HR: 10:05h
AN: S51E-07 INVITED     [Abstracts]
TI: Techniques for Handling Large Data Sets in Global Tomography
AU: * Masters, G
EM: gmasters@ucsd.edu
AF: IGPP,UCSD, 9500 Gilman Drive, La Jolla, CA 92093 United States
AU: Reif, C
EM: creif@ucsd.edu
AF: IGPP,UCSD, 9500 Gilman Drive, La Jolla, CA 92093 United States
AU: Manners, U
EM: umanners@ucsd.edu
AF: IGPP,UCSD, 9500 Gilman Drive, La Jolla, CA 92093 United States
AB: Largely due to the success of the IRIS consortium, it is now possible to easily obtain in excess of 250 high quality three component recordings for moderately sized earthquakes (a total of roughly 100,000 recordings per year). Of course, with the advent of US Array, this number will continue to grow. Many techniques have been developed to handle such data streams but usually in a regional context. Here, we describe a development of a cross-correlation technique combined with a waveform clustering algorithm to semi-automatically measure relative arrival times. This method has already been applied to body waves and here we give an extension to measure the relative group arrival times of globally distributed surface waves. The method works with band-pass filtered envelope functions (as in the standard group velocity analysis) but measures the relative arrival times in a fixed frequency band of all waveforms for an event. The clustering provides an automatic method of grouping similar waveforms and of removing bad waveforms. The method is fast and accurate and leads to very large datasets. For example, application to 50 second Rayleigh waves results in a dataset of over 200,000 relative group arrival times with over 25,000 measurements per year in recent years. Such datasets are capable of giving fine-scale resolution of near-surface structure over much of the globe.
DE: 7255 Surface waves and free oscillations
DE: 7260 Theory and modeling
DE: 7205 Continental crust (1242)
DE: 7207 Core and mantle
SC: Seismology [S]
MN: 2004 AGU Fall Meeting