Nonhyperbolic velocity analysis and improved demultiple for seismic data
Abstract
Reflection seismology is a widely used geophysical method to image and delimit underground oil and gas prospects using seismic data illumination from the surface. The collected information in acquisition contains reflected energy that is used to build a time or a depth image of the subsurface, but is also corrupted with undesired components like multiple reflections and random noise.
The success of seismic imaging heavily relies on building a velocity and anisotropy models that reflect the subsurface features to improve the resolution of the migrated sections. In this thesis, an automatic workflow for nonhyperbolic velocity analysis is presented to estimate the moveout parameters (velocity and anellipticity) in an efficient way using an accurate traveltime approximation, combined with efficient parameter sampling and highresolution coherency estimators that detect the reflected events in the gathers. The proposed estimators are based on differential semblance and use optimal trace sorting to improve resolution in moveout and residual moveout spectra that helps in enhanced parameter tracking.
A modified version of the parabolic Radon transform is also presented with applications to multiple subtraction and trace interpolation. The transform is computed using a minimum-norm solution to an under-determined inverse problem solved by means of a singular value decomposition of a frequency-independent operator. This has benefits like improved separation of primaries and multiples when they severely intefer and preservation of amplitudes along the primaries after demultiple better than do the commonly used frequency-domain high-resolution parabolic Radon transforms.
The presented methods are later applied to a shallow long-offset marine dataset that exhibits amplitude variations with offset (AVO) and have multiple energy in the gathers. The applied sequence leads to improved demultiple and a better amplitude preservation of the primary reflections that is needed for AVO analysis.