A new look at uncertainty in end member mixing models for streamflow partitioning

Abstract

Chemical and isotope based hydrograph separation methods, such as end member mixing analysis (EMMA1), have been utilized in recent decades because they offer a better explanation of stream chemistry than can be obtained from flow partitions derived using graphical or mathematical filter hydrograph partitioning techniques (Chapman and Maxwell, 1996; Pinder and Jones, 1969). However, use of such constituent based hydrograph separation methods results in uncertain streamflow partitions, attributable, in part, to uncertainty in the characteristic chemical or isotope content of source waters (Figure 1) (Soulsby et al., 2003). This paper explores additional causes of uncertainty when using EMMA for flow partitioning. This paper shows that: • hydrograph partitioning based on measured end member constituent concentrations should not be assumed to be correct just because the stream chemistry is adequately explained; • propagating uncertainty estimated from sample variations will underestimate uncertainty in flow partitions; and • numerical problems during computation introduce considerable uncertainty into solved end member contributions. Sampling and measurement errors exacerbate the problem. © 2009 Creative Commons Attribution 4.0 International CC BY License