Parkfield-Hollister Electromagnetic Monitoring Array

Markus Eisel, Gary Egbert College of Oceanic and Atmospheric Sciences, Oregon State University, Corvallis, OR 97331.

In the Fall of 1995, UC Berkeley installed magnetotelluric (MT) observatories at two locations along the San Andreas Fault in California to monitor possible changes in the electromagnetic fields before, during and after earthquakes (e.g. Fraser-Smith, 1994). Since then MT data have been recorded continuously and archived at the UC Berkeley Digital Seismic Network (BDSN). This data set provides the rare opportunity to investigate long term variations of MT parameters at short periods.

Magnetotelluric (MT) Array

The MT observatories are located at Parkfield (PKD1), 300 km south of the San Francisco Bay area (Figure 5.1), and Hollister (SAO), halfway between San Francisco and Parkfield. Both sites are equipped with 3 induction coils and 2 x 100 m electric dipoles connected to a Quanterra digitizer. Time synchronization is provided by GPS. The data are sent to BDSN via telephone lines.

**Figure 5.1:** Map of the central coast of California showing the San Francisco Bay area and the locations of the MT observatories. Red lines idicate the Bay Area Rapid Transit (BART) railway system. Pink lines are the major faults. The distance between the two sites is 150 km.
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig1.eps,width=6cm}\end{center}\end{figure}$

Data and Processing

Five components of the electromagnetic fields are recorded at 40 Hz and 1 Hz sampling rates. The multiple station processing code (Egbert, 1997) is used for the time series analysis. The continuous data stream is subdivided into daily sets from which MT transfer functions (TF) and interstation magnetic TF are estimated to study daily and seasonal variations. The 1 Hz data covers the years 1996 and 1997. Due to system outages 588 of the 731 days of data are available. (Figure 5.2)

**Figure 5.2:** Schedule of equipment changes at MT sites.
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig2.eps,width=6cm}\end{center}\end{figure}$

Seasonal Variations

Sections of signal power (Figure 5.3), noise power (Figure 5.4) and signal to noise ratio (Figure 5.5) of Hy at site PKD1 have been derived from the daily averages of the 1 Hz data. The sections cover the time 1/96 to 12/97 and clearly show a strong seasonal variation. The other field components and those at site SAO show the same seasonal variations.

**Figure 5.3:** Hy signal power at Parkfield
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig3.eps,width=8cm}\end{center}\end{figure}$

**Figure 5.4:** Hy incoherent noise power at Parkfield
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig4.eps,width=8cm}\end{center}\end{figure}$

**Figure 5.5:** Hy signal to noise ratio at Parkfield
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig5.eps,width=8cm}\end{center}\end{figure}$

**Figure 5.6:** Sum of 3 hour Kp indices
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig6.eps,width=8cm}\end{center}\end{figure}$

During the summer, short period (1 - 10 s) signal and SNR are increased while at long periods these values tend to decrease in correlation with the magnetic activity indices. For comparison the daily sum of three-hour Kp indices is plotted (Figure 5.6). The black line gives the median filtered values which emphasize long term variations. A general decrease of magnetic activity from 1996 to 1997 visible in the magnetic indices plot is reflected in the MT data as well.

Transfer Functions

Apparent resistivities and phases and interstation magnetic transfer functions were calculated for 7 days during the night-hours (0am - 4am PST) and 7 days during rush hour (6am - 10am PST) during the Bay Area Rapid Transit (BART) strike from day 251 - 257 of 1997. To emphasize the bizarre nature of the bias effects, we computed transfer function estimates in very narrow bands.

The magnetic transfer functions (SAO predicted by PKD) calculated from the daytime windows (Figure 5.7) shows the enormous bias due to coherent noise between the two sites. TF derived from time windows after midnight (same days) (Figure 5.8) are much smoother, but there is still some bias present.

**Figure 5.7:** Amplitude and phase of TF estimate: Strongly biased interstation magnetic TF at SAO with PKD as the reference station caluclated from ``noisy'' 4-hour time segments (6am - 10am PST) of 7 days during a BART strike.
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig7.eps,width=8cm}\end{center}\end{figure}$

**Figure 5.8:** Amplitude and phase of TF estimate: The same TF from ``quiet'' time segments (0am - 4am PST) of the same days. Still some bias visible.
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig8.eps,width=8cm}\end{center}\end{figure}$

**Figure 5.9:** Apparent resistivity and phase curves at PKD1 from the ``noisy'' segments. There is almost no bias effect.
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig9.eps,width=8cm}\end{center}\end{figure}$

**Figure 5.10:** Apparent resistivity and phase curves at SAO from the ``noisy'' time. A slight bias around 15s period
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig10.eps,width=8cm}\end{center}\end{figure}$

Principal Component Analysis

The plane wave source assumption corresponds to a response space of dimensionality two. Likewise the spectral density matrix (SDM) has 2 dominant eigenvalues corresponding to the two plane wave source polarizations. Other eigenvalues above background noise levels show that the plane wave response space is not entirely appropriate for the Parkfield-Hollister array. These additional eigenvalues of the SDM represent noise, some of which is coherent across the array. The significantly larger third and fourth eigenvalues may be a result of BART (Figure 5.11).

**Figure 5.11:** Eigenvalues from SDM
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig11.eps,width=8cm}\end{center}\end{figure}$

**Figure 5.12:** Eigenvectors from SDM : green = magnetic channels : red = electric channels
$\begin{figure}% \begin{center} \epsfig{file=figs/bsl98_em_fig12.eps,width=8cm}\end{center}\end{figure}$

References

Egbert, G. D., Robust Multiple-Station Magnetotelluric Data Processing. Geoph. J. Int.,it 130,, 475-496, 1997.