A Database Approach for Constraining Geostatistical Reservoir Models: Concepts, Workflow and Examples

The sedimentary architecture of fluvial depositional systems is characterized by heterogeneities - manifested over a wide range of scales - that control hydrocarbon distribution and fluid-flow behavior; thus, subsurface subseismic-scale sedimentological features are often tentatively predicted by means of geostatistical modeling techniques, often conditioned by hard and soft sedimentological data obtained from outcrop successions or modern rivers considered to be analogous to the reservoir. We propose an alternative database approach as a way to derive such constraints from several classified case studies whose boundary conditions or architectural properties best match with the subsurface system that needs to be modeled.

The relational database characterizes the fluvial architecture of classified case studies from the stratigraphic record and modern rivers at three different scales of observation, corresponding to three types of genetic unit (large-scale depositional elements, architectural elements and facies units) that constitute the building blocks of reservoir models. The database case studies can be filtered on their boundary conditions or architectural properties, generating composite datasets consisting of genetic-unit proportions, dimensions and transition statistics with which to inform and condition fluvial reservoir models.

The potential value of the database in providing constraints to stochastic reservoir models is demonstrated by employing both object-oriented and pixel-oriented techniques to generate unconditional idealized models of fluvial architecture, associated to given system parameters (e.g. river pattern), giving a special focus on the aptness of the hierarchically-nested database output to the integration of different modeling techniques into the same reservoir model, with the scope to improve and/or validate predictions. In addition, the simulation outcomes work as graphical representations of stratigraphic volumes of given synthetic depositional/facies models of fluvial architecture and could be employed as training images to constrain multi-point statistics-based reservoir models.

AAPG Search and Discovery Article #90142 © 2012 AAPG Annual Convention and Exhibition, April 22-25, 2012, Long Beach, California