Geology topics

Daniel T. Feinstein

Publications and source records attributed to Daniel T. Feinstein.

At least 19 recordsLinked to original sources

USGS2025

A regional model comparison between MODPATH and MT3D of groundwater travel time distributions

Groundwater quality changes in wells and streams lag behind changes to land use due to groundwater travel times. Two contaminant transport methods were compared to assess differences in their simulated travel time distributions (TTDs) to streams and wells in the Wisconsin Central Sands. MODPATH simulates advective groundwater flow with particle tracking, while MT3D simulates age-mass using a finite difference solution without dispersion to allow for direct comparison of the two methods. MODPATH appropriately simulates groundwater TTDs from the water table to surface discharge but is subject to inaccuracies at weak-sink well cells due to the flow-model grid discretization and imprecise location of well discharge within well cells. MT3D better represents weak-sink well cells since it removes mass in proportion to the prescribed pumping rate, although travel time within well cells is neglected. Conversely, MT3D's treatment of surface water boundary cells is not as accurate as MODPATH because mass should be removed from the water table rather than the full cell volume. MT3D simulations of TTDs can also be confounded by the instantaneous vertical distribution of mass introduced throughout recharge cells instead of at the water table, which initiates mass along deeper flow paths. We evaluated 9 MODPATH and 13 MT3D implementations, generating differences in median travel times of up to 18 years. Both methods have strengths and weaknesses, with MT3D better representing weak-sink well cell behavior and MODPATH better representing surficial recharge and discharge. The effect of these characteristics on simulated TTDs, along with ideas for ameliorating method weaknesses, is discussed.

Wisconsin↗

USGS2023

Groundwater residence times in glacial aquifers—A new general simulation-model approach compared to conventional inset models

Groundwater is important as a drinking-water source and for maintaining base flow in rivers, streams, and lakes. Groundwater quality can be predicted, in part, by its residence time in the subsurface, but the residence-time distribution cannot be measured directly and must be inferred from models. This report compares residence-time distributions from four areas where groundwater flow and travel time were simulated with conventional simulation-inset models (IMs) and with a new automated model-construction method called general simulation models (GSMs). The comparison provides an opportunity to explore controls on travel time and improve the methods used in the creation of GSMs. These models can be useful for three main-use cases: (1) rapid testing of relationships that govern groundwater flow and age, (2) generation of consistent examples for training a machine-learning metamodel, and (3) serving as a starting point for more detailed models. Comparison of the GSMs to IMs indicated a qualified pattern of agreement for residence-time distributions as indicated by the Nash-Sutcliffe efficiency and Spearman’s correlation coefficient. The agreement was best for the median values of the simulated residence times in young fractions of groundwater (defined as the fractions of groundwater in samples less than 65 years old) at the scale of the eight-digit hydrologic-unit code. Generally, the median values of the young fractions in the IMs were correlated with the median values from the GSMs. The relative trends across the four areas also were similar for the other residence-time metrics. The medians of residence-time metrics at finer scales show a fair degree of scatter. The GSM results compared most poorly for median travel times in the older fraction of groundwater (older than 65 years). The GSM approach is intended as a flexible framework for developing models that can be useful individually as screening tools or collectively to support projects in statistical learning. Although one set of GSM algorithms was presented here, the approach can accommodate many types of data and also different categories of prior information. Comparison of GSMs and IMs suggests ways in which the GSMs, while remaining easy to construct and calibrate, can be improved for estimating groundwater travel times. IMs do not yield exact travel times, and matching GSMs to IMs does not guarantee an improvement; however, IMs provide a convenient benchmark against which to explore relations between physical characteristics of watersheds and the distribution of travel times within them. This effort was undertaken as part of the National Water Quality Program of the U.S. Geological Survey to assist in determining the susceptibility of groundwater in glacial aquifers to a variety of natural and anthropogenic contaminants.

Illinois, Indiana, Michigan, Wisconsin↗

USGS2023

Simulation of groundwater flow at the former Badger Army Ammunition Plant, Sauk County, Wisconsin

To help support remedial efforts at the former Badger Army Ammunition Plant the U.S. Geological Survey built and calibrated a transient groundwater flow model using the Newton Raphson formulation (MODFLOW–NWT) of the U.S. Geological Survey’s modular three-dimensional finite-difference code. The model simulates the groundwater flow system at the site from 1984 to 2020. The former Badger Army Ammunition Plant is a 7,275-acre site in Sauk County, Wisconsin. The plant produced smokeless gunpower and solid rocket propellent as munitions components. Peak production periods were during World War II, the Korean War, and the Vietnam War. Subsequent groundwater contamination investigations have found four plumes at the site. A health risk assessment identified at least one contaminant of concern for human health risk present in three of the plumes: the propellant burning ground plume, the deterrent burning ground plume, and the central plume. A cooperative study began between the U.S. Army Environmental Command and U.S. Geological Survey to better understand the groundwater flow system at the former Badger Army Ammunition Plant. Field data, including aquifer tests, streamflow measurements, continuous groundwater elevations, and groundwater gradients with the Wisconsin River were collected and used to inform and calibrate the groundwater flow model. The model was used to assess the variability of the groundwater system over the study period, the components of the groundwater budget, and groundwater flow directions from identified source areas towards the Wisconsin River. Model performance assessment focused on using particle tracking to compare groundwater flowpaths that originate in the contaminant source areas to the observed plume footprints. This focus on plume behavior geometry should help constrain the advective component of a future groundwater transport model of the site.

Wisconsin↗

USGS2023

Hydrologic change in the St. Louis River Basin from iron mining on the Mesabi Iron Range, northeastern Minnesota

This study compares the results of two regional steady-state U.S. Geological Survey Modular Three-Dimensional Finite-Difference Ground-Water Flow (MODFLOW) models constructed to quantify the hydrologic changes in the St. Louis River Basin from iron mining on the Mesabi Iron Range in northeastern Minnesota. The U.S. Geological Survey collaborated in this study with bands of the Minnesota Chippewa Tribe, and the Minnesota Pollution Control Agency to inform management decisions about aquatic resources in the St. Louis River Basin. A model constructed and calibrated to represent average 1995–2015 mining conditions produced regional groundwater heads and flows. A pre-mining scenario model was constructed from this mining model but had the land and bedrock surfaces restored to pre-mining topographies and had modeled mining features (mine pits, tailings basins, waste-rock piles, and mining-disturbed areas) eliminated to represent general pre-mining stratigraphy and hydrogeology. Many of the features important to the hydrology of this mining area (like individual mine pits) are difficult to represent in groundwater models and required the use of modeling tools to indirectly account for their effects. The difference between the results of these two models represents mining’s effects on the hydrology in the Mesabi Iron Range area of the St Louis River Basin. The mining and pre-mining regional models also can provide boundary conditions and initial properties for future local or site-specific groundwater-flow models in the area. Total groundwater flow through the mining model is 171 million cubic feet per day. Areal recharge is the largest source of groundwater (78 and 81 percent of total groundwater flow in the mining and pre-mining scenario models, respectively). Seepage from streams and lakes provides another 17 percent of the total groundwater flow through both models. Water leaves aquifers through seepage to streams (discharge as base flow, 43 percent in both models) and areal seepage to the land surface (surface seepage), for example to wetlands (45 and 49 percent, mining and pre-mining scenario models respectively). Comparison of the results from the mining and pre-mining scenario models shows that iron mining has produced measurable hydrologic changes in the St. Louis River Basin, but that most of those changes and the highest magnitude changes occur near the mining features. Flow changes to and from surface-water bodies like streams and wetlands were analyzed in detail because of their importance in sustaining surface waters and aquatic life. Overall, groundwater flow in the mining model was 3.62 million cubic feet per day (2.2 percent) greater than total pre-mining model groundwater flow. This was caused by an increase in recharge from tailings basins and a decrease in discharge from surface seepage. Groundwater discharge to mine pits was the largest change in groundwater flows between the models (a change representing 2.8 percent of total pre-mining model groundwater flow). Net recharge to groundwater from tailings basins (2.4 percent), net decrease in surface seepage from groundwater (2.7 percent), and net increase in seepage to streams (1.0 percent) were all in this same range of total pre-mining model groundwater flow. Groundwater lost through mine-pit withdrawals was nearly offset by groundwater gained through recharge from tailings basins. However, because losses and gains occurred in different areas, the effect of mining can have more substantial effects on local areas than the model-wide averages represent.

Minnesota↗

USGS2023

Using airborne electromagnetics to improve depth to bedrock estimates in Wisconsin

Depth to bedrock is an important factor in hydrologic systems. The thickness of glacial sediments overlying bedrock in Wisconsin controls the routing of groundwater in surficial aquifers and its Depth to bedrock is often an important factor in hydrologic systems because hydraulic properties of bedrock and overlying sediments are typically appreciably different. For example, the thickness of glacial sediments overlying bedrock in Wisconsin controls the routing of groundwater in surficial aquifers and its connection with surface water bodies such as lakes and wetlands. In fractured bedrock environments, shallow bedrock can be vulnerable to degraded water quality when contaminants at the surface infiltrate quickly through permeable formations. Here, airborne electromagnetic surveys were acquired in three different parts of Wisconsin, totalling more than 5,700 flight line-kilometers, to improve understanding of depth to bedrock, the lithologic composition of overlying sediments, and as input structure for groundwater model development.

Wisconsin↗

USGS2022

New capabilities in MT3D-USGS for simulating unsaturated-zone heat transport

Changes in climate and land use will alter groundwater heat transport dynamics in the future. These changes will in turn affect watershed processes (e.g., nutrient cycling) as well as watershed characteristics (e.g., distribution and persistence of cold-water habitat). Thus, groundwater flow and heat transport models at watershed scales that can characterize and quantify thermal impacts of surface temperature change on groundwater system temperatures are needed to forecast changes to groundwater-linked ecosystems in riparian zones, streams, and lakes. Including unsaturated zone processes has previously been shown to be important for properly determining the timing and magnitude of groundwater recharge (Hunt et al. 2008). Similarly, heat transport dynamics in the saturated-zone, as well as connected surface-water systems, can be appreciably influenced by unsaturated-zone processes; in this way the unsaturated zone forms an inextricable link between land surface where change occurs and the groundwater system that transmit that change. This paper presents new capabilities for the existing MT3D-USGS transport simulator by adding functionality for simulating heat transport through the unsaturated zone. New simulation capabilities are verified through comparison of simulation results with those of the variably-saturated heat transport simulator VS2DH under steady and transient conditions for both water and heat flow. The new capabilities are assessed using a number of conceptualizations and include evaluations of convective and conductive heat flow. These additional capabilities increase the utility for applied watershed-scale simulations, which in turn should facilitate more realistic characterizations of temperature change on thermally sensitive ecosystems, such as stream habitat.

Groundwater↗

USGS2022

Simulation of heat flow in a synthetic watershed: Lags and dampening across multiple pathways under a climate-forcing scenario

Although there is widespread agreement that future climates tend toward warming, the response of aquatic ecosystems to that warming is not well understood. This work, a continuation of companion research, explores the role of distinct watershed pathways in lagging and dampening climate-change signals. It subjects a synthetic flow and transport model to a 30-year warming signal based on climate projections, quantifying the heat breakthrough on a monthly time step along connected pathways. The system corresponds to a temperate watershed roughly 27 km on a side and consists of (a) land-surface processes of overland flow, (b) infiltration through an unsaturated zone (UZ) above an unconfined sandy aquifer overlying impermeable bedrock, and (c) groundwater flow along shallow and deep pathlines that converge as discharge to a surface-water network. Numerical simulations show that about 40% of the warming applied to watershed infiltration arrives at the water table and that the UZ stores a large fraction of the upward-trending heat signal. Additionally, once groundwater reaches the surface-water network after traveling through the saturated zone, only about 10% of the original warm-up signal is returned to streams by discharge. However, increases in the simulated streamflow temperatures are of similar magnitude to increases at the water table, due to the addition of heat by storm runoff, which bypasses UZ and groundwater storage and counteracts subsurface dampening. The synthetic modeling method and tentative findings reported here provide a potential workflow for real-world applications of climate-change modeling at the full watershed scale.

Water↗

USGS2021

Groundwater/surface-water interactions in the Partridge River Basin and evaluation of hypothetical future mine pits, Minnesota

The Partridge River Basin (PRB) covers 156 square miles in northeastern Minnesota with headwaters in the Mesabi Iron Range. The basin is characterized by extensive wetlands, lakes, and streams in poorly drained and often thin glacial material overlying Proterozoic bedrock. To better understand the interaction between these extensive surface water features and the groundwater system, a three-dimensional, steady-state, groundwater-flow model of the PRB was developed by the U.S. Geological Survey in cooperation with the Great Lakes Indian Fish & Wildlife Commission using the finite-difference computer code MODFLOW-NWT. The model simulates steady-state base flow in streams and groundwater interactions using the streamflow routing (SFR2) package. Existing mining features including tailings basins, stockpiles, pumped mine pits, and flooded mine pits were simulated using either high hydraulic conductivity zones or the drain (DRN) package. The unsaturated zone flow (UZF) package was used to better represent the groundwater system in areas with a high water table and for wetlands often associated with such areas. UZF typically is used to represent unsaturated zone processes but also can simulate the rejection of recharge and groundwater discharge to the land surface when the water table is near land surface. The steady-state model used data from the 2011 to 2013 period when 2011 high-resolution land surface (light detecting and ranging [lidar]) data were available that reflected land-surface and water elevations from mining activity in the basin. The parameter-estimation software suite PEST_HP was used to obtain a best fit of the modeled to measured groundwater levels, streamflow, pit inflow rates, and mapped peat deposits. The PEST calibration used the target residuals from two models with the same model parameters and targets from two separate periods: (1) a 1995–2015 calibration model, which provided a larger number of calibration targets, and (2) a 2011–2013 mining conditions model, which included calibration targets that reflected conditions consistent with the modeled mine-workings topography. Calibration of the PRB model resulted in ranges of glacial horizontal hydraulic conductivity parameters that generally agreed with literature values and other models of the region. Horizontal hydraulic conductivity of the bedrock was higher in the upper bedrock layers where numerous and continuous fractures have been observed and lower in the deeper bedrock layers. Average basin-wide calibrated infiltration was 5.3 inches per year. An average of 4.6 inches per year of infiltration crosses the water table and becomes recharge and 0.7 inch per year is rejected by UZF due to saturated conditions at the land surface. Simulated groundwater runoff (the sum of rejected recharge and groundwater seepage to the land surface) can either be routed to streams or removed from the model as evapotranspiration. The calibrated model indicates relatively shallow groundwater-flow paths dominating and approximately 50 percent of the stream base flow coming from groundwater runoff. The 2011–2013 mining conditions model was then used to develop five model scenarios simulating the response of the groundwater and surface-water system to potential hydrologic stress. The purpose of these mine pit scenarios is to present a possible workflow to quantify a model’s uncertainty for a given model forecast and serve as a possible guide for initial data collection that may improve a future model’s ability to make such a forecast. The scenarios included one scenario with the currently existing Peter Mitchell pit at final buildout and flooded to an elevation of 1,500 feet, and four scenarios with a hypothetical, new mine pit plus the flooded Peter Mitchell at final buildout. The five model scenarios were used to forecast streamflow at six locations in the PRB, pit inflow rates for the new mine pits and the flooded Peter Mitchell pit, and the average depth to water in 12 wetlands. A linear uncertainty analysis was performed using information from the PEST calibration and tools in the PyEMU python package to assess model uncertainty propagation to the model forecasts. Streamflows generally were reduced with future mining and the greatest streamflow reductions occurred from the flooded Peter Mitchell Pit, probably due to its large size. Average depth to groundwater in wetlands was most affected the closer the wetland was to a new mine pit. Linear uncertainty methods were also used to evaluate data worth, which is the ability for potential new groundwater elevation observations to reduce the uncertainty in scenario forecasts. Data worth was performed for a grid of new hydraulic head observations. Overall, areas with nonnegligible data worth generally corresponded to wetland areas with no groundwater seepage to land surface from UZF. These model behaviors indicated that the land-surface boundary condition simulated by the UZF package was pinning the groundwater elevations to the land surface in areas with groundwater seepage (33 percent of the 2011–2013 base conditions model) such that the sensitivity to new observations in these areas was minimal. Therefore, representing wetlands as boundary conditions minimized the usefulness of data worth calculations because wetland areas were present over a large part of the model domain. Probabilistic capture zones were estimated for each of the mines in the model scenarios. A capture zone represents the area contributing recharge to a model feature, like a well or a mine pit, and can be calculated by forward tracking particles from the water table. By using Monte Carlo techniques, it is possible to generate estimated capture zones that include the probability of recharge capture given the uncertainty present in the model. Monte Carlo techniques use randomly generated model parameter sets sampled from a plausible parameter range to create many possible realizations. The resulting capture zone arrays were calculated by tallying the total number of realizations in which a particle from a model cell was captured by the feature. Probabilities from the Monte Carlo runs ranged from 1 (captured in 100 percent of the runs) near the pits to 0 (captured in 0 percent of the runs) at the edges of the capture zone. Capture zones were not always spatially continuous; for example, the capture zone for the proposed mine pits south of the flooded Peter Mitchell pit was discontinuous with capture surrounding the proposed mine pit and north of the flooded Peter Mitchell pit. This northern section represents deeper groundwater flow paths that originate in the topographic high, move under the flooded pit, and discharge into the proposed pit. This pattern of capture indicates the possibility of some deeper flow through the upper fractured bedrock when the shallow groundwater flow system is modified. These results underscore that future site-specific applications of the base condition model require the input of site-specific data and recalibration to focus on the site of interest.

Minnesota↗

USGS2020

A non-intrusive approach for efficient stochastic emulation and optimization of model-based nitrate-loading management decision support

Use of physically-motivated numerical models like groundwater flow-and-transport models for probabilistic impact assessments and optimization under uncertainty (OUU) typically incurs such a computational burdensome that these tools cannot be used during decision making. The computational challenges associated with these models can be addressed through emulation. In the land-use/water-quality context, the linear relation between nitrate loading and surface-water/groundwater nitrate concentrations presents an opportunity for employing an efficient model emulator through the application of impulse-response matrices. When paired with first-order second-moment techniques, the emulation strategy gives rise to the “stochastic impulse-response emulator” (SIRE). SIRE is shown to facilitate non-intrusive, near-real time, and risk-based evaluation of nitrate-loading change scenarios, as well as nitrate-loading OUU subject to surface-water/groundwater concentration constraints in high decision variable and parameter dimensions. Two case studies are used to demonstrate SIRE in the nitrate-loading context.

Environmental Modelling and Software↗

USGS2019

A simple method for simulating groundwater interactions with fens to forecast development effects

Protection of fens–wetlands dependent on groundwater discharge–requires characterization of groundwater sources and stresses. Because instrumentation and numerical modeling of fens is labor intensive, easy-to-apply methods that model fen distribution and their vulnerability to development are desirable. Here we demonstrate that fen areas can be simulated using existing steady-state MODFLOW models when the unsaturated zone flow (UZF) package is included. In cells where the water table is near land surface, the UZF package calculates a head difference and scaled conductance at these “seepage drain” cells to generate average rates of vertical seepage to the land. This formulation, which represents an alternative to blanketing the MODFLOW domain with drains, requires very little input from the user because unsaturated flow-routing is inactive and results are primarily driven by easily obtained topographic information. Like the drain approach, it has the advantage that the distribution of seepage areas is not predetermined by the modeler, but rather emerges from simulated heads. Beyond the drain approach, it takes account of intracell land surface variation to explicitly quantify multiple surficial flows corresponding to infiltration, rejected recharge, recharge and land-surface seepage. Application of the method to a basin in southeastern Wisconsin demonstrates how it can be used as a decision-support tool to first, reproduce fen distribution and, second, forecast drawdown and reduced seepage at fens in response to shallow pumping.

Wisconsin↗

USGS2018

Metamodeling for groundwater age forecasting in the Lake Michigan Basin

Groundwater age is an important indicator of groundwater susceptibility to anthropogenic contamination and a key input to statistical models for forecasting water quality. Numerical models can provide estimates of groundwater age, enabling interpretation of measured age tracers. However, to extend to national‐scale groundwater systems where numerical models are not routinely available, a more efficient metamodeling approach can provide a less precise but widely applicable estimate of groundwater age, trained to make forecasts based on predictor variables that can be measured independent of numerical models. We trained gradient‐boosted regression tree statistical metamodels to MODFLOW/MODPATH‐derived groundwater age estimates in five inset models in the Lake Michigan Basin, USA. Using high‐throughput computing, we explored an exhaustive range of tuning parameters and tested metamodels through cross validation, a 20% holdout, and a round robin approach among the five inset models withholding each inset model from training and testing on the held‐out inset model. Forecast skill—measured by Nash Sutcliffe efficiency—was high for age‐related responses in the 20% hold‐out case (ranging from 0.73 to 0.84). The round robin analysis provided the opportunity to explore extending to unmodeled areas and a greater range of skill indicated the need to evaluate when it is appropriate to apply a metamodel from one region to another. We further explored the ramifications of metamodel simplification achieved through removing predictor variables based on their estimated importance. We found that similar metamodel performance was achievable with a fraction of the candidate set of predictor variables with well construction variables being most important.

Water Resources Research↗

USGS2018

Extraction and development of inset models in support of groundwater age calculations for glacial aquifers

The U.S. Geological Survey developed a regional model of Lake Michigan Basin (LMB). This report describes the construction of five MODFLOW inset models extracted from the LMB regional model and their application using the particle-tracking code MODPATH to simulate the groundwater age distribution of discharge to wells pumping from glacial deposits. The five study areas of the inset model correspond to 8-digit hydrologic unit code (HUC8) basins. Two of the basins are tributary to Lake Michigan from the east, two are tributary to the lake from the west, and one is just west of the western boundary of the Lake Michigan topographic basin. The inset models inherited many of the inputs to the parent LMB model, including the hydrostratigraphy and layering scheme, the hydraulic conductivity assigned to bedrock layers, recharge distribution, and water use in the form of pumping rates from glacial and bedrock wells. The construction of the inset models entailed modifying some inputs, most notably the grid spacing (reduced from cells 5,000 feet on a side in the parent LMB model to 500 feet on a side in the inset models). The refined grid spacing allowed for more precise location of pumped wells and more detailed simulation of groundwater/surface-water interactions. The glacial hydraulic conductivity values, the top bedrock surface elevation, and the surface-water network input to the inset models also were modified. The inset models are solved using the MODFLOW–NWT code, which allows for more robust handling of conditions in unconfined aquifers than previous versions of MODFLOW. Comparison of the MODFLOW inset models reveals that they incorporate a range of hydrogeologic conditions relative to the glacial part of the flow system, demonstrated by visualization and analysis of model inputs and outputs and reflected in the range of ages generated by MODPATH for existing and hypothetical glacial wells. Certain inputs and outputs are judged to be candidate predictors that, if treated statistically, may be capable of explaining much of the variance in the simulated age metrics. One example of a predictor that model results indicate strongly affects simulated age is the depth of the well open interval below the simulated water table. The strength of this example variable as an overall predictor of groundwater age and its relation to other predictors can be statistically tested through the metamodeling process. In this way the inset models are designed to serve as a training area for metamodels that estimate groundwater age in glacial wells, which in turn will contribute to ongoing studies, under the direction of the U.S. Geological Survey National Water Quality Assessment, of contaminant susceptibility of shallow groundwater across the glacial aquifer system.

Illinois, Indiana, Michigan, Wisconsin↗

USGS2017

Generalized hydrogeologic framework and groundwater budget for a groundwater availability study for the glacial aquifer system of the United States

The glacial aquifer system groundwater availability study seeks to quantify (1) the status of groundwater resources in the glacial aquifer system, (2) how these resources have changed over time, and (3) likely system response to future changes in anthropogenic and environmental conditions. The glacial aquifer system extends from Maine to Alaska, although the focus of this report is the part of the system in the conterminous United States east of the Rocky Mountains. The glacial sand and gravel principal aquifer is the largest source of public and self-supplied industrial supply for any principal aquifer and also is an important source for irrigation supply. Despite its importance for water supply, water levels in the glacial aquifer system are generally stable varying with climate and only locally from pumping. The hydrogeologic framework developed for this study includes the information from waterwell records and classification of material types from surficial geologic maps into likely aquifers dominated by sand and gravel deposits. Generalized groundwater budgets across the study area highlight the variation in recharge and discharge primarily driven by climate.

Glacial Aquifer System↗

USGS2017

Estimation of the groundwater resources of the bedrock aquifers at the Kettle Moraine Springs State Fish Hatchery, Sheboygan County, Wisconsin

Groundwater resources information was needed to understand regional aquifer systems and water available to wells and springs for rearing important Lake Michigan fish species at the Kettle Moraine Springs State Fish Hatchery in Sheboygan County, Wisconsin. As a basis for estimating the groundwater resources available, an existing groundwater-flow model was refined, and new groundwater-flow models were developed for the Kettle Moraine Springs State Fish Hatchery area using the U.S. Geological Survey (USGS) finite-difference code MODFLOW. This report describes the origin and construction of these groundwater-flow models and their use in testing conceptual models and simulating the hydrogeologic system. The study area is in the Eastern Ridges and Lowlands geographical province of Wisconsin, and the hatchery property is situated on the southeastern edge of the Kettle Moraine, a north-south trending topographic high of glacial origin. The bedrock units underlying the study area consist of Cambrian, Ordovician, and Silurian units of carbonate and siliciclastic lithology. In the Sheboygan County area, the sedimentary bedrock sequence reaches a thickness of as much as about 1,600 feet (ft). Two aquifer systems are present at the Kettle Moraine Springs State Fish Hatchery. A shallow system is made up of Silurian bedrock, consisting chiefly of dolomite, overlain by unconsolidated Quaternary-age glacial deposits. The glacial deposits of this aquifer system are the typical source of water to local springs, including the springs that have historically supplied the hatchery. The shallow aquifer system, therefore, consists of the unconsolidated glacial aquifer and the underlying bedrock Silurian aquifer. Most residential wells in the area draw from the Silurian aquifer. A deeper confined aquifer system is made up of Cambrian- and Ordovician-age bedrock units including sandstone formations. Because of its depth, very few wells are completed in the Cambrian-Ordovician aquifer system (COAS) near the Kettle Moraine Springs State Fish Hatchery. Three groundwater-flow models were used to estimate the water resources available to the hatchery from bedrock aquifers under selected scenarios of well placement and seasonal water requirements and subject to constraints on the effects of pumping on neighboring wells, local springs, and creeks. Model input data (recharge, water withdrawal, and boundary conditions) for these models were compiled from a number of data and information sources. The first model, named the “KMS model,” (KMS stands for Kettle Moraine Springs) is an inset model derived from a published USGS regional Lake Michigan Basin model and was constructed to simulate groundwater pumping from the semiconfined Silurian aquifer. The second model, named the “Pumping Test model,” was constructed to evaluate an aquifer pumping test conducted in the COAS as part of this project. The Pumping Test model was also used to simulate the local effects of 20 years of groundwater pumping from this deep bedrock aquifer for future hatchery operations. The third model, named the “LMB modified model,” is a version of the published Lake Michigan Basin (LMB) model that was modified with aquifer parameters refined in an area around the hatchery (approximately a 5-mile radius circle, corresponding to the area stressed by the aquifer pumping test). This LMB modified model was applied to evaluate regional effects of pumping from the confined COAS. The available Silurian aquifer groundwater resource was estimated using the KMS model with three scenarios—named “AllConstraints,” “Constraints2,” and “Constraints3”—that specified local water-level and flow constraints such as drawdown at nearby household wells, water levels inside pumping well boreholes, and flow in local streams and springs. Each scenario utilized the MODFLOW Groundwater Management Process (GWM) to select three locations from six candidate locations that provided the greatest combined flow while satisfying the constraints. The three constraint scenarios provided estimates of 430 gallons per minute (gal/min), 480 gal/min, and 520 gal/min pumping from three wells—AllConstraints, Constraints2, and Constraints3, respectively. The same three wells were selected for the scenarios that estimated 480 gal/min and 520 gal/min; the scenario that estimated 430 gal/min shared two of these same wells, but the third selected well was different. The available COAS groundwater resource was estimated by two scenarios with each conducted over a period of 20 years with the Pumping Test model and the LMB modified model . The Pumping Test model was used to simulate local effects of pumping, and the LMB modified model was used to simulate regional effects of pumping. The scenarios simulate a range of total and seasonal pumping rates potentially linked to site activities. Scenario 1 simulates two wells completed in the Cambrian-Ordovician aquifer system, each pumping for 8 months at 300 gal/min, followed by pumping for 4 months at 600 gal/min. The average yearly pumping rate of Scenario 1 is 800 gal/min. Scenario 2 simulates three wells completed in the Cambrian-Ordovician aquifer system pumping for 8 months at 200 gal/min, followed by pumping for 4 months at 500 gal/min. The average yearly pumping rate of Scenario 2 is 900 gal/min. The Pumping Test model simulations confirmed that drawdown in the boreholes of the pumping wells at the selected 2-well or 3-well rates will meet the desired condition that the pumping water level remains at least 100 ft above the highest Cambrian-Ordovician unit open to the well. The LMB modified model was used to evaluate the regional drawdown of the pumping from the confined COAS under the same 2-well and 3-well scenarios. At the nearest known existing COAS well, Campbellsport production well #4, the simulated drawdown for Scenario 1 after 20 years of cyclical pumping with two pumping wells averaging a total of 800 gal/min is 16.9 ft, whereas the simulated drawdown for Scenario 2 after 20 years of pumping with three pumping wells averaging a total of 900 gal/min is 19.0 ft. The total deep aquifer thickness at the Campbellsport location is on the order of 620 ft, meaning that the simulated drawdown for either scenario is about 3 percent of the confined aquifer thickness. The models developed as part of this project are archived in the project data release. The archive includes the model input and output files as well as MODFLOW source code and executables. ( Haserodt and others, 2017 ).

Wisconsin↗

USGS2017

Simulation of groundwater flow in the glacial aquifer system of northeastern Wisconsin with variable model complexity

The U.S. Geological Survey, National Water-Quality Assessment seeks to map estimated intrinsic susceptibility of the glacial aquifer system of the conterminous United States. Improved understanding of the hydrogeologic characteristics that explain spatial patterns of intrinsic susceptibility, commonly inferred from estimates of groundwater age distributions, is sought so that methods used for the estimation process are properly equipped. An important step beyond identifying relevant hydrogeologic datasets, such as glacial geology maps, is to evaluate how incorporation of these resources into process-based models using differing levels of detail could affect resulting simulations of groundwater age distributions and, thus, estimates of intrinsic susceptibility. This report describes the construction and calibration of three groundwater-flow models of northeastern Wisconsin that were developed with differing levels of complexity to provide a framework for subsequent evaluations of the effects of process-based model complexity on estimations of groundwater age distributions for withdrawal wells and streams. Preliminary assessments, which focused on the effects of model complexity on simulated water levels and base flows in the glacial aquifer system, illustrate that simulation of vertical gradients using multiple model layers improves simulated heads more in low-permeability units than in high-permeability units. Moreover, simulation of heterogeneous hydraulic conductivity fields in coarse-grained and some fine-grained glacial materials produced a larger improvement in simulated water levels in the glacial aquifer system compared with simulation of uniform hydraulic conductivity within zones. The relation between base flows and model complexity was less clear; however, the relation generally seemed to follow a similar pattern as water levels. Although increased model complexity resulted in improved calibrations, future application of the models using simulated particle tracking is anticipated to evaluate if these model design considerations are similarly important for understanding the primary modeling objective - to simulate reasonable groundwater age distributions.

Wisconsin↗

USGS2016

A semi-structured MODFLOW-USG model to evaluate local water sources to wells for decision support

In order to better represent the configuration of the stream network and simulate local groundwater-surface water interactions, a version of MODFLOW with refined spacing in the topmost layer was applied to a Lake Michigan Basin (LMB) regional groundwater-flow model developed by the U.S. Geological. Regional MODFLOW models commonly use coarse grids over large areas; this coarse spacing precludes model application to local management issues (e.g., surface-water depletion by wells) without recourse to labor-intensive inset models. Implementation of an unstructured formulation within the MODFLOW framework (MODFLOW-USG) allows application of regional models to address local problems. A “semi-structured” approach (uniform lateral spacing within layers, different lateral spacing among layers) was tested using the LMB regional model. The parent 20-layer model with uniform 5000-foot (1524-m) lateral spacing was converted to 4 layers with 500-foot (152-m) spacing in the top glacial (Quaternary) layer, where surface water features are located, overlying coarser resolution layers representing deeper deposits. This semi-structured version of the LMB model reproduces regional flow conditions, whereas the finer resolution in the top layer improves the accuracy of the simulated response of surface water to shallow wells. One application of the semi-structured LMB model is to provide statistical measures of the correlation between modeled inputs and the simulated amount of water that wells derive from local surface water. The relations identified in this paper serve as the basis for metamodels to predict (with uncertainty) surface-water depletion in response to shallow pumping within and potentially beyond the modeled area, see Fienen et al. (2015a).

Illinois, Indiana, Michigan, Wisconsin↗

USGS2016

Evaluating the sources of water to wells: Three techniques for metamodeling of a groundwater flow model

For decision support, the insights and predictive power of numerical process models can be hampered by insufficient expertise and computational resources required to evaluate system response to new stresses. An alternative is to emulate the process model with a statistical “metamodel.” Built on a dataset of collocated numerical model input and output, a groundwater flow model was emulated using a Bayesian Network, an Artificial neural network, and a Gradient Boosted Regression Tree. The response of interest was surface water depletion expressed as the source of water-to-wells. The results have application for managing allocation of groundwater. Each technique was tuned using cross validation and further evaluated using a held-out dataset. A numerical MODFLOW-USG model of the Lake Michigan Basin, USA, was used for the evaluation. The performance and interpretability of each technique was compared pointing to advantages of each technique. The metamodel can extend to unmodeled areas.

Illinois, Indiana, Michigan, Ohio, Wisconsin↗

USGS2016

The 2016 groundwater flow model for Dane County, Wisconsin

A new groundwater flow model for Dane County, Wisconsin, replaces an earlier model developed in the 1990s by the Wisconsin Geological and Natural History Survey (WGNHS) and the U.S. Geological Survey (USGS). This modeling study was conducted cooperatively by the WGNHS and the USGS with funding from the Capital Area Regional Planning Commission (CARPC). Although the overall conceptual model of the groundwater system remains largely unchanged, the incorporation of newly acquired high-quality datasets, recent research findings, and improved modeling and calibration techniques have led to the development of a more detailed and sophisticated model representation of the groundwater system. The new model is three-dimensional and transient, and conceptualizes the county’s hydrogeology as a 12-layer system including all major unlithified and bedrock hydrostratigraphic units and two high-conductivity horizontal fracture zones. Beginning from the surface down, the model represents the unlithified deposits as two distinct model layers (1 and 2). A single layer (3) simulates the Ordovician sandstone and dolomite of the Sinnipee, Ancell, and Prairie du Chien Groups. Sandstone of the Jordan Formation (layer 4) and silty dolostone of the St. Lawrence Formation (layer 5) each comprise separate model layers. The underlying glauconitic sandstone of the Tunnel City Group makes up three distinct layers: an upper aquifer (layer 6), a fracture feature (layer 7), and a lower aquifer (layer 8). The fracture layer represents a network of horizontal bedding-plane fractures that serve as a preferential pathway for groundwater flow. The model simulates the sandstone of the Wonewoc Formation as an upper aquifer (layer 9) with a bedding-plane fracture feature (layer 10) at its base. The Eau Claire aquitard (layer 11) includes shale beds within the upper portion of the Eau Claire Formation. This layer, along with overlying bedrock units, is mostly absent in the preglacially eroded valleys along the Yahara River valley and in northeastern Dane County. Layer 12 represents the Mount Simon sandstone as the lowermost model layer. It directly overlies the Precambrian crystalline basement rock, whose top surface forms the lower boundary of the model. The model uses the USGS MODFLOW-NWT finite-difference code, a standalone version of MODFLOW-2005 that incorporates the Newton (NWT) solver. MODFLOW-NWT improves the handling of unconfined conditions by smoothing the transition from wet to dry cells. The model explicitly simulates groundwater–surface-water interaction with streamflow routing and lake-level fluctuation. Model input included published and unpublished hydrogeologic data from recent estimates of aquifer hydraulic conductivities. A spatial groundwater recharge distribution was obtained from a recent GIS-based, soil-water-balance model for Dane County. Groundwater withdrawals from pumping were simulated for 572 wells across the entire model domain, which includes Dane County and portions of seven neighboring counties—Columbia, Dodge, Green, Iowa, Jefferson, Lafayette, and Rock. These wells withdrew an average of 60 million gallons per day (mgd) over the 5-year period from 2006 through 2010. Within Dane County, 385 wells were simulated with an average withdrawal rate of 52 mgd. Model calibration used the parameter estimation code PEST, and calibration targets included heads, stream and spring flows, lake levels, and borehole flows. Steady-state calibration focused on the period 2006 through 2010; the transient calibration focused on the 7-week drought period from late May through July 2012. This model represents a significant step forward from previous work because of its finer grid resolution, improved hydrostratigraphic discretization, transient capabilities, and more sophisticated representation of surface-water features and multi-aquifer wells. Potential applications of the model include evaluation of potential sites for and impacts of new high-capacity wells, development of wellhead protection plans, evaluating the effects of changing land use and climate on groundwater, and quantifying the relationships between groundwater and surface water.

Wisconsin↗