THE BIG HAMMER: COMPLEXITY, COUPLING, AND CALIBRATION OF GROUNDWATER AND SURFACE WATER SYSTEMS

Some societal decisions require simulation of the coupled response of the groundwater and surface water (gw-sw) systems. The need, however, does not obviate the difficulties in such coupled modeling. First, the runtime of a coupled gw-sw model is considerably longer than uncoupled models, which can limit the exploration of optimal parameters and prediction uncertainty. Second, artifacts from parameter overlap in the uncoupled models can color calibrations of the coupled model. Finally, the hydrologic ‘memory’ of the groundwater system can require long simulation periods to reach dynamic equilibrium. As a result, parameter estimation and uncertainty analysis – techniques well developed for groundwater systems – have had little reported success in coupled gw-sw modeling.

Recent work at the USGS Trout Lake WEBB site in Wisconsin has developed a beginning methodology to address this deficiency. Parameter identifiability statistics provide a quick method to assess what information is, and is not, present in existing or proposed datasets. "Super-observation" determination provides a systematic mechanism to extract salient information from measured data, including time-series data. A “sequentially linked” calibration approach can keep run times relatively short but facilitate more correspondence between the groundwater and surface-water models. This approach appears to be superior to using a completely uncoupled approach for getting in the ballpark. Parameter estimation approaches such as these help ensure that the maximum amount of information is extracted from expensive field data, while also providing a receptacle for expert (soft) knowledge of the system. Moreover, they not only help identify what is not known about the system, they help evaluate the efficiency of potential future data collection conducted to address the deficiency. Such capabilities will be important as more coupled groundwater/surface-water models are used to tackle societally relevant questions such as climate change and ecological or economic decision endpoints. However, simple models that capture the salient details of the complex coupled model will still be needed for some societal applications, such as those required by decision-support systems.