Albuquerque Seismological Laboratory
Results: Data Fit
We assess the fit to the data set in two ways. First, we look at how well we can fit the individual seismograms. In the first figure we show the initial and final fits to seismograms for an earthquake on November 11, 1991. In the second figure we show the initial and final fits to seismograms for an earthquake on March 1, 1991. In each figure the synthetic seismograms are indicated by dashed lines.
The second way of assessing the fit to the overall data set is to look at the paths where we fit the seismograms well and the paths were we fit the seismograms poorly. The important things to note are: (1) we fit most of the paths very well, and (2) the paths with poor fits are randomly distributed and are not associated with a single geographic area or station.
To provide a nice illustration of the procedure outlined in the method section we show the initial fit to the seismogram (for a homogeneous starting model), then the fit after the non-linear inversion (we now fit most of the phase, but don't fit the amplitude at all), and finally we see the fit after the linearized "scattering" inversion (we fit both the phase and amplitude).