The surface wave radiation pattern search uses information in the MFT96 dispersion files that contains measured group velocities and spectral amplitudes. given the mechanism, mean dispersion and anelastic attenuation coefficients can be estimated for the data set. These values represent broad regional averages in the northwestern United States and are not specific to the Yellowstone region.

The program surf96 is used to invert for a model that better fits the observed dispersion. The initial model is that used by the University of Utah Seismograph Stations locations of earthquakes in the Yellowstone area:

Yellowstone Model

The resultant model is Revised Yellowstone Model

The inversion preserves the layer discontinuities of the initial model, but slightly perturbs the velocities in the layers. The models and fit to observed dispersion and anelastic attenuation are

Model and dispersion fit

Model and anelastic attenuation fit

It is always interesting to see how these different models would affect the travel time measurements which are used for location. The following figure compares the predicted first P-wave arrival times for the two models for a surface focus using the program timmod96. The first arrival time are very similar out to the Pn arrival.

The real difference is that this model better predicts the transverse component seismograms and does better with the Rayleigh wave on the Z and R traces than does the CUS model.

Comparison of observed and predicted broadband traces


I would not state that this revised model is realistic, only that it is certainly better than the CUS model for source parameter inversion and especially for waveform matching.