Trimming the UCERF2 hazard logic tree
Keith A. Porter, Edward H. Field, & Kevin R. MilnerPublished 2012, SCEC Contribution #2074
There are many ways to explore the relative importance of the various modeling uncertainties of the UCERF2 earthquake rupture forecast (WGCEP 2007) for the state of California. We set out to better understand which uncertainties matter most to societal economic risk, and performed a tornado-diagram analysis of the sensitivity of expected annualized loss (EAL) experienced by an estimated statewide portfolio of woodframe single-family dwellings. Here, EAL refers to expected value of repair cost for the buildings alone.
We estimated EAL for 1,920 combinations of UCERF2 modeling choice (480 branches) and 4 NGA relationships, using a program we created called the Portfolio EAL Calculator, an extension of the OpenSHA software. The tornado-diagram analysis shows that the uncertainty in the expected annualized loss (EAL) among all California woodframe single-family dwellings is dominated by the choice of recurrence model; in the case of Brownian passage-time model, the aperiodicity parameter value; the choice of ground-motion-prediction equation, and the choice of magnitude-area relationship. They represent 40 of the 1,920 branches, when one fixes all the others at a value arbitrarily selected from the available options.
Treating EAL as a random variable that varies with UCERF2 modeling choices and ground-motion-prediction equation, two probability distributions of EAL were calculated: one considering all 1,920 branches and another where only the key uncertainties vary. The two distributions have the same mean and coefficient of variation, and pass a Kolmogorov-Smirnov goodness-of-fit test, which suggests they come from the same underlying distribution. The remaining epistemic uncertainties, while scientifically important and potentially material to local seismic hazard, local seismic risk, and perhaps low-probability, high-loss events, only modestly affect statewide EAL experienced by single-family woodframe dwellings or by the statewide building stock as a whole.
One implication is that, if one is examining statewide EAL and wishes to save computational time, the top 3 epistemic uncertainties alone sufficiently represent uncertainty in hazard. Considering them alone would reduce the complexity of the logic tree by 50x, from 1,920 to 40 branches: 5 combinations of recurrence model and aperiodicity (counting “empirical” only once among recurrence models), 4 ground-motion prediction equations, and 2 magnitude-area relationships. The reduction in model complexity and computation effort is not accompanied by a loss in uncertainty or a bias in expectation, compared with a model that considers all 1,920 branches.
Another implication is that, for a user of risk information concerned with statewide EAL, gathering more knowledge of any of the key epistemic uncertainties could be more effective in reducing uncertainty than exploring the other sources of uncertainty. These conclusions are unchanged if one uses either the leading expert-opinion-based alternative vulnerability model (ATC-13) or a recent empirical one (Wesson et al. 2004), or if one uses an alternative model of Vs30, namely Wald and Allen (2007), or if one uses a statewide portfolio representing all buildings rather than just woodframe single-family dwellings.
Citation
Porter, K. A., Field, E. H., & Milner, K. R. (2012). Trimming the UCERF2 hazard logic tree. Seismological Research Letters, 83(5), 815-828.
