Unique attributes in sequences of recurring, similar microearthquakes at Parkfield, California, provide a means for inferring slip rate at depth throughout the active fault surface from the time intervals between sequence events. Application of the method using an 11-year microseismicity record revealed systematic spatial and temporal changes in the slip rate that were synchronous with earthquake activity and other independent measures of fault-zone slip. If this phenomenon is found to be generally common behavior in active faults, it forms the basis for a method to monitor the changing strain field throughout a seismogenic fault zone.

The San Andreas fault at Parkfield, California, apparently late in an interval between repeating magnitude 6 earthquakes, is yielding to tectonic loading partly by seismic slip concentrated in a relatively sparse distribution of small clusters (<20-meter radius) of microearthquakes. Within these clusters, which account for 63% of the earthquakes in a 1987-92 study interval, virtually identical small earthquakes occurred with a regularity that can be described by the statistical model used previously in forecasting large characteristic earthquakes. Sympathetic occurrence of microearthquakes in nearby clusters was observed within a range of about 200 meters at communication speeds of 10 to 100 centimeters per second. The rate of earthquake occurrence, particularly at depth, increased significantly during the study period, but the fraction of earthquakes that were cluster members decreased.

Deep fault slip information from characteristically repeating microearthquakes reveals previously unrecognized patterns of extensive, large-amplitude, long-duration, quasiperiodic repetition of aseismic events along much of a 175-kilometer segment of the central San Andreas fault. Pulsing occurs both in conjunction with and independent of transient slip from larger earthquakes. It extends to depths of approximately 10 to 11 kilometers but may be deeper, and it may be related to similar phenomena occurring in subduction zones. Over much of the study area, pulse onset periods also show a higher probability of larger earthquakes, which may provide useful information for earthquake forecasting.

The Hayward fault slips in large earthquakes and by aseismic creep observed along its surface trace. Dislocation models of the surface deformation adjacent to the Hayward fault measured with the global positioning system and interferometric synthetic aperture radar favor creep at approximately 7 millimeters per year to the bottom of the seismogenic zone along a approximately 20-kilometer-long northern fault segment. Microearthquakes with the same waveform repeatedly occur at 4- to 10-kilometer depths and indicate deep creep at 5 to 7 millimeters per year. The difference between current creep rates and the long-term slip rate of approximately 10 millimeters per year can be reconciled in a mechanical model of a freely slipping northern Hayward fault adjacent to the locked 1868 earthquake rupture, which broke the southern 40 to 50 kilometers of the fault. The potential for a major independent earthquake of the northern Hayward fault might be less than previously thought.

Main shocks of the earthquake sequences that occurred on the Parkfield section of the San Andreas fault in central California in 1922California in , 1934California in , and 1966 are characterized by southeast rupture expansion over the same 20-to 30-km-long section of the fault. Whereas the seismic moments for the 1922 and 1934 events are ideniical to within a precision of 10%, the seismic moment for 1966 is 20% greater than for the earlier events to within a precision of 20%. The Parkfield area seismicity, in general, seems well described by a recurring moderate size characteristic earthquake, repeating the same epicenter, magnitude, seismic moment, rupture area, and southeast direction of rupture expansion. An unexplained 10 year advance of the 1934 event is the only discrepancy in the hypothesis that the Parkfield earthquakes in 1857, 1881, 1901, 1922, 1934, and 1966 represent a strictly periodic process. Assuming the strictly periodic model and the absence since 1966 of the perturbations hypothesized for the 1922 to 1934 period, the next characteristic Parkfield earthquake should occur between 1983 and 1993. INTRODUCTION H. F. Reid's discussion of the mechanics of the 1906 California earthquake [Reid, 1910] contains the fundamental ideas underlying the time-predictable earthquake recurrence model: an earthquake occurs when strain accumulated since the preceding earthquake results in fault stress greater than the strength of fault material. Assuming a constant rate of strain accumulation and constant fault strength, Bufe et al. [1977] developed the time-predictable recurrence model to explain the recurrence of small earthquakes on a 9-km-long section of the Calaveras fault zone east of San Jose, California. Shimazaki and Nakata [ 1980] concluded that the recurrence of large thrust fault earthquakes at three sites of plate convergence around the Japan arcs was consistent with a time-predictable model rather than a slip-predictable model. That is, they concluded that the time interval between two successive large earthquakes is approximately proportional to the amount of coseismic displacement of the preceding earthquake and not of the following earthquake. Sykes and Quittmeyer [ 1981] have presented additional evidence in support for the time-predictable model for the recurrence of great earthquakes. Earthquakes on the Parkfield section of the San Andreas fault in 1857, 1881, 1901, 1922, 1934, and 1966 provide an excellent opportunity to test the applicability of recurrence models to regions characterized by recurring moderate-sized earthquakes. The similarity of intensity patterns [Sieh, 1978; Toppozada et al., 1981] and surface fractures along the San Andreas fault trace [Brown and Vedder, 1967] suggest that these shocks ruptured the same 20-to 30-km-long section of the fault (Figure 1). Lack of instrumental data precludes the determination of precise source parameters for the 1857, 1881, and 1901 shocks. In this paper we compare source parameters for of the fault since at least 1857 can be modeled, with t...

Seismic anisotropy and P-wave delays in New Zealand imply widespread deformation in the underlying mantle, not slip on a narrow fault zone, which is characteristic of plate boundaries in oceanic regions. Large magnitudes of shear-wave splitting and orientations of fast polarization parallel to the Alpine fault show that pervasive simple shear of the mantle lithosphere has accommodated the cumulative strike-slip plate motion. Variations in P-wave residuals across the Southern Alps rule out underthrusting of one slab of mantle lithosphere beneath another but permit continuous deformation of lithosphere shortened by about 100 kilometers since 6 to 7 million years ago.

Tomographic imaging techniques were applied to two crosshole data sets to determine the velocity structures and the reliability and resolution of the algorithms on real data. The experiments were carried out at the Retsoff salt mine in New York and at the underground radioactive waste study site at the Stripa mine facility in Sweden. The traveltimes at Restoff were high quality and were obtained over raypaths of up to 500 m in length. The structure was quite complicated with velocity contrasts up to 50 percent. The Stripa site was in granitic rock with velocity contrasts of only a few percent. The dimensions of the experiment were small with maximum ray lengths of just over 10 m. The data at this site were collected with very high accuracy, source and receiver locations were measured to better than 1.0 mm, and traveltimes were read to 0.001 ms. A number of algorithms similar to the algebraic reconstruction techniques (ART) used in medical imaging have been applied to the data. Some modifications of the algorithms, such as the application of weighting schemes, damping parameters, and curved raypaths, were performed. The resulting velocity fields were compared to the known fields and with each other to determine an optimal method. The algorithms were found to be a rapid, reliable means of reconstructing the slowness field of real data. Low‐velocity zones were recovered with accuracy in location and value. It was also found that great care was necessary in application of the techniques to ensure that proper damping parameters are used and the proper number of iterations taken; otherwise poor reconstructions will result.

No abstract