Defensible Model Complexity: A Call for Data‐Based and Goal‐Oriented Model Choice

Abstract

Article impact statement: This article provides a structured discussion of defensible model complexity in view of modeling goals and available data.

Forecast First: An Argument for Groundwater Modeling in Reverse

“…After all, the modeling analysis was undertaken to increase the state of knowledge related to the forecasts, so it is only right that the forecasts get more attention during model construction and usage. Additionally, by focusing on the forecasts, a more robust analysis of the appropriate level of complexity can be undertaken (e.g., Guthke ), where complexity is driven not only by the ability to reproduce the past, but also simultaneously by the need to provide robust estimates of forecast uncertainty.…”

Using Models to Identify the Best Data: An Example from Northern Wisconsin

“…Similarly, if the piecewise‐constant zones already encompass sufficient head observations, the data worth is likely to be low, regardless of the distribution of the existing data, because the observations all inform the same parameter. In this way, overly simplified models can be limited in their ability to incorporate information from new data, because their uncertainty is expressed in the structure of the model, and therefore difficult to quantify (Hunt et al ; Fienen et al ; Anderson et al ; Guthke ). This presents an important consideration for evaluating competing models in the environmental decision‐making process (Ferre ).…”

Using Training Images to Build Model Ensembles with Structural Variability

“…This may have the practical advantage of improving buy‐in of clients for more sophisticated modeling analysis and additional data collection ( MacMillan ). Simultaneously, constructing these images can support conversations regarding the meaning and impact of model complexity (Guthke ) and can be used in explicit efforts to build model ensembles with diverse model structures (de Pasquale ).…”

“…The objective of this article is to describe a method to propose multiple, competing, but geologically realistic structural models. These can form the basis for considerations of model complexity (Guthke ) and can help to balance model structure and parameter uncertainty in model ensembles (de Pasquale ).…”