Validation and improvement of probabilistic seismic hazard analysis models using precariously balanced rocks
File(s)
Author(s)
Rood, Anna Hope
Type
Thesis
Abstract
Globally, earthquake hazard estimates of large, infrequent, and potentially damaging earthquakes are highly uncertain and untested at timescales and ground motions beyond limited instrumental recordings of historical earthquakes. A fundamental issue is that there is currently no accepted method to empirically validate and refine earthquake hazard estimates at the timescale of these large, infrequent earthquakes. This lack of validation causes large uncertainty about the potentially life-threatening hazard to the population and associated infrastructure for millions of people. Here, I establish an empirical validation method at the desired timescales using the observed survival of precariously balanced rocks despite thousands of years of prehistoric earthquakes in seismically active regions. In each study in this thesis, I calculated the probability of toppling of a suite of precariously balanced rocks at defined ground-motion levels and determined the age at which the precariously balanced rocks obtained their current fragile geometries using a novel implementation of cosmogenic 10Be surface exposure dating.
First, I established a new earthquake hazard validation method that used empirical data from precariously balanced rocks preserved in coastal Central California at a site with similar site and source characteristics to the Diablo Canyon nuclear Power Plant. By eliminating the hazard estimates inconsistent with at least a 5% probability of precariously balanced rock survival, the mean ground-motion estimate corresponding to the hazard level of 10-4 yr-1 (10,000 year mean return period earthquake) were significantly reduced by 27%, and the range of estimated 5th-95th percentile ground motions reduced by 49%. Therefore, the hazard estimates for future earthquakes at a site relevant to the safety and security of this nuclear facility can be made with twice the certainty.
Second, I validated earthquake hazard estimates across southern California computed by the OpenQuake engine using the UCERF3 seismic source characterization and the NGA-West2 ground-motion models. The unexceeded level of ground shaking recorded by 20 precariously balanced rocks ranging in age from ~1-50 ka were used to investigate the relative performance of each NGA-West2 ground motion model. I showed that none of the NGA-West2 ground-motion models estimated levels of earthquake ground shaking consistent with the observed survival of all 20 precariously balanced rocks. Moreover, my analysis showed that the NGA-West2 model I14 generated hazard estimates most inconsistent with the precariously balanced rock survival, and it was, therefore, invalidated and removed. At a 2475 year mean return period, the removal of this invalid ground-motion model resulted in a 2-7% reduction in the mean and a 10-36% reduction in the 5th–95th fractile uncertainty of the ground-motion estimates across southern California. Therefore, precariously balanced rocks provide a previously unavailable empirical ground-motion model selection criteria for future earthquake hazard studies.
Third, I validated the earthquake hazard estimates for the South Mojave section of the San Andreas fault against the independent observations of precariously balanced rock survival over the timescales of rare, large prehistoric San Andreas fault earthquakes. The hazard estimates that were validated were generated by my own site-specific, partially non-ergodic hazard model using state-of-the-art seismic source and ground-motion characterizations. These validated hazard estimates were reweighted using Bayesian updating methods and the probability of the precariously balanced rock survival to produce a dramatically different, precariously balanced rock-informed suite of earthquake hazard estimates. Specifically for the 10,000-year mean return period earthquake, the mean ground shaking estimate was significantly reduced 65% and the 5th-95th fractile uncertainty range reduced by 72%. The magnitude of this inconsistency between the original hazard estimates and the precariously balanced rock survival revealed the need for major reconsideration of both seismic source and ground-motion characterizations.
Overall, this thesis establishes the immense value of utilizing precariously balanced rocks as an earthquake hazard validation tool. This research provides undeniable evidence of a significant inconsistency between current earthquake hazard estimates and independent precariously balanced rock survival data across California. The validation and reweighting of earthquake hazard estimates, informed by the probability of the observed precariously balanced rock survival, significantly reduces both the mean hazard estimate and the hazard estimate uncertainty. The hazard estimates inconsistent with the precariously balanced rock survival provide previously unavailable information about the relative performance of the hazard model input parameters that describe the cause and effect of earthquake processes. Such insights into the cause and effect of earthquakes make it possible to more reliably assess the safety and security of infrastructure ranging from domestic houses to critical facilities in earthquake-prone regions worldwide.
First, I established a new earthquake hazard validation method that used empirical data from precariously balanced rocks preserved in coastal Central California at a site with similar site and source characteristics to the Diablo Canyon nuclear Power Plant. By eliminating the hazard estimates inconsistent with at least a 5% probability of precariously balanced rock survival, the mean ground-motion estimate corresponding to the hazard level of 10-4 yr-1 (10,000 year mean return period earthquake) were significantly reduced by 27%, and the range of estimated 5th-95th percentile ground motions reduced by 49%. Therefore, the hazard estimates for future earthquakes at a site relevant to the safety and security of this nuclear facility can be made with twice the certainty.
Second, I validated earthquake hazard estimates across southern California computed by the OpenQuake engine using the UCERF3 seismic source characterization and the NGA-West2 ground-motion models. The unexceeded level of ground shaking recorded by 20 precariously balanced rocks ranging in age from ~1-50 ka were used to investigate the relative performance of each NGA-West2 ground motion model. I showed that none of the NGA-West2 ground-motion models estimated levels of earthquake ground shaking consistent with the observed survival of all 20 precariously balanced rocks. Moreover, my analysis showed that the NGA-West2 model I14 generated hazard estimates most inconsistent with the precariously balanced rock survival, and it was, therefore, invalidated and removed. At a 2475 year mean return period, the removal of this invalid ground-motion model resulted in a 2-7% reduction in the mean and a 10-36% reduction in the 5th–95th fractile uncertainty of the ground-motion estimates across southern California. Therefore, precariously balanced rocks provide a previously unavailable empirical ground-motion model selection criteria for future earthquake hazard studies.
Third, I validated the earthquake hazard estimates for the South Mojave section of the San Andreas fault against the independent observations of precariously balanced rock survival over the timescales of rare, large prehistoric San Andreas fault earthquakes. The hazard estimates that were validated were generated by my own site-specific, partially non-ergodic hazard model using state-of-the-art seismic source and ground-motion characterizations. These validated hazard estimates were reweighted using Bayesian updating methods and the probability of the precariously balanced rock survival to produce a dramatically different, precariously balanced rock-informed suite of earthquake hazard estimates. Specifically for the 10,000-year mean return period earthquake, the mean ground shaking estimate was significantly reduced 65% and the 5th-95th fractile uncertainty range reduced by 72%. The magnitude of this inconsistency between the original hazard estimates and the precariously balanced rock survival revealed the need for major reconsideration of both seismic source and ground-motion characterizations.
Overall, this thesis establishes the immense value of utilizing precariously balanced rocks as an earthquake hazard validation tool. This research provides undeniable evidence of a significant inconsistency between current earthquake hazard estimates and independent precariously balanced rock survival data across California. The validation and reweighting of earthquake hazard estimates, informed by the probability of the observed precariously balanced rock survival, significantly reduces both the mean hazard estimate and the hazard estimate uncertainty. The hazard estimates inconsistent with the precariously balanced rock survival provide previously unavailable information about the relative performance of the hazard model input parameters that describe the cause and effect of earthquake processes. Such insights into the cause and effect of earthquakes make it possible to more reliably assess the safety and security of infrastructure ranging from domestic houses to critical facilities in earthquake-prone regions worldwide.
Version
Open Access
Date Issued
2022-03
Date Awarded
2022-09
Copyright Statement
Creative Commons Attribution-NonCommercial Licence
License URL
Advisor
Stafford, Peter
Latham, John-Paul
Publisher Department
Civil and Environmental Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)