T24B-01 16:00h
Three-Dimensional Thermal and Chemical Structure of the Subduction Zone Upper Mantle Beneath the Philippine Sea
Three-dimensional thermal and chemical structure of the upper mantle beneath the Philippine Sea is investigated using recent seismic velocity [Gorbatov and Kennett, 2003] and attenuation [Shito and Shibutani, 2004] tomography data together with mineral physics model. In order to separately determine the thermal and chemical anomalies, one needs to have a multiple data set (such as Vp, Vs, and Q anomalies) for each point and a relationship between these physical/chemical anomalies and seismological observables. In this study, we use the latest knowledge of mineral physics to derive such a relationship. The mineral physics observations indicate: (1) the major element chemistry has only small effects on Q but has some effects on seismic wave velocities, (2) temperature has important effects both on velocities and Q, and (3) water content has a large effect on attenuation and a modest influence on velocities. Our recent analysis [Matsukage et al., 2004] shows that the influence of major element chemistry is only modest and can be described by a single parameter such as the concentration of magnesium or opx. In this scheme, three unknowns (temperature, water content and major element chemistry) are related to three observables (Vp, Vs, and Q) through a matrix that contains physical parameters. This non-linear equation is solved by iteration to determine the three unknowns. The error analysis shows that temperature anomalies and water contents are well-resolved in most regions, but the major element chemistry is poorly resolved because the velocities are only weakly sensitive to the major element chemistry. High temperature anomalies are found in shallow regions (< 200 km) and the wedge mantle beneath the active volcanic chain. And the maximum high temperature anomaly is +300 degree. Regions of high water content are found in the deep upper mantle (and perhaps in the shallow upper mantle near the slab). The maximum content of water in the deep upper mantle is estimated to be ~10 times higher than normal mantle. This suggests that some water is carried into this depth range (perhaps carried by lawsonite). The inferred thermal and chemical structure is largely consistent with the tectonic history of this region involving a long history of subduction and active magmatism and indicates that the effects of subduction are not limited to the source region of most of arc volcanism and extends to far deeper regions.
T24B-02 16:15h
Mantle Temperature Variations in Active Back-Arc Basins Inferred from Seismology, Petrology, and Bathymetry
Active backarc spreading centers show wide variations in axial depth and petrological characteristics. We investigate the correlation between these characteristics and the seismological structure of the North Fiji, Mariana, Lau, and S. Scotia backarc basins. These basins are chosen because they have large spatial dimensions and good waveform paths allowing accurate determination of the average upper mantle structure. For each backarc, vertical and transverse seismograms are inverted from several representative paths traversing the basin, with source to station distances of 650-1300 km. A variety of earthquake depths permits greater resolution of deeper structure. Synthetics are calculated using a reflectivity method, and the nonlinear inversion is carried out using a niching genetic algorithm, which allows alternative local minima to be explored. The inversion solves for separate path-averaged SH and SV velocity structures, with a penalty function to ensure that any polarization anisotropy is required by the data. The results show that all four backarc systems are characterized by low seismic velocity and moderate levels of polarization anisotropy (1-4.5%), with the anisotropy largely confined to the upper 100 km. Substantial differences in seismic velocity between the backarc regions occur at depths of 40-100 km, with differences of up to 7% between the slowest (Lau) and the fastest (Mariana) structures. The mantle seismological structures correlate with major element systematics and bathymetry of the basins, indicating differences in average upper mantle temperatures. The Lau basin shows shallow bathymetry and low Na8, high Fe8, and high Ca/Al, all indicating higher mantle temperatures (potential temperature of 1450\deg C from the major elements). The Mariana Trough shows the opposite trends in all these variables, indicating low upper mantle temperatures (1360\deg C), and the North Fiji basin is intermediate (1410\deg C). The South Sandwich backarc shows intermediate values of seismological structure and ridge elevation, but major elements indicate colder mantle temperatures (1350\deg C). The petrological indicators suggest a ~100\deg C range in mantle potential temperature, but these differences are not large enough to explain the seismological or axial depth data. Some variation in the percentage of partial melt between the hot and cold spreading centers could enhance the differences in seismic velocity. In addition, some contribution from temperature variations below 100 km depth is required to satisfy the ridge elevation data.
T24B-03 16:30h
Thermal modeling of the central Alaska subduction zone
The Pacific plate subducts beneath southern Alaska and produces most of the features commonly associated with subduction, including great earthquakes at shallow depths, intermediate-depth earthquakes to 150 km, a seismically slow mantle wedge, and upper-plate deformation. However, volcanism is virtually absent along the eastern 350 km of the subduction zone. This may be because temperatures in the mantle wedge are unusually low or because the slab does not devolatalize. The recent BEAAR broadband seismic experiment provided high-resolution images of the slab and wedge beneath central Alaska. These images provide some of the best constraints anywhere on the geometry of subduction at depth and the thermal state of the mantle wedge. Receiver functions show the top of the subducting plate to 150 km depth as a low-velocity channel, relative to surrounding mantle. To achieve such low velocities the subducting crust must not convert to eclogite, perhaps because it remains too cool to do so. Intraslab seismicity lies inside the low-velocity channel and gradually migrating from its top to bottom, suggestive of a dehydration isotherm stepping into the slab as it warms. Attenuation tomography shows that the overlying wedge is warm as other subduction zones. This subduction zone is also notable for its unusually shallow dip through the thrust zone. We developed a thermal model of this subduction zone using the geometry determined by the seismic experiment with a non-Newtonian rheology in the dynamic and a 50 km thick rigid lithosphere on top of the wedge. The hypocenters follow the 600 C contour predicted from our model. Below 150 km temperatures become high enough to allow for the conversion to eclogite and the hottest part of the wedge extends not closer than 450 km from the trench. This suggests that the characteristics of the central Alaska subduction zone can be satisfactorily explained by thermal models similar to those in Honshu and Cascadia (e.g., Van Keken et al., Gcubed, 2001) without invoking a special role for subduction zone fluids.
T24B-04 16:45h
Complex upper mantle anisotropy beneath Japan: shear wave splitting observations and geodynamical modeling
The characterization of seismic anisotropy in the upper mantle has the potential to elucidate patterns of mantle deformation. In the context of a subduction zone, this line of inquiry is both especially intriguing and especially challenging, because the presence of volatiles or melt in the mantle wedge or three-dimensional flow induced by complex slab morphology can complicate the relationships between deformation and anisotropy. In this paper we report the results of an intensive effort to fully characterize the shear wave splitting behavior at broadband seismic stations in Japan due to deformation-induced anisotropy in the upper mantle. Japan is an excellent area to study complex anisotropy associated with subduction zone processes; it overlies two different subducting plates with different ages, thermal histories, plate velocities, and slab morphologies. Therefore, Japan can serve as a natural laboratory to test hypotheses relating to the effects of melt, stress variations, and volatile content on the relationships among tectonic processes, deformation, anisotropy, and splitting observations. We have assembled a database of teleseismic shear wave splitting measurements for F-net, a network of more than 65 broadband stations throughout Japan. We have measured splitting parameters for over 1900 individual S, SKS, and SKKS phases. At many F-net stations, the observed splitting patterns are complex and cannot be explained with simple, single-layer anisotropic models. Complexity in splitting pattern appears to be spatially correlated with complexity in slab morphology. We observe trench-parallel fast directions in the Ryukyu arc, which rotate to trench-perpendicular moving northward along the array into Kyushu and along the downdip direction of the Phillippine slab. This rotation from trench-parallel fast directions close to the trench to trench-perpendicular farther away from the trench is also observed in the Kanto-Shikoku regions, moving downdip above the Pacific slab, and on Hokkaido. We seek to explain this persistent feature, along with other features of our shear wave splitting dataset, by comparing splitting observations with a suite of geodynamical models that take into account the unique slab geometries beneath Japan.
T24B-05 17:00h
Mantle Deformation Beneath the Southern Cascadia Subduction Zone
The objective of this study is to determine the location and extent of seismic anisotropy within the southern Cascadia subduction zone to constrain mantle deformation in the region. Of particular importance is to provide an improved understanding of the role of water in strain development and seismic anisotropy in hydrated mantle. Previous seismic, geodynamic, and petrological studies of the region suggest that the mantle wedge in this region is significantly hydrated and serpentinized. Southern Cascadia is therefore an integral component in the evaluation of the relationship between observations of seismic anisotropy and mantle deformation in subduction zone settings. Data for this study comes from the 1993-1994 IRIS/PASSCAL CASC broadband seismic array, deployed by Oregon State University researchers across a $\sim$300 km swath beginning at the Oregon coast and across the Cascades Mountains. Station spacing ranges from 4 to 8 km, providing an excellent dataset for a detailed investigation of possible small-scale variations in seismic anisotropy. To this end, I apply several shear wave splitting analysis techniques and error estimation methodologies to core shear phases such as SKS to provide constraints on the location, extent, and orientation of mantle deformation across the region. The range of analysis methods, including particle motion linearization, transverse energy minimization, and error surface stacking, is designed to provide confirmation that calculated shear wave splitting parameters from each methodology are robust. Preliminary results indicate that fast polarization directions are oriented ~E-W over the western part of the array, rotating to $\sim$N50E near the eastern end of the array. Resolvable station-to-station variations in fast directions are also evident for some regions, providing a clue of small-scale variations in seismic anisotropy across the array. Splitting times range from 0.8 s to 1.6 s, but currently these variations are not well resolved between regions. While these results are preliminary, a full analysis of this dataset will provide new constraints on the lateral and depth distribution of seismic anisotropy beneath the southern Cascadia region.
T24B-06 17:15h
Sensitivity of Lower Mantle Seismic Anisotropy Beneath Subduction Zones to Mantle Viscosity Structure and Mineral Inherited Deformation
Until recently, the mantle between 410km and the top of the D'' was thought to isotropic. Wookey et al (2002) report shear-wave splitting observations generally between 3 and 6secs from deep focus events in the Tonga-Kermadec subduction zone, recorded in Australia and show that the observations are generated by near-source anisotropy located in the top-most lower mantle. We investigate, using geodynamic modelling and seismic ray tracing, the effect mantle viscosity structure and mineral inherited deformations have on the predicted magnitude of shear-wave splitting in the top-most lower mantle. We use finite element (FE) modelling to predict subduction zone stresses arising from incompressible fluid flow driven by the body forces arising from the negative buoyancy of the subducting slab. We model the flow and stress field at the point in time when the subducted slab reaches the 660km phase transition, and not the development of subduction. The FE model is driven dynamically by the excess density (50kg/m$^{3}$) of the subducting slab within the upper mantle, rather than using plate-like velocity boundary conditions. Large deviatoric stresses (maximum values 40 MPa) are generated in a broad region (lateral wavelength 800km) in the topmost lower mantle. These stresses could induce mineral alignment in a broad region (lateral wavelength 800km) in the topmost lower mantle below the slab. We model finite strain accumulated by a mantle parcel as it propagates through the FE fluid flow models. Strain fields are mapped into seismic anisotropy and we then ray trace from the base of the subducting slab to predict shear-wave splitting along ray paths in the top-most lower mantle to teleseismic distances of $30\deg$-$60\deg$. For a viscosity model proposed by Steinberger (2000), which has viscosity increases at 410km, 660km and in the lower mantle, 5-10secs of shear-wave splitting is predicted compared to only 0-5secs for a uniform mantle viscosity. For the Steinberger viscosity model, we predict that shear-wave splitting reduces from 7-10secs to 4-6secs when deformation above 410km is ignored, and this further reduces to 2-3secs when all deformation accumulated above 660km is zeroed. Similar patterns are seen when other viscosity profiles are used. The dependence on the magnitude of the predicted shear-wave splitting on the mapping of finite strain to anisotropy has been explored, and compared to LPO calculations in the upper 410km of our subduction models. Predicted shear-wave splitting magnitudes using the Steinberger viscosity structure with the effect of mineral inherited deformation included are comparable to those observed from the Tonga-Kermadec subduction zone.
T24B-07 17:30h
Evolution of subduction and Back-arc Extension at a Continental Margin
In central North Island, New Zealand, there exists one of the more complete geological and geophysical records of subduction initiation and evolution. Geological data provide us with evidence of how subduction initiated in late Oligocene times, while geophysical data provide an image of how back-arc spreading and an active mantle wedge have developed beneath central North Island in the past 5 my. Principal information on subduction initation processes come from oil-industry bore hole data located $>$ 500 km from the trench where subduction initiated. These data show a rapid and coeval platform subsidence across a wide area ($> $200 km) of western New Zealand. This is interpreted as being due to a broad subduction-induced, hydrodynamic flow in the mantle. Nearly 20 my of compression and foreland basin development in western North Island followed the platform subsidence. At around 5 Ma a rapid switch in tectonics took place that was marked by the development of lithospheric, dome-like uplift centred on the central North Island. This uplift signal (2500 m of rock uplift or 1200 m of tectonic uplift) is principally derived from mudstone porosities that are used as a proxy for exhumation. From about 4 Ma to the present andesitic and then rhyolitic volcanism developed across the domal uplift, accompanied by extension and tectonic rotation of central North Island. Recent geophysical experiments in the central North Island have begun to define the structure and conditions within the subjacent mantle wedge. Pn wave speeds are exceedingly low (7.4-7.8 km/s) and the attenuation (Q-1) is high. Explosion seismology results show the pre-existing greywacke-schist crust (6 km/s) has been stretched to at least half its original thickness (i.e. 30 to 15 km). Beneath 15 km we see, from wide-angle reflection analysis, a reflective sequence down to about 18 km where velocities increase rapidly from 6.7 to 7 km/s. At 20 $\pm$ 2 km P-wave speeds of $\sim$ 7.4 km/s are measured and these are seen from passive seismic studies to extend down through the mantle wedge. We interpret the rocks from 15 to 20 km as new lower crust and from below 20 km as an anomalous upper mantle. On the basis of our seismic velocity model and the rock uplift data constraints are placed on the buoyancy and degree of partial melt in the mantle wedge. Our best estimate is that the mere replacement of say 70 km of the mantle lid by asthenosphere (density contrast $\sim$ - 40 kg/m3) is insufficient to explain the rock uplift. Instead we require a density contrast more like -70 to -100 kg/m3, which implies an important negative buoyancy contribution from fluids and/or melts in the mantle wedge.
T24B-08 17:45h
Factors Controlling Slab Roll-Back and Back-Arc Extension: Insights From Numerical Models
Although subduction is a first order plate tectonic process, the factors controlling the dynamics of slab roll-back and back-arc formation are still not very well understood. The major driving forces for subduction and slab roll-back are well established as the slab pull and ridge push forces, their relative importance and the relative importance of forces modifying and interacting with these driving forces is, however, not very clear. A number of forces may resist subduction and roll-back of the slab. 1) Normal and tangential forces resisting downwelling of the slab, 2) Bending resistance in the slab at the trench and at the 660 km discontinuity, 3) Resistance to lateral flow of the upper mantle below and above the subducting slab. To investigate the relative contribution of these resisting forces we use 2D plane strain thermo-mechanical finite element models. The model evolution is calculated using an Arbitrary-Lagrangian-Eulerian (ALE) method for the finite element solution of incompressible viscous-plastic creeping flows (Fullsack, 1995). In a first set of models we test the relative roles of bending resistance and upper mantle viscosity with a subducting plate where an overlying plate is not included. The models extend from the surface to 660 km depth. The upper surface of the model is free to move. Upper mantle rheology is linear viscous, whereas the rheology of the subducting slab is either linear viscous or combined linear viscous and von Mises plastic. Reflective and periodic boundary conditions are used. The slab is allowed to sink in the underlying mantle under its own weight. The models indicate that the resisting forces form the primary control on the rate of subduction and roll-back where the velocity of trench retreat depends linear on the viscosity of the upper mantle. Variation of roll-back with viscosity of the subducting slab is minor. Interaction of the slab with the 660 km discontinuity results in a small, $<$ 5 %, decrease in roll-back velocity. This suggests a subordinate role of bending forces. The results are compared with a scaling analysis of relative importance of contribution forces. In a second set of models we investigate interaction of the subducting slab with the overlying plate and specifically focus on factors that may control the opening of a back-arc basin. The down going plate is driven by a kinematic boundary condition, far from the zone of subduction. After an initial stage of far-field driven contraction, the negative buoyant down welling of the mantle lithosphere may drive continued formation of the subduction zone leading to mature subduction. The models suggest that two major factors control whether the subducting system develops an extensional back-arc system: 1) The relative contribution of the imposed far-field velocity and roll-back velocity, 2) The strength of the overlying plate. We investigate the roles of small scale convection, enhanced by corner flow above the subducting plate, and far field plate velocity on the opening of the back arc basin.