Advancing sediment fingerprinting techniques: The importance of considering sediment mixtures data in tracer selection

Abstract

Sediment fingerprinting has become a key tool to identify and quantify sediment sources within a catchment. The technique involves statistical testing of a range of properties of source materials to identify a set of tracers that can effectively discriminate between different potential sources before estimating the source contributions with unmixing models. However, despite its increasing popularity among researchers, there is a lack of standardized procedures for tracer selection, which is crucial to estimating a reliable contribution of sediment sources. The most widespread methodology consisted of an initial mass conservation test, usually termed range test (RT), followed by the use of Kruskal-Wallis (KW) and discriminant function analysis (DFA) tests. However, KW and DFA even though identifies the best combination of tracers that provide the maximum discrimination between sources, do not incorporate the information of the sediment mixtures in the analysis. Novel methods highlight the importance of selecting the right tracers for each individual mixture and avoid the inclusion of tracers out of consensus or with non-conservative behavior by using consensus ranking (CR) and consistent tracer selection (CTS) methods. This contribution addresses the role of selecting appropriate tracers, demonstrating their impact on the results of the unmixing model. The main objectives are to emphasize the importance of considering the information provided by the sediment mixture in the selection of tracers and to pay attention to the impact of having sediment mixtures with values below the detection limit of the tracer being selected for source discrimination. A set of experimental and real sediment mixtures were selected to explore the different tracer selection methods, comparing the tracers selected and the contribution of sources obtained using the FingerPro unmixing model. We present the results of rigorously testing methodologies with the aim of understanding and assessing the suitability of each tracer selection method to select a combination of statistical and process-based criteria to select appropriate sediment properties for the unmixing models. Our findings highlight the importance of considering the information on the sediment mixture information for the selection of potential tracers, an aspect often neglected by conventional methods. This oversight can result in biased findings due to the use of tracers that are either not coherent or not conservative.