The representation of sediment source group tracer distributions in Monte Carlo uncertainty routines for fingerprinting: An analysis of accuracy and precision using data for four contrasting catchments

Summary

This paper systematically compares four approaches to representing tracer distributions in Monte Carlo sediment fingerprinting models, addressing a substantial methodological inconsistency across published studies. Using intensively sampled source materials from four catchments, the authors created virtual sediment mixtures and evaluated model accuracy and precision. The 25th–75th percentile distribution achieved the lowest mean inaccuracy (8.8%) and imprecision (8.5%), whilst normal distributions performed poorly (16.3–16.9% inaccuracy), with implications for improving reliability of sediment source apportionment in hydrological research.

UK applicability

The findings are directly applicable to UK catchment management and soil erosion assessment, where sediment fingerprinting is used to identify and quantify erosion sources. Adoption of the recommended 25th–75th percentile distribution approach could improve the accuracy of sediment source attribution in UK river basin management plans and land-use impact studies.

Key measures

Mean inaccuracy (%) and imprecision (%) of source apportionment estimates across four distribution representation methods; model performance when virtual mixtures comprised 10–100% of retrieved source samples

Outcomes reported

The study evaluated the accuracy and precision of four different statistical approaches for representing tracer concentrations in sediment fingerprinting un-mixing models using virtual mixtures from four contrasting catchments. Results showed that the 25th–75th percentile distribution and sample-based distributions substantially outperformed transformed and untransformed normal distributions in source apportionment accuracy.

Topic tags