Dissolved organic carbon and chromophoric dissolved organic matter properties of rivers in the USA

Authors


Corresponding author: R. G. M. Spencer, Global Rivers Group, Woods Hole Research Center, 149 Woods Hole Rd., Falmouth, MA 02540, USA. (rspencer@whrc.org)

Abstract

[1] Dissolved organic carbon (DOC) concentration and chromophoric dissolved organic matter (CDOM) parameters were measured over a range of discharge in 30 U.S. rivers, covering a diverse assortment of fluvial ecosystems in terms of watershed size and landscape drained. Relationships between CDOM absorption at a range of wavelengths (a254, a350, a440) and DOC in the 30 watersheds were found to correlate strongly and positively for the majority of U.S. rivers. However, four rivers (Colorado, Colombia, Rio Grande and St. Lawrence) exhibited statistically weak relationships between CDOM absorption and DOC. These four rivers are atypical, as they either drain from the Great Lakes or experience significant impoundment of water within their watersheds, and they exhibited values for dissolved organic matter (DOM) parameters indicative of autochthonous or anthropogenic sources or photochemically degraded allochthonous DOM and thus a decoupling between CDOM and DOC. CDOM quality parameters in the 30 rivers were found to be strongly correlated to DOM compositional metrics derived via XAD fractionation, highlighting the potential for examining DOM biochemical quality from CDOM measurements. This study establishes the ability to derive DOC concentration from CDOM absorption for the majority of U.S. rivers, describes characteristics of riverine systems where such an approach is not valid, and emphasizes the possibility of examining DOM composition and thus biogeochemical function via CDOM parameters. Therefore, the usefulness of CDOM measurements, both laboratory-based analyses and in situ instrumentation, for improving spatial and temporal resolution of DOC fluxes and DOM dynamics in future studies is considerable in a range of biogeochemical studies.

1. Introduction

[2] Dissolved organic matter (DOM) plays a multifaceted role in aquatic ecosystems and represents a fundamental player in global carbon budgets. DOM takes part in a range of processes within freshwater environments including biological, chemical and physical transformations [Jaffé et al., 2008]. The flux of DOM derived from terrestrial net ecosystem production on entering aquatic environments represents an essential link between terrestrial and aquatic ecosystems and dissolved organic carbon (DOC) is the most important intermediate in the global carbon cycle fueling microbial metabolism [Cole et al., 2007; Battin et al., 2008]. For instance, riverine export of DOC provides the largest flux of reduced carbon from land to ocean (0.25 Pg C yr−1), as current POC flux estimates are lower (0.18 Pg C yr−1), and underpins biogeochemical cycling in coastal margins [Hedges et al., 1997; Battin et al., 2008]. With respect to human health, DOM is a water quality constituent of concern as it has been shown to play a role in the formation of carcinogenic and mutagenic disinfection byproducts [Weishaar et al., 2003; Chow et al., 2007] and has also been linked to the transport and reactivity of toxic substances such as mercury [Dittman et al., 2009; Aiken et al., 2011; Bergamaschi et al., 2011]. Therefore, understanding the production, transport and fate of DOM in aquatic ecosystems is of direct relevance to studies addressing issues from water quality to bacterioplankton community structure and function [Crump et al., 2009; Krupa et al., 2012]. Consequently, DOC concentration and DOM composition data for rivers and streams are of interest to a diverse range of scientists and engineers across an assortment of environmental disciplines.

[3] The concentration of DOC in streams and rivers typically ranges from approximately 0.5–50 mgL−1 and is linked to climate and watershed characteristics [Mulholland, 2003]. Although DOC concentration is an extremely important measurement for deriving fluxes across the landscape and examining temporal and spatial trends, it provides little information about the biochemical composition or quality of DOM and hence its biogeochemical role [Hood et al., 2006; Jaffé et al., 2008; Fellman et al., 2009]. Colored or chromophoric dissolved organic matter (CDOM) parameters have been linked to DOM molecular weight [Peuravuori and Pihlaja, 1997; Helms et al., 2008] and composition in a number of recent studies [Boyle et al., 2009; Spencer et al., 2010a; Osburn and Stedmon, 2011]. Furthermore, the ability to not only examine DOM quality but also its biogeochemical processing (e.g., photochemical or microbial degradation) has previously been related to CDOM parameters [Cory et al., 2007; Fellman et al., 2009; Mann et al., 2012]. These relatively straightforward and inexpensive CDOM measurements can be undertaken with small volumes of water, and recent developments now allow for the possibility of in situ observations [Spencer et al., 2007; Saraceno et al., 2009; Pellerin et al., 2012]. The prospect of high-resolution in situ CDOM measurements is opening up the potential for analyses at the temporal and spatial scales required to truly understand DOM dynamics and variability in freshwater ecosystems.

[4] Recent studies have examined the utility of CDOM measurements to derive DOC concentration and examine DOM composition in specific catchment types (e.g., northern high-latitude rivers) [Spencer et al., 2009a]. However, their applicability across a gradient of watershed types including watershed size and landscape drained has to date not yet been addressed. This study examined 30 fluvial sites in the U.S. draining all dominant land cover classes within the U.S. and ranging in size from small headwater streams to the mouth of the Mississippi. The aims of this study were twofold: first, we investigated the possibility of relating CDOM to DOC concentration in the comprehensive range of sites studied. We also tried to determine whether there are any unique features of the watersheds where the CDOM-DOC relationship breaks down that could explain these systems' unsuitability for such an approach. Second, we tested the possibility of utilizing CDOM parameters to address DOM composition across the range of watersheds examined and determined which CDOM parameters may be of the greatest utility to future investigators seeking to address DOM quality in fluvial systems.

2. Materials and Methods

2.1. Study Sites

[5] Thirty sites were examined in this study with the aim of covering the diverse range of watersheds found within the United States, with respect to both watershed size and landscape drained (Figure 1, Table 1). For example, watersheds range in size from headwater streams (e.g., Hubbard Brook WS6; 0.132 km2) to the mouth of the Mississippi River (2,926,686 km2). Focus was especially placed on larger watersheds (e.g., Atchafalaya, Colorado, Columbia, Mississippi, Mobile, Potomac, Rio Grande, Sacramento, San Joaquin, St. Lawrence, Susquehanna and Yukon) near their terminus to examine the applicability of utilizing CDOM to derive DOC export to coastal waters. The rivers chosen also include a diverse range of terrestrial ecosystems including permafrost underlain (e.g., Porcupine), forest (e.g., Androscoggin, Evergreen, Penobscot, Pike), agriculturally impacted (e.g., Mississippi, Sacramento, San Joaquin), urban (e.g., Little Wekiva, Oak Creek), swamp (e.g., Edisto, St. Marys), arid and semi-arid highly regulated systems (e.g., Colorado, Columbia, Rio Grande) and rivers draining from the Great Lakes (e.g., St. Lawrence).

Figure 1.

Map of the study sites within the U.S. See Table 1 for site names and details.

Table 1. Riverine Study Site Numbers as Shown in Figure 1
River NumberRiver NameRiver AbbreviationnSampling PeriodWatershed Size (km2)LatitudeLongitudeDaily MaxQ/MinQ
1AndroscogginAND122006–2007889443.92−69.9735.4
2AtchafalayaATC272008–201024168730.69−91.748.3
3ColoradoCOL272008–201063895032.72−114.7212.7
4ColumbiaCUA182009–201066536746.18−123.187.0
5EdistoEDI152005–2008707133.03−80.3911.2
6EvergreenEVR152002–200516745.07−88.6824.1
7Fishing BrookFBR212007–20096543.98−74.27—–
8Hubbard BrookHBR312005–20070.1344.57−72.254699.2
9HudsonHUD212005–200949843.97−74.13—–
10Little WekivaLWK142002–200611528.70−81.39—–
11Lookout CreekLCR122002–20046244.21−122.26125.0
12Lower AtchafalayaLAT322008–201024630829.69−91.228.3
13MississippiMIS292008–2010292668629.86−89.986.1
14MobileMOB252008–201011136931.09−87.987.5
15NeversinkNEV142005–200617241.89−74.59193.2
16Oak CreekOCR122002–20046542.93−87.87420.8
17PassadumkeagPAS272004–200776945.18−68.47—–
18PenobscotPEN622004–20081946044.83−68.7031.2
19PikePIK142002–200466045.50−88.0012.5
20PorcupinePOR352001–20107640566.99−143.14276.5
21PotomacPOT212008–20102996638.93−77.12165.6
22Rio GrandeRIG272008–201045670025.88−97.45191
23SacramentoSAC222008–20105956938.46121.5012.9
24San JoaquinSAJ232008–20103505837.68−121.2758.1
25St. LawrenceSTL282008–201077388845.01−74.801.6
26St. MarysSTM142002–2006181030.36−82.081970.4
27SusquehannaSUS212008–20107018839.66−76.17304.2
28TananaTAN302000–20076630464.57−149.0916.1
29Yukon at Eagle VillageYRE282000–200229396464.79−141.2021.3
30Yukon at Pilot StationYRP572001–201083138661.93−162.8831.1

2.2. Water Sample Collection and Processing

[6] Water samples were collected across the annual hydrograph to encompass the range of discharge conditions for each study site. The majority of samples were collected as part of the U.S. Geological Survey National Stream Quality Accounting Network (NASQAN) and National Water Quality Assessment (NAQWA) programs from 2000–2010 (Table 1). Sample collection took place over a minimum of two years and up to a maximum of ten years and all analyses were conducted in one laboratory. All water samples were filtered in the field through Gelman AquaPrep 600 capsule filters (0.45 μm) that were pre-rinsed with sample water. The hydrophobic organic acid fraction (HPOA) was obtained following established protocols [Aiken et al., 1992; Spencer et al., 2010b]. In brief, samples were acidified to pH 2 using HCl and passed through a column of XAD-8 resin. The HPOA fraction was retained on the XAD-8 resin and then back eluted with 0.1 M NaOH.

2.3. Dissolved Organic Carbon and Chromophoric Dissolved Organic Matter Analyses

[7] Dissolved organic carbon measurements were carried out on a heated persulfate oxidation OI Analytical Model 700 TOC analyzer [Aiken, 1992]. UV-visible absorbance measurements were undertaken within 48 h of collection on a Hewlett-Packard photo-diode array spectrophotometer (model 8453) between 200 and 800 nm using a 10 mm quartz cell. All samples were analyzed at constant laboratory temperature and sample spectra were referenced to a blank spectrum of distilled water. All absorbance data presented in this manuscript are expressed as absorption coefficients,a(λ), in units of m−1 [Hu et al., 2002]. Chromophoric DOM (CDOM) absorption coefficients (Napierian) were calculated from:

equation image

where A(λ) is the measured absorbance and l is the cell path length in meters. The CDOM absorption ratio at 250 nm to 365 nm was calculated (a250:a365) and SUVA254 values were derived by dividing the UV absorbance (A) at λ = 254 nm by the DOC concentration (mgL−1) and is reported in the units of liter per milligram carbon per meter [Weishaar et al., 2003]. The spectral slope parameter (S) was calculated using a nonlinear fit of an exponential function to the absorption spectrum in the ranges of 275–295 nm and 350–400 nm using the equation:

equation image

where a(λ) is the absorption coefficient of CDOM at a specified wavelength, λref is a reference wavelength and S is the slope fitting parameter. The spectral slope ratio (SR) was calculated as the ratio of S275–295 to S350–400 [Helms et al., 2008].

3. Results

3.1. Bulk Dissolved Organic Carbon and Fractionation

[8] Mean riverine DOC concentrations ranged from 1.0 mgL−1 (0.4 ± SD) in Lookout Creek to 42.1 mgL−1 (16.0 ± SD) in St. Marys (Table 2, Figures 2a–2b). The majority of U.S. rivers had mean riverine DOC concentrations between 2.0–10.0 mgL−1. Mean fraction HPOA ranged from 0.28 (0.03 ± SD) in the St. Lawrence to 0.67 (0.06 ± SD) in St. Marys and the bulk of rivers studied had a mean fraction HPOA between 0.40–0.60 (Table 2, Figures 2c–2d). The HPOA fraction has historically been described as primarily composed of fulvic acid with the remainder as humic acid and thus represents the high molecular weight, aromatic carbon-dominated fraction of DOM [Aiken et al., 1979, 1992]. A higher fraction HPOA therefore typically indicates an increased contribution from allochthonous organic matter sources (i.e., terrestrial), whereas a lower fraction HPOA is indicative of organic matter from autochthonous sources (i.e., algal or microbial) or photodegraded DOM [McKnight and Aiken, 1998; Cory et al., 2007]. For example, microbially dominated Antarctic lakes have been shown to have a fraction HPOA of approximately 0.23 [Aiken et al., 1992], whereas allochthonous-dominated aquatic systems have greater fraction HPOA values (e.g., Arctic blackwater stream = 0.47 [Cory et al., 2007] and Suwannee River = 0.58 [Aiken et al., 1992]). Increasing fraction HPOA is also important with respect to toxic substances such as mercury as it acts as a ligand and studies have shown strong positive linear relationships between the fraction HPOA and dissolved mercury concentration [Schuster et al., 2008; Dittman et al., 2009].

Table 2. Mean (±SD) Riverine Dissolved Organic Carbon (DOC), Fraction HPOA and Chromophoric Dissolved Organic Matter (CDOM) Parameters
River AbbreviationDOC (mgL−1)Fraction HPOASUVA254 (LmgC−1m−1)a250:a365S275–295 (×10−3nm−1)S350–400 (×10−3nm−1)SR
AND6.4 (0.9)0.56 (0.01)3.59 (0.15)5.05 (0.35)14.53 (0.76)16.69 (0.62)0.87 (0.04)
ATC5.2 (0.9)0.47 (0.04)3.22 (0.33)5.25 (0.45)14.85 (0.80)16.63 (0.88)0.89 (0.04)
COL3.1 (0.4)0.39 (0.03)1.67 (0.22)9.05 (1.47)21.69 (1.56)18.99 (2.09)1.15 (0.09)
CUA2.0 (0.3)0.42 (0.04)2.62 (0.43)5.89 (1.09)16.33 (1.37)16.87 (1.13)0.97 (0.06)
EDI8.9 (3.4)0.54 (0.04)3.75 (0.29)4.70 (0.23)13.32 (0.41)16.92 (0.64)0.79 (0.03)
EVR4.5 (3.6)0.45 (0.07)3.08 (0.60)4.89 (0.44)13.62 (0.40)15.97 (0.89)0.86 (0.05)
FBR7.3 (1.9)0.54 (0.03)3.89 (0.29)4.81 (0.13)13.85 (0.62)17.31 (0.28)0.80 (0.03)
HBR3.1 (1.1)0.49 (0.04)2.80 (0.27)6.55 (0.86)16.02 (1.14)20.03 (1.56)0.80 (0.05)
HUD5.3 (1.0)0.52 (0.03)3.48 (0.22)5.16 (0.31)14.64 (0.73)17.32 (0.27)0.84 (0.04)
LWK5.3 (2.5)0.46 (0.04)2.85 (0.44)5.50 (0.46)16.02 (1.05)16.98 (0.59)0.94 (0.05)
LCR1.0 (0.4)0.43 (0.09)2.45 (0.35)6.46 (2.32)13.19 (1.10)16.18 (2.56)0.82 (0.09)
LAT5.1 (0.7)0.47 (0.03)3.13 (0.24)5.41 (0.53)15.32 (0.83)16.88 (0.86)0.91 (0.04)
MIS4.1 (0.5)0.45 (0.02)2.99 (0.23)5.45 (0.61)15.14 (0.80)16.56 (1.24)0.92 (0.06)
MOB5.5 (1.2)0.50 (0.03)3.45 (0.34)4.80 (0.53)14.27 (1.63)16.23 (0.75)0.88 (0.08)
NEV2.1 (1.5)0.47 (0.05)2.47 (0.66)6.65 (2.43)15.53 (1.65)17.86 (1.27)0.86 (0.10)
OCR6.7 (1.9)0.44 (0.04)2.86 (0.52)6.19 (0.76)15.46 (1.20)17.92 (1.05)0.86 (0.06)
PAS12.0 (4.3)0.61 (0.04)4.19 (0.30)4.56 (0.17)13.38 (0.59)16.44 (0.24)0.81 (0.04)
PEN9.8 (2.9)0.58 (0.03)3.80 (0.26)4.98 (0.30)14.13 (0.89)17.05 (0.33)0.83 (0.04)
PIK8.0 (5.0)0.52 (0.07)3.71 (0.55)4.87 (0.42)14.28 (1.61)17.04 (0.29)0.80 (0.07)
POR10.0 (5.9)0.51 (0.05)3.02 (0.55)5.72 (1.14)15.54 (2.06)18.17 (1.67)0.86 (0.08)
POT3.3 (0.6)0.40 (0.04)2.31 (0.33)5.73 (0.77)15.74 (1.71)16.58 (1.79)0.96 (0.13)
RIG4.8 (0.5)0.35 (0.03)2.03 (0.24)7.38 (0.84)19.80 (1.45)17.76 (1.36)1.12 (0.11)
SAC2.4 (0.8)0.39 (0.05)2.41 (0.51)5.34 (0.72)15.69 (1.43)16.35 (1.26)0.96 (0.07)
SAJ3.8 (1.1)0.41 (0.03)2.47 (0.25)5.93 (1.21)15.71 (0.73)16.87 (0.87)0.93 (0.04)
STL2.7 (0.2)0.28 (0.03)1.31 (0.16)9.65 (2.30)22.96 (1.76)18.95 (3.17)1.23 (0.16)
STM42.1 (16.0)0.67 (0.06)4.56 (0.28)4.20 (0.27)12.47 (0.78)16.65 (0.57)0.75 (0.06)
SUS2.5 (0.5)0.40 (0.04)2.25 (0.28)5.79 (0.68)15.33 (1.57)17.34 (1.45)0.89 (0.14)
TAN3.2 (1.9)0.46 (0.07)2.68 (0.54)5.66 (1.10)16.13 (1.78)17.77 (1.90)0.89 (0.06)
YRE5.2 (3.0)0.50 (0.06)3.00 (0.64)5.93 (1.12)15.95 (2.02)18.17 (2.14)0.88 (0.07)
YRP8.3 (5.0)0.51 (0.05)3.08 (0.47)5.52 (0.90)15.26 (1.87)17.89 (1.45)0.85 (0.07)
Figure 2.

Box plots of (a–b) DOC, and (c–d) fraction of hydrophobic organic acid fraction (HPOA) for the 30 rivers. Note the different y axis scale between Figures 2a and 2b. The black dotted line and the black solid line represent the mean and the median, respectively. The horizontal edges of the boxes denote the 25th and 75th percentiles, the whiskers denote the 10th and 90th percentiles, and black circles represent outliers.

3.2. Chromophoric Dissolved Organic Matter

[9] Mean SUVA254 values in the rivers examined ranged from 1.31 L mg C−1 m−1 (0.16 ± SD) for the St. Lawrence to 4.56 L mg C−1 m−1 (0.28 ± SD) in St. Marys (Table 2, Figures 3a–3b). The majority of the rivers examined in this study had mean SUVA254 values between 2.00 and 3.80 L mg C−1 m−1 (Table 2, Figures 3a–3b). SUVA254 values have been positively correlated to the percent aromaticity of DOM as measured by 13C-NMR [Weishaar et al., 2003]. The lowest mean SUVA254values observed in U.S. rivers are comparable to values reported for HPOA isolates from microbial-dominated end-members such as Pony Lake (1.7 L mg C−1 m−1) and Lake Fryxell (1.8 L mg C−1 m−1; Weishaar et al., 2003], and aquatic systems with little vascular plant input (e.g., groundwaters: 1.3–1.6 L mg C−1 m−1) [Spencer et al., 2008]. Similarly, the highest mean SUVA254values reported in this study are comparable to values for HPOA isolates from allochthonous-dominated end-members (e.g., Ogeechee and Suwannee Rivers; 3.2–5.3 L mg C−1 m−1) [Weishaar et al., 2003] and aquatic systems with significant vascular plant inputs (e.g., blackwaters: 3.4–4.5 L mg C−1 m−1) [Spencer et al., 2008, 2010a].

Figure 3.

Box plots of (a–b) SUVA254, and (c–d) a250:a365 for the 30 rivers. The black dotted line and the black solid line represent the mean and the median, respectively. The horizontal edges of the boxes denote the 25th and 75th percentiles, the whiskers denote the 10th and 90th percentiles, and black circles represent outliers.

[10] The a250:a365 ratio has previously been related to the aromatic content and molecular size of DOM with increasing values indicating a decrease in aromaticity and molecular size [Peuravuori and Pihlaja, 1997]. Mean a250:a365 values ranged from 4.20 (0.27 ± SD) in St. Marys to 9.65 (2.30 ± SD) in the St. Lawrence, with the bulk of a250:a365 mean values in the rivers examined ranging from 5.00–6.50 (Table 2, Figures 3c–3d). The lowest and highest mean a250:a365values in St. Marys and the St. Lawrence are comparable to allochthonous-dominated blackwaters of the Great Dismal Swamp (4.57–4.64) and coastal waters (e.g., Georgia Bight = 8.7 ± 1.4), respectively [Helms et al., 2008].

[11] The spectral slope parameter (S) describes the spectral dependence of the CDOM absorption coefficient and as a result provides information with respect to the quality of the CDOM [Blough and Del Vecchio, 2002). S has been shown to vary with the source of CDOM and also to be sensitive to biological and photochemical alteration of the source material [Stedmon and Markager, 2001; Obernosterer and Benner, 2004; Osburn and Stedmon, 2011]. Typically, a steeper S value has been related to a decrease in molecular weight and aromaticity of DOM [Blough and Green, 1995; Helms et al., 2008]. Historically, S has been calculated over a range of wavelengths and 275–295 nm (S275–295) and 350–400 nm (S350–400) were chosen because Helms et al. [2008] in their extensive study of S in a range of aquatic ecosystems and DOM sources observed the first derivative of the natural log spectra had the greatest variation in these ranges. The slope ratio (SR) of S275–295: S350–400 has also been shown to be sensitive to characterizing CDOM in natural waters, with lower relative values indicative of DOM of higher molecular weight, greater aromaticity and increasing vascular plant inputs [Helms et al., 2008; Spencer et al., 2010a; Osburn et al., 2011].

[12] Mean S275–295 values ranged from 12.47 × 10−3 nm−1 (0.78 ± SD) in St. Marys to 22.96 × 10−3 nm−1 (1.76 ± SD) in the St. Lawrence and the majority of U.S. rivers exhibited S275–295 values between 13.00–16.50 × 10−3 nm−1 (Table 2, Figures 4a–4b). The shallowest mean S275–295values are comparable to allochthonous-dominated waters such as the Congo River (12.34 × 10−3 nm−1) [Spencer et al., 2009b], the Yukon River at the peak of the freshet (12.28 × 10−3 nm−1) [Spencer et al., 2009a] and the Great Dismal Swamp (12.7–12.8 × 10−3 nm−1) [Helms et al., 2008]. The steepest mean S275–295 values are comparable to data from U.S. coastal waters (e.g., 24.00 × 10−3 nm−1 in the Georgia Bight [Helms et al., 2008] and 22.00–28.00 × 10−3 nm−1 in surface waters of the northern Gulf of Mexico [Shank and Evans, 2011]), and the minimum values for lakes in the Great Plains (e.g., 22.18 × 10−3 nm−1) [Osburn et al., 2011], which represent DOM from autochthonous sources and photochemically degraded allochthonous DOM. Mean S350–400 values followed a similar trend to S275–295 values but covered a narrower range with shallowest values in Evergreen River of 15.97 × 10−3 nm−1 (0.89 ± SD) and steepest values in Hubbard Brook of 20.03 × 10−3 nm−1 (1.56 ± SD) (Table 2, Figures 4c–4d). Most U.S. rivers had S350–400 values between 16.50–18.25 × 10−3 nm−1. As observed for S275–295, the steepest mean S350–400 values are comparable to previously reported data for coastal waters (18.00–19.00 × 10−3 nm−1) [Shank and Evans, 2011] and prairie lakes (22.41 × 10−3 nm−1) [Osburn et al., 2011]. Similarly, the shallowest mean S350–400 values are analogous to data reported from aquatic ecosystems with high allochthonous inputs such as the Congo River (15.21 × 10−3 nm−1) [Spencer et al., 2009b].

Figure 4.

Box plots of (a–b) S275–295, (c–d) S350–400, and (e–f) SR for the 30 rivers. Note the different y axis scale between Figures 4a and 4b and between Figures 4e and 4f. The black dotted line and the black solid line represent the mean and the median, respectively. The horizontal edges of the boxes denote the 25th and 75th percentiles, the whiskers denote the 10th and 90th percentiles, and black circles represent outliers.

[13] The mean SR values ranged from 0.75 (0.06 ± SD) in St. Marys to 1.23 (0.16 ± SD) in the St. Lawrence, with the majority of U.S. rivers ranging from 0.80–0.95 (Table 2, Figures 4e–4f). Lower mean SR values are similar to data reported from Arctic rivers at the peak of the freshet (e.g., Yukon = 0.79; Yenisey = 0.79) when they receive significant vascular plant inputs [Spencer et al., 2009a; Stedmon et al., 2011], or blackwater tropical rivers during the onset of the wet season (0.79) [Spencer et al., 2010a]. The highest mean SR values are comparable to mean values from prairie lakes (1.36) [Osburn et al., 2011] and coastal waters (1.20–1.40) [Shank and Evans, 2011].

4. Discussion

4.1. Deriving DOC Concentration From CDOM in U.S. Rivers

[14] Historically, relationships between CDOM and DOC have principally been examined in coastal waters [Ferrari et al., 1996; Vodacek et al., 1997; Rochelle-Newall and Fisher, 2002]. Although CDOM represents only a fraction of the total DOC pool a number of studies have reported strong correlations between CDOM properties and DOC concentration in coastal waters [see Del Vecchio and Blough, 2004, and reference therein; Mannino et al., 2008]. The investigation of CDOM and DOC relationships in riverine environments is also extremely pertinent to facilitate the development of improved DOC flux estimates through increased temporal coverage via recently developed in situ instrumentation [Spencer et al., 2007; Saraceno et al., 2009; Downing et al., 2009; Pellerin et al., 2012]. Furthermore, a recent study by Griffin et al. [2011] highlighted the potential to derive DOC via CDOM in a major river (Kolyma, Siberia) using Landsat satellite imagery. Therefore, if robust CDOM versus DOC relationships can be derived for rivers they can be utilized to examine shifts in the export and timing of the flux of DOC in highly dynamic periods (e.g., storms, snowmelt) in small watersheds when the majority of DOC export occurs [Inamdar et al., 2006; Saraceno et al., 2009; Pellerin et al., 2012], and also to improve estimates of the land-ocean flux of terrestrial DOC from major rivers [Spencer et al., 2009a]. To assess this objective in the diverse range of U.S. rivers studied here with respect to both watershed size and landscape drained we examined CDOM absorption relationships at a254, a350 and a440 to DOC (Figure 5, Table 3).

Figure 5.

Relationships between dissolved organic carbon (DOC) and chromophoric dissolved organic matter (CDOM) absorption (a254, black circles, black line; a350, gray circles, gray line; and a440, white circles, black dashed line): (a) Mississippi River and (b) Hubbard Brook.

Table 3. Relationships Between Dissolved Organic Carbon (DOC) and CDOM Absorption Coefficients (a254, a350 and a440)
Rivera254a350a440
r2PSlopeStandard ErrorInt.Standard Errorr2PSlopeStandard ErrorInt.Standard Errorr2PSlopeStandard ErrorInt.Standard Error
AND0.930<0.00019.080.829−5.235.380.7690.00042.560.467−2.553.030.4990.01520.6470.216−0.8061.40
ATC0.904<0.000110.80.706−17.13.710.837<0.00013.280.290−7.011.520.699<0.00010.9410.123−2.450.648
COL0.903<0.00017.440.489−11.11.540.740<0.00011.830.217−3.760.6840.1810.02680.4040.172−1.030.542
CUA0.701<0.00019.871.61−7.573.260.4310.00312.420.694−1.931.400.1380.1300.5580.350−0.4760.707
EDI0.969<0.00018.860.438−1.334.170.940<0.00012.410.1690.2811.610.820<0.00010.5230.06790.3110.647
EVR0.994<0.000110.90.226−12.91.290.992<0.00013.090.0756−3.880.4290.961<0.00010.6720.0376−0.7470.214
FBR0.935<0.000110.60.553−11.34.160.930<0.00013.020.196−3.911.470.927<0.00010.6820.0452−1.000.340
HBR0.937<0.00016.990.337−1.581.110.915<0.00011.820.103−1.210.3380.608<0.00010.3230.0482−0.2730.158
HUD0.887<0.00017.670.6271.673.370.791<0.00012.050.2410.1171.300.5490.00010.4820.100−0.2330.538
LWK0.983<0.00018.770.335−9.631.950.973<0.00012.350.114−3.330.6640.949<0.00010.5800.0387−1.020.226
LCR0.913<0.00016.370.623−0.8320.6860.857<0.00012.000.258−0.6340.2840.3440.04510.4980.218−0.3660.240
LAT0.869<0.00019.880.699−13.63.6130.785<0.00013.020.289−6.271.500.495<0.00010.9490.175−2.640.905
MIS0.843<0.00018.910.740−8.253.060.707<0.00012.820.350−4.541.450.3640.00051.100.280−2.761.16
MOB0.823<0.00018.470.819−2.984.590.557<0.00012.130.3970.2782.220.3060.00410.4500.1410.4420.792
NEV0.982<0.00019.300.368−5.510.9340.969<0.00012.590.133−2.020.3390.833<0.00010.5750.0742−0.6030.189
OCR0.957<0.00019.380.629−17.54.360.848<0.00012.380.319−5.502.210.5880.00360.4810.127−1.210.883
PAS0.966<0.00019.680.3620.2874.620.935<0.00012.870.127−0.9361.620.918<0.00010.7130.0428−0.3910.546
PEN0.951<0.00018.520.2492.162.550.934<0.00012.570.0881−2.130.9030.891<0.00010.6410.0289−0.9420.296
PIK0.999<0.000110.60.083−11.60.7660.998<0.00012.970.0342−3.710.3180.915<0.00010.6240.0548−0.4380.509
POR0.988<0.00018.160.156−7.011.810.981<0.00012.290.0560−3.880.6480.949<0.00010.5550.0219−1.250.254
POT0.855<0.00018.190.775−9.192.580.714<0.00012.390.346−3.521.150.5350.00020.6690.143−1.240.477
RIG0.4530.00015.661.24−4.655.960.2650.00601.280.427−1.852.050.01400.5570.09880.1660.3570.794
SAC0.945<0.000110.40.563−10.91.450.929<0.00013.380.209−4.420.5360.868<0.00011.070.0930−1.710.239
SAJ0.961<0.00017.350.325−5.841.280.889<0.00011.750.135−1.440.5320.628<0.00010.4230.0710−0.5380.280
STL0.2060.01542.200.8482.212.330.0370.3280.2750.2750.5460.7570.00340.769−0.06720.2270.3400.623
STM0.975<0.00019.750.45620.421.90.981<0.00013.570.144−13.76.460.960<0.00010.8020.0474−2.192.12
SUS0.822<0.00016.670.712−3.671.800.642<0.00011.650.282−1.030.7130.3360.01570.2880.119−0.09400.301
TAN0.966<0.00018.010.286−4.401.070.915<0.00012.030.117−1.260.4380.501<0.00010.3750.0708−0.09210.265
YRE0.982<0.00019.180.243−8.661.460.952<0.00012.580.114−3.690.6830.779<0.00010.6400.0669−1.250.401
YRP0.985<0.00018.330.138−6.741.330.938<0.00012.430.0844−3.700.8160.617<0.00010.7010.0745−1.760.720

[15] Absorption coefficients at 254, 350 and 440 nm correlated strongly and positively with DOC concentration for the majority of U.S. rivers (Table 3). Examples of relationships between DOC and a254, a350 and a440 are shown for the smallest (Hubbard Brook WS6) and largest (Mississippi) watersheds studied in Figure 5. The relationship between absorption coefficient and DOC concentration varied between the wavelengths studied with typically stronger relationships observed at shorter wavelengths. CDOM absorption spectra typically decrease in an approximately exponential fashion with increasing wavelength, and so the accuracy of CDOM measurements decreases at longer wavelengths resulting in a weakening in the correlation [Baker et al., 2008]. This is particularly the case for samples exhibiting low CDOM absorption values.

[16] In the U.S. rivers examined in this study a number of rivers consistently standout as having statistically weak relationships between CDOM and DOC concentration. The Rio Grande (r2 = 0.453; p = 0.0001) and the St. Lawrence (r2 = 0.206; p = 0.0154) are the only two rivers that do not exhibit a statistically significant relationship at the <0.0001 significance level for DOC versus a254, with the relationship explaining over 70% of the variance in all other rivers (Table 3). With respect to DOC versus a350 the Rio Grande and the St. Lawrence exhibit weak correlations (r2 = 0.265; p = 0.0060 and r2 = 0.037; p = 0.3279, respectively), as does the Colombia (r2 = 0.431; p = 0.0031), with the relationship explaining over 55% of the variance in all other rivers at a significance level of <0.0005 (Table 3). Similarly, for DOC versus a440 the Colombia (r2 = 0.138; p = 0.130), the Rio Grande (r2 = 0.0140; p = 0.557) and the St. Lawrence (r2 = 0.0034; p = 0.769) stand out as having weak correlations, and the Colorado also shows a poor correlation (r2 = 0.181; p = 0.0268) with respect to DOC versus a440. These four rivers (Colorado, Colombia, Rio Grande and St. Lawrence) represent in many ways atypical systems from the other rivers in this study, as they all exhibit values for DOM parameters indicative of autochthonous or anthropogenic sources, or photochemically degraded allochthonous DOM (Table 2, Figures 24). The St. Lawrence (the river with the weakest correlations between DOC and absorption coefficients) exhibits the lowest mean values for fraction HPOA and SUVA254 and the highest a250:a365, S275–295 and SR values of all rivers studied (Table 2, Figures 24). Draining from the Great Lakes the St. Lawrence is dominated by autochthonous DOM [Cotner et al., 2004; Sterner, 2010] and photochemically modified allochthonous DOM [Biddanda and Cotner, 2003; Hiriart-Baer et al., 2008]. DOM produced via autochthonous processes has previously been shown to be decoupled from DOC accumulation, and photodegradative processes are known to remove CDOM preferentially over DOC [Moran et al., 2000; Opsahl and Zepp, 2001; Rochelle-Newall and Fisher, 2002]. Thus, the lack of a relationship between DOC and CDOM absorption coefficients in the St. Lawrence is not surprising due to the dissociation between the CDOM and DOC pools.

[17] The Colorado, Colombia and Rio Grande rivers also exhibit fraction HPOA values and CDOM parameters indicative of systems dominated by autochthonous DOM or DOM derived from anthropogenic sources (e.g., wastewater), or photochemically degraded allochthonous DOM (Table 2, Figures 24). The Colorado is heavily regulated with over 20 major dams and extensive reservoirs along its course, including the two largest reservoirs in the U.S. (Lake Mead and Lake Powell). The Columbia has also been significantly dammed with 14 major dams on the main stem including the Grand Coulee Dam (impounding the sixth largest reservoir in the U.S.). The Rio Grande is also extensively modified with significant water withdrawal, water impoundment (e.g., Amistad Dam, Falcon Dam, Elephant Butte Dam) and levee construction disconnecting the river from its floodplains [Valett et al., 2005]. These three rivers all drain semi-arid to arid areas in the intermontane plateaus of the U.S. western states that receive intense solar irradiance. Therefore, with the significant impoundment of water within these watersheds it seems apparent that an increased relative contribution of groundwater and wastewater derived DOM in these systems, or autochthonous production along with potentially photochemical degradation of allochthonous DOM within reservoirs leads to the DOM exported exhibiting low values for fraction HPOA and SUVA254 and high a250:a365, S275–295 and SR values in comparison to the majority of rivers studied (Table 2, Figures 24), and a decoupling of CDOM and DOC (Table 3).

[18] The strong correlations observed between DOC and a254, a350 and a440 for U.S. rivers except the Colorado, Colombia, Rio Grande and St. Lawrence are indicative of the dominance of predominantly allochthonous DOM in these watersheds (Table 3, Figure 5). The relationship between DOC and CDOM absorption for all U.S. rivers also generally exhibits a negative intercept and is always potentially negative within standard error (Table 3) clearly showing that not all DOC is chromophoric. The DOC versus CDOM linear regression lines typically also become steeper (i.e., increased absorption per unit DOC) in rivers with greater vascular plant derived character. For instance, St. Marys has the highest fraction HPOA and SUVA254 values and lowest a250:a365, S275–295 and SR values, and for DOC versus a350 the steepest slope (3.57 ± 0.144 SE; Table 3). Conversely, the Rio Grande that has a much lower fraction HPOA and SUVA254 values and higher a250:a365, S275–295 and SR values, has a relatively shallower slope for DOC versus a350 (1.28 ± 0.427 SE; Table 3). The greater vascular plant derived character of St. Marys reflect its source in one of the largest freshwater wetlands in the World (Okefenokee Swamp) in comparison to the heavily regulated Rio Grande with little wetland area within its semi-arid and arid watershed. Therefore, future studies may be able to link both DOC versus absorption coefficient relationships and thus improve DOC export fluxes, as well as CDOM compositional parameters to watershed characteristics such as wetland coverage or degree of impoundment.

4.2. Examining DOM Composition From CDOM in U.S. Rivers

[19] The utility of DOC fractionation via XAD-8 resin has been extensively shown since the 1970s and the isolates derived from XAD resins are used by the International Humic Substances Society (IHSS) to produce their well-studied and widely used reference materials (e.g., Suwannee River Fulvic Acid). Such DOM reference materials have been characterized to a degree not historically possible with wholewaters and have been comprehensively studied in controlled experiments linking DOM composition to its properties [Weishaar et al., 2003; Stubbins et al., 2008; Boyle et al., 2009]. As such fraction HPOA represents a useful metric for DOM source and processing and has also been linked to the transport of toxic substances such as mercury [Cory et al., 2007; Dittman et al., 2009]. However, fractionation of DOC is expensive, time consuming, requires large volumes of water (≥1 L) and is analytically demanding, and so is not logistically feasible for examining DOM compositional changes at high temporal resolution. However, DOM compositional changes have been reported in rivers on hourly timescales [Hood et al., 2006; Fellman et al., 2009], even within large rivers such as the San Joaquin [Spencer et al., 2007]. Although a number of studies have examined relationships between components of the DOM pool (e.g., lignin phenols) [Hernes and Benner, 2003] and absorption coefficients, few studies have to date investigated relationships between DOM biochemical and optical properties [Spencer et al., 2010a; Osburn and Stedmon, 2011]. In order to examine the potential of developing a CDOM derived proxy for fraction HPOA and thus understand DOM biogeochemical function (e.g., microbial and photochemical reactivity) we explored relationships between the fraction HPOA and SUVA254, a250:a365, S275–295, S350–400 and SR.

[20] Significant linear correlations were observed between fraction HPOA and SUVA254, a250:a365, S275–295 and SR (r2 = 0.89, r2 = 0.56, r2 = 0.55, r2 = 0.66 respectively; P < 0.0001; Figure 6). No significant relationship was found between fraction HPOA and S350–400 although generally samples with higher mean fraction HPOA exhibited mean shallower S350–400 values (e.g., St. Marys; HPOA = 0.67; S350–400 = 16.65 × 10−3 nm−1), and on the contrary samples with lower mean fraction HPOA had mean steeper S350–400 values (e.g., St. Lawrence; HPOA = 0.28; S350–400 = 18.95 × 10−3 nm−1; Table 3). The lack of a relationship between fraction HPOA and S350–400 could be because the range of S350–400 values is much smaller than observed for S275–295 (Table 2) and so has a smaller statistical dispersion versus fraction HPOA (i.e., the greater range in S275–295 potentially allows for clearer distinction versus fraction HPOA). The S275–295 parameter can also be measured with greater precision, especially in optically clear waters and so is more robust for examination across a broad range of aquatic systems [Helms et al., 2008].

Figure 6.

Relationships between fraction HPOA and CDOM parameters for the 30 rivers: (a) SUVA254, (b) a250:a365, (c) S275–295, and (d) SR. Black circle represents the mean value for an individual river and error bars represent the standard deviation.

[21] SUVA254 exhibits a very robust positive relationship across the 30 rivers studied with fraction HPOA (Figure 6a). This easily derived parameter from DOC concentration and UV absorbance at 254 nm therefore provides a good surrogate for fraction HPOA. Although SUVA254 is a straightforward measurement requiring little sample volume and can have a high throughput, it cannot be measured in situ. Solely CDOM derived parameters such as a250:a365, S275–295 and SR do have the possibility however to be utilized by future in situ studies. These three measurements explain between 55% and 66% of the variance between fraction HPOA and the CDOM parameters (Figures 6b–6d) and thus highlight the ability to examine DOM composition from CDOM parameters in riverine systems. A previous study focusing across a similarly diverse range of surface waters encompassing North American streams, lakes and estuarine environments reported a significant relationship between SUVA254 and DOM C:N ratio [Jaffé et al., 2008]. That study also reported a significant relationship between DOM C:N ratio and fluorescence index (the ratio of emission intensities at 470 and 520 nm at an excitation wavelength of 370 nm [Cory et al., 2010]), which also has the potential to be measured via in situ technologies. It seems apparent that CDOM can be utilized in a broad range of freshwater ecosystems to not only derive improved fluxes of constituents of interest (e.g., DOC, lignin phenols, mercury) but also to examine DOM composition. Such an approach employing recently developed in situ technologies will allow for enhanced understanding of the biogeochemical role that DOM plays in freshwater ecosystems at relevant temporal resolution, as well as opening up the possibility for improving spatial resolution of how DOM is modified within aquatic environments (i.e., from source to sea).

4.3. Future Implications

[22] The ability to derive DOC concentration via CDOM absorption in the majority of U.S. rivers examined has implications for improving flux estimates through increased temporal resolution via in situ instrumentation. The fate of terrigenous DOC in the ocean remains a pertinent biogeochemical question, and the prospect of improving load estimates for terrigenous DOC and biomarkers such as lignin via CDOM has ramifications that may help answer this paradox [Hedges et al., 1997; Spencer et al., 2009a]. Naturally, calculation of the residence time of terrigenous DOC in the ocean is dependent on accurately constraining the size of fluxes and reservoirs, and therefore any technique that can improve flux estimates is of assistance in improving residence time calculations. Recent studies utilizing CDOM derived estimates of terrigenous fluxes have shown increased flux estimates over historical studies due to the ability to easily increase temporal resolution [Spencer et al., 2009a; Osburn and Stedmon, 2011]. If such an approach across a wide range of watersheds led to an increase from past estimates in terrigenous DOC flux to marine environments, this would highlight a potentially greater role for removal mechanisms (e.g., photochemical and microbial degradation) and further underpin recent work showing terrigenous DOC to be more reactive than historically thought [Holmes et al., 2008; Alling et al., 2010; Stubbins et al., 2010].

[23] Aquatic ecosystems are currently facing a range of stressors from climate change, land-use changes (e.g., deforestation, conversion to agriculture, urbanization), engineering (e.g., channelization, impoundment) and the impacts of pollution. As shown in this study (Figure 6), the ability to not just examine DOC concentration but also DOM composition via CDOM holds great potential. For example, DOC flux may not be changing from an impacted watershed but the quality of DOM exported may be fundamentally altered in a watershed in transition [Bernardes et al., 2004]. The biochemical character of DOM is key with respect to its biogeochemical role in downstream ecosystems [Crump et al., 2009; Mann et al., 2012]. Therefore, the ability to improve temporal and spatial resolution of measurements of DOM quality via CDOM parameters provides an extremely useful mechanism by which to scale up DOM compositional measurements in future studies and to assist in delineating the complex role DOM plays in aquatic environments. Particularly in highly dynamic periods within watersheds such as storms and spring freshets when both DOC concentration and DOM composition change rapidly with concurrent shifts in biolability [Fellman et al., 2009; Mann et al., 2012], high resolution in situ CDOM measurements will be of great value for understanding the role and fate of DOM. Deployment of in situ CDOM instrumentation in freshwater ecosystems is still in its early stages but a number of studies have examined DOM variability in watersheds [Spencer et al., 2007; Saraceno et al., 2009; Pellerin et al., 2012]. The data presented in this manuscript clearly show the ability to derive DOC loads from CDOM and discharge measurements for the majority of U.S. rivers studied, and also to examine DOM composition and thus biogeochemical function via CDOM parameters which can be measured in situ (e.g., S275–295). Therefore, the potential of CDOM measurements, both traditional laboratory-based analyses and in situ instrumentation to improve spatial and temporal resolution of DOC fluxes and DOM dynamics in future studies is considerable.

Acknowledgments

[24] The authors gratefully acknowledge the contributions of many U.S. Geological Survey scientists and field personnel who collected the samples reported on here. This study was supported by the U.S. Geological Survey National Stream Quality Accounting Network (http://water.usgs.gov/nasqan), the U.S. Geological Survey National Water Quality Assessment Program (http://water.usgs.gov/nawqa/), the U.S. Geological Survey National Research Program (http://water.usgs.gov/nrp), NASA grant NNX09AU89G and NSF ETBC grant 0851101. The use of brand names in this manuscript is for identification purposes only and does not imply endorsement by the U.S. Geological Survey. For producing Figure 1, we thank Greg Fiske at the Woods Hole Research Center, and for their helpful comments on the manuscript, we thank two anonymous reviewers, the Associate Editor and the Editor.

Ancillary