Isoscapes as a Regional-Scale Tool for Tracing Groundwater Uranium Cycling in the Northern Plains, United States

Abstract

Groundwater uranium (U) contamination poses a significant health risk, particularly in Native American communities reliant on private wells. This study examines groundwater U cycling in a tribal region in South Dakota that participates in the Strong Heart Water Study based on samples from 140 private wells. We measured U concentrations, δ²³⁸U, (²³⁴U/²³⁸U), and redox-sensitive elements including iron (Fe), manganese (Mn), nitrate (NO₃⁻), selenium (Se), and vanadium (V). Uranium concentrations range from 0.4 to 48.2 μg/L, with 5% exceeding the USEPA maximum contaminant level of 30 μg/L. Spatial patterns in δ²³⁸U and (²³⁴U/²³⁸U) delineate distinct redox regimes: oxidizing zones in the northeast show higher U (median=18 μg/L) and positive δ²³⁸U values (0.08 to 0.30‰), while reducing zones in the southwest display lower U (median=10 μg/L) and large negative δ²³⁸U values (−0.61 to −0.30‰). The (²³⁴U/²³⁸U) values (1.53 to 3.07; median 2.03) serve as a tracer of source proximity, with lower values (1.53–1.80) indicate shorter travel distance relative to U source and higher values (1.80–2.50) reflecting U transported farther along flow paths. Cluster and uniform manifold approximation and projection (UMAP) analyses identify three geochemical environments consistent with oxidizing, reducing and intermediate redox conditions. Constructing the first δ²³⁸U and (²³⁴U/²³⁸U) isoscapes for a sandstone aquifer, we show that U is released by oxidative dissolution in the northeast and removed under reducing conditions in the southwest and that the northeastern zone may require continuous monitoring and intervention for exposure reduction.

Graphical Abstract

INTRODUCTION

Safe drinking water is a critical concern in Native American communities due to limited infrastructure and widespread groundwater contamination.^1,2 Approximately 6.5% of Native American households lack access to safe and adequate water sources or waste disposal facilities, compared to less than 1% of households in the general U.S. population.^3,4 Additionally, groundwater contamination from uranium (U) and arsenic (As) poses serious health risks, especially in Native American communities that rely on private, unregulated wells^2,5–9. Communities participating in the Strong Heart Water study (SHWS), located in the Northern Plains, have groundwater U concentrations as high as 60 ppb, nearly double the United States Environmental Protection Agency’s (U.S. EPA) maximum contaminant level (MCL) of 30 ppb.^10,11 Public health studies have shown that residents in these areas experience a 40% higher cancer mortality rate than the national average¹² and have the highest documented U and As exposure among all demographic groups in the U.S.^13,14

While history of U mining in parts of South Dakota has raised concerns about mining-related contamination,^{15–19, 18,19} there are no legacy or active U mines within or near SHWS study sites. Instead, elevated groundwater U concentrations in this region are primarily attributed to geogenic sources, as U mineralized zones are dispersed throughout the subsurface.¹⁹ The dissolution of U-bearing bedrock results in non-point source contamination in aquifers, particularly in areas where residents rely on untreated private wells for drinking, livestock water, and irrigation. Despite ongoing infrastructure improvements, many households in these communities remain dependent on private wells, which increases their exposure to naturally occurring contaminants. Understanding the processes that govern the release or removal of U in groundwater is essential for identifying high-risk areas and developing effective mitigation strategies.

While dissolved U concentrations provide a general indication of U contamination, they alone are insufficient to characterize the geochemical processes that control U mobility in groundwater. A key factor influencing U mobility in groundwater is its oxidation state. Under reducing conditions and near-neutral pH, the highly soluble U(VI) undergoes reduction to insoluble U(IV), leading to the formation of U-containing solids, such as uraninite, which are effectively immobilized. This reduction is primarily driven by microbially mediated terminal electron-accepting processes in aquifers, such as iron [Fe (III)] and/or sulfate (SO₄²⁻) reduction.^20–26 However, reduced U(IV) solids are susceptible to reoxidation and easily oxidized by dissolved oxygen (O₂) or nitrate (NO₃⁻) to form soluble U(VI).^27,28 In aqueous environments, with elevated Ca²⁺ concentrations, U predominantly forms neutral calcium–uranyl–carbonate complexes, including both neutral Ca₂UO₂(CO₃)₃⁰ and negatively charged CaUO₂(CO₃)₃²⁻ species.^29–36 The presence of complexing agents like bicarbonate and organic ligands stabilizes dissolved U(VI) in groundwater and enhances its mobility.^37,38 Therefore, local variations in aquifer geochemistry, mineral composition, organic matter content, and microbial activity create heterogeneous redox zones that further complicates the interpretation of the spatial distribution of dissolved U in groundwater. This challenge can be addressed by using indicators sensitive only to the U redox reactions. Fractionation of U isotopes (i.e. changes in ²³⁸U/²³⁵U) serves as an indicator of environments conducive to U(VI) reduction and mobilization (Table S1 of the Supporting Information).^39–42

Uranium isotope fractionation occurs due to the nuclear volume effect (NVE), which results from differences in nuclear size rather than mass between isotopes.^43–45 Because ²³⁸U has a disproportionately larger nuclear volume than ²³⁵U and therefore a lower electron density at the nucleus, ²³⁸U is preferentially incorporated into reduced U(IV) phases during reduction, which results in a more thermodynamically stable configuration. Consequently, U(VI) remaining in solution becomes progressively enriched in ²³⁵U. Laboratory experiments have demonstrated that U isotope fractionation [expressed as ε, approximately the per mil difference in ²³⁸U/²³⁵U between U(IV) and U(VI) relative to an isotopic standard] during microbial U(VI) reduction, can reach approximately 1‰,^46,47 while abiotic reduction by FeS, a product of microbial sulfate reduction, results in similar fractionation (ε ≈ 0.8‰).³⁹ This consistency underpins the use of U isotope ratios as robust indicators of reduction, regardless of whether the process is microbially mediated or abiotic. In contrast, U(VI) removal by adsorption onto mineral surfaces causes minimal fractionation (ε ≈ −0.2‰), with a slight preference for ²³⁵U during adsorption, which produces an effect opposite to that of reduction-driven fractionation.^48,49

Variation in U activity ratios or (²³⁴U/²³⁸U) in groundwater may serve as an effective indicator for determining U source proximity (Figure S1). In sandstone-hosted U deposits, (²³⁴U/²³⁸U) approaches secular equilibrium over approximately one million years, as the ingrowth of ²³⁴U is balanced by the decay of ²³⁴U, which produces a U activity ratio of 1.⁵⁰ However, in natural systems, this equilibrium is often disrupted by water-rock interactions that preferentially mobilize ²³⁴U. Such disequilibrium arises from α-recoil, in which recoil energy from the decay of ²³⁸U displaces its daughter nuclide ²³⁴Th by approximately 20–30 nanometers in silicate minerals. As a result, some of these atoms are ejected from the mineral into groundwater or relocated toward grain boundaries or microfractures, where they are more susceptible to leaching.^51,52 The extent of ²³⁴U recoil loss is influenced by the grain size and geometry of U-bearing minerals and tends to increase as grain size decreases.^53–55 This process can lead to (²³⁴U/²³⁸U) values below 1 in U-bearing minerals. Importantly, U transport in groundwater directly influences (²³⁴U/²³⁸U). When dissolved U stays close to its source, groundwater (²³⁴U/²³⁸U) reflects the U activity ratio of the source and is maintained near secular equilibrium.^56–59 In contrast, when U is transported over longer distances without precipitation, continued α-decay in groundwater and sustained ²³⁴U release from minerals along the flow path led to progressively elevated (²³⁴U/²³⁸U) values in groundwater.^56–59 This process is largely unaffected by adsorption–desorption of U, in contrast to the pronounced isotopic effects of α-decay of ²³⁸U in water and preferential leaching of ²³⁴U from recoil-damaged lattice sites of U-bearing minerals. Dissolved and surface-bound U(VI) rapidly reach isotopic equilibrium,^60–63 and attain identical (²³⁴U/²³⁸U) values within 2–24 hours^48,49 at ionic strengths typical of groundwater. Overall, U activity ratios in groundwater provide a robust tracer for distinguishing local versus distal U sources.

The present study focuses on sandstone aquifers in the Northern Plains, U.S.A., as a natural system to investigate how redox conditions, microbial activity, and flow dynamics regulate U mobility. These aquifers have well-connected pore networks that sustain advective flow, support microbial populations, and create broad zones where redox conditions gradually change with depth and distance^64–67 unlike the sharp, localized reactions that occur along fractures in granitic aquifers⁶⁸. Within these pore spaces, metal-reducing microbial communities reduce U(VI) to U(IV)^20–26 and promote U precipitation along the grain boundaries, processes that can be reversed when they interact with oxygenated water. Several sandstone aquifers across the western United States, including the entire High Plains, and Colorado Plateau, host extensive U deposits that collectively account for 90 % of U.S. reserves¹⁹ but also supply more than 114 million L/s of groundwater.⁶⁹ Nearly half of this water is used for public supply,⁶⁹ which underscores the risk of U contamination in drinking-water. Therefore, sandstone aquifers represent an important setting to constrain how redox processes regulate U mobility and long-term sequestration in the aquifer. To our knowledge, no study has investigated the spatial pattern of U-isotopic composition (i.e., U isoscapes) in groundwater to characterize U cycling in sandstone aquifers at a regional scale (~200 km scale).

Here, we report total U concentrations, ²³⁸U/²³⁵U, (²³⁴U/²³⁸U) and total dissolved concentrations of redox-sensitive solutes, NO₃⁻, Fe, manganese (Mn), selenium (Se), and vanadium (V), in groundwater from private wells on tribal lands in the Northern Plains. This geochemical dataset, combined with field parameters [temperature, pH, redox potential (Eh), and electrical conductivity (EC)], is analyzed using uniform manifold approximation and projection (UMAP), a nonlinear dimensionality-reduction method that captures complex relationships among isotopic and geochemical variables.⁷⁰ This technique can resolve overlapping geochemical regimes,^71–74 often hidden in traditional data reduction techniques like Principal Component Analysis (PCA) which assume linearity and normal distribution. The objective of this study is to construct the first δ²³⁸U and (²³⁴U/²³⁸U) isoscapes for sandstone aquifers that (i) delineates zones of U reduction and oxidation indicating where U is released or removed from groundwater and (ii) identifies those wells that are proximal to U sources. By combining the geochemical dataset with UMAP-based nonlinear analysis, this study develops a process-level, transferable framework for interpreting U redox dynamics in sedimentary aquifers.

MATERIAL AND METHODS

Study Population

This study was conducted as part of a community-engaged effort in collaboration with Native American communities in the Northern Plains, within the framework of the Strong Heart Water Study (SHWS).^75–81 All research activities including participant recruitment, informed consent, and data collection were carried out in partnership with Missouri Breaks Industries Research, Inc. (MBIRI), a Native American owned medical research organization. MBIRI facilitated community engagement and implementation through established networks of tribal leaders, and local institutions. The study was approved by appropriate Tribal Research Review Boards, and all activities adhered to principles of ethical research conduct and tribal data sovereignty. In accordance with data sovereignty principles, participating tribes retain ownership of the data, exercise decision-making authority over its future use, and are actively involved in guiding how findings are interpreted and disseminated. To uphold these commitments and protect confidentiality, the identities of participating communities and precise sampling locations are not disclosed. Household recruitment was conducted through community outreach led by MBIRI, and groundwater samples were collected only after informed consent was obtained. Participants had the opportunity to receive their household water quality results and were informed about any contaminants exceeding health-based guidelines. This participatory model ensures that research activities are responsive to community priorities and supports respectful engagement throughout the study lifecycle.

Study Area

The study area covers approximately 9,000 km² of tribal land in southwestern South Dakota, southeast of the Black Hills. The region is underlain by the Arikaree Formation, which is the primary aquifer, and the White River Group (Figures S2–S3 of the Supporting Information). Previous studies describe these connected formations as compositionally similar sandstones with interbeds of siltstone, clay, and minor lignite. Their dominant mineral assemblage includes quartz, feldspar, Fe-oxide coatings, clays, and localized pyritic or organic-rich zones that can promote reducing microenvironments.^10,82–85 In some localities, the upper White River Group also yields water to wells, as it is hydraulically connected, and composed of rock types similar to the Arikaree Formation. The basal part of the White River sandstones contains ore-grade U mineralization in the form of roll-front U deposits.⁸⁶ The aquifer is largely unconfined, and groundwater generally flows from south to north (Figure 1). Because the formations are nearly flat lying (regional dip < 1°), the potentiometric surface follows local topography and drainage (Figure 1).⁸³ Well-depth information was not available for the private domestic wells sampled, but previous studies suggests that most domestic wells in the study area draw from the Arikaree Formation, although some wells on the northeastern and southwestern margins may intersect the upper White River Group.^10,11,82,87 Additional information is provided in section S1 of the Supporting Information.

Figure 1:

Spatial distribution of uranium (U) concentrations (μg/L) in groundwater samples (n=140) across the study region. The heatmap is generated using IDW interpolation, with warm colors (red) indicating higher U concentrations and cool colors (blue) representing lower concentrations. Different symbols represent groundwater sampling locations (n=140) based on their cluster assignments. The inset map shows the location of Pine Ridge Reservation in southwest South Dakota. A scale bar and north arrow are included for spatial reference. The black thin arrows indicate groundwater flow path, and the black thin continuous lines indicate potentiometric surface. Gray numerical labels show potentiometric surface elevations (contour interval: 15 m), referenced to the NAVD 88 vertical datum traced from published studies.⁸³

Sampling Approach

Groundwater samples were collected from 140 private wells across the study area as part of a community-based monitoring initiative between April 2023 and October 2023 in multiple batches. Sampling locations were selected based on participant household locations and reflect a non-random, human-centered sampling design. Prior to sampling, wells were purged until the field parameters such as temperature, pH, redox potential (Eh), and electrical conductivity (EC) stabilized. Immediately after collection, the samples were filtered using 0.2 μm cellulose acetate membrane filters (Millipore) and then stored in pre-cleaned high-density polyethylene (HDPE) bottles. The samples were then shipped in insulated shipping containers at 4°C to the laboratory, where samples for trace metal and U isotope analysis were acidified with Optima-grade nitric acid (HNO₃) to a final molarity of ~0.15 M to preserve metals in solution. Measurements of trace metals, anions, and U isotopes were carried out in batches starting in May 2023. Additional methodological details, including sampling protocols, analytical techniques, and quality assurance procedures, are provided in the Supporting Information (sections S2 and S3 of the Supporting Information).

Uranium Isotopic Analysis

The ²³⁸U/²³⁵U measurements in groundwater were carried out using a Nu Plasma 3 MC-ICP-MS (Wrexham, UK) operating in low-resolution mode at the Novel Isotopes in Climate Environment and Rocks (NICER) lab, Lamont-Doherty Earth Observatory, Columbia University using a double-spike technique described in references.^41,88–90 All samples were purified using the UTEVA resin (Eichrom) prior to isotopic measurements (details in section S4 of the Supporting Information). During isotopic analysis, the ²³³U/²³⁶U, ²³⁴U/²³⁶U, ²³⁵U/²³⁶U, ²³⁴U/²³⁸U, and ²³⁸U/²³⁶U ratios were measured simultaneously. The ²³⁸U/²³⁵U ratios are reported as δ²³⁸U, defined by:

$${\delta }^{238}\text{U}(‰)=\left[\frac{{\frac{{}^{238}U}{{}^{235}U}}_{sample}}{{\frac{{}^{238}U}{{}^{235}U}}_{CRM145}}-1\right]\times 1000$$ (1)

The analytical uncertainty of δ²³⁸U measurements, determined as twice the root mean square difference for 19 pairs of duplicate sample preparations (15% of the total samples), is 0.08‰. The (²³⁴U/²³⁸U) values were determined from measured ²³⁴U/²³⁸U using CRM 145 as the bracketing standard [(²³⁴U/²³⁸U) =0.963]. The analytical precision (2σ) of the (²³⁴U/²³⁸U) measurements is 0.01.

Inverse Distance Weighting (IDW) Interpolation

IDW interpolation was used to visualize the spatial distribution of groundwater U concentration, δ²³⁸U, and (²³⁴U/²³⁸U) across the study area. IDW was selected because it performs reliably with irregularly distributed samples and provides a straightforward means of highlighting localized geochemical variability without assuming statistical stationarity or strong spatial autocorrelation. The interpolation was applied using the five nearest neighbors and restricted within the sampled convex-hull boundary to minimize edge effects. Additional implementation details, comparison with alternative methods, and interpolation parameters are provided in the section S6 of the Supporting Information.

Cluster Analysis

Unsupervised clustering techniques, including K-means and hierarchical clustering, were employed to classify groundwater samples based on geochemical characteristics. Numerical variables including U concentration, δ²³⁸U, (²³⁴U/²³⁸U), redox-sensitive indicators (e.g., Fe, Mn, Se, V, As, NO₃⁻, SO₄²⁻), fluoride (F⁻) and chloride (Cl⁻), and field parameters (pH, Eh, EC) were standardized via z-score normalization to ensure comparability across different scales. The optimal number of clusters (k) was determined using the elbow method⁹¹ and silhouette score⁹² (see Figure S4 of the Supporting Information), ensuring well-separated, internally cohesive groupings. Hierarchical clustering (Ward’s method) validated the robustness of the results, and a dendrogram based on Euclidean distance illustrated sample similarity (Figure S5 of the Supporting Information).⁹³ Box plots were used to compare geochemical differences between clusters, providing insights into U mobility and redox-driven variability (Figure 4 and Figure S11 of the Supporting Information). Additional methodological details are provided in section S7 of the Supporting Information.

Figure 4:

Box-and-whisker plots show U concentration (μg/L), δ²³⁸U (‰), and (²³⁴U/²³⁸U) for clusters 1–3. Clusters were identified by K-means on standardized chemistry (K = 5); groups with <2 samples were removed, leaving three retained clusters used throughout. Boxes denote the interquartile range with the median line; whiskers span the 5th–95th percentiles (values outside this range omitted for clarity). Colors match the cluster palette used in other figures. Cluster 1 shows the highest median U concentrations and more negative δ²³⁸U, whereas Cluster 3 exhibits the lowest U concentrations and the highest (²³⁴U/²³⁸U) ratios. They are listed in Table S3 of the Supporting Information.

Uniform manifold approximation and projection (UMAP)

The UMAP was applied to reduce data dimensionality and visualize nonlinear relationships among groundwater geochemical parameters. Unlike linear PCA, UMAP preserves both local and global structure, making it suitable for complex, multivariate datasets with nonlinear correlations. The algorithm was performed on standardized data using 15 nearest neighbors and a minimum distance of 0.1 to balance local continuity and global separation. The resulting two-dimensional embedding captured key geochemical gradients and facilitated comparison with cluster assignments (Figure 5). The UMAP outputs were integrated with redox indicators and isotopic parameters to reveal coherent geochemical patterns associated with U mobility. Further details are provided in section S8 of the Supporting Information.

Figure 5:

Two-dimensional UMAP projection of standardized geochemical variables. K-means was fit with K = 5 in the original feature space; two clusters with <2 samples were removed, leaving three substantive clusters shown (sample sizes as indicated in the legend). Points are colored and shaped by cluster for consistency with other figures. UMAP axes are unitless and used solely for visualization.

RESULTS AND DISCUSSION

Geochemistry and U isotope geochemistry of groundwater:

The concentrations of trace elements, NO₃⁻, SO₄²⁻, F⁻ and Cl⁻ along with the field parameters for the groundwater samples are presented in Table S2 of the Supporting Information. The pH of the groundwater samples ranges from 6.2 to 8.5, with a median value of 7.6. The EC varies between 190 and 927 μS/cm, with a median value of 431 μS/cm. Concentrations of F⁻ and Cl⁻, with median values of 0.3 mg/L and 4.3 mg/L, respectively, remain below the EPA MCLs of 4.0 mg/L for F⁻ and 250 mg/L for Cl⁻. Dissolved U [or U(VI)] ranges from 0.4 to 48.2 μg/L (median: 7.5 μg/L; Table S3), with 5% of samples exceeding the MCL of 30 μg/L. Uranium concentrations of 28–45 μg/L observed in the northeastern of the aquifer align with values reported by previous studies (Figure 1).^10,11 This clustering of high U samples in the northeastern and southwestern sections of the Arikaree aquifer, makes this area a high U zone (Figure 1), where the depth to water is relatively shallow compared to the rest of the aquifer.⁸³ Although >95% of groundwater samples contain U below the MCL, high-U occurrences pose a clear health concern, and many households in the study area rely on private, unregulated wells as their only water source. The U.S. MCL goal for U is 0 μg/L, indicating that no level of exposure is completely safe, and chronic exposure even at low concentrations (>2–10 μg/L) has been linked to kidney damage.^94–101 Understanding U behavior across the concentration range is therefore essential for evaluating long-term risk and U isotopes provide a key advancement in achieving this.

Groundwater δ²³⁸U values provide insights into the processes that control the release and removal of groundwater U in this region (Table S3 of the Supporting Information). The δ²³⁸U values range from −1.06‰ to 0.90‰, displaying a variation of up to 2‰ (Figure 2a). Only 6% of the samples exhibit δ²³⁸U > 0.10‰, while the majority (93%) fall within −0.60‰ to 0.10‰, with a median of −0.19‰ (Figure S6 of the Supporting Information). One sample with a strongly negative δ²³⁸U (−1.06‰) is located at the western boundary of the convex hull, where it is excluded from the isoscape due to spatial constraints, but not from statistical analysis. The δ²³⁸U isoscape (Figure 2a) reveals a regional trend, with higher δ²³⁸U values in the northeastern region of the study area and lower values in the southwestern region. Notably, both regions show high-U concentrations in groundwater (Figure 1).

Figure 2:

Spatial distribution of δ²³⁸U and (²³⁴U/²³⁸U) in groundwater samples (n=140) across the study region. (Top): Groundwater δ²³⁸U isoscapes showing warm colors (red) indicate more positive values and cool colors (blue) represent more negative values. (Bottom): Spatial variation of groundwater (²³⁴U/²³⁸U), with darker brown shades indicating lower ratios close to 1 and dark green shades indicates higher ratios.

Geochemical Evidence for U Reduction

To account for the observed U isotopic fractionation in the groundwater samples, we evaluated the role of key redox processes. These include oxidative dissolution of source rocks, localized U(VI) reduction, U adsorption onto mineral surfaces and remobilization of precipitated ²³⁸U-enriched U(IV) minerals.

Groundwater samples with positive δ²³⁸U from 0.01 ‰ to 0.90 ‰ closely match the δ²³⁸U signature of the source rock. A density plot of compiled δ²³⁸U values from sandstone-hosted U deposits,^102–104 which represent the dominant U sources in this region, is shown in Figure S7 of the Supporting Information. It exhibits a range of −0.26‰ to 0.84‰, with a median of 0.33‰. Notably, 98% of all compiled δ²³⁸U values fall between 0.01‰ and 0.84‰, which indicates a tightly constrained isotopic signature for the source rock. The overlap between source rock δ²³⁸U values and the groundwater samples with δ²³⁸U from 0.01 to 0.90‰ indicates that these samples are consistent with oxidative dissolution of U-bearing rocks present in the aquifer, a process that does not fractionate U isotopes and preserves the original source signature.¹⁰⁵

However, most (93%) groundwater samples with δ²³⁸U ranging from −1.06‰ to 0.01‰, deviate from the expected oxidative dissolution baseline (~0.33‰). This systematic decrease in δ²³⁸U provides strong evidence that portions of the aquifer are undergoing U reduction. Even the least fractionated samples (δ²³⁸U = 0.01‰) likely experienced limited U(VI) reduction, while the most fractionated waters (e.g., δ²³⁸U = –1.06‰) point to more extensive reduction.

Most of the groundwater samples exhibiting reduced signatures, show δ²³⁸U variations (−0.50‰ to −0.30‰) that are smaller than those observed in well-mixed experimental systems (0.00‰ to −2.00‰)^47,89,106, indicating attenuation of isotopic fractionation due to advective–diffusive mixing and redox heterogeneity within aquifers.¹⁰⁷ This pattern suggests that U(VI) reduction may be taking place in diffusion-limited microenvironments, such as organic rich zones or fine-grained sediments, with low permeability. In these isolated systems, reduced U(IV) becomes sequestered and has restricted exchanges with the surrounding U(VI) pool, as most groundwater flow and dissolved U bypass these zones. In these isolated zones, which are connected to the bulk groundwater primarily through diffusion, restricted transport of U(VI) to reduction sites weakens isotopic fractionation in the surrounding groundwater, leading to a dampened δ²³⁸U signal. A similar process has been observed for Se¹⁰⁸ and chromium (Cr)¹⁰⁹ isotope systems, where diffusion limitation within isolated reducing sediments weakens the isotopic fractionation observed in dissolved Se or Cr.

Additionally, the spatial distribution of δ²³⁸U observed in the groundwater samples indicates that U reduction occurs in localized redox pockets rather than uniformly throughout the aquifer. In the southwestern wells, elevated U concentrations under reducing conditions likely reflect slow U(VI) reduction, consistent with experimental observations of kinetically limited U reduction under similar conditions.^110,111 The relatively low δ²³⁸U values (−0.50‰ to −0.30‰) indicate that reduction is occurring but remains incomplete. These findings underscore the sensitivity of δ²³⁸U as a redox indicator, capable of detecting partial U(VI) reduction.

Uranium adsorption is likely a minor contributor to the observed δ²³⁸U variability given the high Ca concentrations (4.0–346 mg/L; median 45 mg/L) and near-neutral to slightly alkaline pH (7.2–8.5) of these groundwaters. Under these geochemical conditions, U(VI) predominantly exists as Ca–uranyl–carbonate complexes that strongly suppress the adsorption of U(VI) onto mineral grains.^29–36,90 Speciation modeling (Table S4 of the Supporting Information) supports this, showing that ~95% of dissolved U occurs as Ca–uranyl–carbonate species, primarily Ca₂UO₂(CO₃)₃⁰ (~61%) and CaUO₂(CO₃)₃²⁻ (~34%). Both the neutral Ca₂UO₂(CO₃)₃⁰ and the negatively charged CaUO₂(CO₃)₃²⁻ sorb weakly onto sandstones dominated by negatively charged quartz surfaces at near-neutral pH, resulting in minimal surface complexation and adsorption-driven isotope effects. Furthermore, the pH remains relatively uniform; 93% of pH data falls in the 7.5–7.7 range across the study area, providing no evidence for localized desorption hotspots. The direction and magnitude of adsorption-induced fractionation are also inconsistent with the large ²³⁵U enrichment reflected by negative δ²³⁸U values (−0.50‰ to −0.30‰). Adsorbed U is isotopically lighter than dissolved U [Δ²³⁸U(adsorbed–dissolved) ≈ –0.15 ‰],⁴⁸ which would enrich groundwater in ²³⁸U rather than ²³⁵U. The observed ²³⁵U enrichment (negative δ²³⁸U) therefore reflects reductive processes rather than adsorption. Interestingly, a field investigation at a U-contaminated site at Rifle, Colorado, showed no measurable isotope effect during U(VI) adsorption–desorption in an aquifer dominated by Ca−uranyl− carbonate species.⁹⁰

Other processes, such as the remobilization of precipitated ²³⁸U-enriched U(IV) minerals, are also unlikely to explain the observed large negative δ²³⁸U values (−0.50‰ to −0.30‰). For example, if U minerals with δ²³⁸U values ranging from 0.33‰ to 4.7‰^112,113 were dissolving, the groundwater would become enriched in ²³⁸U and exhibit positive δ²³⁸U values,¹¹⁴ rather than the negative values observed. Collectively, these lines of evidence indicate that U reduction, rather than adsorption or remobilization of precipitated ²³⁸U-enriched U(IV) minerals, produces the observed large ²³⁸U enrichment or negative δ²³⁸U in groundwater.

The spatial distribution of additional redox-sensitive indicators, including NO₃⁻, and total dissolved Mn, Fe, Se, and V (Table S2 of the Supporting Information), aligns with the reducing environments identified from δ²³⁸U data. In groundwater, microbial metabolism typically progresses from NO₃⁻ reduction to Mn(IV) and Fe(III) reduction, depending on the availability of electron acceptors. In the western portion of the aquifer, low NO₃⁻ concentrations and elevated Fe and Mn levels (Figure 3) indicate that redox conditions have advanced to Mn(IV) and Fe(III) reduction. Most groundwater samples (92%) contain NO₃⁻ below the MCL of 10 mg/L, consistent with redox conditions beyond NO₃⁻ reduction. Whereas isolated samples with NO₃⁻ > 10 mg/L likely reflect localized anthropogenic inputs such as agricultural runoff or wastewater infiltration. The weak correlation between NO₃⁻ and δ²³⁸U values (Figure S8 of the Supporting Information) suggests that NO₃⁻plays only a minor role in U mobilization under prevailing conditions. Although previous studies have shown that microbial NO₃⁻ reduction can oxidize U(IV) to U(VI) and transiently immobilize U through adsorption onto newly formed Fe (III) oxides.^27,28 Such NO₃⁻-driven redox transitions are not evident in our data set, which reflects processes beyond NO₃⁻ reduction, and is dominated by Fe- and Mn-reducing conditions. Concentrations of Mn and Fe also show weak linear correlations with δ²³⁸U (r² < 0.2), their spatial patterns reveal consistent redox-dependent behavior observed in δ²³⁸U. Lower δ²³⁸U values (−0.50 to −0.14 ‰, median of −0.23 ‰) correspond to elevated Mn (228–997 μg/L, median of 501 μg/L) and Fe (24–752 μg/L, median of 28 μg/L), indicative of active Mn(IV) and Fe(III) reduction. Conversely, higher δ²³⁸U values (−0.08 to 0.30 ‰) are associated with low Mn (0.03–41 μg/L, median of 0.4 μg/L) and Fe (0.6–14 μg/L, median of 1.1 μg/L), which reflect oxidizing conditions in the northeast. Additionally, the spatial distribution of Se and V mirrors the redox zones defined by δ²³⁸U (Figure S9 of the Supporting Information). Higher Se and V concentrations occur in the oxidizing northern section, while lower concentrations are found in the reducing southwest. This trend reflects their redox-sensitive mobility, as both elements tend to persist in groundwater under oxidizing conditions but are attenuated under reducing conditions.

Figure 3:

Spatial distribution of manganese (Mn) and iron (Fe) levels with groundwater flow direction. The unit of Mn and Fe concentration is μg/L.

Hydrological Influence On Redox Patterns

The spatial redox zones inferred from δ²³⁸U, Fe, Mn, Se, and V distributions correspond to hydrologic controls associated with recharge, saturated thickness, and horizontal hydraulic conductivity in the Arikaree aquifer. Recharge across the Arikaree aquifer is relatively uniform, but saturated thickness and horizontal hydraulic conductivity vary considerably.⁸² In the northeastern section, the vertical distance between the water table and aquifer base is small, which creates a thin saturated zone with high horizontal hydraulic conductivity, where oxygenated recharge water flows laterally at faster rates and oxidizes and leaches U from the U-bearing parts of the aquifer, as evident from high U concentrations and positive δ²³⁸U near the northeastern section. In contrast, the southern and southwestern parts, characterized by greater saturated thickness, likely have longer water residence times leading to sub-oxic to reducing groundwater with lower U and negative δ²³⁸U.

Uranium Activity Ratio as Tracers of U Source Proximity

The (²³⁴U/²³⁸U) of groundwater provide insights into U source proximity, and groundwater residence time within the Arikaree aquifer. The (²³⁴U/²³⁸U) in groundwater samples range from 1.53 to 3.07, with a median value of 2.03 (Table S3 of the Supporting Information and Figure 2b). Most samples (85%) fall within narrower range of 1.80 to 2.50, while 15% of samples exhibit (²³⁴U/²³⁸U) values between 1.53 to 1.80 (Figure S10 of the Supporting Information). These variations reflect combined influences of geological setting, rock–water interactions, redox processes, and alpha recoil.

Sorption of U is unlikely to account for this variability. The pH across the aquifer is relatively uniform (with 93% data within 7.5–7.8 range). Our speciation calculations (Table S4 of the Supporting Information) indicate that dissolved U(VI) is dominated by 95% of calcium–uranyl–tricarbonate complexes, with 61.3% present as the neutral Ca₂UO₂(CO₃)₃⁰ species and 33.7% as the negatively charged CaUO₂(CO₃)₃²⁻. Minor fractions (<3%) occur as uranyl–carbonate or hydroxide species. Formation of these species weakens U sorption on negatively charged quartz grains in sandstone, by making it reversible, which allows isotopic equilibrium between the dissolved and surface-bound U. Laboratory studies show that isotopic equilibrium between these dissolved and adsorbed U are established rapidly,^60–63 within 2 to 24 hours under typical groundwater conditions,^48,49 so any transient fractionation would dissipate over typical groundwater residence times. Moreover, since ²³⁴U is solely produced by the radiogenic decay of ²³⁸U, the (²³⁴U/²³⁸U) in groundwater is primarily governed by radiogenic ingrowth and the dissolution of U from mineral surfaces, rather than by sorption-related processes. Therefore, sorption has a negligible effect on the observed variation in (²³⁴U/²³⁸U) in this region.

In this region, U occurs in heterogeneously distributed mineralized zones within the host rock exhibiting (²³⁴U/²³⁸U) ranging from 0.60 to 1.60, with a median of 1.0 (Figure S10 of the Supporting Information).^41,115–117 Groundwater with (²³⁴U/²³⁸U) values around 1.53 to 1.80 is concentrated in the northern and southwestern parts of the aquifer, where the Arikaree Formation is thinner and the underlying White River Group, which is host to ore-grade U deposits,⁸⁶ lies closer to the surface (Figure S2 and S3 of the Supporting Information). In these areas, USGS test wells intersect the White River Group at shallower depths (27–72 m) compared to deeper penetrations (91–252 m) elsewhere in the aquifer.⁸² Within these deposits, α-decay of ²³⁸U over geological timescales causes ²³⁴U atoms to be ejected outside U mineral grains via α-recoil. Thus, the surfaces of U minerals often acquire a U activity ratio that is less than 1, while bulk mineral attains a (²³⁴U/²³⁸U) that is near secular equilibrium.¹¹⁸ Subsequent interaction with oxygenated groundwater preferentially leaches U from these surfaces and groundwater inherits the activity ratio of these surfaces which is near or below secular equilibrium. Longer groundwater transport times or distances increase the inherited U activity ratios, with greater transport distances producing more elevated ratios. This effect is evident in the northeastern part of the aquifer, where oxidizing conditions dominate and U concentrations reach 23–48 μg/L with δ²³⁸U from 0.08 to 0.30‰, and (²³⁴U/²³⁸U) values around 1.53 to 1.80. This lower range of (²³⁴U/²³⁸U) indicate shorter travel distance relative to U ores or ore-like enriched zones (e.g., smaller roll fronts). In comparison, samples with (²³⁴U/²³⁸U) ranging from 1.83 to 2.5 reflect U that has been transported from such sources over longer distances and accumulation of ²³⁴U derived from ongoing α-decay of ²³⁸U and recoil loss of ²³⁴U from the ubiquitous, background-level ²³⁸U present throughout the aquifer matrix. This characteristic U isotopic signature helps distinguish groundwater in contact with sandstone-hosted U deposit from groundwater where U has been transported away from its source.

Redox Domains in the Aquifer

By integrating U concentrations, δ²³⁸U values, (²³⁴U/²³⁸U) and redox-sensitive elements (Fe, Mn, Se, and V), we identify two contrasting redox regimes across the study area. This is illustrated by the zones shown in Figures 1–3 for descriptive purposes, to highlight dominant spatial trends in groundwater chemistry and redox conditions, rather than to imply sharply defined boundaries. In the northeastern downgradient section (Zone 1; Figure 2), oxidative dissolution actively mobilizes U, yielding the highest U concentrations (median = 20 μg/L). Elevated δ²³⁸U values (0.08 to 0.30‰), low Fe and Mn concentrations, and high Se and V concentrations together indicate persistent oxidizing conditions that maintain U in its soluble form. Groundwater here shows relatively low (²³⁴U/²³⁸U) values (1.53–1.80), pointing towards shorter travel distance from U ores or ore-like enriched zones. In contrast, the western section (Zone 2) exhibits markedly lower U concentrations (median=10 μg/L), large negative δ²³⁸U values (−0.61‰ to −0.30‰), and elevated Fe (24–752 μg/L) and Mn (228–997 μg/L) concentrations, collectively pointing to reducing conditions that favor U(VI) reduction and immobilization. The concurrent depletion of Se and V further supports this interpretation, as both form insoluble species under reducing environments. The higher (²³⁴U/²³⁸U) of ≥ 1.83 observed in this zone reflects longer travel distances from these sources. Overall, the contrasting redox environments in Zones 1 and 2 dictate whether U remains mobile or is sequestered, ultimately governing its long-term fate in the aquifer.

Three Dominant Redox Environments from Cluster and UMAP Analysis

Our cluster and UMAP analysis identified three major clusters (1–3) across the aquifer (Figures 1–6). Two minor clusters containing single samples were excluded from further interpretation (Figure S5 of the Supporting Information). Both methods resolve coherent geochemical groupings as shown by the box plots (Figure 4 and Figure S11 of the Supporting Information) and statistical analysis (Table S5 of the Supporting Information). The UMAP embedding (trustworthiness = 0.80; Figure S12 of the Supporting Information) reliably preserves local geochemical structure while maintaining global relationships in the data, which allows visualization of gradients among redox-sensitive elements. The smooth color gradients across the UMAP surface illustrate that redox transitions occur progressively rather than as sharp boundaries, and provide an integrated view of redox structure and U mobility across the aquifer (Figure S13 of the Supporting Information).

Figure 6.

Cluster-wise standardized means (z-scores) of major analytes from k-means clustering (K = 5; clusters with <2 samples removed, leaving three clusters). Cells report the mean z-score per variable; red/blue indicate above/below the dataset mean, respectively (center = 0). Cluster 1 is enriched in U, Se, V, As, and moderately in NO₃⁻ and Cl, with lower Fe and Mn and slightly higher δ²³⁸U and lower (²³⁴U/²³⁸U). Cluster 2 shows elevated Fe, Mn, SO₄²⁻, and Cl⁻, with lower Se, V, As and more negative δ²³⁸U. Cluster 3 is generally depleted in metals and anions, with slightly higher (²³⁴U/²³⁸U) and near-mean δ²³⁸U.

Cluster 1 represents an oxidizing regime with high U mobility. It shows the highest U concentrations (median = 18 μg/L), with several samples approaching or exceeding the 30 μg/L MCL. Differences in U concentrations among clusters are highly significant (p = 0.008 between clusters 1 and 3 and clusters 2 and 3) (Figure 4). Cluster 1 is characterized by elevated Se, V, As and SO₄²⁻ concentrations, which are significantly higher than those in Clusters 2 and 3 (p < 0.01 between both clusters 1 and 2 and clusters 1 and 3), and by depleted Fe and Mn, consistent with oxidizing conditions. The (²³⁴U/²³⁸U) values range from 1.53 to 2.30, with most samples (≈70%) falling between 1.53 and 1.80, and indicate that U was mobilized over relatively short travel distances from sources. The median δ²³⁸U is –0.16 ‰, and about 28% of samples show more positive values that indicate oxidation. The UMAP z-scores show strong positive loadings for NO₃⁻, U, Se, V and As and negative scores for Fe and Mn (Figure 6), which together confirm oxidation-driven U mobility.

Cluster 2 defines a reducing regime with high Fe and Mn concentration. Uranium concentrations are moderate (median = 10 μg/L), and δ²³⁸U values are the lowest (median = –0.20 ‰), consistent with isotopic fractionation associated with U(VI) reduction. The (²³⁴U/²³⁸U) values range from 1.60 to 2.20, with roughly one-third of samples between 1.60 and 1.80 and the remainder above 1.80. Both subsets occur under reducing conditions but differ in transport history. Samples with lower (²³⁴U/²³⁸U) values likely represent U that has traveled shorter distances from U sources. In contrast, higher (²³⁴U/²³⁸U) values reflect U transported over longer flow paths. Concentrations of Fe (median = 27 μg/L) and Mn (median = 383 μg/L) are significantly higher than in Clusters 1 or 3 (p < 0.001 between both clusters 2 and 1 and clusters 2 and 3), in agreement with active Fe (III) and Mn (IV) reduction. Lower Se and V concentrations in cluster 2 (p < 0.01 relative to Cluster 1), and in the UMAP space, strong positive z-scores for Fe and Mn and negative values for Se and V (Figure 6) are consistent with reducing conditions.

Cluster 3 represents a transitional regime defined by the lowest U concentrations (median = 6 μg/L) and generally low levels of redox-sensitive parameters such as NO₃⁻, Fe, Mn, Se, V, and As. The δ²³⁸U values in this cluster are slightly higher than those of cluster 2 but lower than cluster 1, consistent with weakly reducing to mixed conditions. The (²³⁴U/²³⁸U) values exceed 1.8 in about 70% of samples (p = 0.0001 compared to Cluster 2), which reflect groundwater that has moved along longer flow paths within the aquifer matrix gradually enrich the dissolved phase in ²³⁴U.

The spatial distribution of clusters across the study area reveals distinct redox zones and U mobility patterns, which provide critical insights into aquifer geochemistry (Figure 1). Cluster 1, associated with oxidizing conditions, occurs in the northeastern region of the study area. In this region, oxidative dissolution mobilizes U, which leads to elevated concentrations and higher δ²³⁸U (−0.08‰ to 0.30‰). Cluster 2 represents reducing conditions and is concentrated in the southwestern part of the study area, where U undergoes removal from groundwater through reduction and precipitation. This process inhibits U transport. Cluster 3 shows no clear spatial pattern and appears dispersed across the study area, which indicates the presence of transitional zones with varying redox conditions.

Environmental Implications

In aquifers like those beneath our study area, where sampling locations are sparse and unevenly distributed, U isotopes serve as a powerful tool for identifying zones of groundwater U release and removal. Spatial variations in δ²³⁸U delineate areas of active reduction, supported by Fe and other redox-sensitive indicators. However, Fe (III) reducing environments do not always coincide with U(VI) reduction, which can be kinetically limited. Unlike U concentrations, which can remain steady during slow U(VI) reduction, isotopic shifts capture ongoing redox transformations even when aqueous U levels remain at steady state. In such cases, partial isotopic shifts in δ²³⁸U provide early evidence of reduction before measurable declines in U concentrations occur. By integrating geochemical, and statistical analyses, this study establishes a practical framework for identifying redox regimes controlling U cycling and for anticipating zones of U persistence or attenuation.

Our results offer a robust spatial snapshot of redox regimes rather than a temporal record of redox dynamics. Long-term evaluation of U(VI) reduction and mobilization requires time-series isotopic and concentration data from private wells across the region. However, the irregular distribution of private wells, dictated by population density, distance between households, water access and affordability, limits the ability to monitor U transport at a fine spatial resolution. Future work should prioritize targeted resampling in the northeastern Arikaree aquifer, where ongoing oxidative dissolution may sustain U release. Despite these challenges, the combined geochemical–statistical approach provides a transferable model for assessing U mobility in data-limited settings. Collaboration with tribal environmental programs and local research organization (MBIRI) will ensure results are shared through community outreach, with risk maps and key messages translated into native languages to support well testing and safe water access initiatives.

Supplementary Material

Groundwater field parameters, trace element and anion concentrations, U isotopic values [δ²³⁸U and (²³⁴U/²³⁸U)] values, U speciation results and t-test results between clusters: Tables S1−S5

Details of the site background, protocol for sample collection, analytical procedures for trace element and anion concentration measurement, information on U isotope analysis, parameters for U speciation calculations, detailed methodology for IDW interpolation, cluster and UMAP analysis and As behavior in relation to U redox cycling.

SYNOPSIS.

Isotopic and statistical analyses identify groundwater uranium release and retention zones, informing contamination risk assessment and mitigation in a South Dakota tribal region.