Least-squares fitting of marine seismic refraction data

Summary. An iterative procedure is presented for fitting waveform data from a marine seismic refraction experiment. The wavefunction from the explosive source is known and the crustal structure is refined using the damped least squares procedure. The damping parameter serves the dual purpose of stabilizing an under-constrained inversion and improving the linearity by suppressing high frequencies. The synthetic seismograms and their model differentials are calculated using the WKBJ seismogram method. Both the synthetic seismograms and the linear algebra are sufficiently straightforward that the computations can be performed on an array processor. The inversion procedure is then sufficiently rapid that interactive computations are possible. The technique is illustrated using the FF2 refraction data from the Fanfare cruise of the Scripps Institution of Oceanography. These data had been interpreted previously by trial-and-error using the reflectivity method. Starting from two different, simple models, the inversion procedure obtains essentially one unique model. Its features are very similar to the previous model.