Abstract: Integrating Facies Models with Geostatistical Models of Aquifer Heterogeneity

Steven F. Carle, Graham E. Fogg

Subsurface architecture of debris flows, stream channel, overbank, and floodplain facies in an alluvial fan environment largely dictates subsurface contamination migration routes at the Lawrence Livermore National Laboratory, Livermore, California. Despite a fairly extensive database consisting of several hundred boring logs, the complexity of the three-dimensional (3-D) facies architecture cannot be adequately described by traditional approaches. Alternatively, we implement a geostatistical technique called probability cokriging (PCK) to stochastically generate "realizations" or simulations having a realistic spatial variability. To account for the spatial interrelationships between the facies, PCK requires 3-D probability cross-covariance (CCV) models, but our database supports only 1-D vertical CCV models. Instead of relying on pure data, we draw on facies models to establish a "realistic" pattern of bedding plane geometry, then synthesize a 3-D CCV model from a sequence of 1-D Markov chain models of transition probability calculated from the vertical data and the pattern. The resulting 3-D PCK realizations of facies architecture honor the heterogeneity prescribed by the bedding plane pattern and the vertical data. We then use these realizations to parameterize numerical studies of flow and transport.

AAPG Search and Discovery Article #90986©1994 AAPG Annual Convention, Denver, Colorado, June 12-15, 1994