Global Biogeochemical Cycles

Relationships between Δ14C and the molecular quality of dissolved organic carbon in rivers draining to the coast from the conterminous United States


Corresponding author: D. Butman, Yale School of Forestry and Environmental Studies, New Haven, CT 06511, USA. (


[1] Dissolved organic carbon (DOC) in natural waters possesses chemical and molecular qualities indicative of its source and age. The apportionment of DOC by age into millennial and decadal pools is necessary to understand the temporal connection between terrestrial and aquatic ecosystems in the global carbon cycle. We measured Δ14C-DOC and chemical composition indices (specific ultraviolet absorbance (SUVA254), fluorescence index (FI), hydrophobic organic acid fraction (HPOA) content) for 15 large river basins in the conterminous United States. Across all rivers the average proportion of HPOA in DOC correlated strongly with SUVA254 (r2 = 0.93 p < 0.001). Individual Δ14C-DOC ranged from a low of −92.9‰ (726 y.b.p.) in the Colorado River to 73.4‰ (>Modern) in the Altamaha River for the year 2009. When adjusted by total discharge, these U.S. Rivers export modern carbon at between 34 and 46‰, a signal dominated by the Mississippi River. The variation in Δ14C correlates to indices of the aromaticity of the DOC measured by the SUVA254 (r2 = 0.87, p < 0.001), and FI (r2 = 0.6; p < 0.001) as well as differences in annual river discharge (r2 = 0.46, p < 0.006). SUVA254 was further correlated to broad scale vegetation phenology estimated from the Enhanced Vegetation Index derived from the NASA Moderate Resolution Imaging Spectrometer (MODIS). We show that basins with high discharge, high proportions of vegetation cover, and low human population densities export DOC enriched in aromatic material that corresponds to recently fixed atmospheric CO2. Conversely old DOC is exported from low discharge watersheds draining arid regions, and watersheds more strongly impacted by humans. The potential influence from fossil carbon from human inputs to aquatic systems may be important and requires more research.

1. Introduction

[2] Rivers and streams connect the terrestrial carbon cycle to estuarine and ocean carbon cycles. In doing so, in addition to passive transport, they also provide conditions conducive to both chemical and biophysical transformations of organic matter. Humans have changed the carbon chemistry within rivers by large scale land cover conversions [Ciais et al., 2006; Raymond et al., 2004], impoundments and hydrologic diversions [Aufdenkampe et al., 2011; Downing et al., 2008; Syvitski et al., 2005] as well as direct inputs of organic pollutants and wastewater [Zeng and Masiello, 2010; Griffith et al., 2009; Harrison et al., 2005; Hopkinson and Vallino, 1995; Spiker and Rubin, 1975].

[3] Although a substantial fraction of terrestrial organic matter transported to rivers and streams remains confined by impoundments or is remineralized and respired along the river reach [Aufdenkampe et al., 2011], the total flux of dissolved organic carbon (DOC) from rivers to the coast is ∼0.2 Pg C yr-1 [Aufdenkampe et al., 2011; Hope et al., 1994; Ludwig et al., 1996; Mayorga et al., 2005; Meybeck, 1982]. This contribution of organic carbon is more than needed to balance the turnover and radiocarbon age of DOC in the world oceans [Opsahl and Benner, 1997]. Many studies have tried to evaluate the age of riverine and stream DOC. A majority of these studies demonstrate that DOC is dominated by a modern, post-bomb carbon pool [Benner et al., 2004; Evans et al., 2007; Hedges et al., 1986; Hélie and Hillaire-Marcel, 2006; Neff et al., 2006; Raymond et al., 2007; Spiker and Rubin, 1975; Striegl et al., 2007] suggesting that the majority of DOC in rivers is of recent origin.

[4] Millennial aged carbon in riverine DOC was reported in a study of North American rivers feeding the North Western Atlantic Ocean [Raymond and Bauer, 2001b]. In this work it was suggested that bacterial utilization of DOC preferentially removes modern material reducing the Δ14C signature of the remaining carbon. Aged organic carbon has recently been determined in a number of river systems [Blair et al., 2003; Dickens et al., 2004; Goñi et al., 1997; Hood et al., 2009; Masiello and Druffel, 2001; Raymond and Bauer, 2001b; Sickman et al., 2010; Striegl et al., 2007]. Thus an older more recalcitrant pool may be present in some rivers and hidden due to the heterogenic nature of DOC that is often dominated by a young 14C-bomb enriched pool. A small number of studies attempt to estimate the contribution of both modern and aged organic carbon to the oceans across the water year [Neff et al., 2006; Raymond et al., 2007; Sickman et al., 2010]. Research demonstrates that basins can export older DOC under both base flow and high flow scenarios [Masiello and Druffel, 2001; Raymond and Bauer, 2001b; Raymond et al., 2004; Sickman et al., 2010]. Few studies provide additional information on the nature or chemistry of the DOC in these rivers. The composition of DOC together with its radiocarbon age may help determine the apportionment between fast (annual to decadal turnover) and slow (>decadal turnover) carbon pools.

[5] Δ14C measurements of natural organic matter in rivers to determine its age must be used with caution. Studies have found that petroleum-based radiocarbon dead (devoid of14C due to decay) DOC can be introduced via rainwater [Avery et al., 2006; Raymond, 2005; Hood et al., 2009; Stubbins et al., 2012] as well as from wastewater treatment plants [Ahad et al., 2006; Griffith et al., 2009]. Although old, a significant proportion of this anthropogenic petroleum fraction can potentially biodegrade [Petrovic and Barceló, 2004; Hood et al., 2009]. Thus, it is possible that the export of modern DOC is modulated by both the bacterial processing of DOC in rivers and estuaries, and by the introduction of petroleum-based DOC that might leave an aged DOC pool for oceanic export. The export of aged organic carbon, therefore, may be less a function of old soil organic matter turnover and more related to the anthropogenic influence of wastewater, atmospheric inputs and other anthropogenic activities.

[6] Regardless, broad continental-scale systematic evaluations of the radiocarbon age of DOC in temperate rivers to determine the relative importance of modern versus aged organic carbon have not yet been completed. Studies to date where the Δ14C of DOC have been measured suggest we do not have a complete understanding of the potential sources of aged material moving through our rivers. To address this we provide an analysis of the quality and radiocarbon age of DOC leaving 15 large rivers basins in the U.S.

2. Site Description

[7] Fifteen major U.S. river basins that drain to the coast were sampled across the hydrograph in this study (Figure 1). In total, these basins drain ∼6.4 million km2 of the conterminous U.S., Mexico and Canada and represent ∼79% of the land area of the contiguous 48 states and ∼90% of all the freshwater discharge in 2009. Basin sizes range from 2.9 million km2 for the Mississippi River to 28,183 km2 for the Connecticut River. Average annual precipitation is lowest in the Colorado River basin at 328 mm yr−1 and highest in the Mobile River basin at 1437 mm yr−1. Water management regimes and overall hydrology differ greatly among these basins, which are reflected in the internal variation in discharge. Mean annual discharge varied during the 2009 sampling period from 20,190 m3 s−1 for the Mississippi River to 6 m3 s−1 for the Rio Grande River (USGS National Water Information System (NWIS); Additional site information is available in Tables 1a1b.

Figure 1.

Map showing the watersheds and sampling site of North American Rivers reported on in this paper.

Table 1a. NASQAN Coastal Sampling Sites Used for This Analysis With the Hudson and Connecticut Addeda
BasinBasin Area (km2)Location of Sampling StationSample Date RangeDischargeb (m3 s−1)
Latitude (dd)Longitude (dd)StartEndMinMaxMean
  • a

    Sample date range for Δ14C-DOC analysis. Basin areas were delineated using NHDPlus and ArcHydro and are reported in square kilometers. Sampling locations are those provided by the USGS National Water Information System ( Discharge measurements from the 2009 sampling period have been converted from cubic feet per second derived from the NWIS database.

  • b

    Discharge measurements have been compiled for the sampling period only.

Lower Atchafalaya252,41729.702708−91.20204822-Jan-0907-Oct-092,3088,4674,828
Rio Grande564,11725.898969−97.49748425-Feb-0931-Aug-092106
San Joaquin72,09337.676041−121.26632907-Jan-0927-Oct-09177839
St. Lawrence775,67345.006160−74.79491021-Jan-0921-Oct-096,2308,2127,622
Table 1b. Physical and Climatic Attributes for the 15 Large River Basinsa
River BasinAvg Precipb (mm yr−1)E.V.I.cL.S.T.d (deg C)% Land Covere
  • a

    Average precipitation does not cover the sampling dates since a seamless high resolution precipitation data set that covers watershed basins that are beyond the conterminous U.S. does not exist yet.

  • b

    Average precipitation is from a global model data set averaged through 2005 [Hijmans et al., 2005].

  • c

    Enhanced Vegetation Index (E.V.I.) derived from monthly MODIS EVI at 0.05 deg resolution is averaged for the sampling period.

  • d

    Land Surface Temperature (L.S.T) is derived from the MODIS LST product at 0.05 deg resolution and averaged for the sampling period. Both E.V.I. and L.S.T. cover the dates of the sampling period.

  • e

    Land Cover data is derived from the North American Land Change Monitoring System at 250-m resolution [Homer et al., 2004].

Lower Atchafalaya10360.050.540.3116.33323.82831.
Rio Grande3810.130.250.1710.744.529.
San Joaquin4630.190.350.257.239.924.515.830.
St. Lawrence8490.160.310.22−11.823.18.937.9183434.5

2.1. Sample Collection

[8] Freshwater samples were collected through the United States Geological Survey National Stream Quality Accounting Network (NASQAN) at 13 coastal locations, along with the Hudson and Connecticut rivers (Figure 1). At each location, 4–10 samples were collected during January–December 2009, and discharge measurements were made at the time of sample collection (Table 1a). For the NASQAN effort, depth-integrated samples were collected at multiple locations across the main channel and subsequently churned to create a mixed sample [U.S. Geological Survey, 1997–1999, chapters A1–A9] (updates and revisions are ongoing and can be viewed at Unfiltered water was shipped on ice to the USGS in Boulder CO within 24 h of collection and immediately filtered using pre-combusted Whatman GF/F filters at 0.7-μm nominal pore size. Subsamples for DOC concentration and composition were stored in precombusted amber glass bottles fitted with Teflon lined caps and refrigerated until analysis. Subsamples for radiocarbon analyses were placed in acid washed 125 ml polycarbonate bottles, acidified to pH 2.5 using high purity 60% H3PO4, immediately frozen, and shipped overnight to Yale University. Grab samples from the Connecticut and Hudson rivers were collected at the surface only at an approximate depth of 0.5 m. For the Connecticut River, all samples were collected from the main channel near Deep River, CT except those collected in winter when the presence of river ice necessitated collecting from a dock extending offshore by 10 m. Hudson River samples were collected from the town landing at Newburgh, NY. Both the Connecticut River and Hudson River samples were collected unfiltered into pre-combusted amber glass bottles, placed on ice, and shipped overnight to Boulder CO. All further handling of samples from these two rivers is similar to that for the NASQAN samples.

2.2. Concentration and Optical Characterization of Organic Matter

[9] DOC concentration was measured utilizing the platinum catalyzed persulfate wet oxidation method on an O.I. Analytical Model 700 TOC Analyzer™ using established methods [Aiken et al., 1992]. Reported values are the averages of duplicate analyses. Standard deviation for the DOC measurement was determined to be ±0.2 mg carbon L−1. Fluorescence and absorbance measurements were made within 48 h of sample filtration on a JY-Horiba Fluoromax-3 fluorometer and an Agilent Model 8453 UV-Visible spectrophotometer, respectively. Fluorescence measurements were obtained and corrected according toMurphy et al. [2010], and the fluorescence index (FI) calculated as the ratio of the emission at 470 nm to that measured at 520 nm when excited at 370 nm [McKnight et al., 2001]. Decadal absorption coefficients were measured at 1 nm wavelength intervals with a path length of 1 cm. Specific ultraviolet absorbance at 254 nm (SUVA254) was calculated according to Weishaar et al. [2003], and reported in units of [L/(mg carbon * m)]. Absorbance at 254 nm was chosen because of its common association with aromatic moieties in a sample [Weishaar et al., 2003].

2.3. Organic Acid Fractionation

[10] Organic acid fractions were chromatographically separated using columns packed with Amberlite™ XAD-8 and XAD-4 resin in a modified version of the method described byAiken et al. [1992]. The method separates DOM in a water sample into 5 operationally defined fractions: hydrophobic organic acids (HPOA), hydrophobic organic neutrals (HPON, calculated by difference), low molecular weight hydrophilic acids (HPI), transphilic organic acids (TPIA), and transphilic neutrals (TPIN, not reported). The distribution of DOM in these fractions can provide information about carbon quality for bacterial mineralization, photodegradation, and reactivity. In brief, 1 L of acidified filtered water was passed through a 20-mL glass column packed with XAD-8 resin at a flow rate of 4 mL/minute. The XAD-8 effluent was then passed through a 20-mL glass column packed with XAD-4 resin at the same flow rate. The effluent from the XAD-4 column was collected as HPI. Both columns were back-eluted with 0.1 N NaOH at a flow rate of 2 mL/minute. The eluates, HPOA from the XAD-8 and TPIA from the XAD-4, were collected and acidified for analyses. Each fraction was weighed to determine sample volume, and analyzed for both UV absorbance and DOC concentration. HPON was then calculated by the mass balance difference between the XAD-8 effluent and the HPOA fraction. Data are reported as averaged percentages from duplicate fractionations, based on mass of organic matter in each fraction with a standard deviation of ±2 percent.

2.4. Isotopic Analysis

[11] Samples were prepared for isotopic analysis using established methods [Druffel and Williams, 1992; Raymond and Bauer, 2001a]. Frozen samples were allowed to thaw in a refrigerator at 4°C for 12 h. For each sample, 120 mL were placed into a 160-mL quartz reaction vessel connected to a vacuum extraction line. The previously acidified samples were sparged with ultra-high purity N2 for 10 min to remove inorganic CO2. Ultra-high purity O2 was then bubbled through the samples for 5 min to add additional O2 as an oxidant. The sample was irradiated (5 h) using a high energy UV light to convert DOC to CO2 [Bauer et al., 1992; Eadie et al., 1978; Williams and Gordon, 1970], which was then trapped and cryogenically purified within the vacuum system using liquid nitrogen. The purified CO2was sealed in a 6-mm combusted Pyrex© tube and sent for accelerator mass spectrometry (AMS).

[12] All isotopic measurements were conducted at the National Ocean Sciences Accelerator Mass Spectrometry (NOSAMS) facility at the Woods Hole Oceanographic Institution (WHOI). For Δ14C analysis for DOC, the CO2 gas samples were converted to graphite targets by reducing CO2 with an iron catalyst under 1 atm H2 at 550°C. The graphite targets were analyzed for Δ14C and corrected for potential fractionation using δ13C results, following the conventions of Stuiver and Polach [1977]; δ13C values were obtained with an Optima mass spectrometer at the NOSAMS facility.

3. Results

3.1. Across Basin Trends

[13] Average DOC concentrations ranged from 10.0 ± 2.6 mg C L−1 in the Altamaha River to 2.0 ± 0.2 mg C L−1 in the Columbia River (Table 2) in 2009. The average DOC concentration for all 15 rivers presented here was 4.2 ± 2.0 mg C L−1 (Table 2). Average Δ14C values ranged from a low of −92.9‰ in the Colorado River to 73.4‰ in the Altamaha. These Δ14C values correspond to DOC calendar ages of 726 ± 30 y.b.p. to modern (Figure 2). Modern carbon is any organic carbon that has been taken up from the atmosphere since 1950 when bomb testing was prevalent and has a positive Δ14C value. Flow weighted values, based on all samples collected during 2009, produce a range from −120.8‰ in the Sacramento to 70.0‰ in both the Altamaha and Mississippi rivers corresponding to 970 Yrs BP to modern ages (Table 2). The flow weighted average Δ14C-DOC exported from the conterminous U.S. is 46‰ driven primarily by the modern carbon signal from the Mississippi River. The Colorado, Potomac, Rio Grande, Sacramento and Susquehanna rivers showed aged DOC across the annual hydrograph while all other rivers exported modern carbon (Table 2). The flow weighted average Δ14C-DOC for systems that exported aged DOM suggest a further depletion of radiocarbon under higher discharge, while those rivers that exported more modern DOM show increases in radiocarbon ratios under higher flow (Table 2).

Table 2. Average Dissolved Organic Carbon Concentration, Specific Ultraviolet Absorbance (SUVA254), Fluorescence Index (FI), δ13C, Δ14C, and Mean Discharge Data During the 2009 Sampling Period for North American Rivers (Table 1a) With Standard Deviation in Parenthesesa
RiverDOC (mg L−1)SUVA254 (L (mg C * m)–1)FIHPOA (%)δ13C-DOC‰Δ14C-DOC‰Calendar Ages (y.b.p)Flow Weighted Δ14C‰ (y.b.p)Discharge (m3 s−1)
  • a

    SUVA254 units are [L (mg C * m)−1] and all isotopic data are in ‰. Calendar ages are calculated as [age = 8033*ln(Fm)] where Fm is the fraction modern [Stuiver and Polach, 1977]. Calendar ages are presented as >Mod when the fraction modern exceeded 1 and represent fixed bomb carbon since 1950.

Altamaha10.0 (2.6)4.5 (0.33)1.32 (0.02)56 (3)−27.79 (1.09)73.46 (12.57)>Mod70.0 (>Mod)680 (905)
Atchafalaya5.9 (0.58)3.5 (0.34)1.41 (0.01)49 (3)−26.82 (0.72)51 (23.11)>Mod59.7 (>Mod)9226 (3359)
Colorado2.9(0.10)1.6 (0.10)1.49 (0.05)38 (2)−25.7 (1.09)−92.94 (20.02)726 (540–1020)−95.5 (748)65 (17)
Columbia2.0 (0.20)2.7 (0.52)1.39 (0.02)43 (3)−25.21 (2.26)18.01 (16.61)>Mod (10−>Mod)26.6 (>Mod)6250 (2496)
Connecticut3.4 (0.59)3.1 (0.13)1.41 (0.02)45 (3)−27.15 (0.46)22.14 (22.06)>Mod46.7 (>Mod)438 (278)
Hudson3.7 (0.37)3.1 (0.17)1.43 (0.04)48 (1)−27.33 (0.66)30.05 (27.01)>Mod51.2 (>Mod)457 (133)
Lower Atchafalaya5.6(0.54)3.3 (0.84)1.42 (0.01)48 (3)−26.41 (1.86)39.25 (29.36)>Mod (115−>Mod)52.6 (>Mod)4828 (2072)
Mississippi4.4(0.39)3.1 (0.28)1.44 (0.01)44 (2)−25.82 (1.01)45.46 (33.83)>Mod70.0 (>Mod)20190 (6771)
Mobile5.9(1.2)3.3 (0.84)1.45 (0.06)50 (2)−20.19 (13.13)23.53 (27.88)>Mod (155−>Mod)29.8 (>Mod)1574 (1249)
Potomac3.3(0.25)2.3 (0.38)1.49 (0.02)39 (5)−26.55 (0.39)−40.6 (107.19)>Mod (2470−>Mod)−111.5 (892)231 (172)
Rio Grande4.8(0.37)1.9 (0.11)1.53 (0.01)34 (3)−27.24 (1.97)−62.53 (46.85)470 (170–1260)−108.2 (862)6 (3)
Sacramento2.6(1.12)2.5 (0.58)1.42 (0.03)41 (6)−21.57 (18.33)−34.7 (82.41)222(1940−>Mod)−120.8 (976)451 (311)
San Joaquin3.8(1.50)2.4 (0.24)1.47 (0.03)40 (2)−25.16 (2.6)3.91 (26.03)>Mod (275−>Mod)22.2 (>Mod)39 (20)
St. Lawrence2.7(0.15)1.3 (0.21)1.41 (0.02)29 (3)−25.38 (1.93)21.75 (19.09)>Mod (60−>Mod)26.1 (>Mod)7622 (767)
Susquehanna2.7(0.14)2.4 (0.24)1.47 (0.03)41 (4)−25.45 (1.63)−22.64 (38.39)126 (545−>Mod)−38.6 (258)1228 (559)
Figure 2.

Box plot of Δ14C-DOC across 15 large rivers in the U.S. Samples are taken from 2009 across the annual hydrograph. In all cases, peak flow is included. Numbers in parentheses represent the number of Δ14C-DOC measurements made. Dark lines are the median values and the red dashed lines are the means. Error bars represent the 5th and 95th percentiles with outliers represented by the black points.

[14] Annual average SUVA254values ranged from 4.5 ± 0.33 L mg-C−1 m−1in the Altamaha River down to 1.3 ± 0.21 L mg-C−1 m−1in the St. Lawrence River with the average across all systems being 2.7 ± 0.80 L mg-C−1 m−1. Annual average Δ14C signatures showed a statistically significant linear increase across a range of SUVA254values from 1 to 5 L mg-C−1 m−1 (r2 = 0.6, p < 0.001 Figure 3). The St. Lawrence River appears as an outlier to the linear trend with very low SUVA254 values and an average modern carbon isotopic signature of 21.8 ± 19.09 ‰. The FI ranged from a high of 1.53 ± 0.01 in the Rio Grande to 1.32 ± 0.02 in the Altamaha River. There was a strong negative relationship between Δ14C-DOC and FI (r2 = 0.6, p < 0.001, Figure 3).

Figure 3.

(a) Average specific ultraviolet absorbance (SUVA254) and (b) fluorescence index (FI) data versus Δ14C-DOC across 15 large rivers in the conterminous U.S. Error bars represent 1 standard deviation in the estimate. Removal of the St. Lawrence improves the relationship significantly to r2 of 0.87 with SUVA254.

[15] The percentage of DOC as the hydrophobic organic acid fraction (HPOA), which is comprised of aquatic humic substances, ranged from 56 ± 3% in the Altamaha to a low of 34 ± 3% in the Rio Grande (Table 2). There is a strong relationship between SUVA254 and the percentage of HPOA in DOM across the basins (r2 of 0.93) (Figure 4). This result illustrates the applicability of a simple measure like SUVA254 toward understanding the compositional structure of DOM in freshwaters.

Figure 4.

The relationship between specific ultraviolet absorbance (SUVA254) and the amount of hydrophobic organic acids as a percentage of whole water dissolved organic carbon concentration. Error bars represent 1 standard deviation in the estimate.

[16] River discharge is highly variable across basins with the Mississippi River having an average of 20,190 ± 6771 m3 s−1 down to 6 ± 3 m3 s−1 for the Rio Grande for 2009. Differences in hydrology correlated with the Δ14C-DOC. The basins with the highest water yield export modern carbon while systems in arid environments (Rio Grande, Colorado), and those potentially impacted by urban systems (Sacramento, Susquehanna and Potomac), showed aged DOC (Figures 2 and 4). Figure 5 shows the log transformed average discharge for each basin along with the average annual precipitation. There is a significant relationship between the Δ14C signature in a river basin and large-scale hydrologic processes such as precipitation and river discharge (r2 = 0.46, p < 0.006) (Figure 5) where wetter basins have younger DOC.

Figure 5.

Relationship between the annual average Δ14C signal to (a) discharge and (b) precipitation across 15 large rivers in North America.

3.2. Within Basin Trends

[17] All measured constituents and their variability are presented in Table 2. In general there were mixed trends within river basins. Δ14C showed positive correlations with DOC concentration in the Altamaha River (r = 0.79), Atchafalaya River (r = 0.78), Connecticut River (r = 0.84) and Lower Atchafalaya River (r = 0.82) while the St. Lawrence River showed a significant negative correlation (r = −0.79) (Table 4). The across system trends between Δ14C and SUVA254 (Figure 6) were not always evident within basins. Δ14C and SUVA254 showed strong positive relationships in the Columbia, Mobile, Potomac, and Susquehanna rivers (Tables 3 and 4). In rivers where aged carbon dominated the Δ14C signature, such as the Rio Grande and the Colorado (Tables 3 and 4), negative relationships were shown between SUVA254 and radiocarbon isotopic ratios. Those river basins that had a large range in SUVA254, across the water year, such as the Mobile River, had the strongest correlations between SUVA254 and Δ14C.

Figure 6.

Specific ultraviolet absorbance (SUVA254), Δ14C-DOC, and discharge for 15 large river basins in the U.S. for 2009. Black lines and points represent SUVA254, red lines and points represent Δ14C-DOC.

Table 3. Linear Relationship Between Specific Ultraviolet Absorbance (SUVA254nm) and Δ14C -DOC
RiverSlopeInterceptr2# of Sample Daysp
Lower Atchafalaya53.5−1380.390.12
Rio Grande−243.9400.90.3360.2
San Joaquin41.3−95.60.1570.4
St. Lawrence40−31.60.1860.4

[18] The within basin correlation between FI, percent HPOA (HPOA%), discharge and Δ14C-DOC showed significant variation (Table 4). FI showed a negative relationship (p < 0.05) with Δ14C in the Altamaha (r = −0.86), Atchafalaya (r = −0.80), Connecticut (r = −0.93), and the Susquehanna (r = −0.86). Other basins exhibited no significant relationships between FI and Δ14C. As HPOA% increased so did the Δ14C signature (p < 0.05) in the Altamaha (r = 0.80), the Potomac (r = 0.78) the Susquehanna (r = 0.73) and the Hudson River (r = 0.98). In all basins except the Colorado, the relationship between the HPOA% and SUVA254 was positive, however, only the Atchafalaya, Columbia, Connecticut, Potomac, Sacramento and Susquehanna rivers were statistically significant at the p < 0.05 level.

Table 4. Pearson's Correlation Coefficients (r) and a Modified Kendall's Tau [Buonaccorsi et al., 2001] Between Δ14C (‰) and Dissolved Organic Carbon (DOC), Discharge, Specific Ultraviolet Absorbance (SUVA254), Fluorescence Index (FI) and Percent Hydrophobic Acid (HPOA%)a
RiverDOC (mg l−1)Discharge (m3 s−1)SUVA254nmFI% HPOA
  • a

    Bold tau values represent those systems where Δ14C (‰) and a measured constituent move in synchrony at each time step.

Lower Atchafalaya0.820.250.28−0.500.55−0.25−0.540.250.200.00
Rio Grande−0.380.20−0.89−0.60−0.57−0.600.14−0.60−0.33−0.20
San Joaquin0.060.000.761.000.380.67−0.61−1.000.510.33
St. Lawrence−0.79−0.60−0.33−0.200.430.200.11−0.200.661.00

[19] Within basins, discharge was positively correlated to Δ14C- DOC in the San Joaquin (r = 0.76) and negatively correlated in the Rio Grande (r = 0.89) only. In general, within system variation in discharge was not related to the age of the DOC.Figure 6 shows Δ14C, SUVA254, and flow for each of the 15 basins. Across each of these basins the annual flow characteristics are very different. The Colorado and St. Lawrence rivers showed limited variability due to episodic events and very small standard deviations for 2009, while others, like the Sacramento and Altamaha, are dominated by significant seasonal variability. Figure 6 and Table 4 depict how flow, SUVA254, DOC, FI and HPOA % correlate for 2009. The modified Kendal's tau parameter [Buonaccorsi et al., 2001] is a measure of synchrony between two entities where a value of −1 corresponds to a consistent negative correspondence while a value of 1 corresponds to consistent synchrony. Table 4 shows each chemical constituent as it moves with Δ14C-DOC. The degree of synchrony represented by the tau values illustrates the limited correlation between Δ14C-DOC and each measured parameter for DOM and discharge.

4. Discussion

4.1. Age of DOC to U.S. Coasts

[20] In terms of the total contribution to coastal waters, the U.S. exports modern carbon (34–46‰) when adjusted for flow. This number is controlled by the river basins with the largest freshwater flows, the Mississippi and Atchafalaya Rivers. Modern organic carbon signatures have been shown in the Amazon Basin [Hedges et al., 1986; Mayorga et al., 2005] and Arctic river basins [Benner et al., 2004; Neff et al., 2006; Raymond et al., 2007] as well as forested and peat dominated systems [Palmer et al., 2001; Schiff et al., 1997] across the globe.

[21] There are regional differences in the Δ14C-DOC in large river basins. DOC exported to the west coast of the U.S. from 4 major river basins (Columbia, San Joaquin, Sacramento and Colorado) adjusted for flow has an average Δ14C signature of 15.6‰ which corresponds to modern carbon. Although the Colorado River is shown to have consistently low ages, it has a low discharge, which results in a negligible contribution to the flow adjusted average Δ14C. Rivers flowing into the Northern Atlantic Ocean (St. Lawrence, Hudson, Connecticut) are exporting DOC that has a flow adjusted average Δ14C signature of 28.5‰, slightly more modern than rivers in the West. The Gulf of Mexico is dominated by modern DOC from the Mississippi, Atchafalaya and Mobile rivers resulting in a flow adjusted average Δ14C signature of 63.2‰ as is the Altamaha (70‰), a southeastern river draining to the South Atlantic region of the U.S. The remainder of the rivers (Susquehanna and Potomac) discharge to the Atlantic Ocean from the Mid Atlantic region of the eastern U.S., and support an average flow adjusted Δ14C-DOC of −50‰. This export corresponds to a calendar age of ∼350 years before present (1950).

4.2. Within Basin Controls on Δ14C-DOC

[22] Relationships across the seasonal hydrograph between the radiocarbon age of DOC and the molecular characteristics of that organic carbon remain difficult to define. There do not appear to be uniform correlations that hold within each basin (Table 4). The finding that Δ14C of DOC showed strong correlations with the concentration of DOC in the Altamaha, Atchafalaya, and Connecticut river systems suggests that these systems may be dominated by two sources of DOC with distinct 14C-DOC signatures. The Altamaha, Atchafalaya, and Connecticut river basins have a relatively high percentage of wetland and forested ecosystems (Table 1b) which could be the 14C-enriched source. However, the Mobile River also has similar land use characteristics and Δ14C-DOC shows a strong negative relationship with concentration, indicative perhaps of an alternate source derived from either mobilized organic matter from soils or non-terrestrial inputs. Regardless, the dominant source of organic matter is of modern origin.

[23] In contrast to the finding that SUVA254 correlates with Δ14C-DOC across basins, the within basin relationships are much weaker. Only three basins, the Mobile, Potomac and Susquehanna rivers, showed strong relationships between Δ14C-DOC and the aromaticity of that carbon. However, DOC in the Mobile River was consistently modern whereas DOC in the Potomac and Susquehanna rivers was, on average, much older. The discrepancy between modern and old does not exclude a possible uniform explanation. All of these basins show significant proportions of agriculture and forest land cover (Table 1b) which could be significant sources of less degraded organic material. If this were the case, however, we would suspect that increases in discharge would also correlated with increases in modern Δ14C-DOC [Raymond and Saiers, 2010], which does not occur within these basins. In the Colorado and Rio Grande rivers, where there were negative relationships between SUVA254 and Δ14C of DOC, SUVA254 values are very low, and FI is the highest for all of the rivers in this study (Table 2). This finding indicates that throughout the year there is a slight shift in the source of organic carbon to include even older and slightly aromatic material later in the season under lower flow conditions. Inputs from agricultural activities could be mobilizing older soil carbon [Sickman et al., 2010] but we would caution this interpretation as the variation in the measurements presented here for these river basins remains very small, and the SUVA254 and FI values themselves suggest that terrestrial organic carbon may not be the dominant source at any point throughout the year.

[24] The seasonal characteristics of flow differ greatly across these river systems (Figure 6). Only the San Joaquin and Connecticut (positive) and Rio Grande (negative) showed substantial correlations between Δ14C-DOC and discharge. Although the annual discharges and basin areas for the San Joaquin and Rio Grande are nearly an order of magnitude apart, these rivers represent relatively arid systems with low forest cover. We hypothesize that there are two different mechanisms that can attribute a relationship between Δ14C-DOC and discharge unique to each basin. For instance, in the San Joaquin River, modern terrestrial organic carbon is mobilized into the river system as precipitation and flow increase, while in the Rio Grande River, agricultural runoff combined with wastewater and groundwater contributions are being mobilized under higher flow conditions [Benke and Cushing, 2005]. The implication of wastewater contribution to Δ14C-DOC is discussed further below. We suggest that the coupling of discharge with Δ14C-DOC in the Connecticut is a direct result of the strong seasonality of vegetation since the basin is 89% forested (Table 1b) [Raymond and Saiers, 2010].

4.3. Across Basin Controls on Δ14C-DOC

[25] The variation of the 14C-age of DOC across these river systems was not predicted by land cover. Our data suggest that DOC age measured at downstream locations of large river basins is modulated by a number of factors including: (1) sources of DOC (natural versus anthropogenic), (2) the seasonal vegetation cycle (climate) and, (3) water residence time (climate and impoundments). The radiocarbon age of DOC found within these rivers separate along a continuum where one or two of these factor appears most influential. We present a framework to identify the dominant controls on the14C-DOC inFigure 7.

Figure 7.

Preliminary framework for the dominant controls on the Δ14C-DOC in large river basins across the U.S. In summary, the dominant drivers are either climate induced (precipitation influencing groundwater inputs and natural land cover), or human induced (direct inputs or manipulation of the landscape). Each set of effects from either changing climate or humans drives differences in residence time, the processing of DOC along the river continuum, and ultimately the signature of Δ14C-DOC. The rivers investigated in this study are placed along both a residence time and DOC age spectrum that changes from short (young) to long (old) respectively. Both the Rio Grande and Colorado Rivers are strongly influence by both climate and anthropogenic activities.

[26] Aged organic carbon dominates the annual DOC in the Rio Grande, Colorado, Potomac, Sacramento and Susquehanna rivers. These results are similar to those reported previously for the Susquehanna, Hudson, and Delaware rivers [Raymond and Bauer, 2001b; Raymond et al., 2004]. Both the Rio Grande and the Colorado rivers, which consistently showed aged DOC, drain arid regions where groundwater and wastewater can contribute significantly to riverine discharge, water extraction activities can reduce discharge, and impoundments can change the quality of organic carbon. In these dry watershed systems, we hypothesize that three things are occurring. First, in these arid systems with large water diversions and low flow, groundwater contribution and long residence time provides a longer period for radiocarbon dead organic chemicals to build up and influence 14C-ages. That is, direct inputs from anthropogenic activities (agricultural diversions and wastewater) could represent a significant proportion of the total water flow at these arid downstream locations [Swietlik et al., 1995]. Groundwater DOC ages have been reported as low as 16,000 y.b.p [Murphy et al., 1989; Purdy et al., 1992] and could contribute a significant proportion of the total discharge in the Colorado [Warner et al., 1985] and Rio Grande systems. High concentrations of organo-chlorine pesticides have been found in water and sediments in the Colorado River Delta [Garcìa-Hernandez et al., 2001]. These pesticides are synthesized from organic compounds derived from fossil fuel and are therefore radiocarbon ‘dead’ [Reddy et al., 2002]. Second, the low terrestrial productivity of these arid systems may lead to lower contributions of aromatic, modern DOC, which is further compounded by low surface-water runoff. For instance, only 40% of the flow in the Colorado River above Cisco, Utah was accounted for by surface runoff [Michel, 1992]. Third, the potential for in situ production of DOC behind impoundments offers opportunities to incorporate inorganic carbon sources from weathering products that have very old carbon signatures.

[27] In addition, wastewater cannot be ignored as a potential source in all of the rivers presented here, but especially in arid, low flow systems, such as the Colorado and Rio Grande where a small percentage of surface recharge from precipitation contributes to flow. Wastewater plants have been shown to be significant contributors of aged organic carbon in coastal watersheds in Northern England and the Hudson River sampled below Newburg NY [Ahad et al., 2006; Griffith et al., 2009]. Radiocarbon based ages of wastewater DOC for measurements along the east coast ranged from modern-2650 y.b.p. but averaged 1630 y.b.p. [Griffith et al., 2009]. Griffith et al. [2009]suggest that these ages are a function of surfactants and the degradation byproducts of various detergents, cleaners, and disinfectants. Isotopic analysis of dissolved organic carbon shows that wastewater is composed of ∼25% petroleum-based DOC and ∼67% C3 vegetation, and ∼8% C4 vegetation [Griffith et al., 2009]. If riverine DOC is composed of only 1% wastewater derived compounds, the apparent reduction in age is ∼160 y.b.p. The average age of DOC for the Colorado and Rio Grande were 726 and 470 y.b.p., respectively.

[28] The composition of DOC in river systems continues to be used as an indication of the source and processing of carbon [Stubbins et al., 2010; Weishaar et al., 2003] and each basin in this study exhibited unique SUVA254 values that are closely tied to the HPOA% of the DOC (Figure 4). Across basins there is a positive relationship between SUVA254 and the average Δ14C (Figure 3) suggesting that systems with a larger proportion of this aromatic organic material might be more enriched in Δ14C-DOC. In addition, there is a positive relationship between the average annual MODIS Enhanced Vegetation Index (EVI) and the annual SUVA254 values for DOC (r2 = 0.4, p < 0.05). This suggests that large watershed systems that have a higher proportion of the watershed in green vegetation throughout the year export more aromatic carbon. Within basins that show strong seasonal shifts in greenness, correlations exist between SUVA254 and Δ14C-DOC that suggest vegetation is again driving organic carbon composition. There is substantial variation in the strength of the Pearson's correlation coefficients (Table 4) between SUVA254 and Δ14C-DOC within a basin. Interestingly, those systems with high EVI have strong positive correlations between Δ14C-DOC and SUVA254, while those systems that have very low EVI have strong negative correlations between SUVA254 and Δ14C (Figure 8). Removal of the Hudson River makes this comparison clearer as the river basins separate more strongly into groups.

Figure 8.

Linear regression of the Pearson's correlation coefficient (r) between Δ14C-DOC and specific ultraviolet absorbance (SUVA254) within a basin and the annual average MODIS Enhanced Vegetation Index (E.V.I.) for the watersheds of 15 large rivers in the conterminous United States. Red points represent those sites with either positive or negative significant correlations between SUVA and Δ14C-DOC.

[29] In systems like the Rio Grande and Colorado rivers there is a negative correlation between SUVA254 and Δ14C-DOC suggesting that, in less productive watersheds, other factors are controlling the relationship. Both the Colorado and Rio Grande rivers are heavily impounded. Reservoirs increase residence time and have been shown to alter the composition of downstream organic matter through carbon inputs from primary production [Parks and Baker, 1997; Westerhoff and Anning, 2000] and removal of terrestrial DOC. The replacement of allochthonous DOC with algal sources, which have low SUVA254, could explain the strong negative relationship between aromaticity of DOC and its radiocarbon age within the Colorado and Rio Grande rivers. High fluorescence index values also support the importance of in situ production of DOC in these systems (Table 2). However, algal DOC can incorporate the isotopic signature from sources that range in ages from modern (atmospheric CO2 pool) to aged (terrestrial weathering) [Mayorga et al., 2005]. There are insufficient data to discern the exact sources of inorganic carbon in the Colorado and Rio Grande rivers but alkalinity within the Colorado were on average 167 ± 9.1 mg L−1 as CaCO3 for 2009 suggesting a strong influence from carbonate rich weathering materials imported from groundwater sources (USGS National Water Information System – NWIS; In addition, large sources of oil and grease exist along the Rio Conchos, a major tributary above the sampling location associated with the NASQAN station at Brownsville, Texas [Benke and Cushing, 2005]. Therefore, in the case of these arid basins, small inputs of radiocarbon dead material may disproportionately affect the age of DOC while measurements of carbon quality reflect an autochthonous source.

[30] While our data show aged Δ14C-DOC values for the Sacramento River DOC (Table 2), our results are enriched compared to those reported in a previous study [Sickman et al., 2010]. Sickman et al. argued old organic matter found within the Sacramento and San Joaquin rivers is the result of contributions of aged organic material derived from upland soils recovering from agricultural disturbance [Sickman et al., 2010]. However, Michel [1992] demonstrated that only 35% of the outflow of the Sacramento River was from runoff held in the basin for less than one year. The remainder was from older sources, such as groundwater. In addition, the San Joaquin River has also been shown to be impacted by agricultural chemicals [Bergamaschi et al., 2001] that may influence the radiocarbon signatures. Our data Δ14C-DOC for the main stems of these rivers are consistent with agricultural and groundwater influences and suggest either less contribution from older soil organic carbon, a larger in situ source of modern carbon mixing with older carbon, or large inter-annual variation.

[31] In watersheds like the Mobile, Potomac and Susquehanna rivers, SUVA254 correlates strongly with its radiocarbon age. These basins also show dynamic vegetation cycles throughout the year indicated by the EVI (Figure 8). In the case of the Mobile River, the lower stretch flows through coastal plain marshes and the strong relationship may be due to contributions of young, aromatic DOC from the marshes during high flow. The situation for both the Potomac and Susquehanna Rivers is likely due to the relative contributions of surface runoff and groundwater in these rivers. Previous research using tritium to trace water residence times determined that groundwater was an important fraction of flow in both the Susquehanna (20%) and Potomac River basins (54%) [Michel, 1992]. In particular, the diminished contribution of modern surface runoff in the Potomac River (46% of flow) could explain why 14C-DOC correlates strongly with SUVA254. Under conditions when groundwater dominates flow, the river contains older, less aromatic DOC. When surface runoff dominates, the river contains more terrestrially derived organic matter that is younger and more aromatic. The Susquehanna River has a similar pattern, but has a greater contribution of surface runoff (80%). Our data suggest that organic carbon quality and its bulk radiocarbon age responds quickly to both changes in the vegetation on the land surface and the subsequent effect that has on the hydrologic cycle, however determining the portion of the watershed that has the most influence on organic carbon requires more research.

[32] The St. Lawrence River is an outlier in Figure 3 with a modern radiocarbon age and very low SUVA254. Samples reported on here were collected at Cornwall, Ontario where 99% of the flow is from Lake Ontario and the contribution of runoff is relatively low [Benke and Cushing, 2005]. In this system, therefore, more aromatic, terrestrially derived organic matter is less of a factor and the SUVA254 signal is dominated by the production of autochthonous DOC in the Great Lakes. In addition, water residence times for this portion of the St Lawrence are substantially longer due to the Laurentian Great Lakes, and photochemical oxidation of terrestrially derived DOC would result in lower SUVA254 values [Osburn et al., 2001].

[33] In the context of the large temperate river basin chemistry explored here, we suggest that there are two major influences on the radiocarbon age of DOC, climate and humans (Figure 7). Climate determines both the terrestrial productivity of organic material as well as transport of organic matter from terrestrial systems under different precipitation regimes, and, to some degree, the balance between surface water runoff and groundwater contributions to flow. Systems with high productivity and precipitation such as the Connecticut, Altamaha, and the Atchafalaya have a 14C-signature that is dominated by young, aromatic DOM. In an arid system, aquatic residence times are longer, allowing for the removal of young terrestrial DOM and the potential for older sources (e.g., groundwater, agricultural and wastewater byproducts) to express themselves as witnessed in both the Rio Grande and Colorado rivers. This effect is seen in the Potomac as well even with higher rates of precipitation. Although preliminary, we present a framework to think about the broad scale controls on Δ14C-DOC sources to temperate watersheds (Figure 7). Until we are able to definitively identify sources to aged organic carbon within a basin, caution must be used when applying the bulk Δ14C-DOC as a tool to understand the turnover of terrestrial carbon in watersheds.

4.4. Caveats With the Use of Δ14C-DOC

[34] The fact that there is a strong separation of river basins with regard to SUVA254, FI and the Δ14C- DOC suggests that broad scale processes and characteristics influence DOC composition in large river basins. Rivers are considered integrators of the entire watershed, and, in general, the concentrations and fluxes of organic carbon from rivers have been correlated with hydrology, land use, and underlying soils and lithology [Likens and Bormann, 1995]. With respect to DOC age, however, there could be the direct influence of anthropogenic inputs of petroleum-based compounds through surface runoff and wastewater that would dramatically alter the isotopic signature of these rivers [Ahad et al., 2006; Griffith et al., 2009]. In addition, relatively small, point source additions of 14C dead petrochemicals could alter the age of the DOC [Spiker and Rubin, 1975].

[35] We tried to evaluate this effect by exploring the distribution of Δ14C related to the population within a basin using the LandScan (2009) high-resolution global population data set © UT-Battelle, LLC (Oak Ridge National Labs). Population density within 10 km of the sampling station and the flow weighted Δ14C signatures for each river basin are presented in Table 5. There is a trend that suggests local population centers in close-upstream proximity to a sampling location may contribute to an aged DOC signal (Figure 9, r2 = 0.55, p < 0.05). Without the Colorado River the relationship improves (r20.7, p < 0.01). These data raise important questions on the impacts of anthropogenic, petroleum-based carbon sources on the overall14C isotopic signatures of whole water samples. Misinterpretation of the bulk Δ14C-DOC as representative of the turnover of terrestrial carbon as it moves to coastal systems in cases where the introduction of radiocarbon dead material contributes to those ages, would have significant effects on the calculated cycling of carbon within estuarine and marine systems [Opsahl and Benner, 1997]. At the scale of the conterminous U.S. over annual time periods, the dominant source of organic carbon to the coastal ocean is the Mississippi River and, at this scale, the use of radiocarbon as a tracer for terrestrial processing is a powerful tool. However, the research community would greatly benefit from the expansion of compound specific applications of radiocarbon dating to tease apart exactly what components of the organic carbon pool are contributing to its age.

Table 5. Population Information Based on the LANDSCAN Gridded Population Product for the Year 2009 High Resolution Global Population Data Seta
RiverPopulation Density (km2–1) (Basin)Population Density (km−2) (10 km Upstream)Ratio 10 km: Basin
  • a

    UT-Battelle, LLC, Oak Ridge National Laboratories.

Lower Atchafalaya13.359.84.51
Rio Grande22.31512.767.73
San Joaquin45.07.30.16
St. Lawrence44.071.31.62
Figure 9.

Relationship between population near the location of Δ14C-DOC samples and the average flow weighted Δ14C across 15 large river basins in the U.S. Population density is calculated from the LandScan (2009) data set. A 10 km buffer was applied to each sampling point in ArcGIS and used to summarize population density upstream of the sampling location only (Table 5).

5. Conclusion

[36] Across the conterminous United States, DOC exported from large river basins showed radiocarbon ages that ranged from modern (>1950 y.b.p.) to 2470 y.b.p. The average Δ14C-DOC ranged from −92.9‰ in the Colorado River to 73.5‰ in the Altamaha River. When adjusted for the total runoff from the terrestrial landscape, the conterminous U.S. exports modern DOC dominated by the large discharge and modern carbon signatures measured in the Mississippi River basin. Regionally, the contribution of Δ14C-DOC to the ocean differs. Modern carbon enters the Eastern Pacific Ocean, North West Atlantic Ocean, and the Gulf of Mexico, whereas aged DOC represents a significant fraction draining to the Mid-Atlantic Ocean. Each of these 15 large river basins showed unique indicators of organic carbon quality and isotopic signatures. River systems with higher proportions of aromatic carbon produced more enriched Δ14C signatures, while systems where the aromatic component of DOC is less corresponded with an aged carbon fraction. Evidence presented here indicates that large river basins with high discharge and high vegetation cover fractions, have younger and more aromatic DOM. Many of the systems that export the most aged carbon are either arid or have high population density upstream of the sampling locations. Both groundwater and anthropogenic pollution are the most likely sources of the depleted 14C-signatures found in those systems, raising the importance of future research into the effects humans can have on the Δ14C-DOC in freshwaters. Further investigation is needed to evaluate the specific compounds that impart the Δ14C-DOC signatures in terrestrial runoff.


[37] The authors gratefully acknowledge the contributions of many USGS scientists and field personnel that collected the samples reported on here. This study was supported by the U.S. Geological Survey National Stream Quality Accounting Network (, NASA grant NNX09AU89G, and the U.S. Geological Survey National Research Program ( Financial support for this research was provided by a NASA Earth and Space Science Fellowship (NNX07AN83h), a NASA Carbon and Ecosystems Program grant (NNX11AH68G), an NSF-CAREER grant (NSF DEB-0546153) and the Yale School of Forestry and Environmental Studies. The use of brand names in this manuscript is for identification purposes only and does not imply endorsement by the U.S. Geological Survey.