SCEC Award Number 14178 View PDF
Proposal Category Individual Proposal (Integration and Theory)
Proposal Title Empirical Evaluation of the Variance Associated with Multiple Ruptures of a Single Fault
Investigator(s)
Name Organization
John Anderson University of Nevada, Reno
Other Participants Gony Biran
SCEC Priorities 6e, 3c, 4a SCEC Groups GMP, CSEP, SIV
Report Due Date 03/15/2015 Date Report Submitted N/A
Project Abstract
 We estimate the variance in ground motions related to repeated large earthquakes occurring on the same fault segment with similar magnitudes. We find eight earthquake pairs for which suitable strong motion records exist. Two are crustal strike-slip earthquakes from California and six are subduction zone earthquakes from Japan. We consider only large earthquakes and deal with frequencies greater than the earthquake corner frequency, so the variability that is considered here is related to smaller scale differences in the rupture process, particularly on the part of the fault nearest the station. We find that the variance of the 5% damped spectral accelerations of these pairs, termed τ2F, averages to about 45% and 80% for the crustal and subduction zone earthquakes, respectively, of τ2, where τ2 is the contribution of source variability to the total variability of ground motion estimated by some recent ground motion prediction equations (GMPEs). We suggest that τ2F is lower than τ2, for the frequencies where τ2F is estimated, because it depends primarily on only local physical properties of a fault that are the same in repeated earthquakes. We therefore suggest that at sites where the hazard is controlled by a single re-rupturing source, one could potentially use a between-event variance that is smaller than τ2, in seismic hazard calculations. Thus these results may help to resolve the inconsistencies that are now present between the national hazard maps, and some precariously balanced rocks in southern California.
Intellectual Merit John Anderson and Jim Brune suggested, in a 1999 paper, the concept of a characteristic ground motion earthquake, in which repeated rupture on the same fault might tend to cause nearly identical ground motions. This project tests that idea with the best available strong motion data. This paper characterizes the differences between ground motions, recorded on the same station from repeated ruptures, with a standard deviation "tau_F". This parameter is best compared with the parameter "tau" that characterizes the uncertainty of ground motion estimates from the source in general, based on single ruptures from multiple faults, including multiple fault geometries and styles of faulting. We estimate that in the frequency range where it can be estimated, tau_F is much smaller than tau for large strike-slip earthquakes in California, and smaller than or equal to tau for repeated subduction zone earthquakes recorded in Japan. To our knowledge, this is the first time anyone has tried to estimate tau_F.
Broader Impacts The results can be used as a constraint on the input to probabilistic seismic hazard analysis. If these results are accepted and used by the larger hazard community, the result could have a significant impact on the outcome of these seismic hazard analyses. Since in the US alone the output of the US National Seismic Hazard Map influences about $1 trillion in construction every year, improvements of any sort have a large impact.
Exemplary Figure Figure 6. Between-event variance (τ2) for SA, as calculated by four NGA-WEST2 models that calculate τ (the gray shaded region), compared with the mean of the maximum likelihood estimate of variance for the trusted range of periods. The GMPEs variance is divided by 2, the factor between the mean of the two-point maximum likelihood estimate of the variance and the true variance. (a) Imperial Valley; (b) Parkfield; (c) Tokachi-oki; (d) Tokachi-oki and Sanriku-haruka-oki; (e) Miyagi-oki; (f) Ibaraki-oki; (g) Iwate-oki; and (h) Kamaishi-oki. The variance from NGA-WEST2 models is used because GMPEs derived from Japanese data do not separate the total variance into its component parts.