Classification of groundwater age (nuclear/mixed/pre-nuclear) based on tritium: development and application of a non-linear, two end-member mixing model to samples obtained from long-screened wells

From May 2004 through June 2010, the USGS sampled ~1900 wells in more than 100 groundwater basins as part of the California Water Board's GAMA Priority Basin Project. Of these wells, ~1450 were sampled for the purpose of obtaining a spatially-unbiased assessment of groundwater quality at the basin-scale and ~450 were sampled for understanding groundwater quality. Most of the wells (~1700) are supply wells that are relatively deep (mean depth ~ 150 m) with relatively long screens (mean length ~ 75m). Some (~200) are observation wells that are relatively shallow (mean depth ~ 90m) with relatively short screens (mean length ~10m). Samples from all wells were analyzed for tritium (3H) and about half were analyzed for carbon-14 (14C). Analysis of 14C as a function of 3H indicates that samples from production wells are comparable to samples from observation wells across a wide range of 3H. A two end-member mixing model was developed to aid in the assessment of groundwater age at the basin-scale. The groundwater system is conceptualized as consisting of nuclear-age groundwater above an interface and pre-nuclear-age groundwater below the interface; the interface can be taken as sharp or diffuse. Wells may be screened above, below or across the interface. The model assumes that there are two characteristic 3H values: an upper threshold, above which a sample is classified as nuclear age, and a lower threshold, below which a sample is classified as pre-nuclear age. 3H values between the thresholds are considered representative of mixed-age groundwater. The lower threshold is based on analytical precision (0.2 tritium units [TU]) and the upper threshold is obtained through optimization of a non-linear, two end-member mixing model applied to samples with both 14C and 3H data. In the model, nuclear-age groundwater (3H>upper threshold) has a constant value of 14C. For mixed-age groundwater, 14C decreases linearly as a function of the fraction of nuclear-age groundwater (F). Samples with 3H below the lower threshold are not evaluated within the model. There are 4 unknowns: upper threshold for 3H, 14C for nuclear age groundwater, 14C for modern, pre-nuclear-age groundwater, and the slope of the linear relationship between 14C and F. Model parameters were optimized at the statewide, regional, and basin scales. Application of the model at statewide and regional scales suggests an upper threshold of 3H of 2.0 TU (normalized to 2008 as the year sampled). Analysis of variance indicates that classification of groundwater age based solely on 3H is appropriate for an ensemble of data points but not appropriate for an individual sample. In addition, application of the model provides a basis for comparing and contrasting basins to one another, and for identifying samples that are potential outliers. Identification of two thresholds for 3H provides a basis for classifying groundwater age for samples for which 14C data are not available (about one-half of our samples).