Real-time monitoring of fine-scale changes in fault and earthquake properties

Abstract:

The high-resolution back-processing and re-analysis of long-term seismic archives has generated new data that provide insight into the fine-scale structures of active faults and seismogenic processes that control them. Such high-precision studies are typically carried out retro-actively, for a specific time period and/or fault of interest. For the last 5 years we have been operating a real-time system, DD-RT, that uses waveform cross-correlation and double-difference algorithms to automatically compute high-precision (10s to 100s of meters) locations of new earthquakes recorded by the Northern California Seismic System. These locations are computed relative to a high-resolution, 30 year long background archive that includes over half a million earthquakes, 20 million seismograms, and 1.7 billion correlation measurements. In this paper we present results from using the DD-RT system and its relational database to monitor changes in earthquake and fault properties at the scale of individual events. We developed baseline characteristics for repeating earthquakes, fore- and aftershock sequences, and fault zone properties, against which we evaluate new events in near real-time. We developed these baseline characteristics from a comprehensive analysis of the double-difference archive, and developed real-time modules that plug into the DD-RT system for monitoring deviations from these baselines. For example, we defined baseline characteristics for 8,500 repeating earthquake sequences, including more than 25,000 events, that were found in an extensive search across Northern California. Precise measurements of relative hypocenter positions, differential magnitudes, and waveform similarity are used to automatically associate new member events to existing sequences. This allows us to monitor changes relative to baseline parameters such as recurrence intervals and their coefficient of variation (CV). Alerting of such changes are especially important for large sequences of repeating events with CV~0 - such as are exclusively found on creeping segments of the San Andreas fault system - as they may indicate changes in local fault properties and/or loading rates.