Geoscience Reference
In-Depth Information
•
Can local stakeholders be involved in defining modelling strategies for local conditions in a way that
leads to more realistic rainfall-runoff models (a socially “bottom-up” approach within the models of
everywhere framework)?
Finally, a last word about the equifinality of models in representing rainfall-runoff processes. A major
theme of this topic has been the apparent equivalence in performance of different model structures or
different sets of parameters within a model structure. This has been referred to as “equifinality” to suggest
that it is a generic problem, not simply a difficulty of finding the “right” model. But equifinality smacks
of relativism, something that is not very acceptable to many scientists, including hydrological scientists.
There should, after all, be an answer to aim for: a single, proper description of the catchment system
that would be generally applicable and recognised as acceptable for predictive purposes. Perhaps such
a description will develop in the future but even if such a model were available, applying it would not
necessarily avoid the equifinality problem unless techniques for the direct measurement of effective
parameters were evolved at the same time. The unique nature of every catchment system will always
require the estimation of parameter values to reflect that uniqueness. Without direct measurement some
form of conditioning on observed responses would be required and, with what would undoubtedly be a
complex model, some equifinality of different parameter sets will undoubtedly ensue.
Personally, however, I see this as a positive concept for the very reason that it leads naturally to a
hypothesis testing framework and to a focus on observational data in testing hypotheses. If multiple
models can be found that are consistent with available observations, allowing for the uncertainty in those
observations, then they can be treated as hypotheses to be tested. That is a methodology that we can call
scientific, even if we have to develop new testing methods to deal with epistemic uncertainties. The nature
of catchment systems and hydrological data is such that there might still be a plurality of models that
are consistent with the data ...but that is the nature of the prediction problem. Looking into the future,
the limitations of observation techniques for subsurface flow pathways and processes, in particular, are
likely to constrain how well we can predict rainfall-runoff processes. That is what makes rainfall-runoff
modelling still challenging and will do so for some considerable time to come. Good luck!
