HR: 10:20h
AN: S31G-01    [PDF]
TI: Testing hypotheses of earthquake occurrence
AU: Kagan, Y Y
EM: ykagan@ucla.edu
AF: UCLA, Dept. Earth & Space Sciences 595 Young Drive East, Los Angeles, CA 90095-1567 United States
AU: * Jackson, D D
EM: djackson@ucla.edu
AF: UCLA, Dept. Earth & Space Sciences 595 Young Drive East, Los Angeles, CA 90095-1567 United States
AU: Schorlemmer, D
EM: DANIJEL@seismo.ifg.ethz.ch
AF: ETH Zurich, Institute of Geophysics ETH Hoenggerberg, HPP P, Zurich, CH-8093 Switzerland
AU: Gerstenberger, M
EM: MATT@seismo.ifg.ethz.ch
AF: ETH Zurich, Institute of Geophysics ETH Hoenggerberg, HPP P, Zurich, CH-8093 Switzerland
AB: We present a relatively straightforward likelihood method for testing those earthquake hypotheses that can be stated as vectors of earthquake rate density in defined bins of area, magnitude, and time. We illustrate the method as it will be applied to the Regional Earthquake Likelihood Models (RELM) project of the Southern California Earthquake Center (SCEC). Several earthquake forecast models are being developed as part of this project, and additional contributed forecasts are welcome. Various models are based on fault geometry and slip rates, seismicity, geodetic strain, and stress interactions. We would test models in pairs, requiring that both forecasts in a pair be defined over the same set of bins. Thus we offer a standard "menu" of bins and ground rules to encourage standardization. One menu category includes five-year forecasts of magnitude 5.0 and larger. Forecasts would be in the form of a vector of yearly earthquake rates on a 0.05 degree grid at the beginning of the test. Focal mechanism forecasts, when available, would be also be archived and used in the tests. The five-year forecast category may be appropriate for testing hypotheses of stress shadows from large earthquakes. Interim progress will be evaluated yearly, but final conclusions would be made on the basis of cumulative five-year performance. The second category includes forecasts of earthquakes above magnitude 4.0 on a 0.05 degree grid, evaluated and renewed daily. Final evaluation would be based on cumulative performance over five years. Other types of forecasts with different magnitude, space, and time sampling are welcome and will be tested against other models with shared characteristics. All earthquakes would be counted, and no attempt made to separate foreshocks, main shocks, and aftershocks. Earthquakes would be considered as point sources located at the hypocenter. For each pair of forecasts, we plan to compute alpha, the probability that the first would be wrongly rejected in favor of the second, and beta, the probability that the second would be wrongly rejected in favor of the first. Computing alpha and beta requires knowing the theoretical distribution of likelihood scores under each hypothesis, which we will estimate by simulations. Each forecast is given equal status; there is no "null hypothesis" which would be accepted by default. Forecasts and test results would be archived and posted on the RELM web site. The same methods can be applied to any region with adequate monitoring and sufficient earthquakes. If fewer than ten events are forecasted, the likelihood tests may not give definitive results. The tests do force certain requirements on the forecast models. Because the tests are based on absolute rates, stress models must be explicit about how stress increments affect past seismicity rates. Aftershocks of triggered events must be accounted for. Furthermore, the tests are sensitive to magnitude, so forecast models must specify the magnitude distribution of triggered events. Models should account for probable errors in magnitude and location by appropriate smoothing of the probabilities, as the tests will be "cold hearted:" near misses won't count.
DE: 3200 MATHEMATICAL GEOPHYSICS (New field)
DE: 3210 Modeling
DE: 7209 Earthquake dynamics and mechanics
DE: 7223 Seismic hazard assessment and prediction
SC: Seismology [S]
MN: 2003 Fall Meeting