What should decision-support groundwater modelling seek to accomplish? Should it try to build a digital replica of what happens underground? Obviously not, because this is impossible. Not nearly enough is known about the subsurface and its hydraulic properties to do this.
This monograph (written by John Doherty and Catherine Moore) is meant for modellers and non-modellers alike. It is easy to read, but rigorous. Its purpose is to assist modellers in choosing a level of model complexity that is appropriate for their decision-support needs. It is also meant to help non-modellers assess the decision-support appropriateness of models built by others.
“Appropriate” implies metrics. The authors provide these. Basically, decision-support modelling is all about quantifying and reducing the uncertainties of predictions that matter. It is the job of decision-support modelling to harvest information that can do this. In many circumstances, the best model for the job may not be too complex at all.
So how complex should a model be? It should be complex enough to glean information from wherever it resides, and to direct that information to predictions that matter. All aspects of a model’s design (including its complexity) should be referred back this primary imperative.