Extracting genetic information from coarse clastic modes

Abstract

Information gained from analysis of the mineralogical and petrological properties of the framework components of coarse-grained sedimentary rocks is central to most exploration and exploitation projects. Such data sets, characterized by the fact that the sum of the framework components equals 1.0 or 100%, pose a major stumbling block in attempting to analyze variation and strength of association among the components being studied. Common sense and knowledge of the principles of physical geology argue that variability among the framework components records information concerning the source area, the slope (dispersal system) or the sink (environment of accumulation) or a combination of some or all of these parameters. However, it is possible that a smooth variation on a framework component plot is due to the fact that percentages or proportions were formed in order to be able to make such comparisons in the first place. In this paper we present the results of the application of a set of procedures designed to overcome difficulties inherent in the statistical analysis of percentages/proportions to modal analyses of a set of sandstones selected from the literature. These procedures allow the investigator to assess systematically the entire data set for statistical independence. Unfortunately, these tests fail when applied to sets of ternary percentages. This argues for abandoning the practice of measuring the amounts of a number of components (for example, monocrystalline quartz, chert, alkali feldspar, plagioclase, igneous and metamorphic rock fragments) and reducing the number of components to three by forming linear combinations of the original variables (for example, total feldspar equals alkali feldspar plus plagioclase). Multivariate procedures, such as principal components analysis, should be examined as they make use of all of the available information.