Ray-Born Modelling and Full Waveform Inversion

Abstract

Seismic forward modelling is used to simulate seismic wave propagation in the subsurface. Common modelling techniques include methods such as the ray-Born approximation and the finite difference method. In this thesis I present a comparison between the ray-Born approximation and the finite difference method. This comparison consists of computing synthetic seismograms using both methods for several seismic velocity models and performing a comparison between them. The main motivation for this comparison is to see if the ray-Born approximation, which is known to be faster, provides enough accuracy to provide an efficient alternative to the finite difference method. The ray-Born approximation was found to be sufficiently accurate in cases where the velocity model did not have large velocity contrasts and multipathing did not occur. In those cases the method can replace the finite difference method in forward modelling algorithms. Full waveform inversion is a technique that is used to compute a model of the subsurface given the seismic data recorded at the surface. In order to perform full waveform inversion, forward modelling is an obvious requirement. Traditional full waveform inversion algorithms use the finite difference method to perform forward modelling and the cost of the computation is therefore high. In order to reduce the cost of the inversion, a full waveform inversion algorithm that uses the ray-Born approximation instead of the finite difference method has been developed. The results from the full waveform inversion when using the ray-Born approximation show that it is possible to achieve an accurate reproduction of the subsurface at a greatly reduced cost compared to the finite difference method. The ray-Born approximation is therefore evaluated to be a method that can replace the finite difference method both in forward modelling and in full waveform inversion algorithms.