HR: 08:45h
AN: S41D-04 INVITED [Abstracts]
TI: Evaluation Techniques for Alarm-Based Earthquake Forecasts
AU: * Zechar, J D
EM: zechar@usc.edu
AF: University of Southern California, Department of Earth Sciences
3651 Trousdale Pkwy, Los Angeles, CA 90089
United States
AU: Jordan, T H
EM: tjordan@usc.edu
AF: University of Southern California, Department of Earth Sciences
3651 Trousdale Pkwy, Los Angeles, CA 90089
United States
AB:
When earthquake forecasts are stated in terms of conditional intensity, a number of well-known methods are available for
hypothesis testing. For example, the Regional Earthquake Likelihood Model (RELM) group of the Southern California Earthquake
Center (SCEC) will soon begin systematic, simultaneous testing of a number of algorithms that forecast earthquake intensity.
The likelihood framework, however, cannot be directly applied to prediction methods that generate alarms rather than
probabilistic forecasts. Because a number of investigators are using alarm-based methods in earthquake prediction
experiments, it is important that they be evaluated and, when possible, compared. Methods for evaluation of deterministic
binary forecasts have been developed from work related to signal detection, weather forecast verification (Jolliffe and
Stephenson 2003), information theory (Jones and Jones 2003), and economic losses (Molchan 1990, Molchan and Kagan 1992). In
this presentation, we provide an overview of the available techniques and highlight theoretical and practical difficulties.
We illustrate the use of these methods by application to two recent pattern recognition prediction algorithms: Reverse
Tracing of Precursors (Keilis-Borok et al 2004) and Pattern Informatics (Tiampo et al 2002, Rundle et al 2003).
DE: 3238 Prediction (3245, 4263)
DE: 3245 Probabilistic forecasting (3238)
DE: 7209 Earthquake dynamics (1242)
DE: 7223 Earthquake interaction, forecasting, and prediction (1217, 1242)
DE: 7230 Seismicity and tectonics (1207, 1217, 1240, 1242)
SC: Seismology [S]
MN: Fall Meeting 2005