Teleseismic P wave residuals from 70 events (1971–1979) for seven stations in central and northern California are investigated as a function of azimuth to model upper mantle velocity structure. Relative residuals are determined for each station using Berkeley as the reference station. Maximum relative residual variation is +0.9 s in the northern Coast Ranges to −1.7 s in the Klamath Mountains. Interpretations of the residual variations represent lateral variations of the P wave velocity in the upper mantle. The modeling technique utilizes a simple ray‐tracing technique to calculate time‐residual equivalent path lengths in the anomalous velocity regions. A model of the paleosubduction zone, Farallon plate, beneath northern California represents a region (40‐ to 150‐km depth) of high‐velocity (8.3–8.5 km/s), which explains the negative residual pattern for the Klamath Mountains, Cascades, and Modoc Plateau region. A low‐velocity zone (7.2–7.4 km/s) beneath the Gorda plate (25‐ to 155‐km depth) produces the observed positive residual pattern of the northern Coast Ranges. The low‐velocity region is related to proposed models of the Mendocino triple junction migration and associated upwelling of asthenospheric material.
Seismic data recorded at Jamestown, California, provide P- and SV-wave spectral corner frequencies for 18 trans-Sierra Nevada events in the range 3.2≤ML≤4.0. These include most of the available earthquakes from 06 July 1974 to 31 August 1975. The ratio R of P- to SV-wave corner frequency was 0.96±0.17 for these events, which sample a wide range of source azimuths and focal mechanisms, Assumptions of a flat far-field source spectrum and frequency-independent attenuation in the 1-20 Hz band lead to Qβ not less than 480. Many recent source theories predict factor-of-two or more variation of R with azimuth, and significant deviation of its average value from unity; we conclude that such source models do not apply to these earthquakes. Previous studies of R by MOLNAR et al. (1973), STUMP (1974), and BAKUN et al. (1975), also based on excellent data, find different results. We interpret this discrepancy not as a contradiction, but as an indication that R may be an extremely useful source-model discriminant, its variation indicating that different models need be applied to different source regions. R may lead to a clearer understanding of the physical processes associated with earthquakes; for example, the only source models that predict a value of unity for R at all azimuths are those in which the corner frequency measures the time duration of motion at the source, thus indicating that this duration is as long as the rupture propagation time across the source.