The unconsolidated Quaternary geologic materials from the land surface to the underlying bedrock units are known from well data and from exposures at the Klein and Conklin Quarries near the City of Coralville, Iowa (Kemmis and others, 1992). These deposits consist of a complex interlayered sequence of glacial till and intertill deposits, including paleosols and associated sediments, capped by loess or alluvium (Hallberg, 1980; Tassier-Surine and others, 2007). The thickness of these deposits varies across the model area from 0 m along segments of the Iowa and Cedar Rivers, where bedrock intersects the land surface, up to 141 m in southeastern and western Johnson County and parts of Cedar, Muscatine, and Louisa Counties, where bedrock valleys are located (fig. 3). The glacial till that forms the bulk of these Quaternary deposits consists primarily of clay with discontinuous lenses of silt, sand, and gravel (Schulmeyer and Schnoebelen, 1998). The upper part of the till is mostly oxidized and fractured, and the deeper unweathered till is unoxidized and unfractured (Schilling and Tassier-Surine, 2006). These deposits, particularly the unweathered till, generally have low permeability and are considered confining units that limit the vertical movement of groundwater. Where these deposits are present, recharge to the underlying bedrock is relatively lower. Where they are relatively thin or absent, recharge to the underlying bedrock is higher. Some private water wells are completed in the Quaternary deposits, mainly in the southern and western parts of Johnson County and usually in shallow alluvium or buried sand and gravel deposits. The City of Coralville has three production wells screened in a buried sand and gravel deposit, but these deposits seem to be isolated units within the glacial tills (Deb Tisor, River Products Company, Inc., written commun., January 2023).

Devonian bedrock is the uppermost bedrock in most of the model area; Pennsylvanian- and Mississippian shales overlie Devonian bedrock in a small part of the model area, and Devonian bedrock has eroded along the Cedar River in the eastern part of the model area, exposing Silurian bedrock. Devonian System bedrock is divided into the Upper Devonian, Middle Devonian, and Lower Devonian Series (Hansen, 1970).

The Upper Devonian Series is the uppermost bedrock group in the southwestern half of Johnson County and is the first bedrock penetrated by wells near the Cities of North Liberty, Tiffin, and Hills (fig. 3). The Lime Creek Formation is the prevalent geologic unit in the Upper Devonian, but the overlying Sheffield and Maple Mill Formations also are present in some areas. These units are composed predominantly of shale, although lesser carbonate interbeds are present. The thickness of the Upper Devonian Series ranges from 0 m in the northeastern part of the county to more than 100 m in the southwestern part of the county. Exposures of this unit are few in Johnson County but present in the Klein Quarry west of Iowa City, and small exposures are present along Coralville Lake. The Upper Devonian Series is a confining layer in the model.

The Middle Devonian Series in the model area consists of the Cedar Valley Group, a thick sequence of limestone present throughout Johnson County, except in the northeast, where Devonian units have eroded and Silurian units are present at the land surface (fig. 4). The Cedar Valley Group contains multiple formations, including from youngest to oldest, the Lithograph City Formation, the Coralville Formation, and the Little Cedar Formation. The Cedar Valley Group is exposed at the land surface along the Iowa River in Iowa City and upstream along Coralville Lake. The Cedar Valley Group is made up of relatively dense, pure limestone. The high calcite mineral content of the Cedar Valley limestones makes them susceptible to dissolution by groundwater, creating karst features such as caves and bedding plane voids. The dense and crystalline nature of the limestone imparts brittleness, and vertical and horizontal fractures are visible in some exposures. These fractures, some of which may be solution-enlarged, are the primary conduits for water movement through this unit (Witzke and others 1988). The Cedar Valley Group and the top part of the Lower Devonian, described in the following paragraph, form the Devonian aquifer, which serves as a source of water for some domestic wells in Johnson County. Where present, the aquifer typically ranges from 29 to 33 m thick.

The Lower Devonian Series in the model area is made up of the Wapsipinicon Group, which includes the Davenport, Spring Grove, and Kenwood Members of the Pinicon Ridge Formation, and the Otis Formation. The Davenport and Spring Grove Members, which consist of limestone and dolostone, make up the lower part of the Devonian aquifer. The Kenwood Member consists of 3.0–4.6 m of shale or shaly dolostone and is a confining unit underlying the Devonian aquifer, separating it from the underlying Silurian aquifer. The elevation of the top of the Devonian aquifer and the top of the shale of the Kenwood Member is shown in figures 5 and 6, respectively. The Otis Formation is a thin layer of fractured limestone and dolomite at the base of the Devonian System. The Otis Formation is hydrologically connected with the underlying Silurian bedrock and is part of the Silurian aquifer. Because the Otis Formation is thin, discontinuous, and not typically delineated in well logs, the elevation of the top of the Silurian aquifer used for this model is the elevation of the top of the Silurian rock.

The Silurian bedrock in the model area consists of four geologic formations (from youngest to oldest): the Gower Formation, the Scotch Grove Formation, the Hopkinton Formation, and the Blanding Formation. The four formations are hydrologically connected and form one hydrogeologic unit called the Silurian aquifer. Witzke (1992) presented a detailed stratigraphy of these units from the Silurian outcrop belt of adjacent Linn and Jones Counties.

The Silurian units are a thick sequence of variably cherty dolostones and demonstrate lateral and vertical variation in texture, porosity, and fracture density (Gannon and others, 2011). Some of these variations are within the subsurface in the model area, but their effect on the hydraulic properties of the aquifer has not been investigated. Substantial variations in water yield have been recorded for many wells completed in the Silurian aquifer, likely related to changes in porosity and fracture density and the presence or absence of karst or solution features. Within the model area, the Silurian rocks composing the Silurian aquifer are mostly confined by overlying shales of the Kenwood Member and glacial till. The one known exception is along the Cedar River, where the Devonian units are absent and the glacial tills are relatively thin. In these areas, the uppermost bedrock is Silurian. The elevation of the top of the Silurian units and the points representing well logs used to define this surface (Iowa Geological Survey, 2022) are shown in figure 7.

In the model area, the Silurian-age stratigraphy is thinnest (~3.0 m) in the southern part and thickest (~140 m) along the Linn and Johnson County boundary and in parts of Cedar County. In the Iowa City and City of North Liberty areas, the Silurian stratigraphy generally ranges in thickness from 61 to 76 m. Some variations in thickness are due to the uneven surface of the underlying Maquoketa Formation. Thinner areas of Silurian rock also are within bedrock valleys in the northern and southern parts of the model area, as shown on figure 8.

The Silurian rock is underlain by low-permeability shales of the Upper Ordovician Maquoketa Formation (fig. 2; Gannon and others, 2011). The Maquoketa Formation is composed of shale strata with secondary dolomite units, especially in the lower and middle parts of the formation. The Maquoketa Formation is between 61 and 76 m thick in the model area. The thick shales form a regional confining unit that separates the overlying Silurian aquifer from the underlying Ordovician and Cambrian aquifers, which are not included in this model. The elevation of the top of the shales of the Maquoketa Formation, which was created using well information from GeoSam (Iowa Geological Survey, 2022), is shown in figure 9.

The uppermost hydrogeologic layer (model layer 1) included the following systems, groups, or formations combined into a single confining unit: the Quaternary System or undifferentiated deposits, Pennsylvanian System, Mississippian System (only present in a small part of the model area), and Upper Devonian System (western half of the model area). The top of layer 1 was defined as the land surface. The purpose of this layer was to provide a source of net recharge and to create confining conditions for model layer 2. The ground surface elevation was sampled at the model-cell scale from light detection and ranging digital elevation data (Iowa Geospatial Data, 2023).

Layer 1 varies in thickness from less than 7.6 m in parts of Cedar and northeastern Johnson Counties to more than 152 m in parts of Benton, Iowa, western Johnson, and Washington Counties. Layer 1 was conceptualized as a regional confining layer over most of the model area. Exceptions exist along parts of the Iowa and Cedar Rivers, where alluvial deposits are in direct contact with the Devonian and (or) Silurian aquifers. For most of layer 1, horizontal hydraulic conductivity values were uniformly assigned as 0.11 meter per day (m/d), and vertical hydraulic conductivity values were assigned as 0.011 m/d (fig. 10). Along the major river systems, Coralville Lake, and Lake Macbride (fig. 1), horizontal hydraulic conductivity values were uniformly assigned as 0.18 m/d (fig. 10). Hydraulic conductivity values were based on drillers’ lithologic descriptions and standard values from literature (Freeze and Cherry, 1979).

Layer 2 represents the Devonian aquifer and includes the Cedar Valley Group (Lithograph City Formation, Coralville Formation, and Little Cedar Formation), and the Spring Grove and Davenport Members of the Pinicon Ridge Formation. The Devonian aquifer crops out extensively along the Iowa River in Iowa City and upstream along Coralville Lake. The Devonian aquifer is mostly composed of dense, relatively pure limestone. Most of the permeability is related to secondary permeability, such as fractures, bedding plane voids, and solution cavities. Cedar Valley Group limestones are susceptible to dissolution by groundwater, which creates karst features such as caves and bedding plane voids. These karst features are the primary conduits for groundwater movement (Witzke, 1992; Ghasemizadeh and others, 2016; Cai and others, 2021).

The top elevation of the Cedar Valley Group rock, as well as the dip of the individual units and field mapping of exposed bedrock units (Gannon and others, 2011), was based on information compiled from well logs in GeoSam (Iowa Geological Survey, 2022). Hydraulic properties were determined from direct testing of the aquifer using controlled aquifer tests or aquifer pump tests on wells completed in the Devonian aquifer using known pumping rates, pumping duration, and water-level measurements (table 1; Iowa Geological Survey, 2023). The range of hydraulic conductivities derived from pump tests was 0.09–4.54 m/d but can be much higher because of karst features (table 1). A hydraulic conductivity of 5.12 m/d was defined for the city limits of the City of Coralville, 0.18 m/d along the major rivers, and 1.12 m/d for the rest of layer 2 within the model (fig. 11).

Layer 3 represents the Kenwood Member of the lower Devonian Pinicon Ridge Formation and is a relatively low-permeability layer confining the underlying rock in most of the model area (fig. 12). The top of the shale of the Kenwood Member (where present) was generated by adding 10 feet to the top of the Silurian surface. The horizontal and vertical hydraulic conductivity of layer 3 was estimated based on the lithology in well logs. Most of layer 3 was assigned a horizontal hydraulic conductivity value of 0.11 m/d and a vertical hydraulic conductivity value of 0.011 m/d. Areas of layer 3 along the major rivers were assigned a hydraulic conductivity value of 0.18 m/d.

Layer 4 represents the Silurian aquifer and consists of rock from the Devonian Otis Formation and Silurian units. The Silurian aquifer is a thick sequence of dolostones and limestone. The top elevation of the Silurian rock surface, as well as the dip of the individual units and field mapping of exposed bedrock units (Gannon and others, 2011), was based on information compiled from well logs in GeoSam (Iowa Geological Survey, 2022). Silurian aquifer thickness was assumed to be the thickness of the Silurian units in the model; although the Silurian aquifer also includes the Devonian Otis Formation, this unit is relatively thin, is not distinguished in most well logs, and was not included in the thickness of the Silurian aquifer.

Data from 27 aquifer pump tests completed partly or entirely in the Silurian aquifer were used to hydraulically characterize the aquifer in the model area (Iowa Geological Survey, 2023). Of the tests, 16 were pump tests using 1 or more production wells and 1 or more observation wells, and 11 tests were recovery tests completed with only the production well (fig. 13; table 1; Williams and Gannon, 2026). Transmissivity estimates from pumping tests of the Silurian aquifer ranged from 8.44 to 346.06 square meters per day (m²/d; table 1). Williams and Gannon (2026) estimated the arithmetic mean transmissivity value for the aquifer was 72.3 m²/d (Williams and Gannon, 2026). The variability in transmissivity is because of the secondary permeability in the Silurian aquifer from fractures and voids, especially in the Gower Formation. The fractures and voids have limited lateral extent and may not be representative of the regional permeability distribution; therefore, transmissivity in untested parts of the aquifer may be much higher than the values listed in table 1. Based on pump test results, hydraulic conductivity in the aquifer ranged from 0.091 to 4.54 m/d with an arithmetic mean of 0.91 m/d and standard deviation of 0.85 m/d (Williams and Gannon, 2026). These hydraulic conductivity values were averaged to zones in the model grid based on the spatial distribution of values observed in pump tests (fig. 14).

Hydraulic storage parameters (dimensionless) also were estimated for the Silurian aquifer from pumping test data. The storage coefficient, or storativity, was estimated for transient model construction and is equal to the volume of water released from a vertical column of the aquifer per unit surface area of the aquifer and unit decline in water level (Freeze and Cherry, 1979). Pumping test data for the storativity of the Silurian aquifer ranged from about 1.0×10⁻⁷ to 7.9×10⁻⁴, and the arithmetic mean storativity value was 1.5×10⁻⁴ (Williams and Gannon, 2026). An average storativity value of 1.0×10⁻⁴ was used for the Silurian aquifer based on the pump test results (Williams and Gannon, 2026). For confined aquifers, specific storage can be estimated by dividing storativity by the aquifer saturated thickness (Freeze and Cherry, 1979). The specific storage distribution in the Silurian aquifer estimated using a uniform storativity value of 1.0×10⁻⁴ divided by the isopach thickness of the Silurian aquifer (fig. 8) is shown in figure 15. Specific storage varied from 2.5×10⁻⁶ per meter across the northern parts of the model area to 1.0×10⁻⁵ per meter across the southern parts of the model area.

Layer 5 is the lowest model layer and represents the Maquoketa Formation, which is characterized as a low-permeability confining layer. The Maquoketa Formation is dominated by shale strata and averages ~61.0 m thick in the model area. This layer was conceptualized as a homogenous bottom confining unit for the model, and hydraulic conductivity for this layer was set uniformly at 0.11 m/d (0.011 m/d vertical hydraulic conductivity) based on published values for shale (Freeze and Cherry, 1979).

An estimated steady-state potentiometric surface was interpolated from water-level measurements collected from known Silurian wells and average river stage elevation with the assumption that the water level in the Silurian aquifer approximately matches river stage at the location of major rivers (fig. 16; table 2). Steady-state conditions were defined using the river stage elevations and precipitation data for time intervals representing average or normal climatic conditions. Water-level data from 46 wells were used, including 39 static water levels from drillers’ logs representing dates from 1899 to 2023 (most from 2000 and later) and averaged time-series data for 7 observation wells with multiple water-level measurements (table 2; Iowa Geological Survey, 2022; Iowa Department of Natural Resources, 2023). Average river stage (nondrought and nonflood) elevation data from 2000 to 2022 for nine USGS streamgages (fig. 17) in the model area also were used to define the groundwater potentiometric surface; the nine USGS streamgages were the Iowa River at Marengo (05453100), Iowa River below Coralville Dam near Coralville (05453520), Iowa River at Iowa City (05454500), Iowa River near Lone Tree (05455700), and Iowa River at Columbus Junction (05465100; USGS, 2023a, b, c, d, i) and Cedar River at Blairs Ferry Road at Palo (05464420), Cedar River at Cedar Rapids (05464500), Cedar River at Cedar Bluff (05464780), and Cedar River near Conesville (05465000) (USGS, 2023e, f, g, h). Measured groundwater levels and stream stage were interpolated to model cells to generate the Silurian aquifer water table surface. Groundwater-flow direction in the Silurian aquifer was estimated by drawing lines perpendicular to equal head contours of the water table.

The regional groundwater head gradient in the model area is generally higher in the northwest and lower in the southeast. Rivers and quarries cause groundwater depressions in the water table (fig. 17). Groundwater depressions also are present at concentrated pumping areas in Johnson County, especially near the City of Coralville and Klein and Conklin Quarries. The water table gradient associated with the Johnson County cone of depression is steeper than the general regional gradient.

Recharge to the Silurian aquifer in the model area is mostly from precipitation and is assumed to be highest where the bedrock is at or near the land surface. Other inflow sources include seepage from the Iowa and Cedar Rivers into the aquifer, seepage from Lake Macbride and Coralville Lake (fig. 1), and groundwater inflow from outside the model area in Benton County and the northwestern part of Linn County. The primary aquifer outflows include withdrawals from wells, dewatering pumping from quarries, discharge to the Iowa and Cedar River systems, and groundwater outflow from the model area.

Net recharge is the amount of precipitation, minus evapotranspiration, that enters model aquifer layers from overlying confining and semiconfining beds. Net recharge to the Silurian aquifer in the northern part of Johnson County was estimated at ~0.30 centimeter per year (cm/yr; Hansen, 1970; Tucci and McKay, 2006). If that recharge rate is applied over the current model area (~3.72 billion square meters), the total volume of recharge per day would be ~0.031 million cubic meters per day (m³/d). The uniform recharge rate used by Tucci and McKay (2006) for the Silurian aquifer in Johnson County is not representative of the entire model area. Spatial variability in the vertical permeability and the thickness of the upper confining units in layer 1 results in a spatial variation of values for recharge or vertical leakage to the underlying aquifers. Because recharge was conceptualized as net recharge, evapotranspiration was not included separately in the model.

For the conceptual model, layer 1 was divided into several recharge zones to allow for spatial variation in recharge rates. Higher recharge values were assigned to areas in central and northeastern Johnson County where Silurian bedrock is within 7.62 m of the land surface (fig. 3). Lower recharge values were assigned to areas with thicker glacial tills (layer 1) and Devonian confining beds (layer 3). The interpolated potentiometric map also was used to delineate areas of higher and lower recharge outside of areas of heavy groundwater withdrawal (fig. 16), where areas with higher groundwater elevations were associated with zones of higher recharge and areas with lower groundwater elevations were associated with zones of lower recharge. The rate of groundwater lost or gained by various reaches of the Iowa and Cedar Rivers was generalized for the model area, and quantifying these values more precisely is outside the scope of the study.

Calibration is the process of adjusting model parameters to minimize the differences between numerical model outputs and hydrologic observations. The MODFLOW 2005 version of the steady-state model was calibrated manually and with PEST calibration software. Calibrated values for recharge and hydraulic conductivity values were used as parameters for the MODFLOW 6 versions of the steady-state and transient models.

The hydraulic conductivity distribution in the Silurian aquifer (model layer 4) was calibrated using PEST calibration software version 12.3.1 (Doherty and Hunt, 2010; Doherty, 2018) included in Visual MODFLOW Flex (ver. 9.0). The PEST package adjusts hydraulic conductivity values for each zone in the Silurian aquifer (model layer 4) using an iterative approach to find a single set of model parameters that results in the lowest root mean square error (RMSE) between simulated and observed water levels.

Horizontal hydraulic conductivity values for layer 4 were generated from the PEST calibration of the steady-state model and ranged from 0.40 m/d over the northeastern third of Johnson County to 5.49 m/d around Cedar Rapids (fig. 19). Vertical hydraulic conductivity was not calibrated but was set to one-tenth of the horizontal hydraulic conductivity value for each zone.

The lowest value for the RMSE during the steady-state calibration was 2.84 m. This error is relatively small compared to the size of the model boundary. By comparison, in the Tucci and McKay (2006) model, the RMSE for the Johnson County Silurian-Devonian aquifer system was 4.15 m, and in the Gannon and others (2011) model, the RMSE for the East-Central Iowa Silurian aquifer was 5.24 m. The average absolute value error in simulated water levels was 2.35 m. Of the 46 measured observed water levels used to compare to simulated water levels, 23 were lower than simulated values, and 23 were higher than simulated values (fig. 20). The range of errors between observed and simulated water levels was 0.009–6.08 m and can be attributed to fluctuations in precipitation recharge, river and streamgage elevation changes, and varying pumping rates from nearby wells.

Sensitivity analysis was completed to determine the effect on the RMSE from adjusting one parameter and holding the other parameters constant. The method used for the steady-state model was to vary one parameter by a certain percentage from the calibrated values and evaluate the resulting RMSE. The changes in RMSE for varying recharge and hydraulic conductivity based on this approach are listed in table 3. The steady-state model was more sensitive to changes in hydraulic conductivity when smaller percentages of change were used and was more sensitive to recharge when larger percentages of change were used. This result is reasonable because the design of other boundary conditions, such as river cells and general-head boundaries used in lakes and quarries, means that as groundwater heads simulated by the model change, more or less water enters the model from these boundary condition features. This design numerically compensates for changed recharge values in the water balance, especially when these changes are small.

The transient model was manually calibrated to 132 observed groundwater levels, including the pumping water levels provided by city staff. In this calibration process, horizontal hydraulic conductivity in layer 4 was adjusted to best match model output water levels to calibration observations. In addition to the production wells, 6 USGS and cooperative USGS and IGS observation wells with more than 20 years of groundwater-level data and 7 wells from GeoSam with groundwater-level data also were used for calibration observations (fig. 21; Bristow and Gannon, 2026). As a result of this process, the layer 4 horizontal hydraulic conductivity values in the refined grid area around the City of North Liberty and Iowa City well fields were adjusted from 1.12 m/d to 0.62 m/d and 0.40 m/d, respectively; the calibrated hydraulic conductivity values from the steady-state model were otherwise retained in the transient model.

Observed and simulated water-level hydrographs for years 2020 through 2022 for 5 groundwater observation wells are shown in figure 22A–E, and 3 production wells are shown in figure 22F–H). The observation well hydrographs (Elmira Depot, Ely North, Ely Northwest, Stockpile, and Lincoln Church) indicate that the model simulated the water table within a few meters of the observed values at these wells but did not capture small-scale water-level fluctuation in areas that are not affected by pumping. The average residual for these five sites ranged from −1.8 to 3.4 m. The hydrographs from production wells have higher errors in water level and indicate more fluctuation over time. The average water-level residual at Solon 4 was −5.6 m, and the simulated water level was lower than observed for most stress periods. Average residuals at Coralville 14 and Iowa City SW4 were 8.2 m and 22.4 m, respectively. Simulated water level at the Iowa City SW4 well was consistently around 22 m too high, but the model accurately simulated water-level fluctuations at this well over time. Iowa City SW4 was the only production well with daily pumping records and daily water-level records. Other cities provided monthly or annual totals rather than daily, which is likely a contributing factor to the higher accuracy of the Iowa City hydrograph because a water-level measurement made once monthly is less representative of the average daily value over that month than a water-level measurement made daily. The higher accuracy of the Iowa City hydrograph indicates that although the model overestimates water levels in some areas of heavy pumping, it still reliably represents drawdowns (changes in water level) over time because of pumping.

A simulated potentiometric map for the final (October–December 2022) stress period of the transient model is shown in figure 23. Lower water-level elevations are in the City of Coralville area near dewatering areas for the quarry and City of Coralville wells.