Datapages, Inc.Print this page

Stratigraphic Stacking Patterns in Downscaling Seismic Data to Fine-Scale Flow Models

Kalla, Subhash 1; White, Christopher D.1; Gunning, James 2; Glinsky, Michael E.3
1 Louisiana State University, Baton Rouge, LA.
2 CSIRO, Melbourne, VIC, Australia.
3 BHP-Billiton, Perth, WA, Australia.

Reservoir architecture may be inferred from geologic models, seismic surveys, and well data. Analogs, conceptual models, sedimentation experiments, and mechanistic simulations give facies successions and geobody correlation lengths, but these results are not conditioned to well or seismic data. Stochastically inverted seismic data are uninformative about fine-scale features (circa 1 m vertically), but aid downscaling by constraining mesoscale properties (circa 10 m) such as total thickness and average porosity. Well data reveal detailed facies and vertical trends, but cannot specify interwell stratal geometry. Consistent geomodels can be generated for flow simulation by considering the scale and density of these diverse data.

The proposed approach integrates stratigraphic data via surface-based models. An ensemble of surface-based models provides a statistical description of the reservoir stratigraphy, which stipulates the probability of all layers being present at all grid locations. This probabilistic prior model removes the need to condition geomodels to one specific stratigraphic model, reducing conditioning artifacts that plague some modeling methods. The prior stacking models are updated using well and seismic data to generate the posterior model. The seismic inversion data provide mesoscale constraints, whereas the well data (encapsulated in a prior) are locally informative at the fine scale (circa 1 m). Markov Chain Monte Carlo methods sample the posteriors to integrate stochastic seismic inversions, well data, and the stacking framework from surface-based geomodels. Because all of these data are uncertain, stochastic ensembles of geomodels are needed to capture variability.

Plausible fine-scale features are introduced into flow models using this Bayesian approach, whilst avoiding overtuning to seismic data, well data, or geologic models. Fully integrated cornerpoint flow models are created, and examined using reservoir simulations. The mesoscale constraints need not be from a seismic survey: any spatially dense estimates of these properties can be used. Similarly, the fine-scale stratigraphic information included in the prior can be estimated using approaches other than surface-based models, such as mechanistic simulations or scaled experiments. The proposed workflow offers a flexible, multiscale framework for integrating stratigraphic, well, seismic, and related data.

 

AAPG Search and Discovery Article #90090©2009 AAPG Annual Convention and Exhibition, Denver, Colorado, June 7-10, 2009