Electrical resistivity surveys identify horizontal and vertical changes in subsurface resistivity that typically correspond to changes in lithology, pore fluid chemistry, and water content of subsurface materials (Sheets, 2002). Land-surface-based two-dimensional (2D) electrical resistivity surveys commonly use many electrodes equally spaced along a straight transect. The spacing of electrodes along a transect generally controls measurement resolution and the depth of investigation. Greater electrode spacing generally corresponds to a greater depth of investigation but lower measurement resolution, whereas finer electrode spacing results in a shallower depth of investigation but higher measurement resolution (Loke, 2000).

Electrodes are connected to stainless-steel stakes spaced at regular intervals along a cable, which is connected to the resistivity meter that controls the measurements. The resistivity meter runs a pre-programmed, data-collection routine that automatically selects two current electrodes and at least two potential electrodes. Electrical current is transmitted into the ground through the current electrodes, and the resulting voltage is measured at the potential electrodes. Electrode groupings and the sequence of measurements are controlled by the type of array specified for the survey. Commonly used arrays include Wenner, dipole-dipole, and Schlumberger (Loke, 2000, 2004). After the survey is completed, electrical resistivity data are downloaded from the resistivity meter and modeled using tomographic inversion software to obtain 2D cross sections of subsurface resistivity.

The parcels identified by the City of Sioux Falls were evaluated by USGS staff before conducting electrical resistivity surveys to identify potential sources of interference (such as underground utilities or infrastructure), verify the presence of borings with lithologic data, and plan the location and length of survey transects. The three survey areas were determined to be free of interference based on underground utility maps (Ted Lewis, City of Sioux Falls Water Division, written commun., 2025), satellite imagery (Esri, 2026a), and field inspection by USGS staff. Lithologic data from nearby borings were obtained from the South Dakota Department of Agriculture and Natural Resources (2025) and the City of Sioux Falls (Ted Lewis, City of Sioux Falls Water Division, written commun., 2025). Lithologic data and parcel dimensions were used to plan the placement of electrical resistivity transects in the three survey areas (figs. 2–4).

One irrigation well and two borings were near the North survey area (not shown in fig. 2). The irrigation well, about 25 meters (m) north of transect NS7, was excluded from analysis because lithologic data were absent and the well was inaccessible to USGS staff. The two borings were more than 200 m from the North survey area; however, both contained lithologic data that were used to estimate the depth of investigation necessary for electrical resistivity surveys. The top of glacial till was about 38 ft (11.6 m) below land surface according to lithologic data from the two borings (South Dakota Department of Agriculture and Natural Resources, 2025). An electrode spacing of 3 m was selected for transects in the North survey area (total transect length of 165 m; fig. 2) to achieve the desired depth of investigation and horizontal and vertical resolution. Transects were arranged in a north-south grid pattern with approximately 30 m of spacing between each (fig. 2). This configuration allowed electrical resistivity results to be combined to produce contours of the top elevation of glacial till.

Several borings and a municipal supply well (well 72) within the Well 72 survey area (fig. 3) had associated lithologic data. Two soil borings were completed before drilling municipal supply well 72 (well 72 #1 and well 72 #2; fig. 3). Well 72 was designed as a specialized horizontal collector well to improve its pumping capacity and extend its operation life. The well consists of a 16-ft-inside-diameter caisson completed to a depth of about 40 ft below land surface, with four 200-ft-long concrete lateral well screens projecting horizontally from the caisson into the Big Sioux aquifer (fig. 3; HR Green, Inc., 2026). Borings surrounding well 72 were drilled to assist design engineers in determining lateral placement. Three of the five transects (W72_3–W72_5) were positioned near borings so that lithologic data could be compared to electrical resistivity results and to fill data gaps between borings. Transect W72_5 was oriented at an angle relative to transects W72_3 and W72_4, which generated intersection points that could be used to check the interpretation of electrical resistivity models from different directions. Transects W72_1 and W72_2 were placed outside of the primary survey area and away from borings for exploratory purposes. The transects were arranged in an irregular grid pattern to enable the generation of contours representing the top elevation of glacial till.

Several borings within the Nose survey area (fig. 4) had associated lithologic data. The top of glacial till below land surface from lithologic data at nine borings (SF14–SF20, SF22, and SF23; fig. 4) varied from 33.6 ft at boring SF15 to 73.7 ft at boring SF23. Because the first transect (L1; total length of 220 m) was positioned near the deepest observed glacial till at SF23, an electrode spacing of 4 m was tested to evaluate the potential depth of investigation in the Nose survey area. Although the 4 m spacing provided sufficient depth to image the deepest glacial till, an electrode spacing of 3 m was used for subsequent transects L2–L7. The 3 m spacing offered adequate depth of investigation and reduced the transect length, which minimized the placement of electrodes in dense vegetation near the ends of transects L2–L7.

The grid pattern of electrical resistivity transects on the eastern side of the Big Sioux River was irregular because data collection was prioritized in areas cleared of vegetation by City of Sioux Falls staff. Transects L1–L3 were arranged in a northeast-southwest orientation along the eastern edge of the Big Sioux River. Transects L4–L6 intersected transects L1–L3, allowing for comparison of results at intersection points and filling data gaps between transects. Transects L4–L6 were not evenly spaced or aligned at consistent angles because USGS staff positioned them based on available cleared areas. Transect L7 was completed for exploratory purposes on the western side of the Big Sioux River to examine the top of glacial till on the opposite bank.

Electrical resistivity data were acquired near the City of Sioux Falls, S. Dak., using an eight-channel, SuperSting R8 resistivity meter (Advanced Geosciences Inc., 2026a) with 56 total electrodes on four cable sections (14 electrodes per section). Electrodes along each cable were connected to stainless-steel stakes placed at evenly spaced intervals (2, 3, or 4 m) along predetermined transects on the land surface. Unless the ground was saturated from precipitation or snowmelt, the land surface was sprayed with freshwater after placing the electrodes to reduce electrical contact resistance before data collection. The cables were then connected to the SuperSting R8, which automatically collected resistivity data by switching between programmed combinations of current and potential electrode pairs using the dipole-dipole array configuration. The dipole-dipole array configuration was chosen because the array setup is relatively simple, requires less survey time than other methods, and has relatively balanced horizontal and vertical resolution. Loke (2000) and Zohdy and others (1974) provide a detailed description of the dipole-dipole array.

Models were prepared for each transect using the EarthImager2D tomographic inversion software suite (Advanced Geosciences Inc., 2026b). The first step in model preparation was the automatic and manual removal of noisy data in EarthImager2D. Noisy electrical resistivity data typically appears as data points with abnormally high or low electrical resistivity values that deviate from surrounding data points. Sources of noisy data can include poor contact between electrodes and the land surface, the presence of nearby conductors above and below land surface (chain-link fences or metal pipes), variations in subsurface material or land-surface topography, and electrical noise from internal or external factors, such as changes in current flow, battery voltage, and air temperature. The automatic data removal used seven criteria specified in the recommended EarthImager2D settings to remove noisy data before inversion. The second step in model preparation was to incorporate elevation information using land-surface elevation data from RTK surveys for each electrode to minimize errors caused by topographic changes.

The third step in model preparation was to choose the method of inversion and remove noisy data identified after inversion. Electrical resistivity data were modeled using a least-squares smooth model inversion (Constable and others, 1987; Farquharson and Oldenburg, 1998) because the surveyed area was best approximated as a simple layered model with only three or four layers, rather than a more complex model with geologic structures or intrusion of conductive waters, such as seawater. Smooth model inversion reduces differences between measured values and model-predicted values in a simple starting model using an iterative process, in which successive iterations attempt to reduce the difference between measured data and predicted values by determining the root mean square error (RMSE). The goal of smooth model inversion is to reduce the RMSE until a desired error tolerance is met rather than to minimize the RMSE, which increases model roughness, causes artifacts, and diverges from the simple layered model (Constable and others, 1987). The maximum number of model iterations was set to eight for all ERT transects, and the accepted model was chosen as the one that best fit the site conceptual model and was closest to the convergence criteria. In some instances, noisy data missed during the first step were removed manually after the first inversion, then a new inversion was completed. For all ERT transects, the accepted model was chosen within the first three iterations and the RMSE, expressed as a percentage, of accepted models was 6 percent or less. Finalized 2D ERT models meeting convergence criteria are provided in appendixes 1–3.

ERT models from the three survey areas (figs. 2–4) displayed consistent vertical distributions of electrical resistivity (apps. 1–3) matching the three or four distinct layers observed in lithologic data from nearby borings (fig. 5; South Dakota Department of Agriculture and Natural Resources, 2025) and in airborne electromagnetic results (Valseth and others, 2018). Electrical resistivity values for geologic materials vary (Telford and others, 1990), so interpretations from ERT models were constrained using lithologic data from nearby borings. Low electrical resistivity values (less than 30 ohm-meters [ohm-m]) near the land surface were interpreted as the soil horizon (layer 1; fig. 5), which was about 2–3 m thick and relatively uniform along the transect. Aquifer materials comprising the Big Sioux aquifer (sands and gravels; layer 2) generally underlie the soil horizon, where present, and are represented in ERT models as the layer of relatively high electrical resistivity values (greater than 60 ohm-m) immediately below the uppermost low electrical resistivity layer (fig. 5). The thickness of the high electrical resistivity layer varies from about 5 m to more than 10 m in some areas (apps. 1–3). Deposits of glacial till underlie the Big Sioux aquifer (layer 3) and appear as the thick (greater than 6–8 m) low electrical resistivity (less than 40 ohm-m) layer below the high electrical resistivity values in ERT models (fig. 5). A fourth layer was observed in ERT models within the North survey area (fig. 2; app. 1). The fourth layer was characterized by relatively high electrical resistivity values and underlies the thick low electrical resistivity layer representing deposits of glacial till (fig. 5). The fourth layer likely represents the Sioux Quartzite bedrock and correlates well with airborne electromagnetic results (Valseth and others, 2018).

The contact between the Big Sioux aquifer and the underlying glacial till was delineated using ERT models (apps. 1–3) and, when available, was constrained using nearby lithologic data from borings (South Dakota Department of Agriculture and Natural Resources, 2025). The contact was manually delineated for all ERT transects near the transition from the relatively high electrical resistivity values of the Big Sioux aquifer (layer 2; 60 ohm-m or greater) to the low resistivity glacial till (layer 3; typically 40 ohm-m or less; fig. 5). Delineated contacts do not span the entire length of the transect because the models are truncated along edges where data are relatively sparse. ERT models, and therefore delineated contacts, are most accurate near the center of each transect where the greatest concentration of data is collected. Greater uncertainty exists for delineated contacts near the edges of ERT models. The top elevation of glacial till from nearby borings was used to constrain delineated contacts if the borings fully penetrated the Big Sioux aquifer, contained lithologic data, and were within 20 m of a transect. The distance between the boring and the ERT transect was considered during delineation; borings that were about 20 m from the transect had less influence on the placement of the delineated contact than those closer to the transect. Additionally, some ERT transects intersected, which allowed for delineated contacts to be further constrained. ERT models were inspected at intersection points to ensure model results matched and that the top elevation of delineated surfaces was within 0.1 m.

Delineated contacts between the Big Sioux aquifer and deposits of glacial till were used to construct contour maps of the top elevation of glacial till and the thickness of Big Sioux aquifer materials for the three survey areas (fig. 1). The delineated top elevation of glacial till was treated as a 2D surface. Its elevation values were extracted at each electrode along ERT transects (56 electrodes per transect) to create a dataset of points to be used in the construction of contours. The thickness of the Big Sioux aquifer was calculated by subtracting the top elevation of glacial till from the land surface elevation at each electrode. Thickness estimates for the Big Sioux aquifer include the soil horizon, which was, on average, about 2–3 m thick (apps. 1–3). The point datasets of the top elevation of glacial till and the thickness of Big Sioux aquifer materials are provided for all 22 transects in the data release accompanying this report (Medler and Anderson, 2026).

Three-dimensional raster datasets were interpolated from the point datasets delineated from ERT models and from borings that fully penetrate the Big Sioux aquifer using the Topo to Raster tool in ArcGIS Pro software (Esri, 2026b). Raster datasets were then contoured using the Contour tool (Esri, 2026c) at various intervals depending on the variability of elevation and thickness in the three areas of interest. Contours generated using the tool were manually adjusted to correct areas of relatively sparse data near the edges of ERT models or where transects did not align in a grid pattern. Contours were considered inferred when more than 20 m from a transect. Shapefiles for contours of the top elevation of glacial till and the thickness of the Big Sioux aquifer for the three survey areas are provided in the data release accompanying this report (Medler and Anderson, 2026).

Single-base RTK surveys were used to measure the three-dimensional positions of ERT electrodes. Single-base RTK surveys use Global Navigation Satellite Systems technology consisting of a single stationary receiver (or “base station”) and a mobile unit (or “rover”) to collect spatial data at survey points. Base stations send real-time differential corrections via radio transmission to the rover so the elevation of a site can be measured. Before the real-time corrections are applied from the base station, the rover undergoes an initialization process that downloads information, such as satellite positions and transmission time delays used to calculate the rover’s position (Rydlund and Densmore, 2012). Traditionally, the base station is set on a surveyed benchmark for comparison with established position data for quality-assurance purposes. In areas without a benchmark, the base station can be set at any location. For surveys where the base station is not set on a surveyed benchmark, the elevation and position of spatial data collected by the rover are determined as a differential from the base station (Rydlund and Densmore, 2012).

Established survey benchmarks were not used for RTK surveys in this study because they were too far from transects in the study area for the base station and rover to maintain a reliable connection. The poor connection was likely caused by topographic obstruction, the distance between the two receivers, or a combination of both. Instead, temporary benchmarks were established by occupying locations close to survey transects for at least 4 hours, when possible, while the rover was used to collect spatial data along survey transects. Static occupation files from the base stations were processed using the Online Positioning User Service (OPUS; National Geodetic Survey, 2014). The coordinates and land-surface elevation data of the electrodes were adjusted based on the difference between the OPUS solution and the position data collected by the base station. The accuracy of the surveyed temporary benchmark elevations was estimated from the overall RMSE calculated by the OPUS solution. During some surveys, the static occupation file was recorded for less than 4 hours. In these cases, the OPUS solution was evaluated by checking the RMSE, plotting the corrected positions and comparing them to plotted data from other survey lines, and comparing elevations from RTK surveys to 1-m digital elevation models (DEMs) of the area (U.S. Geological Survey, 2024).

The estimated uncertainty and RMSE for position information determined from static occupation files processed by OPUS varied by site. In the North survey area (fig. 2), the horizontal (latitude and longitude) uncertainty ranged from 0.003 to 0.012 m and the vertical (elevation) uncertainty ranged from 0.028 to 0.058 m. The overall RMSE for the North survey area ranged from 0.015 to 0.036 m. The average, minimum, and maximum differences between RTK-surveyed elevations at the 560 surveyed electrode points from 1-m DEM were −0.060 m, −0.266 m, and 0.003 m, respectively.

In the Well 72 survey area (fig. 3), the horizontal uncertainty ranged from 0.004 to 0.048 m and the vertical uncertainty ranged from 0.055 to 0.65 m. The overall RMSE for the Well 72 survey area ranged from 0.013 to 0.016 m. Surveyed elevations for the 280 electrode points were compared to elevations from the 1-m DEM; however, the DEM was created before completion of well 72 and did not include the fill material added for the well housing and support buildings (fig. 3). The added fill material caused elevation differences as large as 1.55 m along transect W72_4 at electrodes 36 to 56 (fig. 3). Excluding electrodes 36 through 56 along transect W72_4, the average, minimum, and maximum differences between RTK-surveyed elevations and DEM elevations were −0.187 m, −0.558 m, and 0.072 m, respectively.

In the Nose survey area (fig. 4), the estimated horizontal uncertainty ranged from 0.002 to 0.458 m and vertical uncertainty ranged from 0.071 to 1.872 m. The overall RMSE ranged from 0.016 to 0.220 m. The average, minimum, and maximum differences between RTK-surveyed elevations at the 392 surveyed electrode points and the 1-m DEM were 0.057 m, −0.624 m, and 1.429 m, respectively. Elevation differences greater than 0.5 m were attributed to the Nose survey area being cleared of vegetation (bushes, trees, and so on) and its ground being leveled a few days before ERT and RTK surveys. Uncertainty and RMSE could not be calculated for the static occupation file for transect L6 because the file was corrupted and could not be processed by OPUS. Instead, the average base station position from transects L1–L5 was used to correct horizontal and vertical data for transect L6 because the base station was set up in the same location for each RTK survey.