ABSTRACT
Sediment fingerprinting estimates the proportional contribution of fine sediment from distinct catchment sources delivered to downstream receiving environments. Increased attention has focused on assessing the accuracy of source contribution estimates, particularly in relation to tracer selection and statistical un-mixing procedures. However, no studies have systematically tested the impact of source combination or dominance on the accuracy of source estimates. Here, we assess sensitivity to tracer type, selection, and number of sources, and examine how variations in the dominant sediment source affect the accuracy of source apportionments using numerical mixtures. Sources were sampled according to erosion process and land cover from a New Zealand catchment. Topsoil and subsoil (landslide) samples were collected from pasture, harvested pine, kanuka scrub, and native forest, while banks were sampled along the main channel. Samples were analysed for bulk geochemistry, fallout radionuclides, and compound specific stable isotopes (CSSIs). Source apportionment accuracy tended to decrease as source number increased, which reflected decreasing source discrimination. Tracer selection showed variations in accuracy but exhibited no clear pattern overall. Source combination and particularly the dominant source had the largest impact on accuracy, reflecting the level of discrimination for each source. Notably, channel bank was frequently identified as the dominant source when using CSSI tracers. While this partly reflected lower levels of discrimination, the CSSI apportionment was particularly sensitive to the use of post-unmixing corrections routinely applied to derive soil proportional contributions from isotopic proportions. This sensitivity likely related to the low organic carbon content in bank material and the assumption that source apportionments based on isotopic proportions can be corrected using a linear relationship with organic carbon content. These results indicate that the use of CSSI tracers in catchments where erosion sources exhibit large differences in soil organic carbon content may introduce significant unquantified error in source apportionments.
