We are developing a new method of characterising sedimentary deposits that will allow complete descriptions of the patterns of deposits in an individual system (e.g. Amazon submarine fan) or a system class (e.g. sand-dominated submarine fan) to be encoded in a format that is intuitive to geologists and computer-friendly.
The method is based on formal language theory. Using this method each environment is encoded as a grammar. Grammars are a very useful method for describing classes of complex objects as they provide a precise description in a compact form; furthermore they can be graphically represented, which will make them easily intelligible to the geologist.
In a grammar a complex object is described in terms of its components (which are represented by symbols) and a set of rules that describe how the components are assembled to form a complex object. For a sedimentary system we can represent each architectural element (at the scale of interest) by a symbol. We can then state the rules that describe the valid patterns of elements in a particular depositional system. For example, the rules may state that a channel fill deposit must be flanked by levee deposits; or that a crevasse splay deposit lies adjacent to the outside bend of a channel segment; or that, in a sinuous channel fill, a channel segment with an s-bend should be followed by a channel segment with a z-bend.
Our geological grammars are probabilistic, this means that a probability is assigned to every rule to allow for variation in spatial arrangement of elements. For example, a probability can be used to control how frequently a crevasse splay deposit occurs on the outside bend of a channel instead of a levee deposit. Additionally, the size or shape attributes of a symbol may be described by a probability distribution so that the natural variation inherent in a sedimentary system can be included in the class description.
We are also developing a computer program, called a parser, to generate examples of facies models using the descriptions provided by the grammars. The intended application of this program is to generate multiple realisations of facies models for petroleum reservoir simulation.
In this presentation we will demonstrate how our geological grammars are constructed and show examples of models generated by the parser using specified grammars.