Bitumen reserves in oil sands in Alberta, Canada represent one of the biggest such deposits in the world. The Athabasca region contains the bulk of this resource, and the Lower Cretaceous McMurray formation contains the most significant target interval. Inclined heterolithic strata and associated sand accumulations comprise most of the formation. However, the distribution of bitumen in the formation varies because of the high degree of facies heterogeneity throughout the deposit. This lithological heterogeneity causes difficulties in interpreting geology and estimating the bitumen distribution.

Surface seismic data could play an important role in characterizing the subsurface heterogeneity because they provide lateral and vertical coverage and a link to rock physics through amplitude variation with offset (AVO). However, most (with notable recent exceptions using deterministic lambda-mu-rho, shown by Bellman, 2007 and Evans and Hua, 2008) applications of surface seismic in Athabasca have been to provide attributes for statistical and neural network predictions (Tonn, 2002; Anderson et al., 2005). The relationships between seismic data and lithology are determined at the well control points by multivariate analysis or neural networks and then the lithology between wells is predicted from these relationships. However, interpreters often find these relationships less straightforward than conventional techniques.