Application of multivariate modelling to detect hydrocarbon components foroptimal discrimination between two source rock types

The present study presents a multivariate procedure to reveal light hydroca rbon components which significantly distinguish between source rock thermal extracts. The two source rocks included in this study are the marine shale s of the Late Jurassic Spekk Formation and the coals and paralic shales of the Early Jurassic Are Formation offshore Mid-Norway. Because of the large number of components in the C-4-C-13 hydrocarbon fract ion of source rock extracts a multivariate approach was required. The proce dure consists of three distinct steps: (1) Principal component analysis of the whole data set for detection of non-significant individual components, This reduced the number of individual components from 46 to 22. (2) Separat e principal component analysis of the two source rocks (Are and Spekk) to d etect outliers. (3) Principal component modelling of each of the two source rocks after deletion of outliers and non-discriminating variables to detec t those hydrocarbon components which are most significant and robust for th e separation of the two source rocks. The resulting model shows that there is a definitive compositional differen ce between the source rocks investigated. (C) 2000 Elsevier Science Ltd. Al l rights reserved.