Paper No. 3
Presentation Time: 1:45 PM


LEAF, Andrew T., U.S. Geological Survey, Wisconsin Water Science Center, 8505 Research Way, Middleton, WI 53562, HUNT, Randall J., Wisconsin Water Science Center, U. S. Geological Survey, 8505 Research Way, Middleton, WI 53562 and FIENEN, Michael N., Wisconsin Water Science Center, U.S. Geological Survey, 8505 Research Way, Middleton, WI 53562,

Because groundwater is influenced by system properties that are often unseen and approximately characterized, groundwater models have become the primary vehicle to teach quantitative groundwater science. Recently, many advances have occurred in the application of parameter estimation for calibration and uncertainty analysis using groundwater models. The advance of highly parameterized models is well-suited for hydrogeology as it helps encompass system complexity geologists are well aware of. Highly parameterized models, however, have underscored concepts important to teach to future hydrologists. Non-uniqueness is inherent to almost all quantitative groundwater representations because we do not have the ability to sufficiently characterize system processes, properties, and hydrologic outputs. Likewise, this inability to characterize the system inherently results in uncertainty in model predictions. With decision-makers and resource managers increasingly interested in “how well do you know this?”, it is important to appropriately convey these topics.

However, such concepts are seldom formally taught in groundwater courses. We present two examples from real-world problems, in which readily available open-source software tools were used to investigate and visualize model non-uniqueness and uncertainty for predictions of interest. The first example uses the Pareto feature within the PEST software suite to explore the trade-off between soft-knowledge of the hydrogeologic system and model fit to the calibration dataset. The second example uses the PREDUNC utilities in PEST to demonstrate that not all areas for field data collection have the same utility, by identifying areas where additional data collection would most reduce uncertainty for a model prediction. These examples provide an effective vehicle for teaching non-uniqueness and uncertainty to groundwater students, by allowing them to formally test their quantitative understanding, and subsequent conceptualization of the unseen groundwater system.