T21A-0345 

Spatial and Temporal Patterns of Non-Volcanic Tremor Source Locations All Along the Cascadia Subduction Zone

* Boyarko, D C (boyarko4@gmail.com), Miami University, 114 Shideler Hall, Oxford, OH 45056, United States Brudzinski, M R (brudzimr@muohio.edu), Miami University, 114 Shideler Hall, Oxford, OH 45056, United States

Along the interface of a convergent plate boundary, the potential for megathrust earthquakes lay in the seismogenic zone, where the converging plates are locked and accumulating strain. Downdip from the seismogenic zone, the increasing temperatures and oceanic slab dehydration generate a transitional zone between the locked and free-slip portions of the subduction interface. Episodic Tremor and Slip (ETS), the correlation of slow slip events monitored by GPS observations and non-volcanic tremor (NVT) monitored by seismic signals, is believed to originate in this transitional zone. The processes that govern ETS and its relationship to megathrust earthquakes in space and time remain unresolved, although ETS has been proposed to impact the likelihood of megathrust earthquakes. Given the increased density of seismometers with the initial deployment of the Earthscope Transportable Array over the past 2 years, one can begin to examine the sources of NVT all along the Cascadia subduction zone in detail. We analyze NVT signals following Obara (2002) in which the cross-correlation of filtered envelope seismograms is used to define S-wave arrivals that are then converted to hypocentral locations. We apply this technique to the last ten years of seismic data collected at both short-term and long-term stations along the entire Cascadia subduction zone to determine spatial and temporal patterns in source locations. Specifically, we will address segmentation and recurrence defined by where and when ETS events start and stop, as well as hypocentral migration rates, directions and geographic variability.

T21A-0346 

How Repeatable are Episodic Tremor and Slip Events in Cascadia?

* Sweet, J R (jrsweet@u.washington.edu), Department of Earth and Space Sciences, University of Washington, 4000 15th Avenue NE, Seattle, WA 98195, United States Wech, A G (wech@u.washington.edu), Department of Earth and Space Sciences, University of Washington, 4000 15th Avenue NE, Seattle, WA 98195, United States Creager, K C (kcc@ess.washington.edu), Department of Earth and Space Sciences, University of Washington, 4000 15th Avenue NE, Seattle, WA 98195, United States Rubinstein, J L (justin@ess.washington.edu), Department of Earth and Space Sciences, University of Washington, 4000 15th Avenue NE, Seattle, WA 98195, United States

We employ a cross correlation method to locate deep non-volcanic tremor in western Washington during the July 2004, September 2005, and January 2007 Episodic Tremor and Slip (ETS) events. We use these locations to explore the spatio-temporal evolution of tremor and how consistently it behaves from event to event. We find that the 2004 and 2005 events were remarkably similar from start to end, while the 2007 episode had some similarities and some differences relative to the previous two events. A striking feature that is common to all three ETS episodes is that tremor amplitude was at its maximum when emanating from near Port Angeles, Washington. The 2004 and 2005 sequences were remarkably similar; they initiated in the eastern Straits of Juan de Fuca and migrated southwest towards Port Angeles. Both sequences then bifurcated, with tremor migrating both to the northwest and to the southeast. The 2007 sequence, in contrast, began in southern Puget Sound and migrated northwest to southern Vancouver Island. All three sequences were located within the same geographic area, roughly between the surface projection of the 30 and 45 km depth contours of the plate interface. We analyze tremor amplitude using data from six temporary, small-aperture seismic arrays. Tremor amplitude, corrected for attenuation and geometric spreading, exhibited a nearly linear increase over the first five days of each ETS sequence. The greatest amplitude for the 2004 and 2005 sequences occurred near Port Angeles during bifurcation. Interestingly, the amplitude was highest in the same area during the 2007 sequence, even though it simply migrated through the region. This maximum amplitude could be related to physical properties of the fault surface at that location or to increased stress caused by tidal forcing from the Straits of Juan de Fuca. Indeed, ETS tremor from these three episodes appears to be strongly modulated by tidal forcing.

T21A-0347 

Patches of tremor around the Mendocino Triple Junction

* Tran, A (adtran@seismo.berkeley.edu), UC Berkeley, 215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760, United States Allen, R (rallen@berkeley,edu), UC Berkeley, 215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760, United States

Episodic Tremor and Slip (ETS) has been observed all along the Cascadia Subduction Zone. Here we focus on Non-Volcanic Tremor (NVT) in the southern part of the Cascadia around the Mendocino Triple Junction. Continuous data from the Northern California Seismic Network has recently been made available at NCEDC, allowing for a more extensive study of tremor in the Mendocino area. In this study, data from 56 stations for the January 2006 to June 2007 period was analyzed for tremor using an automated approach similar to that used by Brudzinski and Allen (2007). More than 10 distinct NVT events with durations between 1.5 and 3 weeks were identified during the 18 month period. A distinct event is identified by clear peaks that rise above the noise level in the processed time series. Typically these events are visible on more than five stations. While some events show a single patch of tremor, other events can be divided into phases showing migration of the tremor activity. For example, an event in May 2006 begins as two distinct patches approximately 100 km apart which then migrate towards one another to form a single large patch. In late October 2006 tremor initiates north of the Mendocino triple junction and then propagates south of the triple junction over a period of approximately 5 weeks.

T21A-0348 

Observations of tremor using newly available seismic datasets in Cascadia

* Porritt, R (rwporrit@seismo.berkeley.edu), Berkeley Seismological Laboratory, University of California Berkeley 215 McCone Hall, Berkeley, CA 94704-4767, Allen, R (rallen@berkeley.edu), Berkeley Seismological Laboratory, University of California Berkeley 215 McCone Hall, Berkeley, CA 94704-4767,

Observations of episodic tremor and slip have now been made along the length of Cascadia. Until recently the seismic data available to study tremor has been limited, particularly in Oregon. However, continuous data from the NC network in northern California was recently made available at the NCEDC, Earthscope Transportable Array stations were deployed along the length of Cascadia, and this summer two Earthscope Flexible Arrays have been deployed in the region. The Mendocino Experiment (collaborative between U C Berkeley, University of Oregon, and Rice University) is a dense clustering of 80 broadband seismometers in northern California adjacent to the Mendocino Triple Junction for 1 year to gather continuous data to better image the crust and upper mantle in the region. The Flexarray Along Cascadia Experiment for Segmentation (FACES: collaborative between UC Berkeley and Miami University of Ohio) aims to provide data to expand on observations of segmentation in ETS characteristics along the Cascadia Subduction Zone. The deployment employs 23 broadband seismometers from southern Oregon to northern Washington. Using these datasets we examine the characteristics of tremor throughout the region.

T21A-0349 

Frequency Content of Subduction Zone Tremor and Similarities to Volcanic Tremor

* Sit, S (stefany.m.sit@lawrence.edu), Lawrence University, PO Box 599, Appleton, WI 54912, United States Brudzinski, M R (brudzimr@muohio.edu), Miami University, 114 Shideler Hall, Oxford, OH 45056, United States

Non-Volcanic Tremor (NVT) signals have been correlated in space and time with slow slip movements across several subduction zones in a process known as Episodic Tremor and Slip (ETS). Currently, the process responsible for ETS is unknown, however two possible models have been proposed. The first attributes tremor to temporal strain release caused by episodic slip along the plate interface, while the second suggests that tremor is fluid-related and can occur away from the interface. This study investigates patterns within the frequency domain of NVT in Cascadia and southern Mexico to help constrain the possible models. NVT signals from Cascadia have been collected from a mixture of seismometers including those of the Earthscope Tranportable Array and a host of permanent regional networks that span the subduction zone. A recently deployed network of 8 broadband stations provides the data for analysis along the Oaxaca segment of southern Mexico. Fast Fourier Transforms (FFT) were performed on successive hour-long segments during entire ETS events and showed a range of activity from 1 Hz to up to at least 10 Hz, along with a series of dominant peaks within the domain. Additionally, spectrograms constructed for selected time frames showed discrete banding at specific frequencies verifying the peaks seen with FFT are continuous, stable features. At both subduction zones, we find that primary peaks occur in the 1-4 Hz range, secondary peaks in the 2-7 Hz range, and additional smaller peaks in the 5-9 Hz range. Peaks seen in NVT are not as well-defined as many traditional volcanic tremor signals, but this may be due to superposition of several sharper peaks. Nevertheless, at stations where NVT amplitudes are most prominent, we observe both harmonic and inharmonic peaking common in volcanic tremor. The similarities between the frequency content of NVT and traditional volcanic tremor support a fluid-related model for ETS.

T21A-0350 

Spatial and Temporal Patterns of Non-Volcanic Tremor Source Locations Along the Oaxacan Segment of the Middle America Subduction Zone

* Hinojosa-Prieto, H R (hinojohr@muohio.edu), Miami University, Department of Geology 114 Shideler Hall, Oxford, OH 45056, United States Brudzinski, M R (brudzimr@muohio.edu), Miami University, Department of Geology 114 Shideler Hall, Oxford, OH 45056, United States Cabral-Cano, E (ecabral@igeofcu.unam.mx), Universidad Nacional Autonoma de Mexico, Instituto de Geofisica, Mexico City, DF 04510, Mexico Arciniega-Ceballos, A (maac@igeofcu.unam.mx), Universidad Nacional Autonoma de Mexico, Instituto de Geofisica, Mexico City, DF 04510, Mexico Diaz-Molina, O (oscard@igeofcu.unam.mx), Universidad Nacional Autonoma de Mexico, Instituto de Geofisica, Mexico City, DF 04510, Mexico DeMets, C (chuck@geology.wisc.edu), University of Winsconsin-Madison, Department of Geology and Geophysics, Madison, WI 53706, United States

Convergent tectonic plate boundaries generate large devastating earthquakes when plate motion accumulates tectonic stresses on the locked, seismogenic zone of the subduction zone interface. Downdip from the seismogenic zone, where increasing temperatures, pressure and slab dehydration limit frictional behavior, periodic slow slip events appear to occur in a transitional zone. The slow slip events have been shown to correlate with non-volcanic tremor (NVT), forming so-called episodic tremor and slip (ETS). The Oaxacan segment of the Middle America subduction zone is an ideal area for detailed ETS studies due to its relatively rapid convergent rates, shallow subduction angle, short megathrust earthquake recurrence intervals (decades), and short trench-to-coast distances that bring the seismogenic and transition zones of the plate interface as much as 250 km inland from the coastline. Previously analyzed slow slip events in southern Mexico occur over vast, deep areas of the subduction zone interface, and may even extend updip into the seismogenic zone, potentially playing a role in the timing of location of upcoming megathrust earthquakes. In addition to recent expansion of the permanent GPS station network, a new seismic deployment consists of 8 broadband (high-resolution) seismometers geographically dispersed inland along the Oaxacan segment, providing the means to examine NVT signals in detail for the first time in this region. Our analysis of NVT follows the approach of Obara (2002), using cross-correlation of filtered envelope seismograms to define S arrivals that are then converted to hypocentral locations. We apply this technique to data recorded over the first year of the deployment to determine spatial and temporal patterns in source locations. Specifically, we will address where and when NVT events start and stop, as well as migration rates, directions and geographic variability. A key issue we will investigate is why NVT signals appear more frequently (~2 month recurrence) than slow slip (12-24 month recurrence).

T21A-0351 

Non-Volcanic Tremors in Costa Rica

* Thorwart, M (thorwart@geophysik.uni-kiel.de), SFB574, University of Kiel, Department of Geoscience, Otto-Hahn-Platz 1, Kiel, 24118, Germany Rabbel, W (rabbel@geophysik.uni-kiel.de), SFB574, University of Kiel, Department of Geoscience, Otto-Hahn-Platz 1, Kiel, 24118, Germany Taylor, W (WTaylor@ice.go.cr), OSIVAM, Instituto Costarricense de Electricidad, Apdo. 10032-1000, San Jose, 1000, Costa Rica

Non-volcanic tremors were observed for the first time in Japan. They were also found at other subduction zones like in Cascadia, but also at transform faults like San Andreas fault. Intensity, duration and periodicity differ between the regions. In Cascadia the ocurrence of high non-volcanic tremor activity is related to silent slip events. Silent slip events in the region of Nicoya Peninsula, Costa Rica were observed by continious GPS measurements (Norabuena et al., JGR, 2004). In order to investigate a possible relation to non-volcanic tremor occurrence an array of 6 short period borehole stations was installed on Nicoya Peninsula, Costa Rica, since June 2006 as a part of the Collaborative Research Centre SFB 574 "Volatiles and Fluids in Subduction Zones". Within the first year of operation we observed non-volcanic tremor with a duration of 4 - 20 min and a frequency range of 1 - 5 Hz. A phase of increased activity took place between June and October 2007.

T21A-0352 

Multiple seismic array analysis of low frequency tremors in western Shikoku, Japan

* UENO, T (ueno@bosai.go.jp), National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennodai, Tsukuba, IBARAKI, 305-0006, Japan MAEDA, T (maeda@bosai.go.jp), National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennodai, Tsukuba, IBARAKI, 305-0006, Japan OBARA, K (obara@bosai.go.jp), National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennodai, Tsukuba, IBARAKI, 305-0006, Japan ASANO, Y (asano@bosai.go.jp), National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennodai, Tsukuba, IBARAKI, 305-0006, Japan TAKEDA, T (ttakeda@bosai.go.jp), National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennodai, Tsukuba, IBARAKI, 305-0006, Japan

In southwest Japan, low frequency tremors (LFTs) have been detected at deeper part of the seismogenic zone on the subducting Philippine Sea plate interface by high sensitivity seismograph network Hi-net operated by NIED. The tremor source area migrates during each episode along the strike of the subducting plate with a migration velocity of roughly 10 km/day, which was referred from source locations obtained by the envelope correlation method. In order to investigate the wave field propagation from the tremors in detail, we carried out a seismic observation campaign in western Shikoku, where the active tremor associated with the short-term slow slip event occurs with a recurrence interval of around six months. In this campaign, we deployed three seismic arrays above the belt-like LFT area with a spacing of about 20 km among arrays, during the period from February to May in 2007 because we anticipated the coming LFT episode to occur on March or April, 2007. Each array was composed of 32-channel receivers which mainly consist of vertical-component with natural frequency of 2 Hz. The receivers were placed with an average spacing of 30 m. Waveform data were recorded continuously with a sampling interval of 0.01 s. Since the active tremor episode with the short-term slow slip event occurred from 13 to 15 on March 2007 as expected, we performed frequency-wave number power spectrum analysis for each array recordings by the MUltiple SIgnal Classification (MUSIC) method in the period. At two seismic arrays, the apparent slowness were continuously low during the tremor episode. At the other array, the slowness was relatively high, and the arrival direction of the waves slightly changed with increasing time. In order to ascertain the change to be migration of the tremors, we located the sources of tremors by grid search method by using backazimuth and apparent slowness estimated at each array. We successfully located the sources of tremors, and detected the migration of the sources clearly with a velocity about 10 km/day. Furthermore we recognized that there were many arrival directions within short time window. This indicates that the tremors occur simultaneously at different places, which might suggest the inhomogeneity of the slip distribution on the plate interface.

T21A-0353 

Particle Motion Polarization Properties of Eastern Shikoku Tremor

* Wech, A G (wech@u.washington.edu), Department of Earth and Space Sciences, University of Washington, Box 351310, Seattle, WA 98103, United States Obara, K (obara@bosai.go.jp), Earthquake Data Center, Earthquake Research Department National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennodai, Tsukuba, 305-0006, Japan Maeda, T (maeda@bosai.go.jp), Earthquake Data Center, Earthquake Research Department National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennodai, Tsukuba, 305-0006, Japan Creager, K C (kcc@ess.washington.edu), Department of Earth and Space Sciences, University of Washington, Box 351310, Seattle, WA 98103, United States

Deep non-volcanic tremor or slow slip has been observed at many subduction zones around the world. The best observations, which are characterized by both a temporal and spatial coincidence of the two phenomena known as Episodic Tremor and Slip (ETS), are located along the Nankai Trough in Japan and in northern Cascadia. The eastern Shikoku region of southwest Japan experiences repeated episodes of deep non-volcanic tremor with a recurrence interval of a few months. The minor slow slip event is expected to occur coincident with each eastern Shikoku tremor episode. We analyze the polarization properties of eastern Shikoku tremors during the March 2007 event and compare the observed polarizations to those predicted from tremor emanating from the plate interface. Preliminary polarization analysis of 6 NIED Hi-net seismic records in eastern Shikoku show a consistent trend toward shallow dip angles, suggesting horizontal particle motion, when the tremor sources are beneath the stations. Despite low linearity and significant scatter, probably related to considerable heterogeneity of the continental crust, the polarization azimuths observed at most of the stations align approximately with the direction of plate convergence. Predicted polarization directions of near-vertical upgoing S-waves radiated from shear slip on the plate interface align with the relative motion between the Philippine Sea and Eurasian plates. This finding suggests that eastern Shikoku tremors represent shear slip on the plate interface, consistent with similar polarization observations in Cascadia, and with inferences in Japan from focal mechanisms of low- frequency earthquakes and very low- frequency earthquakes.

T21A-0354 

A new hypocenter determination method using the summation of waveform cross- correlation

* Ohta, K (ohta@eps.s.u-tokyo.ac.jp), Dept. EPS, Univ. Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo, 113-0033, Japan Ide, S (ide@eps.s.u-tokyo.ac.jp), Dept. EPS, Univ. Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo, 113-0033, Japan

Deep low frequency tremor is a swarm activity of low-frequency earthquakes (LFEs) at least in the western Shikoku, Japan [Shelly et al., 2007]. LFEs are considered as shear slip on the plate interface because their focal mechanism has a low-angle thrust fault plane [Ide et al., 2007] and because precisely determined hypocenters are located on the plate interface [Shelly et al., 2006]. However, it is not obvious whether LFEs in other regions have the same characteristics. For example, in the Tokai region and the Kii Peninsula the hypocentral depths of LFEs determined by Japan Meteorological Agency are widely distributed from 20 to 50 km. Precise hypocenters of LFEs are essential for discussion on regional diversity of low frequency tremor. The low signal-to-noise (S/N) ratio of LFE records can be a source of larger estimated error in the hypocenter determination. Although waveform cross-correlation is a useful tool for precise hypocenter determination, it is not reliable in the case of the low S/N ratio because the value of cross-correlation is generally small. However, Shelly et al. [2007] showed that the summation of waveform cross-correlation for many stations can enhance event detectability. Similarly, using the summation of waveform cross-correlation, we develop a new hypocenter determination method that is less sensitive to noises. We apply this method to LFEs in the Tokai region and the Kii Peninsula. The data are 3-component velocity records from the NIED Hi-net, bandpass filtered between 2-8Hz. We first determine the relative hypocenral location between a pair of LFEs in a event list. Assuming a layered structure, we calculate theoretical differential travel times of body waves relative to the arrival times from the one event, After shifting the records of the other event using these differential travel times, we calculate the summation of the waveform cross-correlation coefficient between two events for all stations. The relative location is determined to maximize this summation by a grid search. For some pairs the maximum takes a statistically significant value that is not explained by a Gaussian distribution. Then we invert the relative locations to determine the hypocenter distribution using least square which is weighted according to a value of sums of waveform cross-correlation.

T21A-0355 

Extracting Low Frequency Earthquakes From Tremor

* Brown, J R (jrbrown5@stanford.edu), Department of Geophysics, Stanford University, 397 Panama Mall, Stanford, CA 94305, United States Beroza, G C (beroza@stanford.edu), Department of Geophysics, Stanford University, 397 Panama Mall, Stanford, CA 94305, United States Shelly, D R (dshelly@gmail.com), Berkeley Seismological Laboratory, UC-Berkeley, 307 McCone Hall, Berkeley, CA 94720, United States

Recent studies have shown that deep tremor in Shikoku consists of a swarm of low frequency earthquakes (LFEs) that occur as slow, shear slip on the down-dip extension of the primary seismogenic zone of the Nankai Trough. In this study, we develop a method for identifying LFEs within tremor. The method is a matched-filter algorithm, similar to the technique used to infer that tremor in parts of Shikoku is comprised of LFEs. In our case, however, we do not know the origin times and locations of the LFEs a priori. We search for LFEs using a running autocorrelation of the tremor waveforms for all Hi-Net stations in the vicinity of the tremor source. We apply this method to Hi-Net recordings of tremor and demonstrate that it allows us to extract both known and previously unidentified LFEs. Lags showing a high degree of similarity in the autocorrelation are either repeats, or near repeats, of LFEs within the tremor. Once LFEs are identified, we apply waveform cross-correlation to measure arrival times and locate them. This approach allows us to extend the analysis of Shelly et al. [2007] to parts of the Nankai Trough that have only sparse LFE coverage. It should also allow us to extend our analysis to other regions that experience deep tremor, but where LFEs have not yet been identified. Shelly, D. R., G. C. Beroza, and S. Ide, Non-volcanic tremor and low frequency earthquake swarms, Nature, 446, doi:10.1038/nature05666, 2007.

T21A-0356 

Detection Method of Low-Frequency Earthquakes in the Non-Volcanic Tremor Beneath the San Andreas Fault

* Cannata, A (andrea.cannata@unict.it), Università di Catania, Dipartimento di Scienze Geologiche, Corso Italia 57, Catania, 95129, Italy Hellweg, M), University of California at Berkeley, Berkeley Seismological Laboratory, 215 McCone Hall, Berkeley, CA 94720, United States Nadeau, R M (nadeau@seismo.berkeley.edu), University of California at Berkeley, Berkeley Seismological Laboratory, 215 McCone Hall, Berkeley, CA 94720, United States Gresta, S (gresta@unict.it), Università di Catania, Dipartimento di Scienze Geologiche, Corso Italia 57, Catania, 95129, Italy

Recent studies showed the occurrence of non-volcanic tremor in the Cascadia Subduction zone, in the southwest Japan and along the San Andreas Fault. In the southwest Japan the non-volcanic tremor was accompanied by low-frequency earthquakes (LFEs). Using a method developed on tremor at Mt. Etna volcano, we apply a detection procedure to look for LFEs in the non-volcanic tremor recorded during 2004-2006 along the San Andreas Fault, by a seismic network comprising 13 3-C stations. This method is composed of two steps, trigger detection and trigger selection. The former is based on three algorithms, standard, adaptive standard and adaptive recursive STA/LTA, that evaluate the ratio between short- and long-term energy density to find amplitude transients. The main differences between them consist in the manual or automatic (adaptive) selection of the window length of the STA and LTA and in the rectangular or decaying exponential impulse response (standard and recursive, respectively). Using these algorithms we obtain a trigger list for each station. Trigger selection consists of selecting from these lists the triggers common at many of the stations which are characterised by similar spectral content of the seismic signal following the triggers and a time distribution of triggers consistent with the distribution of the stations. Applying this procedure we were able to find small amplitude transients related to LFE activity.

T21A-0357 

Non-Volcanic Tremor Near Parkfield, CA Systematically Excited by Teleseismic Waves

* Peng, Z (zpeng@gatech.edu), School of Earth and Atmospheric Sciences, Georgia Institute of Technology, 311 Ferst Drive, Atlanta, GA 30338, United States Vidale, J E (john.vidale@gmail.com), Department of Earth and Space Science, University of Washington, Box 351310, Seattle, WA 98195, United States Rubinstein, J L (justin@ess.washington.edu), Department of Earth and Space Science, University of Washington, Box 351310, Seattle, WA 98195, United States Gomberg, J (gomberg@usgs.gov), US Geological Survey, Box 351310, Seattle, WA 98195, United States

Non-volcanic tremor triggered by teleseismic waves was discovered recently along the subduction zones in Japan and Cascadia, and along the transform plate boundary in CA. Here we summarize non-volcanic tremor along the San Andreas fault (SAF) near Parkfield, CA triggered by the surface waves of regional and teleseismic events. We analyze 10 M ≥ 8.0 earthquakes since 2001, the M6.7 Nenana Mountain and M7.9 Denali, Alaska earthquakes in 2002 and the 2005 M7.2 Mendocino, California earthquake. We identify triggered tremor as bursts of high-frequency (~ 3-15 Hz), non-impulsive seismic energy that is coherent among many stations, and has a significant component in phase with the passing of the surface waves. We qualitatively judge the clarity of tremor observations and find the strongest, most coherent examples for the M7.9 Denali, M8.3 Hokkaido, M9.1 Sumatra, and M8.1 Kuril Islands earthquakes. The M6.7 Nenana Mountain earthquake did not trigger visible tremor, and the evidence for triggered tremor for the remaining 8 events is equivocal. The identification of tremor does not correlate strongly with peak ground velocity, but may correlate with cumulative energy density for long- period (≥ 30 s) surface waves. The observations suggest that longer-period waves may be a more effective trigger, most likely due to a better penetration to depth where tremors occur. Our observation, in concert with those of Gomberg et al., Vidale et al., and Rubinstein et al. [this meeting], suggests that non-volcanic tremor triggered by teleseismic waves is much more widespread than previously thought, and the effective stress, or the frictional coefficient is very low at depth along the SAF near Parkfield.

T21A-0358 

Tidal Variation in Non Volcanic Seismic Tremor Activity at the Chile Triple Junction

* Gallego, A (agallego75@gmail.com), Dept. of Geological Sciences University of Florida, P.O. Box 112120, 241 Williamson Hall, Gainseville, FL 32611, United States Russo, R M (rrusso@ufl.edu), Dept. of Geological Sciences University of Florida, P.O. Box 112120, 241 Williamson Hall, Gainseville, FL 32611, United States Comte, D (dcomte@dgf.uchile.cl), Depto. de Geofisica Universidad de Chile, Blanco Encalada 2002, Santiago, 837-0449, Chile Mocanu, V I (mocanu@gg.unibuc.ro), Dept. of Geophysics University of Bucharest, 6 Traian Vuia Str, Bucharest, RO-70139, Romania Murdie, R E (ruth.murdie@ctbto.org), CTBTO, P.O. Box 1200 Vienna International Centre, Vienna, A-1400, Austria VanDecar, J C (jvandecar@hotmail.com), DTM, Carnegie Inst. of Washington, 5241 Broad Branch Rd. NW, Washington, DC 20015, United States

We present evidence that non volcanic seismic tremor detected at Chile Ridge Subduction Project (CRSP) seismic network correlates with the semidiurnal (M2) tidal amplitude variation, enhanced by the lunar monthly (Mm) orbital cycle. Tremors were detected at 14 broadband seismic stations in the Los Chonos Archipelago and Taitao Peninsula by two methods: simple visual inspection of correlated tremor on seismograms at CRSP stations, and an automated search of the tremor frequency band (5-10 Hz). The two methods yield similar results. Analysis of two years of data (2005-2006) reveals a positive correlation (R = 0.76) between daily tremor duration and daily tidal variations, with a lag of 2.5 days between maximum tidal amplitude difference and tremor activity. The 28 day tidal cycle is produced by conjunction (New Moon), generating the highest and lowest tide amplitudes of the orbital cycle. Gravitational traction due to tidal acceleration is maximum near the latitude of the study region, and is enhanced by the ocean loading. At least two hypotheses could explain the correlation: stress cycling on the subducted Taitao and Darwin transform faults due to solid earth tide and ocean loading, leading to tremor generation; or enhanced hydrothermal activity during fluid release in the subducted Chile ridge structure. http://www.clas.ufl.edu/users/russo/

T21A-0359 

Precise relocations of deep low-frequency earthquakes beneath NE-Japan and estimation of 3-D seismic velocity structure in the surrounding areas

* Nii, K (nii@aob.geophys.tohoku.ac.jp), Research Center for Prediction of Earthquakes and Volcanic Eruptions, Graduate School of Science, Tohoku University, 6-6 Aza-Aoba, Aramaki, Aoba-ku, Sendai, 980-8578, Japan Okada, T (okada@aob.geophys.tohoku.ac.jp), Research Center for Prediction of Earthquakes and Volcanic Eruptions, Graduate School of Science, Tohoku University, 6-6 Aza-Aoba, Aramaki, Aoba-ku, Sendai, 980-8578, Japan Ueki, S (ueki@aob.geophys.tohoku.ac.jp), Research Center for Prediction of Earthquakes and Volcanic Eruptions, Graduate School of Science, Tohoku University, 6-6 Aza-Aoba, Aramaki, Aoba-ku, Sendai, 980-8578, Japan Nakajima, J (nakajima@aob.geophys.tohoku.ac.jp), Research Center for Prediction of Earthquakes and Volcanic Eruptions, Graduate School of Science, Tohoku University, 6-6 Aza-Aoba, Aramaki, Aoba-ku, Sendai, 980-8578, Japan Zhao, D (zhao@aob.geophys.tohoku.ac.jp), Research Center for Prediction of Earthquakes and Volcanic Eruptions, Graduate School of Science, Tohoku University, 6-6 Aza-Aoba, Aramaki, Aoba-ku, Sendai, 980-8578, Japan Hasegawa, A (hasegawa@aob.geophys.tohoku.ac.jp), Research Center for Prediction of Earthquakes and Volcanic Eruptions, Graduate School of Science, Tohoku University, 6-6 Aza-Aoba, Aramaki, Aoba-ku, Sendai, 980-8578, Japan Zhang, H (hjzhang@mit.edu), Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA 02139- 4307, United States Thurber, C H (clifft@geology.wisc.edu), Department of Geology and Geophysics, University of Wisconsin-Madison, 1215 West Dayton Street, Madison, WI 53706, United States

It is well known that deep low-frequency earthquakes (DLFEs) with dominant frequency around 2 Hz occur at depths from 20 to 40 km in volcanic areas beneath NE-Japan (Hasegawa and Yamamoto, 1994; Okada et al., 2000). In addition, it has been pointed out that the source mechanism of DLFEs consists of both DC and non-DC components by moment tensor inversions (Okada et al., 2000; Nakamichi et al., 2003). These observations suggest that the generation mechanism of DLFEs is closely related with fluids (or magma). In this study, we precisely relocate the DLFEs occurring beneath Mt. Iwate and Mt. Naruko, determine the 3-D seismic velocity structure of the crust and uppermost mantle in the region surrounding them, and discuss the common characteristics for the two regions. We applied the DD tomography method (Zhang and Thurber, 2003) to data obtained by the seismic networks of Tohoku University, JMA and NIED. Note that we have included events within the subducting Pacific slab in the tomography inversion to increase the resolution of the lower crust structure. We employed the 3D seismic velocity model of the whole NE-Japan by Nakajima et al. (2001) as the initial model. Obtained results beneath Mt. Iwate show that DLFE forming two clusters are concentrated in the two locations to the south and the northeast of Mt. Iwate, respectively. Most of them are located at depths a few km shallower than the Moho and just above the low Vs and high Vp/Vs areas, which probably correspond to regions of partial melting. Note that they are located not within but above the S-wave low-velocity zone. This feature also can be seen for the DLFEs beneath Mt. Naruko.

T21A-0360 

Deep low-frequency tremor and very-low-frequency earthquake as indicator for slow slip event at the transition zone on the plate interface in southwest Japan

* Obara, K (obara@bosai.go.jp AF: AF: AF: AF: AF:

Three types of deep slow earthquakes; low-frequency tremor [Obara, 2002], very-low-frequency (VLF) earthquake [Ito et al., 2007], and short-term slow slip event [Obara et al., 2004] have been detected at the transition zone on the plate interface in the Nankai subduction zone. These slow earthquakes characterized by different frequency properties occur simultaneously and reflect the stress relaxation process at the deeper side of the seismogenic zone. Considering magnitude of each slow earthquake, the short-term slow slip event may be the primary phenomenon as a stick slip on the plate interface and other slow events are triggered by the transient slip. We examine the detail time history of these slow earthquakes to reveal the interaction between each slow event. The tremor source area is divided into some segments along the narrow belt-like zone considering with the space-time distribution of activities. Western Shikoku, northern Kii and Tokai segments are characterized by the recurrence interval of six months. In these areas, the short-term slow slip event can be detected by some tiltmeters and the fault models are estimated. On the other hand, tilt changes are rarely recognized during active tremor stage with a recurrence interval of three months at eastern and central parts of Shikoku and central Kii segments. There are many minor tremor activities without tilt change, however major tremor episodes are always accompanied by the slow slip event and migrate unilaterally or bilaterally. Comparison between the tilting record and tremor activity indicates that the beginning of the detectable slow slip event coincides with activation of tremor. This supports an idea that the slow slip event occurs first and induces the tremor. The front of the migrating tremor activity is very sharp and behind of the front the tremor activity keeps for a while. This observational fact suggests that the migration of tremor is induced by the rupture propagation of the slip event and the tremor behind the migration front keeps highly activity by the slipping. The VLF earthquake with predominant period of 20 seconds has been detected during the episode of tremor and slip. Based on the centroid moment tensor analysis, the VLF earthquake is located at the same position of the tremor source area and the focal mechanism is the reverse fault type coincident with the slow slip event and the plate interface geometry. In many cases, VLF earthquakes occur at the most active stage of tremors and coincident with tremor migration. Sometimes, VLF earthquakes seem to correspond to the acceleration stage of ground tilt. Therefore, tremor and VLF earthquake might reflect the change in slip velocity on the plate interface. The crustal movement caused by the short-term slow slip event is too small to resolve the precise space-time distribution of slip; however the monitoring of tremor and VLF earthquake is considered as the sensitive indicator for plate motion.

T21A-0361 

Precursory slow crustal deformation before short-term slow slip event in January 2006, recorded at Shingu borehole station southern Kii Peninsula

* Fukuda, M (fkdmst@seis.nagoya-u.ac.jp), Nagoya University, Earth and Environmental dynamics course, Earth and Environmental Sciences, Graduate School of Environmental Studies, Nagoya University, Furo-cho Chikusa-ku, Nagoya, 464- 8601, Japan Sagiya, T), Nagoya University, Earth and Environmental dynamics course, Earth and Environmental Sciences, Graduate School of Environmental Studies, Nagoya University, Furo-cho Chikusa-ku, Nagoya, 464- 8601, Japan

In January 2006, a deep low frequency tremor activity and an associated short-term slow slip event occurred in the eastern Kii Peninsula and this coupled activity migrated to the northeast at a rate of 10km/day. We are monitoring crustal deformation at Shingu borehole station in the southeastern Kii peninsula. The Shingu borehole site is located about 100km landward from the Nankai Trough axis, and close to the epicenter of the 1944 Tonankai Earthquake. The borehole is 500 m deep and is equipped with an integrated multi-component borehole monitoring system developed by Ishii et al. (2002), consisting of 6 strain sensors (4 in horizontal, 2 in vertical), 2 pendulum tilt sensors, a magnetic direction finder, and a quartz thermometer. Each signal is originally recorded with a sampling frequency of 50 Hz. We decimated the original data into hourly data, which we decomposed into tidal response, barometric response, smoothed trend and random noise component by applying BAYTAP-G software [Tamura et al., 1991]. In the trend component from November 2005 to March 2006, we did not found deformation signal at the time of the Jan. 2006 tremor event. However, we found three significant slow strain changes from the processed records. Two of them coincide with the occurrence of the tremor activities in the southern Kii Peninsula, and are characterized by N-S contraction (0.019-0.031 ppm) and E-W extension (0.025-0.038 ppm). These are the first evidence of the short-term slow slip event in this area. The third change is characterized by NW-SE extension (0.026 ppm), N-S contraction (0.012 ppm), E-W extension (0.022 ppm), and southwestward tilting (0.23 micro rad). It occurred from December 29, 2005 to January 2, 2006, just before the tremor and slip event in January 2006, but was not accompanied by any tremor activity. We conducted a series of inversion analysis to infer the source of this possible slow slip event. We assumed that the slow slip event was caused by a reverse fault slip on the plate boundary, and found that the fault model with the following parameters successfully reproduces the observation. Here are the estimated parameters; (34.08 (latitude), 136.23 (longitude), 214 degree (strike angle), 22.1 degree (dip angle), 35 km (depth), 4.1 cm (slip), 90.64 degree (slip angle), 34.1 km (length), 24.4 km (width)). The estimated moment magnitude was 5.9. The result strongly suggests that the strain change was caused by a slow slip on the plate boundary at the southwestern extension of the tremor activity in January 2006. Since the January 2006 tremor event migrated to the northeast, the slow transitions seem to be the direct precursory of the tremor activity. This will probably be the first observation of the short-term slow slip event without associated tremor activity along the Nankai Trough. The estimated fault model fills a gap between clusters of tremor activities, indicating lateral heterogeneity along the down-dip end of the seismogenic zone. Also, our result suggests that the slow slip is essential process and tremors are just accompanying.

T21A-0362 

Transient crustal deformation modeled from Japanese GEONET GPS time series

* Liu, Z (zliu@jpl.nasa.gov), Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, United States Owen, S (susan.E.owen@jpl.nasa.gov), Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, United States Dong, D (danan.dong@jpl.nasa.gov), Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, United States Webb, F H (fhw@jpl.nasa.gov), Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, United States Lundgren, P R (Paul.R.Lundgren@jpl.nasa.gov), Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, United States Hetland, E A (eah@gps.caltech.edu), Seismological Laboratory, Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, United States Simons, M (simons@caltech.edu), Seismological Laboratory, Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, United States

The Japanese continuous GPS network (GEONET) with ~1450 stations provides a unique opportunity to study ongoing subduction zone dynamics, and crustal deformation at different space-time scales. Recently a reanalysis of GEONET data using a GIPSY-based Network Processor yields high precision GPS solutions for the whole network [Owen et al., 2006]. We construct the detailed subduction plate interface along the Nankai trough based on the composite model [Wang et al., 2004] and approximate it with a triangular mesh to better represent the complex 3-D geometry. We derived GPS site velocities in SW Japan relative to the stable part of the Amurian Plate and estimate the interplate coupling along the SW Japan subduction zone. Cleaned GPS position time series also reveal new characteristics of the slow slip events (SSE). For example, the Tokai slow slip event appears to affect a larger spatial distribution of stations compared to previous studies. We applied the extended network inversion filter [MacGuire and Segall, 2003] to model the spatial and temporal distribution of this slow slip event using longer data from 2000.0 to 2004.0. Our initial results show that this SSE involves three sub-events since 2000.0. They all initiated at roughly the same location ~(137.6E, 34.7N) in a weakly coupled zone. The maximum cumulative slip locates at the deeper end of the seismogenic zone with an equivalent Mw~7.1 as of Dec. 2003. We estimate the stress change from these sub-events and investigate possible relations with seismicity rate changes in the region. We also investigate other slow slip events including the 2003 Bungo Channel and discuss their relation with plate coupling as well as imaged structural heterogeneities in the region.

T21A-0363 

The 2007 Boso Slow Slip Event and the associated earthquake swarm

* Sekine, S (ssekine@bosai.go.jp), National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennoudai, Tsukiba, 305-0006, Japan Hirose, H (hirose@bosai.go.jp), National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennoudai, Tsukiba, 305-0006, Japan Kimura, H (kimura@bosai.go.jp), National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennoudai, Tsukiba, 305-0006, Japan Obara, K (obara@bosai.go.jp), National Research Institute for Earth Science and Disaster Prevention, 3-1, Tennoudai, Tsukiba, 305-0006, Japan

In the Boso Peninsula, which is located in southeast of the Japan mainland, slow slip events (SSE) have been observed by the GEONET GPS array operated by the Geographical Survey Institute Japan and the NIED tiltmeter network every 6-7 years (Ozawa et al.,2003; NIED 2003). The unique characteristics of the Boso SSE are that earthquake swarm activities have also occurred in association with the SSE. The latest activity of the SSE and the earthquake swarm took place in August 2007. On 13th August, an earthquake swarm began to occur at east off Boso Peninsula and the slow tilt deformations also started. The earthquake sources migrated to the NNE direction, which is the same direction of the relative plate motion of the subducting Philippine Sea Plate with respect to the overriding plate. The largest earthquake in this episode (Mw 5.3) occurred on 16th and the second largest one (Mw 5.2) on 18th. Most of the larger earthquakes show low- angle thrust type focal mechanisms that are consistent with the plate motion and the geometry of the subduction plate interface. The tilt changes seem to stop on 17th and the activity of the swarm rapidly decreases after 19th. The maximum tilt change of 0.8 micro radian with northwest down tilting was observed at KT2H, the nearest station from the source region. Based on the tilt records around Boso Peninsula, we estimate a fault model for the SSE using genetic algorithm inversion to non-linear parameter and the weighted least squares method to linear parameters. As a result, the estimated moment magnitude and the amount of slip are 6.4 and 10 cm, respectively. The size and the location of the SSE are similar to the previous episodes. The estimated fault plane is very consistent with the configuration of the plate interface (Kimura et al., 2006). Most of the earthquakes are located on the deeper edge of the estimated SSE fault area. The coincidence of the swarm and the SSE suggests a causal relation between them and may help us to understand the mechanism of earthquake triggering. On the other hand, in the southwest Japan, tremors which repeat at approximately six months have occurred in association with the SSE. To compare with these two types of SSE may suggest the difference of the boundary conditions on the same subducting plate.

T21A-0364 

The time constant variations of slow slip events in the south Alaska subduction zone

* Ohta, Y (ohta@aob.geophys.tohoku.ac.jp), Research Center for Prediction of Earthquakes and Volcanic Eruptions, Tohoku University, 6-6 Aza-Aoba, Aramaki, Aoba-ku, Sendai, 980-8578, Japan Freymueller, J T (jeff@giseis.alaska.edu), Geophysical Institute, University of Alaska Fairbanks, 903 Koyukuk Drive, Fairbanks, AK 99775-7320, United States Miura, S (miura@aob.geophys.tohoku.ac.jp), Research Center for Prediction of Earthquakes and Volcanic Eruptions, Tohoku University, 6-6 Aza-Aoba, Aramaki, Aoba-ku, Sendai, 980-8578, Japan

Slow Slip events (SSEs), episodic periods of slip on the plate interface that are very slow in comparison to earthquakes, but much faster steady plate motions, have been identified at several subduction zones. One characteristic shared by all of these events is that they occur near or downdip of the base of the seismogenic zone as defined by the slip pattern of great earthquakes. A large slow slip event occurred at the Alaska subduction zone during 1998-2001 [Ohta et al., 2006, EPSL]. The SSE occurs downdip of the Prince William Sound asperity, on a section that accumulated some slip deficit before and after the SSE. During the SSE, >20,000 sq. km of the plate interfae slipped > 10cm, for a cumulative moment magnitude of Mw=7.2. Slip during this event was accompanied by significant non-volcanic tremor, which located in the same area as the slip. The area of the SSE did not have significant slip in the 1964 earthquake, compared to the slip of the main asperity. Our previous work did not attempt to determine the time history of slip in the SSE, because of a lack of continuous GPS data, but a reanalysis of the data that includes a few new sites allows us to estimate the approximate time history. Freymueller and Ohta (2007) found the short-term abnormal displacement in the south Alaska region during 2002-2006. The end of 2005 event may have the biggest event in the short-term abnormal displacements. We assumed the abnormal deformation as short-term SSE at the plate interface. The slip of short-term event reaches a maximum around 12-14 mm at plate boundary depths of 25-45 km, which corresponded to the transition zone of the plate interface. The short-term event fault located the neighboring of the long-term one during 1998-2001 investigated by Ohta et al., (2006.EPSL). The SSEs may play key role of strain accumulation in subduction zone. We will more precisely compare the displacement pattern and time history of the large SSE to the displacements to evaluate these smaller events actually represent SSEs on the same part of the plate interface.

T21A-0365 

Detection of Slow Slip Events Along the Cascadia Subduction Zone Using Plate Boundary Observatory Borehole Strainmeters

* McCausland, W (wmccausland@usgs.gov), Cascades Volcano Observatory, 1300 SE Cardinal Ct Building 10, Suite 100, Vancouver, WA 98683, United States Roeloffs, E (evelynr@usgs.gov), Cascades Volcano Observatory, 1300 SE Cardinal Ct Building 10, Suite 100, Vancouver, WA 98683, United States

In the spring of 2005, the Plate Boundary Observatory (PBO) began installing borehole strainmeters in the Pacific Northwest. Currently there are 18 borehole strainmeters operating from Vancouver Island, Canada to Southern Oregon that are favorably located to detect slow slip along the Cascadia subduction zone. While the longest (> two weeks) subduction tremor episodes are accompanied by slow slip events, there are shorter duration tremor episodes that have not been accompanied by GPS-detected slow slip events. It is possible that a slow slip event does occur, but is below the current resolution of the GPS network. Using data from these PBO strainmeters, we developed a scanning technique to detect strain transients. The strainmeter data are band-passed between 5 and 16 days to eliminate tides and long-term variations. In this band, the instrument responds primarily to atmospheric pressure. Excursions of the gauges from the atmospheric pressure variations are investigated as possible tectonic strain transients. Several features of the strainmeters aid us in distinguishing between tectonic strain transients and other variations in the data. First, tectonic transient strain signals will be recorded similarly on closely clustered strainmeters; there are 4 such clusters along the Cascadia subduction zone. Second, if the areal strain recalculated using two orthogonal gauges is not consistent with the areal strain calculated on three equally spaced gauges, then the signal is not caused by strain. Pumping or rainfall can cause strain variations that we can distinguish from tectonic events by looking at other measured quantities including fluid pressure, precipitation, etc. Thus far, in northern Washington and Canada, all the known transient slow slip events recorded on the strainmeters were also detected by the GPS network. In southern Oregon, our scanning technique found one possible tectonic event on the B035-B036 cluster near Grant's Pass during late July and early August 2007. Other researchers (Szeliga et al, 2004; McCausland et al, 2005; Brudzinski and Allen, in press) have reported recurring subduction tremor and slow slip in this area. Further corroborating our observations, cleaned GPS data from the PANGA website for the east-west component of nearby station DDSN, resembles signals from previous slip events.

T21A-0366 

Constraining transient vertical deformation associated with slow slip events in Cascadia.

* Holtkamp, S (sgh46@cornell.edu), Department of Earth and Atmospheric Sciences, Cornell University, Snee Hall, Cornell University, Ithaca, NY 14853, United States * Holtkamp, S (sgh46@cornell.edu), Geology Department, Miami University, 114 Shideler Hall, Miami University, Oxford, OH 45056, United States Brudzinski, M (brudzimr@muohio.edu), Geology Department, Miami University, 114 Shideler Hall, Miami University, Oxford, OH 45056, United States Pritchard, M (mp337@cornell.edu), Department of Earth and Atmospheric Sciences, Cornell University, Snee Hall, Cornell University, Ithaca, NY 14853, United States

Dense GPS and seismic networks in Cascadia have recorded Episodic Tremor and Slip (ETS) events throughout the entire Cascadia margin and studies of these events consistently place the majority of slip down-dip from the locked portion of the interface in the transition zone to stable sliding along the interface. Motions recovered in the horizontal components of GPS stations along the margin show weeks of transient motion towards the trench during ETS events, consistent with release of strain due to the convergence of the Juan de Fuca and North American plates. In this study, we examine horizontal and vertical motions of permanent GPS sites from the PANGA network for transient motions using a scanning hyperbolic tangent fit algorithm. Larger uncertainties on the vertical component make identification of vertical transients independent of horizontal components difficult, so we rely on horizontal components to establish the presence and timing of slow slip events. Nevertheless, our examination of data over the last ten years reveals 40 events with significant transient vertical motion where the hyperbolic tangent fits surpass a simple linear slope with f-test values greater than 99 percent. Not surprisingly, vertical displacements for nearly all of these events are larger than the maximum value of 5 mm observed at site DRAO which represents typical scatter on stable North America. These events show uplift near the trench and subsidence farther inland with an inflection point at about 40 km above the plate interface, consistent with thrust- type faulting events on the plate interface with a down-dip extent of ~40 km. Examination of the densely sampled Northern Washington/Southern Vancouver Island yields 28 individual observations that show a peak in the uplift at ~25 to 30 km above the plate interface and a peak in the subsidence at ~50 to 60 km above the interface. To further investigate the geographic patterns in vertical motion, we will examine available InSAR data for total line-of- sight (LOS) transient motion. Preliminary modeling of slip on the interface during ETS events incorporating vertical offsets will be discussed as well as consistency between horizontal and vertical data sets in modeling results. Surface representations of vertical transient motion during ETS events will add constraints to the depth extent of slip and whether strain release may be infiltrating into the seismogenic zone.

T21A-0367 

Slow slip events, temperature and interseismic coupling at the Hikurangi subduction zone, New Zealand

* McCaffrey, R (R.McCaffrey@gns.cri.nz), GNS Science, PO Box 30368, Lower Hutt, 5040, New Zealand Wallace, L (L.Wallace@gns.cri.nz), GNS Science, PO Box 30368, Lower Hutt, 5040, New Zealand Beavan, J (J.Beavan@gns.cri.nz), GNS Science, PO Box 30368, Lower Hutt, 5040, New Zealand

Slow slip events (SSE) at the Hikurangi subduction zone east of the North Island, New Zealand, occur within the geodetically-determined transition from locking to free-slip and vary over a wide range of depths (10 to 50 km). While globally, shallow slow slip events are rare, most (5 of 7) observed at the Hikurangi margin are at less than 20 km depth. Shallow slow slip events follow the same scaling laws for moment v. duration as deeper ones. We estimate the temperatures, calibrated by surface heat flow data, on the fault zone where the slow slip events occur. By this estimate, which includes heating from 20 MPa of shear stress, the temperature at the geodetic transition zone where SSEs occur in the north (near 10 - 15 km depth) is estimated to be ~100 to 150°C and in the south it is ~250 to 400°C where SSEs occur at 40 - 60 km depth. At other subductions zones the locked portion of the fault is typically in the 150 to 350°C range and the transition zone occurs in the 350 - 450°C range. Temperatures estimated for the northern Hikurangi subduction zone are much less than at most transition zones and alone cannot explain the positions of the slow slip events or the geodetic transition zone. Tomography results from the Hikurangi margin suggest that trapped fluids within the fault zone are a possible control.

T21A-0368 

Stable and transient motion on Kilauea's south flank from InSAR Persistent Scatterers

* Johanson, I A (ijohanson@usgs.gov), U.S. Geological Survey, 345 Middlefield Rd MS 977, Menlo Park, CA 94025, United States Poland, M P (mpoland@usgs.gov), U.S. Geological Survey - Hawaiian Volcano Observatory, P.O. Box 51, Hawaii National Park, HI 96718, United States Wicks, C (cwicks@usgs.gov), U.S. Geological Survey, 345 Middlefield Rd MS 977, Menlo Park, CA 94025, United States

Seaward sliding of the south flank of Kilauea volcano on the island of Hawaii occurs at a rapid rate (~10 cm/yr) and is associated with major earthquakes, such as a M7.2 in 1975. These events, as well as `interseismic' sliding, are inferred to occur on a decollement at the interface between oceanic sediments and the volcanic edifice at 7-9 km depth. Large landslide blocks, whose headwalls form the Hilina Pali and the Holei Pali, also accommodate seaward motion of the south flank, but whether they terminate at depth against the basal decollement or move independently in the shallow crust is unclear. Recently, several slow earthquakes (SEQs) have been observed by continuous GPS measurements on Kilauea's south flank. Evidence for the source depth of the SEQ events is ambiguous; slip may occur on the basal decollement or on the sliding plane of a landslide block at shallower depths. These possibilities are difficult to distinguish given the continuous GPS station geometry. InSAR provides an increase in the spatial density of measurements, though it is vulnerable to spurious signals produced by tropospheric water vapor. Using the Persistent Scatterer method can mitigate atmospheric error sources; allowing for a more precise determination of the deformation field than is possible from a single interferogram. We used the Stanford Method for Persistent Scatterers (StaMPS) to identify >250,000 persistent scatterers (PSs) from 28 descending-mode SAR scenes (27 interferograms) acquired between 2003 and 2007 by the European Space Agency's ENVISAT satellite. StaMPS isolates pixels that show consistently good spatial correlation with other PS candidates and then extracts phase change information from all 27 interferograms to produce a time history of deformation at the location of each PS. Atmospheric effects are mitigated by applying a spatially correlated noise (SCN) filter, which removes signals that are long-wavelength in space, but short- wavelength in time. The dataset captures the overall seaward motion of the south flank, as well as small-scale deformation features associated with the surface expressions of the Koa'e and Hilina Fault Zones. Evaluation of the time series suggests that the small-scale deformation features near the Hilina Fault Zone are unaffected by the slow earthquakes, implying that the most shallow part of the sliding plane is not involved in the SEQs. We will attempt to construct a 2D model of flank motion that includes slip on the basal decollement and the landslide blocks that effectively partitions slip between shallow, deep-seated, and transient sources.

T21A-0369 

Relationship Between Fault Creep and Shallow Stress Accumulation Rate

* Wei, M (mwei@ucsd.edu), Scripps Institution of Oceanography, 1102 IGPP, La Jolla, CA 93093-0225, United States Sandwell, D (dsandwell@ucsd.edu), Scripps Institution of Oceanography, 1102 IGPP, La Jolla, CA 93093-0225, United States Smith-Konter, B (Bridget.R.Konter@jpl.nasa.gov), Jet Propulsion Laboratory, California Institute of Technology, M/S 183-501, Pasadena, CA 91109, United States

Fault creep, which is non-seismic movement of a fault, is important for earthquake hazard assessment but the factors controlling fault creep are not well understood. Here we investigate the hypothesis that creep rate depends on the product of shallow stress accumulation rate and the creeping depth. Stress accumulation rate at 2 km depth along the San Andreas Fault System (SAFS) is estimated from a 3-D viscoelastic model constrained by GPS observations. The creeping depth is estimated with two approaches: i) assuming a uniform depth along the entire SAFS and ii) using a depth based on the balance between model stress and fault friction. The model with a uniform creeping depth shows a good correlation with observed creep rate (0-30 mm/yr) while the correlation is poor for the model where creep rate depends on accumulated stress.

T21A-0370 

Slow strain steps observed by two Ishii strainmeters within an M3 source area at a 2.9km depth, Mponeng gold mine, South Africa

* Yasutake, G (rp014036@se.ritsumei.ac.jp), Fac Sci Engr, Ritsumeikan U, Kusatsu, Japan, 525, Japan Ogawasara, H (ogasawar@se.ritsumei.ac.jp), Fac Sci Engr, Ritsumeikan U, Kusatsu, Japan, 525, Japan Kawakata, H), Fac Sci Engr, Ritsumeikan U, Kusatsu, Japan, 525, Japan Morishita, K), Fac Sci Engr, Ritsumeikan U, Kusatsu, Japan, 525, Japan Yamamoto, A), Fac Sci Engr, Ritsumeikan U, Kusatsu, Japan, 525, Japan Takeuchi, J), Fac Sci Engr, Ritsumeikan U, Kusatsu, Japan, 525, Japan Shimoda, N), Fac Sci Engr, Ritsumeikan U, Kusatsu, Japan, 525, Japan Naoi, M), U Tokyo, Tokyo, Japan, 113, Japan Nakatani, M), U Tokyo, Tokyo, Japan, 113, Japan Kato, A), U Tokyo, Tokyo, Japan, 113, Japan Ishii, H), Tono Res Inst Eq Sci, Mizunami, Japan, 509, Japan Nakao, S), Kagoshima U, Kagoshima, Japan, 890, Japan Otsuki, K), Tohoku U, Sendai, Japan, 980, Japan Yamauchi, T), Nagoya U, Nagoya, Japan, 464, Japan Iio, Y), Kyoto U, Uji, Japan, 611, Japan Carlsten, R), Mponeng mine, Western Levels, South Africa, 2501, South Africa McGill, R), Anglogold Ashanti Ltd., Jo'burg, South Africa, 2001, South Africa Tony, T), Seismogen CC., Carletonville, South Africa, 2500, South Africa Aswegen, G v), ISS International, Stellenbosch, South Africa, 7613, South Africa Mendecki, A J), ISS International, Stellenbosch, South Africa, 7613, South Africa Lenegan, P), ISS International, Stellenbosch, South Africa, 7613, South Africa Reserch Group SeeSA

The Research Group for Semi-controlled Earthquake-generation Experiments in South African gold mines (SeeSA) has attempted to observe the details of the source fault behavior near the Mw2-3 earthquake source area [e.g., Iio and Fukao, 1992]. At a depth of 2.9km in Mponeng gold mine, one of our observational sites, two Ishii strainmeters were installed near the fault where M3 earthquakes were anticipated [e.g., Ogasawara et al., 2005]. We analyzed the ~8-month strainmeter recordings. During this period, a high seismicity was induced by active mining, more than 5000 earthquakes (-3<Mw<3) being catalogued using mine seismic array within a 500³m³ volume around the strainmeters. In the Mponeng recordings, we found 6 steps with much longer durations than seen in normal earthquakes. In a preceding study, Naoi et al. [2006] found similar steps (slow-strain step) at the Bambanani mine site. The slow- steps found at both sites are possibly aseismic events because they had no corresponding catalogued earthquakes. 5 of 6 slow-steps at Mponeng were recorded with both strainmeters and they are synchronized with each other. Hereafter, we refer to the 5 slow-steps as SstM. The strain changes for one of SstM was on the order of 1E-6 and with a duration of ~8-minute. This is the largest, clearest slow-step ever recorded in SeeSA projects. Hereafter, we refer to this as SstML. The remaining 4 of 5 SstMs was on the order of 1E-7 and with duration of from 10-minutes to 1-hour. We successfully constrained the source distances and Mw of SstM, taking into accounts the ratio of magnitudes of strain changes and geometrical settings of strainmeters. In the result, the source of SstML was within ~3m from one strainmeter and Mw<-1.5. Mws for the other 4 SstMs were also constrained to be between -1.2 and 0.7. There have never been observed these small slow events. Ide et al. [2007] proposed a new scaling law to unify slow events that were recently found at subducting plate boundaries, including ETS, LFE, VFL, SSE and silent EQ. This relationship, that a seismic moment is proportional to their duration, clearly differs from that of regular earthquakes. On the other hand, SstMs follow another scaling relationship that a seismic moment is proportional to the cube of their duration. SstML were preceded by a clear forerunner linearly increasing with time, not accelerating with time. Hagiwara and Otsuki [2007] found similar phenomena in laboratory experiments, and the phenomena are thought to be seen during Riedel shear process. After the forerunner, SstML undergoes the fastest change followed by a logarithmic decay, with a sequence of smaller, but clear, shorter-steps.

T21A-0371 

Continuous GPS Monitoring of Deformation along the Xianshuihe Fault, Southwest China

* Shen, Z (zshen@ies.ac.cn), Institute of Geology, China Earthquake Admin., P.O.Box 9803 Qijiahuozi, Beijing, 100029, China * Shen, Z (zshen@ies.ac.cn), Dept of Earth and Space Sciences/UCLA, 595 Charles Young Dr, Los Angles, CA 90095- 1567, United States Wang, M), Institute of Earthquake Science, China Earthquake Admin., 62 Fuxing Rd, Beijing, 100036, China Gan, W), Institute of Geology, China Earthquake Admin., P.O.Box 9803 Qijiahuozi, Beijing, 100029, China Li, T), Institute of Geology, China Earthquake Admin., P.O.Box 9803 Qijiahuozi, Beijing, 100029, China Liao, H), Sichuan Seismological Bureau, Chengdu, Sichuan, 610041, China

The Xiaoshuihe fault is one of the most actively deforming faults in Chinese continent. Located in the eastern boundary of the Tibetan plateau, it slips more than 10 mm/yr left laterally, and plays an important role in transforming the north-south shortening of the plateau into east-southeastward extrusion. To understand its deformation pattern and mechanism we establish a small continuous GPS network along the Xianshuihe fault to monitor its spatio-temporal deformation behavior. The network is composed of two pairs of stations straddling the central and southern segment of the fault at Daofu and Tagong respectively. More than one and a half years of data have been collected, which are processed as baselines (17 km and 34 km in length respectively) across the fault. Analysis of the daily baseline solutions shows that they are accurate at about 1 mm or better, enough to detect some transient slip on fault. Displacement time series of the station pairs reveal markedly different deformation patterns across the two fault segments: steady across the south and variant in time across the central segment. Employing a dislocation model attributing the transient deformation to slip on fault segments in the brittle-ductile transition zone in central crust, we develop a spatio-temporal slip model to interpret the data. According to the model, the central segment of the Xianshuihe fault splits into east and west branches, which slip alternatively, and the slip on the east branch migrates gradually from south to north. The transition zone of the southern segment, on the other hand, slips steadily over the entire observation period. The difference in deformation style and process between the south and central segments should reflect the evolution status of the fault zone stress/strain, associated with the history of fault rupture and fault zone rheology. The difference in kinematics among fault segments and branches may suggest higher tectonic stress and greater seismic potentials for the south than the north segment, and for the east than the west branch of the central segment of the Xianshuihe fault.

T21A-0372 

Observation and Modeling of Seafloor Flow Rate Response to Slow Slip Deformation at the Costa Rica Forearc

* LaBonte, A L (alabonte@nrcan.gc.ca), Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla, CA 92093-0244, United States * LaBonte, A L (alabonte@nrcan.gc.ca), Neptune Canada University of Victoria, PO Box 1700 STN CSC, Victoria, BC V8W 2Y2, Canada * LaBonte, A L (alabonte@nrcan.gc.ca), Pacific Geoscience Centre, P.O. Box 6000, Sidney, BC V8L 4B2, Canada Fialko, Y (yfialko@ucsd.edu), Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla, CA 92093-0244, United States Brown, K M (kmbrown@ucsd.edu), Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla, CA 92093-0244, United States

Observations of transient flow through the seafloor during the 2000 Costa Rica Seismogenic Zone Experiment off the Nicoya peninsula suggests transient deformation events occurred in the shallow subduction zone, around 10 km arcward from the Middle America Trench. The observed deformation in this region is unexpected, as it is generally believed that accommodation of plate convergence in the shallow plate interface is through stable sliding. This study is an investigation of the type and extent of deformation events that could cause these flow transients. Fluid flow rates in response to displacement on a fault in a porous media are numerically calculated using a fully- coupled poroelastic finite element model. Modeled flow rates at the surface of a half-space, the seafloor, are intended to assist in interpretation of records from seafloor flow meter instrumentation. Model results show seafloor fluxes are a superposition of two effects causing volumetric strain at the surface: 1) compressional and dilational regions that radiate out from the fault tips, and 2) extensional and compressional bending of the free- boundary surface. Solutions of the spatial and temporal flow rate response to sudden slip along the decollement of a subduction zone are presented for near-field to far-field ruptures in a homogeneous crust and a heterogeneous crust with a layer of sediment overlying the basement. The characteristic patterns in flow rate for near-field and far-field ruptures in homogeneous and heterogeneous crust help to determine optimal instrument placement for future subduction zone studies. In addition, comparisons of the fully-coupled half-space model simulations for a homogeneous verses a heterogeneous crust demonstrates situations where it is safe to use existing analytical solutions for finite ruptures in a homogeneous poroelastic half-space instead of finite element modeling. Novel simulations of updip and downdip propagating ruptures in this study demonstrate unique temporal records of flow rate through the seafloor surface. These propagating rupture model results are applied to help constrain rupture characteristics of a flow event recorded on the toe of the Costa Rica prism. The observed flow rate time series is nicely reproduced with a downdip propagating rupture centered below a flow meter instrument 6.5 km arcward from the trench. The observed variability in flow rates recorded at 2 instruments located at along-strike distances of 15 and 30 km could be explained by a single bilaterally propagating event that has an along-strike variation in rupture initiation. This is the first result suggesting episodic slow slip may initiate near the trench and possibly independent of a triggering event further downdip. A shallow slow release of stored energy is also the likely cause for shallow propagation during the nearby 1992 Nicaragua tsunamogenic earthquake. This proposed slow slip event in the frontal Nicoya prism, a section where subducted clay-rich sediments along the decollement are saturated and hydrous, poses important questions concerning processes controlling temporal variation of fault mechanics in the ‘stable-sliding' prism toe. How does stress accumulate at the shallow plate interface? Perhaps the topography associated with the normal faulting of the subducting basement prevents stable sliding from occurring until a critical stress threshold is overcome.

T21A-0373 

Investigating Slow Slip: Development of an Adjoint-Based, Time-Dependent Inversion Scheme

* Tandon, K (kush@coas.oregonstate.edu), College of Oceanic & Atmospherics Sciences, 104 COAS Admin. Bldg. Oregon State University, Corvallis, OR 97331, United States Egbert, G (egbert@coas.oregonstate.edu), College of Oceanic & Atmospherics Sciences, 104 COAS Admin. Bldg. Oregon State University, Corvallis, OR 97331, United States

Investigating the nature of slow slip deep beneath the earth in different parts of the world is beset with multiple uncertainties, for example, noisy and incomplete data, sparse spatial and temporal data sampling with regards to various geodetic datasets, flawed parameterization, and incomplete physics and geological description of the earth processes involved. The same sources for error hold for any study of short-term, active tectonics problems. As a part of IOM (Inverse Ocean Modeling) initiative, a variational data assimilation scheme developed for ocean modeling, we added tools for active tectonic problems as one of its new client. The IOM requires the client model to provide a tangent-linear (TL) of the non-linear dynamic model and an adjoint (ADJ) of the TL in a gradient- based search algorithm. Our client model is GeoFEST (the Geophysical Finite Element Simulation Tool), a 3D visco-elastic finite element forward model that is also a part of QuakeSim, a NASA Earth Science Enterprise Project simulation toolbox. GeoFEST is suited for modeling broad class of active tectonics problems, including slow slip. As a client for IOM, we developed TL and ADJ for GeoFEST. The IOM is a modular implementation of a variational data assimilation scheme that allows for errors in data, the model, adjoint forcing, and initial and boundary conditions. Here, we present this approach as an adjoint-based, time-dependent inversion for slow slip analysis using both IOM and GeoFEST together. Our adjoint-based, time-dependent inversion scheme here for slow earthquakes is one possible initial application from a broad-based tools developed by us for variational data assimilation in active tectonics. We map the spatial sub-surface forcing variations, both in strike-slip and dip-slip in a localized region, and the time history for slow slip using CGPS (Continuous GPS) data. We explore the issue of resolution in inverting for forcing causing the slow slip events and optimal array design using simple models and synthetic CGPS datasets. Some specific tests presented are: 1) Resolvability of the depth of the localization of slow slip, and accurate time detection in presence of different signal-to-noise ratio, and other limitations in the data. 2) Can our technique resolve multiple sources for slow slip within the regional fault or faults from the GPS data? 3) What array design, precision, and quality of CGPS stations is needed to accurately observe and analyze different slow slips? To contribute toward future development of new techniques, our software tools could be used to assist in the design of the ongoing and future observatory experiments in different tectonic settings, such as EARTHSCOPE, HiNET, etc., and also possibly map and predict the migratory pattern of the slow slips in real data.

T21A-0374 

Episodic Slow-Slip Transients and Rate-and-State Friction

* Rubin, A M (arubin@princeton.edu), Department of Geosciences, Princeton University, Princeton, NJ 08540, United States Segall, P (segall@stanford.edu), Department of Geophsyics, Stanford University, Stanford, CA 94305, United States

Numerical simulations dating back to Horowitz and Ruina [JGR 1989] suggest that aseismic slip transients may result if the velocity-weakening portion of a fault is large enough to nucleate an event but too small for that event to reach instability. Models of rate-and-state friction can credibly explain both the large dimensions of slow slip events in subduction zones and their propagation velocity, provided the effective normal stress is very low (of order 1 MPa). However, both standard laws for the evolution of fault "state" fail to explain other first-order features of these events. For lab-like values of a/b, where a and b are the coefficients of the velocity- and state-dependence of the frictional strength, the fault length needed for instability depends upon the effective fracture energy of the expanding nucleation zone. For the "aging" evolution law this fracture energy increases as the square of the logarithm of the velocity excursion, and the range of fault lengths hosting aseismic transients grows in rough proportion to (1-a/b)^-1. Near neutral stability this range is quite large. However, existing lab data strongly favor the "slip" evolution law as a predictor of nucleation style. For this law the effective fracture energy increases only as the logarithm of the velocity excursion, and the range of fault lengths capable of hosting slow slip events appears to be too small to explain why these events are so common. Even for the aging law, the range of fault lengths generating slip speeds reaching those inferred geodetically (~10-100 times the plate rate) is very small. The reason is that over most of parameter space where slow slip events are produced, the bulk of the increasing mechanical energy needed to grow the nucleation zone is supplied by the steady plate motion down- dip. The resulting slip speed cannot greatly exceed the plate rate. For steady-state friction that decreases as the logarithm of the slip speed, stable slip speeds significantly exceeding the driving rate can be reached over a substantial range of fault lengths if the fracture energy increases more rapidly than the square of the logarithm of the velocity excursion. An appealing mechanism is inelastic dilation of the fault zone coupled with pore pressure reduction, which becomes more effective at low effective stress. This is demonstrated in the companion presentation using a simplified representation of pore-pressure diffusion.

T21A-0375 

Explaining postseismic and aseismic transient deformation in subduction zones with rate and state friction modeling constrained by lab and geodetic observations

* Liu, Y (yjliu@princeton.edu), Dept. Geosci., Princeton Univ., 308A Guyot Hall, Princeton, NJ 08544, United States DeDontney, N L (ndedontn@fas.harvard.edu), Dept. Earth Planet. Sci., Harvard Univ., 20 Oxford St., Cambrdige, MA 02138, United States Rice, J R (rice@esag.harvard.edu), Dept. Earth Planet. Sci. and Sch. Eng. Appl. Sci., Harvard Univ., 29 Oxford St., Cambridge, MA 02138, United States

Rate and state friction, as applied to modeling subduction earthquake sequences, routinely predicts postseismic slip. It also predicts spontaneous aseismic slip transients, at least when pore pressure p is highly elevated near and downdip from the stability transition [Liu and Rice, 2007]. Here we address how to make such postseismic and transient predictions more fully compatible with geophysical observations. For example, lab observations can determine the a, b parameters and state evolution slip L of rate and state friction as functions of lithology and temperature and, with aid of a structural and thermal model of the subduction zone, as functions of downdip distance. Geodetic observations constrain interseismic, postseismic and aseismic transient deformations, which are controlled in the modeling by the distributions of a \barσ and b \barσ (parameters which also partly control the seismic rupture phase), where \barσ = σ - p. Elevated p, controlled by tectonic compression and dehydration, may be constrained by petrologic and seismic observations. The amount of deformation and downdip extent of the slipping zone associated with the spontaneous quasi- periodic transients, as thus far modeled [Liu and Rice, 2007], is generally smaller than that observed during episodes of slow slip events in northern Cascadia and SW Japan subduction zones. However, the modeling was based on lab data for granite gouge under hydrothermal conditions because data is most complete for that case. We here report modeling based on lab data on dry granite gouge [Stesky, 1975; Lockner et al., 1986], involving no or lessened chemical interaction with water and hence being a possibly closer analog to dehydrated oceanic crust, and limited data on gabbro gouge [He et al., 2007], an expected lithology. Both data sets show a much less rapid increase of a-b with temperature above the stability transition (~ 350 ^oC) than does wet granite gouge; a-b increases to ~ 0.08 for wet granite at 600 ^oC, but to only ~ 0.01 in the dry granite and gabbro cases. We find that the lessened high-T a - b does, for the same \barσ, modestly extend the transient slip episodes further downdip, although a majority of slip is still contributed near and in the updip rate-weakening region. However, postseismic slip, for the same \barσ, propagates much further downdip into the rate-strengthening region. To better constrain the downdip distribution of (a - b) \barσ, and possibly a \barσ and L, we focus on the geodetically constrained [Hutton et al., 2001] space-time distribution of postseismic slip for the 1995 Mw = 8.0 Colima-Jalisco earthquake. This is a similarly shallow dipping subduction zone with a thermal profile [Currie et al., 2001] comparable to those that have thus far been shown to exhibit aseismic transients and non-volcanic tremor [Peacock et al., 2002]. We extrapolate the modeled 2-D postseismic slip, following a thrust earthquake with a coseismic slip similar to the 1995 event, to a spatial-temporal 3-D distribution. Surface deformation due to such slips on the thrust fault in an elastic half space is calculated and compared to that observed at western Mexico GPS stations, to constrain the above depth-variable model parameters.

T21A-0376 

Precursors to Great Earthquakes Along the Nankai Trough; Slow Earthquakes, Non-volcanic Deep Tremor and Slow Slip

* Sacks, S (sacks@dtm.ciw.edu), Department of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad Branch Road NW, Washington, DC 20015, United States Linde, A T (linde@dtm.ciw.edu), Department of Terrestrial Magnetism, Carnegie Institution of Washington, 5241 Broad Branch Road NW, Washington, DC 20015, United States

Just before (and during) the December 1944 Tonankai great earthquake, the Military Survey Institute of Japan carried out leveling surveys in the anticipated region of the earthquake. Mogi (1985) pointed out that large closure errors, for line segments measured on the day before and the morning of the earthquake, could be indicative of continuing tilting due to precursory slip. Before the 1946 Nankaido great earthquake there were level changes recorded by tide gauges and also large changes in water levels in wells. We have shown (Linde and Sacks 2002) that the pre-earthquake changes for both events are indicative of slow slip on the down-dip extension of the seismogenic zone, a region that can store strain energy but fails with slow slip. The coseismic slip for both earthquakes averages about 4 meters; the down-dip slow slip was determined to be about half the seismogenic value. In his most recent studies of the same area, Obara (2006) reports that small (~cm) slow slip events and non-volcanic tremor occur on the upper surface of the subducting plate. The locations for these events correspond rather closely to the areas we proposed as having slow slip precursory to the great earthquakes. Additionally the slip rate from Obara's work would result in about 2 meters being released in 100 years, the approximate return interval for the great earthquakes. This is consistent with the deficit being released as a large slow event just before those great earthquakes.

Author(s) (2007), Title, Eos Trans. AGU, 88(52), Fall Meet. Suppl., Abstract #####-##.