S53A-0171 1340h
The Seismicity and Structure of Izu-Bonin Arc Mantle Wedge at $31\deg$N Revealed by Ocean Bottom Seismographic Observation
The Izu-Bonin arc system is the subduction zone forming the plate boundary between the downgoing Pacific plate and the overriding Philippine Sea plate. Seismicity along the Izu-Bonin subduction zone is very different in character from other western Pacific subduction zones. Few large earthquakes have occurred at shallow depths (0-100 km), but many large earthquakes have occurred at greater depths ($>$400km). Other unique characteristics of this subduction zone include the existence of serpentine seamounts exposed along the forearc slope and the presence of low-velocity ($<$7.3km/s) material between the two plates in a zone extending from the forearc seamounts to the mantle wedge. To investigate these unique characteristics, we carried out an ocean-bottom seismic experiment in 1999 to estimate the hypocenter distribution and the structure of the mantle wedge simultaneously by performing 3D event locations employed within different velocity models. We obtained the following results: (1) No earthquakes occurred on the upper surface of the subducting plate and some were located as far as 20 km away from the upper surface. Most events occurred within the mantle of the subducting slab. (2) There were no earthquakes in the mantle wedge. (3) The low mantle velocity area in the mantle wedge terminates about 140km west of the trench axis. (4) The subducting slab has a dip of $55\deg$ to the west. From these results we suggest that the low-velocity material between the plates is chrysotile, a low-temperature, low-strength, low friction phase of serpentine, which may act as a lubricant on the plate boundary. The western boundary of the low mantle velocity region in the mantle wedge coincides with the temperature-controlled transition from chrysotile to antigorite (the high-temperature phase of serpentine) along the plate boundary. Our results suggest that chrysotile may migrate upward and eastward along the plate boundary, while antigorite may move downward with the subducting slab.
S53A-0172 1340h
MICROSEISMICITY AROUND THE FOCAL AREA OF THE 1978 MIYAGI-OKI EARTHQUAKE BY OBS OBSERVATION
In the subduction plate boundary off-Miyagi, middle part of the Japan Trench area, M 7 class interplate earthquakes occur repeatedly at intervals of about forty years. Twenty-six years has already passed since the occurrence of the most recent earthquake, the 1978 Miyagi-Oki earthquake (M 7.4), and the Japanese government evaluated that the next large earthquake may occur within 20 years from now with over 80 % possibility. We started a series of ocean bottom seismographic (OBS) observations in September 2002 in order to clarify the spatio-temporal distribution of seismicity around the area where the next large earthquake is supposed to occur. The OBS network is composed of two kinds of pop-up OBSs, long-term (one year) recording type (LTOBS) and short-term ( 3 months) recording type (STOBS). The LTOBSs are deployed at five stations and are replaced every year to maintain the continuous observation until 2005. Additional ten stations with the STOBSs were in operation during the summer of 2003 and 2004. In this paper, we describe the hypocenter distribution of the off-Miyagi area, from September 2002 to October 2003, deduced from the OBS observation. During this period, there were two large earthquakes in off-Miyagi area; one was an interplate thrust earthquake with M 6.1 on November 3 2002 and the other was an intra-slab one of M 7.0 on May 26 2003. Combining the OBS arrival time data with those of the land seismic network data, we relocated hypocenters of the earthquakes whose hypocenters has been located in and around the off-Miyagi region by Japan Meteorological Agency. Owing to the OBS data the accuracies of focal depths was greatly improved and the most of the hypocenters were relocated along a landward dipping plane, the subducting plate interface. The shape of the plate boundary defined by the relocated hypocenter distribution shows a significant change of the dip angle at about 130 km off the coast, where the depth to the plate boundary is about 20 km. Similar geometry of the plate boundary has been reported in the southern and northern part of the Japan Trench subduction zone. Besides the interplate seismicity, it becomes clear that there is a significant seismicity in the overriding land plate as a result of examining the OBS data. It seems that the intraplate seismicity halted for about two months, after the occurrence of the M 6.1 interplate earthquake. It is notable that the period of the quiescence corresponds to that of an aseismic slip along the plate boundary around the rupture area of this earthquake detected by the continuous GPS observation on land (Miura et al.,2003). This coincidence suggests that the intraplate seismicity can be variable according to the stick/slip behavior of the plate interface.
S53A-0173 1340h
Seismicity in the Oceanic Crust and the Uppermost Mantle of the Philippine Sea Plate off Kii Peninsula, the Central Nankai Trough Seismogenic Zone
Along the Nankai trough, southwestern Japan, the Philippine Sea plate (PHS) is subducting beneath the Eurasian plate, and large interplate earthquakes have occurred repeatedly. Recurrence intervals between the earthquakes were about 100-200 years. The latest large thrust earthquake at the eastern Nankai trough off Kii peninsula was the 1944 Tonankai earthquake. A previous ocean bottom seismograph (OBS) experiment showed an absence of the seismicity in the coseismic rupture area of the 1944 Tonankai earthquake and relatively active seismicity near the trough axis. Although the earthquakes near the trough axis were located in the oceanic crust, there were large uncertainties for the hypocenter locations, especially in depth. We conducted a microseismicity observation at the trough axis and the trough landward slope of the eastern Nankai trough off Kii peninsula to obtain accurate locations of the earthquakes near the trough axis and their focal mechanisms. We deployed 30 OBSs with shorter horizontal intervals than those in the previous OBS experiment. Hypocenters were determined using a 3-D seismic velocity structure, which was based on the first-arrival time tomography using the airgun shots as controlled sources. Obtained results show that the most of the earthquakes near the trough axis occurred within the oceanic crust of the PHS and made several clusters. The composite focal mechanisms of the earthquake clusters near the trough axis were strike slip faults with NW-SE direction T-axes. These earthquakes coincide with faults cutting the oceanic crust. The earthquakes in the oceanic crust at the trough axis are considered to indicate the intracrustal deformation of the incoming PHS. On the other hand, the coseismic rupture area of the 1944 Tonankai earthquake is characterized by an absence of the seismicity in the subducting oceanic crust. This may be a result of the uniform coupling between the subducting and overriding plates. Although the seismicity in the oceanic crust is absent in the coseismic rupture area of the 1944 Tonankai earthquake, earthquakes in the uppermost mantle of the subducting PHS were observed beneath the region from the trough axis to the Kii peninsula. The uppermost mantle seismicity may relate to the dehydration of the serpentinized mantle of the PHS.
S53A-0174 1340h
Seismic activity of very low-frequency earthquake on the subducting Philippine Sea plate near the Nankai Trough, southwest Japan
The Nankai trough subduction zone in southwest Japan is characterized by some kinds of _gslow earthquake_h. Around the deep side of the seismogenic zone on the subducting Philippine Sea plate, non-volcanic tremor is distributed in a narrow belt along the strike of the plate (Obara, 2002). On the other hand, on shallower parts of the seismogenic zone, an anomalous seismic activity has been detected. The waveform of the earthquake is characterized by a band-limited low-frequency content of between 10 and 20 seconds. We call the earthquake _gvery low-frequency (VLF) earthquake_h. The filtered seismogram with a passband of 10 to 100 seconds periods operated on the output from high-sensitivity accelerometer (tiltmeter) installed in every NIED Hi-net station is used for detection and location analysis. The waveform of the VLF earthquake looks like to that of teleseismic event, however the spatial pattern of the amplitude and the apparent velocity of the VLF earthquake are quite different from those of teleseismic events. Because waveforms of the VLF earthquake are quite similar in neighbor stations, epicenters are estimated by using a cross correlation analysis. The Hi-net stations in southwest Japan are divided into some groups with a diameter of about 100km. The cross correlation is calculated for every pair of stations in each group in order to measure the time lag which gives the highest cross correlation coefficient. Then, the set of time lags obtained in each group are used to estimate the propagation direction and the apparent velocity. Finally, the epicenter of the VLF earthquake is estimated by focusing the back projection of the ray propagation calculated with good resolution in each group. In the year of 2003, there are two active clusters of the VLF event near the Nankai trough; off Cape Muroto and the Hyuga-nada region. The VLF seismic activity usually lasts for a month in each cluster. Both activities are located on the seaward updip portion of the seismogenic zone on the subducting Philippine Sea plate. Just after the occurrence of 2003 Tokachi Earthquake in northeast Japan on September 26, 2003, the VLF seismic activity in the eastern part of Hyuga-nada region became very active and continued for a month. Considering the reverse fault type mechanism and shallow depth estimated by the centroid moment tensor (CMT) analysis, the VLF earthquakes might occur in the accretionary prism or on the decollement. Moreover, both active clusters correspond to the extension of the chain-like sea mounts on the ocean floor of the Philippine Sea. Therefore, the occurrence of the VLF earthquake might be related to the existence of the subducting sea mount. At present, there is no clear relationship between the episodic tremor and the VLF earthquake activity in their time histories. However, both seismic phenomena are located on transition zones between the locked zone and decoupled aseismic zones at the deeper part and shallower part on the subducting plate boundary. Therefore, these low-frequency families must be representing the subduction process of the young oceanic plate.
S53A-0175 1340h
Hypocenter distribution of plate boundary zone off Fukushima, Japan, derived from ocean bottom seismometer data
The microearthquake observation using twenty-seven Ocean Bottom Seismometers (OBSs) was carried out to obtain the detailed distribution of microearthquakes beneath the area off Fukushima, the middle part of the Japan Trench in the summer of 1997. The hypocenters of 401 earthquakes were obtained during the observation period of about one month. The hypocenter distribution has high spatial resolution (an error of less than 3 km in the horizontal direction and less than 5km in depth), because the structure used for locating events was determined by airgun-OBS surveys in the study area and the hypocenters were finally located by a simultaneous inversion to determine hypocenters and one-dimension velocity structure. Almost all of the hypocenters determined with small spatial errors are in the vicinity of the plate boundary. The seismicity in the area within 100 km distance from the trench is low. The seaward limit of the high seismicity region is close to the western end of the direct contacting zone between the oceanic crust and the overriding landward crust. From focal mechanisms of earthquakes, we estimate that the coupling between the subducting plate and the overriding landward plate is not weak beneath the OBS network and there is intraplate-type earthquake activity in the subducting plate due to an abrupt increase of the dip angle of the subducting plate. Within the overriding landward plate, 29 earthquakes were located. Twelve earthquakes were located about 30km below the plate boundary, forming a landward dipping plane, which seems to be an up-dip continuation of the lower plane of the double seismic zone. The microseismicity characteristics around the plate boundary are interpreted to be controlled by geometry and a property of the plate boundary interface. We suggest that a moment stress due to abrupt bending of the subducting plate and dehydration embrittlement process of serpentinized mantle relate earthquake occurrence in the lower seismic zone beneath the forearc.
S53A-0176 1340h
Fault-plane Identification and Focal Mechanism of the 2003 Chengkung Earthquake Sequence, at East Taiwan
Two subduction systems are taken place around Taiwan Island respectively to the north and to the south. In between, the arc-continent collision proceeds at east Taiwan along the Longitudinal Valley fault. The Chengkung Earthquake, MW 6.5 December 10 2003 at the southern part of the Longitudinal Valley fault, is a typical thrust reflecting release of stress/strain accumulated in the eastern collision framework. Such earthquake sequence contains a seismicity distribution roughly covered a region of 40 km by 40 km, which compose of a thrust fault dipping to the east. According to the focal mechanism of the main shock and the spatial distribution of aftershocks, the mainly crustal movement in the Chengkung earthquake can therefore be recognized as a rupturing with an east-dip plane and thrusting toward the west. However, 15 larger aftershocks occurred 7 days after the main shock showed a more complicated tectonic setting around this area. In the offshore area, three aftershocks illustrated normal-fault-type deformation, which is rarely observed in this area. On land, the larger aftershocks exhibited oblique thrust deformation along the Longitudinal Valley fault. In order to better understand the seismotectonic setting in this earthquake, a waveform inversion was applied to refine the focal mechanism solutions and identify the fault plane from one or multiple seismic recordings at short epicentral distance. Kanamori et al., (1990) and Singh et al., (1997) showed clear examples of local events for which source parameters could be well constrained by using near-field waves, even with a single station. However, when surface ruptures are not observed, as in the case of too small events or blind faults, the fault plane may be undetermined. The method used in this study includes the effect of source finiteness directly in the inversion process, allowing us to invert sparse, near-field data for focal mechanism and fault plane determination simultaneously. Seven parameters, including the strike, dip, rake, and dislocation, were explored with a grid search and minima of the misfit error between the observed and calculated seismograms. The seismograms considered here are in displacement integrated from three-component strong-motion record, organized by the Central Weather Bureau of Taiwan. We have identified three groups of aftershocks showing nearly identical focal mechanisms in the Chengkung sequence, which have initially compiled by the BATS (Broadband Array in Taiwan for seismology), Academia Sinica of Taiwan. Each of these groups have been examined their spatial attitudes of fault plane and refined their focal mechanisms. Our results show a segmented rupture at the southern part of the Longitudinal Valley fault, and a deformation accommodation as function of space can be inferred in the Chengkung sequence. In comparison with the historical earthquake taken place in the late 1951 at the almost location, the Chengkung sequence is probably the reactivated rupture along the same fault segment, suggesting that the temporal and spatial rupture pattern was repeated.
S53A-0177 1340h
Earthquake Rupture Processes Along the Philippine Trench
Earthquakes along subduction zones exhibit spatial and temporal variations in rupture processes. Recent studies have demonstrated a systematic decrease in the moment-normalized rupture duration of earthquakes with depth along the plate interface. Several shallow earthquakes were observed to exhibit anomalously long rupture duration, comparable with the durations of tsunami earthquakes. These observations were suggested to be manifestations of subduction faults with frictional properties capable of generating tsunami earthquakes. The cause of the duration variation was hypothesized to be related to physical changes in properties of subducted sediment. The Philippine trench is a setting where we can study the variations of earthquake source rupture processes along the plate interface. The trench is young ($< $ 5 Ma) with a poorly developed accretionary prism. Bathymetry, seismic reflection studies, and offshore drilling show little or no evidence of accretion of the materials from the subducting Philippine Sea plate. Eighty nine possible interplate earthquakes were selected from the Harvard CMT catalogue for the year 1989 to 2001. From these, about 30 events had adequate signal-to-noise ratios and well-constrained mechanisms to warrant further investigation. We used teleseismic broadband records of vertical component waveforms and applied multi-station deconvolution technique to extract the source time function and depth of each event. Our results showed no systematic trend of decreasing source duration with depth at the Philippine trench. The observed moment-normalized durations of all the shallow earthquakes are shorter than tsunami earthquakes. We observed a significant scatter in the relationship between source time and depth, which may reflect heterogeneity of the materials at the trench interface. The thickness of low-rigidity materials in the trench shallow region appears insufficient to affect the rupture durations and produce anomalously slow ruptures. These null results for the sediment-starved Philippine Trench support the earlier hypothesis that the variations of earthquake duration with depth in other subduction zones is related to sediment properties.
S53A-0178 1340h
Examination of interplate coupling in the southern Alaska subduction zone, based on the GPS data
We reexamined the interplate coupling in terms of backslip rate beneath the Kenai Peninsula in southern Alaska with a 2-dimensional GPS time series during three different time periods. Zweck et al., (2002) investigated the variation in coupling between the Pacific and North American plates, and they estimated interplate coupling as a scalar quantity. With this background, we estimate back-slip distribution on the plate interface as a vector by applying geodetic inversion method devised by Yabuki and Matsu'ura (1992) based on the GPS velocities data. We divided into three time periods, before 1997, 1998-2001 and after 2001. Freymueller et al (2000) suggested that a transient crustal deformation, slow slip event, occurred in this region during 1998-2001. The second time period coincides with a slow slip event on the plate interface. We inverted GPS velocities to infer the interplate coupling before 1998 and after 2001. Our results show that a maximum back slip rate of 60mm/year was found at the depth of 15-25km around 148.8W, 60.0N. According to results compiled by DeMets and Dixon (1999), convergence rate of the Pacific plate has a velocity at a rate of 57 mm/year for this region. This convergence rate suggests that east coast Kenai Peninsula region has an approximately full locking between the North American and subducting Pacific plate. In the contrast, forward slip appears north of the Kenai Peninsula during 1998-2001. By subtracting the averaged back-slip rate during before 1998 and after 2001, we extracted the abnormal deformation caused by slow slip event. Event has a northeastward direction (azimuth 165N). Slip at a 3-year average rate of up to 45mm/year was inferred north of the Kenai Peninsula (149.0W, 61.5N) at a depth of 25-45 km. Over 3-years the cumulative moment magnitude reaches Mw=7.2.
S53A-0179 1340h
The Relationships Between Earthquakes, Faults, and Recent Glacial Fluctuations in Southern Alaska
In southern Alaska, northwestward-directed subduction of the Pacific plate is accompanied by accretion of the Yakutat terrane to continental Alaska. In the tectonically complex region between the transcurrent Fairweather fault and the Alaska-Aleutian subduction zone, active crustal shortening and strike-slip faulting occurs. Since a series of large earthquakes in 1899 (Mw = 8.1, Yakataga; Mw=8.1 Yakutat), there has been only one large event (1979 St. Elias Mw = 7.4) in the Yakutat region between the aftershock zones of the 1964 Prince William Sound (Mw = 9.2) and 1958 Fairweather (Mw = 8.2) earthquakes. In this region, the glaciers are extensive and many of them have undergone significant retreat in the last 100 years. This study investigates the relationships between small to moderate magnitude events, ongoing crustal deformation, active geological structures in the region, and recent glacial fluctuations. To map earthquake locations with respect to current glacier positions, we will incorporate Ice Cloud and land Elevation Satellite (ICESat) data into an updated Digital Elevation Model (DEM) of key glaciated regions that has been created using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images in conjunction with Shuttle Radar Topography Mission (SRTM) data. For the seismological investigation, we focused on relocating events that have occurred since the last large earthquake at St. Elias in 1979 using data obtained from the Alaska Earthquake Information Center (AEIC). P-wave polarity first motion focal mechanisms were generated for the relocated events and evaluated. Our preliminary relocations suggest a dipping slab in cross-section and also show a number of shallow event clusters around local glaciers. The focal mechanisms are quite variable but, in general, indicate strike-slip and oblique-slip focal mechanisms. Some of our highest quality focal mechanisms show dip-slip faulting and are from shallow events located near glacial boundaries. For mapping the correlation of earthquakes to geological structures, we will use 3D visualization software to plot relocated events and correlate them with shallow structures in the region.
S53A-0180 1340h
Investigating active faulting in the south-central Chilean forearc by local seismicity and moment tensor inversions
The seismological ISSA experiment is giving a detailed insight into the seismicity distribution of southern Chile, where major earthquakes (M>8) have repeatedly ruptured the surface, involving vertical offsets of several meters. During a nearly 5-month observation period in 1999 and 2000 a temporary seismic network recorded approximately 350 local earthquakes. Two localized areas, North and South of the Arauco peninsula, showed a very high seismic activity in and above the interplate seismic zone of the Nazca-South America convergent margin. We used a double-difference relocation technique to obtain detailed images of the seismicity distribution in these areas. We also determined fault plane solutions to interpret the observed alignment of earthquakes hypocenters. Due to the low signal to noise ratio reliable first motion reading were difficult to achieve, which only very few clear readings. To overcome this problem we used moment tensor inversions to estimate reliable source mechanisms. However, for small magnitude earthquakes ($<$5) the biggest obstacle is the alignment of synthetic and observed waveforms. Inverting only for the amplitude spectrum, and therefore dropping the information in the phase spectrum can mostly circumvent the alignment problem. The two clusters investigated show high waveform correlation coefficients for most of the earthquakes indicating that possibly changes in fluid pressure can be responsible for triggering the events. After relocation most of the hypocenters in each of the two clusters align on a eastward dipping fault. Source mechanisms obtained indicate thrust faulting, where one of the possible fault planes aligns with the steep eastward dipping fault based on the seismicity distribution. These faults are reaching down to the top of the seismogenic zone and may serve as pathways for ascending fluids released in the subduction process. Active crustal-scale faulting below and active uplift of the coast account for active tectonic segmentation and are discussed in relation with active basal accretion and active shortening in the South-Central Chilean forearc.
S53A-0181 1340h
Nonlinear Source Tomography Of The Mw=8.4, 23 June 2001 Arequipa, Peru, Earthquake.
We investigate on the rupture process of the $Mw=8.4$ tsunami-genic earthquake of June 23, 2001 in Arequipa, Peru, putting special emphasis on the determination of rupture velocity. Since the determination of rupture velocity is nonlinear we use the Neighbourhood algorithm for the inversion. We use 16 broad band far field seismograms band pass filtered between 0.01 and 0.4$Hz$. The goodness of the fit between observed and synthetics is mesured with an $L^2$ norm. The fault is parameterized with a model containing 57 patch of size $9\times 20km$. Because of computer time problems the final steps in the inversion, once the solution has converged to a well defined minimum, are done with a downhill simplex algorithm. The best solution for this unilateral event is found to have propagated southward at a very low speed ($1,6km/s$) with $80%$ of the final moment released as one patch $80s$ after the onset of rupture. The general features of the inversion are compatible with results of linearized inversion and the slow rupture speed is similar to that observed for tsunami earthquakes. The location of the main energy patch is confirmed by a CMT grid search over the whole area of rupture and by tsunami observations.
S53A-0182 1340h
Rupture Process of the June 23, 2001 Mw 8.4 Peru Earthquake
The 2001, Peru, earthquake, the largest earthquake to have happened worldwide since 1965, occured off the coast of southern Peru. The largest previous reported earthquake (Mw $\sim$8.8) in this region occurred in 1868. Of the 996 aftershocks reported by the ISC in the 6 months following the earthquake, we relocate 551 to within an uncertainty of 30km and find that the majority form a rectangular zone of length $\sim$350km, width $\sim$150km and with a strike of $\sim$310$\deg$.We recalculate the moment tensor using mantle waves from 86 stations and 257 channels and find a solution close to that reported in the Harvard catalogue (centroid location 17.20$\deg$S, 72.59$\deg$W, $M_0 \sim 5.66 \times 10^{21}$ Nm, strike 311$^{\circ}$, dip 12$^{\circ}$, rake 68$^{\circ}$). The aftershock area together with the CMT solution indicates that rupture occurred on a shallowly-dipping northwest-southeast trending thrust fault associated with the subduction of the Nazca plate under South America. The rupture zone is contained entirely within that of the 1868 earthquake (estimated from intensity data). Inversion of broad band body wave data shows that most of the moment release occured between 30 and 100 seconds after the initiation of the event, and is located from 100-200km to the southeast of the epicentre. Very preliminary inversions show that the majority of slip is primarily confined to the shallower portions of the crust, providing an explanation for the formation of the associated tsunami. Although this earthquake ruptured a large portion of the plate boundary, there still remains a significant seismic gap of $\sim$700km to its immediate south, off the coast of northern Chile.
S53A-0183 1340h
Temporal and Spatial Correlations Between Interplate and Intraplate Subduction Zone Seismicity
Subduction zone earthquakes on the interface between subducting and overriding plate represent some of the world's most destructive natural disasters. Updip from this interface, the outer rise comprises an upwarping of the oceanic lithosphere just before it descends into the trench. Previous work established the characteristics of the stress regime within the subducting lithosphere and has suggested that a temporal and spatial correlation may exist between interplate and intraplate seismicity, with intraplate earthquakes possibly serving as stress gauges for the large-scale deformation involved in subduction zones. We have compiled a new catalog of outer rise seismicity, using an automated search algorithm applied to the Harvard CMT catalog, and confirmed the preferential occurrence of normal faulting outer rise events after large interplate thrust events, in particular after tsunami earthquakes. However, any possible temporal pattern in compressional outer rise events is not as straightforward, which hints that a more complex physical mechanism may be at work in the seismogenesis of outer rise seismicity than simple elastic plate bending. A more likely candidate is a combination of downward plate bending and in-plane compression, with the former being the main contributor for outer rise tensional earthquakes and an elevated level of the latter being responsible for thrust faulting outer rise events. An inelastic analysis of lithospheric stress distributions predicts similar seismic behavior. We also investigated the spatial occurrence of outer rise events and found a correlation of heightened activity with increasing plate curvature, plate age as well as the dip of the subducting plate. Further investigations will also focus on the orientation of the source mechanisms with respect to the strike and dip of the subducting plate, with preliminary results indicating an unexpected asymmetry in the dips of the two fault planes. We plan to construct a more complete and enhanced catalog of intraslab seismicity, extending our analysis to greater depths (150 km) and lowering the magnitude threshold from 6.0 to 5.5 for the relatively infrequent compressional outer rise events. This catalog will aid in refining our understanding of the stress evolution within the subducting slab and the relationship between seismic coupling and intraplate seismicity. We will apply a teleseismic P-wave modeling technique to refine the depths of the intraplate events to a higher precision and homogeneity. Subsequently a source spectral analysis of the earthquakes with magnitude greater than 6.5 will be carried out to determine dynamic rupture parameters. These investigations will provide us with insight into a myriad of issues, including: the failure mechanism of intraplate and outer rise earthquakes, the state of stress in the subducting lithosphere, the origin of the hydration of the subducting plate and the mode of deformation of the outer rise.
S53A-0184 1340h
Outer Rise Stresses of the Tonga Subduction Zone Inferred from Earthquakes and 2D Models of Flexure
We investigated outer rise intraplate stresses at the Tonga Trench using both observed earthquakes and flexure models. The depths and focal mechanisms for 21 intraplate outer rise earthquakes spanning 1987-2003 were determined using teleseismic P and SH body waveform inversion, and these were combined with 14 other prior earthquakes whose source parameters had been similarly determined. The 35 earthquakes were co-located using a joint-hypocenter decomposition technique in order to obtain precise relative locations with uncertainties generally less than 3 km. A well-defined pattern emerged with distinct groups of 13 shallow (10-26 km) tensional and 15 deeper (30-49 km) compressional events within an approximately 40 km thick seismic lithosphere. The approximately 10 km gap between these two groups provides an estimate of the lithosphere's elastic core thickness and constrains the depth of the neutral surface to about 28 km. Additionally, 3 deep (35-60 km) and shallow (14-30 km) tensional and compressional earthquakes occur greater than 75 and 100 km downdip and seaward of this zone, respectively. These imply that the plate is under inplane compression before it enters the zone of maximum bending and that it is fully under tension past it. We also discuss the temporal and spatial outer rise earthquake patterns and show that compressional and tensional events cluster into regions along the trench and tend not to occur in the same area at the same time. Flexure models of the Pacific Plate were computed using a 2D finite difference technique with a multilayered brittle-plastic rheology and plate cooling models defined by either constant- or depth-dependent thermodynamic parameters. In addition to modeling a general Tonga slab with age 90 Ma, we also separately investigated the areas north and south of the Capricorn Seamount. The parameters modeled included the bending moment, interplate coupling force, inplane force, and strain rate. The best fitting models to bathymetry and free air gravity profiles accurately predict the observed location of the neutral surface and seismic stress patterns. We find that the compressional seismicity observed in the lower lithosphere occurs within the deep plastic portion of the plate, suggesting either that the semi-brittle region extends quite deep within oceanic lithosphere or that some type of ductile faulting is taking place. The best fitting models require a compressional inplane force, which implies that slab pull is not a dominant force in this section of the subducting plate.
S53A-0185 1340h
Global Outer-Rise/Near Trench Seismicity and Focal Mechanisms: Trends and Diversity
Based on well-constrainted hypocenter information and focal mechanism solutions, Seno and Yamanaka (1996) identify two classes of outer-rise/near-trench (OR/NT) earthquakes: 1) Shallow tensional events and 2) Deeper compressional events. They indicate that subduction zones with compressional deep events (DCE) also tend to have double seismic zones (DSZ), and proposed a hypothesis that earthquakes in the lower plane of the DSZ represent reactivation of faults by dehydration embrittlement previously active as DCE_fs. They suggest that plates can be hydrated even at depths as great as 40 km by passing over the superplume volcanic centers. In this study we examine the characteristics of many more earthquakes at the outer-rise/near-trench region using the EHB hypocenter catalogue (Engdahl et al., 2002) and Harvard CMT focal-mechanism. We studied M$>$5.5 OR/NT events in the Circum-Pacific/Indonesian earthquake belts from 1977 to 2002 that occurred at depths shallower than 60 km depth and within 150 km from the trench axis. In order to select trench-outer-rise events, we checked carefully all event locations and CMT solutions in map- and cross-sectional-views superposed on the background seismicity. We also compared event locations with global maps of seafloor bathymetry, gravity and Kawakatsu_fs 1986 investigation of the DSZ in the Tonga subduction zone. Our results are as follows: 1) Eighteen compressional and 93 tensional events were found. 2) Solitary compressional events were found in two areas (Vanuatu and Guam) where a DSZ has not been observed. 3) Thirteen compressional events are concentrated in the Tonga-Kermadec reagion. 4) Tensional events occur at depths of less than about 33 km and the compressional group occurs at greater depths in the Tonga-Kermadec region, corresponding to Kawakatsu_fs upper an lower zones but with opposite focal mechanisms. 5) Many trench-outer-rise events occur where seamount/guyot volcanic chains are subducting. 6) Compressional OR/NT events tend to have mb_es that are larger that those for tensional and interplate events. This implies that DCE events at trench-outer-rise have higher amplitudes at high frequencies.
S53A-0186 1340h
Evidence of Two Source Mechanisms for Cascadia Intraslab Seismicity From an Anomalous Secondary P Arrival
A prominent secondary P arrival is observed at stations on the Olympic Peninsula from 13 intraslab earthquakes in the subducting Juan de Fuca (JDF) slab beneath western Washington. The events occur in two regions at a depth range of 45-55 km. Ten of the events locate beneath south central Puget Sound, and 3 are off the SE tip of Vancouver Island. These two regions are coincident with the flanks of the JDF arch structure. Intraslab events outside the 45-55 km depth range do not show the secondary phase, nor do events within this depth range located between these two regions. First arriving P at Olympic stations for all intraslab events have apparent velocities consistent with upper mantle paths, while the secondary P arrivals have apparent velocities of approximately 6.0 km/s, consistent with crustal paths. Careful comparison of slab arch models to hypocenters suggests that earthquakes with a secondary P originate above the slab Moho in the subducted oceanic crust, and those without the secondary phase are in the upper mantle of the JDF slab. Two independent source mechanisms are proposed for Cascadia intraslab seismicity with most of the earthquakes occurring in the slab upper mantle and relatively few in the subducted crust. The pattern of larger (M $>$ 2.5) events without secondary P displays a linear, NW trend. Magnetic anomaly patterns [Wilson, 2002] reveal an offset in the isochrons coincident with this linear distribution, suggesting a link between these slab upper mantle events and structure. The largest historic Cascadia intraslab events including the 1949 M7.1, 1965 M6.5 and 2001 M6.8 earthquakes occur in the south Puget Sound region where most of the events with the secondary phase are located. Trace data are not available at Olympic Peninsula stations for the two earlier events and is clipped for the 2001 main shock, so the existence of a secondary phase for these events is indeterminate. However, the two largest aftershocks from the 2001 event and all other well-recorded events in this region show the secondary phase and are therefore interpreted to be in the subducted oceanic crust. This suggests that the largest events may originate in the subducted oceanic crust.
S53A-0187 1340h
Source processes of the 1949 Olympia, Washington and other Cascadia intraslab earthquakes
Cascadia intraslab earthquakes have occurred frequently including in 1949 (M 7), 1965 (M 6.9), and 2001 (M$_{w}$ 6.8) and the hazard they pose is equal to other seismic sources for time scales relevant to retrofitting (50%PE in 75 yrs). Reexamination of historical earthquakes and analysis of broadband seismograms from new earthquakes improves strong motion prediction more than stochastic scenarios and provides a better understanding in how stress is released in the subducting slab relative to external tectonic forces and local dehydration. Reanalysis of the 1949 Olympia earthquake using digitized seismograms collected by Barker and Langston [1987], Wiest et al. [2004], long-period WWSSN data from College, Alaska, Bogot\'{a}, Columbia, and Pasadena, California, has yielded a normal-slip mechanism that fits all the available data with $\sim$ E-W extensional-axis. This mechanism better resembles those from previous Puget Sound intraslab earthquakes. The earthquake ruptured southward along a plane with 170\deg strike and 70\deg dip. The hypocenter depth is 60 km with a total M$_{o}$ of 1.9\times10$^{26}$ dyne-cm (M$_{w}$ 6.78). We inverted teleseismic body waves to determine the mechanism and rupture pattern in space and time using a time-domain iterative inversion process. We limited the number of unknowns by assuming only 2 subevents and performed a grid search on the remaining variables including hypocenter depth, rupture direction, subevent rise time and rupture velocity. A rise time of 3 s best fits the frequency content of the waveforms but there is no resolution of rupture velocity with this dataset so we assume 3.5 km/s. The total moment is typically lower because of stricter model assumptions and the use of band limited data. The amplitudes from the Pasadena 1-90 s Benioff records indicate a higher M$_{w}$ $\sim$ 6.8 when compared to those from the 2001 Nisqually earthquake convolved with a similar response. We show from similar analyses of this and other intraslab earthquakes that they have significantly smaller asperity areas relative to other types of earthquakes with the same seismic moment originally identified by Asano et al. [2003] and Ichinose et al. [2004].
S53A-0188 1340h
Review of Episodic Tremor and Slip in Cascadia
Episodic Tremor and Slip (ETS) has now been observed over much of the Cascadia subduction zone, including regions within the main Juan de Fuca Plate segment, the Explorer Plate and Gorda Plate segments at the northern and southern ends respectively. The non-earthquake, tremor-like seismic signals have frequency content between 1-6 Hz, last anywhere from tens of seconds to hours, are made up primarily of S-waves and vary in amplitude over periods of seconds to tens of seconds. These appear similar to the tremor reported in southern Japan. Concentrated bursts of tremor lasting 10 days or more are associated with the systematic displacement of GPS sites over a period of 1 or 2 weeks in a direction opposite to longer-term elastic strain accumulation. This geodetic signal has been successfully modeled as deep slow slip events on the subduction fault interface. From Vancouver Island to southern Washington the ETS zone appears to be continuous along strike. Different segments rupture at different times, but some with remarkable regularity of about 14 months. Over the last 7 years the Explorer Plate region ETS activity has occurred with the same 14 month period as the Juan de Fuca plate region of southern Vancouver Island, but 6 months out of phase. A separate zone in southern Washington seems to have events both synchronous and independent of the Vancouver Island segment. An area in central Puget Sound is overlapped by two different tremor zones and undergoes tremor activity during both sets of slip events as well as short periods of tremor not associated with geodetically detectable slip. Tremor bursts in northern California have been located beneath the northern Great Valley in the Lake Shasta area. Tremor in central and southern Oregon has not been positively identified yet; however the seismic network in this area is not extensive enough to easily and unambiguously identify it. Tremor source depths are not well constrained by regional network location techniques but do seem to vary over a range of depths (15 to 55 km) larger than the depth uncertainty. Thus, the source region is not confined to the plate interface but extends some distance above it.
S53A-0189 1340h
OTIS Flow Meter for Offshore Detection of Silent Earthquakes
The Optical Tracer Injection System (OTIS) is a new marine flow meter developed for the novel application of measuring fluid flow rates sensitive to transient pore pressure gradients and volumetric strain. These meters are designed with the potential to detect aseismic deformation in the offshore environment. Upcoming empirical results from subduction zone studies in Cascadia and Costa Rica are to be processed with dislocation modeling. Knowing the role of fluids in aseismic and seismic events will help to characterize the earthquake cycle: what fraction of built up stress is released in the form of earthquakes vs. silent creep events, and in what sequence these events occur. While bore hole studies give significant insight to subsurface spatial and temporal pressure gradients, they are difficult to execute and costly. On the other hand, monitoring fluid flux at natural focused flow seep sites on the seafloor as an indicator of strain, like monitoring water level changes in well bore studies, is a cost efficient alternative. Coupled to the seafloor, the OTIS optically senses, and electronically logs, the concentration of injected fluorescent rhodamine tracer from which the rate of fluids flowing through the sediment water interface can be determined. It has been developed to detect a large range of flow rates at the sediment water interface, 0.01 to 500+ m/yr, at 5-30 minute temporal resolution. The OTIS is deployed in conjunction with an osmotic sampler for post-recovery chemical analyses of pore fluids expelled at seep sites. Initial instrument tests at cold seeps in Monterey Bay coincided with Chemical Aqueous Transport (CAT) meter results and confirm improved temporal resolution, and low sensitivity to ocean currents. Presently, an OTIS is placed at the toe of the Cascadia forearc at the triple junction with the Nootka transform fault as part of a modem package for uplink of real time data to a surface buoy. In conjunction with land based seismic and geodetic networks, offshore flow meters capturing a hydrologic signature slow earthquake events help to constrain the geophysical and hydrogeologic interpretation of subduction zone processes.
S53A-0190 1340h
Small Aperture Array Resolution Capabilities for Use in Locating Deep Tremor
Three small aperture seismic arrays (diameter ~600m) were operated in Northern Washington during spring and summer, 2004 to monitor an expected episode of Deep Tremor which occurred in July. Each array was composed of six 3-component, short-period seismometers. Besides recording two weeks of strong, deep tremor many earthquakes with a range of magnitudes, distances and azimuths were also recorded. Earthquake signals have been used to test the resolving ability of the arrays for subsequent analysis of the deep tremor signals. Analysis techniques tested were "Zero Lag Cross-Correlation" and classical "Beam Forming" array techniques in time domain, and "High Resolution" array methods in the frequency domain. Calculated backazimuth and slowness values of P and S direct phases were compared with those calculated based on the hypocenters determined by the regional network and the regional velocity structure to give an estimate of the error associated with array techniques. The time domain analysis results show good agreement with the real earthquake locations. The resolution capability associated with the Zero Lag Cross-Correlation method is between 0.01 and 0.02 s/km for slowness and 5-10 degrees in back azimuth, depending mostly on the signal to noise ratio. On the contrary, the results of spectral techniques are less reliable and generally differ more than expected from the true values. This is probably due to the small number of stations composing the arrays which affects the spectral techniques more than the time domain techniques. As a further test, synthetic waveforms were calculated for a variety of parameters including backazimuth, slowness, array geometry and noise level (injected synthetic noise). The synthetic tests results are comparable with those obtained by the analysis of earthquakes. These results will be used for estimating location errors of deep tremor signals as recorded simultaneously on all three arrays.
S53A-0191 1340h
Linearized perturbation analysis of along-strike nonuniformity of slip in 3D fault models with depth-variable properties
In our three dimensional modeling [EOS, 2003; JGR submitted, 2004] of long term loading and earthquake sequences on a shallow subduction fault, with depth-variable rate and state friction properties, we found the response was perturbed into a strongly nonuniform slip mode along strike by introducing small along-strike perturbations in friction properties. Similar results were found in some cases of 3D strike slip modeling by Rice and Ben-Zion [PNAS, 1996]. To explore this further, we report results of linearized perturbation analyses for two versions, ``ageing'' (or ``slowness'') and ``slip'', of the friction laws. The 3D solution vector $S(x,z,t)$, where $x,z$ are the respective along-strike and downdip coordinates in the fault plane, consists of shear stress $\tau(x,z,t)$, slip $\delta(x,z,t)$ and state variable $\theta(x,z,t)$. It can be written as the sum of a 2D solution vector $S_0(z,t)$, which is subject to initial conditions $S_0(z,0)$, and an infinitesimal variation Re$[S_1(z,t) \exp(2 i \pi x / \lambda)]$, where $\lambda$ is a perturbation wavelength. In our case the friction properties and external driving are such that $S_0(z,t)$ describes a sequence of earthquakes separated by long interseismic loading intervals during which slow creep slippage occurs, like in the Tse and Rice [JGR, 1986] type of 2D modeling. Linearizing the governing equations in $S_1(z,t)$ (giving a nonautonomous system, because coefficients depend on $S_0(z,t)$), we can calculate the evolution of $S_1$ for a given unperturbed history $S_0(z,t)$ and initial conditions $S_1(z,0)$. For both pure thrust and pure strike-slip fault geometries, we found that there is a critical ratio $\lambda_{crit}/h^*$, which seems to determine the stability of along-strike response; $h^*$ is the minimum neutrally stable downdip slip patch size, according to rate and state stability theory for perturbation of steady slip. When $\lambda_{crit}/h^*$ is greater than the critical value, $\partial\delta_1(z,t)/\partial t$ and $\theta_1(z,t)$ grow to significantly large values; when less than the critical value, the perturbations decay with time. Our calculations give the critical ratio around 4 to 6. Such a transition is confirmed by our fully nonlinear 3D simulations. However, the perturbation growth history (which is not a simple exponential in $t$) depends on the position at depth. Fault parts which are in the well-locked seismogenic zone have earlier rise time and faster growth rate than those in the velocity-strengthening regions. Remarkable resistance to break up into non-uniform strike slip was observed by Rice and Ben-Zion when using the slip version of friction law. Linearized perturbation analysis with that version shows, as compared to the ageing version with the same $h^*$, that the $\lambda_{crit}$ is similar but that the perturbation growth is much slower. This might explain why along-strike heterogeneity of slip for that version was much less than for the ageing version. However, for the 3D thrust fault case, we have not found significant qualitative difference for the two laws; both show break-up of the slip distribution along strike.
S53A-0192 1340h
Delineation of Rupture Propagation of Large Earthquakes Using Source-Scanning Algorithm: A Control Study
Determination of the rupture propagation of large earthquakes is important and of wide interest to the seismological research community. The conventional inversion method determines the distribution of slip at a grid of subfaults whose orientations are predefined. As a result, difference choices of fault geometry and dimensions often result in different solutions. In this study, we try to reconstruct the rupture history of an earthquake using the newly developed Source-Scanning Algorithm (SSA) without imposing any a priori constraints on the fault's orientation and dimension. The SSA identifies the distribution of seismic sources in two steps. First, it calculates the theoretical arrival times from all grid points inside the model space to all seismic stations by assuming an origin time. Then, the absolute amplitudes of the observed waveforms at the predicted arrival times are added to give the "brightness" of each time-space pair, and the brightest spots mark the locations of sources. The propagation of the rupture is depicted by the migration of the brightest spots throughout a prescribed time window. A series of experiments are conducted to test the resolution of the SSA inversion. Contrary to the conventional wisdom that seismometers should be placed as close as possible to the fault trace to give the best resolution in delineating rupture details, we found that the best results are obtained if the seismograms are recorded at a distance about half of the total rupture length away from the fault trace. This is especially true when the rupture duration is longer than ~10 s. A possible explanation is that the geometric spreading effects for waveforms from different segments of the rupture are about the same if the stations are sufficiently away from the fault trace, thus giving a uniform resolution to the entire rupture history.
S53A-0193 1340h
Compactness vs. Smoothness: Methods for regularizing fault slip inversions with application to subduction zone earthquakes.
We examine inversions of geodetic data for fault slip and discuss how inferred results are affected by choices of regularization. The final goal of any slip inversion is to enhance our understanding of the dynamics governing fault zone processes through kinematic descriptions of fault zone behavior at various temporal and spatial scales. Important kinematic observations include ascertaining whether fault slip is correlated with topographic and gravitational anomalies, whether coseismic and postseismic slip occur on complementary or overlapping regions of the fault plane, and how aftershock distributions compare with areas of coseismic and postseismic slip. Fault slip inversions are generally poorly-determined inverse problems requiring some sort of regularization. Attempts to place inversion results in the context of understanding fault zone processes should be accompanied by careful treatment of how the applied regularization affects characteristics of the inferred slip model. Most regularization techniques involve defining a metric that quantifies the solution "simplicity". A frequently employed method defines a "simple" slip distribution as one that is spatially smooth, balancing the fit to the data vs. the spatial complexity of the slip distribution. One problem related to the use of smoothing constraints is the "smearing" of fault slip into poorly-resolved areas on the fault plane. In addition, even if the data is fit well by a point source, the fact that a point source is spatially "rough" will force the inversion to choose a smoother model with slip over a broader area. Therefore, when we interpret the area of inferred slip we must ask whether the slipping area is truly constrained by the data, or whether it could be fit equally well by a more spatially compact source with larger amplitudes of slip. We introduce an alternate regularization technique for fault slip inversions, where we seek an end member model that is the smallest region of fault slip that can explain the data. In these "compact slip" inversions, we define model simplicity as the spatial compactness of the fault slip distribution. Our measure of compactness allows for multiple regions of slip when they are required by the data. We compare inversions using compactness and smoothness as regularization criteria in several synthetic scenarios. We also apply the compact slip technique to coseismic and postseismic deformation associated with the 1995 Mw 8.1 Antofagasta, Chile, and 2003 Mw 8.1 Tokachi-Oki, Japan, subduction zone earthquakes, using InSAR and GPS data.