Recent developments in quantitative stratigraphy

Abstract

During the past 10 years, about 150 scientists in 25 countries have collaborated under the auspices of the International Geological Correlation Programme Project No. 148: Evaluation and Development of Quantitative Stratigraphical Correlation Techniques. This paper reviews mathematical methods of stratigraphical correlation, mainly in biostratigraphy but also in chronostratigraphy and lithostratigraphy. Sequencing methods treat the relative order of stratigraphic events such as the highest occurrences of fossil taxa as observed in many sections. Intervals between successive events in an ordered sequence can be estimated and the results expressed in linear time if a subgroup of the stratigraphical events can be dated. Recently such methods have been used to develop a new deep-water benthic foraminiferal zonation for the Cenozoic strata of the Central and Viking Grabens, North Sea. Several regional hiatuses of 2 to 5 m.y. in duration, can be recognized and matched to changes in sea level. The same methods have been used for automated isochron contouring with error bars in depth or time units in Cenozoic and Cretaceous basins, off eastern Canada. Such information may be used for automated basin history analysis. Time-successive assemblages of fossils can be established by using multivariate methods applied to co-occurrences of events or by the method of unitary associations in conjunction with graph-theory on the overlap of stratigraphical ranges. Other methods for stratigraphical correlation reviewed in this paper include the composite standard method and estimation of the age of chronostratigraphical boundaries using maximum likelihood and cubic spline interpolations. Attractions of quantitative stratigraphy are the use of rigorous methodology which highlights many properties of the data, the ability to handle large and complex data bases in an objective manner, and statistical evaluation of the uncertainty in the results. Generally, little conceptual orientation is required in order to use these methods and thereby gain more information from a particular dataset.