S42B-01 INVITED 10:20h
Spatial Distribution of Seismic Tremors During Episodic Tremor and Slip Events in Southern Vancouver Island
We study the seismic tremors that occurred during the last two Episodic Tremor and Slip (ETS) events in southern Vancouver Island. A combined dataset, including three-component broadband seismic waveforms recorded by both permanent and temporary stations, is used to delineate the spatial distribution of ETS tremors and their origin times. The majority of ETS tremors are located by the Source-Scanning Algorithm that systematically scans the entire model space for seismic sources consistent with the observed waveforms. For a few relatively large and isolated tremors, the solutions are further verified by conventional earthquake location methods and/or 3-D relocation technique. The epicenters of most tremors are confined to a limited band bounded approximately by the horizontal locations of 30- and 45-km depth contours of the subducting plate interface. The depth range, on the other hand, is wide and spans from the upper crust to within the subducted Juan de Fuca slab, with a peak at the depth of 25-35 km. Detailed uncertainty study rejects the possibility of the wide depth range being an artifact of analysis. There seems to be an anti-correlation between the locations of tremors and background seismicity. Moreover, many ETS tremors appear to coincide with the strong seismic reflectors within the overriding crust where no local earthquakes have been observed. A comparison of frequency spectra for tremors and for local earthquakes shows that tremor spectra are comparable to those of magnitude ~1.5 earthquakes in the frequency range of 1-5 Hz, but their corresponding high-frequency (>5 Hz) content is an order of magnitude smaller. These observed characteristics suggest that the ETS tremors and background seismicity involve different physical processes.
S42B-02 INVITED 10:35h
Causes of co-seismic rupture segmentation and inter-seismic slow slip in the Nankai seismogenic zone revealed from active source seismic studies
In order to obtain structures affected on co-seismic and inter-seismic processes in the Nankai seismogenic zone, we have carried out a series of active source seismic surveys since 1997. The most important finding from the surveys is a ridge and seamount subduction preventing the rupture propagation of the last 1944 Tonankai and the 1946 Nankai mega-thrust earthquakes. Remaining issues to be solved by seismic surveys are causes of a slow slip at a deeper part of the seismogenic zone and a structural factor controlling the segmentation boundary between the Tonankai and Nankai earthquakes. In this paper we present results of recent active source seismic studies to address those questions. Data from an onshore-offshore integrated seismic survey was acquired along a 485 km long profile crossing the eastern Nankai Trough and central Japan from the western edge of the Izu island arc. The observed inter-seismic slow slip event is situated at the middle of the profile. Highly reflective subducting oceanic crust is imaged from those data at 25 - 45 km depth beneath central Japan, and shows a 2 km high doming structure at 30 - 40 km depth, indicating a subducted ridge structure. This highly reflective subducted crust overlaps with the high Poisson_fs ratio zone imaged in a seismic tomography study. We propose, from these structures, that the slow slip can be attributed to a high pore pressure zone, which significantly extends the conditionally stable region in the subducted crust. The most recent wide-angle seismic data for imaging the segmentation boundary between the 1944 Tonankai and 1946 Nankai earthquake were acquired in January to February, 2004 along three margin parallel profiles (150 _E190 km long). We found that a low velocity uppermost mantle underlying a thin subducted oceanic crust at the segmentation boundary. This structure is interpreted as a subducted fracture zone. The low velocity uppermost mantle may be attributed by serpentinized mantle, which is created by a hydration reaction along the fracture zone.
S42B-03 10:50h
Aseismic Slip on the Hikurangi Subduction Zone, New Zealand.
In October 2002, an aseismic slip event of 20--25 mm magnitude was observed on two continuous GPS (CGPS) instruments near Gisborne, North Island, New Zealand. This event was the first of its kind to be documented with CGPS in New Zealand. It occurred over approximately ten days and is interpreted to be the result of about 20 cm of reverse slip on the Hikurangi subduction interface offshore of the north-eastern North Island. Since October 2002, smaller aseismic slip events (of 4--8 mm magnitude) have been recorded on CGPS stations near Gisborne, and Hastings to the south. Using campaign GPS data from the region we are attempting to better constrain the Gisborne 2002 event and quantify the distribution, frequency and magnitude of silent earthquakes on the northern Hikurangi subduction margin. We have analysed nearly ten years of regional campaign GPS records (1995--2004) in conjunction with recent CGPS and seismic data. We use station position time-series as the primary means of identifying transient behaviour, as would be expected from an aseismic slip event. The large October 2002 aseismic slip event is not obviously discernible in any of the campaign time-series, indicating that such events may occur on a more regular basis than previous geodetic data have revealed. Work is underway to investigate a possible correlation between the October 2002 aseismic slip event and an anomalous seismic swarm above the Hikurangi subduction margin offshore from Gisborne. We also intend to analyse broadband seismic records for low-frequency tremor that may be associated with the aseismic slip events. By combining recent, high-quality CGPS data with campaign data spanning ten years and regional seismicity data, we are able to re-evaluate existing models of deformation for the northern Hikurangi margin, and propose new models based on the known occurrence of aseismic slip. Greater understanding of the role of aseismic slip in accommodating relative plate motion on the northern Hikurangi margin will lead to more accurate assessment of seismic hazard posed by this subduction zone to New Zealand.
http://www.geonet.org.nz
S42B-04 11:05h
Fluid Activity Around the Downward Extension of the Seismogenic Fault of the 2000 Western Tottori Earthquake Inferred From Deep Low-Frequency Earthquakes
Low-frequency tremors were newly detected in the forearc region of the Nankai and Cascadia subduction zones recently. They are associated with the subduction of the young plates and attributed to the fluid activity around the plate boundary. On the other hand, there is another example of low-frequency events in the backarc region in southwest Japan that is associated with active faults. One example is the western Tottori area, where we had a Mw=6.7 earthquake in 2000. It is an unusual example because the seismogenic fault is outlined by an intense aftershock activity, beneath which many deep low-frequency (DLF) earthquakes were observed. DLF earthquakes were observed at depths of around 30 km beneath the aftershock activity. A fault model derived from the coseismic crustal movements (Sagiya et al., 2002) indicates that the DLF earthquakes are located around the downward extension of the fault. The DLF events are classified into three groups in features of the waveform. Type-1 are the most commonly observed ones. One of them shows a single-force type source mechanism (Ohmi and Obara, 2002). Type-2 events have larger P-wave onsets compared to those of type-1 events. Magnitudes of the type-2 events are slightly larger than those of type-1 events. They have been observed since mid 2002. Assuming that type-2 events are caused by shear faulting, we estimated the seismic moment and source dimension from the source pulse. Relation between the source dimension and moment indicates that the stress drop of the type-2 events are extremely low compared to those of ordinary earthquakes. It suggests the existence of soft materials such as fluid saturated gauge zone at the fault interface. Type-3 event is a tremor-like event observed in April 2003. We examined the tilt data in the region if the associated slip of the fault is observed. However, it was difficult to detect the tilt change more than 1.0 \times $10^{-7}$ radian, which is apparently equal to 1.3 cm slip on the fault model of Sagiya et al. (2002). As we described, observed features suggest the fluid activity in the focal region of the DLF events and is also supported by the seismic tomography analysis (e.g. Zhao et al., 2004). It shows the existence of low velocity bodies in the focal region of the DLF events, that reflects the fluid related to the dehydration process of the subducting Philippine Sea plate. Recent studies (e.g. Iio and Kobayashi, 2002) proposed that the seismogenic faults have downward extension in the lower crust, whose aseismic slip accumulate stress on the seismogenic faults in the upper crust and controls the occurrence of the earthquake. Hypocenters of the DLF earthquakes discussed in this paper are distributed around the deeper extension of the shallow aftershock distribution and probably located on the downward extension of the seismogenic fault of the Western Tottori earthquake. It is important to understand the nature of DLF events beneath active faults, in relation to the behavior of fluids in the lower crust that might affect the aseismic slip of the downward extension of the seismogenic faults and control the occurrence of the shallow crustal earthquakes.
S42B-05 11:20h
Array observations and analyses of Cascadia deep tremor
The July 8-24, 2004 Cascadia Episodic Tremor and Slip (ETS) event was observed using three small aperture seismic arrays located near Sooke, BC, Sequim, WA, and on Lopez Island, WA. Initial tremor burst epicenters located in the Strait of Juan de Fuca and were calculated using the relative arrivals of band-passed, rectified regional network signals. Most subsequent epicenters migrated to the northwest along Vancouver Island and a few occurred in the central to southern Puget Sound. Tremor bursts lasting on the order of a few seconds can be identified across the stations of any of the three arrays. Individual bursts from distinct back-azimuths often occur within five seconds of each other, indicating the presence of spatially distributed but near simultaneous tremor. None of this was visible at such a fine scale using Pacific Northwest Seismograph Network (PNSN). Several array processing techniques, including beam-forming, zero-lag cross correlation and multiple signal classification (MUSIC), are being investigated to determine the optimal technique for exploring the temporal and spatial evolution of the tremor signals during the whole ETS. The back-azimuth and slowness of consecutive time windows for a one half-hour period of strong tremor were calculated using beam-forming with a linear stack, with an nth-root stack, and using zero-lag cross-correlation. Results for each array and each method yield consistent estimates of back azimuth and slowness. Beam-forming with a nonlinear stack produces results similar to the linear case but with larger uncertainty. Among the arrays, the back-azimuths give a reasonable estimate of the tremor epicenter that is consistent with the network determined epicentral locations.
S42B-06 11:35h
Estimates of Fault Area and Rupture Velocity for Long Duration Subduction Zone Earthquakes
Previous studies of global subduction zone earthquake rupture processes have found examples of shallow earthquakes that exhibit depth dependent durations, with the longest durations observed for the several of the shallowest earthquakes. The long durations, after scaled for earthquake size, are comparable to those durations found for tsunami earthquakes. One model for tsunami earthquakes involves shallow faulting in regions of low fault zone rigidity or other variable friction conditions to explain the observed long earthquake source durations and efficient tsunami generation. A model of variable rigidity, and thus variable rupture velocity, is one end-member explanation for the global earthquake dataset as well. Depth dependent durations can also be explained through a model of variable stress drop. In the case of variable stress drop, rupture velocity and fault zone material properties are constant, suggesting that duration variations indicate variable scaled rupture areas for each earthquake. Here we analyze a subset of the global dataset, including earthquakes in the Alaska-Aleutian, Chile, Kamchatka, Peru, and Mexico subduction zones in an attempt to distinguish between the two end member models. This subset contains the longest duration earthquakes, as well as additional earthquakes with shorter durations for comparison. More refined estimates of the source time functions, durations, and depths are presented with analysis for source finiteness, rupture directivity effects, and variable focal mechanism effects. Fault rupture areas and rupture velocity estimates are also presented in order to describe the observed source duration pattern in terms of either the variable stress drop or rupture velocity models.
S42B-07 11:50h
The rupture process of the 2003 Tokachi-Oki earthquake using 1-Hz GPS data and its afterslip history inferred from GPS data.
We analyzed GPS data to infer the rupture process of and afterslip distribution following the 2003 Tokachi-Oki earthquake. Firstly, we processed 1-Hz GPS data with kinematic GPS analysis [Larson et al., 2003] for 90 minutes (approximately 45 minutes before and after the quake) to obtain the displacement waveform. Then we inverted those waveform to infer the rupture process of the main shock using multi time window inversion method. We use the Freq-Wavenumber (FK) codes developed by Zhu and Rivera [2003] to calculate the green's function. We find the slip is not significant in the epicenter. The rupture propagated downdip, and maximum slip as large as $\sim$ 9 m is found $\sim$ 50 km downdip. This result is consistent with those inferred from teleseismic and strong motion inversions. Secondly, we processed postseismic daily GPS data with precise point positioning, and inverted the time series with the Network Inversion Filter [Segall and Matthews, 1997]. It is notable that afterslip is absent in the area where the complementary spatial pattern of coseismic rupture and afterslip may indicate along strike spatial variations of frictional properties, although other interpretations can not be ruled out. We calculated shear stress change due to afterslip in the direction of plate convergence. The stress change follows a quasi-linear trend with log of slip-rate, suggestive of a decay back to steady-state slip on a velocity strengthening fault.
S42B-08 12:05h
Nonvolcanic Deep Tremors in the Transform Plate Bounding San Andreas Fault Zone
Recently, deep ($\sim$ 20 to 40 km) nonvolcanic tremor activity has been observed on convergent plate boundaries in Japan and in the Cascadia region of North America (Obara, 2002; Rodgers and Dragert, 2003; Szeliga et al., 2004). Because of the abundance of available fluids from subduction processes in these convergent zones, fluids are believed to play an important role in the generation of the tremor activity. The transient rates of tremor activity in these regions are also observed to correlate either with the occurrence of larger earthquakes (Obara, 2002) or with geodetically determined transient creep events that release large amounts of strain energy deep beneath the locked Cascadia megathrust (M.M. Miller et al., 2002; Rodgers and Dragert, 2003; Szeliga et al., 2004). These associations suggest that nonvolcanic tremor activity may participate in a fundamental mode of deep moment release and in the triggering of large subduction zone events (Rodgers and Dragert, 2003). We report the discovery of deep ($\sim$ 20 to 45 km) nonvolcanic tremor activity on the transform plate bounding San Andreas Fault (SAF) in central California where, in contrast to subduction zones, long-term deformation directions are horizontal and fluid availability from subduction zone processes is absent. The source region of the SAF tremors lies beneath the epicentral region of the great 1857 magnitude (M) $\sim$ 8, Fort Tejon earthquake whose rupture zone is currently locked (Sieh, 1978). Activity rate transients of the tremors occurring since early 2001 also correlate with seismicity rate variations above the tremor source region.