S23D-01
Assessment of Well-Constrained Seismicity and Focal Mechanisms in the Andaman- Sumatra-Java Subduction Systems
Most prior studies of regional seismicity in the Andaman-Sumatra-Java subduction zones have been limited to global catalog hypocenters that often have poorly constrained epicenters and depths. More than 8000 teleseismically well-constrained earthquakes occurring along this subduction margin during the period 1918- 2006 are relocated with special attention to focal depth. Reduced uncertainties of epicenters and depths that meet the stringent Engdahl-van der Hilst-Buland (EHB) location quality criteria are on the order of 15 and 10 km, respectively. The patterns arising from these relocations provide important insights into the complex spatio- temporal relationships among earthquakes along these subduction systems. Relocated hypocenters, when combined with CMT focal mechanism data, reveal complex details about subduction zone processes and microplate formation in the region, and help to clarify how strain is accumulated and released. Selected cross sections reveal in finer detail the shallow-dipping interplate zone, back-arc seismicity, and the intraplate seismic zone at depth in the mantle. Broad-scale features such as the abrupt increase in shallow strike-slip activity along the volcanic arc as subduction becomes highly oblique and the decrease in maximum depth of intraslab earthquakes are readily apparent in the global dataset. Offshore central Sumatra, there is a dearth of shallow, underthrusting earthquakes corresponding to the region affected by the 1833 great subduction earthquake; the seismic hazard along this segment of the Sumatra subduction zone remains high. Along Java, reverse fault mechanisms, rather than thrust earthquakes, dominate the shallow subduction zone seismicity. Comparison of relocated aftershocks and slip distributions of two Java tsunami earthquakes (1994 and 2006) in this region shows that most 1994 aftershocks occurred within the slab updip of the mainshock, but the 2006 aftershocks have a more complex pattern. These aftershock locations have implications for models of tsunami earthquake generation.
S23D-02
Tomographic Image of Subducting Lithosphere beneath Indonesia
We present tomography results for Indonesia, with emphasis on the Sumatra region, from the inversion of teleseismic arrival times for regional earthquakes occurring during the period 1918-2006 and global earthquakes occurring 1964-2006. The earthquake locations are well-constrained teleseismically by the Engdahl, van der Hilst, and Buland (EHB) method of single event relocation and cover the region 15 degrees S to 25 degrees N, 90 to 135 degrees E. We invert these data using a finely-gridded regional model (0.5 degrees horizontally) nested within a coarsely-gridded global model (5 degrees horizontally). Cell thicknesses increase with depth from the surface, where they are 35 km and 110 km thick for the regional and global grids, respectively. The use of the coarse global grid helps minimize the effects of lateral heterogeneity from outside the region of interest. We trace rays through the ak135 global velocity model for primary phases and depth phases. Inclusion of data from the 2004-2005 megathrust sequences greatly improves ray path sampling and resolution in the Sumatra region as compared to previous models for the region. Preliminary, single-iteration results show clear recovery of the high- velocity slab in much of the region, with significant variations in the dip of the slab observed along the 2004-2005 rupture zones, consistent with dips inferred by EHB event relocations. In addition, our results illuminate plate boundary structure at the northern end of the trench offshore Burma where oceanic subduction transitions into continental collision. Synthetic and checkerboard resolution tests will be used to assess the quality of the results, and future versions will benefit from an iterative solution and improved phase data.
S23D-03
Imaging the Andaman and Sunda Subduction Zones Using Regional Double-Difference Tomography
We present an extension of the double-difference (DD) local earthquake tomography algorithm to teleseismic scales and show initial results for the Andaman and Sunda subduction systems. The 2004 and 2005 great Sumatra earthquakes and the resulting aftershock sequences generated thousands of globally recorded events and illuminated the shallow seismogenic zone. We focus analysis on earthquakes that meet the stringent Engdahl-van der Hilst-Buland (EHB) location quality criteria, as these events provide the most reliable information for imaging the complexities of the seismogenic zone. Where broadband waveforms exist, we use a frequency- based automatic picking technique to identify additional first arrivals and depth phase onset times for inclusion in the dataset. Waveform cross-correlation is used to reduce relative picking errors between earthquakes with similar waveforms, leading to higher precision differential times. To extend the DD tomography code for use with teleseismic raypaths, we integrated a hybrid ray tracer that combines a finite-difference (FD) travel time calculator and pseudo-bending algorithm to solve for P and S first arrivals. This advancement allows us to use teleseismic data to more precisely determine the relative location of subduction zone earthquakes, especially their focal depths, and solve for 3D velocity heterogeneity. The FD travel time calculator is computationally stable and capable of computing travel times to all points in the model and can locate diffractions in ray shadow zones. It can find the correct solution even in strongly heterogeneous medium. However, its accuracy depends on grid spacing size and computation time may be unacceptable if the grid size is too small for the desired accuracy. Therefore, the FD travel time calculator is used to provide an approximate initial raypath. We then apply the spherical pseudo-bending algorithm of Koketsu and Sekine (1998) to further improve raypath accuracy. Because of non-linearity of the two-point problem, this algorithm may fail to find the true two-point ray path in a heterogeneous medium if the initial ray path to start with is far from the true solution. By combining the two algorithms, we can efficiently and accurately find the ray path solution even in strongly heterogeneous medium.
S23D-04
Structure and Evolution of the Accretionary Margin of Java-Sumatra. Seismic Data and Numerical Modeling Comparisons.
We present a numerical model for the evolution of an accretionary prism along a subduction margin. We find the mechanical partitioning of the growing prism into active region, abutting against a deformable backstop, and a relatively undeformed forearc basin is a function of the double tapered basal geometry, where the dip of the detachment is assumed to be opposite above oceanic or continental lithopshere. Varying properties of both materials and detachment can be used to adjust the surface slope and hence geometry of the system, but mechanical partitioning remains essentially the same with the regions becoming broader or narrower. The model appears to closely reproduce the geometry of the Sumatra-Java prism, where a high accretion margin has produced the same distinct mechanical units. Newly prestack depth-migrated marine seismic data reveal the extent and geometry of the active deformation of the deformable backstop, and give indications of some material passing into a subduction channel below the accretionary complex. The deformable backstop appears to be composed of multiple duplex structures, but present day tectonic activity is mostly in the form of transtensive or transpressive deformation, possibly reactivating older dip-slip, accretionary structures. The numerical approach used in the simulation (distinct elements) shows great promise in modelling large deformation in situations such as accretionary prisms, and has also been adapted to incorporate the role of fluid pressure and migration in tandem with large deformation (shortening of the order of 100's of kilometres).
S23D-05 INVITED
Results from SCS Profiling of the Sumatra accretionary prism: insights into tsnamigenesis
The SEATOS high resolution single-channel seismic reflection survey of the Sumatran accretionary prism depicts a landward-vergent thrust front, with active folding characterizing part of the December 2004 Mw9.2 earthquake rupture zone. Structure and bathymetry co-vary at distinct wavelengths along a 220-km-long profile crossing the prism and the Aceh (forearc) Basin. At the largest wavelength (tens of kms), the prism surface is defined by a steep (8-12 degrees), 55-km-wide outer slope, a 110-km-wide upper slope forming a broad depression between two forearc highs, and a 25 km-wide, steep inner slope between the landward high and the forearc basin. Anticlinal ridges spaced ~13 km apart display landward- and seaward-vergent folds along the inner and outer slopes, respectively; symmetric folding occurs across the upper slope. We suggest that the long-wavelength variations are consistent with the existence of a strong inner wedge beneath the upper slope. The ~13 km anticline spacing implies deformation of a slope apron deforming independently of this stronger wedge interior. Seismic profiles crossing the toe of the prism image a series of landward vergent, fault-related folds, suggesting that the shallow fill of the Sunda Trench is delaminated from the predominantly seaward-vergent plate boundary system and is uptilted along a triangle zone. Profiles crossing the seaward flank of the Aceh Basin reveal a near- vertical, undulatory deformation front that appears to mark the location of the West Andaman-Mentawai right- lateral strike-slip fault zone. Our model for prism architecture based on these geophysical results involves advance of the strong inner wedge during great earthquakes like the 2004 event, which then peels up shallower and less competent trench fill, deforming the toe and the upper slope of the forearc, producing massive uplift that is likely tsunamigenic. Seismic rupture was limited to the megathrust westward of the West Andaman fault and ROV observations showed no evidence for surface rupture along the strike-slip fault system. These results suggest strong strain partitioning along this oblique margin and highlight the hazards represented by the West Andaman and Sumatran faults.
S23D-06
Crustal structure of the 2004 Great Sumatra Earthquake epicentral zone from wide-angle seismic data.
The 26th December 2004 great Sumatra earthquake (Mw=9.1) is among the 4 largest earthquakes ever recorded and the largest of the last 40 years. It initiated at a depth of 20-30 km and ruptured about 1300 km of the Indo-Australian/Sunda plate boundary, from the vicinity of Simeulue Island up to the north of Andaman Islands. During the SAGER-OBS cruise 56 ocean-bottom seismometers were deployed on a profile which crosses the margin north of Simeulue Island, in the epicentral area. The profile is 252 km-long, resulting in a 4.6 km instrument interval. A seismic refraction source of 8,300 in3 was used to fire 2170 shots on the profile. Data quality is excellent with useful arrivals to distances up to 180 km from the shot position. A tomographic inversion of 33127 picked first arrivals allows first insights into the deep structure of the zone of origin of the great earthquake. The subducting oceanic slab can be imaged down to a depth of ~25 km, more than 100 km landward from the trench. Based on seismic velocities, there is no evidence of a crystalline backstop up to ~120 km from the trench axis, below the fore arc basin. This implies that a significant part of the seismogenic interplate would consist of a contact between ancient accreted sediments and the downgoing plate. A high velocity zone at the lower landward limit of the ray-covered domain, at ~23 km depth, may indicate a shallow continental Moho. This observation might imply that the 2004 megathrust earthquake nucleated on a portion of the plate contact that lies between the upper surface of the downgoing plate and the fore arc mantle. At the seaward end of the model, 4-5 km of sediments seem to overlie the oceanic crust at the trench, most likely consisting of trench fill and older hemi-pelagic sediments.
S23D-07
Spatio-temporal variation of seismogenic stress field in the source region of the 2004 Sumatra-Andaman earthquake
The 2004 Sumatra-Andaman earthquake occurred in an oblique subduction zone. The plate convergence direction significantly changes along the trench in the source region. The aftershocks activity and its variety of source mechanisms indicate the stress field in the source region is very complex. We conducted seismological analyses to investigate spatio-temporal variation of the stress filed in detail. We calculate angular deviation of earthquake slip directions from the plate convergence directions based on the Global CMT solutions (1976-2007, 209events). We also compare the coseismic slip direction (Tsai et al. 2005, Chlieh et al. 2007) with the plate convergence direction. Both results indicate that the coseismic slip direction is normal to the trench in the southern portion of the source region. This result implies that the fore-arc sliver motion accommodates the trench parallel component of the oblique plate convergence. On the other hand, in the northern portion around Andaman Islands, the fault slip direction deviates from the plate convergence direction by about 40 degrees, which is not perpendicular to the trench. The result suggests that stress partitioning due to the fore-arc sliver motion is not complete in the northern part. The large width of the fore-arc sliver and an effect of collision with the Eurasian plate at the northern edge of the sliver may be responsible for the incomplete stress partitioning. In order to discuss spatio-temporal stress variation with CMT solutions, we also conduct relocation of hypocenters in the source region by using the modified Joint hypocenter determination (MJHD) method. Initial hypocenter locations are assumed based on the ISC catalogue and Engdahl(2007), and 196300 P wave arrivals of 1807 events are reanalyzed. As a result, the average travel time residual reduced from 8.2 sec to 0.4 sec, and configuration of the subducted slab becomes clear. Combining the relocation result with the Global CMT solutions, various characteristics are identified. For example, afertshocks with high angle reverse faulting mechanism occurred just beneath the trench axis in the southern part. On the contrary, normal fault earthquakes occurred under the trench in the northern part, indicating a significant stress variation along the trench. In the southern part, normal fault events occurred near the plate boundary only after the mainshock, implying significant temporal stress change due to the mainshock.
S23D-08
Earthquake doublet that occurred in a pull-apart basin along the Sumatran fault and its seismotectonic implication
On March 6, 2007, an earthquake doublet occurred around Lake Singkarak, central Sumatra in Indonesia. An earthquake with magnitude (Mw) 6.4 at 03:49 is followed two hours later (05:49) by a similar-size event (Mw 6.3). Lake Singkarak is located between the Sianok and Sumani fault segments of the Sumatran fault system, and is a pull-apart basin formed at the segment boundary. We investigate source processes of the earthquakes using waveform data obtained from JISNET, which is a broad-band seismograph network in Indonesia. We first estimate the centroid source locations and focal mechanisms by the waveform inversion carried out in the frequency domain. Since stations are distributed almost linearly in the NW-SE direction coincident with the Sumatran fault strike direction, the estimated centroid locations are not well resolved especially in the direction orthogonal to the NW-SE direction. If we assume that these earthquakes occurred along the Sumatran fault, the first earthquake is located on the Sumani segment below Lake Singkarak and the second event is located at a few tens of kilometers north of the first event on the Sianok segment. The focal mechanisms of both events point to almost identical right-lateral strike-slip vertical faulting, which is consistent with the geometry of the Sumatran fault system. We next investigate the rupture initiation points using the particle motions of the P-waves of these earthquakes observed at station PPI, which is located about 20 km north of the Lake Singkarak. The initiation point of the first event is estimated in the north of the lake, which corresponds to the northern end of the Sumani segment. The initiation point of the second event is estimated at the southern end of the Sianok segment. The observed maximum amplitudes at stations located in the SE of the source region show larger amplitudes for the first event than those for the second one. On the other hand, the amplitudes at station BSI located in the NW of the source region show larger amplitude for the second event than that for the first one. Since the magnitudes, focal mechanisms, and source locations are almost identical for the two events, the larger amplitudes for the second event at BSI may be due to the effect of rupture directivity. Accordingly, we obtain the following image of source processes of the earthquake doublet: The first event initiated at the segment boundary and its rupture propagated along the Sumani segment to the SW direction. Then, the second event, which may be triggered by the first event, initiated at a location close to the hypocenter of the first event, but its rupture propagated along the Sianok segment to the NE direction, opposite to the first event. It is known that the previous significant seismic activity along the Sianok and Sumani segments occurred in 1926, which was also an earthquake doublet with similar magnitudes to those in 2007. If we assume that the time interval between the earthquake doublets in 1926 and 2007 represents the average recurrence interval and that typical slip in the individual earthquakes is 1 m, we obtain approximately 1 cm/year for a slip rate of the fault segments. Geological features indicate that Lake Singkrak is no more than a few million years old (Sieh and Natawidjaja, 2000, JGR). If the pull-apart basin has been created since a few million years ago with the estimated slip rate of the segments, we obtain roughly 20 km of the total offset on the Sianok and Sumani segments, which is consistent with the observed offset. Our study supports the model of Sieh and Natawidjaja (2000) that the basin continues to be created by dextral slip on the en echelon Sumani and Sianok segments.