HR: 0800h
AN: IN21A-0078    [Abstracts]
TI: Detection of regional events using streaming GPS
AU: * Granat, R
EM: granat@jpl.nasa.gov
AF: Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Dr., Pasadena, CA 91109,
AU: Pierce, M
EM: mpierce@cs.indiana.edu
AF: Indiana University, 105 S. Morton Ave., Bloomington, IN 47404,
AU: Gao, X
EM: gao4@indiana.edu
AF: Indiana University, 105 S. Morton Ave., Bloomington, IN 47404,
AU: Bock, Y
EM: ybock@ucsd.edu
AF: Scripps Institution of Oceanography, University of California, San Diego, 9500 Gilman Dr., San Diego, CA 92093,
AB: In recent years, GPS measurements of surface displacement have proved critical to understanding earthquake fault systems. As well as measuring the deformation resulting from tectonic processes and seismic events, there has been increasing evidence that GPS sensors are capable of detecting signals associated with "slow earthquakes" and stress transfer between faults. Our hypothesis is that there may be subtle signals of these processes that have remained undetected at the level of individual sensors but can be detected by using the aggregated information from multiple sensors. To this end, we present a method for detecting regional events that combines statistical analysis of individual sensor measurements to detect signals within a region of interest covered by a GPS network or sub-network. This method is applicable not only to archived data but also to real-time streaming data, and so offers the possibility of being used in alert-type service. In our approach, time series from individual sensors is first segmented into discrete modes according to statistical properties of the data. This is accomplished using an algorithm for robust, unconstrained fitting of hidden Markov models (HMMs); tests indicate that this significantly method outperforms standard approaches for a variety of data types, including GPS time series. The resulting segmented time series from individual stations are then compared to one another and incidences of correlated mode changes between sensors are tallied. A significant signal at a particular time is indicated by a high number of correlated mode changes between network members. This technique has been integrated into a web portal/web services environment that facilitates the simultaneous analysis of data from multiple GPS stations, provides a web-baed interface to the method, and direct access to streaming real-time as well as archived GPS data sources. We demonstrate our approach using GPS position data provided by the Scripps Orbit and Permanent Array Center (SOPAC). A map-based visualization interface allows for use as a interactive analysis tool.
DE: 1207 Transient deformation (6924, 7230, 7240)
DE: 3270 Time series analysis (1872, 4277, 4475)
DE: 7230 Seismicity and tectonics (1207, 1217, 1240, 1242)
SC: Earth and Space Science Informatics [IN]
MN: 2007 Fall Meeting