U33A-0013
Exploiting the Frequency Content of Non-Volcanic Tremor to Characterize its Prevalence
Non-volcanic tremor (NVT) and slow slip episodes have now been well documented and studied in several subduction zones, and they are associated with the transition from stick-slip to stable sliding on the plate interface. Previous studies have exemplified Northern Cascadia for its 2-4 week episodes of correlated tremor and slip occurring regularly every 14±1 months. However, other regions of Cascadia show more variable recurrence and duration, and other subduction zones (Japan, Mexico) have some cases where tremor and GPS-detected slip occur separately. In an attempt to improve classification and understanding of episodic tremor and slip (ETS), this study examines the prevalence of NVT in Cascadia in greater detail by exploiting the signature low frequency spectrum of NVT. Tremor signals are collected from a mixture of seismometers including those of the EarthScope Transportable Array and a host of permanent regional networks that span the subduction zone. Seismic data are bandpass filtered into three categories, 10-15 Hz, 2-5 Hz, and 0.2-0.5 Hz, where we expect prominent signals from microseismicity, NVT, and surface waves, respectively. We then apply a simple detection algorithm that compares amplitudes of each passband looking for elevated energy in the tremor passband relative to surface wave and microearthquake passbands. When applied to the Cascadia region, we find this algorithm can successfully detect previously identified ETS events. Moreover, our technique finds additional episodes of NVT that vary in duration, recurrence, and spatial extent, many of which do not appear to have a coincident GPS-detected slow slip event. By lowering the detection threshold of NVT, we hope to achieve a more complete characterization of NVT behavior that sheds new light on temporal and possibly causative relationships with slow slip and traditional earthquakes.
U33A-0014
Rock-Physics Interpretation of Fluid-Driven non Volcanic Seismic Tremor
Any rock can fail by hydraulic fracture if pore pressure exceeds the confining pressure, because elevated pore fluid pressures leads to weakening and embrittlement through the principle of effective stress. This process can be monitored in detail in laboratory, measuring the output of seismicity, as acoustic emission (AE). Here we illustrate results from AE measurements during dehydration experiments conducted while heating gypsum, diasporite, serpentinite etc. samples to beyond their equilibrium dehydration temperatures. Experiments were performed typically on 30 mm long cores with a diameter of approximately 15 mm, under a hydrostatic stress of 200 and 300 MPa in a Paterson high-pressure/high-temperature internally-heated gas apparatus. AEs were recorded via three piezoelectric transducers embedded in the sample end caps, which allowed to localize the hypocenters of the AEs. At 200 MPa, serpentinites dehydrated to talc + olivine and water around 500 Celsius, as confirmed by microstructural and geochimical post experiment analyses (EPMA, FE-SEM, XRPD). Microseismicity in the form of high-energy AE events was confined to a narrow temperature interval just above the equilibrium dehydration temperature. This overstep is due to the heating rate being faster than for equilibrium studies in our experiments. Very similar results were achieved during dehydration of gypsum and diasporite, where a 20 to 50 Celsius overstepping was necessary to initiate the dehydration reaction. Two main types of AEs were recorded (localized inside the sample): (1) high frequency, high amplitude and short duration AEs and, (2), low amplitude, lower frequency and very long duration AEs. The first type was interpreted as isolated microcracking due probably to differential thermal expansion. The second type coincided with the onset of the dehydration reaction, and was interpreted as a cascade of short event induced by the movement and escape of the fluid from the sample. These long lasting AEs were accompanied by sudden drop of temperature recorded at the sample end by the thermocouple. Our results conclusively show that seismicity can be generated by dehydration reactions even in the absence of a deviatoric stress. This has potentially important implications for earthquake nucleation in subducting lithospheric plates. Moreover, the cascade of events that followed the onset of dehydration may well be related to the low-amplitude long-duration (tremor) seismic events that characterize the seismic activity in subduction zones and has been tentatively interpreted as caused by dehydration of the subducting plate.
U33A-0015
What can we Learn From Dynamic Triggering of Low-Frequency Earthquakes?
Remote triggering of small low-frequency seismic events near the seismic-aseismic transition zone of subduction zones, by surface waves from large distant earthquakes, has been reported in southwest Japan and the Cascadia region. Recent observed triggering in southwest Japan from three large earthquakes (2003 Tokachi-oki (Mw 8.1), 2007 Solomon (Mw 8.1), and 2008 Wenchuan (Mw 7.9)) covering wide azimuthal information that is necessary to distinguish the triggering processes, shows significant triggering from Rayleigh waves rather than Love waves. This observation provides strong evidence for the influence of fluids in the source area because fluid can be affected by normal stress changes and not by shear stress changes. In the Cascadia region, it has been reported that seismic tremor associated with episodic slip is triggered by shear stress changes from Love waves. The low-frequency events both in southwest Japan and the Cascade region are though to be fluid related events. The Coulomb failure stress analyses suggest the effective friction coefficient is large for southwest Japan and small for the Cascadia region, which could be related to the amount of fluid in the source regions of the low-frequency events.
U33A-0016
Episodic Tremor and Slip: Insights From a Lab-scale Subduction Zone
We have devised a laboratory experiment to investigate the frictional and acoustic patterns of a salt slider with cumulated slip. We record a continuous change of the frictional behavior of the slider from stick-slip to the stable sliding regime, similar to the transition from earthquake to slow slip event along subduction zone. The Acoustic Emission associated with the movements of the slider evolves with cumulative displacement and interface ageing. It follows a trend from strong impulsive events, similar to earthquake seismic signals, to a collection of smaller amplitude and longer duration signals, similar to NVT. Our results favor the hypothesis of a common origin to SSE and NVT, which would be found in the slip with friction of a deformable interface.
U33A-0017
Investigating ETS using the GRACE time variable gravity data
Geodetic measurements above subduction zones have revealed a range of transient deformations. The mechanism by which slow slips occur remains uncertain. In particular, does surface deformation represents the integrated deformation from a large number of discrete seismic slip events? Moreover, little is known about displacements of materials having different densities as well as changes in the intrinsic density of the surrounding material, which is induced by its dilatation or contraction. The GRACE twin satellite mission has provided detection of glacial isostatic adjustments, variable snow and ice levels on the polar caps and elsewhere, surface water flux on continent, and seafloor pressure fluctuation. It has also detected co seismic and post seismic gravity changes after the great Sumatra earthquakes. We proposed a method based on the Empirical Orthogonal Function (EOF) decomposition to extract meaningful signal from the GRACE data. The main advantage of this method is that it does require less a priori information on the event, which makes the detection more objective. Indeed, the other methods used up to now consider that we know when and where the event occurred, and use that information to separate the earthquake signature. Applying the EOF method at major subduction zones, significant signals in the GRACE time series are found above the Cascadia, South-America (rupture zone of the 1960 Chile EQ) and Japan subduction zones. In particular, a significant change in gravity appeared north-east of the Vancouver Island end 2006-beginning 2007. This event as well as other ones may be related to ETS or other transient processes in the mantle, further away from the downdip end of the locked and transition zones.
U33A-0018
Monitoring the 2008 Northern Cascadia ETS Episode With Absolute Gravity Measurements
The monitoring of subduction zone Episodic Tremor and Slip (ETS) has been carried out primarily using seismic data for tremor and continuous GPS observations for transient slip. More recently, the establishment of the Plate Boundary Observatory has added borehole instrumentation consisting of Gladwin borehole strainmeters (BSM) as well as pore pressure gauges. For sites on southern Vancouver Island, the regularity of ETS episodes has allowed us to schedule extended periods of continuous absolute gravity (AG) measurements to augment these other data and thereby help in understanding the fundamental physical processes involved in the generation of ETS. For the 2008 ETS event in the northern portion of the Cascadia Subduction Zone, AG observations were carried out at the Pacific Geoscience Centre (PGC), near Victoria, BC, Canada. PGC is also the site of three BSM's and a continuous GPS station whose data have been used in the past to constrain models of slip on the subduction plate interface. This contribution focuses on the analysis of the approximately 15 week-long continuous series of AG observations at PGC. Preliminary results show that a subtle transient gravity signal is observed but its origin is not clear. Little or no uplift/subsidence is observed in the GPS data although its vertical resolution is 3 or 4 mm at best. The lack of a change in elevation is not surprising since PGC lies close to the hinge-line for vertical deformation. Strainmeter data from the 3 co-located BSM's show discrepancies that indicate interfering signals of likely non-tectonic origin.
U33A-0019
Searching for displacement associated with Cascadia slow slip events with InSAR.
We have detected deformation associated with nearly 40 episodic tremor and slip (ETS) events in Cascadia using GPS time series spanning 1997-2007. Despite these successes, several factors cause the characterization of the ETS events based on GPS to be incomplete. There is a low spatial density of GPS observation points, low GPS sensitivity to vertical deformation (which is critical to assessing the down-dip extent of slip on the subduction interface), and the time series observations start in 1995 or later. To overcome some of these problems, we examine InSAR as a possible supplementary data set to GPS. We have compiled a list of ETS events in the Puget bay region dating back to 1992 and in northern California dating to 1993. We used this to generate a complete catalog of SAR before and after scenes spanning each event, from the following satellites: ERS 1 and 2, Envisat Image mode 1 to 7, Radarsat fine and standard beam modes, and ALOS. We limited temporal baselines to ~3-4 months spanning any slow slip event for two reasons: 1) to maximize interferometric correlation, and 2) to reduce contamination of the ETS events by other deformation signals. For all modes except Radarsat finebeam and ALOS, we limited perpendicular baselines to <500m. Conventional two pass interferometry (using the ROI_PAC processing suite) had limited results. Common problems included widespread decorrelation and small (and long wavelength) tectonic signal. Nearly complete spatial correlation is demonstrated for some scenes in the Puget bay region and northern California, but these represent <5% of the initial catalog we tested. Synthetic tests using forward models of slip events inverted from GPS displacement fields show LOS deformation on the order of 1 to 1.5 radians for C-band systems, a level comparable to expected residuals due to atmospheric and orbital errors. We also conducted synthetic tests showing how the large wavelength deformation signal would affect the quadratic ramps that we remove from the interferograms in order to reduce the impact of orbital errors. Because of the difficulties associated with conventional two pass interferometry, we examine how additional techniques, such as persistent scatterer or time series analysis may improve our results.
U33A-0020
Spatial and temporal patterns of nonvolcanic tremor and segmentation along the southern Cascadia subduction zone
Episodic tremor and slip (ETS), the spatial and temporal correlation of slow slip events (SSE) monitored by GPS observations and non-volcanic tremor (NVT) monitored by seismic signals, is a new type of deformation occurring down-dip from the seismogenic zone along several subduction margins. We apply a semi- automated algorithm to seismic data from southern Cascadia to detect the most prominent pulses of NVT and invert arrivals for source locations. We investigate in detail ten NVT episodes from 2005-2007 with source locations over a 650 km along-strike region from Oregon to northern California. NVT epicenters occur within a limited band corresponding to the 30-45 km contours of the plate interface and migrate at 2-13 km/d via clusters of locations, with occasional large jumps (100-200 km). The cluster edges and terminations of NVT episodes appear to be repeatable features that support previously proposed segmentation of ETS and suggest an even finer detail reminiscent of asperities. NVT displays a wide depth distribution using a 1D velocity model, but peak distributions are within 10 km of the plate interface. We also detect over a 100 low- frequency earthquakes (LFEs) similar to observations from southwest Japan. Considering the occurrence of LFEs, patchiness of source locations, slow migration with jumps >100 km, the tight epicentral band of NVT, and associated regions of increased pore-fluid pressure and coupling along the plate interface, NVT in southern Cascadia is consistent with interplate shear slip.
U33A-0021
The Slip History and Source Statistics of Major Slow Slip Events along the Cascadia Subduction Zone from 1998 to 2008
We estimate the time dependent slip distribution of 16 prominent slow slip events along the northern half of the Cascadia subduction zone from 1998 to 2008. We process continuous GPS data from the PBO, PANGA and WCDA networks from the past decade using GAMIT/GLOBK processing package. Transient surface displacements are interpreted as slip on the plate interface using the Extended Network Inversion Filter. Of these 16 events, 10 events are centered north of Puget Sound, 4 events are resolved around the Columbia River and 1 event is located near Cape Blanco. The February 2003 event is complex, extending from Portland to southern Vancouver Island. Other smaller events beneath Northern Vancouver Island, Oregon and Northern California are not well resolved because of the limited station coverage. We identify two characteristic segments based on the along-strike extent of individual transient slip events in northern Washington. One segment is centered around Port Angeles. Another segment is between the Columbia River and the southern end of Puget Sound. The propagation direction of slow slip events is variable from one event to the next. The maximum cumulative slip for these 16 events is ~ 27 cm, which is centered beneath Port Angeles. This indicates that the strain release by transient slip is not uniform along-strike. In northwestern Washington where cumulative slip is a maximum, the subduction zone bends along-strike and dip of the plate is lower compared to the north and south. We hypothesize that the geometry of the slab plays an important role for focusing transient strain release at this location along the subduction zone. We explore the relationship of source parameters of slow slip using our catalogue of 16 events. The estimated moment magnitude ranges between 6.1 and 6.7. The average stress drop of 0.06-0.1 MPa is nearly two orders of magnitude smaller than that found for normal earthquakes (1-10 MPa). Standard earthquakes follow a scaling relationship where rupture length is proportional to slip amplitude resulting in a nearly constant stress drop. Slow slip events display the similar scaling law up to a possible limit in slip amplitude. We also explore the relationship of event duration to other source parameters.
U33A-0022
Cycles within cycles - Repetitive Tremor between Cascadia 14-month ETS episodes
We have automatically searched Northwestern Washington for non-volcanic tremor in all 5-minute time windows during the inter-ETS period from February, 2007 through April, 2008. Tremor was detected in nearly 3000 windows, which overlap by 50%, so tremor was seen 2% of the time. The catalog of 5-minute tremor locations cluster in time and space into groups we call tremor swarms. This sorting revealed 35 tremor swarms during the inter-ETS period. The number of hours of tremor per swarm ranged from about one to 50 hours, adding up to 193 hours. The inter-ETS tremor swarms generally locate along the downdip side of the major ETS events, and account for approximately 45% of the time tremor has been detected during the last entire ETS cycle, which includes the May, 2008 ETS episode. Many of the inter-ETS events are near carbon copies of each other in duration, spatial extent and propagation direction, as is seen for the larger 14-month- interval events. These 35 inter-ETS swarms plus one major ETS episode follow a power law relationship such that the number of swarms, N, exceeding duration τ is given by N = c τ-0.6. If we assume that seismic moment is proportional to τ as proposed by Ide et al. [Nature, 2007], we find that the tremor swarms follow a standard Gutenberg-Richter logarithmic frequency-magnitude relation, log10N = a 10-bMw, with b = -0.9, which lies in the range for normal earthquake catalogs. Furthermore, the major ETS events fall on the curve defined by the inter-ETS swarms, suggesting that the inter-ETS swarms are just smaller versions of the major 14-month ETS events.
U33A-0023
Comparing Observations of Low Frequency Earthquakes From Tremor at Three Subduction Zones
We present a comparison of low frequency earthquake (LFE) detections from three subduction zones: southwest Japan, south Vancouver Island, and northern Costa Rica. In southwest Japan it has been shown that near-repeating swarms of LFEs are the primary constituent of non-volcanic tremor and their mechanism reflects shear slip on the plate interface between 30-35 km depth. In contrast, previous studies of non- volcanic tremor in south Vancouver Island lack LFE detections preventing a similar analysis, although shear slip as the tremor mechanism has also been proposed as the tremor mechanism in other portions of the Cascadia margin. Unlike southwest Japan, the location of non-volcanic tremor in south Vancouver Island is widespread particularly with depth. More recently, non-volcanic tremor was observed in Costa Rica and it is proposed that this process may be similar to non-volcanic tremor in southwest Japan. Preliminary tremor locations in Costa Rica, obtained from cross-correlation of tremor envelopes have poor depth resolution; however, they appear to concentrate at shallower depths that overlap the active seismogenic zone. In order to acquire a better understanding of non-volcanic tremor in south Vancouver Island and Costa Rica, we employ an autocorrelation approach to detect and locate LFEs in all three locations. Preliminary results suggest non-volcanic tremor in all three subduction zones consists of a swarm of near repeating LFEs indicative of shear slip down-dip of the primary seismogeic zone, and that this phenomenon can occur in both hot and cool subduction zones.
U33A-0024
Low-Frequency Earthquakes in Cascadia Using Texan Array
Low-frequency earthquakes (LFEs) were first reported in Japan and have been observed to occur coincidently with non-volcanic tremor in both space and time. Compared to similarly-sized ordinary earthquakes, LFEs are deficient in frequencies above 5 Hz. The frequency spectrum of LFEs mirrors the spectrum of tremor. Indeed Shelly et al. (2006, 2007, Nature) have suggested that tremor is simply the superposition of many individual LFEs. Accordingly, LFEs have been used to constrain the location of tremor. In Japan, LFEs are routinely identified by their S-waves, while their P-waves are typically below noise levels. In March 2008 we deployed a 1km aperture array on the Olympic Peninsula of Washington State that recorded a small tremor swarm prior to the main episodic tremor and slip event in May. Our array consisted of approximately 80 geophones paired with Texan recorders and was laid out on a 100m spaced grid. Initial analysis of one hour of data reveals many LFE-like events with similar spectra to locally observed tremor. Unlike LFEs in Japan, P-waves are clearly seen on many individual stations. Using a clear LFE as a template event, nearly 100 matching events have been found with S minus P times that differ by less than a few hundredths of a second from event to event suggesting that they are all within a few hundred meters of each other. Preliminary locations of this cluster indicate that the LFEs are near the plate interface east of our array. For many of the LFEs, stacks of the P-wave are very similar to stacks of the S-wave, suggesting that these stacks are reasonably good approximations of source time functions. These source time functions vary in complexity from one LFE to another. Synthetic waveform modeling suggests a source focal mechanism consistent with thrust faulting on the plate interface. The location and focal mechanism of these LFEs support the notion that tremor is associated with slip on the plate interface. This LFE cluster provides a promising new opportunity for precisely characterizing the location, focal mechanism, and perhaps the moment-duration scaling of tremor in Cascadia.
U33A-0025
Recent Findings on the Nature of Episodic Tremor and Slip Along the Northern Cascadia Margin
Episodic Tremor and Slip (ETS), as observed along the northern Cascadia margin, has been defined empirically as repeated, transient ground motions at a plate margin, roughly opposite to longer-term interseismic deformation, occurring synchronously with low-frequency, emergent seismic signals. Although the exact causal processes are still a matter of debate, recent improvements in the monitoring of these transient events provide clearer constraints for the location and the migration of both tremor and slip. In areal distribution, the tremors continue to occur in a band overlying the 25 to 55 km depth contours of the nominal subducting plate interface. The previously reported extended depth distribution of tremor is also observed for the most recent tremor episodes, as is the coincidence of peak tremor activity with a band of seismic reflectors that is commonly interpreted to be positioned above the plate interface. In these episodes, tremors migrate along strike of the subduction zone from the southeast to the northwest at speeds ranging from 5 to 13 km/day. Tremor data also show changes in migration speed during the course of a single episode. No systematic migration in depth has yet been resolved. Denser GPS monitoring and the introduction of borehole strainmeters have also led to a better definition of the ETS surface deformations patterns, including those derived from the vertical GPS component. Inversion of the GPS data, constrained by limiting slip to the currently accepted plate interface, results in an area of slip that parallels the strike of the subduction zone, overlapping with but narrower than the band of tremor distribution and displaced slightly seaward. Inversion constrained by a shallower occurrence of slip, on or near the reflector band, results in a broader distribution of slip with reduced magnitudes. This would be more commensurate with the wider distribution of tremor. The current GPS deformation data are unable to tell whether the slip could be distributed over a thick shear zone. However, the time series of borehole strain indicate that the along-strike propagating slip patch likely has a sharp propagating front, supporting the notion that slip occurs along a thin slip zone or a single decollement. Consequently, a working model for ETS is that the dominant slip occurs repeatedly along a well-defined weak zone but the synchronous tremor occurs not only within this zone but also in a surrounding volume whose material properties have been altered by an abundance of fluids and the presence of high pore pressures.
U33A-0026
Seismic and geodetic constraints on Cascadia slow slip
We present a comprehensive view of tremor activity for the past 4 episodic tremor and slip (ETS) episodes in northern Cascadia. Automatically detected and located tremor epicenters provide a high-resolution map of Washington's slow slip region. Thousands of epicenters from each of the past four ETS events from 2004—2008 provide detailed map-view constraints that correlate with geodetic estimates of the simultaneous slow slip activity. Analysis of the latest 15-month inter-ETS period also reveals ageodetic tremor activity equivalent both in duration and extent to ETS tremor. Epicenters from both ETS and inter-ETS tremor are bounded between the 30—45 km plate interface depth contours and locate approximately 75 km east of the locked portion of the subducting Juan de Fuca plate. Based on the high spatio-temporal correlation tremor and slip, the tremor duration and slip magnitude relationship and the similarity in map view an duration of ETS and inter-ETS tremor, we suggest that the mapped tremor source area demarcates the freely slipping portion of the subducting slab. These findings also reveal a well-resolved, sharp updip edge of tremor epicenters that may both reflect a change in plate interface coupling properties as the plate transitions from periodic free slip to the locked zone and map a region accumulating stress with the potential for coseismic shear failure during a megathrust earthquake.
U33A-0027
Very-low-frequency earthquakes accompanied with episodic tremors and slips on the plate boundary in the transition zone from locked to aseismic interface
Anomalous very-low-frequency (VLF) signals in the frequency range 0.02--0.05Hz have been recorded by a dense broadband seismic network in southwestern Japan. One of signals is generated from shallow VLF earthquakes that occur in the accretionary wedge along the Nankai subduction zone. Another signal originates from the region of deep VLF earthquakes in the region of transition from a locked zone to an aseismic slip region. Here we provide a detailed seismicity of VLF earthquakes in the transition zone in the down-dip part of the Nankai subduction zone, including the calculation of their moment tensor solutions, for the period from April 2003 to May 2008. We also consider the processes responsible for the seismic moment release during VLF earthquakes in the transition zone on the plate boundary. Using continuously recorded seismograms for a five-year period obtained from a very dense broadband seismic network, we performed grid moment tensor inversion and centroid moment tensor inversion to calculate 242 moment tensor solutions with moment magnitudes ranging between 3.1 and 3.8. There were at least 5--10 sequences of repetitive activity during the five years. The VLF events formed clustered distributions along the 35-km isodepth contour of the subducting plate surface. The nodal planes, which dip landward, determined from the moment tensor solutions of the VLF events reflected the configuration of the subducting plate interface. The slip vectors were consistent with the direction of movement of the subducting plate; the dip and strike corresponding to the slip vectors clearly reflected the configuration of the upper surface of the subducting plate. These observations suggested that VLF seismic signals are radiated from shear slips on the upper surface of the subducting Philippine Sea Plate. We also found that the rates of seismic moment release per unit area associated with five major VLF clusters were very similar, although both the seismic magnitudes and sizes of the clusters varied considerably. The rate of seismic moment release from detectable VLF sources was 0.1 percents of the rates for short-term slow-slip events, suggesting that the source areas occupied only 0.1 percents of the fault segment on which the short-term slow slip occurred.
U33A-0028
Regionality of deep low-frequency earthquakes associated with subduction of the Philippine Sea plate along the Nankai Trough, southwest Japan
The Fukuoka District Meteorological Observatory recently logged three possible deep low-frequency earthquakes (LFEs) beneath eastern Kyushu, Japan, a region in which LFEs and low-frequency tremors have never before been identified. To assess these data, we analyzed band-passed filtered velocity seismograms and relocated LFEs and regular earthquakes using the double-difference method. The results strongly suggest that the three events were authentic LFEs, each at a depth of about 50 km. We also performed relocation analysis on LFEs recorded beneath the Kii Peninsula and found that these LFEs occurred near the northwest-dipping plate interface at depths of approximately 29–38 km. These results indicate that LFEs in southwest Japan occur near the upper surface of the subducting Philippine Sea (PHS) plate. To investigate the origin of regional differences in the occurrence frequency of LFEs in western Shikoku, the Kii Peninsula, and eastern Kyushu, we calculated temperature distributions associated with PHS plate subduction. Then, using the calculated thermal structures and a phase diagram of water dehydration for oceanic basalt, the water dehydration rate (wt%/km), which was newly defined in this study, was determined to be 0.19, 0.12, and 0.08 in western Shikoku, the Kii Peninsula, and eastern Kyushu, respectively; that is, the region beneath eastern Kyushu has the lowest water dehydration rate value. Considering that the Kyushu–Palau Ridge that is subducting beneath eastern Kyushu is composed of tonalite, which is low in hydrous minerals, this finding suggests that the regionality may be related to the amount of water dehydration associated with subduction of the PHS plate and/or differences in LFE depths. Notable dehydration reactions take place beneath western Shikoku and the Kii Peninsula, where the depth ranges for dehydration estimated by thermal modeling agree well with those for the relocated LFEs. The temperature range in which LFEs occur in these regions is estimated to be 400-500°C.
U33A-0029
Seismic exploration of deep low-frequency tremor area in western Shikoku, Japan
Low-frequency tremor, first found in southwestern Japan, is a seismic phenomenon that would be attributed to subduction system of the Philippine Sea plate. We carried out a seismic reflection survey in March 2008 to reveal the structure around the source region of tremor in western Shikoku, Japan. The survey had two reflection profiles, Line-SN and Line-WE, whose lengths were 75 km and 85 km, respectively. Six explosive sources with 500 kg were fired in this survey. Such large sources enable us to detect reflected waves from the Philippine sea plate at depths of ~40 km. 341 seismometers with the natural frequency of 2 Hz (L-22D) were deployed with the average separation of 333 m for Line-SN and 500 m for Line-WE, respectively. We retrieved high S/N data for all shots. The data was analyzed with typical seismic processing; (1) CDP editing, (2) filtering, (3) muting, (4) NMO correction, and (5) depth conversion. We assumed a homogeneous medium with velocity of 6.4 km/s for (4) and (5). From the obtained seismic profiles, we identified abundant reflectors. In the Line-SN, we can trace a strong and continuous reflector, which has a north dip of 8 degrees at depths of 22-35 km. We interpret the reflector as the upper boundary of the Philippine Sea plate. Another reflector, whose amplitude is smaller than that of the plate boundary, is identified at depths of 30-42 km. The reflector is interpreted as the Moho discontinuity in the Philippine Sea plate because it is below and parallel to the plate boundary and is consistent with the oceanic Moho-discontinuity estimated by receiver function analysis. These interpretations show that the thickness of the oceanic crust is about 8 km and the hypocenters of tremor concentrate near the plate boundary. Along the dip of the Philippine Sea plate, the reflection from the plate boundary around the tremor is not clear, whereas we observe a strong reflector from down-dip extent of the slab. This difference in the reflection intensity may be a key to understand the physics of the tremor.
U33A-0030
Detailed hypocentral distribution and activity of small repeating earthquakes at the Kanto region, central Japan
At the Kanto region, central Japan, the Philippine Sea plate (PHS) and the Pacific plate (PAC) are subducting, where numerous small repeating earthquakes were found based on analysis of waveform similarity of seismograms recorded by Kanto-Tokai network (1979-2003) (Kimura et al., 2003; 2006) and a portion of Hi-net data (2003-) of National Research Institute for Earth Science and Disaster Prevention (NIED) (Kimura et al., 2007). Highly similarity of seismograms shows that repeating earthquakes at Kanto can be regarded as repetition of rupture at the same fault and can be regarded as an indicator of relative plate motion, the same as at other regions (Nadeau et al., 1995; Igarashi et al., 2003; Matsubara et al., 2005). To understand detailed configuration and interaction of plates, we studied detailed hypocentral distribution and activity of repeating earthquakes. We determined high-precision relative hypocenter distribution by Double-Difference method (Waldhauser and Ellsworth, 2000) with waveform correlation for seismic clusters of the PHS and PAC and obtained detailed hypocentral distribution near the plate boundary. Most M3-class repeating earthquakes are located within 40 m and 140 m, at largest, for the PHS and the PAC, respectively. Larger distances for the PAC would be caused by location of the target region at the edge of the seismographic network. Considering fault size of used events, they are likely to occur at the same location. Hereafter, we will explain about results from two distinct seismic clusters. At the southwestern Ibaraki, north of Kanto, significant linearly distributing seismic clusters are observed and repeating earthquakes also occur. Repeating earthquakes are distributed along a plane and its dip angle is almost the same with a dip angle of low-angle nodal plane of focal mechanism of repeating earthquakes. Dip angles of seismic plane are 36° and 26° for northern and southern part of the cluster, respectively. Widths of the seismic cluster are 3 km and 6 km for northern and southern part, respectively. The changing locations of dip angle and width are almost the same. At the other cluster, significant temporal change of activity is observed. Off the Boso Peninsula, southeastern Kanto, activations of repeating earthquakes synchronizing with the Boso slow slip events (SSEs), which repeat with time interval of 5-7 years, were observed. Recurrence time averaged for each sequence of M3- class repeating earthquakes off the Boso Peninsula is 1.5 - 19.4 years with total average of 9.2 years. These events have close relation with SSEs. At offshore region, no repeating earthquakes occurred after the previous SSE on 2002. However, numerous repeating earthquakes occurred synchronizing with the 2007 SSE. Since repeating earthquakes at Kanto can be regarded as an indicator of plate motion, this result indicates that plate interface is locked between SSEs and interplate slip is accelerated synchronizing with the Boso SSE, causing repeating earthquakes.
U33A-0031
Wavefield and source spectra of non-volcanic low-frequency tremors in a southwest Japan subduction zone
We investigate wave-field properties and source spectra of non-volcanic low-frequency tremors in a southwest Japan subduction zone, which have not been fully understood due to a low signal-to-noise ratio of tremor signals. Geological Survey of Japan, AIST, has recently started an integrated borehole observation of water levels, strains, tilts, water temperatures and seismic waves in southwest Japan for monitoring of the anticipated Tonankai and Nankai earthquakes. At each observation site, we have drilled three boreholes with different depths (about 30 m, 200 m and 600 m) and installed high-sensitivity seismometers at a bottom of every borehole. In this study, we use borehole seismic data recorded by this vertical seismic array network as well as NIED Hi-net. In order to investigate wave-field properties of non-volcanic low-frequency tremors, the vertical array data are used. During tremor activities, high semblance values continuously appear at a specific apparent velocity. Because the S-wave velocity estimated by borehole logging is consistent with the apparent velocity, we conclude that the non-volcanic low-frequency tremors are composed of near vertically incident S-waves. This is supported by polarization analysis, where the particle motion is linearly polarized and the incident angle is steep. The polarization analysis also reveals that the polarization angles are almost same during each tremor activity. Because the polarization angle will have various values if the tremor signals contain large amount of scattered waves, it is likely that they are composed mostly of seismic waves coming directly from the source. A comparison of source spectrum between regular earthquakes and non-volcanic low-frequency tremors is a key to understand the difference in their physical processes. We first estimate model source spectrum for co- located regular earthquakes by Multi-Window Spectral Ratio method [Imanishi and Ellsworth, 2006]. By averaging the ratios between observed and model spectrum over all events in a cluster, we obtain the average attenuation function, which is assumed to represent the path and site effects between the source and station. The source spectrum of each tremor is then calculated by the ratio between the observed spectrum and the attenuation function. The estimated velocity spectra of tremors are flat for higher frequency, which clearly differ from those for regular earthquakes that follow f-1 decay at high frequency. It is noted that this feature of the velocity spectra can be explained by the Brownian walk model for slow earthquakes that was recently proposed by Ide [2008].
U33A-0032
Quiescent Period Variation of Deep Tremor Prior to Slow-slip Events in Southwest Japan
Temporal variation of the deep low-frequency tremor (LFT) activity along the strike of the subducting Philippine Sea plate in southwest Japan was investigated with a focus on the Quiescent periods between LFT activations (QPLFT). The QPLFT fluctuated according to the activity level of the LFT. While the LFT was active, the QPLFT became short. The QPLFT was considered to be a useful indicator of LFT activity. The investigation was based on the low-frequency earthquake catalog by the JMA and continuous seismic data at a single station. Gradual reductions of the QPLFT were observed in most of the areas of the beltlike distribution of the LFT in southwest Japan. The gradual reductions were related to SSE and some preceded the observed crustal deformation of the SSE. The QPLFT preceded the short-term (up to ten days) SSE by one to three months. A preceding gradual reduction of the QPLFT was also observed for the long-term SSE in the Bungo Channel. This reduction began eight months before the crustal deformation observed with the GPS network. Gradual recovery of the QPLFT was also observed in the slowdown stage of the aseismic slip. Temporal variation of LFT activity since 1997 related to the Tokai long-term SSE was investigated using continuous seismic data at a single station. Reduction of the QPLFT was observed while the SSE was ongoing, and the QPLFT fluctuated during the long-term SSE. While the the crustal deformation was rabid in 2003 and 2004, the QPLFT became short. It was also observed that the change of the QPLFT began one year before the observed crustal deformation of the SSE. A similarity was observed between the gradual reduction of the QPLFT and the slip speed related to nucleation of an earthquake simulated based on the rate- and state-dependent fault property of a single spring-slider system. It was assumed that the gradual reduction of the QPLFT was related to the SSE which could not be detected with the current observation system.
U33A-0033
A causal relationship between the slow slip event and deep low frequency tremor indicated by strain data recorded at Shingu borehole station
In the southwest Japan, synchronized deep low frequency tremors and short-term slow slip events occur repeatedly in several regions such as Tokai, northern Kii Peninsula and western Shikoku areas, and these activities are partitioned by 'gaps' of tremors. Although concurrent occurrences of slow slips and tremors have been detected at various subducting plate boundary, their physical mechanism is still not well understood. We are monitoring crustal deformation at Shingu city on the southeastern coast of Kii Peninsula, with an integrated multi-component borehole monitoring system developed by Ishii et al. [2002]. The borehole sensor unit consists of 6 strain sensors (4 in horizontal, 2 in vertical), 2 pendulum tilt sensors, magnetic direction finder and a quartz thermometer and installed at 510m depth. Fukuda et al. [2007] reported two types of strain changes, one associated with deep low frequency tremors and the other without tremors from November 2005 to March 2006. We extend the analysis period to 41 months, from January 2004 to September 2007. We identified 11 episodic strain changes. One of them was caused by heavy rainfall but the rest of the changes are considered to be slow slips. Among all the slow slips identified, five events occurred associated with reported tremor events, but the rest 5 changes were not accompanied by tremors. These slow slip events are characterized by N-S compression (0.017 - 0.063 ppm), and E-W extension (0.013 - 0.071 ppm), NW-SE extension (0.008 - 0.097 ppm), and last 4 - 9 days. We estimate a fault model for each event by forward modeling, and find that the all the strain changes can be attributed to reverse faulting on the plate boundary beneath the Kii Peninsula. An interesting strain change occurred from 26 Dec. 2004 to 2 Jan. 2005. In this period, a tremor activity propagated southwestward on central Kii Peninsula and the level of activity remarkably drops when the activity propagated into the tremor gap zone. After that, the tremors were activated again at the southwestern side of the tremor gap. During this activity, the main strain change at the Shingu borehole was recorded between two separated tremors, implying that the tremor gap is the source of the slow slip. Observations of slow slips preceding or without tremors indicate that a slow slip and tremor are independent phenomena. These results give support to a hypothesis that the slow slip smoothly propagates on the plate boundary and is the main process of simultaneous occurrences and tremor is a triggered phenomenon under special conditions.
U33A-0034
Crustal movement and groundwater changes associated with the episodic deep low frequency tremors and slow slip events in the southern part of Kii Peninsula, Japan
Geological Survey of Japan (GSJ), National Institute of Advanced Industrial Science and Technology (AIST)
has been constructing a new borehole observation network composed of 12 stations for forecasting the
Tonankai and Nankai earthquakes since 2006. Each station has 3 wells where water temperature, water level
(pressure) and ground motion are observed. At 1 of 3 wells, crustal strain is also observed by the multi-
component borehole strain meter. In 2007, we started the observation at 2 stations, named HGM and ICU, in
the southern part of the Kii Peninsula. In this area, the deep low frequency activities lasting a few weeks
occur periodically with an average recurrence interval of three of four months. According to Automatic
Tremor Monitoring System (ATMOS) of Hiroshima University, 4 major tremor activities occurred in the
southern part of the Kii Peninsula since the observation started at these 2 stations. We detected strain
changes related to some of the 4 tremors. We also detected possible groundwater changes associated with
one of them. Especially during the tremor activity in June 2008, clear strain changes were observed at ICU.
These changes can be explained by slow lip events occurring at several segments of the plate boundary
whose locations are consistent with the tremor occurrence area. Before stated our new observation, crustal
movement associated with the deep low frequency tremors were not detected clearly in the southern part of
the Kii Peninsula. Therefore, this result is meaningful.
GSJ, AIST will also start observation of crustal movement and groundwater changes at the other 10 stations
in Sikoku (6), Kii Peninsula (3) and Tokai (1) districts by the end of 2008. Those highly precise monitoring
system will have much contribution to clarify the mechanism of the slip in the plate boundary and be useful to
forecast the Tonankai and Nankai earthquakes in future.
http://staff.aist.go.jp/itaba-s
U33A-0035
Phenomenology of deep slow earthquake family in southwest Japan —Spatiotemporal characteristics and segmentation—
The phenomenology of slow earthquakes in the transition zone at the deeper side of the mega-thrust zone on the subducting plate interface in southwest Japan is investigated on the basis of the data obtained from the dense seismic and geodetic network, NIED Hi-net. Slow earthquakes include deep low-frequency tremors [Obara, 2002], deep very-low-frequency earthquakes [Ito et al., 2007], and short-term slow slip events [Obara et al., 2004]. Low-frequency tremors and very-low-frequency earthquakes are characterized by long- period seismic waves having a dominant frequency of 2 Hz and 0.05 Hz, respectively. A short-term slow slip event is recognized by ground tilting lasting for several days. Based on the space-time activity of tremors, the belt-like distribution of tremors is divided into some segments bounded by aseismic gaps. The repetition property of slow earthquakes depends on the segment size. In large segments longer than 100 km, the coupling phenomena of slow earthquakes occur with a recurrence interval of approximately 6 months. In smaller segments, tremor episodes that occur with a recurrence interval of a few months occasionally accompany the slight ground tilting caused by a small-sized slow slip event. Assuming all tremor episodes to be caused by a slow slip event, the apparent seismic moment of a slow slip event is estimated from the number of detected tremors and the area of each tremor episode. The apparent slip history estimated from all tremor episodes is almost constant in each segment. This indicates that the tremor activity is a good proxy for slow slip events. Some segments indicate an along-strike variation of tremor seismicity and VLF earthquakes. Along-strike inhomogeneity on the plate interface might affect the tremor migration. The initiation and direction of tremor migration have some variations within a segment. In many cases, the tremor migration that corresponds to the rupture front of a slow slip event starts or stops at the edge of a segment bounded by the gap. The tremor occasionally propagates beyond the gap to a neighboring segment. Considering the successive activity in both segments of the gap, an aseismic slip might propagate through the gap area. Various migration patterns of tremors might be applied to resolving the rupture process of mega-thrust earthquakes at the plate interface.
U33A-0036
Remote Triggering of Non-volcanic Tremor Around Taiwan
Non-volcanic tremor triggered by teleseismic surface waves was recently discovered in the Island of Taiwan by Peng and Chao (GJI, 2008). They reported clear bursts of tremor on the weak basal detachment fault beneath the Central Range triggered by the surface waves of the 2001 Mw7.8 Kunlun earthquake in Tibet. Here we perform a systematic survey of tremor triggered around Taiwan for the 45 teleseismic earthquakes with Mw ≤ 7.5 and depth < 100 km since 1998. The data analyzed in this study was recorded by the Broadband Array in Taiwan for Seismology (BATS) and the short-period seismometers by Central Weather Bureau (CWB). We identify triggered tremor as bursts of high-frequency, non-impulsive seismic energy that are coherent among many stations and during the passage of teleseismic body and surface waves. So far we have found 8 teleseismic events associated with clear triggered tremor around Taiwan, including the disastrous Mw7.9 Wenchuan earthquake in 2008. The epicentral distances between these 8 events and BATS station TPUB are from 1800 to 4000 km. The measured vertical peak ground velocity (PGV) ranges from 0.1 to 2.2 cm/s, corresponding to a dynamic stress of 0.009 and 0.19 MPa (with a nominal shear rigidity of 30 GPa and a surface wave velocity of 3.5 km/s). Our next step is to locate the tremor triggered by these teleseismic events, and quantify the relationship between surface waves and tremor occurrence. Our systematic survey of triggered tremor in Taiwan would help to quantify the triggering mechanisms and necessary conditions of tremor generation in an arc-continental collision environment, and improve our understanding of the fundamental processes of deep fault zones in active fold-thrust belts.
U33A-0037
Slow Slip and Co-Shock Microseismicity Beneath the Mahia Peninsula, Hikurangi Subduction Zone, New Zealand
Geodetically-detected episodes of slow slip in several subduction zones are accompanied by bursts of low-
frequency coherent noise known as seismic tremor. It remains unresolved, however, whether a single
physical process governs this association or even whether slow slip is invariably accompanied by tremor.
Detailed analysis of broadband seismic data spanning a slow slip episode near Gisborne (northern Hikurangi
subduction zone, New Zealand) in 2004 reveals that slow slip was accompanied by distinct reverse-faulting
microearthquakes, rather than tremor. The timing, location, and faulting style of these earthquakes are
consistent with stress triggering down-dip of the slow slip patch, either on the subduction interface or just
below it. These results indicate that tremor is not ubiquitous during subduction zone slow slip, and that slow
slip in subduction zone environments is capable of triggering high-frequency earthquakes near the base of
the locked subduction thrust. Near Gisborne and in other locations (Hawaii, Boso Peninsula) where slow slip
is accompanied by triggered microseismicity --- "co-shocks" --- the estimated upper extent of the slow slip is
shallower (less than ~20 km) than in those locations from which tremor has been reported. This
suggests that ambient temperature- or pressure-related conditions govern the character of the seismic
response to slow slip on subduction thrusts and other large faults, with low temperatures or pressures
triggering high-frequency microearthquakes and higher temperatures or pressures triggering seismic tremor.
http://www.geonet.org.nz
U33A-0038
Searching for tremor in seismic noise on the 84 OBS (Ocean Bottom Seismometers) and 40 Land Seismometers, 3 months deployment in the Lesser Antilles subduction
THALES, L.A.S.T., stands for Lesser Antilles Subduction zone Team of the THALES WAS RIGHT project (Coord. A. Hirn) of the European Union FP6, which gathers the scientific teams of a cluster of surveys and cruises in 2007. This cluster comprises the German cruise TRAIL with the vessel F/S Merian (PI E. Flueh and H. Kopp, IFM-GEOMAR), the French cruise SISMANTILLES 2 with the IFREMER vessel N/O Atalante (PI M. Laigle, IPG Paris and JF. Lebrun, Univ. Antilles Guyane), and French cruise OBSANTILLES with the IRD vessel N/O Antea (PI P. Charvis, Geoazur, Nice, France). Presentation T53A-1109 at last year, 2007 AGU Fall Meeting, of THALES, L.A.S.T. summarized the goals and first results of these experiments dedicated specifically to image at depth the seismic structure and activity of this subduction zone segment, which comprised: - MCS, multi-channel reflection seismic profiles as well as coincident multi-beam bathymetry that have been collected for 3700 km along a grid comprising 300 km along strike from North of Guadeloupe to Martinique islands and extending 150 km offshore over the forearc and accretionary wedge. - 84 OBSs at the nodes of this grid of profiles and 40 land stations. These instruments recorded the marine shots for a coincident refraction survey. They recorded local seismicity for precise location and focal mechanisms. The recovery of the continuous recording at this dense and extensive set of temporary sensors (1), recently completed allows to initiate an analysis of the continuously recorded seismic noise. Changes and transients of the noise character, and their possible correlations among instruments in the array will be searched for in the view of checking evidence of possible seismic tremor episodes or seismic transients, as have been described elsewhere with the specific aspect that most observations were acquired at sea-bottom as the forearc extends here broadly offshore. Preliminary results will be documented. (1) During these cruises and surveys, 84 Ocean Bottom multi-components Seismometers (OBS) have been brought together from several pools (Geosciences Azur, INSU-IPGP, IFM-GEOMAR, AWI), with up to 40 land stations (CSIC Barcelona, IPG-Paris, INSU-RLBM and LITHOSCOPE), for 3 months in early 2007, with a lesser number of instruments for similar period before and after. Support for the surveys came principally by ANR Catastrophes Telluriques et Tsunamis (SUBSISMANTI) to IPGP, by the EU SALVADOR Programme of IFM-GEOMAR, the OBSISMER CPER project of IPGP, Région Martinique and EU-FEDER, as well as by the EU project THALES WAS RIGHT on the Antilles and Hellenic active subductions to which contribute IPG- Paris and Geosciences Azur (France), IFM-GEOMAR (Germany), ETH Zurich (Switzerland), CSIC Barcelona (Spain), Univ. Trieste (Italy) and NOA Athens (Greece).
U33A-0039
Comparison of slip distribution of large slow slip events in Guerrero subduction zone
Aseismic slow slip events (SSEs) have been reported in most of the well geodetically instrumented subduction zones worldwide (Japan, Cascadia, Mexico, New Zealand, Costa Rica, Alaska). For most of the observed SSEs, the slip distribution on the subduction interface was inferred from the surface GPS displacements to be located at the downdip extension of the seismogenic zone, in the conditionally stable frictional regime. Hence, the stress transfer due to the SSEs on the seismogenic zone should advance the occurrence time of the next strong subduction earthquake. Since the installation of the GPS network, the Guerrero gap area, Mexico, has been affected by large SSEs in 1998, 2002, 2006. We have analyzed the Mexican permanent GPS data describing the 2002 and 2006 SSEs, so that to determine and compare their slip distribution. The resolution analysis made on those data shows that the slip distribution is dependent on the number and location of the GPS stations. The slip distribution inferred for the 2006 SSE, for which the GPS network was denser, is thus much better resolved than that for the 2002 event. The inversion result is not unique. However, for each event, all possible solutions draw a similar, overall slip distribution pattern. Both events are similar in terms of equivalent moment magnitude (7.4 to 7.6) and of maximum slip amplitude on the interface (up to 20 cm). By contrast, the size and location of the slip zone differ from one event to the next: In 2002, the zone sustaining slow slip spreads through the entire Guerrero subduction segment plus the western part of the Oaxaca subduction segment (~400 x 150 km); the slip is quite evenly distributed over the entire zone, though maximum slip concentrates both at the down dip limit of the seismogenic zone and in the stably slipping region. In 2006, slip occurs only in the Guerrero gap thus over an area twice smaller (~230 x 150 km) than before; in the west, the maximum slip is located both in the seismogenic zone and at its downdip extension, whereas in the east, the maximum slip localizes in the transition and stably slipping zone. Our results thus show that slip distribution and location of maximum slip markedly differ from one large SSE to the next. This suggests that the 2002 and 2006 events are parts of a complex multi-events sequence. Furthermore, in both events, slow slip occurred both in the seismogenic zone and below it. This demonstrates that SSEs play a role in the seismic cycle that is more complex than previously thought. This makes them a crucial element to be taken into account in seismic hazard assessment.
U33A-0040
Subducting slab ultra-slow velocity layer coincident with silent earthquakes in southern Mexico
Silent earthquakes, or episodic slow slip events and non-volcanic tremor have been observed in a few shallow subduction zones such as Cascadia1, Southwest Japan and southern Mexico. In general, the majority of the slow slip and tremor activities take place at the transition zone down-dip of the strong coupling section where great thrust earthquakes occur. High reflectivity and high Poisson's ratio, at the top of the subducting plate have been used to suggest the presence of a high pore-fluid pressure layer, which reduces the effective normal stress and promotes episodic slow slip. This inference is supported by dynamic triggering of non-volcanic tremors from passing surface waves and tides, and by numerical modelling result reproducing the duration and interval of the episodic slow slip events. However, no low seismic velocity layer associated with such a high pore-fluid pressure layer has been reported and it is not clear how it spatially and temporally relates to the episodic slow slip events and non-volcanic tremor. Here we show that locally observed converted SP arrivals and teleseismic underside reflections arrivals sampling the top of the subducting plate in southern Mexico reveal a spatially varying ultra-slow velocity layer (USL) (3-5 km, S wave velocity ~ 2.0-2.7 km/s), most likely a high pore-fluid pressure layer. The majority of reported slow slip patches coincide with the presence of the USL while tremor activities primarily concentrate further down-dip where the USL ends sharply. The persistence of the USL before, during and after the slow slip events suggests its longevity, whereas the spatial extent of the USL delineates the zone of transitional frictional behaviour.
U33A-0041
Improved GPS Analysis Including Modelling of non-Tectonic Signals Confirms Simple 4-4.5 Year Period for SSEs in Guerrero, Mexico, and Allows Prediction of the Next Event in 2010
The precise knowledge of the occurrence time of aseismic slow slip events (SSEs), their spatio-temporal evolution and the amount of deformation they produce are prerequisites to any deeper understanding of the loading/unloading process on faults and thus on earthquake cycles. We re-process and re-analyze observations from permanent GPS stations located in the Guerrero subduction zone, Mexico, where the world's largest SSEs have been observed in 2002 and 2006. Our objective is to refine the characteristics of these two major SSEs as well as smaller quasi-annual SSEs that have previously been reported in this zone and that have been interpreted as climate-driven SSEs. First, we choose to process the data using a double- difference processing strategy that shows a noise reduction of ~40% with respect to the published PPP analysis. Second, using the VMF1 tropospheric mapping function, applying an up-to-date ocean loading model, and applying at the observation level atmospheric pressure loading corrections, the variability of the position time series reduces by ~30%. Third, we remove a posteriori elastic deformation related to surface loading of hydrologic origin to reduce the annual perturbations that remained in the vertical position time series. We generate the elastic deformations by characterizing changes in surface loads from gravity anomalies estimated from GRACE data, assuming that the anomalies were caused entirely by changes in surface hydrologic loads. Our new GPS analysis has three main results. (1) Most of the time series corrected by the modelled hydrologic loading are different from the raw time series at 80% confidence level. (2) There is no longer clear evidence of the small, quasi-annual, periodic SSEs suggesting that the previously identified signals may actually be artifacts of the GPS analysis strategy. However, non-periodic anomalous displacements with amplitude <~3mm are present in some time series. (3) The large SSEs do not recover completely the accumulated strain suggesting that a part still has to be released. About the characteristics of the large SSEs, we show that the start and end of the SSEs are roughly simultaneous on the 3 components for most of the stations located inland whereas for the stations located on the coast (closer to the trench and above the seismogenic part of the subduction interface) the temporal slip evolution is complex and shows three stages with different rates. Using a 1-D elastic dislocation model and the total 3D surface displacements, we invert for slip on the subduction interface and find that a large amount of slip (up to 20cm) is needed both in the seismogenic and transition zones for the 2002 and 2006 SSEs. For the 2006 SSE, the amount of slip in the transition zone is smaller than in the seismogenic zone. The models are simple and the solutions are not unique but they strongly suggest that the seismogenic zone is involved in the SSE generation and that both SSEs do not have the same slip distribution on the interface; thus are not similar. The improved constraints on the Mexican GPS position time series permit us to predict that the next SSE should occur in 2010.
U33A-0042
Non-Volcanic Tremor along the Oaxaca segment of the Middle America Subduction Zone
Convergent plate boundaries generate great earthquakes when tectonic stresses accumulate on the plate interface. Down-dip from the seismogenic zone where increasing temperatures, pressure and dehydration affect frictional behavior, episodic tremor and slip (ETS) has been shown to occur in the transitional zone. The Oaxaca subduction zone is an ideal area for detailed ETS studies as rapid convergent rates, shallow subduction, and short trench-to-coast distances bring the seismogenic and transition zones of the plate interface up to 250 km inland. Previously analyzed slow slip events occur over large areas in southern Mexico, and may even extend up-dip into the seismogenic zone, potentially playing a role in future megathrust earthquakes. A new seismic deployment consisting of 7 broadband seismometers geographically dispersed inland along the Oaxaca segment provide the means to examine non-volcanic tremor (NVT) signals in detail for the first time. In this study, we investigate the spatial and temporal distribution of NVT in this region, including the duration, initiation and termination points, migration rates and directionality, and recurrence. We analyze NVT signals with a semi-automated process for identifying prominent bursts, and analyst-refined relative arrival times are inverted for source locations. We find week-long NVT episodes with location clusters that migrate at typical speeds, but NVT episodes occur much more frequently than month- long GPS-detectable slow slip episodes. We also observe a clear spatial relationship indicating a downward progression of deformation with the subducting plate from megathrust earthquakes to microseismicity to slow slip events to NVT to intraplate seismicity.
U33A-0043
The sources of the shallow, upper plate earthquakes in Central Mexico, and their possible triggering by the subduction earthquakes, 'normal' and slow
Large earthquakes have broken the Central Mexico upper plate in the historical time (1887, M5.3; 1912, M6.9; 1920, M6.5; 1976, M5.3; 1979, M5.3), but the faults responsible for those earthquakes are not precisely known. Nor is their link with the subduction, especially in terms of possible triggering between subduction and shallow earthquakes. To address those issues, we first identify the major active faults that cut the upper plate, and determine their slip mode and overall organization. For that, we conduct a morphological analysis of the region, based on the use of satellite and topographic images. The upper plate appears dissected by a dense network of hundreds of major active faults, which overall form 2 distinct large-scale systems, named Jalisco and Mexico. The Jalisco system is made of the N-S Colima normal fault system which runs from the coast to the Transmexican Belt (TMB) where it ends in a large fishtail centered on the Chapala Lake. The western branch of the fishtail is made of NW-striking, N-dipping faults that are both normal and left-lateral, while the eastern branch is made of ENE-trending, N-dipping normal faults. The Mexico system resembles a large-scale horsetail. It is made of a major NNW-trending left-lateral strike-slip fault that runs from the coast to the TMB, at the eastern edge of the Oaxaca region. As it enters the TMB, that NNW system connects to a series of E-W, N-dipping, normal-right-lateral faults, while extending further north through a series of NNW, E-dipping normal-left-lateral faults. Together these faults bound to the south and west a series of rhomboidal half-grabens, among those are the basins of Mexico City and Acambay. Mexico City is thus bounded by large active faults but also dissected by smaller ones, both E-W and NNW. The E-W fault system that bounds the Acambay half-graben to the south is likely the one to have ruptured in 1912. Second, we examine whether the subduction interface and the upper plate active faults interact due to static stress transfer. Hypothesizing that the Guerrero gap entirely breaks in a M8.1 earthquake on the subduction interface, we calculate that the static Coulomb stress change that such an earthquake would produce at 12 km depth on the E-W, N60°-dipping Mexico City and Acambay normal faults, would be 0.5-1 bars, hence large enough to trigger the rupturing of those faults. The 1912 Acambay earthquake may thus have been triggered by the 1911 large earthquake that apparently broke the Guerrero gap. Using the slip distribution provided by Cotton et al. (this issue) for the 2006 slow earthquake on the Guerrero subduction interface, we calculate that the static stress change that such a slow event would produce on the same Mexico City and Acambay faults, would be 0.05-0.1 bars. Because slow earthquakes repeat frequently and possibly every 4 yrs in Guerrero (Cotte et al., submitted), only 10 events (over 40 yrs?) similar to the 2006 slow earthquake would be enough to produce a stress change similar to the M8 subduction events, hence possibly play a part in the trigger, and advance the time of occurrence of the crustal Mexico and Acambay faults. Seismic hazard in Central Mexico thus depends on multiple sources, including the slow subduction earthquakes.
U33A-0044
Updated Episodic Tremor and Slip on the Cocos-Caribbean Subduction zone as measured by a GPS and Seismic Network on the Nicoya Peninsula, Costa Rica
The close proximity of the Nicoya Peninsula to the Cocos-Caribbean Subduction zone plate boundary makes it a prime location to use GPS to study episodic tremor and slip. Nicoya Peninsula currently has an operating network of 13 continuous GPS (CGPS) and 12 seismic stations designed to identify and characterize the pattern of episodic tremor and slip (ETS) events along the seismogenic zone under Costa Rica's Pacific Margin. The CGPS stations have varying degrees of equipment and communication. Of the 13 sites, five are equipped with SIM cards and modems for direct download capabilities, two of which are connected to a router for direct internet access. These sites, LMNL, located in Limonal, and LEPA, located in Lepanta are being monitored on the PBO network by UNAVCO with all of the quality and systems checks that this implies. All sites with NetRS receivers are currently partitioned to record both 5Hz and 30 second position data. The advantage to recording at a 5 Hz interval is that the receivers may be used to record long period dynamic events during large earthquakes along the Cocos-Caribbean plate boundary. The occurrence of slow slip events has been previously postulated in this region based on correlated fluid flow and seismic tremor events recorded near the margin wedge in 2000 and from sparse GPS observations in 2003. Paucity of data prevented details of these events from being resolved. In May 2007 a slow slip event was recorded on our densified GPS network and accompanied by seismic tremor. We will present the refined GPS time series and correlated seismic tremor for both likely slow slip events in September 2003 and May 2007. We will also present the inferred pattern of slip on the plate interface. Future plans include installation of additional sites, including the interior of the peninsula and reference sites on the stable Caribbean plate.
U33A-0045
A Search in Strainmeter Data for Slow Slip Associated with Triggered and Ambient Tremor near Parkfield, California
We test the hypothesis that, as in subduction zones, slow slip accompanies triggered and ambient tremor in the transform boundary setting of California. Our study builds on the results of the study of Peng et al. [2008] of triggered and ambient tremor near Parkfield, California during time intervals surrounding 31 posited triggering, M7.5 and greater, teleseismic earthquakes. Waves from 10 of these earthquakes triggered tremor and 29 occurred during the period covered by the ambient tremor catalog, which shows activity between 10 days before and after the passage of each posited triggering wavetrain. We look for slow slip during 3-month windows that include 11 of these triggering and non-triggering teleseisms, using continuous strain data recorded on two of the original NEHRP borehole Gladwin tensor strainmeters (GTSM) located within the distribution of tremor epicenters. We model the GTSM data assuming only tidal and 'drift' signals are present, and find no evidence of detectable slow slip, either ongoing when the posited triggering teleseismic waves passed or triggered by them. We infer a conservative detection threshold of about 10-20 nanostrain for abrupt changes and about twice this for more slowly evolving signals. This could be lowered by a factor of 2-3 by adding analyses of other data types, theoretical predictions of the temporal evolution of slow slip signals, and GTSM data calibration. The GTSM detection threshold we infer and strainfields predicted by simple dislocation models show that significant slip may go undetected. Detection depends strongly on the slipping fault's location (particularly depth) and size, which we describe in terms of equivalent earthquake moment magnitude, M. In the best case of the GTSM immediately above a very shallow slipping fault, detectable slip events must exceed M~2, and if the slow slip is beneath the seismogenic zone even M~5 events are likely to remain hidden.
U33A-0046
Nonvolcanic Tremor Activity in the Parkfield-Cholame region of California and the 2003 M6.5 San Simeon and 2004 M6.5 Parkfield Earthquakes
We investigate the possible relationship between nonvolcanic tremor (NVT) activity and fault zone processes
responsible for the generation of earthquakes. NVTs correlate with deep slow slip events along subduction
zones in Japan and Cascadia and recently it has been observed that tidal stresses as well as dynamic stress
changes produced by the passage of surface waves from large teleseisms are capable of inducing NVT
activity. These observations suggest that tremor rate changes are closely related to and even stimulated by
stress changes in deep fault zones. The computed values of tidal and dynamic stress changes also appear
to be small, on the order of a few kilopascals, indicating that NVTs are more sensitive to stress changes than
triggered earthquakes.
We present a multi-year analysis of the NVT activity since August 2001 along the San Andreas Fault near
Cholame, California detected and located using envelope cross-correlation methods. During this period two
strong earthquakes occurred in proximity of the tremor-generating region: the 2003 M6.5 San Simeon and
2004 M6.0 Parkfield earthquakes. We qualitatively and quantitatively show that since these two events, long-
term background NVT activity rates have increased. Our calculations of static Coulomb stress changes
induced by the two earthquakes show that the amplitude of the static stress changes is small (below a few
tens of kilopascals) but that they are enough to activate tremor activity and that the degree of stress change
transmitted into the tremor zone correlates with the degree of tremor activation. Finally we show that a new
pattern in the NVT activity has also emerged following the 2004 Parkfield event with the occurrence of quasi-
periodic tremor episodes that have persisted well into 2008. Our results suggest that NVT activity can provide
clues to variations in fault zone stress levels at tremor depth.
U33A-0047
Complex Non-volcanic Tremor Near San Andreas Fault Around Parkfield Triggered by the Great 2004 Sumatra Earthquake
In several instances, the passing surface waves from large earthquakes have ignited non-volcanic tremor
(NVT) on major faults. Still, the mechanism of tremor and its reaction to the dynamic stressing from different
body and surface waves is poorly understood. We examine tremor around Parkfield, California near the San
Andreas Fault triggered by the Mw 9.2 2004 Sumatra earthquake. The prolonged shaking reveals the
richest, and the most varied observations of dynamically-triggered tremor to date. The tremor appears in at
least three distinct locations and shows activity pulsing with encouraging stress, as has been observed in
other cases. The greatest amount of triggering and tremor modulation accompanies the long-period Love
waves. Also, at times, the tremor stops before the surface waves are complete, at other times it is still
quivering after the waves have passed. While tremor is found to be sensitive to small stress changes, there
are times when stresses of comparable magnitudes do not trigger noticeable tremor. Some tremors in this
NVT sequence appear to be associated with the passage of P waves, which is unusual and surprising
given the small stresses they impart.
http://staff.washington.edu/aghosh1/AGhoshParkfield.html
U33A-0048
Moving, Deep Non-volcanic Tremor at San Andreas Fault, CA, Imaged by Small-aperture Seismic Arrays
Non-volcanic seismic tremors at the San Andreas fault (California, USA) near Cholame has been reported earlier from analysis of regional network data. In order to determine precise locations we deployed 4 small-aperture seismic arrays in the tremor region for a time period of 6 weeks during fall 2007. All arrays consist of 10 seismic stations located within an area of roughly 1 km diameter, each. Distances between the arrays was between 30 and 50 km. The arrays are designed for optimal performance. All seismic stations were equipped with a short period (4Hz) three-component seismic sensor and a GPS-synchronized data logger, recording at a sample rate of 200 SPS. During 6 weeks of continuous recording we identified more than 70 tremor events of different magnitude. Instead of using travel time information derived by cross correlating envelopes we developed a method to absolutely localize tremor sources using seismic array techniques in combination with a source scanning technique. We found that almost all tremors occur in a surprisingly narrow, a few km wide zone striking with the San Andreas Fault and being offset to the surface trace by ~14km toward South-West at a depth centered at ~40km. Another interesting observation was the discovery that the tremor sources move in space during a tremor event, as was seen for subduction zone tremors in Japan.
U33A-0049
Time Reversal location of non-volcanic tremor in California.
Gomberg et al. (2008) recently reported the observation of tremor along the San Andreas transform plate boundary triggered by the surface waves from the 2002 Denali earthquake. This observation implies that the conditions needed for tremor generation are not unique to subduction zones where most tremor observations were previously made (Obara, 2002; Rogers&Dragert, 2003). Insight into the physical process of origin will certainly be provided by accurate location of tremor sources, especially their relative position to the transform fault in the case of California. Locating tremor is difficult because tremor signals are long lived and have few distinctive timing features. For such signals Time Reversal has several advantages; (1) it relies on no a priori assumption of the form of the source and (2) it employs the entire waveform of the received signal using the complete Green function for back-propagation. In this study, we propose to locate the tremor source near the San Jacinto fault in Southern California that was identified by Gomberg et al. (2008) using broad-band and high-broad-band data from local stations. An initial exploratory (purely numerical) study is necessary as the tremor case has important features not encountered in previous Time Reversal studies. The tremor source is not of impulsive nature which increases the complexity of the coda and of the noise generated by mode conversion in the Time Reversal process. The success of Time Reversal relies on a large number of receivers and/or significant wave scattering. Because the receivers are confined to the surface, the receiver aperture is limited. In addition, scattering is limited due to the smoothness of the numerical model and the presence of absorbing boundary conditions. Finally, the velocity model contains low velocity zones which trap seismic energy. This hampers our focus detection capability which relies on monitoring the amplitude of the reverse wavefield. Following this study Time Reversal treatment of the recorded data will be undertaken.
U33A-0050
Spectral Attributes of Tremor Sequences at SAFOD
Prior to the M6 Parkfield earthquake of 2004, a 46-channel surface array deployed at the SAFOD site recorded nonvolcanic tremors, low frequency acoustic emissions, microearthquakes, and earthquakes. Although the 46-channel array of geophones was deployed for the SAFOD drill bit seismic experiment, the array length of 1275 meters was sufficient to capture numerous earthquakes cataloged by the Northern California Seismic Network as well as other seismic activity. During the three and one half months of deployment, the surface array recorded a series of tremors which were often associated with microearthquake or earthquake activity. The great majority of the tremors generally have an approximate bandwidth of 0-15 Hz. Occasional short tremor events of an apparently different type have an approximate bandwidth of 0-50 Hz. Low frequency acoustic emissions with high frequency content exceeding 250 Hz were also detected by the array in the month preceding the M6 Parkfield earthquake. This presentation examines three tremor cycles in particular. One of undetermined length with a notable periodicity in amplitudes bursts within the tremor. A second with an apparent length exceeding an hour, and the third which apparently lasts more than four hours with a length that is yet to be determined.
U33A-0051
Regional characteristics of deep low-frequency earthquakes reproduced in a 3-D simulation for a subduction plate boundary
Recently, occurrences of deep low-frequency tremors/earthquakes have been recognized at depths of about 30 km in southwest Japan [Obara et al., 2002]. Similar deep low-frequency tremors have been detected also in Cascadia [Dragert et al., 2001] and southern Mexico [Kostoglodov et al., 2003]. The tremors occur sometimes independently and sometimes go into chain-reaction propagating at a speed of about 10 km/day, which may result from interaction among nearby small asperities. In this study, we formulate a 3-D subduction plate boundary model with two types of small asperities (r=2.5 or 5 km) chained along the trench at the depth of 30 km. Our simulations have successfully represented both independent occurrence and the chain-reaction type rupture propagation with various propagation directions (from eastward to westward, for example) along the trench with propagation speed of the chain reaction is about several km/day. Larger asperities (r=5 km) tend to propagate aseismic slip unilaterally with longer distance, while smaller (r=2.5 km) bilaterally with shorter distance, which suggests that we can estimate the size of localized asperities from characters of the propagation process.
U33A-0052
Synthesis of reference broadband seismogram of Very Low Frequency Earthquake: Direct measurement of VLFE source process
Western Japan and Ryukyu islands regions are very active region of slow seismic and geodetic event like slow slip, low frequency tremor and very low frequency earthquake (hereafter described as VLFE). The VLFE events excite the longer period seismic wave greater than 20sec mainly. And most VLFE events that detected Japanese seismic network originate at shallow depth near Nankai trough and Ryukyu trench. Seismicity pattern is very similar to earthquake swarm activities (Ishihara, 2002). In this research, we focus two swarms, the one is South-east off Kyuhsu and the other is around cross point of East-off Taiwan and South of Ishigaki island. The two regions have advantage for further research because of relative wider azimuth stationsf distribution. Generally seismic records for data analysis are used with high cut filtering to extract high S/N signals. Precursive analysis shows VLFEs are located in subduction wedge and the mechanism has steep dipping fault (Ito et. al. 2005). Most signals recognized as surface wave, so that hypocenter, especially source depth, and fault mechanism estimated by extracted signals remain some uncertainties. To understand more real source process, body wave phase seismic record is very valuable. We synthesized reference seismic records excited by VLFE events and we will discuss about source process of VLFE seismic events. We performed signal stacking of original VBB records. The time windows of stacking are determined by cross correlation in long period fine S/N signals. The data set for stacking are grouped by relative travel time between stations. We assumed that grouping seismic signalsf source are same location and have same source mechanism. We have succeeded the synthesis for one group which located SE off Kyushu near Nankai trough. The stacked records show some seismic phases before major surface wave packet. VLFE also excites higher frequency, greater than 1Hz, and small seismic waves (Ishihara, 2007). The higher frequency signal corresponds to precedent phases in stacked reference seismogram. Considering hypocenter location and apparent velocity, we believe these precedent phases are body waves. The reference seismograms show that source duration is apparently about 20sec. We will try to do modeling theoretical synthetic seismograms and inversion. Remarkably three components fine reference seismograms are obtained. More stable inversion and modeling are expected from these records.
U33A-0053
Tremor-tide correlations at Parkfield, CA
Recent studies linking non-volcanic tremor activity with tidal forcing have focused on short-term episodic tremor in subduction zone environments. In this study, we analyze the effect of tidal stresses and stress rates on tremor occurrence and magnitude at Parkfield, CA to determine if the Parkfield tremor, as a more continuous phenomena and in a different tectonic regime, demonstrates the same relationship to tides as its transient counterpart. For each of the 1777 tremor events in the 6-year catalog, we compute the volumetric, normal, shear, and coulomb stresses and stress rates for the central tremor zone at the initiation time of each event. Assuming that tremor and tides are uncorrelated, the percentage of tremor that occurs under a specific loading condition should be directly proportional to the percentage of time (normalized over the duration of the catalog) spent under those loading conditions. Using this as our null hypothesis, we assess the statistical significance of the deviations from this hypothesis to determine which conditions have the most profound influence on NVT.