干旱区生态环境知识资源中心

Probabilistic seismic hazard analysis (PSHA) attempts to answer a simple question: at what rates are different levels of strong ground motion exceeded? The reliability of PSHA results depends on the quality of the input. Primary interest is in exceedance rates of less than 10-2 per year so instrumental verification is not possible. However, in some places precariously balanced rocks (PBRs), with ages of similar to 10(4) years, can be used to test the predictions of exceedance rates in this range of interest. A PBR is a type example of a fragile geological feature (FGF) that is suitable for testing PSHA. Another potentially valuable class of FGF is speleothems, since their preservation in caves is not dependent on climate.

In several important cases, tests on PBRs have indicated that the input to a PSHA requires improvements. PBRs in the deserts of the southwestern United States provide type examples for these tests, particularly in Nevada and southern California. This paper identifies several approaches to resolve these inconsistencies. These include modifying the ground motion prediction equations (GMPEs) used to estimate the hazard to reduce dependency on the ergodic assumption, truncating the large-amplitude tails of the GMPEs, revising the activity rates on individual faults, revising the background seismicity, revising the distribution of earthquake sizes on faults, use of the PBR points to adjust weights on logic tree branches, switching to scenario-based PSHA, and living with the discrepancies when collecting the data to eliminate them is cost prohibitive.

As the prediction of PSHA for any ground motion amplitude is the result of an integral over all possible earthquake sources, improvements in the input that resolve inconsistencies at any point on the hazard curve will affect the entire hazard curve, both at the site of the test and at nearby sites. Tests at low probabilities afforded by PBRs are particularly important for two distinct reasons. One is that the resolution of tests at high probabilities is low. The other is that low-probability hazards to critical facilities are very important. Thus we conclude that to gain confidence in PSHA inputs and statistical assumptions it is important to use PBRs to test hazard curves at low probabilities whenever data are available.