fingR: A Framework for Sediment Source Fingerprinting

Tracer selection for sediment source fingerprinting among measured properties according to the most conventional approach, the so-called three-step method.

The three steps are: (1) assessing conservative behaviour with range test, (2) assessing the capacity of properties to discriminate source with a non-parametrical test (Kruskal-Wallis H-test or two-samples Kolmogorov-Smirnov test), (3) a discriminant function analysis (DFA) forward stepwise selection based on Wilk's Lambda criterion to identify the best subset of predictors among the identified tracers (i.e. conservative and discriminant properties). The selected tracers are then used for source contribution modelling.

To assess the accuracy of source contribution modelling, virtual mixtures are used as mixture with known contributions. Here, some functions help to generate their contributions and property values as simple proportional mixtures of sources.