Statistical modelling of compositional trends in sediments

Abstract Trends in compositional diagrams are frequently used to describe and interpret the imprint of certain processes on sediment composition. Most of these trends are intuitive and “hand-drawn”, although a numerical description of compositional changes involved with these processes would allow for a quantitative assessment of the observed sediment compositions and the proposed processes. This paper suggests a statistical approach to model compositional linear trends, which allows for a mathematically rigorous description of changes in sediment composition related to a specific process or a combination of processes. Three case studies dealing with different aspects of the sedimentary cycle demonstrate the usefulness of the approach. In the first example, a modelled average global chemical weathering trend is used to calculate an appropriate measure for the degree of chemical weathering of the suspended load from some of the world's major rivers. In the second example, data on precipitation and relief from different watersheds are quantitatively related to the petrographic composition of fine-grained sands from low-order streams derived from these watersheds. The third example is an attempt to quantify the interdependence of chemical sediment composition and grain size in the absence of chemical weathering. Superposition of individual compositional linear trends allows for modelling sediment composition in more complicated case studies. Although primarily a forward modelling approach, it can be also used inversely, and, therefore, has strong potential for tracing back the sediment-forming processes in order to reconstruct the original source rocks of the sediment.