Equifinality

Equifinality is the finding that many different models, or many parameter sets within one model, can fit the observed data about equally well. The term comes from general systems theory, where it means that the same end state can be reached by different paths. Keith Beven brought it into hydrology in the 1990s, and it is now central to how modellers think about uncertainty. It arises because hydrological models usually have more parameters than a flow record can pin down. Parameters interact: a larger soil store with faster drainage can produce nearly the same hydrograph as a smaller store with slower drainage. Calibration then finds a long, flat ridge of good scores rather than one clear peak. A simple case shows the idea. Suppose runoff = a × b × rainfall, where a and b are two separate parameters. With a = 0.5 and b = 0.6, 100 mm of rain gives 30 mm of runoff. With a = 0.6 and b = 0.5, it also gives 30 mm. Flow data alone cannot tell the two sets apart, yet they may mean very different things about the catchment, and they may give different answers once conditions change. This matters because a model that fits the past for the wrong reasons can fail when used to predict under new land use or climate. Equifinality is why many hydrologists report a range of results instead of a single best run. The GLUE method (Generalized Likelihood Uncertainty Estimation) is one response: it runs many parameter sets, keeps those that score above a chosen threshold as behavioural, and uses them all to set prediction bounds. Equifinality can be reduced, though rarely removed, by using more kinds of data (groundwater levels, soil moisture, snow, tracers), by using simpler models with fewer free parameters, and by testing parameter sets on periods and catchments not used in calibration.