Joint Inversion of Surface Waves and Ps-converted Waves

Abstract

To put constraints on the crustal and upper-mantle structure, seismologists have often made use of the sensitivity of surface wave dispersion observations and teleseismic P-wave receiver functions to the shear velocity structure of the medium.

Both types of datasets have their own limitations in resolution. In this study, it was investigated if the resolution gaps can be bridged (and hence leading to tighter constraints on the shear wave velocity structure), by doing a joint interpretation of surface wave dispersion data and teleseismic P-wave receiver functions. This joint inversion of surface wave dispersion data and Ps-receiver functions was based upon fundamental mode Rayleigh wave phase and group velocity measurements and teleseismic P-wave receiver functions.

To solve the joint inverse problem, two different solution techniques were closely investigated: a Monte Carlo-search and an iterative linearized least-squares inversion.

The results of this study clearly show that performing a joint inversion with surface wave phase velocity dispersion data and receiver function data significantly improves the retrieved shear wave velocity structure. Using group velocity dispersion data in the joint inversion (in addition to phase velocity dispersion data and receiver functions), improves the retrieved shear wave velocity structure even more.

The non-uniqueness problem of the receiver functions is tackled remarkably well by the joint inversion. Nevertheless, the joint inverse problem remains highly non-linear. Therefore, with the iterative linearized least-squares inversion, a final solution which resembles the true shear wave velocity structure perfectly is almost impossible to obtain even with the application of a general approach for the iterative linearized least-squares inversion developed in this study. During this development, it has become clear that the choice of the starting model does influence the end-result obtained, but the most important influencing part is formed by the Vp/Vs ratio of the starting model, which was kept fixed in the inversion.

A Monte Carlo-search is in general the best solution technique, while this results in a range of possible models that give a clear indication of the real shear wave velocity structure. However, also the iterative linearized least-squares inversion results in quite good shear wave velocity-depth plots.