Predictability of repeating earthquakes near Parkfield, California

Abstract

We analyse sequences of repeating microearthquakes that were identified by applying waveform coherency methods to data from the Parkfield High-Resolution Seismic Network. Because by definition all events in a sequence have similar magnitudes and locations, the temporal behaviour of these sequences is naturally isolated, which, coupled with the high occurrence rates of small events, makes these data ideal for studying interevent time distributions. To characterize the temporal predictability of these sequences, we perform retrospective forecast experiments using hundreds of earthquakes. We apply three variants of a simple algorithm that produces sequence-specific, time-varying hazard functions, and we find that the sequences are predictable. We discuss limitations of these data and, more generally, challenges in identifying repeating events, and we outline the potential implications of our results for understanding the occurrence of large earthquakes. © 2012 The Authors Geophysical Journal International © 2012 RAS.