On the retrieval of pure path velocities from great circle data

Summary. Synthetic great circle Rayleigh wave phase velocities are generated from a 3-D upper mantle model for two great circle distributions: a uniform random distribution and an experimental distribution arising from a real set of sources and receivers. Given a tectonic regionalization, the retrieval of the pure path velocities from these great circle data sets leads to a strongly biased solution when the uniform distribution is used, whereas the results from the experimental distribution are closer to the true values. An analysis of the data importances and the information distribution in the data indicates that the great circles introduce coupling between tectonic regions; thus the degeneracy of the solution inherent in great circle data cannot be completely overcome by introducing a tectonic regionalization. It is shown that a satisfactory solution is obtained by selecting data as pure as possible, i.e. data that over-sample one tectonic region without oversampling the others. This condition is better fulfilled by the experimental distribution than by the uniform distribution.