INTRODUCTION

Global seismology tomography faces a different set of problems than those faced in typical oil exploration tomography projects. Energy travels at significantly wider range of angles in global seismology and the velocity-depth ambiguity of reflection based seismology is not present. On the other hand, global seismologists have orders of magnitude less data. In addition, this data usually has a much lower signal to noise ratio than in the typical oil industry project.

The result is that both types of tomography operators have a large null space. In global seismology, this problem is usually is addressed by limiting the number of model components, either through some type of global harmonic parameterization or simply larger grid cells. Another approach is to introduce an inverse model covariance operator as a regularization operator into the inversion scheme. This regularization operator can be a non-stationary operator that introduce a desired, or hypothesized, structure to the velocity model Clapp (2001).

In this paper we apply a series of non-stationary regularization operators, a steering filter Clapp (2001), to a global seismology tomography problem. We show how these filters can produce a more aesthetic pleasing image that still fits the data, and we hypothesize they can be used to help evaluate different model hypotheses.