Acquisition of cores is common in both geological studies of the shallow subsurface as well as in hydrocarbon exploration and production. Descriptions of cores are usually based on destructive analyses of small samples and on continuous records acquired by visual inspection. Such descriptions are not fully integrated, and their resolution, precision and accuracy tend to vary downcore. This research project is aimed at developing a fully quantitative method of core description with uniform (high) downcore resolution, and specification of the uncertainties of all measured and predicted quantities. XRF core scanning provides a fast, non-destructive and relatively inexpensive method to acquire high resolution data of a core. XRF core scanning is essentially a geochemical logging technique which aims at measuring element abundances through the acquisition of fluorescence spectra. By using a limited set of carefully selected calibration samples, analyzed using any well-established chemical technique, it is possible to turn element intensities derived from fluorescence spectra into chemical element abundances (Weltje & Tjallingii, 2008). Experiments have shown that this method is accurate (unbiased) and its precision is comparable to that attained in conventional destructive XRF analysis.
The current workflow in XRF spectrometry requires that spectra are first interpreted in terms of element intensities to enable calibration. Our aim is to perform calibration by a direct empirical approach which relates spectra to sample composition. Prediction of sediment properties (such as chemical composition) may be formulated as a generalized multivariate statistical estimation problem. By doing so, processing is no longer hampered by the interpretability of XRF spectra in terms of discrete element peaks. This allows us to reduce the prediction uncertainties and to predict abundances elements which are considered to be undetectable using XRF methods, but have been recorded in the calibration samples (work in prep.). The feasibility of this approach follows from the redundancy of geochemical compositional data: groups of elements are often closely associated because they reside in specific minerals with limited compositional variability.
Integration of the XRF spectra with secondary core information such as high-resolution digital images is a crucial part of the research. By combining various multivariate data sets, automatic interpretation of cores can be extended from purely (chemical or petrographic) composition-based classification, to a full-blown geological interpretation, which involves poro-perm prediction, grain-size estimation, classification of sedimentary structures, and further exploitation of color to support interpretations in terms of redox conditions. A powerful application of our data-processing scheme is the ability to identify common trends in multivariate data sets, e.g. grain size and geochemical composition. Such common trends may be used to filter stratigraphic records, as illustrated by the removal of grain-size related variation from geochemical data (Bloemsma et al., 2011). This innovative technique lies at the heart of the first statistically rigorous algorithm for chemostratigraphic correlation (work in prep.). Other potential applications include the simultaneous processing of XRD and XRF data, in order to facilitate interpretation of sediment composition in terms of internally consistent geochemical and mineralogical models.
References:
Weltje, G.J., Tjallingii, R., 2008. Calibration of XRF core scanners for quantitative geochemical logging of sediment cores: Theory and application. Earth and Planetary Science Letters, 274(3-4): 423-438.