Consistency of patterns in concentration‐discharge plots

Concentration‐discharge (c‐Q) plots have been used to infer how flow components such as event water, soil water, and groundwater mix to produce the observed episodic hydrochemical response of small catchments. Because c‐Q plots are based only on observed streamflow and solute concentration, their interpretation requires assumptions about the relative volume, hydrograph timing, and solute concentration of the streamflow end‐members. Evans and Davies [1998] present a taxonomy of c‐Q loops resulting from three‐component conservative mixing. Their analysis, based on a fixed template of end‐member hydrograph volume, timing, and concentration, suggests a unique relationship between c‐Q loop form and the rank order of end‐member concentrations. Many catchments exhibit variability in component contributions to storm flow in response to antecedent conditions or rainfall characteristics, but the effects of such variation on c‐Q relationships have not been studied systematically. Starting with a “baseline” condition similar to that assumed by Evans and Davies [1998], we use a simple computer model to characterize the variability in c‐Q plot patterns resulting from variation in end‐member volume, timing, and solute concentration. Variability in these three factors can result in more than one c‐Q loop shape for a given rank order of end‐member solute concentrations. The number of resulting hysteresis patterns and their relative frequency depends on the rank order of solute concentrations and on their separation in absolute value. In ambiguous cases the c‐Q loop shape is determined by the relative “prominence” of the event water versus soil water components. This “prominence” is broadly defined as a capacity to influence the total streamflow concentration and may result from a combination of end‐member volume, timing, or concentration. The modeling results indicate that plausible hydrological variability in field situations can confound the interpretation of c‐Q plots, even when fundamental end‐member mixing assumptions are satisfied.