Streambed temperature mapping surveys were completed during three site visits in 2016 and one site visit in 2017. Streambed temperature was measured using a Traceable Control Company digital thermometer with a 10-cm probe. Measurement locations were surveyed with submeter accuracy using a Trimble R1 Global Navigation Satellite Systems (GNSS) receiver (Trimble, Inc., Sunnyvale, California). Additional details on the temperature mapping methods are in Naftz and others (2019).

Electrical resistivity tomography (ERT) is a surface-based, direct-contact geophysical method for estimating subsurface electrical conductivity (resistivity equals 1 divided by the conductivity). A series of electrodes are implanted into the surface (or in boreholes), and these electrodes are used to measure electrical potential in response to injected currents. The potential fields generated from current injections depend on the distribution of electrical conductivity in the subsurface and the geometry of the electrodes. For more details about the ERT method, see (for example) Binley and Slater (2020). For this study, we used an AGI Supersting R8 system (Advanced Geosciences, Inc., Austin, Texas) with 56 stainless steel electrodes and placed them in roughly linear transects using 1- and 2-m spacings between the electrodes.

A total of 1,063 measurements were made during the electrical resistivity surveys completed at the study site during August 2017. Typical ERT survey geometries were used for data collection including Wenner and dipole-dipole measurements for each 56-electrode survey (Binley and Slater, 2020). Contact resistance was high (greater than [>] 5 kilohms [kΩ]) in pure sand areas. Contact resistances were reduced to less than 3 kΩ in all cases by packing a slurry of clumping clay (for example, cat litter) and stream water around problematic electrodes. Once each survey line was installed, data were collected for about 2 hours.

Five separate ERT transects were collected on a sand bar next to the Little Wind River. The two longer transects (W1W2W3 and W4W5W6) used 2-m electrode spacing in a “roll along” style of collection, wherein overlapping data are collected to generate longer datasets with limited length cables. Data with 1-m electrode spacing were collected along three tie-in lines (W7, W8, W9) nearly perpendicular to transect W1W2W3. Additional details on the ERT survey methods, location of survey lines, electrode spacing, and raw/processed data are in Terry and Briggs (2019).

Electromagnetic induction (EMI) includes several related techniques for measuring subsurface electrical conductivity by generating a primary electromagnetic field within a wire loop that induces secondary fields in conductors within the subsurface. These secondary fields are measured by a receiver coil. In this study, we specifically used a hand-carried frequency domain instrument (GEM-2, Geonics Ltd., Mississauga, Ontario, Canada). The GEM-2 operates at several user-defined frequencies to provide effective sensitivities at various depths (Huang and Won, 2003); however, the actual investigation depths are affected by the distribution of subsurface electrical conductivity.

Roughly 10 line-km of horizontal coplanar EMI data were collected over several days during August 2017. Most of the EMI survey lines were land based; however, one survey line was conducted along the Little Wind River with the instrument strapped to a kayak. Data were collected at five frequencies ranging from 450 to 63,030 hertz (Hz). For most surveys, data were briefly collected at a common reference station to account for slight drifts in values caused by temperature changes in the internal circuitry of the instrument. Additional details on the EMI survey methods, location of survey lines, and raw/processed data are in Terry and Briggs (2019).

From August 6 to September 24, 2015, fiber-optic distributed temperature sensing (FO–DTS) data were collected in the study reach using a 1-km long armored cable (about 5-millimeter [mm] outer diameter) containing multiple multimode optical fibers. The cables were installed along the sediment-water interface of the streambed; one cable was installed through a side channel area and within about 5 m of the larger river shoreline, and the other generally was installed along the center channel of the Little Wind River. The FO–DTS data were collected at 1.01-m linear resolution with an Oryx model SR Remote Logging DTS Unit manufactured by Sensornet Ltd. (Imperial Way, Watford, United Kingdom). The FO–DTS system was programmed to integrate temperature measurement at every cable position every 40 minutes. However, solar power to the control unit failed intermittently during the deployment period, especially later in the record, so the data were not recorded using a consistent timestep. Only data from the near-bank cable are discussed in this report. Additional details on the FO–DTS methods are in Terry and Briggs (2019).

iButton Thermochron band-gap temperature sensors and loggers (Maxim Integrated Products, Inc., Sunnyvale, Calif.) were embedded horizontally in hollow 2.5-centimeter (cm) diameter stainless steel pipes about 40 cm long and secured and waterproofed with silicone caulking. Each stainless-steel pipe contained three temperature sensors and is referred to herein as a “sensor array.” Sensors were synchronized before deployment so that measurements were recorded on the same schedule. A total of seven sensor arrays were deployed during 2016 at as many as three locations each. A total of 11 sensor arrays were deployed at single locations during 2017. During the 2016 deployments, three iButtons were used in the construction of each sensor array, and the pipes were installed vertically in the streambed such that that temperature was recorded at about 0.04, 0.07, and 0.11 m depths below the streambed surface. Sensor arrays deployed during 2017 were designed for iButton installation depths at 0.01, 0.07, and 0.11 m or at 0.01, 0.06, and 0.16 m to provide flexibility in postprocessing. Additional details on the vertical thermal sensor array methods are in Naftz and others (2019).

Tube seepage meters (TSMs) (Solder and others, 2016) were installed in streambed sediments along the left bank of the study reach during 2016 to collect vertical hydraulic head differences and falling head time series data for calculation of vertical seepage rates (q, in meters per day [m/d]) and during 2017 to collect head time-series data for direct measurement of vertical seepage rates. Seepage meters were installed by pounding 7.6-cm inner diameter clear polycarbonate tubing or 7.6-cm inner diameter, stainless steel tubing into streambed sediments at two sites in 2017.

In 2016, hydraulic head differences and falling head time series were measured manually. Hydraulic head differences between the inserted base of the TSM and stream surface level (ΔH) was measured inside the TSM after closing the TSM to the stream by placing a rubber stopper in the drilled hole and allowing the interior water level to equilibrate. Hydraulic head gradient (J, in meters per meter) was calculated as ΔH divided by the depth of TSM insertion into streambed sediments. A falling head test was used to estimate the vertical hydraulic conductivity (K_v, in meters per day) of the streambed sediments inside the TSM using the methods of Genereux and others (2008).

In 2017, TSM measurements were completed using a semiautomated system. The semiautomated system consisted of an interior head measurement device, a valve system that opens and closes the tube interior to the stream, a microprocessor controller, a battery pack, and a data logger to record hydraulic head measurements and associated date/time stamps. Additional details on the manual and semi-automated vertical seepage measurement methods are in Naftz and others (2019).

Four sampling trips were completed during the summers of 2016 and 2017. During each sampling trip, surface-water samples were collected for radon-222 analysis. River radon-222 samples were collected at 10 intervals spaced 200 m apart beginning at the upstream location of the study reach and moving down the reach. Field parameters including temperature, specific conductance, dissolved oxygen, and pH were measured at each site using an Aqua TROLL 600 multiparameter sonde (In Situ, Inc., Fort Collins, Colorado) that was calibrated daily. Bottles were submerged near the thalweg and uncapped, and at least three bottle volumes were flushed through the sample container using a peristaltic pump. Samples were then capped underwater to prevent degassing to the atmosphere.

Shallow streambed drive points and DOE monitoring wells also were sampled to obtain the endmember radon-222 concentration of the groundwater in the shallow aquifer. Shallow aquifer wells were selected throughout the plume to provide a sufficient representation of the groundwater composition. Wells were sampled using a peristaltic pump, and three well volumes were purged before sample collection. The water was pumped into a galvanized bucket to allow for bottle submersion. Radon-222 sample bottles were submerged, and then the pump discharge was inserted into the bottom of the bottle. At least three bottle volumes were pumped through, and then bottles were capped underwater to prevent gas exchange with the atmosphere. Field parameters were measured with an Aqua TROLL 600 multiparameter sonde equipped with a flow-through chamber. Drive points (DPs) were installed in the river near the left bank next to the plume discharge zone. Stainless-steel DPs screened over the lower 30 cm attached to a section of 1.9-cm diameter polyethylene tubing were driven to 1 m in depth; however, some DPs met refusal at shallower depths because of the presence of river cobbles. The depth of refusal ranged from 60 to 100 cm. The DPs were pumped to develop the well and then sampled in a similar manner to the monitoring wells. During sampling, pumping rates were kept low (less than 1,000 milliliters per minute) to avoid contamination of hyporheic water with overlying stream water. Additional radon-222 samples were collected from smaller diameter (0.25 cm), nested, stainless-steel DPs at 30-, 50-, and 70-cm depths next to the 1-m depth DP sites. All radon-222 samples were collected in clear, 250-milliliter (mL) Boston round glass bottles. All radon and supporting water-quality data from the surface-water and groundwater sampling sites are available in Naftz and others (2019).

Temporary bank operated cableways were established across the Little Wind River at three transects along the study reach during August 9, 2016, and August 24–25, 2017. Streamflow was measured multiple times at each transect using an acoustic Doppler current profiler (ADCP) according to established USGS methods (Turnipseed and Sauer, 2010). Stream discharge data from each transect were used to define three equal discharge increments at each transect. Water-quality samples were collected at the midpoint of each equal distance increment at each transect on August 11, 2016, and August 27, 2017, according to established USGS methods (USGS, 2012). Water sample aliquots included filtered and acidified, filtered and unacidified, and unfiltered and acidified. Water samples were acidified with trace-metal grade concentrated nitric acid to 1 percent (volume to volume). Field filtration was performed using a peristaltic pump and 0.45-micrometer (μm) capsule filters. Additional details on the equal distance increment sampling methods are available in Naftz and others (2019).

Pore water was sampled in the upper 100 cm of streambed across the zone of the contaminant plume at sites with groundwater upwelling through the streambed identified by streambed temperature surveys (see the “Streambed Temperature Mapping” section). Samples were collected from DPs installed at 30, 50, and 70 cm and 1 m (nominal depth) below the streambed, and from the USGS MINIPOINT minipiezometer (MP) array (Duff and others, 1998; Naftz and others, 2019) installed at 3-cm intervals between 3 and 15 cm deep below the streambed. The DPs installed at 1-m nominal depths had an inside diameter of 2.5 cm (referred to as large-diameter DPs), and the DPs placed at 30-, 50-, and 70-cm depths had an inside diameter of 0.9 cm (referred to as small-diameter DPs). A sample of overlying stream water was sampled from a MP sampling port placed 3 cm above the streambed surface. The DPs and MPs were sampled for field parameters (pH, specific conductance, and dissolved oxygen), major cations and anions, and selected trace elements including uranium and molybdenum, the primary elements of concern in the legacy contaminant plume (see the “Introduction”). DP and MP samples were collected following methods described in Fuller and Harvey (2000). All MP depths were sampled simultaneously, and small-diameter DPs were sampled sequentially immediately after MP sampling.

Streambed sediment cores were collected during August 2017 by inserting a 2.5-cm diameter butyrate clear plastic core tube into the sediment until refusal. Core lengths ranged from 14.5 to 26 cm. Plastic end caps were taped on each end of the core tubes and frozen within 3 hours of collection. The location of each core site and details on how the cores were subsampled and prepared for chemical analysis are available in Naftz and others (2019).

Diffusive gradients in thin-films (DGT) and diffusive equilibrium in thin-films (DET) are methods of measuring fine-scale (centimeters to millimeters) solute concentrations in surface waters and sediment pore waters (Davison and Zhang, 2012). DET devices are typically used to measure equilibrium surface and pore-water concentrations by allowing solutes to diffuse through a membrane layer (0.45 μm) into an inner hydrogel layer. DET probes typically sample a miniscule volume of sediment pore water (1-cm depth intervals equal 0.1 cm × 1 cm ×1.8 cm, or about 0.18 cubic centimeter [cm³]) perpendicular to the probe interface. Whereas DET relies on establishing equilibrium between solutes in the surrounding water environment and in the device, in DGT, solutes continuously diffuse across a diffusion gel layer and progressively accumulate in a resin or binding gel layer (Davison and Zhang, 2012). Furthermore, DGT only binds free ions or weakly complexed ions and thus is considered a good general measure of solute speciation and bioavailability (Zhang and others, 1995; van Leeuwen and others, 2005; Davison and Zhang, 2012). In addition, as the DGT probe continually removes solutes from solution, it effectively perturbs the system and therefore can provide information on solid-solution interactions in the sediments (Lehto, 2016).

DGT samplers were deployed at seven locations in the streambed sediment along the left bank of the study reach during August 2016. DGT samplers were deployed in the upper 15 cm of the streambed sediments for about 24 hours and then transported to a laboratory in Riverton, Wyo., for processing within 2 hours of retrieval. During August 2017, a network of 10 DGT and 10 DET probes were inserted into the sediment on the left bank of the study reach along the Little Wind River. The probes were 15 cm long and were retrieved after about 72 hours and processed within 2 hours after retrieval at the laboratory in Riverton, Wyo. (Byrne and others, 2021).

Surficial streambed sediment samples were collected from depositional areas along the channel margins of the study reach during 2016 and 2017. Areas where there seemed to be relatively recent (1–2 years) bank failure were avoided. Submerged sediment was collected using a polypropylene hand scoop and placed into a modified 10-cm polypropylene Buckner funnel fitted with a 64-µm nylon mesh. Sediment was sieved directly into 500-mL polypropylene bottles using site water. The bottles were stored on ice and shipped to the USGS Metal Bioavailability Laboratory in Menlo Park, Calif., for further processing and analysis. Additional details on the locations, collection, and processing of the surficial streambed sediment samples are available in Naftz and others (2019).

Samples of attached aquatic algae (Spirogyra spp. [Link, in C.G. Nees, 1820]) and aquatic insects were collected from natural streambed substrates during the site visit in August 2016. Algal filaments were removed from the substrate using plastic forceps and placed in opaque, acid-washed, plastic containers with a small volume of ambient stream water and immediately placed in an ice-filled container. Immature larval and nymph stages of insect taxa were collected at each site using a nylon-mesh kick net. Organisms were then placed into acid-washed plastic containers with ambient stream water and frozen within 30 minutes of collection using dry ice. Additional details on the locations, collection, and processing of the algal and macroinvertebrate samples are available in Naftz and others (2019).

During the August 2017 site visit, periphyton were collected using floating-rack samplers installed in free-flowing parts of the stream channel at five locations at a depth of 10 cm below the water surface. Flow velocity and incident solar radiation were determined at each sampling location on the day of sampler installation. Samplers were deployed from August 28 to September 11. At the end of the colonization period, slides were removed from each sampler, stored in sealed opaque containers on dry ice, and shipped to the USGS Metal Bioavailability Laboratory in Menlo Park, Calif. Additional details on the locations, collection, and processing of the periphyton samples are available in Naftz and others (2019).

Surface-water and groundwater samples collected from the Riverton Processing site were analyzed on-site for radon-222 using a Durridge RAD7 solid-state alpha decay detector with RAD H₂O accessory (DURRIDGE Company, Inc., Billerica, Mass.). The detector allows for the measurement of the radon-222 concentration in water with a detection limit of less than (<) 10 picocuries per liter (pCi/L) to >400,000 pCi/L. Four, 5-minute counting cycles were averaged to calculate the radon-222 concentration in water samples. Additional details on the analytical methods used for the analysis of radon-222 during the study are in Goble (2018) and Naftz and others (2019).

The concentration of calcium, magnesium, potassium, sodium, silica, strontium, aluminum, iron, and manganese in water samples collected from the Riverton Processing site were determined using inductively coupled plasma-atomic emission spectroscopy. Additional trace elements (copper, cobalt, molybdenum, uranium, and zinc) were determined by direct injection inductively coupled plasma-mass spectrometry. Concentrations of chloride, sulfate, and bromide were determined by ion chromatography. Dissolved organic carbon was measured by wet oxidation, and total alkalinity was determined by Gran titration. All analyses were performed at the USGS research laboratories in Menlo Park, Calif. Additional details and references on the analytical methods used for these chemical constituents are in Naftz and others (2019).

Three replicates of dried surficial streambed sediment samples were initially digested by hot acid reflux in 16 normal (N) nitric acid (HNO₃), and then evaporated to dryness. For the first replicate, 10 mL of 0.6N hydrochloric acid (HCl) was added, whereas 10 mL of 1N HNO₃ was added to the second and third replicates. After equilibration, sample solutions were filtered (0.45 µm) and analyzed. The first replicate was analyzed for major and trace metals, exclusive of molybdenum and uranium, using inductively coupled plasma-optical emission spectrometry. The second and third replicates were analyzed for molybdenum and uranium using inductively coupled plasma-mass spectrometry. All analyses were conducted at the USGS Metal Bioavailability Laboratory in Menlo Park, Calif.

Dried subsamples from each sediment core were sieved to <1 mm and digested using a mixture of HCl, HNO₃, perchloric, and hydrofluoric acids at low temperature. The decomposed samples were analyzed by inductively coupled plasma-atomic emission spectrometry and inductively coupled plasma-mass spectrometry through an analytical contract administered by the USGS (Analytical Chemistry section of the Central Mineral and Environmental Resources Science Center, Denver, Colo.). Additional details and references on the methods used for the analysis of the chemical constituents in the surficial streambed sediment and sediment core samples are in Naftz and others (2019).

Filamentous algae samples were cleaned of external debris by rinsing and examined under magnification to confirm taxonomic identification. Cleaned samples were oven-dried at 70 degrees Celsius (°C), and part of the dried sample was digested using concentrated HNO₃. Samples were evaporated to dryness, reconstituted in 0.6N HCl, filtered through a 0.45-µm pore-size filter, and analyzed for metal content using inductively coupled plasma-optical emission spectrometry. An additional part the dried sample was combusted at 400 °C, re-dried to a constant mass, and reweighed to determine ash-free dry mass. Aquatic insect samples were inspected by microscopy to ensure sample integrity and sorted to their lowest practical taxonomic level (usually genus). After sorting, each sample was oven-dried at 70 °C, weighed, and digested by reflux using concentrated HNO₃. After digestion, samples were evaporated to dryness, reconstituted in 0.6N HCl, filtered through a 0.45-µm pore-size filter, and analyzed for metal content using inductively coupled plasma-optical emission spectrometry. Additional details and references on the methods used for the analysis of the chemical constituents in algae and macroinvertebrate samples are in Naftz and others (2019).

Autotrophic (chlorophyll a) and heterotrophic (ash-free dry mass) biomass accrual was determined using a subset of slides from each site location using methods described in Naftz and others (2019). Slides selected for chemical analysis from each sample location were scraped with a Teflon spatula into a metal-free beaker with a small volume of deionized water for subsequent filtration, digestion, and chemical analysis. Additional details on sample digestion and analysis are in Naftz and others (2019).

After slicing the DGT and DET probes into 1-cm sections, each slice was eluted for 24 hours with 1N HNO₃. The solutions were then analyzed by inductively coupled plasma-mass spectrometry for target analytes (DGT was used for uranium and molybdenum; DET was used for uranium, molybdenum, calcium, potassium, iron, manganese, magnesium, and strontium). The DGT solute concentrations were calculated as follows: CDGT=M×Δg÷Dmdl×Ap×t(1)where

Both EMI and ERT data were inverted to produce depth models of electrical conductivity. Briefly, inversion is the process of estimating an earth conductivity model from the geophysical data. The inverse problem (geophysical data to earth model) is typically nonlinear, so various numerical optimization methods are used. Further constraints are typically needed to stabilize the inversion algorithm to provide realistic results. The most common procedure is a spatial regularization parameter that enforces a degree of consistency between adjacent model cells, which gives resulting images a blurry appearance. For a summary of some common approaches for inversion, see Menke (2012).

EMI data were first corrected for instrument drift by using the median difference from each frequency recorded at the reference station before and after individual surveys and applying a linear scaling factor based on those differences (for example, Sudduth and others, 2001). Data were then imported into the Aarhus Workbench software (Aarhus GeoSoftware, 2021) using the ground conductivity module and were resampled to 1-m-spaced soundings using a 3-m moving average window. Data quality was assessed visually, and a few areas with extremely high standard deviations (computed from the moving average) were removed; however, over 99 percent of the data were retained. Inspection of reference station data indicated an approximately 75-part-per-million (ppm) standard deviation for all frequencies, so this value plus 1 percent was used to provide error estimates to the inversion.

The inversion considered 20 layers logarithmically increasing in thickness to 10 m depth. A laterally constrained inversion was used, wherein a one-dimensional (1D) model is estimated at each data point, regularized to the surrounding inversion results. “Medium” vertical and horizontal constraints were used to enforce smoothness between adjacent models, and the inversion successfully converged after a few iterations. For additional details on the inversion approach used in Workbench, see Auken and others (2015). Given the relatively high density of measurements at the site, results were linearly interpolated to aid in interpretation.

ERT data having stacking errors >2 percent, negative apparent resistivities, or apparent resistivities >5,000 ohm meters (Ω•m) were removed before inversion. Measurements were of overall high quality because >95 percent of the data retained after these filtering criteria were applied. Data were assumed to follow a 0.02 Ω•m plus or minus (±) 5-percent error model. The filtered ERT data were inverted using R2 software (Binley, 2021). Further details on ERT inversion are in Binley (2015). Starting resistivity models based on the inverted EMI results were used. All inversions converged successfully within seven iterations. Topography was added to resistivity models using the USGS 3D Elevation Program 1/3 arc-second resolution digital elevation model (USGS, 2013).

To better understand the approximate depth of sensitivity for geophysical data, depth of investigation calculations were done. Depth of investigation for the EMI inversion was calculated using a routine built into the Aarhus Workbench software package (Aarhus GeoSoftware, 2021) and based on Christiansen and Auken (2012). ERT depth of investigation was computed using the method of Oldenburg and Li (1999) using 50 and 1 Ω•m reference models.

FO–DTS data were extracted from the instrument files, geolocated, and analyzed for summary statistics in Matlab (Mathworks, 2021). Georeference information was collected in the field with a handheld global positioning system device at known cable distances whenever the cable installation deviated from a straight line (number of samples [n] =21 points), and unknown cable positions were interpolated between reference points.

The concentration of radon-222 along the study reach that was receiving groundwater inflow was modeled with the equations for 1D steady state, advective stream transport with gas exchange and radioactive decay (Cook and others, 2006): Q ∂c ∂x =I c i −c +wEc−kwc−dwλc (2) ∂Q ∂x =I x −L x −E x (3)where

Groundwater inflow along the study reach was estimated by matching the observed stream concentrations of radon-222 with modeled radon-222 concentrations by varying the groundwater inflow along the reach. Groundwater inflow steps were assigned as varied parameters to coincide with the stream sampling intervals. Groundwater inflow for each step, I, from equation 2 was varied iteratively using the Levenberg-Marquardt algorithm (Marquardt, 1963) to obtain a least-squares fit by minimizing χ-squared residual. The resulting inflow value at each step provides an estimate of the spatial and volumetric distribution of groundwater inflow along the study reach with a change in stream discharge. Additional details on radon-222 modeling methods are in Goble (2018).

Analysis of TSM data collected in 2016 and 2017 was completed using different methods. For 2016 data, vertical seepage rates (that is, vertical specific discharge) were calculated using Darcy’s equation: q= ΔH L  Ky (4)where q

is the specific discharge, in meters per day;

Data collected in 2017 were analyzed to determine q following a modified version of the method of Solder and others (2016). Described briefly, the time-series of interior head after isolation (that is, valve closed) is fit with a second-order polynomial by least-squares regression, with increased deference toward measurements at the start of the period, to describe the change in head through time. The first derivative of the fitted equation evaluated at time equals 0 is equal to q. This differs from Solder and others (2016) in that the time at which to evaluate the fitted polynomial (that is, the interior head equals the stream head) does not need to be estimated. With the zero-displacement valve, the interior head at the time of valve closure (that is, time equals 0) is equal to the stream head. Positive values of q indicate groundwater discharge to the stream. Measurements of q can be converted to a volumetric discharge rate (Q, in cubic meters per day) by multiplying q by the cross-sectional area of the tube (A=0.456 m²). Qualitative seepage direction (gain, loss, or neutral) also is reported for 2017 measurements in which q could not be reliably calculated because of a lack of significant difference (less than the variability of measurements close in time) between water levels at the start and end of the interior head measurement period. A qualitative indicator of gain or loss was assigned when a clear trend in the interior head time series was present, and indicator of neutral was assigned when no trend was observed.

Vertical seepage dynamics at subdaily timescales were inferred with independent logging thermistors at known vertical spacing installed within saturated streambed sediments. The effects of seepage on depth-specific temperature time series are predictable based on the governing 1D advection-conduction model if the thermal properties of the streambed are estimated or measured (Lapham, 1989). Briggs and others (2014) showed that the vertical temperature-based inference of upward seepage is particularly sensitive to the sediment thermal parameter model inputs. Recent reworking of the 1D advection-conduction model solution based on paired amplitude ratio and phase shift of the diurnal signal between two depths in the vertical has allowed in situ evaluation of saturated thermal diffusivity, the master conductive parameter (Luce and others, 2013). Following these principles, Irvine and others (2017a) developed an automated workflow addition to the Matlab-based software VFLUX (Gordon and others, 2012) that uses the geometric mean of the thermal diffusivity time series derived from diurnal signals in the amplitude-ratio based analytical models of vertical water flux. We adopted that approach for the streambed temperature profile-based measurements of seepage within the study reach. During the modeling process early and late time (up to a few days) at the beginning and end of each deployment was removed to reduce edge effects of the diurnal signal extraction process as discussed by Gordon and others (2012).

The net loading of uranium to the Little Wind River between upstream and downstream transects is calculated using mass balance considerations. A mass balance equation for the study reach is given by: Q_dC_d=Q_uC_u+ΔQC_L(5)where C

is the dissolved uranium concentration given by the average of the three equal discharge increment samples at a given transect,

The extent of surface and groundwater mixing in the hyporheic zone was estimated as a function of depth in the streambed and attenuation of reactive solutes using a two-end member mixing of overlying surface water and the groundwater following the approach provided in Fuller and Harvey (2000). The fraction of surface water as calculated using major ion concentrations (for example, sodium) was used as conservative, nonreactive tracers of surface and groundwater instead of an in-stream injection of a conservative tracer. The large difference between major ion chemistry in groundwater and surface water allows calculation of the fraction of each water type. For example, concentrations of sodium in groundwater within the contaminant plume range from 10 to 50 times higher than sodium concentrations in the Little Wind River. The fraction of surface water was calculated for each depth above the sample depth used to define the groundwater component as follows: [Na_z]=F_sw×[Na_sw]+(1−F_sw)×[Na_gw](7)where Na_sw and Na_gw

are the conservative ion (sodium) concentrations of surface water and groundwater, respectively;

The concentration of reactive solutes (for example, uranium and molybdenum) in the streambed resulting from groundwater and surface water mixing at each depth was then calculated in the absence of reaction following Fuller and Harvey (2000) and using the F_sw values, as follows (for uranium): [U_pred,z]=F_sw×[U_sw]+(1−F_sw)×[U_gw](8)where U_sw and U_gw

are dissolved uranium concentrations of surface water and groundwater, respectively, and

A variety of techniques and equipment have been used to identify and quantify groundwater discharge to surface water and include seepage meters, heat flowmeters, DPs, MPs, vertical streambed temperature profiling, tracer tests, measuring differences in streamflow along a stream reach, application of flow and chemical hydrograph separation techniques, and FO–DTS (Conant, 2004; Selker and others, 2006; Henderson and others, 2009; Mwakanyamale and others, 2012). Groundwater discharge to streams and rivers also can be associated with contaminant plumes; however, many of these techniques cannot provide the finer scale (meter to centimeter) information in a cost-effective manner to evaluate how these contaminant plumes interact with the stream or streambed (Conant, 2004; Conant and others, 2004). In addition, application of many of the more obtrusive methods at meter to submeter scales can alter the natural system (Conant, 2004). Streambed temperature mapping provides an inexpensive and unobtrusive method for identifying and measuring groundwater fluxes through a streambed at small spatial scales (Conant, 2004) and can guide the placement of semipermanent monitoring equipment in the streambed, such as vertical thermistor arrays and seepage meters.

Streambed temperature measurements have been applied in a variety of settings with various objectives. Streambed temperature mapping has been used to identify areas of upwelling groundwater supporting biologically critical thermal regimes in a Canadian stream and to guide the placement of streambed piezometers (Drake and others, 2010). Mapping of winter season streambed temperatures was used by Franssen and others (2013) to determine if brook trout (Salvelinus fontinalis [Mitchill, 1814]) spawning site selection in northern latitudes was affected by the presence of upwelling groundwater. Temperature mapping also has been used to estimate groundwater discharge through streambeds on stream subreach to reach scales (Conant, 2004; Schmidt and others, 2007). Methods used to quantify groundwater discharge from streambed temperature mapping include analytical (Schmidt and others, 2007) and empirical (Conant, 2004) approaches.

Streambed temperature maps were constructed for selected areas along the left bank of the Little Wind River during four periods: (1) June 30, 2016; (2) July 28, 2016; (3) August 10, 2016; and (4) August 15–17, 2017. Mapped areas were constrained to potential areas where the legacy groundwater plume intersected the left bank of the Little Wind River. Mapped areas were further constrained by stream channel material that would allow full penetration of the temperature probe. Transition from sand/silt to cobble bottom material prevented the extension of temperature maps beyond about 10 to 25 m from the left bank of the study reach. Migration of the sand channel during calendar years 2016 and 2017 and seasonal changes in river stage substantially changed the areas and relative locations available for streambed temperature mapping (fig. 5). Comparison of the aerial photographs of the study reach taken on May 2, 2014, and April 5, 2018, indicated a 23.5-m extension of the left bank associated with sand channel location A (fig. 5) and a 7.2-m shortening of the left bank associated with sand channel location B (fig. 5). Channel migration can change streambed elevation and associated stream depth, which in turn could change the direction and (or) magnitude of groundwater discharge through the streambed (Conant, 2004). The aerial photograph taken April 5, 2018 (USGS, 2022), is used as a base map in subsequent report sections to show streambed temperature measurement locations, as well as the locations of sites where other types of data (for example DPs, MPs, DETs, and DGTs) were collected inland from the location of the left streambank. To access the air photo data collected during April 5, 2018 (https://earthexplorer.usgs.gov), select the “Data Sets” tab, expand “UAS,” and select “UAS – Ortho.” Under “Additional Criteria,” expand “Project” and enter “USGS_WY_Little_Wind_River.” Hit the “Results” radio button to view the data. Despite their appearance on the base maps, temperature measurement and other data collection sites were consistently within the areas of active streamflow during each water year (2016 and 2017) of the study period.

Data used to construct the streambed temperature maps are provided in Naftz and others (2019). Variations in streambed temperature at the Riverton Processing site were used to help guide the selection of monitoring and sampling sites including TSMs, DPs, MPs, continuous vertical thermal profilers (iButtons), and DET/DGT samplers. Boxplots were used to compare the streambed temperature data distributions collected during the four monitoring periods during June 2016 to August 2017 (fig. 6). Median streambed temperatures during the mapping periods ranged from 17.5 to 21.9 °C. The highest median streambed temperature was during July 2016 and the lowest median streambed temperature was during August 2017.

Contour maps of streambed temperatures were constructed for the three sampling periods in 2016 (fig. 7) and the single sampling period in 2017 (fig. 8). The less than or equal to (≤) 25th percentile of streambed temperature was plotted on each contour explanation to provide a point of reference to identify areas with a higher potential for groundwater discharge through the streambed. The coldest streambed temperatures during the June 2016 mapping period were measured at the mouth of the side channel between DPs DP–2 and DP–9 (fig. 7A). Streambed temperatures in this area were between 13 and 14.5 °C. Streambed cobbles along the mainstream channel prevented streambed temperature measurements in a 50-m long section of the left bank of the main channel that contained DPs DP–1 and DP–2 (fig. 7A). June 2016 streambed temperatures within the area containing DPs WR3–L, WR–5, and WR3–R were mostly elevated relative to the 25th percentile; however, there were two areas containing colder (≤25th percentile) streambed temperatures along the left bank between DPs WR3–L and DP–1 (fig. 7A).

Cold (≤25th percentile) streambed temperatures during July 2016 were in narrow (<2-m width) areas along the left bank, north of DPs WR3–C and WR3–L (fig. 7B). Like the June 2016 results, cold (≤25th percentile) streambed temperatures were measured at the mouth of the side channel and extended southeast to DP–2 (fig. 7B). It is likely that cold streambed temperatures were present along the left bank between DPs DP–1 and DP–2; however, cobbles in this section of the channel prevented any measurements of streambed temperatures. The increased extent of the cold streambed temperature anomaly at the mouth of the side channel relative to June 2016 was likely attributable to a decrease in river stage. A third area of cold (≤25th percentile) streambed temperatures was measured in a section of ponded water about 30 m west of drive point DP–1 (fig. 7B). The coldest temperatures in this area were confined to an area along the left bank of the side channel.

August 2016 streambed temperatures were mapped in areas 1 and 2 along the study reach (figs. 7C and 7D). Decreasing water levels in the Little Wind River during the August 2016 mapping period resulted in two stream channels near DP WR3–C (fig. 7C). The coldest (≤25th percentile) streambed temperatures were along the left bank of the split channel immediately west of DP WR3–C and east of DP WR3–L. A similar pattern of cold (≤25th percentile) streambed temperatures exhibited by the July 2016 mapping results also was measured at the mouth of the side channel and persisted in a southerly trend to DP–2. Streambed temperatures were not measured in ponded water west of DP–1 (fig. 7C) during August 2016; however, it is likely that similar conditions resulted in a similar streambed temperature anomaly to what existed during July 2016.

Cobbles in some sections of the streambed in area 2 limited temperature mapping to two sections (fig. 7D). Like area 1, the coldest (≤25th percentile) streambed temperatures were generally along the left bank of the Little Wind River. The cold streambed temperatures near DP WR–10 were likely affected by preferential groundwater flow through an abandoned oxbow intersecting the active channel of the Little Wind River (fig. 7D).

Streambed temperature mapping in 2017 was limited to a single period during August and was conducted in areas 1 and 2 (figs. 8A and 8B). Parts of the active stream channel in the central part of area 1 shifted to the northeast between the 2016 and 2017 mapping periods (fig. 8A). Because of the shifting stream channel, the clustering of cold (≤25th percentile) streambed temperatures also shifted to the northeast. Like the 2016 results, the colder streambed temperatures were mostly limited to narrow (<5 m) areas along the left bank of the active channel where DPs WR17–2, 3, 4, 5, and 6 were installed (fig. 8A). The 2017 streambed temperature mapping area was extended upstream in the side channel in the southwest corner of area 1 (fig. 8A). Streambed temperatures in the southwest corner of area 1 were generally greater than or equal to 25th percentile of the streambed temperatures except for one small area near DP WR17–1 (fig. 8A). In contrast to the 2016 results, the mouth of the side channel in the north part of area 1 did not contain cold (≤25th percentile) streambed temperatures (fig. 8A). The cause for this is not known; however, continued erosion of the delta at the mouth of the side channel between the 2016 and 2017 mapping periods may have changed the movement of groundwater into the near-surface parts of the streambed. As noted previously, about 7 m of the side channel delta in area 1 was removed between May 2014 and April 2018 (fig. 5). The cold streambed temperatures that extended south along the left bank of the main channel to DP–2 during 2016 (figs. 7A–C) also were measured during the 2017 monitoring period and were between DPs WR17–6 and 7 (fig. 8A).

Changes in bed material along the left bank of the stream channel in area 2 between August 2016 and August 2017 prevented streambed temperature mapping in the exact same area. For reference, figure 8B compares the boundary of area 2 in 2016 (white) with the boundary of Area 2 in 2017 (yellow). No areas of cold (≤25th percentile) streambed temperatures were mapped in area 2 during August 2017 (fig. 8B). The absence of cold anomalies in the area 2 streambed sediments during 2017 likely was caused by the upstream shift in the survey area that did not capture groundwater input from the abandoned oxbow intersecting the active channel of the Little Wind River. Groundwater inflow from the abandoned oxbow channel was about 40 m east (downstream) of the area with the coldest streambed temperatures during 2017 in area 2 (fig. 8B). For comparison, a DP (WR17–10) was placed along the left bank of the active stream channel that contained the coldest streambed temperature measurements observed in area 2 during the 2017 mapping period (fig. 8B).

Both geophysical methods for sensing electrical conductivity (EMI and ERT) have advantages and disadvantages. EMI data collection is relatively straightforward. Because the EMI instrument does not require direct ground contact, data can be collected at walking speed, unlike ERT where electrodes and cables must be set up, and measurements may take several hours to complete. However, EMI generally cannot detect low conductivity targets. Further, the depth resolution of EMI is generally poor when compared to ERT. Both methods are subject to decreasing sensitivity with depth. For the configurations used in this study, we did not expect any sensitivity for either method beyond about 10 m below ground surface. Our hope was to use ERT to gather electrical conductivity information at relatively fine depth resolution and to extend these findings to spatially extensive EMI data.

The distribution of electrical conductivity inverted from EMI data is shown at 1-, 3-, and 5-m depths in figure 9. In most areas, the depth of investigation did not extend below 5 m, so data analysis was limited to the upper 5 m. High electrical conductivity values (>1,500 microsiemens per centimeter at 25 °C [µS/cm]) mainly were present in the northern area near the oxbow and toward the southern area, in the general area where the contaminant plume likely is present. At the 3-m depth, high electrical conductivity in both areas seems to migrate closer to the Little Wind River (fig. 9A). The southern high conductivity zone is still relatively extensive at the 3-m depth, whereas the northern zone is diminished in size (fig. 9B). At the 5-m depth, small, localized zones of high electrical conductivity are present near and below the Little Wind River (fig. 9C), generally seeming to line up with the path of high conductivity migration toward the river observed at the 1- and 3-m depths (figs. 9A and 9B). We interpret these results as groundwater flow transporting ions, either from evaporite deposits or related to the contaminant plume, into or below the Little Wind River.

ERT inversions are compared to EMI inversions in figure 10. ERT depth of investigation was limited to about 6 m in most areas, so results were limited to this depth. In general, results indicate a lower conductivity layer overlying a higher conductivity layer. These layers are interpreted as unsaturated shallow aquifer sand deposits (lower conductivity layer) overlying more-conductive saturated sediments (higher conductivity layer). EMT inversions from the 2-m spaced electrode roll-along style surveys (figs. 10F and 10G) seem reasonably consistent with the EMI inversions (figs. 10L and 10M) but overall indicate more vertical and horizontal variability. This variability is particularly apparent in the 1-m spaced lines (figs. 10C–10E), which show discrete zones of high electrical conductivity (low resistivity) surrounded by lower electrical conductivity (higher resistivity) and only are faintly comparable to the more layered (low conductivity over high conductivity) appearance of results obtained from EMI inversions (figs. 10I–10K). Substantial effort was put forth to determine if the cause of these anomalous results was a result of the data processing, including directly incorporating topography and geographic location into the inversion, stricter filtering of the data, and numerous inversion and data weighting options; however, results were typically similar in appearance. It is possible that these anomalies represent isolated lenses of high conductivity materials and (or) high salinity groundwater, and that the decreased resolution/sensitivity of the EMI data causes these anomalies to be smoothed out in space.

Temperature has long been used as an indicator of groundwater discharge throughout surface-water systems, as thoroughly reviewed by Constantz (2008). Spatial and temporal continuity in direct-contact environmental temperature measurements was improved when FO–DTS techniques were introduced to the hydrogeological community (for example, Selker and others, 2006), where temperature is measured over discrete sections along kilometer-scale fiber-optic cables. Raman-spectra backscatter FO–DTS techniques rely on the temperature-dependent backscatter of laser light pulses emitted along an optical fiber that return to the control unit within the anti-Stokes frequency range (Dakin and others, 1985). The most common application of FO–DTS in river systems is for armored cables that contain one or more optical fibers to be deployed directly along the sediment-water interface. Temperature data are collected over time by a control unit that is plugged into the cable, typically at a 0.25 to 1 m minimum linear spatial resolution along cable lengths of ≤2 km. Discrete or spatially “preferential” zones of groundwater discharge are then recognized by their characteristic effect on various temperature metrics, such as mean, minimum, and (or) standard deviation by cable location over time (Briggs and Hare, 2018). Temperature anomalies along FO–DTS cables deployed in flowing water of similar depth to the Little Wind River recently were determined to directly relate to preferential groundwater discharge points that showed on average five times the vertical discharge rate compared to other streambed locations only meters away (Rosenberry and others, 2016).

Mean temperature anomalies often are used to indicate preferential groundwater discharge zones over longer-term FO–DTS deployments (for example, Hare and others, 2015); however, the submerged sections of cable at the Little Wind River site had little spatial variation in this metric. As might be expected, locations along the shallow side channel and toward the distal end of the deployment where the cable was partially exposed to air had the largest standard deviation in interface temperature (fig. 11A). Downstream from the side channel, several discrete locations had slightly reduced standard deviation from the ambient interface condition that are paired with enhanced minimum temperature (figs. 11B and 11C). These locations are interpreted as discharge zones of relatively warm, shallow groundwater (about >14 °C), an interpretation supported by the point-in-time streambed temperature mapping (figs. 7 and 8). Shallow groundwater along these flowpaths may be expected in systems with low lateral hydraulic gradients, such as the Little Wind River corridor, and warm primarily through downward conduction of surface heat over the summer. Focused groundwater discharges that are directly intersected by the FO–DTS cable along the streambed interface buffer daily temperature swings and moderate nighttime cooling, creating the paired standard deviation and minimum temperature anomalies that were observed.

Figure 12 is a comparison between anomalously low streambed temperature standard deviations and anomalously high electrical conductivity at 1 m deep measured with EMI. Four zones of low temperature standard deviation (<2.28 °C) were identified (labeled a1 to a4 in fig. 12), which correspond closely with high electrical conductivity areas (>600 µS/cm). Extremely high electrical conductivity zones (1,000 µS/cm) are also observed in the corresponding areas of the ERT inversion results; zone a2 corresponds to the 0–20-m region of W1W2W3 (fig. 10F), and zone a1 corresponds to the 0–20-m region of W9 (fig. 10E). These areas are interpreted as zones of discharge to the Little Wind River.

Traditional seepage meters have been widely used to measure groundwater discharge in streams and rivers (Libelo and MacIntyre 1994; Landon and others, 2001; Rosenberry, 2008). Decades of use and testing have revealed that traditional seepage meters are subject to errors related to flow constriction, resistance to filling of collection bags, and pressure effects on exposed bags (Murdoch and Kelly, 2003; Rosenberry and others, 2008), although careful design, installation, measurement technique, and meter calibration (Rosenberry and Menheer, 2006) can provide reliable field results. In flowing surface-water settings, groundwater seepage measurements may be biased if the relatively large frontal area and rigid structure of traditional meters induce seepage via the Bernoulli effect (Cable and others, 2006; Rosenberry, 2008), if the large footprint causes substantial sediment disturbance (Rosenberry 2008; Rosenberry and others, 2010), or both.

TSMs (Solder and others, 2016) are a newer method for direct measurement of groundwater-surface water exchange, designed to reduce some of the errors associated with traditional seepage meters. Vertical seepage rates (q) were measured by TSM during four site visits along the study reach during 2016 and 2017 (figs. 13 and 14). TSM installation was limited by a pervasive cobble layer in the stream bed below more mobile sands and fine-grain materials. The cobble layer prevented TSM full depth penetration (about 1 m) at many sites and prevented any TSM installation in 2016 between sites TSM6 and TSM7R (fig. 14C). Two stainless-steel TSMs were installed in the cobble stream bed section in 2017 (TSM7 and TSM8) and polycarbonate tubes in sand deposits near the left bank (sites TSM9 and TSM13; fig. 14D).

Measured specific discharge, q, in 2016 represents the time-averaged q over an interval related to the temporal stability of the change in hydraulic head (ΔH), whereas the rates measured in 2017 are individual instantaneous measurements of q. Comparison of summarized q measurements for June, July, and August 2016, and August 2017 (figs. 13 and 14) is useful for characterizing broader hydrologic conditions across the study site and through time. The lowest median q was observed in 2017 (fig. 13). The 2016 mean q for all months is 0.45 m/d, ranging from −0.02 to 1.55 m/d with extreme outliers removed; and the 2017 mean q is 0.01 m/d, ranging from −0.04 to 0.05 m/d. Extreme outliers of q in 2016 data (June, WR3C=12.3 m/d, WR3R=10.3 m/d, and WR4=4.36 m/d; and August WR3C=3.90 m/d) are likely the result of increased uncertainty in estimated vertical hydraulic conductivity (K_v), in meters per day, associated with rapid falling water levels (>2.5 cm in 25 seconds), resulting in lower accuracy of timed head measurements at these sites during falling head tests. Spatially, q is consistently higher in 2016 and 2017 at the junctions between the ephemeral side channels and main channel at the upstream and downstream ends of the intensive study reach (fig. 14).

Comparison of TSM-measured mean q for all site visits and mean q from 2016 only to stream stage (USGS streamgage 06235500 [fig. 1]; USGS, 2019) indicated no statistically significant relationship. A statistically significant, albeit weak, negative correlation was determined between hourly measured q at site TSM13 (site with the longest continuous record in 2017) and stream stage (correlation coefficient=−0.27, statistical significance [p-value] <0.01).

Preliminary site-scale estimates of q in the area directly upgradient from the study area were calculated based on horizontal hydraulic head gradients and published estimates of K_v ranging from 45.7 to 54.9 m/d (Knowlton and others, 1997; DOE, 1998). Hydraulic head gradients were calculated from water levels measured in August of 2016 and 2017 at sites between Rendezvous Road and the Little Wind River (DOE, 2021) and averaged 7.3×10⁻⁴ and 6.7×10⁻⁴ meter per meter (m/m) in 2016 and 2017, respectively. Resulting q ranged from 0.11 to 0.13 m/d in 2016 and from 0.10 to 0.12 m/d in 2017.

Patterns of groundwater-surface water exchange are highly variable in space and over time (Boano and others, 2014), often needing methods that can characterize the temporal dynamics of seepage rates. In addition to direct physical approaches (for example, TSMs), vertical seepage dynamics at subdaily timescales can be inferred from vertical thermal sensor arrays installed within saturated streambed sediments. As reviewed thoroughly by Irvine and others (2017b), depth-specific streambed temperature measurements are typically collected with independent logging thermistors at known vertical spacing, which is often achieved by embedding thermistors within a solid rod or pipe. The vertical advection of water (seepage) affects the thermal gradient between shallow and deeper sediments, including the downward propagation of natural surface temperature diurnal signals (Constantz, 2008). The effects of seepage on depth-specific temperature time series are predictable based on the governing 1D advection-conduction model if the thermal properties of the streambed are estimated or measured (Lapham, 1989).

Vertical thermal profiles were collected in saturated streambed sediments during periods from June 29 to September 15, 2016, or from August 23 to August 31, 2017. In 2016, the sensor arrays were relocated to various streambed locations two to three times over the period of deployment, whereas in 2017 the sensor arrays were left in place during the shorter period of deployment (fig. 15). In general, 2016 sensor array deployment locations were more broadly spaced and guided by existing ancillary data and information, whereas in 2017, the profilers were installed in a relatively small area of suspected paleo-channel groundwater discharge to the Little Wind River. As the downward propagation of surface temperature signals is highly attenuated in groundwater discharge zones (target of this study), the custom temperature sensor arrays were designed to collect high spatial resolution data just below the sediment-water interface.

Visual inspection of the raw 2016 streambed temperature sensor array data indicated poor coupling and (or) strong hyporheic downwelling at several locations over the varied deployment periods such that many locations showed little difference in temperature dynamics with depth. When there is effectively no vertical gradient in streambed temperature for a given period, there is little information that can be extracted for seepage flux modeling. Therefore, quantitative flux modeling using data collected in 2016 was only possible for three streambed locations (table 1). Saturated streambed thermal diffusivity as derived from the vertical transport of natural diurnal temperature signals ranged from 0.03 to 0.06 square meter per day (m²/d) (geometric mean) for the 2016 deployments where strong coupling between the streambed and the profiler was indicated. This range in thermal diffusivity values is generally consistent with other reported values for fine-grained river sediments (Rau and others, 2012), and these values were used for the diurnal signal amplitude-ratio base modeling of vertical seepage rates. A constant streambed porosity of 0.35 was assumed for all deployment locations. Modeled flux rates were modest to circum-neutral and generally ranged from 0.2 to −0.2 m/d (positive sign indicates upward flow, or upwelling), depending on location (table 1). The most consistent modeled record of upward groundwater discharge is in the general area where the 2017 temperature sensor arrays were concentrated.

In contrast to 2016, the 2017 streambed temperature sensor array data indicated universally strong coupling with streambed sediments except at site 6 where quantitative seepage modeling was not attempted. A thermal diffusivity of 0.6 m²/d was assumed for all vertical flux models based on the 2016 data and typical reported thermal properties of riverbed sediments, and a porosity of 0.35 was assumed. The modeling results for the 10 remaining streambed sites are listed in table 1. Five streambed sites indicated circum-neutral vertical seepage flux conditions where ±1 standard deviation of subdaily modeled flux values included 0. The other five streambed locations showed mean upwelling values where 0 was not included within 1 standard deviation, and these mean values ranged 0.11 to 0.23 m/d (table 1). Temporal variance was generally small (0.01 to 0.02 m/d standard deviation), except for site 7, which showed large changes in vertical seepage rate over the deployment period.

Radon-222 (²²²Rn) is a soluble, colorless, gaseous, inert, unstable isotope produced by the decay of radium-226 (²²⁶Ra) contained in uranium-bearing subsurface materials. With a half-life of 3.8 days, radon-222 reaches secular equilibrium with radium-226 activity in the subsurface materials within 2–3 weeks (Cook, 2013). Water in equilibrium with the atmosphere has no radon-222 concentration, and the presence of ²²²Rn in surface waters is indicative of subsurface interaction. Radon activities in groundwater are typically two to three orders of magnitude higher than those in surface waters; therefore, points of elevated ²²²Rn in streams indicate zones of groundwater inflow (Yu and others, 2013).

Figure 16 shows the groundwater and surface-water sampling locations where ²²²Rn samples were collected within the study area. Specific conductance ranged from 306.2 to 18,511 μS/cm, with a mean of 8,002 μS/cm (fig. 17A). Concentrations of ²²²Rn expressed as activity per volume in samples collected from wells and drive points between 2016 and 2017 ranged from 87.9 to 1,583 picocuries per liter (pCi/L), with a mean concentration of 456 pCi/L (fig. 17B). Concentrations of ²²²Rn in water samples collected from the Little Wind River during 2016 ranged from 10.9 to 19.7 pCi/L in June; 8.9 to 25.7 pCi/L in July; and 14.8 to 25.7 pCi/L in August. During 2016, the average stream concentrations of ²²²Rn increased each month and were 15.1 (June), 18.7 (July), and 19.6 (August) pCi/L (fig. 18). Instream ²²²Rn concentrations measured during August 2017 were substantially lower, ranging from 5.4 to 9.8 pCi/L with a mean of 7.9 pCi/L (fig. 18). Instream specific conductance values ranged from 479 to 492 μS/cm in June 2016; 1,029 to 1,056 μS/cm in July 2016; and 1,047 to 1,067 μS/cm in August 2016 (fig. 19). Specific conductance values increased in 2016 between each sampling period, reflecting decreases in stream stage (fig. 19). Specific conductance ranged from 700 to 722 μS/cm during the sampling period in August 2017 (fig. 19).

The radon-222 data were used to estimate the groundwater inflow along the study reach (fig. 16) during the four sampling periods in 2016 and 2017 using a method described in Goble (2018). Results of the groundwater discharge modeling indicated an evolution in inflow patterns with fluctuations in river stage and flow conditions (fig. 20). Estimated inflow in June 2016 was large at the upper (0–0.4 km) and central (0.8–1.2 km) parts of the reach (fig. 20B) with a total stream contribution of 19,016 cubic meters per day (m³/d). By July, the inflow was limited to the central (0.8–1.0 km) and the lower (1.2–2.0 km) parts of the reach (fig. 20D) with a cumulative discharge of 14,218 m³/d. Two weeks later in August 2016, the inflow was again confined to the upper central (0.4–1.0 km) and lowermost parts (1.7–2.0 km) of the reach (fig. 20F) with a cumulative discharge of 11,858 m³/d. In August 2017, however, low river radon-222 concentrations resulted in substantially less estimated inflow along the reach. There was an initial pulse of water at the uppermost part (0–0.4 km) of the reach (fig. 20H) contributing 9,536 m³/d to the Little Wind River.

During 2016, consistent inflow of groundwater was observed in the central part of the study reach, between 0.7 and 1.7 km downstream (figs. 20B, D, and F), which was congruous with the center of the previously mapped groundwater plume discharge zone. Cumulative declines in groundwater inflow during June through August 2016 were determined. These declines in groundwater discharge were concurrent with declines in the stream stage (fig. 4) and the associated reduction in aquifer storage. The groundwater contribution along the reach as a percentage of streamflow was around 1 percent in June 2016, increasing to about 6 percent during the baseflow periods in July and August 2016 (fig. 21). Groundwater contribution during August 2017 decreased to about 1 percent of streamflow, indicating higher streamflow conditions relative to August 2016 (fig. 4) and that typical late season baseflow conditions had not been attained.

In June 2016, dissolved uranium concentrations ranged from 2 to 1,260 micrograms per liter (µg/L) in 1-m DPs sampled (fig. 7A), and the highest concentration was measured in the DP–2 sample (fig. 22). Molybdenum concentrations ranged from 1 to 660 µg/L, and the highest concentration was also measured in DP–2. Uranium and molybdenum concentrations were as much as 450 and 650 times greater than surface water, respectively. DP–2 was downgradient from the contaminant plume and had elevated specific conductance, major ion, and uranium and molybdenum concentrations characteristic of the legacy contaminant plume (DOE, 2016).

In July 2016, uranium concentrations at most 1-m DPs installed in June had increased three to eight times, as did specific conductance, likely in response to a decrease in river stage. Molybdenum concentrations increased by as much as a factor of five times between June and July. In contrast, uranium and molybdenum concentrations in DP–2 did not change. DP WR–7, installed in July (fig. 7B), had elevated uranium, molybdenum, and major ion concentrations consistent with the legacy groundwater contaminant plume (fig. 22). The other DPs had elevated concentrations of uranium and molybdenum ranging from 8 to 180 times surface water, along with elevated major ion concentrations, defining the location of the legacy plume in the shallow streambed of the left bank.

Concentrations in 1-m DPs in August 2016 were essentially unchanged from July (fig. 22). The uranium, molybdenum, and major ion chemistry data indicate that the plume was between sites WR3–L and WR–9, and the highest concentration of contaminants was measured between DP–2 and WR–9. Extremely low concentrations of uranium, molybdenum, and other constituents that are indicative of the contaminant plume were measured in DP WR–10 in area 2 next to the abandoned oxbow (figs. 7 and 22).

The MP array was deployed and sampled in tandem with 30- and 50-cm DPs at five sites in August 2016 within 1 m of the left bank of the river within the zone of contaminant plume (fig. 7). Elevated uranium and molybdenum concentrations were observed at 50 cm at all sites consistent with adjacent 1-m DPs (fig. 22). Uranium and molybdenum decreased toward the sediment-water interface. The decreases at sites MP–4 and MP–5 are largely the result of near complete hyporheic exchange to 12 cm evident from similar major ion chemistry to the overlying stream water and the calculated fraction of surface water (F_sw) of 1. F_sw was 0.43 and 0.56 at 30 cm at sites MP–4 and MP–5, respectively. No reactive loss of uranium or molybdenum was calculated at any depth at these sites (eq. 7). Combined, these results suggest little or no flux of contaminants to the river at sites MP–4 and MP–5 because of the strong hyporheic exchange of surface water into the streambed. At WR–3, hyporheic exchange extended to 30 cm, where the F_sw was 0.78, and an F_sw of 1 was calculated between 3 and 15 cm. Little or no reactive loss of uranium or molybdenum was calculated during hyporheic mixing of surface water with the contaminant plume at MP–3. In contrast, much less hyporheic exchange was observed at sites MP–1 and MP–2, where an F_sw of 0.07 to 0.12 was observed between 3 and 15 cm. These sites were at the edge of a small, shallow embayment at the mouth of a dry meander channel. Little or no surface flow was evident at these sites, suggesting low hydrodynamic forcing of hyporheic exchange (Boano and others 2014) consistent with shallow penetration of surface water in the streambed and greater groundwater upwelling. At MP–1, reactive losses of 15 to 19 percent of groundwater uranium and 22 to 25 percent of molybdenum were calculated between 3 and 15 cm. Reactive loss at MP–2 was negligible (5 percent or less) for both constituents.

In August 2017, groundwater uranium concentrations at 100 cm below the streambed increased downstream from sites WR17–9 to WR17–6, and uranium concentrations were similar at sites WR17–6 to WR17–8 (fig. 23). This trend in uranium paralleled major ion chemistry and is indicative of the presence of the legacy contaminant plume (Naftz and others, 2019). DP molybdenum concentrations also increased downstream along this transect, but the highest concentrations were measured at the farthest downstream site, WR17–8, of area 1. Uranium and molybdenum concentrations exceeded 2,000 and 700 µg/L, respectively, and were within the range of the highest concentrations measured in the legacy plume in monitoring wells installed in the adjacent floodplain and dry channel in 2015 (DOE, 2016). Uranium, molybdenum, and major ion concentrations at site WR17–1 in the dry side channel to the southwest were substantially lower, indicating a negligible component of the legacy plume. Low concentrations also were measured in DPs at site WR17–10 (fig. 23), consistent with August 2016 data that indicated the absence of the plume at this site (fig. 8, area 2).

Pore-water concentrations of uranium, molybdenum, and major ions are relatively constant between 30 and 100 cm below the sediment water interface at the sites intersecting the plume, although decreases in uranium and molybdenum were observed at the intermediate depths at site WR17–7 (figs. 23A and 23B; Naftz and others, 2019). Lower pore-water concentrations of uranium, molybdenum, and major ions were observed at depths shallower than 30 cm. The extent of decrease varied among sites. For example, uranium concentrations in MP samples between 3 and 15 cm approached surface-water concentrations at sites WR17–9, 2, 3, 7, and 8, whereas little or no decrease was observed approaching the stream interface at sites WR17–4, 5, and 6. Similar trends with depth were observed for molybdenum (fig. 23B). pH increased at shallower depths and approached the surface-water pH. These differences in pore-water uranium and molybdenum profiles among sites can be explained in part by dilution of groundwater by mixing with surface water in the hyporheic zone. The extent of hyporheic exchange of surface water is illustrated by the contour plot of F_sw versus depth (fig. 23C). The trends in the fraction of surface water with depth generally mirror the concentration contour plots for uranium and molybdenum (figs. 23A and23B), indicating the decrease in the solutes is, in part, because of dilution by surface water. The hyporheic zone, defined by 10 percent or more surface water, ranged in depth from 3 to 70 cm among these sites, and the deepest penetration of surface water was observed at site WR17–9. The extent of hyporheic exchange varied inversely with grain size. For example, the sediments at sites WR17–4, 5, and 6 were finer grained with an average of 45 percent <125 µm by mass (Naftz and others, 2019). In contrast, sediments averaged 10 percent <125 µm by mass at the other sites that had deeper and more extensive hyporheic exchange.

Dilution of uranium and molybdenum by surface water in the hyporheic zone was estimated by calculating their concentrations in the absence of reactive loss or gain using equation 7. The 1-m DP and the surface-water concentrations at each site were used for end-member components in the nonreactive mixing model. The difference between the calculated nonreactive concentration and the measured concentration of uranium and molybdenum is defined as reactive loss or hyporheic mixing of surface water and groundwater in the streambed. The contour plot of uranium reactive loss shows greatest attenuation of uranium at sites WR17–7, 8, and 9 between 6 and 15 cm, and as much as 90 percent uranium attenuation from the upwelling groundwater (note that uranium loss at WR17–9 is not evident in fig. 24A because absolute uranium loss is plotted). Intermediate levels of uranium attenuation (30 to 50 percent) were estimated at sites WR17–3 and WR17–4. A lower extent of uranium attenuation was calculated for sites WR17–5 and WR17–6 (10 to 30 percent). Similar trends in molybdenum attenuation extent were observed except that greater loss (40 to 50 percent) compared to uranium was calculated at sites WR17–5 and WR17–6 (fig. 24C).

Contour plots of sediment uranium and molybdenum total concentrations in the <1-mm fraction of streambed sediment cores show elevated concentrations for sites spanning the legacy contaminant plume relative to outside of the plume (figs. 24B and 24D). Uranium and molybdenum at sites WR17–9 and WR17–2 vary little with depth and are within measurement uncertainty of sediment concentrations of uranium and molybdenum from site WR17–1 upstream from the legacy plume. The average of all intervals at these three sites (WR17–1, 2, and 9) was used to define uranium and molybdenum background concentrations of 1.0 and 0.16 microgram per gram (µg/g), respectively. The higher uranium concentrations ranging from 1.9 to 2.3 µg/g measured at site WR17–10 (Naftz and others, 2019), downstream from the legacy contaminant plume (fig. 8), are likely because of a larger fraction of fine-grained sediments. Sediment uranium concentrations at sites within the legacy contaminant plume were 2–10 times higher than background with the exception of sites WR17–3 and WR17–8 below 3 cm, where little or no enrichment was observed. The highest uranium concentration was observed at sites WR17–5 and WR17–6. Molybdenum concentrations at sites WR17–3 through WR17–8 ranged from two to nine times above background. The highest molybdenum sediment concentrations were observed at WR17–5 and WR17–6. The variation in sediment uranium and molybdenum resulting from differences in grain-size distribution among core intervals may account for some of the observed differences among sites. However, the proportional increase in uranium and molybdenum with increases in the <63 and <125 µm grain-size fractions for background sites does not explain the extent of the variability observed in plume site sediments. For example, uranium increases linearly from 0.9 to 2.3 μg/g with increase of the <63 µm fraction from 1.2 to 45 percent at background sites (WR17–1, 2, 9, and 10). This range in the <63 µm fraction encompasses the grain size distribution in sediments from plume sites (Naftz and others, 2019).

Overall, higher sediment concentrations and higher calculated aqueous uranium and molybdenum reactive loss were observed in the zone where the legacy plume is emerging into the river. However, no spatial correlations are evident between sediment uranium or molybdenum and calculated reactive loss from groundwater in the streambed (fig. 24). Instead, the spatial differences are attributed to the time scales of the sediments and the pore-water profile measurements. The pore-water profiles represent conditions over a short period (<2 hours), whereas the core samples represent accumulation of uranium and molybdenum onto the bed sediments since sediment deposition, which is likely on the order of a few months or longer. Nonetheless, pore-water reactive loss estimates and sediment concentration profiles are consistent with attenuation of uranium and molybdenum from groundwater during hyporheic mixing of surface water with the legacy plume during groundwater upwelling into the river.

The processes resulting in uranium and molybdenum attenuation were not investigated in this study but are topics for ongoing investigation. Uranium removal may result, in part, by sorption onto iron oxyhydroxides forming in the streambed during hyporheic mixing. Dissolved iron concentrations in the legacy plume groundwater defined by the 1-m DPs ranged from 100 to >10,000 µg/L. At shallower depths at some sites, measured iron concentrations were lower than predicted by conservative mixing with surface water. The loss of pore-water iron is attributed to iron oxidation and precipitation in response to increased dissolved-oxygen concentrations in the hyporheic zone. Total uranium and iron sediment concentrations were not correlated, in part, because the amount of iron-bearing minerals dissolved by the strong acid digestion used for sediment analyses is likely much higher than iron precipitated from groundwater. A weaker chemical extraction would be needed to measure the iron precipitated from groundwater on sediments to allow better comparison with sediment uranium concentrations. Significant correlations between sediment molybdenum and total iron (p<0.005, Pearson’s correlation coefficient=0.687, n=56) or total sulfur (p<0.005, Pearson’s correlation coefficient=0.746, n=56) suggest molybdenum attenuation may result from sorption to iron and (or) sulfur phases, although covariation in these constituents does not confirm an attenuation process.

Contour plots of DET uranium and molybdenum pore-water concentrations in the shallow sediments (top 15 cm) along the Little Wind River study reach during August 2017 are shown in figures 25A and 25B. Site WR17–1 (above the groundwater plume) and site WR17–10 (below the groundwater plume) are not shown; however, mean concentrations of uranium (WR17–10, 11 µg/L; WR17–1, 22 µg/L) and molybdenum (WR17–10, 7 µg/L; WR17–1, 4 µg/L) were low, and both sites did not seem to be affected by contaminated groundwater, which was consistent with DP and MP data. The DET probes deployed at sites WR17–2 to WR17–9 clearly identified the location and focus of contaminated groundwater in the shallow sediments of the Little Wind River. The highest uranium and molybdenum concentrations seem to be near sites WR17–5 and WR17–6 (uranium) and WR17–7 (molybdenum). Mean concentrations within the sampling zone were 403 µg/L for uranium (ranged from 4 to 1,320 µg/L) and 60 µg/L for molybdenum (ranged from 1 to 237 µg/L). Between sites WR17–4 and WR17–7, uranium and molybdenum concentrations decreased towards the surface in the upper 3 to 6 cm.

Figure 25C is a contour plot of strontium pore-water concentrations in 2017 from the DET samplers. In this system, strontium is considered a conservative solute, so any decrease in concentration in the sediment pore waters might be related to mixing with surface waters that would contain lower strontium concentrations. Unlike uranium concentrations, elevated strontium pore-water concentrations at sites WR17–5, WR17–6, and WR17–7 (range 4.6–7.4 milligrams per liter [mg/L]) persisted closer to the surface (surface water=1.1 mg/L), but concentrations decreased at site WR17–7 (to 2.8 mg/L) (fig. 25C). These slightly different patterns in reactive and conservative solute concentrations indicate dilution by infiltrating low uranium concentration surface water (table 2) and reactive uptake of uranium in groundwater by sediments during hyporheic mixing, which may account for the observed decrease in uranium concentrations.

DGT pore-water concentrations and fluxes (into DGT probe) for uranium in August 2017 are shown in figure 26. Uranium concentrations determined using DGT measurements were substantially lower than concentrations determined using DET measurements. The DGT mean uranium concentration was 15 µg/L (range 1–121 µg/L), and DGT molybdenum concentrations were generally below detection limits. It is not uncommon for bulk pore-water concentrations derived from DET or DP samplers to be higher than DGT concentrations. Typically, there are three possible explanations for this difference in concentrations. First, re-supply of solutes from sediments or pore waters to the DGT is not sufficient to keep pace with the demand from the DGT (Lehto, 2016). Second, the solutes of interest may not exist in bioavailable phases and are instead partially labile or inert to DGT (Degryse and Smolders, 2016). Complexes with large dissolved organic ligands often have small diffusion coefficients and can be less labile than dissolved inorganic complexes. Third, other solutes may outcompete uranium and molybdenum for binding sites on the DGT gel (Turner and others, 2012).

The capability of the sediment to re-supply solute to the DGT can be determined by calculating the sediment reactivity coefficient (Rc), the ratio of the DGT concentration to the bulk pore water solute concentration (from DET measurements) (fig. 27) (Lehto, 2016). Values of Rc above 0.8 would indicate a well-buffered system with continuous supply from the solid and solution phases at a rate almost equal to the flux into the DGT. Values of Rc from 0.2 to 0.8 would indicate some supply from the solid phase that is insufficient to satisfy the demand from the DGT. Most DGT measurements are <0.2, suggesting a poorly buffered system where diffusion is the major process supplying solute to the DGT. This also may suggest that re-supply of pore water solute through advective transport via groundwater does not occur at a rate fast enough to satisfy demand from the DGT. Isolated pockets of elevated DGT uranium concentrations and fluxes (figs. 26A and 26B) may indicate regions of the sediment with more reactive sediment, faster re-supply rates by advective transport, or both. However, competition for binding sites on the DGT, and (or) the presence of DGT-inert or partially labile uranium and molybdenum species, also could contribute to the low observed Rc values.

Comparison of DET (fig. 25) and MP (fig. 23) uranium and molybdenum concentrations from shallow sediments (top 15 cm) indicates good general agreement in the overall trend. That is, concentrations generally decreased towards the surface of the sediment, supporting the idea of solute attenuation in the shallow sediments. However, there is some disagreement in the actual solute concentrations measured by each method. This disagreement is possibly explained by the finer-resolution sampling of the DET device and the small sampling volume compared to the MP system, which results in minimal disturbance to the pore-water concentration and the chemical gradient. At some sites (WR17–5 and WR17–6), the DET data seem to indicate a sharper attenuation gradient than the MP data centered around 10 cm deep.

Bank operated cableways were established at three locations on the Little Wind River during the 2016 and 2017 field seasons to facilitate streamflow measurements using an ADCP. Bank operated cableway transects were chosen to bracket the area where the contaminant plume enters the Little Wind River; the upstream transect (RIV–US) was upstream from the contaminant plume, the midway transect (RIV–MID) was immediately downstream from the highest plume concentrations, and the downstream transect (RIV–DS) was downstream from the plume (fig. 28). ADCP streamflow measurements were conducted on August 9, 2016, and August 24–25, 2017 (Naftz and others, 2019).

Streamflow during 2016 was below average. ADCP measurements at the field site (about 85 cubic feet per second [ft³/s]) were about 40 percent of median values for USGS streamgage on the Little Wind River (Little Wind River near Riverton, Wyo.) with 77 years of record and located about 2 km below the study reach (USGS, 2019). In contrast, streamflow during 2017 was above average. ADCP measurements (about 380 ft³/s) were about 225 percent of median values. Changes in streamflow between the upstream and downstream transects were variable between years and within the same year (2017). Streamflow at the downstream transect was about 4 percent less than at the upstream transect (about 81 and 85 ft³/s, respectively) on August 9, 2016, and about 0.3 percent less on August 24, 2017 (381 and 382 ft³/s, respectively). These small streamflow losses are likely an artifact of measurement error given their small magnitude, known errors in ADCP measurements, and the fact that the upstream and downstream transects bracket the contaminant plume. In addition, streamflow measurements at the transects were made sequentially (as opposed to simultaneously), and additional errors caused by temporal variability are possible. Streamflow at the downstream transect on August 25, 2017, was about 7 percent greater than streamflow at the upstream transect (390 and 364 ft³/s, respectively), consistent with the entry of the contaminant plume and the increase in constituent concentrations described below.

Streamflow data were used to define three EDIs across each transect, with the left (L), center (C), and right (R) increments each representing 1/3 of the transect streamflow (USGS, 2012). Water-quality samples were collected at the midpoint of each increment on August 11, 2016, and on August 27, 2017, and were analyzed for anions, cations, alkalinity, and dissolved organic carbon (Naftz and others, 2019). The EDI sampling scheme was designed to detect the longitudinal changes in concentration as the Little Wind River passes by the contaminant plume and changes in the transverse direction (perpendicular to streamflow) at each transect. Overall longitudinal changes in concentration are assessed by averaging concentrations from the L, C, and R increments at each transect; changes in the transverse direction are assessed by plotting the L, C, and R concentrations for each transect. Under this sampling scheme, concentrations at the upstream transect are expected to be uniform in the transverse direction (L, C, and R concentrations being comparable) because of the transect being upstream from the contaminant plume. Concentrations for the L increment at the midway and downstream transects are expected to exceed those of the C and R increments as the plume enters along the left bank of the river.

Average streamflow values at the upstream and downstream transects in 2017 indicate a 2.9-percent increase in streamflow along the study reach, and equation 6 can be used to calculate the net load and average concentration. In contrast, average streamflow values from 2016 suggest a small decrease in streamflow along the study reach, an observation that is at odds with the observed increase in stream uranium concentrations. The small decrease is likely attributable to measurement error. As such, net loading calculations for 2016 assume a streamflow increase that is comparable to that observed in 2017 (average streamflow at the downstream transect is set to be 2.9 percent higher than at the upstream transect).

Instream concentrations for all constituents were substantially lower in 2017 compared to 2016 because of the dilution effect of above average streamflow (table 2). Longitudinal and transverse trends in dissolved uranium concentrations were consistent between both years, despite large differences in hydrologic conditions (2016 streamflow was substantially below average, and 2017 was substantially above). Dissolved uranium concentrations increased in the downstream direction in both years; average concentrations at the downstream transect exceeded those of the midway transect, and concentrations at the midway transect exceeded those at the upstream transect (average, fig. 29). The relative percent difference for sequential replicates that were analyzed for dissolved uranium ranged from 0.7 to 3.8 percent (Naftz and others, 2019). In addition, dissolved uranium concentrations for the L increment at the midway and downstream segments were elevated relative to the upstream segment in both years (left, center, and right, fig. 29). Both of these observations are indicative of the addition of water and solute mass to the Little Wind River from the contaminant plume, a conclusion that is at odds with the measured streamflow losses on August 9, 2016, and August 24, 2017, but consistent with the streamflow gain observed on August 25, 2017. As noted above, the streamflow losses are likely an artifact of measurement error, and the August 25, 2017, streamflow gain is a more plausible scenario. Overall, the streamflow data suggest relatively small additions of plume water can cause detectable changes in instream concentration, a situation that arises because groundwater plume concentrations are orders of magnitude higher than surface-water concentrations at the upstream transect.

Mass balance calculations (see the “Data Processing and Modeling” section) indicate that the net uranium load entering the stream during the 2016 field effort is about 290 grams per day (g/d), with an average inflow concentration of 47 μg/L. Mass balance results for 2017 indicate a net load of approximately 435 g/d, with an average inflow concentration of 16 μg/L. The higher loading rate observed in 2017 might be attributable to a higher groundwater table and increased groundwater inflow associated with wetter conditions. Similarly, the lower average inflow concentration observed in 2017 may be due to increased dilution of plume waters. Average inflow concentrations for both years are well below observed concentrations within the groundwater plume, an observation that might be explained by the fact that the mass-balance calculations include all of the water entering between the upstream and downstream transects, rather than strictly plume waters. In addition, uranium concentrations might be attenuated as waters leave the groundwater system through the hyporheic zone and cross the groundwater/surface-water interface, as inferred from the modeling of pore-water chemistry profiles (see previous discussion in the “Drive Points and Minipiezometers” section).

Other constituents that illustrate the effect of the plume on the Little Wind River include chloride, molybdenum, and sulfate. All three of these constituents generally exhibit an upstream to downstream increase in both years, and the addition of plume water is evident from small increases in the L increment concentration at the midway and downstream transects (table 2). The remaining constituents do not exhibit these two trends in a consistent manner, as they lack increases from upstream to downstream and from right to left (calcium, cobalt, copper, fluoride, potassium, magnesium, manganese, sodium, silica, and strontium, table 2).

Longitudinal sampling was performed at multiple sites spanning the contaminated groundwater plume in 2016 and 2017 to identify its location and chemical and biological effect (fig. 30). The samples collected during 2016 included streambed sediment, a diversity of benthic macroinvertebrates (larval aquatic insects), and filamentous green algae. Based on results from 2016, the study design was revised in 2017. Streambed sediments were again sampled, but macroinvertebrates and filamentous algae were not. Instead, a series of samplers were deployed to assess uranium uptake by periphyton and effects of aqueous uranium on primary productivity. Details on sampling and analytical methods are provided by Naftz and others (2019).

In 2016, fine sediment (<63 µm) was collected from the streambed at eight sites spanning regions upstream and downstream from the contaminated groundwater plume (fig. 31). Site 1 was a reference site about 3 km upstream from the plume (not shown on fig. 30). Other sites were located using stream water temperature data for guidance (fig. 7) and are shown on figure 30. Site 2 was in a side channel on the left side of the main channel, and site 3 was in the main channel, both considered to be upstream from the groundwater plume. Sites 4 and 6 were near the upstream perimeter of the plume, sites 7 and 11 were within the plume, and site 14 was downstream from the plume. Samples were collected along the left bank at all sites and additionally along the right bank at sites 1, 6, 11, and 14.

Mean concentrations of uranium and molybdenum in sediment samples collected in 2016 along the left bank ranged from 2.02 to 2.72 milligrams per kilogram (mg/kg) and from 0.21 to 0.37 mg/kg, respectively (fig. 31). On the right bank, uranium and molybdenum concentrations were less variable, ranging from 1.98 to 2.19 mg/kg uranium and from 0.20 to 0.27 mg/kg molybdenum. The effect of the contaminant plume was evident in uranium and molybdenum concentrations at sites 11 and 14. Sediment collected along the left bank exhibited significant enrichment in uranium at both sites (analysis of variance, p-value<0.001; Tukey’s Honest Significant Difference test). The molybdenum concentration at site 11 also was elevated along the left bank, whereas the concentration at site 14 was not. Uranium and molybdenum concentrations among samples collected from the right bank were comparable except at site 14 where uranium concentrations were slightly greater (analysis of variance, p-value<0.01). Other elements varied modestly among sites with no change evident at or near sites 11 and 14 (Naftz and others 2019). Neither uranium nor molybdenum correlated spatially with aluminum, iron, or manganese, suggesting absence of a natural lithogenic source of enrichment.

Samples of streambed sediments were collected from 10 locations in 2017. Five of these were co-located with periphyton samplers at sites 3 and 12–15 (fig. 30). The remaining five sediment samples were collected between site 3 and 12, within the area where the groundwater plume contacted the stream channel. Comparison of left and right bank samples collected in 2016 indicated that uranium concentrations in streambed sediment were generally highest along the left bank; accordingly, sediment samples collected in 2017 were limited to left bank regions of the stream channel.

Mean uranium and molybdenum concentrations ranged from 1.02 to 4.06 and from 0.26 to 2.03 mg/kg, respectively. In 2016, concentrations of both elements were more variable, and maximum concentrations were greater. As shown in figure 32, concentrations increased downstream from site 3, reaching their maximum concentration at site 10. Iron and arsenic concentrations also were elevated at site 10 (Naftz and others 2019) where a surface layer of iron floc was visible, indicating a region of groundwater input. The formation of ferric oxides likely would provide sorption sites for soluble hexavalent uranium. The groundwater signal was highly localized because concentrations of all elements at site 11 were comparable to those at site 1, and uranium and molybdenum concentrations were roughly one-half those measured at the same location in 2016. Further downstream, uranium and molybdenum concentrations trended downward, although with considerable among-site variation that possibly was related to transport and heterogeneous mixing of sediment.

Sediment quality guidelines directly applicable to the Little Wind River are not available. To place the data into context, sediment concentration thresholds were used and derived using lowest effect levels and severe effect levels from the co-variance of benthic invertebrate community structure and contaminant concentrations of sediments from uranium mining and milling sites in Canada (Thompson and others, 2005). The lowest effect level, the uranium concentration at which abundance and species richness of invertebrate communities would be reduced by <20 percent, was estimated to be 104 micrograms per gram based on a weighted method, which is more than an order of magnitude greater than the highest concentrations observed in the Little Wind River bed sediment data.

Samples of attached filamentous green alga Spirogyra spp. were collected during August 9–11, 2016, from natural streambed substrates at select locations above and below where the legacy groundwater plume intercepted the left bank of the Little Wind River (fig. 30; sites 2–6, 11, and 14). To ensure that exposure periods were relatively consistent among sampling locations, care was taken to select relatively undisturbed algal clusters exhibiting new growth with minimum evidence of senescence. The farthest upstream site (site 1) was not included as part of these analyses because of the paucity of new-growth algae at that location.

There was no significant difference (p-value>0.30; paired t-test, Sokal and Rohlf, 2003) in algal uranium or molybdenum concentrations between left and right bank samples (collected at sites 5, 6, 11, and 14), and results presented here are limited to samples collected along the left bank. Algal tissue concentrations of uranium and molybdenum were generally low among left-bank sites and ranged from 0.45 to 2.10 µg/g dry weight and 0.55 to 0.98 µg/g dry weight, respectively (fig. 33). Mean tissue concentrations and corresponding confidence intervals (CIs; ±95 percent CI) upstream of the contaminant plume (sites 2–6) were compared to concentrations downstream at sites 11 and 14. Confidence intervals represent ranges of expected tissue concentrations for evaluating downstream effects. Tissue concentrations that are within the range of expected values for uranium or molybdenum (±95 percent CI) are considered unaltered. Comparison among sites indicate that concentrations of uranium and molybdenum in Spirogyra spp. downstream from the groundwater contaminant plume (sites 11 and 14) were within the expected range of values for sites upstream of where the groundwater plume entered the river (fig. 33). Increased bioaccumulation of dissolved uranium or molybdenum at locations downstream from the groundwater plume was not indicated at the time of sampling.

Immature larval and nymph stages of aquatic invertebrates were collected from natural streambed substrates at six locations in the Little Wind River in August 2016. Four of the sampling locations were located upstream (sites 1–4) of where the groundwater plume entered the river and at two sites (11 and 14) downstream of the plume location (fig. 30). A total of eight invertebrate taxa were collected for tissue analysis and included six Ephemeroptera (mayfly) genera (Choroterpes [Eaton, 1881], Ephoron [Williamson, 1802], Heptagenia [Walsh, 1863], Traverella [Edmunds, 1948], Tricorythodes [Ulmer, 1920], and Baetis [Leach, 1815]), one Odonata (dragonfly) genera (Ophiogomphus [Selys, 1854]), and one Trichoptera (caddisfly) genera (Hydropsyche [Pictet, 1834]). Because of variability in stream habitat quality and availability among sites, not all taxa were present at all locations or in sufficient numbers for processing. Accordingly, among-site comparisons of tissue metal concentrations presented here are limited to the three most common taxa (Choroterpes, Ephoron, and Heptagenia). Tissue metal concentrations for all invertebrate taxa are reported in Naftz and others (2019).

To determine whether concentrations of bioavailable uranium and molybdenum increased near and downstream from where the groundwater contaminant plume entered the Little Wind River, whole-body tissue concentrations in the mayfly nymphs Choroterpes, Ephoron, and Heptagenia were compared at sampling locations upstream and downstream from the groundwater plume. Overall tissue concentrations of uranium and molybdenum tended to be higher in Choroterpes and Ephoron compared to Heptagenia, although concentrations among taxa were generally low with uranium concentrations ranging from 0.28 to 2.15 µg/g dry weight and molybdenum from 0.50 to 3.99 µg/g dry weight. Comparison of mean tissue concentrations at upstream (sites 1–4) and downstream sites (11 and 14) indicated that tissue concentrations of uranium in Choroterpes and Ephoron at site 11 significantly exceeded (p-value<0.05; single observation mean comparison technique, Sokal and Rohlf, 2003) the expected range of values (±95 percent CI) for sites upstream from the groundwater contaminant plume (fig. 34). Uranium concentrations in Heptagenia showed a nonsignificant increase at site 11 relative to upstream sites. Molybdenum concentrations in all three taxa were significantly higher at site 11 (p-value<0.05) compared to tissue concentrations upstream. Constituent concentrations in all three mayfly nymphs decreased farthest downstream at site 14 to within expected levels observed upstream at sites 1–4, indicating that the highest concentrations of bioavailable uranium and molybdenum were likely limited to the region next to and immediately downstream from the contaminant plume.

In August 2017, floating-rack periphyton samplers were installed in free-flowing portions of the stream channel at five locations representing conditions upstream, adjacent to, and downstream of where the legacy groundwater plume intercepts the left bank of the Little Wind River (fig. 30, sites 3 and 12–15). Site 12 was next to where the groundwater plume intercepted the receiving system, but partial bank failure at that location resulted in high sediment loads proximate to the sampler location and extensive sediment buildup on the sampler substrates. Thus, the results from site 12 were not included in the final analysis and summary.

Ambient nutrient concentrations, amount of solar radiation, and flow velocity are primary determinants of algal growth rates in most riverine systems (Genter 1996). Although nutrient concentrations in the study reach were not determined as part of this investigation, it was felt given the limited spatial segregation of sampling locations (about 700 m distance between furthest upstream and downstream periphyton sites) that differences among sites in nutrient concentrations would be relatively minor. Among-site differences in incident solar radiation during the sampling period were not remarkable and ranged from 5.54 to 6.24 kilowatt-hours per square meter per day (Naftz and others, 2019). Near-sampler flow velocity at site 3 upstream from where the groundwater plume entered the receiving system was 0.318 meter per second (m/s) at the time of sampler installation compared to 0.314 m/s (0.200–0.481 m/s) averaged for sites downstream from the plume.

Concentrations of uranium and molybdenum in periphyton were generally low overall and ranged from 1.11 to 1.52 µg/g dry weight and 0.26 to 0.47 µg/g dry weight, respectively. Median uranium concentration in periphyton for sites downstream from the groundwater contaminant plume (sites 13, 14, and 15) was 1.28 µg/g compared to 1.29 µg/g upstream at site 3 (fig. 35). Median molybdenum concentration in periphyton for sites downstream from the contaminant plume was 0.39 µg/g compared to 0.44 µg/g at site 3. There was no evidence of significant uranium or molybdenum enrichment in the receiving system nearby or downstream from the groundwater influx.

In addition, there was no evidence of reduction in net primary productivity at sites downstream from where the groundwater plume entered the study reach. Net autotrophic biomass accrual rates were lowest upstream from the groundwater plume at site 3 (0.16 milligram [mg] chlorophyll a per square meter per day [Chl_a/m²/d]) and highest downstream with concentrations of 0.22, 0.25, and 0.29 mg Chl_a/m²/d at sites 13, 14, and 15, respectively (fig. 36).