Annual Report, 1997

Integrated Los Angeles Area Velocity Model

Harold Magistrale
San Diego State University

To obtain a unified southern California velocity model we must develop methodologies to merge the detailed rule based 3D model of the major southern California basins constrained by geologic and structural data with the information contained in earthquake and explosion travel times. One approach is to use the basin sediment velocities as a priori constraints in travel time inversions for crustal P wave velocities, and to explicitly specify model error estimates. During the past year I used the most recent basins model developed by Working Group B in an inversion of earthquake and LARSE blasts travel times for P wave velocity. Each input contributes to an integrated model: the basins model incorporates geologic constraints, provides control of shallow sediment velocities and locations of the contacts between geologic units, while the travel times provide 3D control of deep sediment and basement rock velocities. Any changes of the basin seismic velocities required to fit the travel time data can suggest updates to be incorporated into the basins model, so this method is useful for basins model verification.

The invasion velocity model includes the area covered by the major populated basins in the Los Angeles area (Los Angeles basin, Ventura basin, San Gabriel Valley, San Fernando Valley, Chino basin, and San Bernardino Valley). The model is parameterized as 6380 constant velocity blocks (Figure 1), mostly 8 by 8 km in area and 2 to 4 km thick. Basin sediments are represented by 456 blocks assigned velocities from the basins velocity model. Blocks representing basement rocks outside and between the basins are assigned velocities from a vertically smoothed 1D Hadley-Kanamori model.

Because the basins model include geological, structural, and preliminary waveform modeling constraints, the basin velocities are assigned smaller uncertainties than the background basement velocities. The basin sediment P wave velocities are assigned uncertainties of 3% at the surface, increasing with depth to 7% at 10 km depth (reflecting fewer constraints with depth). The basement rocks are assigned uncertainties of 16%. (This corresponds to an uncertainty of 1 km/s for a 6.3 km/s velocity). The model blocks are individually damped proportionately to the velocity uncertainties: the basin sediment blocks are more strongly damped than the basement rocks dunng the inversion.

The travel time data are 273,000 P wave arrivals from 8800 selected earthquakes (Figure 1) recorded on the southern California seismic array (SCSN) and the Northridge and Landers portable deployments. Also, 1250 P wave arrivals from 48 LARSE line 1 blasts recorded on the SCSN are used; each blast arrival is given five times the weight of an earthquake arrival. Travel time picks are assigned uncertainties appropriate to the quality assigned during routine SCSN processing. The inversion procedure (Roecker, 1982, with various mod)fications) minimizes the travel time residuals by iteratively adjusting the seismic velocities of the blocks and the locations of the earthquakes. Station corrections are not used. During the inversion, the variance of the travel time residuals is reduced by 77%.

The final crustal P wave velocities (Figure 2) are broadly consistent with other recent inversions (e.g., Hauksson and Haase, 1997). The final model preserves the slow shallow basin velocities where the inversion has poor resolution.
At 5 km depth and above, the basement rocks away from the basins are similar in the start and final models. Near the basins, the basement rocks have slightly lower velocities, due in part to an averaging of the basin sediment and basement rock velocities within the model blocks that straddle the basin edges. Below 5 km depth there are several basement velocity heterogeneities: the San Gabriel Mountains are faster than the Mojave area (5 to 7 km depth); the Santa Monica Mountains are relatively fast; and the east San Fernando Valley is relatively slow.

The final basin sediment velocities in the San Gabriel Valley (1 to 3 km depth) are slower than in the starting model, consistent with recent waveform studies. At 5 to 10 km depth, the Los Angeles basin sediments are slightly slower in the final model relative to the starting model, and the slowest parts of the Ventura basin are further east in the final model than in the starting model. San Fernando Valley observations are hindered by the relatively large model block size relative to the short wavelength variations in structure. There, the basin model verification will require a smaller model element size parameterization.

This method of accounting for the varying amounts of a priori knowledge of Los Angeles area seismic velocities is successful: basin sediment velocities from the geology-based model are altered where required by travel time information, details of the geology-based sediment velocities are maintained where the travel times have poor resolution, and basement velocity heterogeneities are imaged.

1997 Publications:

Magistrale, H., 1997, Geology based 3D seismic velocity models of populated southern California basins (abstract), EOS Trans. AGU, 78, p. F432.

Magistrale, H., 1997, A southern California tomographic crustal velocity model with geologic constraints (abstract), Seismol. Res. Lett. 68, p. 321.

References:

Hauksson, E., and J. S. Haase, 1997. Three-dimensional Vp and Vp/Vs velocity models of the Los Angeles basin and central Transverse Ranges, California, J. Geophys. Res. 102, 5423-5472.

Roecker, S. W., 1982. Velocity structure of the Pamir-Hindu Kush region: Possible evidence of a subducted crust, J. Geophys. Res. 87, 945-959.