SCEC Award Number 17134 View PDF
Proposal Category Individual Proposal (Integration and Theory)
Proposal Title Testing Fault-Based Earthquake Models
Investigator(s)
Name Organization
David Jackson University of California, Los Angeles Yan Kagan University of California, Los Angeles
Other Participants 1 SCEC Intern, not yet named
1 UCLA undergrad student, unnamed
SCEC Priorities 5c, 2e, 5d SCEC Groups CISM, CSEP, EFP
Report Due Date 06/15/2018 Date Report Submitted 06/14/2018
Project Abstract
 The paleoseismic hiatus in California is statistically inconsistent with the rate of events estimated from paleoseismic data, almost all of which were estimated before seismic networks began operating in the early 20th century. Hypotheses for the discrepancy were extreme luck, an unknown (and counter-intuitive) physical mechanism for coordinating rupture statewide, and over-estimation of earthquake rate by including non-earthquake events before seismic networks could verify true seismic events. Each of these hypotheses has quite different implications for use of paleoseismic data in earthquake forecasts. We’ve discovered that a similar hiatus occurred in New Zealand. Explaining the joint observations for California and New Zealand pretty much rules out the first two hypotheses, implicating over-estimation.

We’ve tested both the 1988 and 1995 WGCEP forecasts retrospectively and prospectively using modifications of the CSEP N-test and S-test. The 1988 forecast is for large earthquakes on the major faults, assuming quasi-periodic characteristics on prescribed fault segments. Since 1988 only the Parkfield earthquake of 2004 matches the prescribed segment description, and the forecast fails the N-test at 95% confidence. If the segment boundaries are retrospectively adjusted to match the rupture of the 1989 Loma Prieta Earthquake, the N-test is barely met at 95%. The forecast is consistent with the S-test of earthquake location. The 1995 WGCEP associates each earthquake with one of 65 seismotectonic zones, and includes both a time dependent and time independent forecast. Both significantly over-predict the number of expected events after 1995, and the time-dependent forecast significantly over-predicts the number from 1932 till now.
Intellectual Merit Our project explores the extent to which prior earthquake forecasts, based on common assumptions, agree with earthquake occurrence before and after the forecasts. Thus is contributes to the understanding of earthquake processes, a major goal of the Earthquake Forecasting and Prediction project in SCEC. We also developed methods for associating finite ruptures on faults with prescribed descriptions of such events, a necessary step in evaluation forecasts of such events, and we tested some of these ideas on forecasts of the Working Group on California Earthquake Probabilities (WGCEP). Association and testing of finite ruptures have been major goals of the Collaboratory for Study of Earthquake Predictability (CSEP).

Broader Impacts Our research affects seismic hazard estimation, which has implications for public safety, risk management, and public policy. We’ve presented our results to students and faculty at universities, including UCLA, UCR, ETH Zurich, Victoria Univerity Wellington, and at the 2017 SCEC Annual meeting, showing how Physics, Geology, Geodesy, and Statistics can be combined to solve real world problems. We’ve discussed our analysis with students and faculty in Statistics, showing them applications and encouraging their participation in seismological research.
Exemplary Figure Figure 1. Earthquakes after 1932 occurring within the 65 "seismo-tectonic" zones used in the 1995 WGCEP report. There are 28 events over magnitude 6, but only three (Hector Mine 1999, San Simeon 2003, and Parkfield 2004) occurred from 1995 to 2018. The 1995 WGCEP time-dependent model forecast over 13 such events, and the time independent model forecast over 9 such events.