U31B-01 INVITED
Cascadia Episodic Tremor and Slip events observed on GPS, seismic, and strain/tiltmeter arrays
The rapidly expanding GPS networks along the greater Cascadia forearc have enabled identification of nearly 40 isolated Episodic Tremor and Slip (ETS) events from 1992 through 2008. ETS events are observed throughout the forearc, from northern California to southwestern British Columbia, with station density generally increasing towards the north. Events located in well-instrumented regions can be tracked as they migrate laterally north-south along the plate boundary, but increasing station density has resolved many smaller transients that could not previously be confidently identified. At the specific latitude of the northern Washington State and southwestern British Columbia, the 14-month average recurrence interval still holds true, 6 events after first recognition. Elsewhere, this periodicity is not observed. Along central Oregon, an 18-month recurrence is evident, while in northern California (Yreka) the 11-month periodicity previously documented still holds true. Sporadic smaller events appear frequently throughout the subduction zone, including within the region known for the 14-month periodicity. For the most recent events that have the best GPS and seismic coverage, there is an increasingly strong correspondence between GPS-inversions for slip and tremor epicentral locations. Moreover, the 2007 and 2008 ETS events were also recorded by the PBO borehole strain and PANGA long-baseline tiltmeter arrays. These instruments show up to 200 nr of induced tilts and strain signals whose evolution are temporally coincident with tremor and whose total amplitudes are consistent with that predicted by GPS inversions for slip. GPS offsets for the largest 23 events inverted for slip show moment magnitudes ranging from 6.3 (smallest resolvable with GPS) to 6.8, and typically 2-3 cm of slip. The largest spatial extent of the events resolved to date is just under 500 km along strike, and maximum duration is seven weeks, which lies in marked contrast to other subduction zones. Averaged over many ETS events, the upper limit of transient slip in the vicinity of Seattle, WA lies just west of the heavily urbanized Puget Sound region, suggesting that the lower limit of megathrust seismic rupture may extend much closer to this area than previously thought. A comparison of GPS with tremor analyses of 23 well-recorded events over a ten-year period yields a highly linear relationship between moment release, as estimated from GPS, and total duration of non-volcanic tremor, as summed from regional seismic arrays. All Cascadia events detected since 1997 for which seismic data is available, which collectively span the Cascadia arc from northern California to Vancouver Island, Canada, release moment during tremor at a rate of 5.3 ± 0.38 x 1023 dyne-cm per hour of recorded tremor. This empirical tremor magnitude scale enables continuous estimation of moment dissipation via tremor monitoring along the deeper Cascadia subduction zone that poses the greatest threat to its major metropolitan centers.
U31B-02 INVITED
Northern Cascadia Episodic Tremor and Slip: A Decade of Observations from 1997 to 2007
We analyze continuous seismic and GPS records collected in the last decade (1997-2007) to establish the most comprehensive observational basis for northern Cascadia episodic tremor and slip (ETS) events. A simple ˇ§ETS scaleˇ¨ system, using a combination of a letter and a digit, is proposed to quantitatively characterize the spatial and temporal dimensions of ETS events. Clear correlation between GPS and tremor signals is observed for all major episodes with lateral dimension >150 km (i.e., A- or B-class), but the GPS signature is less obvious for minor ones. Regular ETS recurrence can be established only for A-/B-class episodes in southern Vancouver Island. Halting and jumping are very common in ETS migration patterns, and along-strike migration can happen in both directions. A prominent tremor gap is observed in mid island around 49.5N. This gap coincides with the epicenters of the only two large documented crustal earthquakes in the region. ETS tremors also tend to occur in places where the local seismicity is relatively sparse. The tremor depth distribution shows a peak in the 25-35 km range where strong seismic reflectors (i.e., the E- layer) are documented. Existence of tremors in the vicinity of E-layer is also confirmed by an independent waveform analysis. More significantly, we have found a few very-low-frequency earthquakes (VLFE) at the depth of E-layer showing low-angle thrust faulting mechanisms. Our results suggest that a significant portion of the tremor activity and perhaps associated shearing are taking place along well-developed structures such as the E-layer, while a reduced number of tremor bursts are generated elsewhere in response to induced stress variation throughout the source volume.
U31B-03 INVITED
Are slow slip events more than the cumulative sum of slip in Tremor?
Following the discovery of deep low-frequency tremor, a diverse range of other, related phenomena were found to occur in western Japan. These include: low frequency earthquakes (LFEs), very low-frequency earthquakes (VLFs), and slow slip events (SSEs). We study these unusual earthquakes mainly in western Shikoku where seismic network coverage is good and tremor is most active. Among these phenomena, LFEs have the most impulsive nature, which allows us to apply developed seismological methods to them. Double difference tomography and event location revealed that LFEs are located on the subducting plate interface. P-wave first motion analysis and empirical Green tensor inversion both show that low-angle thrust faulting is the LFE source mechanism. All these results point to a model of LFEs as tiny, somewhat slow, slip events on the plate interface. Deep tremor waveforms match LFE waveforms, which in turn indicates that deep tremor occurs as a swarm of slip events. VLFs and SSEs have also been analyzed, independently, and found to arise from low-angle thrust slip. Therefore all these unusual phenomena share the same low angle thrust mechanism, which is consistent with overall plate motion and the most recent Nankai megathrust earthquake. The seismic moment and event duration of these phenomena are proportional, with a characteristic moment rate of 10**12-13 Nm/s. Taken together, this scale invariance and proximity of location and time, suggests that these phenomena are different manifestations of a single slow earthquake process. These slow earthquakes were identified as separate phenomena due to a combination of the scaling law and limitation on observations due to the interference of microseismic and tidal noise. In favorable situations where S/N is high, however, we can find slower, and larger, events than VLFs, that have a duration 20-200s, and satisfy the same scaling law. These events radiate seismic energy in direct proportion to their seismic moment rate, which implies that scaled energy of slow earthquakes is constant, about 10**-10. These observations, together with other features such as tremor source migrations and tremor amplitude statistics, can be explained by a simple Brownian walk model, in which short-period random fluctuation and long-term average motion represent tremor/LFE and SSE, respectively. In this sense, SSE is accountable by slip in tremor if the detection is perfect.
U31B-04
Recurring slow slip on the Hikurangi subduction interface, New Zealand
A long-duration slow slip event has been occurring since late 2007 at 30-50 km depth on the Hikurangi subduction interface, near the bottom of the zone of inferred high interseismic coupling beneath the southern North Island, New Zealand. The event is centered about 50 km northwest of Wellington, New Zealand's capital city, and has been recorded by more than 10 continuous GPS stations. The observed surface deformation was most rapid from December 2007 through February 2008, but is still continuing in September 2008. The maximum slip in the event has been more than 200 mm, representing about 6 years of accumulated plate motion. The event appears in part to have re-ruptured the area that slipped during a previous 1-year duration slow slip event in 2003-04, though the earlier event was poorly constrained. Slip of 200 mm in the 2007-08 slow slip event is somewhat larger than the amount of plate motion that has accumulated in the time since the 2003-04 event. We have observed a series of short-duration (1-2 week) slow slip events since 2002 at 10-15 km depth in the Gisborne region of the northeastern North Island, also near the bottom of the zone of inferred high interseismic coupling. These occur about every two years, and their sizes imply that all the plate motion in this depth range is taken up by the slow slip events. The most recent Gisborne slow slip event appears to have been triggered by a nearby Mw 6.7 normal-faulting earthquake within the subducted slab.
U31B-05
Nonvolcanic Tremor Activity is Highly Correlated With Slow Slip Events, Mexico
Significant activity of nonvolcanic tremor (NVT) has been observed in the central Mexico (Guerrero) subduction zone since 2001 when continuous seismic records became available. Although the quality of these records is poor, it is possible to estimate a temporal variation of energy in the range of 1-2Hz (best signal/noise ratio for the NVT). These clearly indicate a maximum of NVT energy release (En) during the 2001-2002 and 2006 large aseismic slow slip events (SSE) registered by the Guerrero GPS network. In particular En is higher for the 2001-2002 SSE which had larger surface displacements and extension than the 2006 SSE. A more detailed and accurate study of NVT activity was carried out using the data collected during the MASE experiment in Mexico. MASE consisted of 100 broad band seismometers in operation for ~2.5 years (2005-2007) along the profile oriented SSW-NNE from Acapulco, and crossing over the subduction zone for a distance of ~500 km. Epicenters and depths of individual tremor events determined using the envelope cross-correlation technique have rather large uncertainties, partly originated from the essentially 2D geometry of the network. The 'energy' approach is more efficient in this case because it provides an average NVT activity evolution in time and space. The data processing consists of a band pass (1-2Hz) filter of the raw 100 Hz sampled N-S component records, application a 10 min-width median filter to eliminate the effect of local seismic events and noise, and integration of the energy and normalization of daily En using an average coda amplitude from several regional earthquakes of M~5. A time-space distribution of En reveals a strong correlation between NVT energy release and the 2006 SSE, which also replicates the two-phase character of this slow event and a migration of the slow slip maximum from North to South. There are also a few clear episodes of relatively high NVT energy release that do not correspond to any significant geodetic signal in GPS time series. More detailed analysis of both the seismic and GPS data will probably answer these questions: Do SSE and NVT represent manifestations of the same physical process (episodic tremor and slip (ETS)) in Mexico? Or are those two phenomena only space-time correlated?
U31B-06
Slow Slip and Tremor Detected at the Northern Costa Rica Seismogenic Zone
A slow slip event accompanied by seismic tremor was recorded in May 2007 on a GPS and seismic network installed on the Nicoya Peninsula, Costa Rica. A maximum of about 1 cm of surface displacement was observed over a period of about 30 days, corresponding to a maximum of 9 cm of slip on the plate boundary. A constrained inversion using a simple dislocation model and the limited available data indicates a slip patch in a region of abundant microseismicity, down-dip from the shallow locked patch observed by earlier campaign GPS data. Seismic tremor was observed near the start of this event, but not throughout its ~thirty day period. Tremor bursts located using an envelope cross-correlation and earthquake location method have poor depth resolution but appear to overlap the region of GPS determined maximum slip and streak in the dip direction with a slight downdip time progression. An important aspect of Costa Rica tremor that differentiates it from tremor occurring in other locations is its relatively short extent compared with the duration of the slow slip event and its apparent location within the seismogenic zone. Tremor in both Cascadia and SW Japan is reported as typically occurring 75% of the time during slow slip events and at the down dip edge of the seismogenic zone. Tremor duration in Costa Rica rarely exceeded 10% of the time during the 2007 slow slip event and most tremor events appear to be located within the seismogenic zone. While we do not as yet have an explanation for these variations, we note that they may represent a fundamental difference in tremor activity between subduction zones with different characteristics. Both the Cascadia and SW Japan seismogenic zones are relatively hot and strongly locked with little or no interplate seismicity. In contrast, the northern Costa Rica seismogenic zone is much cooler and has abundant interplate seismicity.
U31B-07
Precise Relative Location of San Andreas Fault Tremors Near Cholame, CA, Using Seismometer Clusters: Slip on the Deep Extension of the Fault?
Non-volcanic tremor, similar in character to that generated at some subduction zones, was recently identified beneath the strike-slip San Andreas Fault (SAF) in central California (Nadeau and Dolenc, 2005). Using a matched filter method, we closely examine a 24-hour period of active SAF tremor and show that, like tremor in the Nankai Trough subduction zone, this tremor is composed of repeated similar events. We take advantage of this similarity to locate detected similar events relative to several chosen events. While low signal-to-noise makes location challenging, we compensate for this by estimating event-pair differential times at 'clusters' of nearby temporary and permanent stations rather than at single stations. We find that the relative locations consistently form a near-linear structure in map view, striking parallel to the surface trace of the SAF. Therefore, we suggest that at least a portion of the tremor occurs on the deep extension of the fault, similar to the situation for subduction zone tremor. Also notable is the small depth range (a few hundred meters or less) of many of the located tremors, a feature possibly analogous to earthquake streaks observed on the shallower portion of the fault. The close alignment of the tremor with the SAF slip orientation suggests a shear slip mechanism, as has been argued for subduction tremor. At times, we observe a clear migration of the tremor source along the fault, at rates of 15-40 km/hr.
U31B-08
What does tremor really look like? Initial results from an 84-element array
Aspiring to see more intimate details, we placed an 84-element short-period vertical-component array with an aperture of 1km on a hard rock mountain over the path of Cascadia tremor. This site is coincident with a stellar 6-station three-component CAFE array (see talk by K. Creager). Texans, which are convenient to deploy but require recycling for fresh batteries every four days, recorded the seismograms. We recorded 8 days in March and 17 days in May 2008. We find most of the arrivals at high frequencies, especially in the stacks, are P-waves, due to the network constitution. The March week contains only six intermittent hours of tremor detectable by the usual envelope analysis of data from the regional network, but array beamforming shows much more continuous activity, and extending about a half day longer. We also pick up a later episode of weak tremor that contains probably the first glance of low-frequency earthquake in Cascadia (see abstract by J. Sweet). The May field season recorded full-blown tremor passing directly underneath in startling detail. The tremor source region in preliminary images is more compact than the cloud of locations determined from envelope correlation, but also with an apparently persistent patchwork of regions that do and do not generate tremor. Further analysis and future deployments with multiple dense arrays show great promise for getting to the bottom of the issue of tremor generation.