Conveners: Ned Field and the Working Group on California Earthquake Probabilities
Dates: May 8-9 2012
Location: Davidson Conference Center, University of Southern California, Los Angeles, CA

The preliminary UCERF3 Model will be formally reviewed by the Scientific Review Panel (SRP) on May 8-9, 2012. The meeting will include an overview of the preliminary model by Working Group, and a formal assessment by the SRP. The results of the review will be submitted to the California Earthquake Authority for their approval.

The primary focus of this meeting is a review the Grand Inversion methodology for defining long-term models, and to judge the viability of input elements.  Given this approach represents a major departure from past practice, plus the fact that interpreting results is also much more challenging, we will need the full meeting to focus on the inversion ingredients and results.  While the time-dependent components have been implemented in UCERF3, more time is needed to further evaluate their implications, and it does not make sense to do so until the long-term models are finalized (given that some tuning will be in order).  An exception is the Empirical model, which will be discussed at this meeting due to a tight coupling with earthquake-rate-model assumptions.

An explicit goal of this meeting is to establish consensus on a final set of logic-tree branches for the long-term model, the weighting of which will be determined by follow-up calculations and analyses.  Our final model, due to CEA by June 30th, will constitute a complete long-term model with proposed logic-tree branch weights.  This “final” model will then be reviewed by the SRP (and others) and revised accordingly.  The model will be also reviewed for use by the NSHMP at a public workshop in Oct. 2012. The extent to which the final model includes time dependence remains to be seen, but the WGCEP remains optimistic that it can deliver what was articulated in the project plan.

As described in the Preliminary Model Report, the Grand Inversion methodology is now up and running, and we can conduct runs for any and all of our presently defined logic tree branches.  The much bigger challenge is interpreting (and defending) the results of any one branch, mostly because we now have more than 200,000 ruptures defined in the fault system.  Indeed, the recent formal review raised several important questions with respect to the Grand Inversion.  To improve our understanding of the results, our recent efforts have focused on applying the methodology using UCERF2 ingredients, which avoids ambiguities of whether differences are caused by the new methodology or new data.  Considerable time will be spent presenting these results, with the goal of convincing the SRP that we can indeed construct a model that represents a clear improvement over UCERF2.

Consequently, the implications of many proposed logic-tree branches remain under-evaluated from the grand-inversion perspective, so their viability will therefore be discussed more from a theoretical and consistency-with-data perspective.  We acknowledge (and emphasize) that some branches that look good now may later be found incompatible with other inversion constraints (which itself reflects the power of the inversion methodology relative to previous approaches).