Less is more? A novel method for identifying and evaluating non-informative tracers in sediment source mixing models

Abstract Purpose Accelerated soil erosion poses a global hazard to health. Understanding and sediment behaviour through fingerprinting enables the monitoring identification of areas with high delivery. Land-use specific source apportionment is increasingly determined using Bayesian mixing model MixSIAR compound-specific stable isotopes (CSSI). Here, we investigate CSSIs fatty acid (FA) tracer selection novel method identify effect non-informative tracers on performance. Methods To evaluate CSSI selection, mathematical mixtures were generated soils (n = 28) from Rhine catchment upstream Basel (Switzerland). Using continuous ranked probability (CRP) skill score, MixSIAR’s performance was evaluated for 11 combinations FAs 15 δ N as line offset tracer. A scaling discrimination analysis (SDA) also developed non-unique spaces. Results FA only overestimated pasture contributions while underestimating arable contributions. When compared models tracers, utilizing resulted in 28% improvement CRP score. + 13 C 26 optimal set resulting 62% relative all FAs. The SDA demonstrated how have space thus behave tracers. Importantly, inclusion decreased Conclusions These results indicate that did not handle effectively. Accordingly, it may be advantageous remove where feasible, permutations should assessed optimize selection. Application these steps can help improve advance ultimately aid improving mitigation management strategies.