T12B-01 INVITED 10:20h
Onshore-Offshore Correlation of Geologic Evidence for Great Cascadia Earthquakes--Permissive Agreement Between Washington Estuaries and Cascadia Deep-Sea Channel
Geologic dating permits one-for-one correlation between coseismic subsidence in coastal Washington and turbidity currents in Cascadia deep-sea channel in the past 4000 years. The correlations make sense if plate-boundary rupture off the Washington coast accounts for the coastal subsidence. Such rupture would radiate seismic shaking directly beneath the submarine canyonheads that feed turbidity currents to Cascadia channel. The correlation strengthens the basis for using the turbidites as proxy records of great Cascadia earthquakes in the early Holocene, beyond the typical reach of the region's coastal subsidence records. For the past 4000 years in southwest Washington, the estuarine record of coseismic subsidence contains eight events at irregular intervals. The record can be seen at low tide as buried marsh and forest soils each capped with tidal-flat mud and some also coated with sand from tsunamis or sand blows. Successive soils differ consistently in stratigraphic patterns that provide field criteria for correlating among outcrops and for estimating relative lengths of recurrence intervals. The field correlations have survived chronological tests that include 32 sample ages with one-sigma counting errors less than 20 radiocarbon years. Most of these were measured on inner rings of the roots of earthquake-killed trees, from which the difference between sample age and event age can be counted with annual growth rings. Other materials set limiting ages that constrain event ages less exactly. Event age ranges at >95-percent confidence, in calendric years before AD 1950: 3550-4150 [8 events ago]; 3310-3390 [7]; 2845-2925 [6]; 2420-2620 [5]; 1540-1610 [4]; 1229-1264 [3]; 760-1190 [2]. The most recent event [1] dates from AD 1680-1720 (radiocarbon), 1699-1700 (ring-width patterns), and January 27, 1700 (Japanese documents). Well-defined intervals between successive events range from a few centuries [4 to 3] to a millennium [5 to 4]. One interval [7 to 6] approximates the mean of about 500 years. Radiocarbon dating provides limiting-maximum ages for five of the last eight turbidity currents in Cascadia channel. Two such limiting ages were measured for each of two turbidites [1, 6], and one age was obtained for each of three other turbidites [3, 4, 8]. Each sample age was measured on planktonic foraminifera that settled to the seafloor before the turbidity current arrived. The difference between the sample age and the event age depends largely on the depth of erosion (if any) by the current. For simplicity we assume no such erosion. In addition, using standard carbon reservoir corrections, we assume that the foraminiferal carbon had an initial age of 800 radiocarbon years. Resulting maximum ages for the turbidites, at two-sigma in calendric years before AD 1960: 3540-3840 [8], 2660-2840 [6], 1600-1830 [4], 1300-1460 [3], and 0-200 [1]. One-for-one correlation provides the simplest explanation for these findings. The correlation is violated only by the limiting-maximum ages for turbidites 1 and 6, which are slightly younger than the tree-death dates for subsidence events 1 and 6. The limiting age for turbidite 6 agrees better with a coastal event age if turbidite 6 correlates with coastal event 5. In that case, the coastal record is incomplete: turbidite 5 lacks an onshore correlative and divides the longest coastal interval (5 to 4). Stratigraphic evidence onshore and offshore casts doubt on this alternative interpretation.
T12B-02 INVITED 10:35h
Earthquake Potential Inferred From Recurrence Intervals of Turbidites in the Southeastern Margin of the Japan Sea
The eastern margin of the Japan Sea, situated along the western coast of the northeast Japan, has been regarded as a new plate boundary. Several damaging earthquakes whose magnitudes were 7.5-7.8 took place along the eastern margin of the Japan Sea during the last century. However, the recurrence intervals of earthquakes in this margin have been poorly known since they were too long to have been covered by historic records. Although several attempts have been made to evaluate earthquake potential using recurrence intervals of deep-sea turbidites in the northern half of this margin, southern half has remained uninvestigated. Assessment of seismicity in this part of the Japan Sea has been urgently required, as a part of this portion (i.e. Sado Ridge) has been suspected to be a seismic gap. In this study, the earthquake potential in the southeastern margin of the Japan Sea was studied based on the recurrence intervals of turbidites. The average recurrence interval of turbidites was estimated around 1000, 500 and 250 years in the southern Sado Ridge (38-39deg.N, 138-139deg.E), the northern Sado Ridge (39-40deg.N, 138-139deg.E) and the northernmost, hypocenter region of 1983 earthquake (40-41deg.N, 139deg.E), respectively. In the Sado Ridge, the last event was estimated at A.D.1200-1600 in the southern part and A.D.1500-1800 in the northern part, respectively. These results suggest that the earthquake potential in the southeastern margin of the Japan Sea is higher than in the northeastern margin where turbidite recurrence intervals of 500-3000 years have been reported. The results also support the notion that the Sado Ridge is a seismic gap and a high-risk area of near future earthquake/tsunami hazards. However, more study is required to reveal the exact seismicity of the eastern margin of the Japan Sea since fault systems there are complex and segmented.
T12B-03 INVITED 10:50h
Cascadia Great Earthquake Recurrence: Correlation Evidence for Repeated Margin Wide Rupture
We are now testing correlation between turbidite event records at widely separated sites in Cascadia with radiocarbon ages and physical properties of the core sediments. We focus here on physical property correlations between sites to test for connections between sites independent of radiocarbon ages. Gamma density, magnetic susceptibility, and P-wave velocity data were routinely collected for all cores at a 2 cm interval. We find that a good stratigraphic correlation can be made between Juan de Fuca Channel (JDF, a tributary of Cascadia Channel) and Cascadia Channel based on individual event signatures, and upon the sequence of unique signatures through the Holocene record of 18 events. 16 individual event density-magnetic signatures between JDF and Cascadia Channel have correlation coefficients of 0.6-0.9, with two scores (0.16 and 0.32) for events with similar, but out of phase characteristics. Non correlated events have lower coefficients in most cases. The event sequence includes a mix of many unique signatures and some events that are similar to others. The unique sequence of these signatures limits the possibility of miscorrelation because rarely do similar events occur adjacent in time. These correlations further support the temporal correlation to great earthquake records in Willapa Bay (this session), as well as extend the record to ~ 9800 years. We find that many events can be correlated in this way between JDF, Cascadia, Hydrate Ridge (central Oregon margin), and Rogue (southern Oregon margin) core sites. The signatures vary more between sites that are not directly connected, but many robust features are preserved, including in many cases, the relative volume of the turbidite, and the number of coarse pulses in the turbidite. These features allow direct correlation between sites independent of other methods. That we are able to correlate physical property "wiggle" plots along channel systems is intriguing, given the expected chaotic nature of deposition controlled by turbid flow. However, the inter-site correlations between turbidite channels that are not connected, implies that the deposition process is influenced by more than long distance transport of event characteristics. We suggest that such correlation implies that something of the earthquake shaking signal may be contained in these records. While not all sites can be correlated on the basis of physical properties, we find that key events in the event sequence have characteristics observable at most if not all sites. Events T6, T8, and T16 are large triplets at all sites, T4 and T9 are single events, T10, T12 are small events at all sites, and T11 is a massive doublet event at all sites. These observations strengthen the correlation, and thus the inference of earthquake origin for these events. The correlations further support the un-segmented nature of most Cascadia ruptures, since direct correlation effectively eliminates this option. Currently, not all events can be correlated margin wide, leaving open the possibility of several segmented ruptures. Strengthened correlations further support a repeating pattern of Great Earthquakes in Cascadia. The pattern appears to have repeated at least four times, with the most recent AD 1700 event being the third of three events following a long interval between events T4 and T5.
T12B-04 11:05h
The giant subduction earthquakes of 1797 and 1833, West Sumatra: Characteristic couplets, uncharacteristic slip
Rupture of a large patch of the Sumatran subduction zone produced a giant earthquake in 1797. Roughly the same patch ruptured again in 1833, just 36 years later. The magnitude of slip in 1833 was several times greater than that in 1797. Two large earthquakes dominate the historical seismic record of the Sumatran subduction zone south of the Equator. In addition to anecdotal reporting of widespread strong shaking, a few reports of large tsunamis and uplift of the outer-arc islands, about 100 to 120 km northeast of the Sumatran trench, have also survived. We have used growth patterns and U-Th dating of coral microatolls to map in detail the vertical deformation associated with these two earthquakes. The patterns of deformation constrain models of slip on the subduction interface, which dips about 12 degrees northeastward about 25 km below the islands. Sipora, North Pagai and South Pagai Islands, which span a 160-km length of the outer-arc ridge, have corals that display evidence of uplift during the 1797 earthquake. Uplift ranges from zero to 70 cm, and a distinct northeastward tilt, away from the trench, is apparent. The same islands also have corals that record vertical deformation during the 1833 earthquake. Uplift values during that event range from 100 to 230 cm. As in the case of the 1797 deformation, these corals show a pronounced tilt away from the trench. Corals on neighboring Siberut Island, farther northwest, display no evidence of uplift or submergence during either the 1797 or 1833 earthquake. Our first models of these data, which assume rectangular elastic dislocations, uniform slip, and rupture almost to the trench, fit the data moderately well. They yield slips of ~3 m for the 1797 event and ~12 m for the 1833 event. Estimated Mw of the two events are about 8.4 and 8.7. For both events, rupture beneath all three islands is required to produce the observed uplift. The northwestern limit of both ruptures must be no more than 40 km northwest of Sipora (the northernmost of the three islands). The southeastern limit of the 1797 rupture is well constrained to beneath the southern part of South Pagai (the southernmost of the three). The southeastern limit of the 1833 rupture is poorly constrained, but extended well beyond South Pagai Island. The downdip limit of rupture in 1833 is appreciably deeper (farther northeast) than that of the 1797 event. The giant 1797 and 1833 earthquakes involved rupture of the same patch of the subduction interface beneath the islands, but slip extended farther downdip and southeast in 1833. Slip on the patch beneath the islands was also several times larger in the latter event, far more than could have accumulated in the 36 years between events. Thus it appears that slip magnitudes can vary by a factor of four or so on the same fault patch and that not all accumulated strain need be relieved in a giant earthquake. Paleoseismic evidence indicates that over the past millennium, the islands have risen during giant earthquakes or earthquake couplets about every 230 years. At least three of the episodes appear to have been couplets separated by just a few decades.
T12B-05 11:20h
Developing Tools for Submarine Earthquake Geology Along the North Anatolia Fault Zone in the Marmara Sea, Turkey
The late Holocene behavior of the North Anatolia Fault beneath the Marmara Sea is being investigated by using two complimentary approaches. First, earthquake ruptures are resolved in time and space along small fault basins (equivalent to sag-ponds on land) and larger transform basins. Second, Holocene deformation and slip rates are quantified by using datable horizontal makers, such as shorelines and turbidites, and piercing points, such as offset channels and slumps. High-resolution subbottom profiling (CHIRP) and multibeam bathymetry acquired on the shelf highlight the linear trace of the Ganos segment of the NAF connecting `sag pond'-like basins. Such small basins along a fault trace are choice localities for earthquake geology because they subside and fill rapidly and can thus accumulate a complete and decipherable record of local fault ruptures. Their drainages are often very restricted and they are thus isolated from other distal submarine ruptures, as well as input from fluvial and alluvial fan that may include weather-related events. Sedimentation on the floor of the larger, deeper basins is punctuated by erosional events that removed thousands of years of sedimentation. Cores from these basins preserve homogenites linked by 14-C dating to the 740AD earthquake, whose epicenter has been located in Cinarcik Basin [Ambrassey and Finkel, 1995]. Homogenites are thick deposits of reworked mud that overlay finely laminated sands and were probably deposited by earthquake-induced seiche-like currents. A complete record of sedimentation of the larger transform basins can be extracted. However, it is critical to sample the precise depocenters because rapid tilting can drastically reduce the area of active deposition during high sea level and low turbidite flux. Depocenters may be eccentric because tilting oblique to the border fault is typical of transform basins. Vertical displacement has been documented along the Imrali Shelf and Izmit Gulf. In Imrali the horizontal marker is an undeformed sedimentary layer containing fresh water mollusks, diatoms and foraminiferal assemblages that document the lacustrine to marine transition of Marmara Sea at 12ky BP. A vertical displacement of 40m offsets this horizon. We are evaluating if the displacement resulted from motion along the Imrali Fault and slumping, or if the level of the Marmara Lake was much lower than the 85m Dardanelle sill at the time of marine flooding. These two alternate explanations could be distinguished with additional coring Initial findings demonstrate that submarine earthquake geology methods can characterize the seismic and neotectonic activity along fault segments and has the potential for extending the earthquake records further back in time than in paleoseimic land investigations. Critical data include very high resolution bathymetry, closely spaced grids of seismic profiles (less than50m apart), high precision coring, long cores in the deepest part of a basin, closely-spaced transects of cores, and very good chronology.
T12B-06 11:35h
Active Normal Fault Behaviour and Continental Rift Geometry in the Corinth Rift, Greece
The Gulf of Corinth continental rift, central Greece extends at up to 15 mm/yr with regular M6+ earthquakes. However, rapid geodetic extension rates in the western rift cannot be accounted for by displacement on onshore faults alone, where slip rates determined from uplifted terraces and paleoseismological trenching are significantly lower. High resolution seismic reflection and multibeam bathymetric data were collected to survey offshore faults contributing to extension and quantify their displacement. In the western rift, a basement horst on the northern margin is uplifted by the N and S Eratini faults and the axial channel is fault-controlled. Subsided lowstand shorelines in the hangingwall of the N Eratini and the well-studied Aigion fault suggest that both faults have similar displacements. Summed extension from the four major faults across this part of the rift (Eliki, Sub-channel, S Eratini, N Eratini) is of the order of 8-12 mm/yr, thereby reconciling geologic and geodetic datasets. Geomorphology indicates that the rift geometry changes along axis, with a model of distributed deformation across multiple faults proposed for the western rift. The high resolution seismic data linked to sea level history within the gulf (isolated during lowstands) potentially allow changes in slip rate to be determined on a 10000 year timescale. These results compliment the often shorter (100's-10000's years) timescales of onshore fault trenching and uplifted terrace sequences in terms of temporal fault behaviour. Ultimately, seismic hazard can be refined based on fluctuations in past fault behaviour within the rift.
T12B-07 11:50h
Middle to Late Holocene Earthquake Chronologies for the Cascadia Subduction Zone From Two Estuaries in Southern Oregon
Comparisons of regional Cascadia earthquake records to earthquake chronologies at the Coquille River and Sixes River estuaries in southern Oregon suggest that, in contrast to the {\bf M} 9 A.D. 1700 earthquake, some events ruptured shorter segments of the subduction zone and failed to trigger offshore turbidites. Between 2,000 and 4,700 years ago, earthquakes occurred on average every 350 to 415 years at the Sixes River. Over a similar time period at the Coquille River estuary, earthquakes occurred on average every 525 to 650 years. When compared to the average recurrence interval of tsunamis and strong shaking reported for Bradley Lake (240 to 280 years), these different return periods at sites within 30 km of one another suggest segmented rupture of the subduction zone. Corroborative evidence for segmented rupture includes 2-sigma age ranges for coseismically subsided soils at the Sixes River that fail to overlap age ranges for soils at the Coquille River estuary, and regional comparisons that suggest the penultimate earthquake recorded in southwestern Washington did not rupture south of Coos Bay, Oregon. Five earthquakes that occurred between 3,700 and 4,700 years ago are alternately recorded at the Sixes River and the Coquille estuary and each earthquake correlates in time with evidence for a tsunami or strong shaking at Bradley Lake. The alternating pattern of this earthquake sequence suggests that rupture termination of adjacent segments coincide as an inferred segment boundary at Cape Blanco. The Cape Blanco anticline, a transverse structural high that separates the Coquille estuary from the Sixes River, is inferred to be a possible segment boundary because it coincides with the boundary between two forearc basins that may reflect separate long-term asperities. We attribute faster relative sea-level rise at the Coquille estuary, accelerated by 0.5 to 1.3 mm/yr compared to the Sixes River, to differential uplift accommodated by one or both upper-plate structures, the Coquille fault and the Cape Blanco anticline, that separate the sites. This Holocene uplift rate agrees with long-term uplift rates derived from marine terrace studies. Variations in relative sea-level curves indicate that at least two plate-boundary earthquakes triggered slip on a blind upper-plate fault that underlies the Cape Blanco anticline. Differential uplift of 0.4 to 1.7 m resulted from an earthquake 3,560 to 3,390 cal yr BP. Another earthquake produced 1.9 to 3.9 m of differential uplift about 2,470 to 2,150 cal yr BP.
T12B-08 12:05h
Holocene Turbidite Recurrence Frequency off Northern California: Insights for San Andreas Fault Paleoseismicity
Numerous turbidites along the northern California continental margin are influenced by the northern San Andreas Fault (SAF). The fault parallels the coast near San Francisco Bay and further north underlies the California margin. Multiple tributary slope canyons and proximal channels join downstream into large channels, and all systems are dominated by the deposition of turbidite silt and sand beds. Our research aims to: 1) test the hypothesis that synchronous turbidites along the margin result from turbidity currents triggered by great earthquakes on the SAF and 2) thus define a paleoseismic record. Most important, we want to outline the recurrence history of paleoseismic events. Several lines of evidence suggest that there is synchronous SAF triggering of turbidites. Channels below tributary confluences are characterized by many single-event turbidite beds with multiple coarse-grained sediment pulses that contain different mineralogy from tributary source canyons. The rate of turbidite bed deposition (number/m) above and below channel tributary confluences typically is the same and not additive in downstream channels. Geotek log signatures of turbidites from different channel systems correlate along the margin and our present limited number of 14C ages suggest correlative events. The most complete and reliable turbidite record is found in Noyo Channel where the canyon head source of turbidites is directly underlain by the SAF. The five youngest turbidite 14C ages of Noyo show general agreement with the SAF paleoseismic record on land.This apparent correlation suggests that Noyo Channel may provide a much longer paleoseismic record for 24 events on the SAF during the past 6,000 yr. We utilize multiple cores with 24 correlative turbidite events from the channel to define event recurrence time between turbidites. We base this time on two independent methods: 1) hemipelagic sediment thickness (H) between two consecutive turbidites (i.e. H/sedimentation rate =recurrence time)(24 events), and 2) 14C ages (i.e. difference in ages between two consecutive turbidites= recurrence time)( 8 events). The average recurrence time we find between events is 210 yr (H method) and 180 yr (14C age method). Both methods show a minimum recurrence time of 140 yr and a maximum time of 275 yr with 75 percent of the recurrence times between 150 to 225 years. With two major assumptions that 1) Noyo Channel turbidites represent great earthquakes on the SAF and 2) the Noyo recurrence pattern continues into the future, the Noyo recurrence data suggests that we are not yet in a window for another great earthquake on the northernmost SAF. This statement is based on present evidence indicating that in the Noyo Channel area: 1) an earthquake greater than 7.2 magnitude is necessary to trigger a turbidity current, 2) minimum recurrence times are140 yr, and 3) the great earthquake in 1906 triggered the youngest turbidite.