Multi-attribute seismic analysis - tackling non-linearity

AVO inversion for Lamé parameters (λρ and µρ) has become a common practice as it serves to enhance identification of reservoir zones. Also, integration of AVO-derived attribute volumes with other non-AVO derived seismic attribute volumes can provide meaningful geologic information when tied back to well data and verified as correlating with rock properties. Computation of reservoir properties for determination of mathematical relationships between variables derived from well logs, for example, is usually done with non-linear multivariate determinant analysis using neural networks. This paper provides a case study of a 3D seismic survey in southern Alberta, Canada, where a probabilistic neural network solution was first employed on AVO attributes (Pruden, 2002, Chopra & Pruden, 2003). Using the gamma-ray, acoustic and bulk density log curves over the zone of interest, gamma-ray and bulk density inversions were derived from the 3D attribute volumes. This methodology was successful, in that two new drilling locations derived from this work encountered a new gas charged reservoir, that not only extended the life of the gas pool but added new reserves as well. Later, instead of neural networks, a different mathematical approach using cubic b-splines was utilized for the same purpose. The results were found to be similar, suggesting that apart from neural networks, the cubic b-splines could be used as a tool for tackling non-linearity in multi-attribute seismic analysis.