Automated Search for Tremor in Parkfield and the Bay Area, CA.

Robert W. Porritt, Richard M. Allen, Robert Nadeau
University of California Berkeley

Mike Brudzinski
Miami University

AGU Fall Meeting 2006
Eos Trans. AGU, 87(52), Fall Meet. Suppl., Abstract T41A-1549

Non volcanic tremors have been observed with different properties in two main fault zones. The Cascadia Subduction Zone tremors are large emergent signals which last days and have amplitudes several times the daily mean. Wrench fault tremors in Parkfield are smaller emergent signals with durations on the order of a few minutes. While the large amplitude and long lasting Cascadia Subduction Zone tremors can be identified by automated methods, the significant differences in wrench fault systems require a more finely tuned detection method. Therefore, we have developed an automated processing sequence which can detect the majority of tremors in the Parkfield region and have applied that to the San Francisco Bay Area. This program is designed to be altered by the user for the tremor signals in the area being evaluated and then run over the time-span the user wishes to evaluate. The program is relatively simple; it downloads seed volumes, writes files of minimal daytime noise, bandpass filters the frequencies, takes an absolute value normalized to a percentile, keeps only values within a range likely to correspond to tremor, correlates over multiple stations, and smoothes over the average length of a tremor outputting a time-series correlating to potential times of tremor. The process was found to be almost 87 percent accurate at identifying Parkfield tremors and it was fast enough to complete about two and a half years worth of data (August 2001 through December 2003) from 8 borehole stations in about four hours. The processing sequence was then utilized with identical parameters and 6 borehole stations for the San Francisco Bay Area and no tremors were found from 1 January 2004 to present. Possible reasons for this include: there is no tremor on the Hayward Fault in the Bay Area, the currently available real-time data is not sufficient temporally and spatially, there is a very high noise variance due to cultural noise, the periodicity of the tremors is beyond the search time window, large magnitude distant earthquakes replicate tremor signal, or the parameters of Bay Area tremor are significantly different.

© Richard M Allen