Quantitative sediment fingerprinting using a Bayesian uncertainty estimation framework

Numerous studies have shown how selected physico-chemical properties of sediments may be used as tracers to identify catchment sediment sources. Sediment fingerprinting has value in elucidating the linkages between erosion and downstream sediment delivery and potentially offers the opportunity to validate deterministic erosion models. However, quantitative fingerprinting is subject to considerable uncertainty throughout the research process, i.e. the inherent variability of source group properties, the number and distinctiveness of source groups, the relative discriminating power of different tracers, numerical issues associated with the un-mixing models, and further complications associated with nonlinear additivity, tracer transformation and enrichment. A Bayesian statistical framework was employed to assess two of the sampling issues using laboratory-based and synthetic data sets. The analysis shows source group contributions can be robustly derived, but source group variability and number of samples collected are key issues influencing performance.