Layered crustal velocity model and station correction analysis for the Los Angeles basin

A simultaneous inversion for hypocenters, station corrections and a one dimensional velocity model was performed on P and S arrival times from 129 earthquakes located in a 45 x 55 km region centered in the Los Angeles basin area in southern California. The velocity inversion was done using a computer program called VELEST which uses a damped least squares inversion technique with the damping parameter calculated from the ratio of the data variance to the a priori uncertainty of model parameter. The a priori uncertainty of the hypocentrallocation was assumed to be 3 km, the uncertainty of the initial one dimensional velocity model was assumed to be 10% from the true solution, and the station corrections were assumed to have an mean of 0.0 sec and a standard error of 0.5 sec. The velocity and station correction inversion reduced the root mean square residual for 129 events from 0.68 s to 0.052 s. The velocity inversion benefited from a data set covering the general Los Angeles basin, thus providing a fairly well-resolved velocity model. The results from the velocity inversion indicate a 3-4% decrease for the upper 5 km and a 2% velocity increase for 5-15 km with respect to the starting model. The final velocity model with the smallest sum of squares of residuals has layers from 0 to 2 km (3.43 km/s), 2 to 5.5 km (3.88 km/s), 5.5 to 10 km (6.07 km/s), 10 to 16 km (6.54 km/s), 16 to 32 km (6.62 km/s) and a half space starting at 32 km (7.8 km/s). The best fit velocity model has a third layer boundary at 10 km which does not correlate with any geologic feature in the retrodeformable cross section by Davis and others (1989), however, this study strongly suggests that there is an 8% velocity increase that occurs at 10 km. A possible interpretation of the velocity discontinuity at 10 km is a lithology contrast due to faulting at depth. The new velocity model was then used in a joint earthquake location and station corrections inversion on a clustered data set of a selection of 103 earthquakes from the Whittier Narrows aftershock sequence. This inversion process provided a higher resolution of the spatial characteristics of the earthquake cluster. The station corrections accounted for lateral velocity variations not accommodated by a velocity model varying only with depth. The relocated cluster of earthquakes is grouped into two northwest linear trends which have been interpreted by various authors (Hauksson and Jones, 1989 and Magistrale, pers. comm., 1989) as possible tear faults bounding the thrust sheet identified after the Whittier Narrows main shock. The focal mechanism for the western cluster shows both right-lateral strike slip fault and thrust fault plane solutions. The focal mechanisms for the eastern cluster consist of thrust, right-lateral and left-lateral strike-slip fault-plane solutions. Also, just to the east of the proposed left-lateral tear fault is the location of a ML 4.7 aftershock with right-reverse motion. The hypothesis that the eastern linear trend of epicenters is a left-lateral tear fault is not consistent with the fault plane solutions along this linear trend nor the fault plane of the ML 4.7 just to the east of this trend. However, I suggest that some of the focal mechanisms from the Whittier Narrows aftershock sequence support a thrust block bounded by tear faults and that the complexities observed in the variety of fault plane solutions in the Whittier Narrows aftershock sequence are due to the fact that thrust, and strike-slip coexist and together accommodate oblique motion in the Los Angeles basin (Hauksson, 1990).