We followed the approach described by *Aoi* (2002). The observation
equation is first linearized and then solved iteratively by singular value
decomposition. Synthetic waveforms as well as sensitivity functions are
needed to obtain the solution. We used the elastic finite-difference code E3D
(*Larsen and Schultz*, 1995) to calculate the waveforms and obtain the
sensitivity functions numerically, by taking the difference of waveforms from
perturbed and unperturbed models.

The results of the inversion using the waveforms for the =6.4 Near Coast of Central Chile event are shown in Figure 2.21. The top two rows show the first 20 s of the waveforms as only this time window was used in the inversion. The bottom two rows show a 60 s time window. In addition to the P-wave, the pP-wave can be seen arriving just after the first 20 s. The results show that although the waveforms for only the first 20 s were used in the inversion, the final model also better describes the P-wave coda following the pP arrival.

Figure 2.22 compares the velocities in the SCV basins
for the USGS v2 model (below station 238), initial, and final model
after 3 iterations. Density and density gradient were not free
parameters but were determined from P-wave velocity using the
Nafe-Drake equation (*Brocher*, 2005).

Berkeley Seismological Laboratory

215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760

Questions or comments? Send e-mail: www@seismo.berkeley.edu

© 2007, The Regents of the University of California