Hydrogeology Journal

, 19:779

A comparison of recharge rates in aquifers of the United States based on groundwater-age data

Authors

    • US Geological Survey
  • L. N. Plummer
    • US Geological Survey
  • J. K. Böhlke
    • US Geological Survey
  • S. D. Shapiro
    • US Geological Survey
  • S. R. Hinkle
    • US Geological Survey
Paper

DOI: 10.1007/s10040-011-0722-5

Cite this article as:
McMahon, P.B., Plummer, L.N., Böhlke, J.K. et al. Hydrogeol J (2011) 19: 779. doi:10.1007/s10040-011-0722-5

Abstract

An overview is presented of existing groundwater-age data and their implications for assessing rates and timescales of recharge in selected unconfined aquifer systems of the United States. Apparent age distributions in aquifers determined from chlorofluorocarbon, sulfur hexafluoride, tritium/helium-3, and radiocarbon measurements from 565 wells in 45 networks were used to calculate groundwater recharge rates. Timescales of recharge were defined by 1,873 distributed tritium measurements and 102 radiocarbon measurements from 27 well networks. Recharge rates ranged from < 10 to 1,200 mm/yr in selected aquifers on the basis of measured vertical age distributions and assuming exponential age gradients. On a regional basis, recharge rates based on tracers of young groundwater exhibited a significant inverse correlation with mean annual air temperature and a significant positive correlation with mean annual precipitation. Comparison of recharge derived from groundwater ages with recharge derived from stream base-flow evaluation showed similar overall patterns but substantial local differences. Results from this compilation demonstrate that age-based recharge estimates can provide useful insights into spatial and temporal variability in recharge at a national scale and factors controlling that variability. Local age-based recharge estimates provide empirical data and process information that are needed for testing and improving more spatially complete model-based methods.

Keywords

Groundwater ageGroundwater recharge/water budgetUSA

Comparaison des taux de recharge d’aquifères basée sur des données d’âge de nappe, Etats Unis

Résumé

On présente une vue d’ensemble des données disponibles sur l’âge d’une nappe et leurs implications pour évaluer les taux et temps de recharge d’aquifères libres sélectionnés, Etats Unis. Les distributions d’âge apparent dans des aquifères, déduites du dosage de carbone chloro-fluorés, hexafluorure de soufre, tritium/helium-3 et radiocarbone dans 565 puits de 45 réseaux, ont été utilisés pour calculer les taux de recharge de la nappe. Les échelles des temps de recharge ont été définies par 1,873 mesures de tritium distribuées et 102 mesures de radiocarbone sur un réseau de 27 puits. La distribution des temps de rechargea été établie par 1,873 mesures du tritium et 102 mesures du radiocarbone sur un réseau de 27 puits. Les taux de recharge s’échelonnent de moins de 10 mm/an à 1,200 mm/an dans des aquifères sélectionnés sur la base des distributions verticales d’âge mesurées et supposant des gradients d’âge exponentiels. Sur une base régionale, les taux de recharge donnés par des traceurs sur une eau de nappe jeune montrent une corrélation inverse forte avec la température annuelle moyenne de l’air et une corrélation directe forte avec la précipitation annuelle moyenne. La comparaison de la recharge déduite des âges de la nappe avec la recharge déduite du débit de base montre des caractéristiques générales similaires mais des différences locales notables. Des résultats de cette compilation démontrent que des estimations basées sur l’âge peuvent fournir des indications utiles sur la variabilité spatiale et temporelle de la recharge à une échelle nationale, et sur les facteurs contrôlant cette variabilité. Des estimations de recharge basées sur un âge local fournissent donnée empirique et information requises pour tester et améliorer des méthodes d’avantage basées sur la modélisation spatiale.

Una comparación de los ritmos de recarga en acuíferos de los Estados Unidos basados en datos de edad del agua subterránea

Resumen

Se presenta una visión general de los datos existentes de edad de agua subterránea y su implicancia para evaluar los ritmos y las escalas temporales de la recarga en sistemas acuíferos no confinados seleccionados de Estados Unidos. Se utilizaron las distribuciones de edad aparente en los acuíferos determinada a partir de medidas clorofluorocarbonos, hexafluoruro de azufre, tritio/helio-3, y medidas de radiocarbono de 565 pozos en 45 redes para calcular los ritmos de la recarga de las aguas subterráneas. Las escalas temporales de la recarga fueron definidas a través de 1,873 medidas de tritio distribuidas y 102 medidas de radiocarbono a partir de 27 redes de pozos. Los ritmos de recarga variaron de < 10 a 1,200 mm/año en los acuíferos seleccionados en base a la distribución de medidas de edades verticales y suponiendo gradientes exponenciales de edad. Sobre una base regional, el ritmo de recarga basado en trazadores de agua subterráneas jóvenes, exhibieron una correlación significativa inversa con la temperatura media anual del aire y una correlación significativa positiva con la precipitación media anual. La comparación de la recarga a partir de las edades de agua subterránea con la recarga a partir de la evaluación del flujo base de las corrientes mostraron un patrón general similar pero con diferencias locales sustanciales. Los resultados de esta compilación demuestran que las estimaciones de la recarga basada en la edad pueden proveer puntos una comprensión útil de las variabilidades espaciales y temporales en la recarga en una escala nacional y los factores que controlan esa variabilidad. Las estimaciones de la recarga basada en edades locales provee datos empíricos e información de procesamiento que son necesarios para probar y mejorar más espacialmente los métodos completos basados en modelos.

Uma comparação das taxas de recarga em aquíferos dos Estados Unidos, baseada em dados de datação da água subterrânea

Resumo

Apresenta-se uma revisão de dados de datação de água subterrânea existentes, com vista à avaliação de graus e escalas temporais de recarga numa selecção de sistemas aquíferos livres dos Estados Unidos. Para calcular as taxas de recarga de água subterrânea foram usadas distribuições de idade aparente de aquíferos, determinadas com recurso a medidas de clorofluorocarboneto, hexafluoreto de enxofre, trítio/hélio-3 e radiocarbono em 565 poços de 45 campos de captação. As escalas temporais da recarga foram definidas por 1,873 medições distribuídas de trítio e 102 medições de radiocarbono em 27 campos de captação. As taxas de recarga variaram de <10 até 1,200 mm/ano em aquíferos seleccionados com base em medidas da distribuição vertical da idade e assumindo gradientes de idade exponenciais. Numa base regional, as taxas de recarga calculadas com base em traçadores em águas subterrâneas recentes mostrou uma significativa correlação inversa com a temperatura média anual do ar e uma correlação positiva significativa com a precipitação média anual. A comparação da recarga derivada das idades da água subterrânea com a recarga avaliada a partir do escoamento fluvial de base mostrou padrões gerais similares mas também substanciais diferenças locais. Os resultados desta compilação demonstram que as estimativas de recarga baseadas na idade podem fornecer perspectivas úteis sobre a variabilidade espacial e temporal da recarga a uma escala nacional e sobre os factores que controlam essa variabilidade. As estimativas de recarga baseadas na idade local fornecem dados empíricos e informação sobre processos que são necessários para testar e melhorar os métodos baseados em modelos espacialmente mais completos.

Introduction

Well-constrained water budgets are needed to assess groundwater availability and manage aquifers sustainably throughout the world (Healy et al. 2007; Reilly et al. 2008). Recharge is perhaps the most difficult water-budget component to quantify because of its spatial and temporal variability (Tyler et al. 1996; Wolock 2003; Scanlon et al. 2006; Crosbie et al. 2010). Several tools, including environmental tracers of groundwater age, are available for quantifying recharge and each has advantages and limitations (see review by Scanlon et al. 2002). Groundwater-age distributions giving vertical groundwater velocities provide relatively direct measures of recharge (Solomon and Sudicky 1991), whereas many other techniques for estimating recharge such as environmental and applied tracers in the unsaturated zone, numerical hydraulic modeling, water-table fluctuations, stream base-flow separation, and various other types of water-budget analyses, are relatively indirect measures of recharge. Groundwater ages also define timescales of recharge processes that could be used as relative measures of aquifer sustainability (Darling et al. 1997; Douglas et al. 2007; Bethke and Johnson 2008; Gates et al. 2008) or aquifer susceptibility (Böhlke 2002; Manning et al. 2005; Osenbrück et al. 2006; Burow et al. 2007; McMahon et al. 2008a). The interpretation of environmental-tracer data can be complicated by processes that affect tracer concentrations in recharge and groundwater along flow paths such as degradation, contamination, sorption, degassing, mixing, gas and water transport in thick unsaturated zones, rock–water interactions, and a decline or variability in atmospheric concentrations of tracers (Solomon and Cook 2000; Kalin 2000; International Atomic Energy Agency 2006). A further limitation is the lack of readily available methods for measuring groundwater ages between about 50 and 1,000 years, although argon-39 may have some applicability for dating groundwater in this age range (Loosli 1982).

Numerous studies have used groundwater-age distributions to estimate recharge in aquifers at the local scale (for example, Vogel 1967; Schlosser et al. 1989; Verhagen 1992; Solomon et al. 1995; Böhlke and Denver 1995; Cook et al. 1995; Le Gal Le Salle et al. 2001; Puckett et al. 2002; Plummer et al. 2004; Böhlke et al. 2002, 2007a, b; Green et al. 2008a; McMahon et al. 2004a, 2007). Relatively few studies have compiled groundwater-age data for the purpose of comparing age-based estimates of recharge between aquifers (for example, Böhlke 2002; Coes et al. 2007; Green et al. 2008a). The purpose of this paper is to compare recharge rates in aquifers across the United States based on groundwater-age data. This paper represents an initial attempt at synthesizing groundwater-age data on a national scale for this purpose. An analysis of all published age-dating studies in the United States was beyond the scope of this study. Rather, the focus is on studies that provided relatively robust analyses of tracer data and, equally important, had sufficient ancillary information to place the data in hydrologic context.

Data sources and analysis

Recharge rates in unconfined aquifers

Recharge rates in unconfined aquifers were calculated from groundwater-age and depth data in two types of well networks: multi-level (nested) wells and wells screened near the water table. In most cases, nested wells in a given aquifer were aligned along the same or similar flow paths, whereas the water-table wells were randomly located. Individual well networks of both types generally were located in areas dominated by a single land use (for example, agriculture, urban, or forest). Apparent ages for both types of networks were obtained from the literature and unpublished reports of the US Geological Survey (USGS) Chlorofluorocarbon Laboratory in Reston, Virginia. Radiocarbon (14C) or a combination of chlorofluorocarbons (CFCs), sulfur hexafluoride (SF6), and tritium/helium-3 (3H/3He) data were used in each of the selected studies to determine apparent groundwater age. The apparent ages and supporting ancillary information are given in the electronic supplementary material (ESM; Tables ESM1 and ESM2). Some of the published reports included age-based estimates of recharge. In order to provide some consistency to the recharge rates reported in this study, they were recalculated using the published apparent ages and methods outlined in the following.

Nested wells

Recharge rates for the nested wells were calculated using the approach of Vogel (1967), which assumes a logarithmic age-depth distribution (Solomon et al. 1995; Cook and Böhlke 2000):
$$ R{\text{ }} = \left( {n\cdot Z/t} \right)\cdot \ln \left( {Z/\left( {Z - z} \right)} \right) $$
(1)
where R is recharge rate, n is porosity, Z is thickness of the aquifer saturated zone, t is the tracer-based groundwater age of a sample, and z is depth of the sample (midpoint of the well screen) below the water table. In most cases, published values of n and Z for the study areas were used, but in some instances assumed values had to be used (published and assumed values are distinguished in Tables ESM1 and ESM2). A single value of R was calculated for each nested-well network. Nonlinear least squares regression was used to fit the computed groundwater-age profile from Eq. (1) to the data (Green et al. 2008a). All of the selected nested-well networks had sufficient information on well location within the flow system to select only those wells located in recharge areas. This approach for calculating recharge is subject to several assumptions, including that the unconfined aquifers received distributed recharge and have constant porosity and thickness, and that individual samples have discrete ages. Mean age or residence time (τ) of groundwater in the sampled flow systems is given by:
$$ \tau = n \cdot Z/R $$
(2)

Water-table wells

Recharge rates for the water-table wells were calculated by assuming age was 0 at the water table and was equal to the tracer-based age at the well midpoint, and that the age gradient between the two points was linear (Cook and Böhlke 2000):
$$ R = \left( {n\cdot z} \right)/t $$
(3)
A value of R was calculated for each well in the water-table well networks (Table ESM2). It was assumed that the wells were located in recharge areas but information was not available for most of the water-table well networks to verify that assumption. Application of Eq. (3) to shallow wells in discharge areas would underestimate recharge rates because, presumably, shallow water in discharge areas would have older apparent ages than shallow water in recharge areas. Mean age or residence time of groundwater was not calculated for the water-table wells because of the general lack of data on aquifer thickness (Z) at those sites.

Tracer comparability

No one type of tracer data was collected at all of the well networks containing young groundwater (< 50 years old), which means comparisons were made between recharge rates derived from various combinations of CFC, SF6, and 3H/3He apparent ages. Studies have shown that ages based on these tracers generally compared well to each other as long as the tracers were suited to the environment in which they were collected (Ekwurzel et al. 1994; Szabo et al. 1996; Busenberg and Plummer 2000). In other words, there was no inherent bias associated with any of the tracers in many unconsolidated aquifers. Geologic heterogeneity and hydraulic dispersion can cause different age tracers to give significantly different apparent ages in some situations. These effects are most likely in fractured rock and karst carbonate aquifers, and some heterogeneous unconsolidated aquifers. Tracer comparability also could be problematic if data for one or more of the tracers were collected from an unsuitable environment. For example, CFCs generally are not suitable age tracers in highly anoxic groundwater systems because of their susceptibility to degradation under those conditions, a process that could result in ages being biased old and recharge rates being biased low. Thus, known environmental limitations associated with the CFC, SF6, and 3H/3He tracers (Solomon and Cook 2000; International Atomic Energy Agency 2006) were used as screening criteria in the data compilation process.

Timescales of recharge in unconfined aquifers from radiocarbon and tritium data

In addition to the age-gradient data described in the previous, databases of 14C and 3H concentrations were used to assess timescales of groundwater recharge. The 14C and 3H timescales for groundwater studies generally span the past 30,000 and 60 years, respectively. Data for both tracers were obtained from published reports and the USGS National Water Information System (NWIS) database. The 14C and 3H data, data sources, and supporting ancillary information are listed in Tables ESM3, ESM4a–d, and ESM5. The data were grouped by principal aquifer. The USGS has defined 62 principal aquifers in the United States (Reilly et al. 2008; US Geological Survey 2009), and those definitions are used in this paper.

Radiocarbon timescale in unconfined aquifers

Apparent groundwater ages derived from 14C data were used to determine groundwater velocities and first occurrences of pre-Holocene recharge in each sampled flow system. Pre-Holocene recharge is defined here as groundwater with a radiocarbon age > 10,000 14C years (Stuiver et al. 1998; Orndorff et al. 2007). Given this focus, groups of wells located along the same or similar flow paths were chosen for the analysis. Wells used in this assessment were required to have information on their depth below the water table to the top of the open interval. Six of the eight selected data sets had published 14C ages and those ages were used in this analysis without further refinement. The adjustment models of Ingerson and Pearson (1964), Mook (1972), Tamers (1975), Fontes and Garnier (1979), and Eichinger (1983), as implemented in NETPATH (Plummer et al. 1994), were used to calculate 14C ages in the two remaining data sets, which were from the Coastal Basins and Central Valley aquifer systems of California. Both aquifers are composed of unconsolidated sediments. Details of those calculations are given in the ESM (Tables ESM4a–d).

Two issues need to be recognized when comparing previously published 14C data. The first issue has to do with the 14C activity units 1) absolute percent Modern (pM) and (2) percent modern carbon (pmC). For pM, 14C activity measurements have been normalized for 13C fractionation, whereas for pmC, 14C activity measurements have not been normalized for 13C fractionation. The preferred measurement unit for groundwater dating is pmC (Mook and van der Plicht 1999; Plummer et al. 2004), but it is not always indicated in the literature which unit has been used. Fortunately, the difference in calculated ages for the two units typically is less than 500 14C years, which generally is less than the uncertainty in the 14C age calculation (Plummer et al. 1990). The second issue has to do with the 14C half-life. Use of the Libby half-life (5,568 years) allows for direct comparison of 14C ages to radiocarbon calibration scales (Stuiver et al. 1998; Plummer et al. 2004). The modern half-life (5,730 years) appears to be the most commonly used half-life in the published data compiled for this study, although not all studies indicated which half-life was used. Radiocarbon ages based on the two half-lives are related by (Plummer et al. 2004):
$$ {t_{\text{Libby}}} = 0.{972}{t_{{{573}0}}} $$
(4)

For 10,000-year-old water, this difference in half-lives only amounts to 280 years (2.8%), which again is less than the uncertainty in most 14C age calculations. Thus, although there could be inconsistencies in the data sets with respect to 14C measurement units and half-lives used in age calculations, they are not considered to be large enough to change the interpretations presented here. Table ESM3 includes information about the 14C units (pM or pmC) and assumed half-lives (5,568 or 5,730 yrs) where available.

Tritium timescale in unconfined aquifers

Tritium data were used to examine the occurrence and distribution of post-1950s recharge in selected unconfined aquifer systems. For the most part, wells in the 3H data sets were widely distributed and not necessarily located along the same or related flow paths. Water samples used in this study were collected between March 1975 and October 2007 so the 3H data were assessed in two ways. First, measured 3H concentrations were decayed to 1 January 2010 to enable broad comparisons of 3H concentrations between aquifers. Before the onset of widespread atmospheric testing of nuclear weapons in late 1952, 3H in precipitation of the United States probably ranged from about 2 to 8 TU (Kaufman and Libby 1954; Thatcher 1962). The maximum pre-bomb 3H concentration of 8 TU would be < 0.5 TU in January 2010; therefore, water samples with decayed 3H concentrations ≥ 0.5 TU were interpreted as having components of post-1950s recharge. Second, measured 3H concentrations were grouped in 5-year intervals of the date of sample collection and plotted against depth below the water table to the top of the open interval of the well to examine vertical distributions of post-1950s recharge in the aquifers. For wells sampled repeatedly, this approach can detect downward migration of the 3H bomb peak in some cases.

Results and discussion

Data from 18 of the 62 principal aquifer systems of the United States (plus selected alluvial aquifers) were used in this comparison of recharge rates. General locations of the well networks in the principal aquifers are shown in Fig. 1. None of the well networks was completely representative of an entire aquifer system, but the geographic distribution of sites provided the opportunity to compare rates and timescales of recharge between aquifers for a variety of climatic and hydrogeologic conditions. Reported data consisted of either (1) “apparent” groundwater ages (assuming no mixing or dispersion) derived from various combinations of measurements of CFCs, SF6, 3H/3He, and 14C for individual wells, or (2) occurrences of age-related tracers such as 14C and 3H.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig1_HTML.gif
Fig. 1

Location of principal aquifer systems and well networks included in this study. See Tables 14 for names of aquifers associated with each well network. For the most part, the tritium networks consisted of widely scattered wells. See Fig. ESM1 (electronic supplemental material) for the location of wells in those networks

Recharge rates based on age data from nested-well networks in unconfined aquifers

The most direct and potentially most accurate estimates of vertical groundwater velocity and recharge rate summarized in this paper were derived from environmental-tracer data from short-screen wells at various depths in unconfined aquifers. In the absence of detailed flow and transport simulations, recharge rates were estimated from vertical distributions of apparent groundwater ages in unconfined aquifers represented by 29 nested-well networks containing 327 wells (Fig. 1 and Table 1). Recharge rates, calculated using Eq. (1), ranged from < 10 mm/yr at semi-arid rangeland networks in the Rio Grande and central High-Plains aquifer systems to 1,200 mm/yr at an irrigation-dominated agriculture network in the northern High Plains (Table 1). The mean residence time (τ) of groundwater, calculated using Eq. (2), ranged from 6 to 19,000 years at the irrigation-dominated agriculture and rangeland networks in the northern High Plains and Rio Grande aquifer systems, respectively (Table 1).
Table 1

Recharge rates based on groundwater-age data from selected nested-well networks in unconfined aquifers. Map reference numbers refer to Fig. 1. See Table ESM1 for apparent ages and supporting information. MAP mean annual precipitation, 1971–2000; National Climatic Data Center 2010

Map reference number

Principal aquifer system

General land usea

Age tracer

Recharge rateb (mm/yr)

Recharge as percentage of MAP

Mean residence time (yr)

Data source

1

Alluvium

2

CFC

620

64c

7

McMahon and Böhlke 1996

2

Central Valley

2

CFC

580

52c

39

Burow et al. 1999, 2007

3

Central Valley

2

SF6, CFC

420

28c

28

Green et al. 2008a, b

4

Glacial Deposits (eastern glaciated region)

1

SF6, CFC, 3H/3He

470

38

24

Böhlke et al. 2009

5

Glacial Deposits (eastern glaciated region)

3

3H/3He

690

56

17

Solomon et al. 1995

6

Glacial Deposits (eastern glaciated region)

3

3H/3He

520

44

13

Mullaney and Grady 1997

7

Glacial Deposits (central glaciated region)

3

3H/3He

750

75

23

Shapiro et al. 1998

8

Glacial Deposits (central glaciated region)

3

3H/3He, SF6

170

22

62

USGS unpublished data; Thomas 2000

9

Glacial Deposits (central glaciated region)

1

CFC

95–180

14–27

40–76

Puckett et al. 2002; Puckett and Cowdery 2002

10

Glacial Deposits (central glaciated region)

1

CFC

230

34

17

Delin et al. 2000; Böhlke et al. 2002

11

Glacial Deposits (central glaciated region)

1

CFC

120

18

38

Delin et al. 2000; Böhlke et al. 2002

12

Glacial Deposits (central glaciated region)

1

SF6, CFC

72

9.5

180

Green et al. 2008a, b

13

High Plains (northern)

3

SF6, CFC

100–160

14–23

42–68

Landon et al. 2008

14

High Plains (northern)

5

14C

20

3.9

2,700

McMahon et al. 2007

15

High Plains (northern)

2

3H/3He

380–1,200

--

6–20

Böhlke et al. 2007a

16

High Plains (central)

5

14C

7

1.4

3,800

McMahon et al. 2004a

17

Northern Atlantic Coastal Plain

1

CFC

100

8.3

33

Tesoriero et al. 2005, 2007

18

Northern Atlantic Coastal Plain

1

SF6, CFC

270

24

13

Lindsey et al. 2003

19

Northern Atlantic Coastal Plain

1,4

CFC

64

5.6

39

Böhlke et al. 2007b

20

Northern Atlantic Coastal Plain

1

CFC

300

26

31

Dunkle et al. 1993; Böhlke and Denver 1995

21

Northern Atlantic Coastal Plain

1

SF6, CFC

270

24

22

Green et al. 2008a, b

22

Northern Atlantic Coastal Plain

3

3H/3He

400

35

31

Szabo et al. 1996; Stackelberg et al. 2000; Kauffman et al. 2001

23

Pacific Northwest basin fill

2

SF6, CFC

200

21c

190

Green et al. 2008a, b

24

Pacific Northwest basin fill

6

CFC

54

18

210

Hinkle et al. 2007

25

Puget-Willamette Lowland

1

3H/3He

650

56

12

Wassenaar et al. 2006

26

Puget-Willamette Lowland

3

3H/3He

290

31

80

Hinkle 2009

27

Rio Grande (eastern mountain front, Albuquerque vicinity)

5

14C

8

3.3

19,000

Plummer et al. 2004

28

Rio Grande (eastern mountain front, south of Tijeras Arroyo)

5

14C

2d

0.3

--

Plummer et al. 2004

29

Surficial

1

CFC

150

13

26

Puckett and Hughes 2005

a1 precipitation-dominated agriculture; 2 irrigation-dominated agriculture; 3 residential/commercial; 4 forest; 5 rangeland; 6 rural residential

bCalculated using Eq. (1), except where noted

cMean annual precipitation plus applied irrigation water. Amounts of applied irrigation water from the indicated references

dAssumes linear age gradient

Recharge rates are not likely to be at steady-state at the timescale of CFC, SF6, and 3H/3He tracers because recharge can be affected by decadal, annual, and shorter-term variations in climate and land use. Similarly, recharge rates are not likely to be at steady-state at 14C timescales because recharge also can be affected by longer-term variations in factors such as climate, sea level, and glacial ice cover. Thus, the apparent recharge rates summarized here are considered average values at the timescales of the selected age tracers.

Median recharge rates in networks grouped by land use increased according to rangeland (8 mm/yr), precipitation-dominated agriculture (190 mm/yr), residential/commercial (400 mm/yr), and irrigation-dominated agriculture (580 mm/yr). Recharge rates at the rangeland networks were significantly (α = 0.05) smaller than rates at networks in the other three land uses on the basis of the nonparametric Wilcoxon rank test (Helsel and Hirsch 1992). Recharge rates at the precipitation-dominated agriculture networks were significantly smaller than rates at the irrigation-dominated agriculture networks. Recharge rates at the precipitation-dominated agriculture and residential/commercial networks were not significantly different from each other at α = 0.05 (but they were at α = 0.10). Likewise, recharge rates at the residential/commercial and irrigation-dominated agriculture networks were not significantly different from each other.

Excluding the irrigation-dominated agriculture networks, the median recharge rate as a percentage of mean annual precipitation (1971–2000; National Climatic Data Center 2010) was 22% and ranged from < 5% at the semi-arid rangeland networks to > 50% at three networks in New England, the upper Midwest, and the Pacific Northwest where the climate was relatively cool and wet (networks 5, 7, 25; Table 1). For the irrigation-dominated agriculture networks, recharge was 21–64% of precipitation plus applied irrigation water, with a median of 40%.

Comparison of age gradients with other methods of recharge estimation

Estimates of recharge based on groundwater apparent-age profiles in unconfined aquifers were compared to published estimates of recharge for the same general study areas based on four independent methods; water balance, stream base-flow separation, local groundwater-flow models, and water-table fluctuation (Fig. 2). Overall, there was a positive correlation between age-based estimates of recharge and recharge determined using the other methods, but with considerable scatter. Recharge estimates from local groundwater-flow models appeared to be in closest agreement with the age-based estimates. In some cases, this is because flow models were calibrated in part with environmental-tracer data (Reilly et al. 1994; Mullaney and Grady 1997; McMahon et al. 2004a). Agreement between age-based recharge and recharge from the water-table fluctuation and base-flow separation methods was relatively poor, but there did not appear to be any systematic bias between these estimates (Fig. 2), although in the case of base-flow separation there were only data from three networks. In contrast, various water-balance estimates of recharge exceeded the age-based estimates at seven of eight networks.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig2_HTML.gif
Fig. 2

Comparison of recharge rates based on groundwater age with recharge rates from four alternate methods. Numbers in parentheses are map reference numbers (Table 1 and Fig. 1). Data sources for alternate recharge methods are as follows: water balance (Burow et al. 1999; Cox and Kahle 1999; Delin et al. 2000; Dugan and Zelt 2000; Steenhuis et al. 1985), Stream base-flow separation (Charles et al. 2001; Lacombe and Rosman 1995; Cushing et al. 1973; Coes et al. 2007), Local groundwater-flow model (Gannett et al. 2001; McMahon et al. 2004a; Morgan et al. 2007; Mullaney and Grady 1997; Reilly et al. 1994; Snyder et al. 1994; Szabo et al. 1996), Water-table fluctuation (Cox and Kahle 1999; Delin et al. 2000; Fisher and Healy 2008; Green et al. 2008a)

Independent local estimates of recharge were not available at all the networks so age-based estimates of recharge from all the networks were compared to national-scale estimates of natural recharge determined using a base-flow index method (Wolock 2003). Base-flow index is the ratio of base flow to total flow in a stream. Wolock (2003) calculated natural recharge at a national scale by multiplying a gridded set of base-flow index values from unregulated streams in the United States by a gridded set of long-term average streamflows. As expected, age-based estimates of recharge at irrigation-dominated agriculture networks were larger than the base-flow index estimates in all cases (Fig. 3) because the streamflow data used by Wolock (2003) were from watersheds with relatively undisturbed water balances. Overall, there was a general positive correlation between groundwater age-based recharge and recharge from the base-flow index method at the precipitation-dominated networks (Fig. 3), although with considerable scatter as was observed with the local estimates of recharge in Fig. 2. For the precipitation-dominated networks, the median difference between recharge based on groundwater age and base-flow index was positive (Fig. 3), implying an underestimation of total recharge on the part of the base-flow index method.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig3_HTML.gif
Fig. 3

Comparison of recharge rates based on groundwater age with recharge rates from the base-flow index method of Wolock (2003). Note: Data of Wolock (2003) were derived from “natural” stream flows

Although it is difficult to evaluate the significance of this apparent bias given the difference in scale of the two methods, the groundwater-age and base-flow methods do have important differences in what they represent so additional local comparisons could help resolve distinctions between different definitions of “recharge” and “base flow.” Groundwater age-based estimates of recharge would be expected to be larger than base-flow estimates if they included recharge that was partitioned to deep flow systems, as the latter estimate might only measure the component of recharge partitioned to the shallow flow system, or if recharge was lost to evapotranspiration prior to reaching a stream. In addition, although the contribution of groundwater discharge during high-flow conditions is highly variable, there is evidence to suggest it can be significant (Sklash and Farvolden 1979; Genereux and Hooper 1998). Automated flow-based hydrograph separation approaches may not always represent those contributions accurately.

As noted by Schaller and Fan (2009), most current climate models route basin recharge (precipitation–evapotranspiration) directly to local streams, lakes, and wetlands. Data from this study, and results from the study of Schaller and Fan (2009), indicate that this assumption may not be valid in some environments. Schaller and Fan (2009) found that many basins they studied in the Atlantic and Gulf Coastal Plains had a component of recharge partitioned to deep flow systems in regionally dipping aquifers of the coastal plain. In this study, age-based recharge exceeded recharge from the base-flow index method in four of five nested-well networks in the Atlantic Coastal Plain for which data were available. Several studies have identified important links between regional groundwater flow and continental climate dynamics (Bierkens and van den Hurk 2007; Anyah et al. 2008; Schaller and Fan 2009), so quantifying regional versus local components of recharge could be important for climate modeling. Local comparisons of recharge from shallow groundwater-age and base-flow methods may be a useful approach for detecting recharge to deeper flow systems, provided the base-flow methods accurately capture the ground-water contribution.

Effect of regional climate variability on recharge

Nested-well networks in precipitation-dominated settings were grouped by mean annual precipitation and air temperature to evaluate effects of regional climate variability on recharge. Ten networks located along the east coast of the United States had roughly similar mean annual precipitation (1,093–1,259 mm/yr) but a range of mean annual air temperatures from 9.9 to 17.5°C (Fig. 4). Thirteen networks located in the northern United States had mean annual air temperatures < 12°C (5.4–11.9°C) and mean annual precipitation amounts ranging from 298 to 1,230 mm/yr (Fig. 5). Recharge along the east coast exhibited a significant inverse correlation with air temperature, presumably because of decreasing evapotranspiration with decreasing temperature and (or) growing season length (Fig. 4). Recharge across the northern United States exhibited a significant positive correlation with precipitation (Fig. 5), although irrigation-dominated agriculture networks did not fit that trend. The apparently strong regional climate-recharge correlations in Figs. 4 and 5 may be due in part to the fact that the age-based method used to estimate recharge integrates processes over large spatial and temporal scales compared to some other methods such as those focused on the unsaturated zone, thereby smoothing to some extent the effect of local variability in factors like sediment texture, topography, and vegetation on recharge.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig4_HTML.gif
Fig. 4

Recharge rate in relation to mean annual air temperature at nested-well networks along the east coast of the United States. Climate averages are for 1971–2000 (National Climatic Data Center 2010). Networks were divided into three major sediment-texture categories based on descriptions of aquifer sediment in the published reports from which the data were compiled. Spearman correlation statistics are shown. The data sets are in Table ESM1. Sources of groundwater-age data are shown next to each network name

https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig5_HTML.gif
Fig. 5

Recharge rate in relation to mean annual precipitation at nested-well networks in the northern United States. Climate averages are for 1971–2000 (National Climatic Data Center 2010). Networks were divided into three major sediment-texture categories based on descriptions of aquifer sediment in the published reports from which the data were compiled. Spearman correlation statistics are shown. The data sets are in Table ESM1. Sources of groundwater-age data are shown next to each network name

Networks in Figs. 4 and 5 were further divided into three major sediment-texture categories (sand/silt/clay, sand, sand/gravel) based on descriptions of aquifer sediment in the published reports from which the data were compiled. The recharge-climate trends within individual texture categories generally appear to be consistent with the overall regional patterns in Figs. 4 and 5. A notable exception to this is network 19, which plots below the overall trend in Fig. 4. Networks 19 and 20 are both located in agricultural watersheds in the Atlantic Coastal Plain of Maryland, but the hydraulic conductivity of surficial aquifer sediment at network 19 is at least two orders of magnitude smaller than at network 20 (Reilly et al. 1994; Böhlke et al. 2007b). Age-based recharge at network 19 was about five times smaller than at network 20 (Fig. 6a).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig6_HTML.gif
Fig. 6

Apparent groundwater age in relation to depth below the water table for nested-well networks in the a Northern Atlantic Coastal Plain aquifer system of Maryland and b Glacial Deposits aquifer system of Minnesota. The lines represent model fits (Eq. 1) to the data (see text for details). Network numbers are map reference numbers (Fig. 1). The data sets are in Table ESM1. See Table 1 for data sources

Local differences in topography apparently contributed to the variability in Fig. 5. At networks 10 and 11, located in the same agricultural watershed in Minnesota, recharge was about 2 times larger at a lowland site than at an upland site (Fig. 6b), presumably because of focused recharge in the topographic depression (Delin et al. 2000; Böhlke et al. 2002).

Additional unexplained variability in the correlations in Figs. 4 and 5 could be caused by land-use effects because the data were from networks in both residential/commercial and precipitation-dominated agricultural settings. However, as discussed previously, there was not a significant difference in recharge rates between those two land uses so their effects on the overall patterns in Figs. 4 and 5 are expected to be relatively small compared to the climate effects. That may not be the case for some other land uses. As reviewed by Scanlon et al. (2006), numerous studies have shown land-use change can have a greater impact on recharge than climate variability, particularly where natural ecosystems are replaced by cropland in semi-arid regions. In this study, the apparent climate effects on recharge illustrated in Fig. 5 were overwhelmed at irrigated sites in semi-arid settings.

The regional climate-recharge correlations presented here are consistent with those of Nolan et al. (2007) for the eastern United States. In that study, recharge determined by saturated-zone chloride mass balance at 108 sites in nine states was significantly correlated with precipitation and air temperature. Recharge also was significantly correlated with soil characteristics such as percent sand and with the amount of residential land use (Nolan et al. 2007), but multivariate nonlinear regression analysis by Nolan et al. (2007) indicated both those factors were less influential than the climate factors on recharge at the regional scale of their analysis. The study presented here extends the findings of Nolan et al. (2007) westward from eastern Nebraska to the Pacific Northwest.

Comparison of mean groundwater residence times

Mean groundwater residence times for the 29 nested-well networks, calculated using Eq. (2), are summarized in Fig. 7. Networks in the semi-arid High Plains and Rio Grande aquifer systems had the largest residence times (2,700–19,000 years). Residence times were considerably smaller in networks from the other aquifer systems (7–210 years). The range of residence times in aquifer systems containing more than 2 networks increased as follows: Northern Atlantic Coastal Plain (1339 years), Glacial (13–185 years), and High Plains (6–3,750 years) (Fig. 7). It is worth noting that these residence-time estimates represent unconfined flow systems that are primarily shallow and contain young groundwater. They are not presumed to be representative of the entire principal aquifer in which they are located. Some regional aquifer systems contain water recharged at multiple timescales, a topic which is discussed later in this paper.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig7_HTML.gif
Fig. 7

a Mean groundwater residence time in relation to aquifer thickness at the nested-well networks (shown in b). Delineated areas in the plots show corresponding recharge rates for a selected range of porosity values. See Table 1 for key to index numbers and data sources

Besides variations in aquifer thickness, which in some cases are not well constrained, the ranges of residence time reported here also reflect the effects of climate, land use, and other factors on recharge, as previously discussed. Regional climate variations appear to have had a larger effect on residence time at networks in the Glacial aquifer system than in the Northern Atlantic Coastal Plain aquifer system on the basis of data in Figs. 4 and 5, possibly because of the broader geographic distribution of networks in the former aquifer system (Fig. 7). Variations in land use (from semi-arid rangeland to flood-irrigated agriculture) had a large effect on recharge/residence time at networks in the High Plains (Table 1 and Fig. 7).

Groundwater age or residence time has been used in several studies to evaluate groundwater sustainability and susceptibility (for example, Nelms et al. 2003; Plummer et al. 2004; Sanford et al. 2004; Zongyu et al. 2005; Rupert and Plummer 2009). The data in Fig. 7 imply that, of the aquifers studied here, the High Plains and Rio Grande aquifers in the vicinity of networks 14, 16, and 27 may be the least sustainable under the pressure of groundwater pumping because of their long residence times. Large groundwater-level declines have been documented in the vicinity of networks 16 (irrigation pumping, McGuire 2009) and 27 (public-supply pumping, Bexfield and Anderholm 2002). The High Plains aquifer near network 14 is relatively undeveloped at this time (2010). Aquifers like the South Platte, High Plains, and Puget-Willamette in the vicinity of networks 1, 15, and 25, respectively, may be some of the most susceptible to contamination from surface sources because of their short residence times (Fig. 7). Short residence times may also limit the reaction time of processes like denitrification, thereby reducing its ability to attenuate nitrate contamination in groundwater (Böhlke et al. 2007a; McMahon et al. 2008b). On the other hand, aquifers with short groundwater residence times may respond relatively quickly to management practices at the land surface intended to improve water quality (Wassenaar et al. 2006; McMahon et al. 2008a). In the case of the Puget-Willamette aquifer system in the vicinity of network 25, with a mean groundwater-residence time of 12 years, a change in agricultural management practices apparently had the unintended consequence of causing an increase in groundwater nitrate concentrations within a decade of implementation because under the new management plan growers shifted from using manure as fertilizer to using more leachable inorganic fertilizers (Wassenaar et al. 2006).

Recharge rates based on age data from water-table well networks

Recharge rates were estimated from apparent ages in 238 wells screened at or near the water table in 16 water-table well networks (Fig. 1 and Table 2). Median values of recharge in the water-table well networks ranged from 80 mm/yr at network 36 (precipitation-dominated agriculture in Wisconsin) to 580 mm/yr at network 33 (residential/commercial in Ohio), both of which were located in the central glaciated region of the United States (Fig. 1). The networks with the next three largest median values of recharge (430–540 mm/yr) were located in a precipitation-dominated agricultural setting in the Pacific Northwest (43) and irrigation-dominated agricultural settings in the northern High Plains (38) and Northern Rockies Intermontaine basins (42; Fig. 1 and Table 2). Overall median recharge rates for the two land uses with at least three networks were 163 mm/yr (precipitation-dominated agriculture) and 181 mm/yr (residential/commercial). Recharge rates in those two land-use settings were not significantly different from each other at α = 0.05, which is consistent with the land-use comparisons based on nested-well data. The median recharge rate for the two irrigation-dominated agriculture networks was 429 mm/yr. These estimates of recharge are potentially less reliable than those from the nested-well networks because age gradients derived from wells screened at the water table would be expected to have large uncertainties because of water table fluctuations and large fractional uncertainties in the age of young water. Moreover, recharge estimates from the water-table well networks may be more variable than those from the nested well networks because local-scale recharge variations associated with sediment texture, topography, vegetation, land use, and other factors are not spatially and temporally as well integrated at the water-table wells.
Table 2

Recharge rates based on groundwater-age data from selected water-table well networks. Map reference numbers refer to Fig. 1. See Table ESM2 for apparent ages and supporting information

Map reference number

Principal aquifer system

General land usea

Age tracer

Median recharge rateb (mm/yr) [range; number of sites]

Data source

30

Basin and Range basin fill

3

3H/3He

190 [39–1,200; 24]

Thiros 2003

31

Glacial Deposits (eastern glaciated region)

3

3H/3He, SF6

260 [47 to >850; 15]

USGS unpublished data

32

Glacial Deposits (central glaciated region)

3

3H/3He

170 [65–1,500; 15]

USGS unpublished data

33

Glacial Deposits (central glaciated region)

3

3H/3He

580 [23–>1,100; 9]

Shapiro et al. 1998; Rowe et al. 1999

34

Glacial Deposits (central glaciated region)

1

CFC

160 [83–300; 5]

USGS unpublished data

35

Glacial Deposits (central glaciated region)

3

3H/3He

170 [36–1,500; 11]

Morrow 2003

36

Glacial Deposits (central glaciated region)

1

CFC

80 [10–120; 7]

Saad 1997; Saad 2008

37

Glacial Deposits (central glaciated region)

1

CFC

210 [9 to >1,000; 18]

Saad 1997; Saad 2008

38

High Plains (northern)

2

3H/3He

430 [160 to >1500; 9]

Böhlke et al. 2007a

39

Northern Atlantic Coastal Plain

1

SF6

170 [42–630; 28]

Debrewer et al. 2007

40

Northern Atlantic Coastal Plain

1

CFC

140 [20–1,400; 47]

Dunkle et al. 1993

41

Northern Atlantic Coastal Plain

3

3H/3He, SF6

370 [160 to >1,000; 9]

Szabo et al. 1996; Stackelberg et al. 2000; Kauffman et al. 2001

42

Northern Rockies Intermontaine Basins

2

3H/3He, CFC

430 [98–9500; 10]

Pope et al. 1999

43

Puget-Willamette Lowland

1

SF6

540 [67–6,200; 12]

Wassenaar et al. 2006; USGS unpublished data

44

Southeastern Coastal Plain

3

SF6, CFC

120 [26–540; 12]

Robinson 2002; USGS unpublished data

45

Surficial

3

SF6

120 [8–250; 5]

Katz et al. 2007

a1 precipitation-dominated agriculture; 2 irrigation-dominated agriculture; 3 residential/commercial

bCalculated using Eq. (3)

Nested- and water-table well networks were co-located in six of the study areas (Fig. 1). Overall, there was a generally positive correlation in recharge rates between the networks, but rates from the nested-well networks were larger than those from the water-table well networks in five of the six study areas (Fig. 8a). The largest differences in recharge were between networks 15 and 38 and networks 20 and 39. For networks 15 and 38, co-located in an irrigation-dominated agricultural setting in western Nebraska, the difference in recharge rates appears to be related to the relative contributions of distributed recharge in irrigated fields (less recharge, network 38) and focused recharge near irrigation canals (more recharge, network 15) to the wells (Böhlke et al. 2007a). The deeper nested wells were more likely to intercept both types of recharge, whereas the water-table wells were more likely to intercept distributed recharge unless they were located adjacent to a canal. For networks 20 and 39, co-located on the Delmarva Peninsula in the Northern Atlantic Coastal Plain, the difference in recharge rates appears to be related in part to differences in soil drainage in the two areas. Network 39 covered an area of about 15,000 km2 (Debrewer et al. 2007), and soil in that area ranged from being poorly to well drained (Fig. 8b). The wells in network 20 only covered about 25 km2 (Böhlke and Denver 1995) of that larger area, and soil there was generally well drained (Fig. 8b). There was a significant inverse correlation between recharge from the water-table wells and soil-drainage index within a 500 m radius of the wells, indicating that the larger recharge rate at network 20 could be related to better soil drainage in that part of the study area compared to the study area as a whole.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig8_HTML.gif
Fig. 8

a Comparison of recharge rates based on groundwater age at co-located nested- and water-table well networks. b Recharge rate in relation to STATSGO soil-drainage index within a 500-m radius around wells for nested-well network 20 and water-table well network 39. The solid line (b) represents a LOWESS smooth of the water-table well data. Spearman correlation statistics are for the water-table wells. STATSGO data from Schwarz and Alexander (1995)

Radiocarbon timescale for recharge in unconfined aquifers

Several of the most heavily pumped aquifers in the United States such as the High Plains, Central Valley, Basin and Range basin-fill, and Rio Grande aquifer systems, are unconfined and contain pre-Holocene recharge at depth because of low natural recharge rates in these semi-arid to arid regions of the country (Maupin and Barber 2005; McMahon et al. 2004a; Izbicki and Michel 2004; Plummer et al. 2004) (Fig. 1 and Table 3). In these settings, the recharge rates measured using tracers of young groundwater at the nested and water-table well networks (Tables 1 and 2) often represent relatively recent changes in the hydrologic system associated with the onset of irrigation. Thus, it is important to recognize that these aquifer systems are influenced by at least two timescales for recharge. The older timescale for recharge measurable with radiocarbon tracers is examined for some of these western aquifer systems.
Table 3

Apparent vertical groundwater velocities and depths to the first measured occurrence of pre-Holocene recharge in unconfined aquifers. Map reference numbers refer to Fig. 1. See Tables ESM3 and ESM4a–d for apparent ages and supporting information

Map reference number

Principal aquifer system

Local aquifer name

Apparent vertical groundwater velocity (m/yr)

Depth below water table to first measured occurrence of pre-Holocene recharge (m)

Data source

46

Basin and Range basin fill

Eastern Morongo groundwater basin

< 2.5 × 10–4 a

125

Kulongoski et al. 2005; Izbicki and Michel 2004

47

California Coastal Basins

Southern California Coastal basin

> 4.8 × 10–2

> 130

USGS unpublished data; Hamlin et al. 2002

48

Central Valley

Central Valley

4.7 × 10–3 b

51

USGS unpublished data; Jurgens et al. 2008

14

High Plains

Northern High Plains

1.3 × 10–2 b

176

McMahon et al. 2007

16

High Plains

Central High Plains

7.0 × 10–3 b

83

McMahon et al. 2004a

27

Rio Grande

Santa Fe Group (eastern mountain front, Albuquerque vicinity)

2.0 × 10–2 b

195

Plummer et al. 2004

28

Rio Grande

Santa Fe Group (eastern mountain front, south of Tijeras Arroyo)

3.8 × 10–3 b

13

Plummer et al. 2004

49

Rio Grande

Santa Fe Group (west-central zone)

< 1.0 × 10–4 a

< 5

Plummer et al. 2004

aNo apparent recharge during the past 4,000 14C years at network 46 and the past 10,000 14C years at network 49

bApparent velocities based on slopes of linear regression lines

A total of 102 analyses in eight unconfined flow systems provided 14C apparent ages that could be interpreted as local vertical groundwater velocities (Fig. 1 and Table 3). The vertical distributions of 14C and apparent 14C ages of dissolved inorganic carbon in each flow system are shown in Fig. 9. Apparent vertical velocities in the aquifers ranged from < 1.0 × 10–4 to > 4.8 × 10–2 m/yr (Table 3 and Fig. 9). Maximum velocities were reported for networks 46 (Basin and Range basin fill aquifer system) and 49 (west-central zone of the Rio Grande aquifer system) because of the apparent lack of recharge in those areas during the past 4,000–10,000 years (Fig. 9). The maximum vertical velocities at networks 46 and 49 would equate to recharge rates < 1 mm/yr, assuming 30% porosity. A minimum velocity was reported for network 47 (California Coastal Basin) because the relatively flat vertical age gradient in that area could not be resolved using 14C data. The minimum velocity at network 47 would equate to a recharge rate > 14 mm/yr, assuming 30% porosity. At the five remaining networks vertical velocities were determined from the slope of linear regression lines through the age-depth data. Corresponding recharge rates at High Plains networks 14 and 16 and Rio Grande network 27 would be 4, 2, and 6 mm/yr, respectively, assuming 30% porosity. For comparison, recharge rates at the same networks calculated using exponential age-gradient models were 20, 7, and 8 mm/yr (Table 1), respectively, also assuming 30% porosity. All the networks contained some pre-Holocene recharge with the exception of network 47 in the California Coastal Basin (Fig. 9). In the Rio Grande aquifer system, New Mexico, depths below the water table to the first measured occurrence of pre-Holocene recharge ranged from < 5 m at network 49 (west-central zone) to 195 m at network 27 (eastern mountain front, Albuquerque vicinity; Table 3). In the remaining networks, depths below the water table to the first measured occurrence of pre-Holocene recharge ranged from 13 to 176 m.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig9_HTML.gif
Fig. 9

Vertical distributions of 14C and apparent 14C ages of dissolved inorganic carbon in selected unconfined aquifer systems. The solid lines represent LOWESS smooths of the 14C activities. The slopes of the dashed lines represent apparent vertical groundwater velocities. Numbers next to aquifer names are map reference numbers (Fig. 1). The data sets are in Table ESM3. See Table 3 for data sources

The large vertical velocity and apparent lack of pre-Holocene recharge at network 47 in the California Coastal Basin (Fig. 9) implies a relatively dynamic flow system that may be related to artificial recharge facilities and extensive pumping for public supply in that area (Hamlin et al. 2002). Data from networks 27, 28, and 49 in the Rio Grande aquifer system are consistent with recharge occurring on the margins of the basin such as mountain-front recharge (networks 27 and 28), and then moving out into the basin where there is no further recharge in areas away from surface-water sources (network 49; Plummer et al. 2004). This process, illustrated schematically in Fig. 10a, results in progressively older water at the water table with distance from recharge areas (compare networks 27 and 49 in Fig. 9). Under these conditions isochrons may approach vertical in the absence of geologic confinement. Such pseudo-confinement also could account for the extrapolated groundwater age at the water table of about 4,000 years at network 46 in the western Mojave Desert, California (Fig. 9), because it too was distal to recharge areas at the basin margin and intermittent stream channels (Izbicki and Michel 2004).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig10_HTML.gif
Fig. 10

Schematic representations of groundwater movement in the a Rio Grande aquifer system, New Mexico (adapted from Plummer et al. 2004 and Anning et al. 2009) and b Central Valley aquifer system, California (adapted from Faunt 2009). Arrows indicate directions of groundwater movement. In a, dashed arrows indicate flow coming out of the page (going to the south). vv vertical groundwater velocity; res. time groundwater residence time

The vertical distribution of pre-Holocene recharge at network 48 in the Central Valley aquifer system, California, appears to have been influenced by substantial downward movement of young (0–60 years) recharge on the basis of the 3H and 14C data (Fig. 9). Peaks in the 3H and 14C values essentially overlapped at depths of 12–14 m below the water table and could be related to peak bomb-derived atmospheric 3H and 14C concentrations of the early 1960s. Almost all the water sampled below the base of those peaks appeared to be pre-Holocene in age (Fig. 9). These data are consistent with widespread downward movement of young groundwater in the past ∼60 years in this semi-arid region, displacing considerably older recharge below it (Jurgens et al. 2008). The most likely source of this recent recharge in the Central Valley was irrigation water. The shift from natural recharge along the valley margins to aerially diffuse recharge across the valley changed the aquifer from a horizontal-flow dominated system toward a more vertical-flow dominated system (Burow et al. 2008; Faunt 2009). The effect of this change in recharge patterns on apparent vertical groundwater velocities was evident when CFC/SF6 derived velocities from shallow agricultural networks 2 and 3 (1–2 m/yr) were compared to the 14C derived velocity from network 48 (0.005 m/yr), which may be more representative of long-term natural recharge (Fig. 10b). Apparent velocities increased by more than two orders of magnitude from pre-development to post-development times. A similar large increase in vertical velocity with the advent of irrigation was simulated in a regional groundwater flow model of the Central Valley aquifer (Williamson et al. 1989). Assuming linear age gradients, the positions of the 3H and 14C peaks in this flow system (Fig. 9) would correspond to somewhat smaller vertical velocities for young recharge of around 0.30.4 m/yr, respectively, given sample collection dates between 2003 and 2006. For a mean porosity of 30%, vertical velocities from the various tracers would equate to apparent recharge rates of about 2 mm/yr during pre-development times and 90–600 mm/yr after the onset of irrigation.

The data in Fig. 9 imply a rather sharp transition from overlying modern to underlying pre-Holocene recharge as a result of the change in recharge patterns in the Central Valley aquifer. At the local scale, however, the transition zone is likely to be less well defined because of mixing due to pumping long-screen irrigation and public-supply wells. Mixing of this type was apparent in 14C age profiles from selected individual well nests in the High Plains aquifer (Fig. 11), another aquifer in which the change in recharge patterns with the onset of irrigation disrupted the natural vertical age gradients and shifted the aquifer toward a more vertical-flow dominated system (McMahon et al. 2007). In areas with few irrigation and public-supply wells, groundwater ages increased approximately logarithmically with depth, as expected for distributed recharge (Fig. 11a). Age gradients flattened considerably in areas with many irrigation and public supply wells (Fig. 11b), presumably because of the combination of increased recharge from irrigation-return flows and mixing by the pumping wells. This mixing of young and old water in the depth interval of production wells has important water-quality implications for those wells (Burow et al. 2007; McMahon et al. 2008a).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig11_HTML.gif
Fig. 11

Groundwater age in relation to well-screen depth below the water table in areas of the High Plains aquifer with relatively a small and b large numbers of pumping wells for irrigation and pubic supply. In a, network numbers are map reference numbers (Fig. 1). Data in b from McMahon et al. (2007). The error bars represent 1 standard deviation in the age estimate

Although pre-Holocene recharge in unconfined aquifers is largely restricted to the western United States, pre-Holocene recharge is commonly found in confined aquifers throughout the country (Aeschbach-Hertig et al. 2002; Clark et al. 1998; Douglas et al. 2007; Kennedy and Genereux 2007; Klump et al. 2008; Lopes and Hoffmann 1997; Ma et al. 2004; Pearson and White 1967; Phillips et al. 1989; Plummer et al. 1990; Plummer 1993; Plummer and Sprinkle 2001; Stotler et al. 2010). Examination of the confined aquifer systems is beyond the scope of this study, but a compilation of data describing age (14C, 4He, or other tracers) distributions in these aquifer systems is needed.

Tritium timescale for recharge in unconfined aquifers

A total of 1,873 3H analyses were used to examine the occurrence and distribution post-1950s recharge in 19 unconfined aquifer systems (Fig. 1 and Table 4). The percentage of water samples with decayed 3H concentrations ≥ 0.5 TU ranged from 11 at network 64 in the Rio Grande aquifer system near Albuquerque, located in a semi-arid setting, to 100 at networks 62 and 67 in the Piedmont and Blue Ridge crystalline-rock aquifer and Valley and Ridge carbonate/sandstone-rock aquifer (Table 4), both of which are located in humid settings. The median percentage was 90. Data in Table 4 were decayed to 1 January 2010 to facilitate comparison between samples collected on different dates, and those with decayed 3H concentrations ≥ 0.5 TU were assumed to contain a component of post-1950s recharge. The apparent widespread occurrence of post-1950s recharge in networks 62 and 67 could indicate relatively rapid movement of groundwater through karst features or fractures in those rocks (Cook et al. 1996; Plummer et al. 2001). Other networks with small percentages (< 50%) of post-1950s recharge included networks 57 and 58 in the High Plains, which were located in relatively dry climates. Large percentages (≥ 95%) of post-1950s recharge were present in networks 55 and 56 (eastern and central United States glacial deposits), 61 (Ozark Plateau carbonates in the central United State), and 66 (Surficial aquifer system in the southeastern United States). Those networks were located in humid climates.
Table 4

Percentage of groundwater samples with tritium concentrations ≥ 0.5 tritium units, for selected unconfined aquifer systems. Tritium concentrations were decayed to 1 January 2010 to normalize samples collected on different dates. Only aquifers with at least 30 samples are listed, with the exception of the Rio Grande aquifer system. Map reference numbers refer to Fig. 1 and Figure ESM1. See Table ESM5 for tritium data and supporting information

Map reference number

Principal aquifer system

Percentage of samples with tritium concentrations ≥ 0.5 TU (sample count)

Data source

-

Alluviuma

95 (150)

USGS unpublished data

50

Basin and Range basin fill

82 (105)

USGS unpublished data; Thiros 2003

51

California Coastal Basins

76 (80)

USGS unpublished data

52

Cambrian-Ordovician

77 (30)

USGS unpublished data; Kolpin et al. 1993

53

Central Valley

90 (170)

USGS unpublished data; Burow et al. 1998; Burow et al. 1999; Jurgens et al. 2008

54

Edwards-Trinity

90 (77)

USGS unpublished data

55

Glacial Deposits (eastern glaciated region)

97 (39)

USGS unpublished data

56

Glacial Deposits (central glaciated region)

95 (482)

USGS unpublished data; Kolpin et al. 1993

57

High Plains (northern)

47 (175)

USGS unpublished data; McMahon et al. 2007; Landon et al. 2008: Stanton and Qi 2007; Stanton and Fahlquist 2006

58

High Plains (central)

32 (101)

USGS unpublished data; McMahon et al. 2004a

59

High Plains (southern)

56 (68)

USGS unpublished data; Fahlquist 2003; McMahon et al. 2004b; Stanton and Fahlquist 2006

60

Northern Atlantic Coastal Plain

86 (72)

USGS unpublished data

61

Ozark Plateau

96 (45)

USGS unpublished data; Kolpin et al. 1993

62

Piedmont and Blue Ridge crystalline rocks

100 (35)

USGS unpublished data

63

Puget-Willamette Lowland

79 (72)

USGS unpublished data; Hinkle 1997

64

Rio Grande (eastern mountain front)

11 (19)

Plummer et al. 2004

65

Snake River Plain

93 (45)

USGS unpublished data

66

Surficial

96 (71)

USGS unpublished data; Katz et al. 2007

67

Valley and Ridge

100 (37)

USGS unpublished data

aNot shown in Fig. 1

Tritium profiles in the networks containing the smallest percentages of post-1950s recharge (networks in the Rio Grande and High Plains aquifer systems) generally were characterized by relatively rapid decreases in 3H concentration below the water table (Fig. 12). The data indicate those aquifers generally contained relatively thin veneers of post-1950s recharge overlying thicker zones of older water, some of which was at least 10,000 years old (Fig. 9). Tritium concentrations plotted in Fig. 12 were not decayed to 1 January 2010, but were grouped in 5-year intervals of the date of sample collection to facilitate the comparison of 3H distributions between networks and with time. Tritium profiles at networks 60 and 53 in the Northern Atlantic Coastal Plain and Central Valley aquifer systems, respectively, were the only ones to contain relatively well developed peaks over multiple sampling periods that could be related to peak 3H concentrations in precipitation of the early 1960s (Fig. 12). The zone of post-1950s recharge in sampled parts of those two aquifer systems was thicker than it was in the Rio Grande and High Plains aquifer systems. Assuming linear age gradients because of the lack of data on aquifer thickness for these widely dispersed wells, and mean porosities of 30–40%, the positions of the 3H peaks in the Northern Atlantic Coastal Plain and Central Valley aquifer systems would correspond to recharge rates of around 95–160 mm/yr and 120–380 mm/yr, respectively (Fig. 12). In general, these rates are less than those determined from age data at the nested and water-table well networks in those aquifer systems (Tables 1 and 2). This difference might be anticipated because the 3H data are a composite from wells located in various land-use settings, possibly including natural settings, whereas the nested and water-table well networks targeted specific land uses (for example, agriculture) that generally have higher recharge rates than other settings like rangeland and forest. The lower recharge rates could also be a result of using the linear model, whereas higher recharge rates might be given by the exponential model for the same profiles. In the two well networks containing 100% post 1950s recharge, 3H concentrations remained fairly constant with depth (Fig. 12). The widely dispersed 3H data summarized here only allowed for general comparisons of post-1950s recharge between sampled parts of several unconfined aquifer systems. However, interpretations based on the 3H data provide important constraints on the residence times, sustainability, and contamination potential of various hydrogeologic and climatic settings.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-011-0722-5/MediaObjects/10040_2011_722_Fig12_HTML.gif
Fig. 12

Vertical distribution of 3H in selected unconfined aquifer systems, grouped in 5-year intervals. Tritium data in these plots were not decayed to 1 January 2010. The solid lines represent LOWESS smooths of the 3H concentrations. The shaded areas show tritium concentrations less than 0.5 TU. Numbers next to aquifer names are map reference numbers (Fig. 1 and Figure ESM1). The data sets are in Table ESM5. See Table 4 for data sources

Implications

Spatially continuous recharge maps such as those derived from geographical information system (GIS)-based water-balance models, are gaining in accuracy and detail. Concurrently, environmental-tracer data are revealing increasingly complex spatial and temporal variations in recharge processes and rates. These approaches have different advantages and limitations, depending in part on the scale of interest and the ultimate objective. Ideally, these approaches would be applied iteratively or tracer data would be integrated directly into model parameter estimation. In addition, distributed environmental-tracer data provide essential (unique) information about recharge rates, groundwater velocities, and flow patterns in various situations, for example: (1) where recharge is affected locally by artificial-recharge basins, irrigation return flow, pumping, topographic-recharge focusing, or hydraulic-conductivity contrasts, physically based or GIS-based models and regressions may not provide accurate predictions; (2) where recharge is localized (e.g., in playas, ephemeral streams, fractures), unsaturated-zone water balance and solute-transport profiles may fail to detect regional-scale aquifer recharge; (3) where changes in climate or land use have caused temporal changes in recharge (e.g., glacial-interglacial changes at 14C time scales; agricultural or urban effects at modern anthropogenic-tracer time scales), those effects on recharge may not be known from first principles; (4) where groundwater-residence time distributions are important (e.g., for predicting responses to contaminant inputs), accurate characterization of watershed water balances and fluxes may fail to reflect groundwater volumes and age distributions (e.g., the “age” of the old component in stream discharge hydrographs, or the extent of vertical groundwater mixing in pumped aquifers). Results from this compilation demonstrate that groundwater age-based recharge estimates can provide useful insights into spatial and temporal variability in recharge at a national scale and factors controlling that variability. Local age-based recharge estimates provide important empirical data and process information needed for testing and improving more spatially complete model-based methods.

Conclusions

This paper represents an initial attempt to synthesize existing datasets of groundwater age into a consistent analysis of the rates and timescales of recharge in aquifer systems of the United States. Recharge rates were based on age distributions in aquifers determined from 565 CFC, SF6, 3H/3He, and 14C measurements in 29 nested and 16 water-table well networks. Timescales of recharge were defined by 102 14C measurements in 8 well networks and 1,873 distributed 3H measurements in 19 networks. The reliance on published “apparent” tracer-based ages could be subject to various sources of error and bias. Nevertheless, the composite dataset generally conforms to patterns that are either predicted by statistical models or rationalized by local phenomena, providing support for the use of empirical environmental-tracer techniques in recharge studies.

Recharge rates ranged from < 10 to 1,200 mm/yr in selected unconfined aquifers in the United States on the basis of measured vertical groundwater-age distributions. Mean groundwater residences times calculated using these recharge rates and data on aquifer thickness ranged from 6 to 19,000 years. Recharge represented 22% (median) of mean annual precipitation at non-irrigated sites, whereas it represented 40% of precipitation plus applied irrigation water at irrigated sites. Infiltration of excess irrigation water increases the rate of entry of agricultural contaminants into aquifers and reduces the mean age of groundwater, potentially decreasing effects of natural remediation. Relations between groundwater recharge and various environmental factors indicated climate, geology, land use, soil type, and topography were important influences on recharge. On a regional basis, recharge exhibited a significant inverse correlation with mean annual air temperature and a significant positive correlation with mean annual precipitation. Empirical characterizations of climate effects on recharge can provide guidance for understanding and modeling how recharge might change in response to future climate scenarios.

Comparison of groundwater age-based recharge to recharge from stream base-flow separation methods showed similar overall patterns but substantial local differences, for example in areas of irrigated agriculture where the age-based estimates were larger than base-flow estimates. Recharge based on groundwater age was greater than recharge from base-flow separation in most non-irrigated settings as well, consistent with underestimation of total recharge on the part of base-flow separation methods. One explanation of these data is underestimation of groundwater discharge by some base-flow separation methods that do not account for increasing groundwater discharge in high-flow conditions. Another possible explanation is that recharge estimated from shallow groundwater-age profiles includes recharge to both shallow (local) and deep (regional) flow systems, whereas recharge estimated from stream base flow only captures the component of recharge partitioned to the shallow (local) flow systems.

Radiocarbon-based groundwater velocities and estimates of the depth to the first occurrence of pre-Holocene recharge provided useful perspective on long-term timescales for recharge in the selected unconfined aquifers. Estimates of depth to the first occurrence of pre-Holocene recharge provided qualitative measures of the distribution of old water in the aquifers that presumably is not renewable on a meaningful timescale for water-resource management in the absence of actions like artificial recharge. Widely dispersed 3H data summarized here only allowed for general comparisons of post-1950s recharge between sampled parts of several unconfined aquifer systems. However, interpretations based on the 3H data such as more post-1950s recharge in (1) karst and fractured-rock aquifers than in clastic-rock aquifers and (2) aquifers in humid climates than aquifers in semi-arid climates, provided important constraints on the residence times, sustainability, and contamination potential of various hydrogeologic and climatic settings. Fundamental shifts in long-term recharge patterns and flow directions were revealed through combined use of CFC, SF6, 3H, and 14C data in parts of the Central Valley and High Plains aquifers where irrigated agriculture caused shifts in predominant flow directions from horizontal to more vertical.

Acknowledgements

We thank D. K. Solomon and L. I. Wassenaar for sharing data and ancillary information with us. We would also like to thank the many US Geological Survey colleagues who shared data and local expertise on aquifer flow systems with us. K. Belitz, J. Denver, F. Leaney, T. Tokunaga, and two anonymous reviewers provided helpful comments on earlier versions of the manuscript. This work was funded by the following programs of the US Geological Survey: Groundwater Resources Program, National Water-Quality Assessment Program, and National Research Program.

Supplementary material

10040_2011_722_MOESM1_ESM.pdf (3 mb)
ESM(PDF 3099 kb)

Copyright information

© Springer-Verlag (outside the USA) 2011