Deaggregation of Uncertainty in Seismic Hazard

Abstract eng:
The U.S. National Seismic Hazard Model (NSHM) is composed of multiple logic trees that capture the epistemic uncertainty inherent in any estimate of seismic hazard at a site. Logic trees and associated weights describe, but are not limited to, variability of fault geometry, maximum magnitude, and regional seismicity rate. They are also used to represent different models of ground motion. With each new edition of the NSHM, the number of logic trees and associated branches grows. While this does not pose insurmountable computational demands, the computation of mean hazard maps does grow proportionally, particularly when computing associated uncertainties on a map scale, and it is difficult to determine which branches are the most influential. It is also increasingly important that the NSHM program produce maps of uncertainty; to this end, any reductions in computation time are a benefit. Using the California component of the NHSM, the Uniform California Earthquake Rupture Forecast version 3 (UCERF3), as an example, we examine uncertainty when computing seismic hazard at a site. Whereas analyses of uncertainty commonly use Monte Carlo sampling to compute fractiles about a mean hazard curve, for this analysis we iterate over the entire logic tree. Although this approach maintains correlations between branches, we are able track individual logic tree branch choices through a hazard calculation that would otherwise be obscured. We then deaggregate the range of possible ground motions or return periods and use tornado diagrams and stacked histograms to illustrate the relative influence of each logic tree branch on hazard at a site. In the case of UCERF3, fault slip-rate and ground motion models are the most influential, but models of fault slip distribution and the maximum magnitude of background seismicity sources are not.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket