An array of 14 irregularly spaced seismograph stations with an aperture of approximately 1 km has been installed near the middle of the rupture zone of the 1966 Parkfield, California, earthquake (M = 5.5). The primary purpose of the array is to image sources of high-frequency energy radiated by the next characteristic Parkfield earthquake, using recordings of its aftershocks at the array as empirical Green's functions. In addition, it will be used to characterize coda Q in the Parkfield region using beam-forming techniques to isolate arrivals from specific volumes of the earth.
Each of the 14 stations consists of a three-component velocity transducer and an accelerometer package to ensure on-scale recordings of the full range of ground motion from microearthquakes to large damaging shocks. Transducer output is digitized at each field site and telemetered to a computer at a central recording facility for event detection and storage. Each component of ground motion is sampled 200 times per second by a 16-bit analog-to-digital converter. Time code received from the GOES West satellite synchronizes sampling across the array. Data telemetry is by ARCNET — a token-passing local area network developed for PCs — using a single coaxial cable. ARCNET is two-way, meaning that data or commands can be sent from any field station to the microcomputer or any other field station over the same medium using the same hardware and programming techniques.
Specific site locations were chosen by a simulated-annealing algorithm that chose the final site locations from a set of about 100 possible sites to optimize the beam pattern of the array. The algorithm minimized the power of peaks in the beam pattern with slownesses less than 1 sec/km relative to the central lobe. The array configuration was required to include several embedded tripartite arrays to allow the examination of coherence across the array.
Seismograms are shown for a M = 2.3 earthquake, which was located at Middle Mountain, 12.9 km from the array. The azimuth to the hypocenter is computed several ways using the array data and compared to the backazimuth from the epicenter location. A least-squares best fit to the arrival times of the first peak yields an azimuth of 7.9°, which is similar to the azimuth of the peak in slowness computed using the MUSIC algorithm of 10.3° (at 9 Hz). These angles are similar but consistently greater than the backazimuth to the hypocenter of 3.2°. The amplitude of the first arrival is approximately 6.4 times greater for a stacked trace (or beam) computed for a ray aimed along the ray path from the hypocenter compared to a beam 180° opposite to that. The amplitude of the first arrival of stacked traces diminishes over a range in azimuth of 80° on either side of the azimuth for a beam aimed along the UPSAR-hypocenter ray path. Coda Q calculated for a range of 0° to 350° and slownesses of 0.1 to 0.24 sec/km show no dominant peaks in the frequency bands of 4 to 8 and 8 to 16 Hz. A large, late arrival in the seismogram in the 2- to 4-Hz band yields high Qs at azimuths of around 0° with large errors. Coda Q increases with frequency from about 150 in the 2- to 4-Hz band to about 500 in the 8- to 16-Hz band.