HR: 13:40h
AN: NG32A-01 INVITED     [PDF]
TI: Long-Term Prediction of Large Earthquakes: When Does Quasi-Periodic Behavior Occur?
AU: * Sykes, L R
EM: sykes@ldeo.columbia.edu
AF: Lamont Observatory, Columbia University, Rt. 9W, Palisades, NY 10964
AB: I argue that the prediction of large earthquakes for time scales of a few decades is possible for a number of fault segments along transform and subduction plate boundaries. A key parameter in ascertaining if forecasting is feasible is the size of the coefficient of variation, CV, the standard deviation of inter-event times of large earthquakes that rupture all or most of a given fault segment divided by T, the average repeat time. I address only large events, ones that rupture all or most of the downdip width of the seismogenic zone where velocity-weakening behavior occurs. Historic and paleoseismic data indicate that the segment that ruptured in the great 1946 Nankaido, Japan, earthquake broke 9 times in the previous 1060 years yielding T=118 years and CV=0.16. The adjacent zone that broke in 1944 exhibits similar behavior as does the Copper River delta, the site of 8 paleoseismic events dated by Plafker and Rubin (1994) above the rupture zone of the 1964 Alaska earthquake. Lindh (preceding abstract) finds that many fault segments in California have similar small values of CV. Paleoseismic data for inter-event times at Pallet Creek and Wrightwood, however, indicate a large CV. Those sites at situated along the San Andreas fault near the end of the 1857 rupture zone where slip was much smaller than in the Carrizo plain, rupture in large events to the northwest and southeast overlap and deformation is multibranched as plate motion is transferred in part to the San Jacinto fault. Plate boundary slip is confined to narrow zones along the 1944 and 1946 segments of the Nankai trough but is more diffuse in the Tokai-Suruga Bay region where the Izu Peninsula is colliding with the rest of Honshu and repeat times appear to be longer (and CV perhaps is larger). Dates of uplifted terraces likely give repeat times of inter-plate thrust events that are too long and large estimates of CV since imbricate faults within the upper plate that generate terraces do not rupture in every great earthquake. The 2002 Working Group on large earthquakes in the San Francisco Bay region followed Ellsworth et al. (1999) in adopting much larger values of CV for several critical fault segments and underestimating their likelihood of rupture in the next 30 years. The Working Group also gives considerable weight to a Poisson model, which is in conflict with both renewal processes involving slow stress accumulation and with values of CV near 0.2. The failure of the Parkfield prediction has greatly influenced views in the U.S. about long-term forecasts. The model of the repeated breaking of a single asperity is incorrect since past Parkfield shocks of about magnitude 6 likely did not rupture the same part of the San Andreas fault.
DE: 6620 Science policy
DE: 7209 Earthquake dynamics and mechanics
DE: 7215 Earthquake parameters
DE: 7223 Seismic hazard assessment and prediction
DE: 7230 Seismicity and seismotectonics
SC: Nonlinear Geophysics [NG]
MN: 2003 Fall Meeting