Syntactic simulation of 1-D sedimentary sequences in a coal-bearing succession using a stochastic context-free grammar

Syntactic pattern recognition (SPR) provides a new symbolic computing paradigm that has a great potential for application in sedimentologic and stratigraphic theory. In an SPR system, a formal language consists of a set of sentences; each sentence is a string of terminal symbols (or words); a grammar is used to characterize syntactic structures of the language, which tells whether or not a specific sentence belongs to the language. A language represents a class of patterns that are similar to each other in some sense. A sentence from the language then represents an individual pattern of the class. The grammar defines the legality of members of the class in general. There are several key subdomains in this methodology, such as grammar inference, language generation or simulation, and language parsing. The objective of this paper is to show the possibility of using geological knowledge, obtained by domain experts (here sedimentologists), to construct valid grammars characterizing the vertical structure of different orders of sedimentary sequences, by applying the methodology to the analysis of a coal-bearing succession of the upper Limestone Coal Group, Pendleian (E1) in the Kincardine basin, Central Scotland. First, complete successions from two boreholes are encoded into syntactic strings of lithofacies symbols. Five groups of different high-frequency sedimentary sequences have been recognized. These high-frequency sequences combine to form a lower-frequency sequence. This geological knowledge is used in a heuristic sense to manually construct grammars for each of the five groups of sequences and for the whole succession. Finally, the constructed grammars, together with a thickness simulation for each lithofacies, are used to generate similar sedimentary sequences to test the validity of the grammars and therefore their further application to a formal analysis of sedimentary sequences. Comparison of simulated sequences with real sequences indicates that SPR is an encouraging technique in the analysis of sedimentary sequences. Copyright © 1996, SEPM (Society for Sedimentary Geology).