliths have a strong preferred orientation. The strength of the axis alignment, along with the relation between the crystallographic axes and the lineations, will help to reveal the conditions in the mantle wedge in Costa Rica. Since plate motions have not changed appreciably since the eruption of the xenoliths approximately 1 My ago, the degree of anisotropy and type of preferred orientation in the xenoliths likely reflects current conditions behind the Central American Arc.
T21B-0517 INVITED 0800h
Multiple Layers of Anisotropy in the Chile-Argentina Subduction Zone, South America
We examine shear wave splitting in teleseismic and local phases to observe seismic anisotropy in part of the South American subduction zone. The data is from the CHARGE network, which traversed the Andes Mountains of Chile and Argentina across two transects between 30$\deg$ and 36$\deg$ S. Beneath the southern part of our network, fast polarization directions from teleseismic phases are consistently trench-parallel, while in the northeastern part of the network fast directions are trench-normal. This trend appears to be correlated with a changing geometry of the subducted slab. Subduction zones may exhibit multiple layers of mantle anisotropy, corresponding to the mantle wedge, subducting slab, and the asthenosphere in the upper mantle below the slab. Based on previous work, we have suggested that the largest source of the anisotropy sampled by teleseismic phases is localized below the subducting slab. Preliminary measurements of anisotropy in the mantle wedge (sampled by local S-waves) reveal quite variable patterns in azimuth and magnitude, similar to other subduction zones. There is also an observed variation in splitting parameters in the teleseismic events with backazimuth that may be explained by a combination of wedge and below-slab anisotropy. However, preliminary two layer models with reasonable values for the wedge and below-slab components do not generally fit observed trends with backazimuth, therefore there may be three layers of anisotropy: the mantle wedge, the subducting slab lithosphere, and an asthenospheric layer. Because the magnitude of anisotropy due to the slab and to the wedge is relatively small compared to total anisotropy, removing it's effect from the total anisotropy leaves an asthenospheric component that is very similar to the total observed anisotropy by teleseisms. Calculations attempting this removal to date have made small adjustments to the telseismic values that actually make the asthenospheric component look more consistent across the network.
T21B-0518 0800h
Shear-wave Polarization Anisotropy in the Mantle Wedge Beneath Hokkaido and Tohoku, NE Japan
We investigated shear-wave polarization anisotropy in the mantle wedge beneath northeastern (NE) Japan. Waveforms of intermediate-depth earthquakes recorded by numerous seismic stations were analyzed by the cross-correlation method [e.g., Ando et al., 1983], and 2528 splitting parameters were obtained. In Hokkaido, obtained results show that most of the leading shear-wave polarization directions (fast directions) at stations in the back-arc side are nearly perpendicular to the trench-axis, whereas those at stations in the fore-arc side are sub-parallel to it. A similar pattern of the shear-wave splitting has been also observed in Tohoku. We infer that the anisotropy caused by lattice-preferred orientation of olivine, which is probably attributable to the upwelling flow potion of the mechanically-induced convection in the mantle wedge [Hasegawa and Nakajima, 2004], is a likely candidate for the shear-wave splitting in the back-arc mantle wedge. If this is the case, the present results for Hokkaido may indicate that the upwelling flow direction is sub-parallel to the maximum dip-direction of the subducted slab, oblique to the relative plate motion direction. Although it is not clear what causes the anisotropy in the for-arc side, where the trench-parallel fast directions are observed, similar features have been observed in other areas of Tohoku [Okada et al., 1995; Nakajima and Hasegawa, 2004]. Average delay times between leading and following shear-waves observed at stations in the back-arc side are 0.1 _E0.4 s, and those at stations in the fore-arc side are smaller (0.05 _E0.15 s). The distribution of average delay times observed in Hokkaido dose not show such a clear spatial pattern as that observed in Tohoku, where delay times apparantly increase toward the back-arc side.
T21B-0519 0800h
Shear Wave Splitting and Seismic Anisotropy in the Mariana Mantle Wedge from a Combined Land-Sea Deployment
We investigate anisotropy and mantle flow in the Mariana mantle wedge using data from the Mariana Subduction Factory Imaging Experiment. The passive component of this experiment consisted of 20 broadband PASSCAL seismographs deployed on the island chain and 58 ocean-bottom seismographs across the arc and the backarc. This dense seismic network recorded from May-June, 2003 until April-May, 2004 and 66 instruments returned useful data. Analysis of shear wave splitting from S waves propagating from intermediate and deep earthquakes in the Mariana slab to this array provides an unparalleled opportunity to study the pattern of anisotropy and mantle flow in the arc and backarc region. We analyze local S wave anisotropy using the shear wave splitting method of Silver and Chan (1991). To check the robustness of our results, we also utilize the cross-correlation method (Bowman and Ando, 1987) and visually inspect particle motions. Preliminary results mostly based on the 20 land seismic stations indicate a complex pattern of anisotropy in the region. Fast splitting directions and magnitudes are highly dependent on source location and depth. Along-arc delay times generally range from small for southern hypocenters near Guam ($<$ $\sim$0.4 s) to larger ($>$ $\sim$0.6 s) for sources in the northern part of the arc. Fast directions generally range from N-S to NW-SE at southern stations, and N-S to NE-SW at northernmost stations. We interpret these results in terms of mantle flow using the conventional assumption that the fast anisotropic direction results from lattice preferred orientation and denotes the direction of maximum extension. There is little evidence of a large-scale convergence-parallel mantle circulation patterns previously reported for this region based on observations from a single station at Guam. Rather, these results indicate that along-strike flow patterns predominate in most regions of the arc. Complete analysis of splitting directions and magnitudes in the region will provide a rare opportunity to observe seismic anisotropy across a subduction zone forearc, arc, and backarc spreading center.
T21B-0520 0800h
Travel Time Residuals and Anisotropic Attenuation in the Central Volcanic Region, NZ
New Zealand's Central Volcanic Region is characterized by the subduction of the Pacific Plate underneath the Australian Plate. SKS splitting results show large delay times and trench parallel fast directions. Teleseismic broadband data were acquired in 2001 along a trench perpendicular line (striking northwest to southeast). An unusual attenuation and travel time pattern was observed on an SKS and SKKS record comparing fast and slow components. Going from south-east to northwest across increasing path length in the mantle wedge, the amplitude of the fast component decreases and it appears to arrive earlier with respect to the ak135 propagation model, whereas the amplitude and arrival time of the slow component stays more consistent. To examine the phenomena we determine travel time residuals relative to arrival times estimated from the ak135 propagation model. For earthquake to station distances from 20 to 120 degrees, P waves residual times vary between -3.66 s to 3.24 s. Values for fast and slow S waves range between -5.00 s and 5.00 s. Results are dependent on backazimuth. Waves arriving from west to northwest show a decrease of residual time and therefore an increase of apparent/mean velocity from west to east along the line. This is consistent with what would be expected due to the increasing length of the travelpath through the higher velocity slab. On the other hand there is an increase in travel time residuals for waves arriving from eastern directions, consistent with longer paths through the mantle wedge. The high frequency fast and slow shear waves examined by this technique show similar trends. We attribute the apparent early arrivals of the fast wave in the SKS/SKKS records to waveform modification during attenuation. We are currently examining the spectral characteristics to quantify the anisotropic attenuation. The anisotropic attenuation cannot be caused by fluid filled cracks, as this would cause the amplitude and arrival time of the slow component to decrease. Another hypothesis to be tested is that the attenuation could be due to the loss of energy during phase conversions from aligned heterogeneities.
T21B-0521 0800h
Mantle Tectonics of a Plate Boundary: the North Island of New Zealand
We constrain the extent of lithospheric and asthenospheric deformation by measuring and modelling anisotropy in the North Island of New Zealand through shear wave splitting of teleseismic waves, petrophysical analysis of mantle xenoliths and finite-element modelling of the mantle deformation in a back-arc spreading environment. We use seismic data collected on the GeoNet broadband stations deployed between 2001 and 2004 and a portable deployment in the Northland region. The results for the eastern part of the North Island confirm the earlier pattern. The fast directions are roughly NE/SW, which is parallel to the strike of the Hikurangi subduction zone. In the north of the North Island, two stations present back-azimuthal dependence of the splitting parameters. This dependence suggests a broader and more complex anisotropic effect, such as a dipping axis of symmetry or multiple layers, might be present beneath New Zealand, influencing the measurements on a regional scale. Results from the Northland deployment suggest that splitting from SKS phases decreases as the effects of present plate boundary deformation weaken and allow us to quantify the extent of the plate boundary deformation zone. The petrophysical analysis of mantle xenoliths from the Raglan (North Island) region allows us to constrain the lithospheric contribution to the observed seismic anisotropy. The spinel lherzolites analysed represent mantle material from depths shallower than 75 km and temperatures in excess of 1000°C up to 1150°C. Their extraction ages are less than 2 Myr. The maximum intrinsic anisotropy on the S-wave measured, 3.5%, seems too low to explain the delay times. This indicates that asthenospheric deformation plays the major role there. Through modelling of the lithospheric deformation, derived from surface deformation rates, in the North Island transtensive Central Volcanic Region, we will compare the anisotropy values obtained from the xenoliths to the modelled regional values derived from geodetic movement rates.
T21B-0522 0800h
Global and Local Variations in Slab Depth Beneath arc Volcanoes
It is commonly accepted that the volcanic front at subduction zones occurs above where the slab reaches depths of 100-120 km. To test this assertion, we reanalyzed a variety of teleseismic and local seismicity catalogs for global variation, and have found that the actual range in slab depths beneath the arc front is from 60 to 225 km, with an average depth of 110 km. We have compiled data for 33,000 km of the global arc system, incorporating depths and other subduction parameters for 732 arc volcanoes. Slab dips vary globally from 20 to 70 degrees, convergence velocities range from 10 to 140 km per Ma, not including back-arc spreading, and incoming seafloor ages vary from 0 to 160 Ma. With the exception of some arcs with complex Quaternary histories, we have found that slab depth varies inversely with respect both to convergence velocity and to descent rate, at least when averaged along strike over arc segments, in agreement with the similar measurements of England et al. (2004). However, over short distances some arc segments exhibit abrupt variations in arc-slab geometry. The most notable cases are the 90-km increase in depth from Sumatra to Java, which occurs over an along-strike distance of 300 km, and the 60-km increase in depth from Nicaragua to Costa Rica, which occurs in 50 km. Several smaller variations occur in the Aleutians and Central America, among groups of 4-10 volcanoes. Most of these local variations correspond with offsets in the arc, not slab contortions, indicating that upper-plate tectonic processes may exert considerable control on the ascent of magmas from the mantle, at least over distances of a few hundred km. These data show that the geometry of the mantle melting region beneath volcanic arcs must vary considerably.
T21B-0523 0800h
Deep Seismic Discontinuity Structure Beneath New Zealand
We have created receiver functions using broadband seismic stations on either side of Cook Strait, New Zealand. These waveforms have been stacked in common conversion point (CCP) bins and reveal lateral variations in seismic structure above 700 km depth. The tectonic relationships beneath New Zealand vary from westward subduction in the north through transform along the Southern Alps to eastward subduction in the very south. The precise geometry of plate interactions at depth is not well known and is only loosely constrained by seismicity, shear wave splitting, seismic tomography, gravity studies, and plate reconstructions. Several outstanding questions include whether the northern subducting plate continues to subduct south of recorded seismicity in the Nelson region and over what lateral extent does the subducting plate penetrate the mantle transition zone. We image the top of the subducting Pacific Plate as well as several mid mantle seismic discontinuities including the 410 km and 660 km upper mantle transition zone discontinuities. Synthetics are generated for our ray geometry and compared with the CCP stacks and suggest that the subducting plate dips steeply under the southern extent of our stations, the southwestern edge of the Marlborough Fault System. The upper mantle transition zone discontinuities are consistent with the penetration of a cold slab southwest beyond recorded seismicity, i.e. the transition zone is thickened in the region of the cold subducted slab. We interpret the significant northeast deepening of the 410 km seismic discontinuity along the subducted plate to be due to differences in the downward component of subduction rate and the kinetics of the olivine phase transition. We observe splitting of the P to S-wave conversion from the 250 km and deeper discontinuities which indicates seismic anisotropy above 250 km depth oriented approximately north-northeast. The orientation reflects transform relative plate motion and/or mechanisms producing a trench parallel fast axis of anisotropy.
T21B-0524 0800h
Imaging the Subduction Factory Beneath Central America: The TUCAN Broadband Seismic Experiment
The subduction factory processes the solids and volatiles entering oceanic trenches, and produces volcanic arcs, deep earthquakes, and long-term modifications to the deep earth. Central America exhibits some of the global extremes in the factory operation: in the Nicaragua volcanic arc, fluxes of geochemical tracers associated with subducting sediment are among the highest on the planet, while in the adjacent Costa Rica arc, many of the same tracers are weak to absent. The factory must be behaving differently at depth over short distances. To understand this behavior, we deployed in July-August 2004 a seismic array termed TUCAN (Tomography Under Costa Rica And Nicaragua), 48 broadband PASSCAL seismographs operating until early 2006, funded by the NSF Margins program. Deployment has been rapid and successful. The array design features dense transects across the two arc segments and sparser along-strike arrays. This design allows for high-resolution imaging of the subducting plate and mantle wedge across the transition from the high-flux regions of central Nicaragua to the low-flux Costa Rica arc and backarc. Velocity and attenuation tomography, together with measurements of anisotropy, provides constraints on the thermal structure and melt pathways within the mantle wedge. Receiver functions, scattered waves and tomography allow both upper and lower-plate structure to be assessed. Initial inspection of these data show marked variations in signal amplitudes and travel times across the arc, as expected for a hot mantle wedge. Both regional and teleseismic signals show the presence of strong secondary phases emerging from the slab in Costa Rica, both from regional and teleseismic signals, indicating the presence of pronounced low-velocity layers associated with subducting crust. Previous work has indicated the low velocity layer beneath Nicaragua is unusually slow by global standards, perhaps indicating high volatile content. These data will allow a comparison with velocities within the Costa Rica slab, to test the notion that the variations in arc geochemistry reflect differences in slab water content.
T21B-0525 0800h
Preliminary Earthquake Locations From the Costa Rica Component of the TUCAN Network
The TUCAN (Tomography Under Costa Rica And Nicaragua) network is a multi-institutional and international project funded by the US National Science Foundation as part of the Subduction Factory Initiative of the MARGINS program. The project's objective is to tribute to the understanding of the local tectonic processes that cause extreme changes in the fluxes of geochemical tracers associated with subducting sediments from Nicaragua to Costa Rica. The observed variations hint that the subduction factory must be behaving differently at depth over short distances. The TUCAN network is a 48 station broadband seismic array composed of two dense (~10 km spacing) lines across the volcanic arcs in Nicaragua and Costa Rica and two other lines (~30 km spacing) parallel to the trench, one in the forearc, at the foothills of the volcanic arc, and the other along the back arc ~70 km from the arc. This network was designed and deployed this summer to collect, for 18 months, seismic data that will be analyzed in many different ways including precise earthquake locations, local velocity and attenuation tomography, receiver functions, anisotropy studies and local, regional and teleseismic wave phases analysis. We will present preliminary earthquake locations recorded at the Costa Rica component of the TUCAN network. The cross line in Costa Rica extends from where the seismogenic zone of large underthrust earthquakes occur along the Nicoya segment at depth of 30 km, to 40 km into the backarc from where the deepest portion (~150 km) of the seismic slab projects to the surface. This geometry combined with OVSICORI's permanent seismic network promises precise local earthquake locations.
T21B-0526 0800h
Dynamics of Subduction and Plate Motion in Laboratory Experiments
3-D laboratory experiments have been designed to investigate the way slab-bearing plates move during subduction inside the mantle. The boundary conditions are as simple as possible: a viscous plate rests in the center of a large tank filled up by honey and subducts under its negative buoyancy once a small instability at the plate edge is created. Varying thickness, width, viscosity, density of the plate and mantle, three characteristic modes of subduction are observed: a retreating trench mode (Mode I), a retreating trench mode following a transient period of advancing trench (Mode II), and an advancing trench mode (Mode III). These modes are characterized by different partitioning of the amount of subduction into plate and trench motion. Our experiments show that the velocity of subduction can be roughly modeled by the dynamic interaction between acting and resisting forces and that some parameters such as the slab viscosity or thickness have the opposite influence than the one usually suggested in the literature. This result is interpreted as the consequence of the dependence (measured in the experiments) of the slab radius of curvature on the plate viscosity and thickness. However, it is still far from being simple to predict how the trench and plate move. Our results suggest that the complexity of the style of subduction could also be controlled by simple geometrical rules of a plate bending inside a stratified mantle: our planet system is in the crucial range where the length of the slab pulling down the plate is about the double of its radius of curvature.
T21B-0527 0800h
3-D Laboratory and Numerical Models of Mantle Flow in Subduction Zones
Analogue and numerical studies are powerful tools to gain insight on the subduction process. Here we investigate some results from both approaches in order to characterize the induced flow triggered in the mantle by slab motion. The fluid velocity field in our 3-D laboratory experiments is reconstructed and analyzed through the PTV (Particle Tracking Velocimetry) image analysis technique, which provides a set of velocity vectors centred with particle centroid positions. Numerical investigation is approached by means of the finite element code Citcom (e.g. Moresi & Solomatov, 1995, Zhong et al., 1998; obtained from geoframework.org), solving the equations for conservation of mass, momentum and energy for an incompressible viscous spherical shell.
T21B-0528 0800h
Constructing a 3D Dynamic Model of Mantle Wedge Under the NE Honshu Subduction Zone, Japan
In our previous studies, we propose an existence of small-scale convection under the NE Honshu subduction zone, Japan, to explain the origin of hot fingers. In this presentation, we refine our model by comparing the model results with the geophysical and geologic data observed there. Assuming that the temperature anomalies have a close connection with the seismic anomalies, we constrain the geometry of the low viscosity wedge (LVW) overlying the slab, which may be produced by the water dehydrated from the subducting slab. Our preferred model suggests step-like low velocity anomalies above the subducting slab rather than smooth anomalies subparallel to the subducting slab, since the shallow nature of the LVW is required for the small-scale convection to occur at the back-arc end of the LVW. A movement of cold plumes generated at the back-arc end of the LVW may be related to a possible migration of volcanism from back-arc to volcanic front side. To be consistent with the observation, which may suggest the migration rate of $_Lsim$ 2 cm/yr, models require weak couplings between the mantle wedge and the underlying subducting slab, whose speed of subduction is $_Lsim$ 10 cm/yr. 3D calculations of our final model show fairly continuous and strong temperature anomalies (several hundreds degrees) under the volcanic front and weak finger-like temperature anomalies (several tens degrees) behind them, which are similar to the pattern of the seismic tomography. The time-dependent behavior of the temperature field shows that the pattern of fingers flip-flops with a time-scale equal to the ratio of the horizontal extent of the LVW to the migration speed of cold plumes. This change of pattern of fingers may have an important implication for understanding the past distribution of volcanism.
T21B-0529 0800h
Tomographic simulation of cold hydrous plumes above slabs
We have calculated the hetergeneous seismic velocity fields taken from numerical modeling of thermal-chemical plume structures above subducting slabs, using 50 million active tracers. Our 2-D numerical model of subduction zones combines the effects of thermal-chemical buoyancy as well as the effects of both hydration and melting. Numerical modeling suggests that partially molten hydrated upwellings above the subduction slab can produce a colder thermal anomaly of 300 to 400 degrees within the mantle wedge! Due to an assortment of crustal rocks and penetration of aqueous fluids, cold plumes may have an upward velocity greater than one meter per year. Initiation of these plumes is most likely due to massive release of water from a subducting slab due to putrefaction of serpentine at depths greater than 100 kilometers. The seismic velocities are computed from a thermodynamic model that accounts for phase transitions. These include melting, for the various lithologies of interest including dry and hydrated mantle, sediments, as well as basaltic and gabbroic oceanic crust. The model provides estimates for the proportions, compositions and seismic velocities of the stable phases as a function of pressure and temperature. Aggregate seismic velocities are computed using the Voight-Reuss-Hill averaging scheme. According to our simulation of 2-D seismic velocity field during the propagation of hydrated but cold plumes, water and temperature will interact producing either positive or negative seismic anomalies depending on water/temperature ratios. These ratios cause the seismic signal to change during plume propagation with positive anomaly been characteristic for the area of cold plume initiation close to the slab.
T21B-0530 0800h
The influence of the basalt-eclogite transition on incipient subduction dynamics
Subduction becomes self-sustaining when the negative buoyancy of the slab is larger than the sum of elastic bending, fault friction, and viscous drag. The oceanic lithosphere, consisting of basaltic crust overlying mantle which is depleted in iron, is compositionally buoyant relative to the asthenosphere and will resist subduction. This is not a critical impediment for mature subduction zones because the basaltic crust should metamorphose into denser eclogite, with the result that the slab has nearly neutral compositional buoyancy at depths greater than this phase transition. During subduction initiation, however, compositional buoyancy may offer significant resistance to subduction and initially limit subduction to plates which have gained sufficient negative buoyancy from cooling. We have shown that while including a 6 km thick basaltic crust does not greatly affect the maximum force required to bend the lithosphere, it substantially increases the net work required to reach a self-sustaining state. These earlier models did not consider the basalt-eclogite transition. We use a visco-elastoplastic numerical method to model the evolution of the force balance during incipient subduction occurring at a fixed convergence rate. Phases are tracked using a particle advection scheme. In some cases, the buoyant basaltic crust buoyantly detaches from the mantle lithosphere and rises to the base of the over-riding lithosphere. Buoyant detachment removes most of the compositional impediment to subduction initiation, but will deny the slab of the additional driving force that would occur after the transformation of basalt to eclogite. Crustal detachment is encouraged by mechanical thickening during convergence, and therefore may depend on the degree of coupling and fault strength at the plate interface.
T21B-0531 0800h
Subduction Following Terrane Accretion: Insights From Two-Dimensional Numerical Experiments
Oceanic subduction is a consequence of the negative buoyancy of the cold and dense lithosphere. Development of subduction with a continental sliver is more problematic, because continental lithosphere is positively buoyant, owing to the light, thick crust. In a subduction process, the collision between a continental fragment and a continent may result in different modes: either the subducted slab breaks off and the plate boundary ceases to be active or the plate boundary reorganizes to continue plate convergence. One possibility for continued convergence is that the dense mantle of the fragment detaches from the buoyant crustal part. If such detachment does not take place the slab will break-off or the continental fragment will be subducted with the oceanic lithosphere. Our general aim is to identify the parameters that act as switches between the different modes of collision. We use a finite element method to solve the mechanical equilibrium equation with a plastic-powerlaw-elastic rheology and the heat equation, since the viscosity is temperature dependent. The model is driven by buoyancy forces. The subduction process is absolutely sensitive to the viscosity structure. We adopt the rheology of Karato and Wu (1993), with dislocation creep in the hot and shallow mantle (non Newtonian rheology) and diffusion creep in the deep upper mantle and lower mantle (Newtonian rheology). We introduce a critical ingredient, a low viscosity wedge (LVW) which has been argued to result from added volatiles and especially water. The reason is that a LVW decouples the lithosphere and the slab and reduces the excessive dynamic topography in the overriding plate since the suction pressure is lower (Billen and Gurnis 2001). We present results of numerical experiments aimed at establishing the most significant parameters, like relative strengths and age of the oceanic lithosphere.
T21B-0532 0800h
Thermal Models for Kamchatka and the Position of the Volcanic arc
A finite element method is applied to model the thermal structure of the subducted Pacific plate and overlying mantle wedge beneath the southern part of the Kamchatka peninsula. A numerical scheme solves a system of 2D Navier-Stokes equations and a 2D steady state heat transfer equation with strong temperature dependent viscosity. A model with the isoviscous mantle exposed very low temperatures ($\sim$800$^{o}$ C) in the mantle wedge, which cannot account for the magma generation below the volcanic belt. Instead, a model with the strong temperature dependent viscosity shows an essential rise of the temperature in the wedge. At the temperature of more than 1300$^{o}$ C beneath the active volcanic chain, the melting of the wedge peridotite becomes possible. Albeit the subducting slab below the Kamchatka peninsula is rather old ($\sim$70Myr), some frictional heating along the interplate between the subducting oceanic slab and the overlying Kamchatka peninsula lithosphere would be enough to result in the melting of the subducted sediments and basaltic oceanic crust. Dehydration ($>$5$%$wt.H$_{2}$O release) occurs in the subducting slab as a result of metamorphic changes. As a consequence, the hydration of the mantle wedge peridotite might produce its melting. The melted material may rise to the base of the continental crust as diapir like blobs. Considering that the melting processes in the subducting plate generate the most of the volcanic material, we developed a dynamic model which simulates the migration of the partially melted buoyant material in a form of blobs in the viscous mantle wedge flow. The blobs with the diameter of 0.4 - 10.0 km rise up to the base of the continental lithosphere within 0.002 - 10 million years depending on the blob diameter and surrounding viscosity. Whether we set up the starting point for the blobs to be on the slab surface just below the volcanic chain, then the blobs touch the base of the continental crust in the same point beneath the volcanic chain. The thermal structure obtained in the model with temperature dependent viscosity is used to estimate the seismic P-wave velocity anomalies (referenced to PREM) associated with the subduction beneath Kamchatka. A low velocity zone ($\sim$-7$%$ velocity anomaly) is obtained beneath the volcanic belt and a high velocity anomaly ($\sim$4$%$) for the cold subducted lithosphere. These results agree with the seismic tomography inversion inferred from P-wave arrivals.
T21B-0533 0800h
Finite Element Models of Viscous Flow in the Mantle Wedge Above a Subducting Slab for Different Relative Subducting and Overriding Plate Motions
A series of models of viscous flow in the mantle wedge above a subducting slab are generated with a two-dimensional finite element code used to solve for time-dependent, viscous deformation of the mantle in response to relative motion of the slab and the overriding plate. Resultant velocity fields are used to compute finite strain for mantle particles traversing the wedge; these patterns of finite strain may serve as a proxy for mantle fabrics formed through lattice preferred orientation (LPO) of olivine crystals. Boundary conditions applied to the wedge model to represent relative motions of the subducting and overriding plates have a strong influence on the appearance of mantle wedge flow. In particular, distinctive flow patterns result from an overriding plate velocity that is either faster or slower than the subducting plate velocity. The resulting geometry of mantle flow is further modified in a three-plate scenario, where two overriding plates move at different velocities along the top of the mantle wedge. Flow fields generated from these boundary conditions are computed for both simple, isoviscous and isothermal cases as well as more complex rheologies such as a power-law relationship between stress and strain rate with temperature-dependent viscosity. For the case of isoviscous and isothermal flow fields, migration of the overriding plate away from the trench results in mantle material being drawn from deeper depths than the case of a stationary overriding plate. In addition, preliminary results indicate that temperature-dependent viscosity may result in even steeper particle paths from deeper in the mantle into the wedge corner. This distinction may have important implications for the thermal structure of the mantle wedge. Finite strain is computed for particles traversing each of these flow fields, yielding a predicted olivine LPO for each case. Important differences in these LPO fields imply that seismic waves traversing the mantle wedge above a subducting slab may yield waveforms with a distinctive signature reflecting the specific geometry and characteristics of the mantle flow field through its particular pattern of olivine LPO. Combining seismic anisotropy measurements with flow modeling may thus provide important constraints for flow in the mantle wedge above a subducting slab.
T21B-0534 0800h
Three-Dimensional Numerical Models of Subduction and Subduction-Induced Mantle Flow
The kinematics of subduction and its influence on mantle convection and plate-scale deformation have been the focus of numerous geodynamic studies [{\it e.g. Garfunkel et al., 1986; Gurnis and Hager, 1988; Zhong and Gurnis, 1995; Christensen, 1996; Olbertz et al., 1997; Conrad and Hager, 1999; Eberle et al., 2002}]. Most geodynamic models have considered only two-dimensional aspects of subduction dynamics by incorporating the assumption that subduction zones are infinite in trench-parallel extent. However, natural subduction zones are intrinsically three-dimensional, due in part to their limited lateral extent. Lateral length scales of natural subduction zones vary from only a few hundred kilometres (e.g. the Calabrian, Hellenic and Scotia slabs) to several thousand kilometres (e.g. the Aleutian, Indonesian, Northwest Pacific and South American slabs). Here, we present results from three-dimensional numerical experiments that simulate lithospheric subduction and subduction induced mantle flow for slabs with a varying lateral extent.
T21B-0535 0800h
Variable Thermal Conductivity, Slab Mineralogy, and Subduction Rates
Models of subducting slabs have suggested that metastable olivine (MO) in the slabs' cold interiors can act like parachutes and reduce their density ($\rho$), lowering the slabs' velocities by up to 30% [{\it Marton et al.}, 1999; {\it Schmeling et al.}, 1999] These models, however, use constant values of thermal conductivity ({\it k})to solve the heat flow equation. Models that use {\it k}s that are functions of {\it P, T,} and mineralogy have wedges of MO that are 20-30% smaller in cross-sectional area and whose maximum extents are 30-50 km shallower [{\it Hauck et al.}, 1999; {\it Marton et al.}, 2004]. As a result of the decreased amount of MO, the parachute effect should also be decreased. Using the same thermo-kinetic model as {\it Marton et al.} [2004] I determine the mineralogy of three sets of subducting slabs and calculate the driving forces acting on them. Terminal velocities ($v_t$) are found via balances of the driving forces and the opposing viscous drag forces. For the same two sets of slabs examined using the previous constant {\it k} model [{\it Marton et al.}, 1999] (thermal parameter $\varphi$ = 3500-12000 km), this model shows reductions that are approximately half those of the former, with maximum reductions of 1-1.25 cm/yr or $\sim$15%. A third set, with older lithosphere (100 My) and $\varphi$ = 5200-17000 km, show reductions of up to $\sim$2 cm/yr or 25%. In addition, there should be a negative feedback between the amount of MO and subduction velocity, narrowing the range of subduction rates [{\it Marton et al.}, 1999; {\it Tetzlaff and Schmeling}, 1999]. This is tested by feeding $v_t$s back into the model, adjusting the durations of the iterations. The resultant $v_t$s and amounts of MO achieve a steady state within a few My after the slabs' tips exit the transition zone. In this case, the size of the parachute effect is dependent on lithospheric age. One set that has constant starting velocities and variable ages shows the same trend as without the feedback, but with $v_t$ changes $\sim$1% smaller. The other two sets with constant ages each have $v_t$ changes that are constant, +2.2% for 70 My old lithosphere, -1.3% for 100 My old lithosphere, that are in-line with the trends of the first group.
T21B-0536 0800h
Numerical simulation for spontaneous generation of one-sided subduction with hysterisis-dependent rheology
Two-dimensional numerical simulations are conducted to build a self-consistent subduction model using a newly developed code based on finite volume method. This code can perform simulation with non-uniform grid spacing to solve fine structure at the plate boundary efficiently. We utilize a thin lubricating layer as thin as an oceanic crust that depends on the hysterisis of the past fracture (Honda {\it et al}., 2000; Nakakuki and Hamada, 2003) with an extended Boussinesq fluid in a 2-D rectangular box. A pre-existing weak zone is given at the oceanic-continental plate boundary as an initial condition. Otherwise, no artificial forces are applied to the surface boundary and plates in the modeling. Results show that a subduction process of an oceanic plate was produced in an one-sided regime spontaneously, with a realistic narrow low viscosity zone formed at the plate boundary. We also examined the effects of tensional and compressional strengths of the subducting and/or overriding lithosphere accounting for some recent laboratory experiments. As the tensional strength is weaker than the compressional strength in the lithosphere, the ratio of tensional strength to compressional strength (1, 1/2, or 1/3) was tested systematically in the yielding condition. When a steep dip angle (45$^{\circ}$) of the initial weak zone (IWZ) was set, weak tensional strength in the subducting lithosphere promoted subduction initiation more effectively than in the cases with a shallow dip angle (18.4$^{\circ}$). Nonetheless, this strength did not affect the shape of subducting slab that directs downward. The initiation of subduction was constrained by the imposed IWZ at the plate boundary but the subduction was driven by negative buoyancy. Trench retreat took place when the thin overriding lithosphere had weak tensional strength. In this case, the shallow dipped subduction style was generated.
T21B-0537 0800h
Mantle anisotropy above the deflected Pacific slab beneath Northeast China
Systematic spatial-variations of SKS splitting parameters (fast polarization direction PHI and splitting time DT) have been observed at global and regional broadband seismic stations in NE China and Mongolia. Based on the characteristics of the resulting splitting parameters, the study area can be divided into western and eastern regions by the western boundary of the Song-Liao Basin (SLB). A DT of about 1.5 s is observed at the eastern stations, corresponding to an anisotropic layer of about 150 km thick. Most western stations show significantly smaller DT (ranging from 0.5 to 1.0 s). The dominant PHI measurements at stations in the eastern region are about 100 degrees from the north, which are almost exactly the same as the relative plate motion directions between the Eurasian and Pacific plates calculated at each of the stations. Spatially inconsistent PHI measurements are observed at stations in the western region. The boundary between the eastern and western regions is consistent with the western limit of the subducted Pacific plate in the mantle transition zone (Widiyantoro, 1997). The slab is deflected horizontally by the 660 km discontinuity to the west by several hundred kilometers. While the difference in lithospheric deformation between the eastern and western regions could result in the observed spatial variations in anisotropy, the observations can be best explained by the nearly E-W flow in the mantle transition zone induced by the subducting slab.