Estimation of Thomsen's anisotropy parameters from compressional and converted wave surface seismic traveltime data using NMO equations, neural networks and regridding inversion
Amber Camille Kelter
To gain a better understanding of the earth's subsurface anisotropy should be considered. This thesis aims to quantify the anisotropy parameters, and , that define compressional and converted waves. It is investigated whether a better approximation can be found from inversion of compressional wave data, converted wave data or the use of these in conjunction. A synthetic data set is used to develop and evaluate a number of inversion algorithms that estimate and . Algorithms include NMO equations, neural networks and regridding inversion. Neural networks are the most robust when applied to compressional wave data. In particular, it is found that is best estimated using P-wave neural networks that solve for , while is best estimated using P-wave neural networks that solve for both and .
Having attained quality results from the synthetic data set, the optimal inversion techniques are applied to the Blackfoot data set. The results are encouraging and consistent with that of Elapavuluri (2000) and Thomsen (1986) where the coals and shales displayed a greater degree of anisotropy than the sands.