The search for repeating earthquakes in the northern San Francisco Bay area Nader Shakibay Senobari and Gareth J. Funning University of California, Riverside

Repeating earthquakes (REs) are sequences of identical repeating events, which are recurring either irregularly (aperiodic) or nearly regularly (quasi-periodic). There are two important characteristics of the events in a RE sequence: they have the same source characteristics (i.e. magnitude and mechanism) and they have the same location and therefore their waveforms at the same stations are extremely similar. Several authors have proposed that the quasi-periodic REs result from recurrent rupture of a small locked patch on a fault surface surrounded by a larger area of creep. The implication is that any detection of characteristic REs along a fault can be interpreted as a signature of the creep at depth on that fault. In addition, REs can be used for locating faults and determining their geometries at depth. In this study, we are looking for REs on the northern Rodgers Creek and southern Maacama faults near Santa Rosa, CA. There is some observational evidence for creep along portions of these faults (e.g. from InSAR, alignment arrays and offset cultural features) but the depth of creep on both faults is still unknown. Finding the locations of REs in this area, and combining them with geodetic data, will help us to place stronger constraints on the distributions of aseismic slip on both the Rodgers Creek and Maacama faults. In order to identify such events, we use data from the Northern California Seismic Network (NCSN) for the period from 1984 to July 2013. We used earthquake waveforms from 2080 events located inside a 25 by 30 km area around Santa Rosa. We choose 7 stations located both inside and outside the selection area at distances of up to 50 km. We calculate the coherence for all pairs of events for each station during the time that the sensor has not been changed (mostly from 1987 to 2013). We align the waveforms using P arrival times from the NCSN catalog. The time windows for the seismogram analysis are set at 0.2 sec before and 10.2 sec after the P-wave arrivals. This time window always contains the S phase, which guarantees that the waves have the same S-P time (i.e. the same location) if they have high cross-correlation coefficients. We calculate the cross-correlation function by allowing a maximum time shift of 0.5 sec (50 samples) in the time domain. An earthquake pair is chosen to be a candidate for REs when the maximum cross-correlation coefficients at 1-15 Hz are larger than 0.95 at four or more stations. We then link a pair of REs with another pair if the two share the same earthquake. We find 49 such clusters of highly correlated events that are candidates for RE sequences. Using the northern California double-difference earthquake catalog (Waldhauser and Schaff, 2008), we find that they are mostly located close to the Rodgers Creek fault trace. In order to confirm whether these clusters are indeed REs, additional tests (e.g. phase and amplitude coherence, time recurrence, visually inspect) must be performed.