Long-Term Monitoring of ULF Electromagnetic Fields at Parkfield CA

Karl Kappler

Berkeley Seismological Laboratory

Electric and magnetic fields in the (10^-4 -1.0) Hz band were monitored at two sites adjacent to the San Andreas Fault near Parkfield and Hollister, California. Observed fields typically comprise natural magnetotelluric fields, with cultural and instrument noise. A data window [2002-2005], enclosing the September 28, 2004 M6 Parkfield earthquake, was analyzed to determine if anomalous electric or magnetic fields, or changes in ground conductivity, occurred before the earthquake. The data were edited, removing intervals of instrument malfunction, leaving 875 days left in the four-year period. Frequent, local spike-like disturbances were removed. The distribution of these spikes was not biased around the time of the earthquake. Signal to noise ratios, estimated via magnetotelluric processing techniques, provided an index of data quality. Plots of signal and noise amplitude spectra, showed the behavior of the ULF fields to be remarkably constant over the period of analysis. From these first-order plots, it is clear that most of the recorded energy is coherent over the spatial extent of the array. Three main statistical techniques were employed to separate local anomalous electrical or magnetic fields from the dominant coherent natural fields: transfer function estimates between components at each site were employed to subtract the dominant field, and look deeper at the 'residual' fields; the data were decomposed into principal components to identify linear combinations of array channels, which are maximally uncorrelated; the technique of canonical coherences was employed to distinguish anomalous fields which are spatially broad from anomalies which occur at a single site only, and furthermore to distinguish anomalies which are present in both the electric and magnetic fields form those which are present in only one field type. Standard remote reference apparent resistivity estimates were generated daily at Parkfield. Most of the variation was observed to be seasonal, and frequency independent, suggesting a local seasonal distortion effect. Once corrected for distortion, nearly all of the variability in the apparent resistivity was removed. In all cases, high levels of sensitivity to subtle electromagnetic effects were demonstrated, but no effects which can be described as precursors to the Parkfield earthquake were found.