Metamodels to bridge the gap between modeling and decision support
Explore the source record for details and available documents.
Groundwater↗
Geology topics
Daniel T. Feinstein
Publications and source records attributed to Daniel T. Feinstein.
Optimization techniques using MODFLOW-GWM
An important application of optimization codes such as MODFLOW-GWM is to maximize water supply from unconfined aquifers subject to constraints involving surface-water depletion and drawdown. In optimizing pumping for a fish hatchery in a bedrock aquifer system overlain by glacial deposits in eastern Wisconsin, various features of the GWM-2000 code were used to overcome difficulties associated with: 1) Non-linear response matrices caused by unconfined conditions and head-dependent boundaries; 2) Efficient selection of candidate well and drawdown constraint locations; and 3) Optimizing against water-level constraints inside pumping wells. Features of GWM-2000 were harnessed to test the effects of systematically varying the decision variables and constraints on the optimized solution for managing withdrawals. An important lesson of the procedure, similar to lessons learned in model calibration, is that the optimized outcome is non-unique, and depends on a range of choices open to the user. The modeler must balance the complexity of the numerical flow model used to represent the groundwater-flow system against the range of options (decision variables, objective functions, constraints) available for optimizing the model.
Conference Paper↗
MODFLOW-NWT – Robust handling of dry cells using a Newton Formulation of MODFLOW-2005
The first versions of the widely used groundwater flow model MODFLOW (McDonald and Harbaugh 1988) had a sure but inflexible way of handling unconfined finite-difference aquifer cells where the water table dropped below the bottom of the cell—these "dry cells" were turned inactive for the remainder of the simulation. Problems with this formulation were easily seen, including the potential for inadvertent loss of simulated recharge in the model (Doherty 2001; Painter et al. 2008), and rippling of dry cells through the solution that unacceptably changed the groundwater flow system (Juckem et al. 2006). Moreover, solving problems of the natural world often required the ability to reactivate dry cells when the water table rose above the cell bottom. This seemingly simple desire resulted in a two-decade attempt to include the simulation flexibility while avoiding numerical instability.
Ground Water↗
Since “Groundwater and surface water–A single resource”: some U.S. Geological Survey advances in modeling groundwater/surface-water interactions
Field and interpretive studies conducted by T.C. Winter and U.S. Geological Survey colleagues, and summarized in the 1998 publication “Groundwater and Surface Water – A Single Resource”, inspired a new generation of research centered on extensions of the groundwater-flow code MODFLOW to more sophisticated simulation of coupled groundwater and surface-water systems. Guided by emerging concerns with water availability, safe yields from wells, health of aquatic habitat, and climate change, the changes to MODFLOW involve: 1) the ability to more precisely and accurately represent the interface between surface and subsurface flows and 2) the consideration of a variety of mechanisms that influence their interaction. A review of the most important changes to the code is supplemented in this article by selected case studies in an effort to show the scope of the advances. The updates discussed include the Streamflow Routing (SFR), Lake (LAK), and Unsaturated-Zone Flow (UZF) Packages in MODFLOW-2005 and the Groundwater Management (GWM), Local Grid Refinement (LGR), and Newton (NWT) formulation versions of MODFLOW-2005. New developments feature the integration of rainfall-runoff modeling with MODFLOW in GSFLOW, coupling of GFLOW and MODFLOW in a hybrid code, and the forthcoming unstructured grid version of MODFLOW. They promise continued advances in the ability to use science to protect the single water resource.
Acque Sotterranee: Italian Journal of Groundwater↗
A hybrid finite-difference and analytic element groundwater model
Regional finite-difference models tend to have large cell sizes, often on the order of 1–2 km on a side. Although the regional flow patterns in deeper formations may be adequately represented by such a model, the intricate surface water and groundwater interactions in the shallower layers are not. Several stream reaches and nearby wells may occur in a single cell, precluding any meaningful modeling of the surface water and groundwater interactions between the individual features. We propose to replace the upper MODFLOW layer or layers, in which the surface water and groundwater interactions occur, by an analytic element model (GFLOW) that does not employ a model grid; instead, it represents wells and surface waters directly by the use of point-sinks and line-sinks. For many practical cases it suffices to provide GFLOW with the vertical leakage rates calculated in the original coarse MODFLOW model in order to obtain a good representation of surface water and groundwater interactions. However, when the combined transmissivities in the deeper (MODFLOW) layers dominate, the accuracy of the GFLOW solution diminishes. For those cases, an iterative coupling procedure, whereby the leakages between the GFLOW and MODFLOW model are updated, appreciably improves the overall solution, albeit at considerable computational cost. The coupled GFLOW–MODFLOW model is applicable to relatively large areas, in many cases to the entire model domain, thus forming an attractive alternative to local grid refinement or inset models.
Groundwater↗
Hypothetical modeling of redox conditions within a complex ground-water flow field in a glacial setting
This report describes a modeling approach for studying how redox conditions evolve under the influence of a complex ground-water flow field. The distribution of redox conditions within a flow system is of interest because of the intrinsic susceptibility of an aquifer to redox-sensitive, naturally occurring contaminants - such as arsenic - as well as anthropogenic contaminants - such as chlorinated solvents. The MODFLOW-MT3D-RT3D suite of code was applied to a glacial valley-fill aquifer to demonstrate a method for testing the interaction of flow patterns, sources of reactive organic carbon, and availability of electron acceptors in controlling redox conditions. Modeling results show how three hypothetical distributions of organic carbon influence the development of redox conditions in a water-supply aquifer. The distribution of strongly reduced water depends on the balance between the rate of redox reactions and the capability of different parts of the flow system to transmit oxygenated water. The method can take account of changes in the flow system induced by pumping that result in a new distribution of reduced water.
Scientific Investigations Report↗
Development and application of a screening model for simulating regional ground-water flow in the St. Croix River basin, Minnesota and Wisconsin
A series of databases and an accompanying screening model were constructed by the U.S. Geological Survey, in cooperation with the National Park Service, to better understand the regional ground-water-flow system and its relation to stream drainage in the St. Croix River Basin. The St. Croix River and its tributaries drain about 8,000 square miles in northeastern Minnesota and northwestern Wisconsin. The databases contain information for the entire St. Croix River Basin pertaining to well logs, lithology, thickness of lithologic groups, ground-water levels, streamflow, and well pumpage. Maps and generalized cross sections created from the compiled data show the lithologic groups, extending from the water table to the crystalline bedrock, through which ground water flows. These lithologic groups are: fine-grained unconsolidated deposits; coarse-grained unconsolidated deposits; sandstone bedrock; carbonate bedrock; and other bedrock lithologies including shale, siltstone, conglomerate, and igneous intrusions. The steady-state screening model treats the ground-water-flow system as a single layer with transmissivity zones that reflect the distribution of lithologic groups, and with recharge zones that correspond to general areas of high or low evapotranspiration. The model includes representation of second- and higher-order streams and municipal and other high-capacity production wells. The analytic-element model code GFLOW was used to simulate the regional ground-water flow, the water-table surface across the St. Croix River Basin, and base-flow contributions from ground water to streams. In addition, the model routes tributary base flow through the stream network to the St. Croix River. The parameter-estimation inverse model UCODE was linked to the GFLOW model to select the combination of parameter values best able to match over 5,000 water-level measurements and base-flow estimates at 22 streamflow-gaging stations. Results from the calibrated screening model show ground-water contributing areas for selected stream reaches within the basin. The delineation of these areas is useful to water-resource managers concerned with protection of fisheries and other resources. The model results also identify the areas of the basin where ground-water travel time from the water table to streams and wells is relatively short (less than 50 years). Ninety percent of the simulated ground-water pathlines require travel times between 3 and 260 years. The median pathline distance traversed and the median pathline velocity were 1.7 mi and 177 ft/y, respectively. It is important to recognize the limitations of this screening model. Heterogeneities in subsurface properties and in recharge rates are considered only at a very broad scale (miles to tens of miles). No account is taken of vertical variations in properties or pumping rates, and no provision is made to account for stacked ground-water-flow systems that have different flow patterns at different depths. Small-scale (hundreds to thousands of feet) flow systems associated with minor water bodies are neglected, and as a result, the model is not useful for simulating typical site-specific problems. Despite its limitations, the model serves as a framework for understanding the regional pattern of ground-water flow and as a starting point for a generation of more targeted and detailed ground-water models that would be needed to address emerging water-supply and water-quality concerns in the St. Croix River Basin.
Scientific Investigations Report↗
Ground-water modeling of pumping effects near regional ground-water divides and river/aquifer systems - Results and implications of numerical experiments
Agreements between United States governors and Canadian territorial premiers establish water-management principles and a framework for protecting Great Lakes waters, including ground water, from diversion and consumptive uses. The issue of ground-water diversions out of the Great Lakes Basin by large-scale pumping near the divides has been raised. Two scenario models, in which regional ground-water flow models represent major aquifers in the Great Lakes region, were used to assess the effect of pumping near ground-water divides. The regional carbonate aquifer model was a generalized model representing northwestern Ohio and northeastern Indiana; the regional sandstone aquifer model used an existing calibrated ground-water flow model for southeastern Wisconsin. Various well locations and pumping rates were examined. Although the two models have different frameworks and boundary conditions, results of the models were similar. There was significant diversion of ground water across ground-water divides due to pumping within 10 miles of the divides. In the regional carbonate aquifer model, the percentage of pumped water crossing the divide ranges from about 20 percent for a well 10 miles from the divide to about 50 percent for a well adjacent to the divide. In the regional sandstone aquifer model, the percentages range from about 30 percent for a well 10 miles from the divide to about 50 percent for a well adjacent to the divide; when pumping on the west side of the divide, within 5 mi of the predevelopment divide, results in at least 10 percent of the water being diverted from the east side of the divide. Two additional scenario models were done to examine the effects of pumping near rivers. Transient models were used to simulate a rapid stage rise in a river during pumping at a well in carbonate and glacial aquifers near the river. Results of water-budget analyses indicate that induced infiltration, captured streamflow, and underflow were important for both glacial and carbonate aquifers; however, in many cases, traveltimes from the river to the well will limit river water from physically entering the well.
Scientific Investigations Report↗
The value of long-term monitoring in the development of ground-water-flow models
As environmental issues have come to the forefront of public concern, so has the awareness of the importance of ground water in the overall water cycle and as a source of the Nation’s drinking water. Heightened interest has spawned a host of scientific enterprises (Taylor and Alley, 2001). Some activities are directed toward collection of water-level data and related information to monitor the physical and chemical state of the resource. Other activities are directed at interpretive studies undertaken, for example, to optimize the location of new water-supply wells or to protect rivers and lakes fed by ground water. An important type of interpretive study is the computer ground-water-flow model that inte- grates field data in a mathematical framework. Long-term, systematic collection of hydro- logic data is crucial to the construction and testing of ground-water models so that they can reproduce the evolution of flow systems and forecast future conditions.
Wisconsin↗
Simulation of stage and hydrologic budget for Shell Lake, Washburn County, Wisconsin
A model that simulates lake stage was developed to test the current understanding of the hydrology of Shell Lake, Wisconsin and to provide a tool for predicting the effects of withdrawing lake water on future lake stages. The model code is written in Fortran and simulates daily lake stage by summing estimates of hydrologic-budget components - precipitation falling on the lake surface, water evaporating from the lake surface, runoff (consisting of overland flow to the lake and intermittent streams flowing into the lake), and ground-water flow out of the lake. The model was calibrated to intermittent lake stage measurements for the period 1948-98. The hydrologic budget model was coupled to UCODE, a parameter estimation model, to aid in estimating runoff coefficients. Trends in stage simulated by the calibrated model compare reasonably well with historical stage trends. The root mean square of the differences of simulated and measured daily lake stage for the period 1948-98 is 0.54 foot. Predictive simulations indicate that withdrawing lake water is an effective way of reducing lake stage. Several years of pumping for at least 200 days per year at rates of 1,000 to 2,000 gallons per minute would have been required to reduce 1990's high stages by about one foot.
Wisconsin↗