A new approach to reservoir heterogeneity modelling: conditional simulation of 2-D parasequences in shallow marine depositional systems using an attributed controlled grammar

Abstract

An attributed controlled grammar (ACG) has been formally used to represent the parasequences of a clastic shallow-marine system. The lithofacies distribution has been conditionally simulated in two dimensions using the ACG. In knowledge representation, the ACG has been shown to have several advantages over context-free, programmed and attributed grammars. The ACG for the parasequences is manually constructed by domain experts based on a conditioning dataset, combined with related sedimentological knowledge. The dataset includes several geological sections measured from outcrops and interpreted from boreholes. A parasequence is decomposed into coastal plain, foreshore, upper shoreface, lower shoreface and offshore facies tracts and their boundaries. Within each tract, lithofacies distribution is described by the facies transition relationship, which can be constructed directly from the dataset and adjusted in terms of related sedimentological knowledge. The boundaries between the tracts are represented by point chains, whereas the facies transitions are controlled by a transitional probability matrix and both vertical and horizontal extensions of the corresponding lithofacies. The simulation results show the following features: (1) the simulation honors the conditioning dataset, (2) the lithofacies distribution simulated from the ACG shows increased variability compared to traditional interpolations between geological sections and (3) the simulated lithofacies distribution is controlled mainly by the uncertainty of the vertical and horizontal extension of each lithofacies, which cannot usually be obtained directly from the conditional dataset, and is not formally considered in traditional geological correlation and interpretation. Work is underway to quantify such lateral and vertical extension in present-day systems.