Over a third of groundwater in USA public-supply aquifers is Anthropocene-age and susceptible to surface contamination

Understanding the distribution of groundwater ages within samples and aquifers is critical for assessing susceptibility to anthropogenic (surface-derived) and geogenic contamination, and for evaluating sustainability under overdraft and changing recharge. The study focuses on identifying the fractions of Anthropocene (post-1953), Holocene (75–11,800 years), and Pleistocene (>11,800 years) water in major U.S. public-supply aquifers. Prior single-tracer approaches (e.g., only tritium or only carbon-14) provide limited information (classifying as modern/mixed/old or yielding a single apparent age) and cannot resolve full age distributions. Here, the authors compute full age distributions at 1,279 public-supply sites across 21 U.S. Principal Aquifers (PAs), using multiple complementary tracers and lumped-parameter models (LPMs). An aquifer-scale cumulative distribution function (ACDF) aggregates individual sample CDFs to represent the spatially averaged age distribution for each PA, enabling estimation of Anthropocene/Holocene/Pleistocene fractions at aquifer and national scales. PAs are grouped into hydrogeologic classes (Western unconsolidated, Coastal clastic, Interior sandstone-carbonate, Glacial, Carbonate, Igneous/metamorphic) to interpret regional patterns. The study aims to provide a comprehensive national assessment of groundwater age distributions to inform contamination susceptibility and resource sustainability.

The paper builds on prior work highlighting the importance and complexity of groundwater age determination and its implications (e.g., Ferguson et al., 2020; Zalasiewicz et al., 2017). Global assessments have used single tracers to infer vulnerability and modern groundwater proportions (Gleeson et al., 2016; Jasechko et al., 2017), but such approaches lack full age distributions, particularly for mixed waters containing both Anthropocene and older components. Lumped parameter models (Małoszewski & Zuber, 1982; Cook & Böhlke, 2000; Kazemi et al., 2006; Jurgens et al., 2012, 2016) provide a tractable way to estimate age distributions and have shown agreement with more intensive particle-tracking/numerical models (Eberts et al., 2012). The sampling design uses equal-area grids to yield spatially representative ensembles at aquifer scale (Scott, 1990; Belitz et al., 2010). Regional hydrogeologic context for U.S. Principal Aquifers and pumping statistics are provided by USGS sources. Prior regional studies link aquifer confinement, recharge areas, and lateral flow distances to age distributions and susceptibility (e.g., Solder, 2020; Masterson et al., 2016).

Study design and sampling: 1,279 groundwater sites (mostly wells; 10 springs) used for public supply across 21 U.S. Principal Aquifers were sampled between 2004 and 2018. Each PA was divided into ~60 equal-area grid cells and one public-supply well per cell was sampled to obtain spatially representative, unbiased ensembles for each PA and the conterminous U.S. Environmental tracers: Multiple tracers with complementary dating ranges were measured, including tritium (3H), tritiogenic helium-3 (3He from 3H decay estimated via excess helium, He_ex), sulfur hexafluoride (SF6), chlorofluorocarbons (CFC-11, CFC-12, CFC-113), carbon-14 (14C) of dissolved inorganic carbon, and radiogenic helium-4 (4He_rad). Not all tracers were available at every site; at least one Anthropocene tracer and one pre-Anthropocene tracer were available at >95% of sites. Reported detection/absence limits included: 3H 0.1 TU; SF6 1 TU; CFC-11 0.5 pptv; CFC-12 10 pptv; CFC-113 50 pptv; 14C 0.1 pmC; He_ex 9.0×10−8. Tracer processing and corrections: - 14C: Corrected for dilution by 14C-dead carbon using analytical correction models, inverse geochemical modeling (NETPATH/NetpathXL), or scaling of the atmospheric 14C record guided by independent Anthropocene tracers to avoid overcorrection. Typical corrections were 30–46 pmC with uncertainty <20%. - Gases: SF6, CFCs, He_ex, and He_rad were corrected for solubility equilibrium, excess air, gas fractionation, and terrigenic helium using DGMETA, incorporating model/analytical uncertainties. Terrigenic He 3He/4He ratios were estimated per PA from low-3H samples; PAs with >10% mantle helium were excluded from He_rad-based age estimation. Where needed, crustal 4He flux corrections were constrained by matching 14C and He_rad depth profiles. Modeling age distributions: - Age definition includes travel time since infiltration; unsaturated-zone travel time was included for Anthropocene waters where relevant. - Lumped-parameter models (LPMs) implemented with TracerLPM were calibrated by inverse methods to minimize chi-square misfit across tracers, using regional atmospheric histories (3H precipitation by site, NH air histories for SF6/CFCs, combined IntCal13 and tropospheric records for 14C). In some areas, 3H histories were scaled to match observed relationships for Anthropocene waters. - He_rad production was estimated following Andrews & Lee using PA-specific U/Th, porosity, and bulk density. - Model selection: Primarily a single dispersion model (DM) for unimodal Anthropocene or pre-Anthropocene samples, or a bimodal mixture of two DMs (BMM-DM-DM) for samples indicating mixtures of Anthropocene and pre-Anthropocene water. Alternative LPMs were considered where appropriate; multiple LPMs often yielded similar inferred distributions. - Fit evaluation: Tracer errors were 10% for 3H and 20% for other tracers to reflect processing uncertainty. Acceptable models had chi-square <10; 97% met this criterion, ~80% had chi-square <1 (potentially under-constrained). For under-constrained cases, multiple solutions were computed and the median mean age was used. Median mean-age error was 9% (IQR 5–25%). Aquifer-scale cumulative distribution functions (ACDFs): The ACDF for each PA was computed as a linear combination (spatial average) of individual sample CDFs (1-year age bins), preserving the full age spectra and enabling estimation of Anthropocene, Holocene, and Pleistocene fractions. National distributions were computed by area-weighting PA ACDFs. Classification and interpretation: PAs were grouped into six hydrogeologic classes (Western unconsolidated; Coastal clastic; Interior sandstone/sandstone-carbonate; Glacial; Carbonate; Igneous/metamorphic) to interpret controls such as climate, confinement, and flow system geometry on age distributions.

- National age composition (area-weighted across PAs): 38% Anthropocene (since 1953), 34% Holocene (75–11,800 years), 28% Pleistocene (>11,800 years). - Anthropocene fraction spans <5% to 100% among aquifers, indicating a wide range of susceptibility to surface contamination. - Across hydrogeologic classes and PAs: • Western unconsolidated PAs: Predominantly Holocene. Highest Anthropocene fractions in California Coastal Basins (CACB), Central Valley (CVAL), and High Plains (HPAQ), reflecting human-accelerated recharge (e.g., irrigation, managed recharge). Many wells show bimodal age distributions. • Coastal clastic PAs (eastern U.S.): Predominantly Pleistocene; ACDFs are similar across PAs. Confinement and long lateral flow paths dominate; small Anthropocene fractions occur in inland outcrop recharge belts. • Interior sandstone–carbonate PAs (COPL, EDTR, CMOR): Low Anthropocene proportions; COPL mostly Holocene, EDTR and CMOR mostly Pleistocene. Deep CMOR contains the oldest waters identified; >25% older than 100 ka. • Glacial aquifer system: Young overall (≈60% Anthropocene, 30% Holocene, 10% Pleistocene) due to humid climate, thin deposits, and short flow paths; older waters where deposits are thicker. • Carbonate PAs: Diverse and often bimodal ACDFs (dual porosity). Confined OZRK and FLOR are older (Holocene/Pleistocene-dominated). Unconfined BNRC (arid) older than BISC (humid); BISC is 100% Anthropocene. • Igneous/metamorphic PAs: CLPT (arid) primarily Holocene; PDBR (humid) primarily Anthropocene (~95%). - Sample-level age characteristics: Mean ages range from ~1 to ~750,000 years (median ~2,500 years). About 30% of samples are bimodal mixtures of Anthropocene and pre-Anthropocene waters. 60% of samples had 14C <90 pmC, and ~25% of those also had detectable 3H (>0.1 TU), revealing mixed-age waters only identifiable via multi-tracer modeling. - Depth relation: Mean age positively correlates with well depth in all but two fractured-rock PAs (BNRC, PDBR), consistent with expected vertical age stratification or dipping formations; lack of correlation reflects fracture-dominated flow and discontinuous outcrops. - Unsaturated-zone travel time: For predominantly Anthropocene samples in western aquifers, UZ travel times were 1–10 years in about half and 10–25 years in the other half, indicating UZ contributions to apparent age. - Error/uncertainty: Median mean-age error 9% (IQR 5–25%); typical 10% mean-age shifts change epoch fractions by <5%.

The multi-tracer, LPM-based ACDF approach reveals systematic spatial patterns in groundwater age that directly inform contamination susceptibility and resource sustainability. Aquifers with high Anthropocene fractions (e.g., BISC, PDBR, Glacial PA, parts of CACB/CVAL/HPAQ) are more susceptible to surface-derived contaminants; observed pesticide occurrence varies with land use intensity despite similar Anthropocene fractions. Aquifers dominated by Holocene/Pleistocene water (e.g., Coastal clastics, Interior sandstone-carbonate, confined carbonates) may be less vulnerable to modern contaminants but can be prone to geogenic issues (e.g., radium in CMOR Holocene/Pleistocene waters; arsenic/fluoride in Pleistocene waters of Coastal PAs), reflecting redox and geochemical evolution along long flow paths. Sustainability assessments benefit from ACDFs: predominantly Holocene age in HPAQ aligns with documented groundwater mining, whereas CACB demonstrates that managed recharge can replace pumped older water with Anthropocene recharge, maintaining hydrologic balance. The ACDF framework preserves multimodality and full age spectra lost in mean-age or single-tracer summaries, enabling nuanced, aquifer-specific management insights.

This study provides a nationwide, aquifer-resolved assessment of groundwater age distributions for U.S. public-supply aquifers by integrating multiple environmental tracers with lumped-parameter modeling and aggregating results via aquifer-scale CDFs. Nationally, public-supply groundwater comprises 38% Anthropocene, 34% Holocene, and 28% Pleistocene water, with strong, systematic variation by hydrogeologic setting, climate, and confinement. The ACDF methodology enables robust evaluation of susceptibility to anthropogenic and geogenic contaminants and supports sustainability assessments by distinguishing recharge dynamics and storage contributions. Future work could integrate transient recharge histories, refine corrections for 14C dilution and helium sources, couple ACDFs with groundwater flow models to quantify capture and leakage, and establish ongoing age monitoring to track shifts under climate variability, land-use change, and managed aquifer recharge.

- Model non-uniqueness: Different LPMs can fit tracer data similarly; although shared features of age distributions are robust, some parameter estimates remain non-unique. - Under-constrained cases: ~80% of models had chi-square <1, indicating potential under-constraint; median-of-solutions was used where appropriate. - Tracer corrections: 14C dilution by dead carbon and helium source partitioning (radiogenic vs mantle) introduce uncertainty; some PAs with significant mantle helium precluded use of He_rad ages or required corrections for crustal helium flux. - Bimodality and error: Higher uncertainties are associated with bimodal age distributions; median mean-age error was 9% (IQR 5–25%). - Representativeness: ACDFs represent spatially averaged age distributions at the depth zone used for public supply; they are not predictive for individual wells and may not represent other parts of the aquifer system. - Scope: Sampling targeted public-supply wells with generally deep and long screened intervals; findings may differ for shallow domestic wells or different completion depths. Unsaturated-zone travel times were incorporated primarily for Anthropocene waters and may vary locally.