Assessing the value of removing earthquake-hazard-related epistemic uncertainties, exemplified using average annual loss in California

To aid in setting scientific research priorities, we assess the potential value of removing each of the epistemic uncertainties currently represented in the US Geological Survey California seismic-hazard model, using average annual loss (AAL) as the risk metric of interest. Given all the uncertainties, represented with logic-tree branches, we find a mean AAL of $3.94 billion. The modal value is 17.5% lower than the mean, and there is a 78% chance that the true AAL value is more than 10% away from the mean, and a 5% chance that it is a factor 2.1 greater or lower than the mean. We quantify the extent to which resolving each uncertainty improves the AAL estimate. The most influential branch is one that adds additional epistemic uncertainty to ground motion models, but others are found to be influential as well, such as the rate of M >= 5 events throughout the region. We discuss the broader implications of our findings, and note that the time dependence caused by spatiotemporal clustering can be much more influential on AAL than the epistemic uncertainties explored here.