Improved Spatial Constraints for Seismic Impedance Inversion [article]

Haitham Hamid, University Of Calgary, University Of Calgary, Laurence R. Lines, Adam Pidlisecky
Most of the modern conventional inversion packages that are available to universities and industry use a 1D forward model with 1D dataset being inverted independently and then combined together to form a pseudo 2D or 3D image. I refer to this approach as 1D Laterally Unconstrained Inversion (1D-LUI). The simplicity of a 1D forward model enables geophysicists to obtain results and model scenarios in a short time but sometimes at the cost of a low quality inverted image. In this thesis, I present
more » ... a 1D laterally constrained inversion (1D-LCI) where the 1D datasets are inverted simultaneously. This algorithm produces layers in a section with laterally smoothed transition zones and sharp layer boundaries, while maintaining computational efficiency. We demonstrate the effectiveness of this approach on a series of poststack synthetic 2D models as well as a 2D poststack field dataset. The results of 1D-LCI results show clear improvement over the conventional method. This algorithm should be of interest to a wide variety of applied geophysicists looking to improve impedance-imaging results with a small increase in computational cost. While the results of the 1D-LCI are promising, it fails to image steeply dipping structures clearly and those events were smeared out laterally. In this work, I develop Structurally Constrained Inversion (SCI). SCI involves simultaneous inversion of all seismic traces of poststack seismic data using a regularization operator that forces the solution to honor local structure. The proposed method involves a multi trace-based impedance inversion and a rotation of an orthogonal system of derivatives operators. We illustrate the effectiveness of 1D-SCI algorithm on a synthetic 2D model as well as a field seismic dataset. The inversion results were able to image clearly the steeply dipping event, recover complex structures and produce low noise images in the estimated model parameters. The results look more geologically realistic when compared to trace-by-trace and Laterally Constrained Inversion r [...]
doi:10.11575/prism/26481 fatcat:h45hfgqrz5b4nj6pt6te6xg5xi