S14C-01
Probabilistic Seismic Hazard Maps for Seattle, Washington, Based on 3D Ground-Motion Simulations
We have produced probabilistic seismic hazard maps for Seattle using over 500 3D finite-difference simulations of ground motions from earthquakes in the Seattle fault zone, Cascadia subduction zone, South Whidbey Island fault, and background shallow and deep source areas. The maps depict 1 Hz response spectral accelerations with 2, 5, and 10% probabilities of being exceeded in 50 years. The simulations were used to generate site and source dependent amplification factors that are applied to rock-site attenuation relations. The maps incorporate essentially the same fault sources and earthquake recurrence times as the 2002 national seismic hazard maps. The simulations included basin surface waves and basin-edge focusing effects from a 3D model of the Seattle basin. The 3D velocity model was validated by modeling several earthquakes in the region, including the 2001 M6.8 Nisqually earthquake, that were recorded by our Seattle Urban Seismic Network and the Pacific Northwest Seismic Network. The simulations duplicate our observation that earthquakes from the south and southwest typically produce larger amplifications in the Seattle basin than earthquakes from other azimuths, relative to rock sites outside the basin. Finite-fault simulations were run for earthquakes along the Seattle fault zone, with magnitudes ranging from 6.6 to 7.2, so that the effects of rupture directivity were included. Nonlinear amplification factors for soft-soil sites of fill and alluvium were also applied in the maps. For the Cascadia subduction zone, 3D simulations with point sources at different locations along the zone were used to determine amplification factors across Seattle expected for great subduction-zone earthquakes. These new urban seismic hazard maps are based on determinations of hazard for 7236 sites with a spacing of 280 m. The maps show that the highest hazard locations for this frequency band (around 1 Hz) are soft-soil sites (fill and alluvium) within the Seattle basin and along the inferred trace of the frontal fault of the Seattle fault zone. Sites on more consolidated soils within the Seattle basin generally exhibit higher hazard than those on similar soils outside the basin, because of basin surface waves and focusing of S-waves by the geometry of the edges and bottom of the basin.
S14C-02
Lessons from the Review of the PEGASOS Results: Limitations for the Applicability of the Traditional PSHA Methodology
The Swiss Nuclear Power Plants sponsored the PEGASOS project, a large scale Probabilistic Seismic Hazard Analysis (PSHA) following the SSHAC (Senior Seismic Hazard Analysis Committee) procedures (SSHAC, 1997) at their most elaborate form, a Level 4 analysis. After completion of the project in 2004 an independent review by the sponsor led to the rejection of the results based on the outcome of a suite of simple plausibility tests. A subsequent detailed analysis revealed some problems inherent to the PSHA-methodology, indicating the existence of some limitations of the methodology's applicability. These limitations have to be considered in practical decision making to avoid inappropriate design solutions. The limitations discussed are related to the following features of the traditional PSHA methodology: · the use of the Poissonian approximation for the superposition of the stochastic point processes of earthquake recurrence in different sources based on the theorem of Khintchine, · the summation of unequal statistical characteristics for the probability of exceedance of a specified ground motion level, · the hazard deaggregation based on spectral accelerations. It is demonstrated that the hazard frequency (sometimes related to the return period for ground motion at a given site), for which the method can be applied, is constrained because of these features. It shall not be extrapolated to the range of low frequencies as it has been tried in the PEGASOS study. It is shown that the method causes problems with respect to energy conservation principles and an appropriate choice of scenario earthquakes based on the hazard deaggregation. The frequently observed statistical importance of low magnitude near site earthquakes is shown to be a mathematical artefact of the methodology.
S14C-03
Statistical Assessment of Earthquake Recurrence for the W-HV Segment of the Wellington Fault, New Zealand
New paleoseismic results from a data-rich trench on the Wellington-Hutt Valley (W-HV) segment of the Wellington Fault have allowed for a more robust statistical approach into earthquake recurrence on that fault. The paleoseismic data is used to generate Monte Carlo simulations of the event age and recurrence interval distributions. Based on a combination of stratigraphic (peats buried by colluvium/scree) and structural (faulting) evidence we recognise up to 4 dateable, co-seismic events in the Te Kopahou-1 trench, at the south coast of Wellington. We have developed 6 different earthquake event age models that draw increasingly on trench interpretations, data from previous trenches, and/or less certain data in order to extend the length of the paleoseismic record. In addition, for 4 of these examples we have run simulations in parallel with a Slip Rate-SED model of recurrence interval for the fault, using a lognormal distribution of a c. 140 kyr slip rate (6-7.6 mm/yr) in combination with a single event displacement (SED) of 4.2 +/- 1.3 m (i.e. SED with CoV of 0.3). We present here two variations of the results from 20,000 random number results; in which both of the distributions include the effects of the Slip Rate-SED model. The first, which includes the last 3 dated paleo- earthquake events with some interpretative weighting gives a mean, median, and 95% lower and upper bounds of 641, 605, 332, and 1107 yr, respectively for earthquake recurrence. The second model includes the last 4 dated paleo-earthquake events (3 inter-event times) with increased interpretive weighting. This model yields recurrence times with parameters of 625, 591, 352, and 1074 yr, respectively. These same 2 models run using the paleoearthquake data alone yield mean intervals of c. 903 and 742 yr based on 3 and 4 events, respectively. We recognise that the recurrence interval distributions are lognormal in form (as is the Slip Rate model). The Slip Rate distribution, the CoV on SED, and the paleoearthquake record each have significant uncertainties that drive the recurrence interval results. The beauty of our technique is that all of these uncertainties are taken into account. Further work will be undertaken to improve these models to output the most robust recurrence model for New Zealand's highest seismic risk fault segment.
S14C-04
Source Modeling of Subduction-Zone Earthquakes and Long-Period Ground Motion Validation in the Tokyo Metropolitan Area
The national seismic hazard map of Japan indicates 30 year probability in the Tokyo metropolitan area controlled by megathrust earthquakes along the Philippine sea plate. Four major subduction-zone earthquakes are Kanto (Mw 7.9), northern Tokyo-bay (Mw 7.0 or greater), Tokai (Mw 8.0), and Tonankai (Mw 8.2) earthquakes. We have experienced the 1923 Kanto and 1944 Tonakai earthquakes, however the rest of two are hypothetical earthquakes. Source modeling and realistic ground-motion prediction for the earthquakes are quite important for disaster mitigation and hazard assessment in the Tokyo metropolitan area. We have carried out ground motion validation for the Kanto earthquake [Miyake et al., 2005] using physics-based source model along the new geometry of the plate, and 3D velocity model based on geophysical surveys. We here present ground motion validation for the Tokai and Tonankai earthquakes under the framework of the Special Project for Earthquake Disaster Mitigation in Urban Areas (2002-2006) and the Special Project for Earthquake Disaster Mitigation in Metropolitan Tokyo ‚`rea (2007-2011). The source process of the 1944 Tonankai earthquake was inferred from strong-motion and teleseismic data by Ichinose et al. [2003] and Yamanaka [2004]. For the hypothetical Tokai earthquake, Matsumura [2002] proposed locked zones figured out from seismicity which could be candidates for future asperities. Using the locations of the asperities, we constructed a characterized source model consisting of asperities and background area. The size and slip of the asperities are constrained by the source scaling of asperities for subduction-zone earthquakes, where the scaling is based on the compilation of past slip inversion results. The stress drop was adjusted so that asperities for long-period ground motions behave as strong motion generation areas for short- period ground motions. We adopted the 3D velocity model beneath the Tokyo metropolitan area constructed by integrating refraction, reflection, borehole, microtremor, and gravity data as well as ground motion spectra [e.g., Tanaka et al., 2005]. This model is upgraded with the velocity models of plate and crustal structures as well as subsurface-shallow structure. Ground motion validation for the Tokai and Tonankai earthquakes are performed. Long-period components are simulated by the FDM, and short-period ones by the stochastic Green's function method with the site amplification factors. Broadband ground motion time histories are reproduced by the hybrid method with the matching period of 3 s. We confirmed the distribution of simulated long-period ground motions are well reflected to the basement depth of the velocity structure, and oceanic structure with the sedimentary wedge which contributes to further propagation of the long-period ground motions. Even the seismic intensity is limited to 4 or 5- in the JMA scale, long-period ground motions are significantly developed in the Kanto basin. This indicates significant impact of long-period ground motion from the future subduction-zone earthquakes against the Tokyo metropolitan area. http://www.eri.u-tokyo.ac.jp/hiroe/
S14C-05
Predictability of Ground Motion: Simulations for the 2003 Tokachi-oki-like Earthquake
We have conducted a joint simulation from the plate subduction to the generation of seismic waves through the earthquake dynamic process. To make a practical estimation of seismic hazard, it is important to evaluate source effects on ground motions based on several earthquake rupture scenarios because the source processes such as rupture directivity and asperity locations leave huge uncertainties comparing to the crustal structures. In this presentation, we focus on the ground motion simulation using the finite difference method with a 3-D crustal structure [Aoi and Fujiwara, 1999] for the 2003 Tokachi-oki earthquake, which occurred in a subduction zone off Hokkaido, Japan. We examined five rupture scenarios to evaluate the rupture directivity effect under a set of dynamic parameters which is given to reproduce the 2003 event. Rupture models are computed by the integral equation method where initial stress and constitutive relation are given by the simulation of plate subduction. The rupture nucleations are assumed, respectively, at a hypocenter of this earthquake in model S, at the shallowest point on the fault area in model A, at its deepest point in model B, at its western end in model C and its eastern end in model D. In all models, seismic moments are the same, which is constrained by the amount of stress drop allowed in the dynamic models. Generally, we observe a clear correlation between amplification of ground motions and the basin structure. In model S, the synthetic seismograms appear to reasonably reproduce the observations. We could observe about factor of 2 differences as the source effect between the models, which is caused by the different rupture scenarios, and the site effect amplifies them at about factor of 5. These variations should be important for the seismic hazard estimation.
S14C-06
3D Spectral Element Method Simulations Of The Seismic Response of Caracas (Venezuela) Basin
We present here 3D numerical simulations of the response of the Caracas (Venezuela) valley up to 5 Hz for different scenarios of plane wave excitation based on the regional seismicity. Attention is focused on the effects of the 3D basin geometry and of the adjacent regional topography. The simulations are performed using Spectral Element method (SEM) together with an unstructured hexahedral mesh discretization and perfectly matched layers (PML). These simulations show 3D amplification phenomena associated with complex wave reflexion, diffraction and focalisation patterns linked to the geometry of the basin. Time and frequency analysis reveal some interesting features both in terms of amplification and energy residence in the basin. The low frequency amplification pattern is mainly controlled by the early response of the basin to the incident plane wave while the high frequency amplification patterns result mainly from late arrivals where complex 3D wave diffraction phenomena are dominating and the memory of the initial excitation is lost. Interestingly enough, it is shown that H/V method correctly predict the low frequency amplification pattern when apply to the late part of the recorded seismograms. The complex high frequency amplification pattern is shown to be associated with surface wave generation at, and propagation from, sharp edges of the basin. Importance of 3D phenomena is assessed by comparison with simple 2D simulations. Significant differences in terms of time of residence, energy and amplification levels point out the interest of complete 3D modeling. In conclusions some of the limitations associated with the use of unstructured hexahedral meshes will be adressed. Despite the use of unstructured meshing tool, modeling the geometry of geological basins remain a complex and time consuming task. Possible extensions using more elaborate techniques like non conforming domain decomposition will be also discussed in conclusion.
S14C-07
Compaction Can Limit Peak Vertical Velocity at Yucca Mountain
Compaction is irreversible decrease of volume of a porous material subject to compressive stress, and it is an effective mechanism to attenuate strong P waves. Compaction of porous unsaturated tuffs at Yucca Mountain can occur if compressive stress reaches a level never reached before since the deposition of the tuff. Compaction can place a physical limit on peak vertical ground velocity at the site of the proposed radioactive waste repository. A constitutive model for compaction is formulated that fits laboratory data on porous sandstone from Wong et al. (JGR 102, 3009-3025, 1997). In stress space, yielding occurs on a Mohr-Coulomb surface at low effective pressure and on an elliptical end cap at higher effective pressure. Stress in the model scales with P*, the effective pressure at crushing under hydrostatic loading, which depends on porosity, grain size, cementation, and other grain-scale properties of the sandstones. The data are fit with a non-associative flow rule, in which plastic strain has a compactive volumetric component at effective pressure greater than 0.1P* and a dilatant component at effective pressure less than 0.1P*. The model is used in dynamic calculations of normal-faulting earthquakes on the Solitario Canyon fault with complete stress drop through the seismogenic depth of the crust and with strong directivity of P waves toward the repository. Compaction occurs in the very porous Calico Hills Tuff unit below the repository. If the yield surface is chosen to be only slightly above the initial stress state in the Calico Hills Tuff, then peak vertical velocity at the repository is limited to 3.6 m/s. In that case, the value of P* used for tuff is 0.3 times P* for sandstone with equal porosity. More laboratory data are needed on compaction of tuffs at Yucca Mountain to establish the velocity limit.
S14C-08
Developing a physics-based finite source characterization tool for strong ground motion simulation of large earthquakes
Accurate prediction of the intensity and variability of near-field strong ground motion for future large earthquakes strongly depends on our ability to simulate realistic earthquake source models for those events. We developed a pseudo-dynamic source modeling method that generates kinematic, but physically self-consistent, finite-source models that include important characteristics of dynamic rupture. This approach was first initiated by Guatteri et al. (2004) for moderate size events and we extend its magnitude coverage to larger strike-slip crustal earthquakes (M > 7.2). We base the new pseudo-dynamic source models on 15 spontaneous dynamic rupture models that were constructed with different slip realizations and hypocenter locations for large, Mw 7.5, strike-slip events. These earthquakes have very long and narrow rupture dimensions (150 km long and 15 km wide), which leads to substantially different source scaling behavior than the smaller events used by Guatteri et al. (2004). We also allow greater latitude for supershear rupture in the present study, because it has been observed recently for a large strike-slip earthquakes (Bouchon et al., 2001; Bouchon and Vallèe, 2003; Dunham and Archuleta, 2004; Das, 2007; Song et al., 2008). Supershear rupture is widely observed in our modeling, and we find that instantaneous (local) rupture velocity correlates strongly with local slip amplitude. We also find that the limited fault width of these long, narrow ruptures exerts a critical control on rupture behavior as predicted by Day (1982). Our improved pseudo-dynamic modeling method can be used to generate realistic finite-source earthquake models for simulating near-field ground motions from large strike-slip events. This will help us to improve our limited knowledge about near-field ground motion characteristics because of the lack of observed data.