SCEC Award Number 11117 View PDF
Proposal Category Collaborative Proposal (Integration and Theory)
Proposal Title Final Evaluation of the Regional Earthquake Likelihood Models Experiment
Name Organization
Maximilian Werner University of Southern California J. Douglas Zechar Danijel Schorlemmer
Other Participants Liukis, Maria
SCEC Priorities A6, A4, A10 SCEC Groups CSEP, WGCEP, EFP
Report Due Date 02/29/2012 Date Report Submitted N/A
Project Abstract
The objective was to obtain final results for the Regional Earthquake Likelihood Models (RELM) five-year, prospective earthquake forecasting experiment, which ended January 1, 2011. We measured the success of the seventeen submitted forecasts in forecasting the number of earthquakes, their magnitudes and locations, and compared the performance amongst each other. Sample results are: a smoothed seismicity model that adaptively smoothed locations of M2+ earthquakes since 1981 performed best, GPS-based forecasts also contained some predictive skill, but a physics-based earthquake simulator did poorly.
Intellectual Merit The RELM experiment set a new standard for how to evaluate hypotheses of earthquake occurrence. The results we obtained during this project showed that, among the submitted forecasts, the best predictive skill for M5+ earthquakes during the 2006-2011 period (including the M7.2 El Mayor Cucapah event) came from smoothing the locations of M2+ earthquakes since 1981. This might indicate further evidence for the importance of small earthquakes. The project also led to new insights into how to conduct earthquake forecasting experiments, including how to specify and evaluate a forecast.
Broader Impacts The RELM experiment serves as a blueprint for many CSEP experiments in California and around the world (New Zealand, Japan, Italy, China). The lessons from the RELM experiment are therefore very relevant for current and future CSEP endeavors. Most importantly, the results and lessons from RELM are expected to influence the scientific consensus for how to assess seismic hazard (WGCEP), which in turn sets building codes, policy and insurance rates.
Exemplary Figure Figure 1. Results of consistency tests for mainshock forecasts. Red points indicate that the forecast failed the test. Horizontal black lines delimit the 95% confidence “pass” region. a) Results of the two-sided N-test, where the forecast number is indicated as the middle vertical dash on each horizontal line, and the observed number is indicated by the filled circle. The circle is not in the same place for all forecasts due to masking. b-d) Results of the one-sided L (S,M)-test.

Zechar, Schorlemmer, Werner, Gerstenberger, Rhoades and Jordan (2012), in preparation.