HR: 16:10h
AN: S44B-01 INVITED     [Abstracts]
TI: Creating a Virtual Seismologist for Seismic Early Warning
AU: * Heaton, T H
EM: heaton_t@caltech.edu
AF: Caltech, MS 104-44 Caltech, Pasadena, CA 91125
AU: Cua, G
EM: gcua@uprm.edu
AF: Univ. of Puerto Rico, Mayaguez, PO BOX 9041, Mayaguez, PR 00681-9041
AU: Yamada, M
EM: masumi@caltech.edu
AF: Caltech, MS 104-44 Caltech, Pasadena, CA 91125
AB: Earthquakes share some aspects with meteorological natural disasters; they both evolve from an initial small state into what may become a major disaster. For instance, hurricanes develop over days and scientists and emergency managers must make response decisions as the storm progresses. If earthquakes occurred over similar time scales, seismologists could make similar analyses . " an earthquake started on the San Andreas fault near Parkfield yesterday and it is proceeding towards the Los Angeles area. It is scheduled to arrive in Los Angeles later today and we suggest that citizens move to open ground." Unhappily, earthquakes happen much faster. What are the prospects for automating the type of common sense analysis used in hurricane tracking to the earthquake problem? VS (the virtual seismologist) is a new paradigm for event identification modeled on the thought process of experienced seismologists; it uses a Bayesian framework for the integration of information existing prior to the earthquake (e.g. seismicity in preceding days, the current topology of the seismic network, etc.) together with a data set that is increasing with time as the earthquake proceeds. The goal is to provide the best characterization of the earthquake with the data that is available at any time. This characterization includes uncertainty estimates associated with the corresponding estimated location and magnitude. Formal parameterization of uncertainty is critical for appropriate decision making on the part of specific applications. That is, response decisions must be made in the context of the cost of initiating an action (for both real and false alarms) compared to the benefit of taking some action when appropriate. The Bayesian framework allows a continuous update of the likely source parameters, including updates of uncertainty. The VS method is described in considerable detail in Cua's Ph.D. thesis (http://etd.caltech.edu/etd/available/etd-02092005-125601/). We will also discuss strategies for characterizing the rupture parameters of very large earthquakes that have long rupture dimensions. While these earthquakes may be especially damaging to society, they are also the events which have the best potential for giving longer warning times (10's of seconds for seismic shaking to 10's of minutes for tsunami warning). While large events may provide us with large warning times, it is a great challenge to automate seismic analysis so that we can interpret large earthquakes automatically as they happen.
DE: 1217 Time variable gravity (7223, 7230)
DE: 7212 Earthquake ground motions and engineering seismology
DE: 7294 Seismic instruments and networks (0935, 3025)
SC: Seismology [S]
MN: Fall Meeting 2005