Stochastic modelling of a fluvial reservoir: a comparative review of algorithms

A typical characterization of a clastic reservoir starts with the modelling of its major compartments (layers, fault blocks) followed by the modelling of the distribution in space of its dominant facies, or facies associations. Many alternative approaches to the modelling of such facies exist. This paper proposes a comparative review of three algorithms representative of the spectrum available: (1) the truncated Gaussian algorithm, (2) a nested indicator (binary) simulation approach and, (3) an object-based algorithm. A common data set originating from a fluvial reservoir of the Statfjord formation (North Sea) is used. The impact on flow simulation of alternative facies models is evaluated through the performance of a tracer flow simulation using a newly developed streamlines-based simulator which does not require any prior upscaling. The crisp geometry obtained from object-based modelling of facies does not always result in significantly different (or better) recoveries.