T12C-01
Non-volcanic Tremor and Earthquakes Driven by the Large Transient Shear Stresses of the 2002 Denali Earthquake
We identify bursts of tremor that radiated from the Cascadia subduction zone near Vancouver Island, Canada, during the strongest shaking from the moment magnitude Mw = 7.8, 2002 Denali, Alaska earthquake. Tremor occurs when the Love wave displacements are to the southwest (the direction of plate convergence of the overriding plate) strongly suggesting that the Love waves trigger the tremor. We show that these displacements correspond to shear stresses of approximately 40 kPa on the plate interface, which suggests that the resistance to sliding on the plate interface is very low. These observations indicate that tremor and possibly slow slip can be instantaneously induced by shear stress increases without changes in normal stress on the subduction interface, effectively a frictional failure response to the driving stress. In addition to the Denali earthquake, we found 3 other teleseisms that triggered tremor on Vancouver Island and 25 that did not. As one would expect, larger amplitude waves appear necessary to trigger non-volcanic tremor, but other conditions also appear important. In particular, the timing of the triggering waves may be important, as tremor is more likely to be triggered shortly after an ETS event than at other times. We also find evidence that distant earthquakes may trigger ETS. Specifically, the ETS episodes with the longest lapse times since the previous ETS all occur within a few days of large teleseismic events, which is not the case for ETS episodes with more typical lapse times. The Denali earthquake also triggered small earthquakes on Vancouver Island at locations far from the triggered tremor. None of the other large teleseisms we examined produced a clear, immediate increase in seismicity rate, but a few others show possible increases. http://earthweb.ess.washington.edu/~justin/trigtrem.html
T12C-02
Widespread Triggered Non-volcanic Tremor Along the California Transform Plate Boundary
The recent discovery of non-volcanic tremor has led to great excitement and research activity in the earth-science community. To date, tremor has almost exclusively been found in subduction zones. The only observation outside a subduction-dominated region is on the strike-slip San Andreas Fault in Parkfield, California. In Japan and Cascadia non-volcanic tremor lasting days to months occurs coincident with aseismic slip along the plate interface, and recent studies have identified short bursts of tremor triggered by the seismic waves of distant earthquakes. Since non-volcanic tremor is mostly observed in subduction zones, nearly all causative mechanisms proposed appeal to conditions expected in the subduction zones. We show that the conditions required for non-volcanic tremor generation must exist in a variety of tectonic environments, by presenting observations of non-volcanic tremor at seven sites along the transform plate boundary in California triggered by the 2002 Mw 7.8 Denali Fault, Alaska earthquake. These observations come from recordings of the Denali earthquake waves from broadband and short period seismometers archived at the Northern and the Southern California Earthquake Data Centers. Some models of non-volcanic tremor in subduction zones invoke frictional behaviors expected in regions transitional between where the fault is locked and slipping freely (creeping). Such transition zones likely occur in California at shallow depths below and adjacent to documented creeping fault segments. We find no clear correlation of tremor locations with creeping, locked, or transitional fault behavior. Other explanatory models relate non-volcanic tremor to the release of fluids from dehydration of the subducting plate. We find numerous triggered earthquakes in two geothermal fields but no tremor, implying that high fluid pressure and/or temperatures may be necessary for tremor generation, but they are not alone sufficient.
T12C-03
Teleseismically-induced tremor near Parkfield, CA - a cacophony or a symphony?
The tremor triggered near Parkfield, CA by the 2002 Denali and 2004 Sumatra earthquakes was strong and well recorded by the dense regional CISN and the borehole HRSN networks. Peng et al. (this meeting) survey tremors triggered by a larger set of 12 regional and teleseismic events, providing a broader context. In the case of both the 2002 M7.9 Denali and 2004 M9.1 Sumatra earthquakes, the tremor emanates from at least two source regions deep within the SAF. The first source region is 40 km NW of the SAFOD in the creeping section of the SAF, and the second region is 40 km SE of the SAFOD near Cholame, close to the location where most of the non-triggered tremor has been found previously (Nadeau and Dolenc, Science, 2005). The Denali earthquake triggered tremor is in phase with the surface waves for about 400s. The northern region started tremoring first by about 100s, and both regions quieted before the end of the surface waves. The wavetrain for the 2004 M9.1 Sumatra earthquake was long enough that tremors were also excited by the weak diffracted P waves, and tremor turned up the volume for an hour upon the arrival of the surface waves, underwent a sudden and curious hiatus for 500s before the end of the surface waves, then re-started and continued for at least an hour after the passage of the surface waves. It is easy to suggest that the tremor was accompanied by deep slip on the SAF, but creep and strain data indicate any slip was too small to generate a detectable surface deformation. These observations suggest a component of driven, instantaneous, perhaps Coulomb-friction response with an added dose of self-sustaining, dribbling activity more suggestive of the oozing of fluids.
T12C-04
Multi-scale Quasi-periodic Rate Changes of Nonvolcanic Tremor at Cholame, CA Following the 2004 Parkfield Mainshock.
In Cascadia and Japan nonvolcanic tremors (NVTs) have shown a remarkable correlation between their activity rates and GPS and tiltmeter measurements of transient deformation in the deep (sub-seismogenic) fault zone. There is also evidence to suggest that stress changes from passing surface waves of large earthquakes may induce NVT activity. These observations suggest that tremor rate changes may be closely related to transient deformation and/or stress changes in the deep fault zone, and that induced stress changes from near-by larger earthquakes could stimulate tremor activity. Here we present observations of the evolution of NVT activity rates in the Cholame, CA area over a multiple year period ending in late 2007 and spanning the occurrence of the 28 September 2004 Parkfield magnitude 6 earthquake. We show: 1) that NVT rates increased significantly following the 2004 Parkfield event, 2) that immediately following the mainshock, rates were highest and then decayed rapidly for several weeks, 3) that following the rapid decay period, overall rates decay much more slowly, and 4) that a pattern of multi-scale quasi-periodic variation in the rates has evoluted over the three year period since the Parkfield mainshock. These observations support the argument that near-by moderate magnitude earthquakes can stimulate deep NVT activity and that such events may have a significant impact on the longer term evolution of near-by NVT activity.
T12C-05 INVITED
Occurrence Of Deep Low-frequency Tremors Synchronized To Earth Tides
Swarms of deep low-frequency tremors occur with slow slip events along the subduction zone of the Philippine Sea plate in southwest Japan. These episodic events are considered to be linked in a stress relaxation process in the transition zone of the subducting plate interface. Tremor swarms often exhibit occurrence with a period of about 12 or 24 hours. Although it is easy to surmise that such periodicity arises from effects of the Earth tides caused by gravitational perturbations from the Moon and Sun, it has not yet been given a quantitative explanation. Here we show that the observed periodic tremor occurrence can be reproduced by seismicity rate change calculated from the periodic stress due to the Earth tides combined with the transient background stress due to a slow slip event. The result shows that although tremor source regions are always being perturbed by the periodic tidal stress, tremor swarms are not triggered until the additional transient stress is imposed. The seismicity rate theory used here is based on the rate- and state-dependent friction law. The value of Aσ, the fault constitutive parameter times the effective normal stress, is on the order of 1 kPa. This is an order of magnitude smaller than that for ordinary earthquakes, indicating a quick and sensitive response of tremor occurrence to stress changes. Observing deep low-frequency tremors is therefore effective for monitoring the stress relaxation process in the transition zone.
T12C-06
Strong Tidal Modulation of Non-Volcanic Tremor in Cascadia
The July 2004, September 2005 and January 2007 episodic tremor and slip events in the Puget Sound/Vancouver Island ETS source region, were exceptionally well recorded by temporary deployments of small-aperture seismic arrays. Analysis of stacked, filtered envelope functions across these arrays shows a clear pulsing of tremor activity with strong, isolated peaks at periods of 12.4 and 24-25 hours, coincident with the periods of the principle lunar and lunisolar tides. The amplitudes of these peaks rise above noise levels, measured at times without active tremor, by factors of 2 to 20. During each ETS episode the amplitude of tremor is modulated by ±30% at the 12.4-hour period. The strong dependence of tremor amplitude on the tides indicates that the small stresses associated with the solid earth and ocean tides influence the genesis of tremor much more than they do "normal" earthquakes at comparable depths. Because variations in calculated tidal stresses are approximately 105 times smaller than the lithostatic loads compressing deep faults, tremor likely occurs on very low-stress faults, perhaps in the presence of nearly lithostatic pore-fluid pressures. Preliminary modeling of variations in shear and normal stresses acting on the interface between the subducting Juan de Fuca and the overriding North American plates, indicates that the most important contributions to these stresses are from water tides in the inland waterways of the Straits of Juan de Fuca, Puget Sound and Georgia Straits, while the Pacific Ocean and solid earth tides are generally less important. Each of these ETS events had durations of 2-3 weeks and occurred in regions near inland waterways that produce tidal variations in stress.
T12C-07
Quasi-Periodic Slow Earthquakes and Their Association With Magmatic Activity at Kilauea Volcano, Hawai`i
Since 1998 the mobile south flank of Kilauea volcano, Hawai`i, has been the site of multiple slow earthquake (SE) events recorded principally with continuous GPS. One spatially coincident family of these SEs exhibited a high degree of periodicity (774 +/- 7 days) from 1998 to 2005 suggesting the next SE would be in mid-March, 2007. In fact, no anomalous deformation occurred there until the June 17 Father's day dike intrusion that caused up to 1m of opening along Kilauea's east rift zone. We analyzed deformation related to the Father's day event using GPS, tilt, ALOS and Envisat interferometry, microseismicity, and elastic dislocation modeling. Our analysis reveals significant motions of far-field sites that cannot be explained by dike-related deformation and that are very similar to previous SE displacements of the same sites, strongly suggesting that a SE occurred. Inclusion of this event in the overall time series yields SE repeat times of 798 +/- 50 days, apparently maintaining the quasi- periodicity of the Kilauea events. Furthermore, the timing of dike- and SE-related deformation and stress modeling suggest the Father's day dike triggered the slow earthquake. We explore the connection between magmatism and SEs at Kilauea and find a potential correlation between SE-timing and eruptive activity since 2000. This suggests the possibility that a mechanistic understanding of Kilauea SEs may require consideration of magmatic processes in addition to fault zone processes.
T12C-08
Detecting Aseismic Fault Slip and Magmatic Intrusion From Seismicity Data
Seismicity triggered by aseismic deformation, such as magmatic intrusions or afterslip, can be used to detect the occurrence of these otherwise difficult to observe processes. Recent studies suggest that aseismic deformation can trigger large amounts of seismicity in a variety of plate tectonic settings. We have developed a new technique that takes advantage of this triggered seismicity to estimate the time-history of aseismic stressing rate on a fault- zone by combining the rate and state dependent friction and the Epidemic Type Aftershock Sequence (ETAS) models of seismicity-rate [ Dieterich, 1994; Ogata, 1988]. In the rate-state model, the integration of an observed seismicity rate results in an estimate of the stress rate acting in a given space-time window. However, the seismicity rate observed in any catalog comes from 3 primary sources: coseismically-triggered seismicity (aftershocks), tectonically-triggered seismicity (i.e., from long-term tectonic loading), and aseismically-triggered seismicity (e.g., from dike intrusion, aseismic slip transients, or fluid migration). In catalogs dominated by directly triggered aftershocks (i.e., ETAS branching ratios >~0.7), the coseismically-triggered seismicity rate will be much larger than the aseismically-triggered rate and will dominate the estimate of stressing-rate, obscuring the aseismic transient of interest if the rate-state method is applied directly. The challenge therefore lies in isolating the aseismically-triggered seismicity rate from the coseismically-triggered seismicity rate. The ETAS model [ Ogata, 1988] provides a natural way to separate the aseismic and coseismic seismicity rates, as the ETAS parameter μ essentially reflects the aseismically-triggered rate (as well as the background tectonically-triggered rate). To develop a method that can resolve the magnitude and time history of aseismic stress transients even in high branching ratio regions, we combine the rate-state and ETAS models into a single data assimilation algorithm. For a given earthquake catalog, we produce maximum likelihood estimates of the ETAS parameters and use an extended Kalman filter to estimate the temporal evolution of the underlying state variables (stress, stress rate and γ in the Dieterich formulation). We have tested the algorithm with a number of synthetic catalogs and can successfully detect order-of-magnitude changes in stressing rate. Additionally, we can detect large fault creep events detected independently from geodetic data. Ultimately, we aim to map spatial as well as temporal variations in aseismic stressing rates from seismicity data. With this tool we can then identify the space-time evolution of such processes as afterslip, fluid migration, or magmatic intrusion. Moreover, algorithms that can detect when aseismic transients are occurring should have direct applications in real-time seismicity and hazard forecasts.