U.S. Geological Survey, M.S. 977, Menlo Park, California 94025, USA

Geology Department, University of California, Davis, California 95616, USA

The initial stages of fault development have been recognized in laboratory experiments, but not at seismogenic depths. To fully understand how faults initially develop and begin to evolve, one needs to first recognize and then observe the emerging structure. Recognizing the birth of a fault requires a priori knowledge of the regional tectonics, geologic structure, and seismic history. This information is needed to assess the likelihood that earthquake hypocenters that occur far from known faults represent new fault formation rather than slip at depth on existing structures. Furthermore, young faults may begin as blind structures, which makes them difficult to recognize.

The 1983 M_W (moment magnitude) 6.4 Coalinga, 1987 M_W 6.0 Whittier Narrows, and 1994 M_W 6.7 Northridge earthquakes revealed the significance of blind thrust faults in California. Unlike the San Andreas fault, blind faults are difficult to recognize because they lack visible, surface features. Even after a blind fault is identified, it is difficult to determine whether the fault is active . However, no active blind strike-slip faults have been documented anywhere, with possible exceptions at the Long Valley caldera and the Mount Lewis seismic trend in California . We present evidence for an incipient, blind, strike-slip fault in California's southern Sierra Nevada that is associated with the 1946 Walker Pass earthquake.

We studied the diffuse band of seismicity that is located between two large, historic, earthquakes: the 1946 M_W 6.1 Walker Pass earthquake on the northern end of the lineament and the 1952 M_W 7.3 Kern County earthquake that ruptured along the White Wolf fault on the southern end of the lineament (hereafter we refer to the seismic lineament as the Scodie lineament — Fig. 1). The idea that these earthquakes were possibly tectonically connected was loosely suggested by , and the existence of the lineament was later observed by . Previous seismicity studies and geologic maps of the area do not show a surface expression of a fault, geologic structure, or tectonic landform corresponding to the Scodie lineament. The regional geology is a mix of Mesozoic granitic rocks with isolated, north-trending pre-Cretaceous metamorphic terranes. The trend of the lineament is at about a 40° angle to the prominent north-trending fabric of the Sierra Nevada ; therefore the local geology provides several unambiguous piercing points along the lineament at which displacement could be resolved.

We relocated and calculated focal mechanisms for the earthquakes between the northern end of the White Wolf fault and the 1946 main shock. Both the seismic networks in the vicinity of the Scodie lineament, and the techniques used to relocate earthquakes have significantly improved since the last detailed seismic studies were conducted for the area. The Scodie lineament falls on the border between the Northern California and Southern California Seismic Networks, and both networks operate seismographic stations in the Sierra Nevada. We took advantage of the rich data sets from these overlapping seismic networks by merging short-period P-wave phase data from both networks for 7274 earthquakes that occurred between September 1976 and March 1998. To better resolve the characteristics of the seismic lineament, we used the method of joint hypocentral determination to obtain new seismic station traveltime corrections and then relocated the earthquakes using the velocity model derived by Jones and Dollar (1986). To minimize erroneous earthquake locations, we excluded earthquakes that were recorded at fewer than 15 seismic stations, had an RMS uncertainty of >0.1 s, had a relative horizontal position uncertainty of >0.35 km, and a relative vertical position uncertainty of >0.9 km. Similarly, focal mechanisms were calculated for all of the magnitude 2.0 and greater earthquakes that had a minimum of 12 phase readings . To minimize potential problems with low amplitude or improper ray traced refracted waves, we gave lower weight to poor quality arrivals. All focal mechanisms were visually and statistically inspected, and unresolveable focal mechanisms were removed from the data set.

The improved hypocenters define a narrower Scodie seismic lineament (Fig. 1A), with earthquakes clustered at the ends and in the center of the lineament. Most earthquakes along the Scodie lineament have strike-slip fault-plane solutions, and their focal depths are greater than 4 km (Fig. 1, B and C). If these earthquakes are aligned with the overall trend of the seismic lineament, then the preferred direction of slip would be left-lateral strike slip along the northeast-striking fault planes (Fig. 2A).

Seismicity in the Scodie seismic lineament can be characterized as three sections (Fig. 1, A-C). The southwest section contains a mixture of left-lateral strike-slip events and reverse events that are located in the aftershock zone of the 1952 Kern County earthquake. Most (84 %) of these shocks are strike slip events, presumably left-lateral slip associated with the White Wolf fault. The central region of the lineament forms a zone of diffuse events with clusters of earthquakes near the axis of the lineament. Nearly all earthquakes in this section are consistent with left-lateral strike slip subparallel to the lineament. Earthquakes along the northern section of the Scodie lineament are predominately strike-slip events with scattered normal-slip events (Fig. 2). The majority of the earthquakes in this region are in the vicinity of the 1946 Walker Pass earthquake and suggest a complex series of en echelon fault segments. Two such fault segments coalesce at depth as well as along strike (Fig. 2).

The microseismicity in the Walker Pass earthquake region suggests that the main shock may have occurred on a northeast—trending southeast-dipping fault, thereby helping to constrain the variety of focal mechanisms that have been calculated for the event. Focal mechanisms for the 1946 Walker Pass earthquake have been calculated by a variety of techniques, all trying to exploit the few available seismic readings . Conclusions from these studies suggest that the main shock occurred on either a north-trending normal fault or a northeast-trending left-lateral strike-slip fault (Fig. 2). On the basis of trends in the microseismicity near the main shock, we favor the strike-slip focal mechanism and suggest that the northeast-trending nodal plane is the fault plane.

On the basis of the seismicity pattern and focal mechanisms, we infer that the Scodie lineament is likely a left-lateral shear zone. Although the central section contains only diffuse seismicity, this could be consistent with an incipient fault, and the clear activity at both ends should evolve, with time, toward a connected throughgoing fault. The lack of a surface expression associated with the seismic lineament may be because it is a young fault that has had little throw.

To determine whether the Scodie lineament is an incipient feature, we searched for evidence of Riedel shear; faults and fractures that form at an oblique angle (about 10°-35°) to the maximum direction of shear in material without a continuous fault surface. In laboratory experiments, Riedel shear fractures form during the early stages of a fault’s development, with the angle between the maximum shear plane and the Riedel shear fractures a function of the angle of internal friction . For well-developed faults, we would expect that the T-axis orientations for earthquakes near the fault to be oriented 45° from the direction of maximum shear. However, if Riedel shear is present, then we would anticipate as much as a 35° counterclockwise rotation of the T-axis orientations from this expected azimuth (Fig. 1D). For the Scodie lineament, we calculated T-axis azimuths from strike slip focal mechanisms that lie within 10 km of the Scodie lineament and found that the majority of the T-axes are not parallel to the lineament— instead they are rotated counterclockwise (Figs. 3, 1B, and 1D).

To evaluate the significance of the observed T-axis rotation within the Scodie lineament, we conducted a statistical test of skewness of the T-axis data about the expected T-axis azimuth. We assume that errors in the focal mechanisms are distributed symmetrically about the true T-axis, an assumption that could be invalid if the distribution is not symmetric because of poor station distribution, improperly traced rays, or noisy data. We binned the data into two groups that contained T-axis trends less than and greater than the expected T-axis azimuth. We also truncated the distributions at ± 35° from the expected direction. Without Riedel shear, the errors should be symmetric about the expected direction (45° from the strike of the fault), so the number of events in the two bins should be described by the binomial distribution. In contrast, if Reidel shear was present, it would show up as skewness in the distribution; i.e., the number of events in either bin is larger than would be found under the null hypothesis of no Reidel shear.

We found that the observed T-axis orientations are inconsistent with pure left-lateral slip along the Scodie lineament, but they are consistent with Riedel shear along the lineament (Fig. 3A, Table 1). The two trends at the northern end of the Scodie lineament (Fig. 2) broaden the distribution of T-axis azimuths, but are not responsible for the rotation of the T-axes from the expected trend. In contrast, at the Parkfield section of the San Andreas fault, observed T-axis orientations for strike-slip earthquakes calculated with both a one-dimensional (1-D) (Northern California Earthquake Data Center) and a three-dimensional (3-D) seismic velocity model are not skewed (Fig. 3B and Table 1), indicating that Reidel shear is not present on this older, well-developed fault. Because the 1-D and 3-D seismic velocity models produced nearly identical results even though the Parkfield region has significant 3-D structure , we do not expect that uncertainties in ray tracing affected the Scodie lineament results. Thus, at the 99.9% confidence level, there is evidence for Reidel shear along the Scodie lineament. This finding suggests that the Scodie lineament has the characteristics of a young, anastomosing fault and not a continuous structure.

The close proximity and orientations of the Scodie seismic lineament and the White Wolf fault suggest that these two structures are related. The seismic lineament may represent an extension or propagation of the White Wolf fault towards the northeast. The big bend segment of the San Andreas fault is a restraining bend that is not oriented parallel to the Pacific—North American plate motion. The White Wolf and Garlock faults both accommodate the big bend transpression; oblique reverse slip on occurs the White Wolf fault, and left-lateral slip occurs on the Garlock fault (Fig. 1A-inset). It is possible that the Garlock fault has shifted south, with the North American plate, beyond the portion of the big bend whose orientation deviates the most from the Pacific—North American plate motion. If so, then this kink in the San Andreas fault might have caused the White Wolf fault to lengthen to accommodate further slip, extending the White Wolf fault to form the Scodie seismic lineament. No older analogues to the Scodie seismic lineament and Garlock fault are known, perhaps because the big bend is not old enough to have produced a prior fault with a similar orientation.

As discussed above, the Scodie lineament may represent an incipient fault that is a product of the tectonic evolution in the big bend region. We propose that the diffuse band of strike slip earthquakes that connects the White Wolf fault to the Walker Pass seismicity is an embryonic fault. The 15° rotation of the focal mechanisms from the trend of the lineament is consistent with laboratory observations of fault development. Thus we have extended the Riedel shear model from laboratory-scale experiments to in situ observations at seismogenic depths.

Support provided by National Science Foundation grant EAR-97-06690 and a Presidential Faculty Fellowship to L. Kellogg. We thank R. Stein, C. Jones, A. Nur, L. Jones, E. Moores, J. Savage, and W. Thatcher for reviews of the manuscript. We also thank R. Bürgmann, A. Donnellan, K. Hafner, L. Jones, D. Oppenheimer, J. Sauber, and R. Twiss for comments, discussions, and technical assistance. The seismic data were obtained through the Northern California Earthquake Data Center, a joint project of the Northern California Seismic Network, U.S. Geological Survey, Menlo Park, and the Seismological Laboratory, University of California, Berkeley, and through the Southern California Seismic Network, a joint project of the U.S. Geological Survey, Pasadena and the Seismological Laboratory, Pasadena.