Fingerprinting sub-basin spatial sediment sources using different multivariate statistical techniques and the Modified MixSIR model

Abstract

Information on the relative contributions of sediment from different sources is needed to target sediment control strategies to prevent excess sediment delivery to receptors like dam reservoirs. The overarching scientific objective of this study was therefore to apportion sub-basin spatial source contributions to the supply of fine sediment in an erodible mountainous basin in north-eastern Iran to inform management. The technical objective was to satisfy the scientific objective using a source fingerprinting procedure based on composite signatures selected by different statistical tests. Nine potential geochemical tracers were measured on 21 sediment samples collected to characterise the three sub-basin spatial sediment sources and seven sediment samples collected at the outlet of the main basin. The statistical analysis employed to select three different composite fingerprints for discriminating the sub-basin sediment sources comprised: (1) the Kruskal–Wallis H test (KW-H), (2) a combination of KW-H and discriminant function analysis (DFA), and (3) a combination of KW-H and principal components & classification analysis (PCCA). A Bayesian un-mixing model was used to ascribe sub-basin source contributions using the three composite fingerprints. Using KW-H, the respective relative contributions from sub-basins 1, 2 and 3 were estimated as 45.6%, 3.8% and 50.6%, compared to 46.8%, 18.8% and 34.4% using KW-H and DFA, and 61%, 2.5% and 36.5% using KW-H and PCCA. Kolmogorov-Smirnov test pairwise comparisons of the distributions of predicted source proportions generated using different composite signatures confirmed statistically significant differences. The root mean square difference between the predicted source proportions based on different composite signatures was ~12%. This study therefore provides more evidence that source tracing studies should deploy a number of composite signatures selected using independent statistical tests to permit appraisal of the consistencies or otherwise in predicted source contributions based on the tracers used. The outputs of this preliminary study will be used to inform the spatial targeting of sediment mitigation.