HR: 0830h
AN: S21D-0328    [PDF]
TI: Applying Massive Waveform Cross Correlation and Double-Difference Location to Northern California and China
AU: * Schaff, D P
EM: dschaff@ldeo.columbia.edu
AF: Lamont-Doherty Earth Observatory, Columbia Univ., 61 Route 9W, Palisades, NY 10964 United States
AU: Waldhauser, F
EM: felixw@ldeo.columbia.edu
AF: Lamont-Doherty Earth Observatory, Columbia Univ., 61 Route 9W, Palisades, NY 10964 United States
AU: Richards, P G
EM: richards@ldeo.columbia.edu
AF: Lamont-Doherty Earth Observatory, Columbia Univ., 61 Route 9W, Palisades, NY 10964 United States
AB: In recent years substantial advances in earthquake location have realized one to two orders of magnitude improvement over seismically broad areas due to the complementary benefits of waveform cross correlation and double-difference (DD) relative location in reducing both measurement and velocity model errors respectively. We report on results from applying these techniques on a massive scale to locally recorded seismic data across all of Northern California and to regional and teleseismic data in China. We are in the process of performing 2.5 billion correlation measurements on 280,000 events in Northern California from 1984 to present. The waveforms comprise the entire digital archive at the NCEDC recorded by 900 short period vertical component stations totaling 225 GB of data. We use a correlation detector instead of a correlation function to allow lags greater than 1/2 of the window length (typically 1 to 2 s) to be searched. At station JST, which includes 35,000 events from Bay area faults such as the Hayward, the Calaveras, and the San Andreas faults, 40% of the events have at least one other event with cross correlation coefficients (CC) greater than 0.9 (62% for CC $>$ 0.8, 77% for CC $>$ 0.7). These percentages of similar events are surprisingly high, but they include known areas of repeating events on the Calaveras and San Andreas Faults. For a station in Long Valley Caldera recording 72,000 events, the distribution is 18% for CC $>$ 0.9, 43% for CC $>$ 0.8, and 67% for CC $>$ 0.7. A station including 20,000 events in the different tectonic settings of Mendocino Triple Junction and Geysers Geothermal Fields has 16% of the events with CC $>$ 0.9 with at least one other event, 32% have CC $>$ 0.8, and 49% have CC $>$ 0.7. The lower numbers of correlated events observed at the latter two stations most likely reflect the different faulting processes that take place in these areas, compared to the (mostly) strike-slip events recorded at JST. To improve the accuracy of inter-event distances from which we determine pairs of events for correlation measurements we have relocated about 240,000 events using the double-difference method together with about 5 million NCSN P-phase picks. The mean shift between routine NCSN locations and DD relocations is about 300 and 500 m in the horizontal and vertical directions, respectively. The relocated seismicity shows a substantially increased level of detail across most of the Northern California region, which can be significantly enhanced by incorporating the cross correlation differential time measurements, once they are available. In China, we applied cross correlation methods to seismograms recorded at regional and teleseismic distances from $\sim$14,000 earthquakes and explosions, to measure relative arrival times to an accuracy of $\sim$0.01 s, obtaining location precision around a few 100 m --- approximately a hundredfold improvement over current methods. From the results it is discovered about 10% of earthquakes in and near China, reported for the period 1985 to 2000 with $M \geq 3$, are repeat events generating essentially the same signals from sources each of which cannot be more than about 1 km from another event.
DE: 7215 Earthquake parameters
DE: 7230 Seismicity and seismotectonics
DE: 7294 Instruments and techniques
SC: Seismology [S]
MN: 2003 Fall Meeting