Supporting Information for: Natural Uranium Contamination in Major US Aquifers Linked to Nitrate Authors: Jason Nolan† and Karrie A. Weber†‡*
Department of Earth and Atmospheric Sciences, University of Nebraska-Lincoln, Lincoln, NE, USA. ‡
School of Biological Sciences, University of Nebraska-Lincoln, Lincoln, NE, USA.
*Correspondence to: [email protected]
; 232 Manter Hall, University of Nebraska-Lincoln, Lincoln, NE, USA 68588-0118, (402) 472-2739
This file includes: Supporting Information Figures S1 to S8 Tables S1 to S9 References 36-51
Supporting Methods: HP and CV Aquifer Geological Description: The geological units underlying the High Plains (HP) aquifer range from Permian to Quaternary. These units consist of silts, clays, cemented clays, dolomite, gypsum, bedded salts, anhydrite and chalks. 36 The aquifer consists of multiple interconnected formations of unconsolidated course to fine sand and gravel with an average saturated thickness of ~61 m with depths to groundwater ranging from on average less than 30 m in half of the aquifer, to up to 120 m (in southwestern Kansas). 37 Generally these sediments are part of major formations including the Brule formation, the Arikaree Group, and the Ogallala Group. Overlying Quaternary deposits consist of valley-fill alluvium, dune-sand, or eolian sands.36 Granites, shales, and various products of volcanism contribute to natural U in sediment.38 The NURE sediment-testing program revealed a maximum sediment U concentration in Texas of 62.8 mg/kg, well below the classification of a low-grade U ore (1,000 mg/kg). Specific formations such as the Arikaree Group and Brule Formation contain U with an average concentration of 3.2 mg/kg and 4.78.9mg/kg in sediment, respectively. 39 One identified U reserve located in western Nebraska lies below the HP aquifer in the Chadron Sandstone and is separated from the HP aquifer by clay rich confining units (~150m). Similar to the HP aquifer the Central Valley (CV) aquifer consists of unconsolidated sands and gravels. The aquifer originated as a basin in the Mesozoic era, later filling with weathered parent material of the Rocky Mountains and foothills during the Cenozic age. Overlying alluvium was later deposited during the quaternary period from further weathering of parent materials.40 Clay and silts typically comprise about 50% of valley fill sediments with additional volcanic ash incorporation. 41 In the CV the depth to groundwater is extremely
variable and has been increasing with increased withdrawals.41 However the majority of freshwater is found in the upper 300 m of the aquifer as the water becomes more saline with deeper depths. 40 Sediment U concentration did not exceed 25 mg/kg in the CV aquifer based on NURE sediment testing data.
Groundwater Nitrate and Uranium (U) Concentration Metadata: Geochemical data from the HP and CV aquifers was collected from 35 federal, state, and local government agencies and academic institutions (Table S2). The data was then assembled into one metadatabase. Latitude and longitude coordinates were linked to site identification numbers obtained from the relevant agency or institution. Location coordinates of groundwater data received in public land survey system format were converted to latitude and longitude using US Bureau of Land Management centroid locations and truncated. As requested by state agencies sharing data, latitude and longitude coordinates for private and public water supply wells were also truncated. The metadata can be downloaded from the UNL data repository at http://dx.doi.org/10.13014/K2MW2F2N. Data received from state agencies and city municipalities have had location of wells (latitude and longitude) and personal identifier information removed or truncated to conform with security requirements. The un-redacted files with locations can be requested by contacting the responsible agency directly with a formal written request. Well site locations for data received from NWIS requires a separate location file that can be downloaded at: http://waterdata.usgs.gov/nwis.
NWIS Groundwater Geochemical Data: Additional geochemical data was downloaded from the NWIS database, http://waterdata.usgs.gov/nwis, to include geochemical parameters (U, pH,
alkalinity, dissolved oxygen, nitrate, iron, manganese, and calcium) measured in groundwater samples. Filtered samples were utilized when appropriate. Samples that did not have a value for each parameter of interest were excluded from further analysis. Non-detect (zero) values were converted to ½ the LDL. Data was then joined to the separate latitude and longitude coordinate file retrieved from the NWIS database with the site identification number in ArcGIS. Any wells that were not within the HP or CV aquifer were excluded. This database was then used for subsequent statistical analyses.
Statistical Analyses: Spatial interpolation. Estimated groundwater U and nitrate concentrations were established using an interpolative model. Four different methods of interpolation were executed to identify the model resulting in the lowest root mean square error calculated using supplemental equation S1. Error within the model was minimized by adjusting spatial power and weighting decisions and then the model was validated (see Supplementary Information for details). 42 The model was cross validated by removing 20% of known data points from the metadatabase and reinterpolating. 42 Five categories of interpolated groundwater U and nitrate concentrations were established and plotted to denote spatial area coverage. Community identification and population calculations were completed using georeferenced census information. Historical and current U mining locations were spatially mapped in reference to contaminated areas.
Spatial Correlation Analysis. Spatial correlation grids were created using normalized U and nitrate interpolations (Eq. S2) for calculation of a Pearson correlation coefficient (rxy). 43
Interpolated nitrate and U concentrations (xindependent and ydependent) within a 1-km by 1-km area (1 km2) were used, dividing the covariance (Sxy) of the predicted concentrations by the product of their standard deviations Sx, Sy using supplemental equations 3, 4, 5, and 6. 44 This method has been previously applied in determining groundwater spatial distributions. 45 Interpolated groundwater U and nitrate concentrations associated with the 1-km by 1-km region calculated above were extracted using raster masking in Arc GIS to calculate mean and median concentrations associated with each correlation category.
Descriptive Statistics and Correlation Analyses. Metadatabase. Descriptive statistics, interpolation, spatial correlation, and tests of differences in mean concentrations of U and nitrate by depth, were conducted using the nitrate and U metadatabase. Descriptive statistics of the metadatabase were computed for groundwater U and nitrate concentrations using mean, max, median and standard error. 46 Histograms were created showing the distribution of groundwater U and nitrate concentrations in each aquifer in GraphPad Prism.
NWIS dataset. To test the correlation of groundwater U concentrations (Figure S9b) between geochemical parameters (see above), data was obtained from the NWIS dataset. Samples in which did not have an analytical report for each geochemical parameter (see above) were excluded from our analyses. Normality was tested using a D'Agostino-Pearson omnibus test. The test indicated non-normal distributions or parameters. As a result, non-parametric methods (Spearman’s rho) were used to test correlation and correlation significance. Correlation analyses conducted in shallow and deep wells were conducted by separating the dataset accordingly.
Scatterplots of wells categorized by depth of the correlation between nitrate and U were created in GraphPad Prism.
Eq. S1 𝑍𝑖 =
(𝑋𝑖 −𝜇𝑖 )
𝑆𝑥𝑥 = ∑𝑛𝑖=1( 𝑥𝑖 − 𝑥̅ )(𝑦𝑖 − 𝑦�) 𝑛
Eq. S2 Eq. S3 Eq. S4
𝑆𝑥 = � ∑𝑛𝑖=1(𝑥𝑖 − 𝑥̅ )2
𝑆𝑦 = � ∑𝑛𝑖=1(𝑦𝑖 − 𝑦�)2
2 3 4 5 6 7 8 9
Figure S1. Groundwater U concentrations measured in wells across the United States. Uranium values