Fingerprinting Sources of Fine-grained Sediment Deposited in a Riverine System by GLUE

Identifying sediment provenance in the catchments is essential to the mitigation of sediments' negative effects. This study is the first attempt to quantify the uncertainty associated with contributions supplied by four individual sediment sources (e.g., gully erosion, sheet erosion, rill erosion, and channel bank) by the Generalized Likelihood Uncertainty Estimation (GLUE) with two statistical processes consisting of BRT-KW-DFA (Basic Range Test – Kruskal–Wallis—Discriminant Function Analysis), and RTM-KW-DFA (Range Test based on the Mean—Kruskal–Wallis—Discriminant Function Analysis). Then, the GLUE performance was validated by Virtual Sediment Mixtures (VSMs) samples. A total of 60 samples consisting of 20 samples from the materials deposited in the river bed and 40 source samples were collected from the individual sources consisting of rill erosion (n = 10), sheet erosion (n = 10), gully erosion (n = 10), and channel bank (n = 10). The samples were analyzed for 52 geochemical elements. The performance of GLUE with the final tracers was selected by two statistical combinations consisting of BRT-KW-DFA, and RTM-KW-DFA assessed by the statistical criteria and VSMs. Based on the results, the contributions modeled by GLUE with RTM-KW-DFA statistical process (with a mean value for the MAF (Mean Absolute Fit) = 89; GOF (Goodness Of Fit) = 98; MAE (Mean Absolute Error) = 16, and RMSE (Root Mean Square Error) = 19) are more accurate than the GLUE with BRT-KW-DFA (with a mean value for the MAF = 85; GOF = 97; MAE = 19; and RMSE = 23). Overall, the Sediment Source Fingerprinting (SSF) within the GLUE framework is a useful tool for identifying the sediment source in a wide range of catchments.