Simulated strong ground motions for magnitude 8 earthquakes on the cascadia subduction zone

Strong ground motions from subduction-zone thrust earthquakes in western Washington and Oregon are estimated using a semi-empirical simulation procedure. The procedure is validated for large subduction earthquakes by modeling recorded acceleration seismograms and response spectra from the magnitude (M) 8 Michoacan, Mexico, and Valparaiso, Chile, earthquakes of 1985. We find that slip-distribution models derived from strong-motion and teleseismic velocity seismograms of these two earthquakes also explain higher frequency motions of the recorded accelerograms. Quantitative measures of the misfit between recorded and simulated response spectra are used to estimate the modeling uncertainty. Ground motions are computed for two fault models representative of the two different subduction-zone geometries beneath Washington and Oregon. The most critical geometrical source parameter controlling ground motions in the urban regions of Puget Sound and Portland is the depth of the rupture on the plate interface. We used a geometry based on earthquake locations that places the downdip limit of rupture about 50 km west of both Seattle and Portland. A geometry in which the plate interface is arched at depths shallower than 40 km beneath western Washington would place the downdip limit beneath Seattle and result in larger ground-motion estimates in the Puget Sound area. Given the assumed fault location, the largest cause of uncertainty in the estimated ground motions is due to the distribution of slip (asperities) with depth on the fault plane. As their depth increases, the asperities approach the urban regions of Puget Sound and Portland and the ground motions increase. Also, with increasing asperity depth, the attenuation of peak acceleration with distance is more gradual. The fault dip of the Oregon model is about twice that of the Washington model, but the estimated ground motions are similar. The attenuation of horizontal peak acceleration (PGA) with distance r from the fault asperity is given by In(PGA)=15.6-3.34In(r+128)+0.79γ where 1n is the natural logarithm and γ is a site term (0 for rock and 1 for soil). This relation is appropriate for r greater than 25 km but less than 175 km and for M 8. When r is defined as the closest distance to the fault plane, the attenuation relation is 1n(PGA)=2.8-1.261n(r)+0.79γ This relation is appropriate for r greater than 30 km but less than 100 km and for M 8. Formal estimates of uncertainty in the calculated ground motions are obtained by considering both parametric uncertainty (estimated from the range of source models of hypothetical Cascadia subduction-zone earthquakes) and modeling and random uncertainty (estimated from the misfit between recorded and simulated ground motions of the Michoacan and Valparaiso earthquakes). For periods less than 1 s, the estimated response spectral velocities at soil sites in the Seattle-Olympia region are about twice those recorded during the 1949 M 7.1 Olympia and 1965 M 6.5 Seattle earthquakes, and the durations of strong motion are significantly longer (45-60 s versus 10-15 s for rock sites).