Transparency in Modeling: Mylar Overlays Might Be out of Fashion, but We Still Need to Shed Light on the System

Processor-capacity increases have followed Moore’s law for the last forty years. Software, data accessibility, and numerical methods have also experienced considerable improvements. The number of new MODFLOW packages, their sophistication, and the rate at which they are produced exhibit a similar trend. GIS programs and GUIs are more powerful, allow additional features to be represented, but sometimes seem to eliminate the insight or expertise prerequisites for building ground-water models. The availability of tools and resources has allowed ground-water models to incorporate more detail spatially, temporally and in terms of processes. Regardless of whether the additional detail is warranted, increased complexity requires corresponding efforts to decipher the simulated processes, and verify that results are consistent with the conceptual intent of the simulation. Existing tools provide the means, but their implementation seems to lag model-construction complexity. This work focuses on penetrating model complexity in order to unmask any unintended simulation results. For transient simulations, assessing the relation between stresses and observations before even using a model provides a good first step. Statistics generated from model runs, including correlations and sensitivities, can demonstrate the numerical model’s ability to follow the conceptual model. Analysis of simulated values, whether statistically or lumped according to the physical system, should answer the question: "are these results expected?" Regardless of model fit, answering this question requires understanding of the physical system and numerical simulation. Aspects of several models from the Western United States are used to demonstrate the benefits of applying well-established tools, and some less common techniques. The emphasis is on making certain that the simulated processes are as intended, as a critical step towards robust model development. Specific examples include comparing farm budget changes to water level hydrographs, using sensitivities to identify problematic observations, and visualizing fluxes to improve water budget assessments.