Probabilistic object-based waveform tomography

Abstract

We applied a probabilistic approach to a full waveform tomographic inversion problem using surface wave data and spectral-element simulations of seismic wave propagation. An object-based parametrisation of the Australian upper mantle is used to obtain a very low dimensional model space, allowing us to effectively use a direct search method - the Neighbourhood Algorithm. We present results concerning structural properties of the Austalian upper mantle and the capability of waveform data to resolve different types of heterogeneities.