ESTIMATION AND OPTIMIZATION OF RESIDUAL MOVEOUT USING KALMAN FILTER

Abstract

Normal Moveout (NMO) correction in seismic data processing, is one of the first few steps which provide an insight into the subsurface velocity variations, thus, it is very necessary to determine the velocities in a manner which generates minimum depth uncertainty. One element of depth uncertainty is introduced geophysically after the process of normal moveout correction i.e. in the form of residual moveout (RMO) at far offsets. The objective of this work is to reduce the RMO due to velocity variations through forward modeling in known anisotropic scenarios with the help of Kalman filter. Residual Moveout can occur due to many reasons; one of them being subsurface anisotropy. A successful attempt in reducing RMO can result in an improved stacked section. Solutions are available to optimize anisotropy related RMO in depth domain however if RMO optimization is successful in known anisotropic setting in time domain itself, then fewer number of iterations will be required in depth domain. The modified three-term approximations of Tsvankin's non-hyberbolic moveout equations (Tsvankin and Thomsen, 1994) and Kalman filter equations have been used in this dissertation for writing code for RMO optimization. Kalman filter estimates future states of a process from the initial states. If initial state is considered to be a point where moveout is zero, it will estimate future states by taking small perturbations around that initial state. So, the estimates will be near to the zero moveout state, thus minimizing RMO with each iteration. In order to study the problem, three synthetic earth models have been created and synthetic shot gathers have been generated by Ray tracing method in OMNI. Henceforth, these gathers have been processed in OMEGA followed by velocity analysis. These gathers and picked velocities are input to the RMO optimization code which estimates new offset-based velocities. The horizontal velocity component estimated at offsets for different events using the MATLAB code result in a reduction of the RMO at various CMP gathers.