De-Convoluting Mixed Crude Oils in Prudhoe Bay Field Using Chemometrics

Peters, Kenneth E.¹, L. Scott Ramos², John E. Zumberge³, Zenon C. Valin¹, Kenneth J. Bird¹ (1) U.S. Geological Survey, Menlo Park, CA (2) Infometrix, Inc, Bothell, WA (3) GeoMark Research, Inc, Houston, TX

Chemometrics is multivariate statistics applied to chemical problems. It is a powerful tool for genetic classification, correlation, prediction of physicochemical properties, and de-convolution of mixtures. Source- and age-related biomarker and isotopic data were measured for >1000 crude oil samples from wells and seeps north of ~55^oN latitude. A unique, multi-tiered chemometric decision tree was created that allowed automated classification of 31 genetically distinct Circum-Arctic oil families. This new method, which we call decision-tree chemometrics, utilizes PCA (principal components analysis) and many tiered KNN (K-Nearest Neighbor) and SIMCA (soft independent modeling of class analogy) models to classify and assign confidence limits for newly acquired oil samples and source-rock extracts. Geochemical data for each oil sample were also used to infer the age, lithology, organic matter input, depositional environment, and identity of its source rock. The data identify nine petroleum systems on the North Slope of Alaska. Crude oils from the giant Prudhoe Bay Field were studied to assess the relative volumetric contributions from different source rocks. Contrary to conventional interpretations, the results of ALS (alternating least squares) analysis show that oils in the field originate mainly from Cretaceous Hue Shale rather than Triassic Shublik Formation source rock. These results were confirmed using both ratios and concentrations of biomarkers. PLS (partial least squares) analysis of the biomarker and isotope data allows accurate predictions of the physicochemical properties of these oil samples, such as sulfur content.