Historical seismicity rates, from various networks operating over the years, suggest that there is sufficient seismicity, perhaps an event per day at a detection threshold around magnitude 0.5 to 1.0 to allow high-resolution imaging of fault-zone processes. Such a detection sensitivity requires high-gain instrumentation and sophisticated noise mitigation techniques, possible in the East Bay Area only with borehole-installed seismometers. A network of borehole-installed wide dynamic range seismographic stations - the Hayward Fault Network (HFN) - is being developed cooperatively with the US Geological Survey with support from USGS, EPRI, Caltrans, the University of California Campus/Laboratory Collaboration (CLC) program, LLNL, and LBNL (Figure 3.1 and Figure 3.2). The purpose of the network is to lower substantially the threshold of microearthquake detection for events along the fault and to record much higher frequency waveforms in order to increase the resolution of fault-zone structure and spatio-temporal characteristics in the seismicity. This new data base will improve working models for the Hayward fault, and provide enhanced data to the real-time monitoring and alert systems functioning at the Berkeley Seismological Laboratory.
The basic concept for the HFN is that, in an urban environment swamped with cultural background noise (traffic and industry vibrations) orders of magnitude above average levels in remote sites, any individual station running on-site detection software has a high probability of being desensitized from the noise and thus will miss many of the small events that we need to record at all network stations for high-resolution study of Hayward Fault activity. It is impractical to telemeter all of the data at the necessarily high sample rates (500 sps, up to 4 components) to a central site for processing in real time. To circumvent this problem we employ two countermeasures: (1) the sensors are placed in boreholes as deep as possible (preferably 100 m or more) and in bedrock, for a significant reduction in surface noise, and (2) representative signals (a single 100 sps vertical component from each station) are processed centrally for the detection of legitimate microearthquakes, and recovery of the full data set for the event from the entire network is accomplished by command from the central site, in a fully automated system. The detector has been running for 18 months now in an offline processing mode successfully. The magnitude threshold is approximately 0.0 for small events along the Hayward Fault. With a false detection rate of about two-thirds, on average more than one event per day is recovered, satisfying the design goals of the network. As more stations in boreholes are added, the quantity and quality of the data will improve.
Six-component borehole sensor packages designed and fabricated at LBNL's Geophysical Measurement Facility by Don Lippert and Ray Solbau, with 3 channels of acceleration (Wilcoxin 731A) and 3 of velocity (OyO HS-1 4.5 Hz geophones) are being installed in the entire network. The 0.1-400 Hz accelerometers have lower self-noise than the geophones above about 25-30 Hz, and remain on scale and linear to 0.5 g. Sensors are installed at depths of 100-300 m and provide signals to the on-site Quanterra dataloggers (Q4120 and Q730 systems). Signals digitized at a variety of data rates (up to 500 Hz (Table 3.1) at 24-bit resolution. 7 of the HFN stations use digital frame relay telemetry to transmit continuous data at a reduced rate (100, 20, and 1 sps) to the central controller for event detection. The central controller declares valid network events, extracts appropriate data segments at full resolution (up to 500 sps), and collects other associated information from the Southern Hayward Fault Network (SHFN).
Similar to BDSN sites, station processors are capable of continuous on-site recording and local storage at full resolution for more than one day and have batteries to provide backup power. On-site detectors can monitor critical conditions and preset changes in operating mode can be invoked. The new UltraSHEAR communication software devised for this network (and now widely used) allows control and change of parameters from the central site. MultiSHEAR provides for multi-station nodes using the Q730 satellite acquisition platforms around a Q4120, with a single 4-station telemetry link to the central site. Data are archived in the NCEDC archive established jointly at Berkeley by USGS and UCB.
As originally planned, the network will consist of of 24 to 30 stations, 12-15 each north and south of San Leandro, managed respectively by UCB and USGS. This is not happening quickly, although west of the fault we have several Caltrans bridge borehole sites. Caltrans plans to provide some holes of opportunity away from the bridges in the coming years, so we are planning two new borehole sites in data gaps east of the fault or near the Oakland Airport, bringing the geometry to a more effective state for imaging and real-time monitoring of the fault.
Central site triggering, one of the design goals of the Hayward Fault Network, is crucial to maximizing the performance of a modern borehole seismic network located in a noisy urban environment. Telemetry bandwidth limitations preclude the possibility of sending continuous 500 Hz or higher sampled data, and the high urban environment noise levels renders local detection of seismic events at each station unreliable. Single-station false triggers at the desired low detection threshold can easily amount to 99 of the recorded data. We have developed and tested a central site triggering algorithm that processes continuous 100 Hz EP1 (vertical geophone) data feeds from all telemetered HFN sites. Detection thresholds were adjusted until each station produced detections at a rate of about one detection per hour. The number of associated detections averages about 5 per day. Of these coincident detections, about one-third are local and regional earthquakes (and an occasional teleseism) and the rest are coincident noise associations. Subsequent analysis of the earthquakes identified by the prototype detector/associator indicates that:
The observed signal-to-noise ratios for a sample of the smallest detected events typically exceeds 10 dB at 20 Hz. The ML threshold, calculated on the assumption that detector will trigger at two or more stations when the signal-to-noise ratio exceeds 10 dB at 20 Hz (STA/LTA ratio of 3.5), is somewhere between 0.0 and -0.6 for Berkeley local events (distances less than 10 km). Inferred from the observed magnitude distribution and rate of seismicity at the ML 1.5 and larger level, the expected rate of seismicity within a 10 km radius of Berkeley, is 3 per month at the ML 0.0 threshold and 10 per month at the ML -0.6 threshold. Based upon the detected/associated number of local events, the detection threshold appears to be closer to ML -0.6 than to ML 0.0.
We are preparing to implement the final detection/association algorithm on the central site computer as we install the last of the Quanterra 4120 and 730 telemetered systems and complete the MultiSHEAR upgrade of the software on the remote dataloggers.