Reducing Uncertainty in Paleoecological Models Using Fuzzy Logic

Anthony Gary, Glenn Johnson, and Ed Yu
Energy & Geoscience Institute, University of Utah, Salt Lake City, UT

Paleoecology, because it is based on biological populations, cannot be quantitatively defined by laws or formulas with the same precision as physical phenomena (e.g., Newtonian mechanics). Further, the manner in which industry biostratigraphic data are collected often makes it inappropriate for traditional statistical analysis. Biostratigraphers have therefore relied primarily on logical models. Unfortunately, the effectiveness of a logical model in representing a particular phenomenon is often severely compromised by associating too much precision to imprecisely-defined concepts. Examples include the concept of “high abundance”, and hard classification of continuous environmental gradients, such as subdividing bathymetry into discrete zones. The effect of misrepresenting precision and arbitrarily subdividing continuous systems in a model is that they contribute to error or uncertainty in the results. To address this problem the Technical Alliance for Computational Stratigraphy (TACS) with industry support has developed a fuzzy inference system (FIS) specifically for application to biostratigraphic data. The FIS uses a graphical interface to enable biostratigraphers to create functions that quantitatively represent their concept of linguistic variables, such as “high abundance of Hyalina balthica” and construct these variables into logical statements. Biostratigraphic data can then be analyzed by these logical statements, and the degree to which the data conform to these statements is represented by a value between zero and one. A value of one represents the “truth” of the statement, whereas zero represents no “truth” in the statement, and values between these extremes represent partial “truth”. By allowing logical statements to have partial memberships ambiguity can be effectively resolved and not contribute to uncertainty in the results.

AAPG Search and Discover Article #90078©2008 AAPG Annual Convention, San Antonio, Texas