Relocation of microearthquakes, using seismic waveform cross-correlation, reveals organized patterns of fault sub-structure (Nadeau et al. 1995; Nadeau & McEvilly 1999; Rubin et al., 1999; Waldhauser & Ellsworth, 1999). These patterns lead us to ask questions regarding fault complexity, similarity and the triggering mechanisms between microearthquakes.
Drawing on the US Geological Survey Northern California Seismic Network dataset from the previous fifteen-year period, we relocate over 15,000 California microearthquakes in various regions extending from Cape Mendocino in the north to Coalinga in the south. Microearthquakes on the San Andreas, Calaveras and Sargent faults define near-vertical faults with strikes that typically mimic the major fault strike (e.g. Figure 1b). However, relocated microearthquakes near Mt. Lewis show an intricate melange of vertical planes with strikes both parallel and perpendicular to the trend of seismicity. On a smaller scale, relocation results exhibit patterns that we term: (1) "Repeaters," which are sets of 2-12 similar magnitude earthquakes that repeatedly re-rupture the same fault area; (2) "Streaks," in which microearthquakes define slip-parallel lineations up to 1-2 km in length but only tens of meters in the vertical and across-strike dimensions; and (3) "Planes," which, seen in cross-section, viewed across-strike define narrow features (tens of meters) but viewed along-strike exhibit no discernable pattern.
Three of our data subsets (two along the San Andreas fault south of the Loma Prieta rupture extending into the creeping section and one along the Calaveras fault near the 1984 Morgan Hill mainshock) exhibit microearthquakes that primarily "streak" and "repeat," while relocated data on the Sargent fault define primarily "planes." We investigate if the seismic signatures of these neighboring faults truly differ or if these apparent differences stem from our limited observation time, which does not allow us to examine similar absolute slip distances due to slip-rate variations (presumably slower on the Sargent fault). We partition our data in space and time so that we may compare data sets with similar overall fault area and total seismic rupture area. We then compute the percentage of seismic area occupied by either multiple events that overlap (not necessarily repeaters) or by "repeat" earthquakes. We hypothesize that similarity in these percentages indicates similar fault behavior. Results show that the absence of "repeaters" on the Sargent fault may be an artifact of the low total slip, whereas the lack of "streaks" is a fault characteristic.
A right-lateral ML5.7 event, the largest recorded to date in the Mt. Lewis region, occurred on March 31, 1986, near the thin north-south trending seismicity. In the following months, seismicity migrated both north and south away from the hypocenter (Zhou et al., 1993), mapping an "hourglass" pattern delineated primarily by east-west trending faults with presumably left-lateral slip. Surprisingly, for events within one fault length of the mainshock, seismicity patterns on either side of the north-south axis are very symmetric. Coulomb failure models with modest frictional coefficients are not able to predict this symmetry due to significant normal stress changes that skew contoured positive Coulomb stress changes towards the tensile sides of the mainshock fault (i.e. NE and SW quadrants of the map). Seismicity at greater distances exhibit this skewed behavior and are better predicted with typically used frictional coefficients. We interpret the migration of seismicity and positional symmetries as indicating time-dependant failure in the "process zones" beyond the ends of the mainshock fault.
