T31A-01
The Role of Slab Width in Determining Trench Migration Velocity and Subduction Zone Curvature: Insight From Numerical Modeling and Global Kinematic Calculations
Subducting slabs provide the main driving force for plate tectonics and flow in the Earth's mantle. Geodynamic modeling provides insight into the kinematics and dynamics of subduction. Many previous numerical studies modeled subduction either in 2D space or in 3D space with essentially static models. Many early laboratory simulations of subduction only focused on 2D aspects of the modeling. In nature, subduction zones and their associated slabs are three-dimensional features with curved geometries and a limited trench-parallel extent (i.e. limited slab width). In addition, the trench velocity varies significantly along individual trenches (up to 20 cm/yr). Only recently have laboratory and numerical models started to focus on subduction processes in three- dimensional space. Here we present results from numerical simulations of free subduction in a stratified mantle reservoir, showing the progressive evolution of subduction in three-dimensional space. The models specifically investigated the influence of slab width on the subduction process, which was varied between 300 km and 7000 km. The sub-lithospheric upper and lower mantle reservoirs were modeled with linear viscosities, while the plate was modeled with a viscoplastic upper half and a viscous lower half. The results show that the trench migration velocity decreases with increasing slab width. The evolution of the trench geometry is also found to depend on the width of the slab with observed geometries ranging between concave, sub-linear and convex. The numerical results are compared to a new global compilation of trench velocity calculations and to subduction zone geometries on Earth, and are found to explain the first-order global trench migration patterns and subduction zone geometries for subduction zones on Earth ranging from narrow (300 km) to wide (7400 km). It is concluded that the simple rheologies and geometries adopted in the 3D numerical simulations can explain a wide variety of subduction zone characteristics as observed on Earth.
T31A-02
Subduction Zone Plate Boundary Coupling, and Subduction Angle Changes
Data from the ocean floor drilling programs, volcano flux, volcanic front migration, mountain building and gravity residual anomalies, allow modeling of the varying geometry and the forces involved in subduction zones. The Pacific plate subduction beneath North-East Japan has been relatively constant in direction and fairly steady in rate over more than 20 my. However, the tectonics and volcanism have changed markedly. The westward migration of the volcanic front over the last 20 my, and the change in tectonic stress from tension to compression at about 14 ma is well documented. We postulate that subduction geometry changed from "Marianas-type" with steep dip and weak coupling to its present 30 degree dip and strong coupling. Once the present geometry is approached, the apex of the mantle wedge is cooled sufficiently by conduction to the cold subducting slab and overlying plate, that its viscosity increases, flow avoids the corner, leading to further conductive cooling. Convection modeling (Kincaid and Sacks, 1997) suggests that the growth of a stiff, cold region trenchward of the volcanic front takes of order 3 million years. The depth range over which strong interplate coupling occurs would therefore increase. Finite element deformation and faulting modeling constrained by observed gravity residual anomalies, topography and heat flow (Huang et al, 1998) show mountain building and erosion strongly affected by this coupling force. The overlying plate is warped, and mountains (Kitakami range) build, and erosion increases. The cores retrieved from holes drilled into the sea floor during leg 186 of the ODP and legs 56 and 57 of the DSDP off the east coast of Japan show a spatially consistent increase in sedimentation rate at about 7 ma. The timing is consistent with the modeling and the migration of the volcanic front.
T31A-03
Seismogenic characteristics of the Mariana shallow thrust zone: What controls plate coupling?
The Mariana arc has been the type example of an "aseismic" or "decoupled" subduction zone since the concept was first proposed by Uyeda and Kanamori [1979]. Most past studies of the seismogenic zone characteristics have focused on "coupled" subduction zones that show large thrust earthquakes. Comparative studies of "decoupled" zones are essential for understanding the factors controlling the occurrence of large earthquakes. We use land and ocean bottom seismograph records from the 2003-2004 Mariana Subduction Factory Imaging Experiment to study the seismicity of the Mariana forearc. This 11 month experiment consisted of 20 broadband stations deployed on islands and 58 semi-broadband ocean bottom seismographs (OBS), with 14 OBSs located in the forearc. Events were detected and arrival times picked with Antelope package and relocated with a hypocentroidal decomposition relative location method. Overall, we have located ~ 500 events in the Mariana forearc between 16-19 N with mb 2 - 5. We determined the focal mechanisms of the largest events using a regional waveform inversion method. We have recently also detected episodic tremor that may be similar to that found at Cascadia, Japan, and several other subduction zones. The results indicate a lack of significant seismicity in the outer forearc beneath Big Blue and Celestial serpentinite seamounts. A few well-located events suggest that the shallow thrust zone is located at a depth of 15-25 km beneath the seamounts. Patches of high seismicity in the shallow thrust zone are located to the west of the seamounts at depths of 30-40 km. The patches are approximately 20 km in diameter and suggest heterogeneous faulting properties along the shallow thrust zone. Another zone of earthquakes at depths of about 60 km initiate beneath the seamounts and extend towards the west, representing faulting within the subducting plate. These events indicate that the lower zone of the Mariana double seismic zone initiates in the forearc and extends continuously to intermediate depths. We also detect episodic tremor that may be radiating from the downgoing plate interface deeper than the zones of high seismicity. We suggest that the relatively aseismic character of the outer forearc of the Mariana islands results from large-scale serpentinization of the outermost mantle wedge, as serpentinite is characterized by stable sliding rather then stick-slip behavior. In addition, the lack of large subduction zone thrust earthquakes may also result from the extreme heterogeneity of the shallow thrust interface, as revealed by the patches of high seismicity and aseismicity at depths of 30-40 km.
T31A-04 INVITED
Structural factors controlling inter-plate coupling and earthquake rupture process
Recent availability of a large number of ocean bottom seismographs (OBSs), a large volume of air-gun array and a long streamer cable for academics provide several new findings of lithospheric scale structures in subduction seismogenic zones. JAMSTEC has acquired long-offset seismic data using a super-densely deploy OBS (i.e. 1 - 5 km spacing OBSs along 100 - 500 km long profiles) in the Nankai seismogeinc zone, SW. Japan, since 1999. Long-offset multichannel seismic (MCS) data by a two-ship experiment, as well as conventional 2D MCS data, have been also acquired at a part of the profiles. Some of those profiles have been designed as combined onshore - offshore profiles for imaging a land-ocean transition zone. One of the most striking findings is an image of several scales of subducted seamounts/ridges in the Nankai trough seismogenic zone. We detected the subducted seamount/ridges, which are 50 - 100 km wide, distributing from near trough axis to ~ 40 km deep beneath the Japanese island. From a point of seismogenic process, an important aspect is that those structures are strongly correlated with slip zones of magnitude 8-class earthquakes, i.e.; subducted seamounts/ridge control the rupture propagations. Moreover, the most recent seismic study crossing the segmentation boundary between M=8 class earthquakes detected a high seismic velocity body forming a strongly coupled patch at the segmentation boundary. The numerical simulation incorporating all those structures explained the historic rupture patterns, and shows the occurrence of a giant earthquake along the entire Nankai trough, a distance of over 600 km long (Mw=8.7). The growth processes of a rupture revealed from the simulation are; 1) prior to the giant earthquake, a small slow event (or earthquake) occurs near the segmentation boundary, 2) this accelerates a very slow slip (slower than the plate convergent rate), at the strong patch, which reduces a degree of coupling, 3) then a rupture easily propagates through the strong patch when the next earthquake is nucleated near the segmentation boundary; consequently growing into a giant earthquake. The most recent result of a series of seismic surveys at the incoming plate of the eastern edge of the Philippine sea plate shows subduction and collision of fore arc ridge system. This structure may provide remarkably different inter-plate coupling at the easternmost part of the Nankai trough where the great Kanto earthquake (M=7.9) occurred in 1923 beneath Tokyo.
T31A-05 INVITED
Interseismic plate coupling, seismic rupture, and postseismic slip at the 2003 Tokachi-oki segment of Kuril Trench
We investigate interseismic plate coupling, the 2003 Tokachi-oki rupture and its afterslip using Japanese nationwide GPS network. The 2003 Tokachi-oki earthquake (M ~ 8) occurred at the western end of the Kuril Trench on September 25, 2003 (UT). We processed 1-Hz GPS for this earthquake, and inverted GPS displacement waveform to infer the coseismic slip distribution. Then we used daily GPS time series after the quake to infer the space-time distribution of the afterslip following this earthquake. We found that the high-speed rupture and afterslip are not overlapping, and this result support the idea that the frictional property of the plate boundary is significantly different between those two regions. Because we have cumulative slip at every epoch, we are able to calculate shear stress change using dislocation theory. We investigated the slip velocity dependence of the shear stress change, and found that the frictional property of the afterslip area would be steady-state velocity strengthening. On the other hand, it is widely believed that high speed rupture occurs at the steady-state velocity weakening region. Hence we suggest that the spatial inhomogeneity of the frictional properties caused the difference between slip styles in rupture region and afterslip reagion. Interseismic coupling inferred from GPS data also support this hypothesis.
T31A-06
Kinematic Coupling Along the Mexican Subduction Zone From GPS Data
Kinematic coupling (α) between the subducting Cocos and Rivera oceanic plates, and the continental North America plate is estimated as a ratio between the back slip on the fault plane and the convergence rate (NUVEL1A model). To obtain a distribution of the back slip ("negative" displacement at the plate interface) along the coast of Jalisco, Guerrero, Oaxaca and Chiapas we performed an inversion of the surface crustal deformation data from the permanent GPS network "SSN-Sismologia-UNAM" and campaign GPS measurements. The displacements for each cell of the plate interface model were calculated using closed form expressions developed by Okada (1992). The inversion based on a simulated annealing algorithm reveals a high coupling ratio, α ~ 1, at the seismogenic zone of 40-50 km width in Guerrero and Oaxaca, just beneath the coast line. Downdip from this area, a transition or partially coupled zone, α ~ 0.28, extends for a distance of about 200 km. For the Isthmus of Tehuantepec zone, at the southeastern Oaxaca, the inversion model shows a very low coupling ratio. In Chiapas, at the southern Mexico Pacific coast, the sparse distribution of GPS stations does not allow us to constrain a satisfactory model. Nevertheless, some preliminary results show a coupled area close to the trench. At the northwest, along the coast of Jalisco, the observed surface crustal deformation is mostly a result of subduction of the Rivera Plate beneath the North America Plate. For this area, the campaign GPS data were obtained from the UNAVCO database. Preliminary results show a highly coupled seismogenic zone. Strongly coupled areas that were determined from the inversion models are in good agreement with the seismogenic zones defined with the rupture areas of large Mexican subduction thrust earthquakes (M > 6.5) occurred last century.
T31A-07 INVITED
Modeling the Interseismic Coupling Phase, Silent Events and Subduction Geometry for the Guerrero Seismic Gap
The coast of a large portion of Guerrero state (Mexico) is a well known mature seismic gap. As part of monitoring this seismic gap, from late 90's decade, some permanent GPS have been installed and continuously operated in this zone recording the deformation regime associated to the steady state coupling phase. At the beginning of 2002, permanent GPS stations detected a drastic change of direction due to a slow slip event which lasted around 4 months (January-April) and had a seismic moment release equivalent to an earthquake of Mw~7.6. Recently (mid 2006), a new slow slip event was detected with a pattern of deformation that seems much more complicated than for the 2002 event. On the other hand, geometry of subducted plate and shallow crustal structure for this region remain not well constrained. For this reason an inter-institutional (UNAM-CALTECH and UCLA) project (MASE) was proposed in 2005. The main propose of this project is to constrain subduction geometry from an experiment consisting of a network of one hundred portable broadband seismic stations located on a line perpendicular to the Middle American Trench and crossing from Acapulco, Guerrero to Tempoal, Veracruz, Mex. In this work we present data modeling for the steady state and slow slip phases as well as some preliminary results of MASE experiment.