M.S. Blondes, Mark A. Engle, N.J. Geboy
In the geosciences, radiogenic and stable isotope data are often used along with chemical concentrations to help determine sources of and processes affecting geochemical variation. Though some work combining isotopic data with concentration data has been done using compositional data analysis (CoDa), a number of questions remain. Tolosana-Delgado and others (2005) showed that concentrations (e.g. SO4) can be split into separate parts (e.g. 34SO4, 32SO4) based on their isotopic ratios. They also showed that an isometric log-ratio (ilr) transform of a stable isotope ratio is proportional to values expressed in classical delta notation and therefore the delta notation can simply be scaled to compare isotopic data and log-ratio transformed elemental/ionic data simultaneously. Puig and others (2011) applied this result to discriminant analysis of groundwaters by separating centered log-ratio (clr) transformed compositional data from untransformed isotopic data and scaling both sets to an equal variance. We are expanding on this idea to show the utility of ilr transformations and clr-biplots for combining elemental data with not only stable isotope data, but also with radiogenic isotopes, which are generally not presented using delta notation. In the present study, we use multiple data sets of deep formation brines and compare traditional mixing models to their CoDa counterparts to delineate fluid movement between reservoirs. Concentrations of individual isotopes are calculated using isotopic ratios and global mean isotopic abundances. One key result is that isotope parts (e.g. 18O, 17O, 16O, 2H, 1H, 87Sr, 86Sr) can simply be modelled by the major ion concentration (H2O, Sr) in a clr-biplot as they are perfectly dependent. Another important result is that an ilr transformation of radiogenic isotope parts (e.g. 86Sr and 87Sr in 87Sr/86Sr) can, like stable isotopes in delta notation, be treated as a linear function of the isotopic ratio, scaled only by a constant.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados