|
Applications to the Solution of Real Geological Problems |
Delft University of Technology, Faculty of Civil Engineering and Geosciences, Department of Geotechnology, Applied Geology Section, PO Box 5028, NL-2600GA Delft, The Netherlands g.j.weltje{at}citg.tudelft.nl
A popular model proposed by W. R. Dickinson and co-workers in the early 1980s relates the composition of sandstones to the plate-tectonic setting of the sedimentary basins in which they were deposited. The present study is devoted to revision and testing of the ‘Dickinson model’ based on the original data which comprise 11 000 thin sections point-counted by hundreds of different operators over a period of three decades. Statistical analyses based on Aitchison’s additive log-ratio transformation are used to obtain an optimal partitioning of ternary compositional spaces into ‘provenance fields’ and combined with stochastic simulation to assess the success ratio of the optimized ‘Dickinson model’. Results indicate that differences between the grand means of each of the three major provenance associations (continental block, magmatic arc and recycled orogen) are highly significant, whereas overall inferential success ratios range from 64% to 78% in the four ternary systems studied. Current methods of dealing with sands of mixed provenance are unsatisfactory. To improve provenance models, the use of ternary subcompositions should be replaced by analyses of the full six-part (Qm, Qp, P, K, Lv, Ls) composition, and their covariance structure could be employed to ‘unmix’ samples into end-member provenance types.
