Surface wave phase velocity maps from multiscale wave field interpolation

Availability of spatially dense broadband observations of seismic surface wave fields allows to derive phase velocity maps on regional scale by non-tomographic means. I present a multiscale interpolation scheme for surface wave fields where for each observed wave field, a multiscale phase velocity map and its standard deviation are calculated from subsets of the available data. Isotropic phase velocity (with an estimate on its standard deviation) is then found as the weighted mean of multiscale maps from different observations. Comparison of different interpolation schemes on synthetic wave fields shows that bivariate linear interpolation of phases in a triangulated region, in conjunction with multiscale stacking, is not inferior than methods involving higher-order polynomials. The recovered phase velocity maps are, in the presence of random uncertainties with realistic magnitude, largely free of artifacts and restore a synthetic input model reasonably well. The method is applied to a subset of Rayleigh wave observations at USArray which yields phase velocity maps well within the range of published results. Multiscale interpolation seems to be a practical alternative to tomographic inversion for regional broadband seismic networks (a parts per thousand yen30 stations). It requires no choice of arbitrary parameters and is insensitive to number of earthquakes and their azimuthal distribution.