RECHARGE AND REGIONAL GROUND-WATER MODELS – THREE COMMON PROBLEMS

Mary Stoertz pioneered efforts to use ground-water models to estimate recharge magnitude and recharge/discharge zones. More commonly, recharge is an input to models that seek to account for sources of water to sinks. While it appears a simple matter to estimate recharge beforehand using techniques that are independent of the model, experience has shown that problems repeatedly emerge that complicate the process. There are at least three groups of problems to confront: 1) estimating recharge at the proper spatial and temporal scale for the objectives of the model; 2) reconciling the density of the surface-water network, which represents the dominant discharge mechanism in most ground-water systems, to the recharge estimates; and 3) decoupling the recharge estimates from estimates of the flux targets that are used to calibrate the model. If these considerations are ignored, it is possible that an otherwise acceptable recharge estimation technique could be applied in ways that undermine the integrity of the model. Even if they are accommodated, it is likely that tradeoffs will occur during model construction and calibration that are not always explicitly recognized.

A USGS pilot study of the water system in the Great Lakes Basin provides an opportunity to confront en masse the common problems of recharge estimation in regional flow models. A two-million-node transient flow model is being constructed for the Lake Michigan Basin and adjacent areas with grid spacing about one mile on a side. One objective of the model is to account for all the recharge to the system that is available to surface-water bodies and wells. Particular challenges include: 1) estimating recharge over rural and urban areas with spatially-varying geologic and climate conditions, 2) estimating changes in recharge related to development and climate change, 3) designing a surface-water network for the model that accommodates the circulation of all available recharge without inserting so many head-dependent discharge zones that the water table becomes fixed rather than a solved surface, 4) comparing recharge estimates with stream baseflow targets that are independently calculated, and 5) balancing recharge uncertainty against uncertainty in hydrogeologic properties during the parameter estimation phase of calibration. Partial and interlocking solutions to these problems are offered.