Complex or Simple—Does a Model Have to be One or the Other?

The primary tasks of decision-support modelling are to quantify and reduce the uncertainties decision-critical model predictions. Reduction predictive uncertainty requires assimilation information. Generally, this information resides in two places: 1) expert knowledge emerging from site characterization 2) field measurements present historical system behavior. former is uncertain should therefore be expressed stochastically a model. range parameter possibilities can then constrained through history-matching. Implementation these Bayesian principles places conflicting demands on level structural complexity. A high complexity facilitate expression by establishing details that recognizable experts, supporting parameters bear close relationship real-world hydraulic properties. However, such models often run slowly numerically delicate; history-matching becomes difficult or impossible. In contrast, if endowed with enough parameters, structurally simple achievement good fit between outputs measurements. values which may looser properties less receptive born knowledge. design process one compromise. paper we describe methodology reduces cost compromise allowing inform use complementary strategic, but not excessive, stochastic, fast-running no We demonstrate approach using case used support management stressed coastal aquifer. empirically validate synthetic