Amounts, isotopic character, and ages of organic and inorganic carbon exported from rivers to ocean margins: 1. Estimates of terrestrial losses and inputs to the Middle Atlantic Bight

Authors

  • Katie Hossler,

    Corresponding author
    1. Department of Evolution, Ecology and Organismal Biology, The Ohio State University, Columbus, Ohio, USA
    • Corresponding author: Katie Hossler, Aquatic Biogeochemistry Laboratory, Department of Evolution, Ecology and Organismal Biology, The Ohio State University, 1314 Kinnear Road, Columbus, OH 43212-1156, USA. (hossler.3@osu.edu)

    Search for more papers by this author
  • James E. Bauer

    1. Department of Evolution, Ecology and Organismal Biology, The Ohio State University, Columbus, Ohio, USA
    Search for more papers by this author

Abstract

[1] Rivers transport carbon (C) from terrestrial ecosystems to the coastal ocean, providing significant heterotrophic support within both rivers and receiving coastal waters. The amounts and ages of these terrestrial-river-coastal ocean C fluxes, however, are still poorly constrained. To address this uncertainty, a study of eight rivers discharging to the Middle Atlantic Bight (MAB) was undertaken. The rivers were sampled periodically over 2 years for concentrations and δ13C and Δ14C signatures of particulate organic C (POC), dissolved organic C (DOC), and dissolved inorganic C (DIC). For the watersheds draining to the MAB, it was estimated that ∼3800 Gg of terrestrial organic C (OC) and 700 Gg of terrestrial inorganic C was removed annually by fluvial transport. Of the terrestrial OC loss, ∼64% was contemporary C representing approximately 1% of the annual terrestrial net primary productivity. Net fluvial C inputs to the MAB shelf were estimated to be ∼70 Gg·yr − 1 of POC, 280 Gg·yr − 1 of DOC, and 800 Gg·yr − 1 of DIC. Terrestrial C, as opposed to in situ produced river C, comprised the majority of the riverine POC and DOC flux and around half of the total C flux. A smaller but significant fraction (<25%) of the river C flux was further composed of aged materials deriving from fossil C and aged soil OC. The timing of fluvial OC inputs to the MAB, which appear to be temporally offset from peak MAB primary production, could help support the net heterotrophy that has been observed there during periods of low productivity.

1 Introduction

[2] Rivers play an integral role in the global carbon (C) cycle by actively transporting different forms of organic and inorganic C (OC and IC, respectively) from terrestrial environs to coastal ocean margins [Cole et al., 2007; Battin et al., 2009; Tranvik et al., 2009; Aufdenkampe et al., 2011; Bauer and Bianchi 2011]. On the terrestrial side, rivers and other inland waters represent a significant fate of terrestrial net primary productivity (NPP), with global NPP estimates being on the order of 56 Pg·yr − 1 of C [Cramer et al., 1999; Imhoff et al., 2004a; Ito, 2011], of which ∼3 Pg·yr − 1 of C is exported to inland waters (i.e., ∼5% of terrestrial NPP) [Cole et al., 2007; Battin et al., 2009; Tranvik et al., 2009; Aufdenkampe et al., 2011]. Loadings of terrigenous, or allochthonous, organic matter (OM) support the predominant heterotrophy observed in inland aquatic ecosystems [Cole and Caraco, 2001; Duarte and Prairie, 2005; Bauer and Bianchi, 2011], resulting in a global net CO2 efflux from these waters on the order of 1 Pg·yr − 1 of C [Cole et al., 2007; Battin et al., 2009; Tranvik et al., 2009; Aufdenkampe et al., 2011]. Of the remaining C (including some in situ produced, or autochthonous, C), a portion is stored via sedimentation within freshwater systems and the remainder—approximately 1 Pg·yr − 1 of C globally—is exported to coastal ocean margins [Meybeck, 1982, 1993; Cole et al., 2007; Battin et al., 2009; Tranvik et al., 2009; Aufdenkampe et al., 2011].

[3] Of the exported riverine C, approximately half is OC in the forms of particulate and dissolved OC (POC and DOC, respectively) [Meybeck, 1982, 1993; Cole et al., 2007; Bauer and Bianchi, 2011]. The majority of the exported OC may be remineralized in nearshore waters [Cai, 2011; Bauer and Bianchi, 2011; Bianchi and Bauer, 2011], supporting both the net heterotrophy observed in estuaries and inner continental shelves (i.e., nearshore) and the net autotrophy (e.g., through release of nutrients during remineralization) reported for middle and outer continental shelves, [Smith and Hollibaugh, 1993; Chen and Borges, 2009; Borges and Abril, 2011; Cai, 2011]. Some fraction of refractory fluvial OC will be further exported to the middle and outer shelves and still smaller amounts ( < ∼2% of river exported OC) exported to the open ocean [e.g., Meyers-Schulte and Hedges, 1986; Opsahl and Benner, 1997; Hernes and Benner, 2006]. Riverine IC comprises the other half of exported riverine C [Meybeck, 1982, 1993; Cole et al., 2007]. Data from a limited number of estuarine systems suggest that 0% to 90% of the initial riverine dissolved IC (DIC) pool may be lost through degassing during transit through nearshore waters [Cai and Wang, 1998; Abril et al., 2000; Hunt et al., 2011]. However, for DIC as well as POC and DOC, significant C additions may also occur during transport through nearshore (e.g., estuarine) to shelf waters [e.g., Fisher et al., 1998; Abril et al., 2002; Raymond and Hopkinson, 2003; Cai, 2011], resulting in net fluvial loads composing 10% to 60% of the total (i.e., net fluvial + nearshore) C exported to the coastal ocean margins [Kemp et al., 1997; Cai and Wang, 1998; Raymond and Hopkinson, 2003; Cai, 2011].

[4] Despite previous efforts to constrain components of the terrestrial-river-coastal C budget, there remains significant uncertainty in the amounts and characteristics of terrestrial C exported annually to rivers and the fluvial C loads ultimately reaching the coastal ocean. Also of growing interest are the relative contributions of aged (e.g., older than ∼60 years; note that following convention in radiocarbon dating we define 1950 A.D. to be the dividing year between aged and modern [Stuiver and Polach, 1977]) versus modern forms of C involved in riverine C cycling and export. While it has been well established that fossil IC (e.g., carbonates) can comprise a significant fraction of riverine DIC and particulate IC (PIC) pools [e.g., Meybeck, 1993; Telmer and Veizer, 1999; Raymond et al., 2004; Zeng et al., 2011], the importance of fossil OC, as well as aged soil OC, to riverine POC and DOC pools has only been relatively recently documented [Spiker and Rubin, 1975; Meybeck, 1993; Kao and Liu, 1996; Cole and Caraco, 2001; Leithold and Blair, 2001; Blair et al., 2003; Masiello and Druffel, 2001; Raymond and Bauer, 2001a, 2001b; Raymond et al., 2004; Longworth et al., 2007; Sickman et al., 2010]. Hence, rivers connect not only contemporary, short-term stores of C between terrestrial and coastal ocean ecosystems, but also relic, long-term C stores. This latter process is potentially significant given the enormous sizes of fossil OC and IC reservoirs on land (∼1500 × 104 Pg and > 6000 × 104 Pg of C, respectively) relative to terrestrial biomass (∼2000 Pg of C) and atmospheric CO2 (∼720 Pg of C) [Falkowski et al., 2000]. There is also growing evidence that, once mobilized, even highly aged terrestrial OM is not necessarily recalcitrant, but may be readily remineralized in rivers, estuaries, and coastal waters [Cole and Caraco, 2001; McCallister et al., 2004; Schillawski and Petsch, 2008; Caraco et al., 2010; Griffith and Raymond, 2011]. For these reasons, a better understanding of riverine C sources and characteristics is necessary for evaluating the role of aged OC and IC materials in terrestrial and ocean margin C cycles.

[5] Here, we present findings from a study designed to characterize the amounts, sources, and ages of POC, DOC, and DIC for eight rivers of the northeastern U.S. The rivers discharge to a common, well-studied ocean margin, the Middle Atlantic Bight (MAB) [e.g., Walsh et al., 1988; Biscaye et al., 1994; Verity et al., 2002] and represent ∼80% of the water discharging to the MAB annually. Estimates of total, allochthonous, and aged POC, DOC, and DIC exports (see also Hossler and Bauer [2012]) provide new constraints for terrestrial C losses and coastal ocean terrigenous imports for the MAB drainage and other ocean margin regions. Natural and anthropogenic controls on these processes are addressed in our companion paper [Hossler and Bauer, 2013].

2 Materials and Methods

2.1 Study Area

[6] The eight rivers included in this study were the Connecticut, Hudson, Delaware, Schuylkill (a tributary of the Delaware), Susquehanna, Potomac, Pamunkey and Roanoke (Figure 1). The rivers covered a range of hydrogeomorphologies, underlying lithologies, land uses, and other anthropogenic impacts (see Tables 1-3 in Hossler and Bauer [2013]). Across the study region, mean annual temperatures ranged from 6.4°C to 13.5°C and annual precipitation ranged from 979 mm to 1160 mm (Table 1 and section S3 in Hossler and Bauer [2013]).

Figure 1.

Map of the U.S. east coast indicating the sampled watersheds (sample locations are marked with yellow squares). The subregional watersheds (see section S3 in Hossler and Bauer [2013]) are outlined in light blue. The red outline indicates the approximate region draining to the MAB. The base map coloring depicts land relief and bathymetry with blue coloration indicating lower elevations, green coloration indicating intermediate elevations, and yellow coloration indicating higher elevations. The smaller inset map shows the region of study with respect to North America. (See also section S1 for details on how the map was generated.)

2.2 Sample Collection and Processing

[7] Details of sample collection and processing can be found in Hossler and Bauer [2012]. Briefly, each river was sampled at approximately 3–4 months intervals between 2005 and 2007 at the point farthest downstream that was accessible by small boat and above the reach of tidal influence. Surface water samples (∼0.1 m depth) were collected near the mid-point of each river and filtered through pre-baked Whatman quartz fiber filters (QFF; 0.8 μm). (Note that one caveat to our sampling strategy is that by collecting surface water samples (although common practice) as opposed to width- or depth-integrated samples we have likely underestimated the C pools, particularly the POC fraction (Curtis et al., 1979; Martin et al., 1992; Raymond et al., 2007). Another potential source of sampling bias is the small number of sampling events per river (i.e., 7); however, we did capture a representative range of hydrologic conditions (Figure S3 in Text S1 of the supporting information), including several high discharge events that are estimated to contribute to the majority of riverine C export [see e.g., Raymond and Saiers, 2010]. In a trade-off with finer scale sampling within a single river, this study focused on sampling a spectrum of rivers within a single region.)

[8] POC was considered to be the fraction collected directly on the QFF filters, and DOC was considered the fraction passing through the QFF filters. Samples for DIC were collected directly from the rivers using gas-tight syringes, then injected immediately into gas-tight glass serum bottles and fixed with saturated HgCl2 (0.2% v/v). (Note that the DIC fraction is a measure of the total DIC species: i.e., dissolved CO2, carbonic acid, bicarbonate, and carbonate. At the pH range (5.67 to 8.76) of the rivers in this study, the predominant form of DIC is bicarbonate.)

[9] Acidified POC and DOC samples from which inorganic carbonates had been removed were oxidized to CO2 (by sealed tube combustion and high-energy UV irradiation, respectively), then purified on a vacuum extraction line [Sofer, 1980; Bauer et al., 1992; Druffel et al., 1992; see also Hossler and Bauer, 2012]. Acidified DIC samples were sparged to extract the CO2 gas which was then purified on a vacuum extraction line [Sofer, 1980; Druffel et al., 1992; see also Hossler and Bauer, 2012]. Aliquots of CO2 were collected in sealed Pyrex tubes and analyzed for δ13C and Δ14C at the National Science Foundation-Arizona Accelerator Mass Spectrometry (AMS) Facility. Concentrations of POC and DIC were determined by quantification of the CO2 yield directly on the vacuum extraction line. DOC concentrations were analyzed independently by high temperature catalytic oxidation.

[10] Note that PIC (another form of riverine C) was not quantified in this study. However, based on available C data from U.S. Geological Survey (USGS) water monitoring stations for the Connecticut, Hudson, Delaware, Schuylkill, and Pamunkey rivers from 2001 through 2012 [U.S. Geological Survey, 2012], the PIC fraction appears to be only a minor constituent of the total C load (∼0.1%) for these northeastern U.S. rivers. Hence, by quantifying only the POC, DOC, and DIC pools, we still account for the vast majority of riverine C transported by the eight study rivers.

2.3 Data Analyses

[11] All calculations, estimations, and statistical analyses were performed in R 2.9.1 [R Development Core Team, 2009]. In addition to the following analytical methodologies (subsections 2.3.1–2.3.3), section S2 details the calculations for extrapolating to the entire MAB region which will be presented in section 3.4.

2.3.1 Weighted Mean Concentrations and Isotopic Signatures

[12] For each river, discharge-weighted mean concentrations were calculated from the two-year sample set as

display math(1)

where inline image is the discharge-weighted concentration for site s (i.e., Connecticut, Hudson, Delaware, Schuylkill, Susquehanna, Potomac, Pamunkey, or Roanoke) and carbon pool p (i.e., POC, DOC, or DIC); Cs,p(t) is the concentration for site s and carbon pool p measured on sampling date t; and Fs(t) is the mean daily discharge for site s on sampling date t. (Mean daily discharge data for each site and sampling date (i.e., Fs(t)) were obtained from the publicly available USGS water data [U.S. Geological Survey, 2011]; see also section S3 in Hossler and Bauer [2013].)

[13] Mean Δ14C signatures were calculated in a similar manner, but using flux-weighted data

display math(2)

where Īs,p is the flux-weighted isotopic signature (i.e., Δ14C) for site s and carbon pool p; Is,p(t) is the isotopic signature for site s and carbon pool p measured on sampling date t; Cs,p(t) is the concentration for site s and carbon pool p measured on sampling date t; and Fs(t) is the mean daily discharge for site s on sampling date t.

2.3.2 Total C Export Estimates

[14] Total (i.e., sum of allochthonous, autochthonous, aged, and modern) POC, DOC, and DIC exports were calculated for each watershed from the discharge-weighted mean concentrations (equation (1)) and mean annual discharge

display math(3)

where inline image is the mean annual export for site s and carbon pool p; inline image is the discharge-weighted concentration for site s and carbon pool p (equation (1)); and inline image is the mean annual discharge for site s. (Mean daily discharge data for the entire 2003–2010 period were also acquired from the publicly available USGS water data [U.S. Geological Survey, 2011] to determine the mean annual discharge for each site (i.e., inline image); see also section S3 in Hossler and Bauer [2013].) Estimation of confidence intervals for total C exports is described in section S3.

2.3.3 Allochthonous and Aged C Contribution Estimates

[15] Allochthonous C was considered to be any OC or IC material originating external to the river (i.e., terrestrially-derived), as opposed to autochthonous materials produced within the river (e.g., photosynthetically fixed CO2). Aged C was defined to be any OC or IC material older than ∼60 years (i.e., not “modern”; note that following convention in radiocarbon dating, we define 1950 A.D. to be the dividing year between aged and modern [Stuiver and Polach, 1977]).

[16] To determine the allochthonous and aged C contributions to POC, DOC, and DIC exports, we first estimated fractional contributions from six potential sources for POC and DOC (i.e., C3 plant material (C3 OC), C4 plant material (C4 OC), algal material (algal OC), slow-turnover soil OC (slow SOC; turnover time 25 years), passive-turnover soil OC (passive SOC; turnover time 5000 years), and fossil OC) and four potential sources for DIC (i.e., atmospheric CO2 exchange, carbonate dissolution, POC remineralization, and DOC remineralization) using a time-varying isotope mixing model (see Hossler and Bauer [2012] (e.g., Figures 1 and S2 and Tables S6–S8) for details). Allochthonous contributions to POC and DOC were then the combined proportions of C3 OC, C4 OC, slow SOC, passive SOC, and fossil OC to the respective POC and DOC pools. Allochthonous contributions to DIC were the combined proportions of IC deriving from carbonate dissolution, remineralization of allochthonous POC, and remineralization of allochthonous DOC. Note that for POC and DOC, we assumed that all C3 and C4 plant materials were terrestrially derived, although some portion of these contributing materials were likely from emergent aquatic vegetation and therefore autochthonous. For DIC, we assumed that the allochthonous proportions of remineralized POC and remineralized DOC were identical to the allochthonous proportions of the initial POC and DOC pools (i.e., no preferential remineralization of allochthonous versus autochthonous OC; note that this assumption should be a reasonable first approximation given the much greater proportion of allochthonous OC versus autochthonous OC despite the presumably greater lability of the latter [Kritzberg et al., 2005, 2006]).

[17] Aged C sources included passive SOC (Δ14C =  ∼− 540‰; see section S8 in Hossler and Bauer [2012]) and fossil OC (Δ14C = − 1000‰) for POC and DOC, and carbonate dissolution (note that this IC source was modeled to be an even mixture of carbonate rock (with Δ14C = − 1000‰) and the weathering agent carbonic acid (with an estimated aged proportion equivalent to that of the DOC pool)) and aged fractions of remineralized POC and DOC for DIC. Note that as for allochthonous DIC, we assumed that the aged proportions of POC and DOC were exactly represented in the contributions of remineralized POC and remineralized DOC to the DIC pool. For POC, DOC, and DIC, we also assumed that the “modern” sources were comprised entirely of modern C, when in fact because of the cycling of C, these modern sources would also contain some proportion of aged C. Estimation of confidence intervals for both allochthonous and aged C contributions is described in section S3.

3 Results and Discussion

3.1 River C Concentrations and Isotopic Character

[18] Of the total riverine C (i.e., sum of POC, DOC, and DIC; see also section 2.2) in the rivers of the present study, DIC comprised the majority of the total C, with discharge-weighted mean percentages ranging from 63% to 87% in seven of the eight rivers (Table 1). DOC comprised the second largest fraction (11% to 29%), and POC comprised the smallest fraction (2% to 9%) of the total C. For the eight rivers combined, the majority of the total riverine C was in the form of DIC (∼75%), followed by DOC (∼19%), then POC (∼6%; Table 1). Considering the OC separately, the dominant component was DOC which comprised ∼76% of the total OC in the eight rivers combined, and ranged from 61% in the Potomac River to 85% in the Delaware and Schuylkill rivers.

Table 1. Discharge-Weighted Concentrations and Percentages of POC, DOC, and DIC in the Eight Study Riversa
 Concentration (mg·L − 1)Percent of Total (%)
POCDOCDICPOCDOCDIC
  1. aThe percent contribution to total was determined by dividing the concentration of each of POC, DOC, and DIC by their sum concentration. The range of measurements over the seven sampling dates is also indicated for each mean (“range” row, with values in parentheses). The final two rows provide discharge-weighted means and ranges for all eight rivers combined.
 Connecticut River
Mean0.72.96.672864
Range(0.2, 1.2)(2.0, 3.9)(5.5, 8.8)(2, 14)(20, 35)(60, 78)
 Hudson River
Mean0.73.113.641878
range(0.5, 1.4)(2.4, 4.5)(11.8, 15.6)(3, 7)(13, 23)(70, 84)
 Delaware River
Mean0.52.210.831780
range(0.3, 1.2)(1.5, 3.0)(8.7, 15.0)(2, 8)(10, 25)(73, 87)
 Schuylkill River
Mean0.42.419.121187
range(0.2, 0.5)(1.6, 3.6)(18.1, 26.0)(1, 2)(8, 16)(82, 90)
 Susquehanna River
Mean0.72.112.851482
range(0.2, 1.5)(1.4, 2.8)(9.9, 19.6)(1, 9)(7, 20)(76, 92)
 Potomac River
Mean2.33.619.691477
range(0.4, 3.2)(1.6, 4.1)(13.3, 27.8)(1, 11)(7, 22)(70, 91)
 Pamunkey River
Mean2.77.04.5174836
range(0.2, 3.8)(2.7, 8.8)(3.9, 10.5)(2, 23)(31, 54)(24, 65)
 Roanoke River
Mean1.13.88.182963
range(0.3, 2.8)(2.5, 5.8)(6.3, 9.2)(2, 17)(20, 36)(47, 74)
 All Eight Rivers Combined
Mean0.93.012.561975
range(0.6, 1.4)(2.2, 3.9)(11.3, 15.7)(3, 7)(14, 25)(68, 82)

[19] The river OC concentrations of the present study (Table 1) were comparable to previous observations from the Hudson, Delaware, and Susquehanna rivers and from the geographically proximate Parker, York, and Neuse rivers of 0.2 mg·L − 1 to 9.1 mg·L − 1 POC (∼26% of the total OC) and 1.4 mg·L − 1 to 12.0 mg·L − 1 DOC (∼74% of the total OC) [Malcolm and Durum, 1976; Findlay et al., 1996; Fisher et al., 1998; Mannino and Harvey, 1999; Raymond and Bauer, 2000; Raymond and Hopkinson, 2003; Raymond et al., 2004; Longworth et al., 2007]. The river OC concentrations from the present study also coincided with reported estimates of 1.0 mg·L − 1 to 10.0 mg·L − 1 POC and 0.6 mg·L − 1 to 18.1 mg·L − 1 DOC for North America [Meybeck, 1982; Ludwig et al., 1996; Dai et al., 2012; Lauerwald et al., 2012] and of ∼ 4.7 mg·L − 1 POC and 1.0 mg·L − 1 to 20.0 mg·L − 1 DOC for temperate and temperate-wet bioclimates [Meybeck, 1982; Ludwig et al, 1996].

[20] The DIC concentrations observed in the present study (Table 1) were typically over twice the 2.7 mg·L − 1 to 7.7 mg·L − 1 DIC range reported from an earlier study that included the Hudson, Delaware, and Susquehanna rivers [Raymond et al., 2004]. However, the samples from the earlier study had been collected much farther upstream (40 km to 200 km) than the samples from the present study, which may account for the discrepancy. On the other hand, the DIC concentrations in this study compared well with observations from other temperate-wet rivers (e.g., the St. Lawrence River and major tributaries in Canada and the Ouseburn, Tern, and Tyne rivers in Great Britain), with reported DIC concentrations ranging from 6.3 mg·L − 1 to 64.0 mg·L − 1 [Hélie and Hillaire-Marcel, 2006; Baker et al., 2008]. (Note also, for the rivers in the present study, the DIC pool was typically composed of ∼20% free dissolved CO2 and carbonic acid and ∼80% bicarbonate.)

[21] The Δ14C signatures of the three major C pools varied considerably across the eight rivers. For POC, the flux-weighted mean Δ14C signatures ranged from − 163‰ to 89‰; Δ14C-DOC ranged from − 55‰ to 509‰; and Δ14C-DIC ranged from − 142‰ to 405‰ (Figure 2 and Table S1). The extreme variation in Δ14C was due almost exclusively to just two of the rivers: the Schuylkill and Pamunkey. These two rivers had exceptionally enriched Δ14C signatures in all three C fractions (Figure 2 and Table S1). The most likely explanation was some form of radiocarbon contamination from upstream nuclear power plants (see section S5 in Hossler and Bauer [2013]).

Figure 2.

Barplots of (a) Δ14C-POC, (b) Δ14C-DOC, and (c) Δ14C-DIC for the eight rivers of this study (lightly shaded bars); the flux-weighted means of all eight rivers combined and (*) all except the Schuylkill and Pamunkey rivers combined (darkly shaded bars); and literature values (open bars) for the north, central, and south regions of the MAB shelf and slope [source: Bauer et al., 2001, 2002]. For each C pool, the vertical bar boundaries indicate the range of observed values, while the horizontal black bar within each vertical bar indicates either the flux-weighted mean (this study) or the unweighted mean (MAB values). The solid and dashed horizontal lines indicate conventional radiocarbon years before 1950 A.D. (yr BP; solid line is 0 yr BP and dashed lines are 500 yr BP, 1500 yr BP, and 4500 yr BP). (Note that the exceptionally enriched Δ14C signatures in the Schuylkill and Pamunkey rivers were likely a result of radiocarbon contamination from upstream nuclear power plants; see section S5 in Hossler and Bauer [2013].)

[22] For the eight rivers combined, flux-weighted mean Δ14C signatures were − 92‰, 17‰, and − 86‰ for POC, DOC, and DIC, respectively (Figure 2 and Table S1). (Note that because the two radiocarbon contaminated rivers (i.e., Schuylkill and Pamunkey) were much smaller, the other six rivers, excluding the two outlying rivers, changed the combined flux-weighted mean Δ14C signatures only slightly: − 98‰, 7‰, and − 90‰ for POC, DOC, and DIC, respectively; see also Figure 2.) Both the POC and DIC fractions contained a substantial portion of aged C based on the equivalent mean 14C ages of ∼800 yr BP (conventional radiocarbon years before 1950 A.D.) and ∼700 yr BP, respectively (and ∼800 yr BP for both POC and DIC excluding the Schuylkill and Pamunkey rivers; Figures 2a and 2c). In contrast, the DOC fraction was primarily modern in character (Figure 2b). Across diverse river systems, Δ14C signatures have been shown to range from highly aged to fully modern depending on such factors as hydrogeomorphology, lithology, temperature, and land-use [e.g., Hedges et al., 1986b; Kao and Liu, 1996; Raymond and Bauer, 2001b; Raymond et al., 2004; Zeng et al., 2011]. Controls on inputs from aged C sources in this study are discussed in the companion paper [Hossler and Bauer, 2013].

[23] The mean age of POC from these eight rivers overlaps with POC ages observed for MAB coastal waters to which these rivers discharge (Figure 2a) [Bauer et al., 2001, 2002]. Shelf water circulation patterns suggest that much of MAB POC derives from water inflowing from the north (i.e., Gulf of Maine and Georges Bank; (Beardsley and Boicourt, 1981; Townsend et al., 2006)), with additional contributions from contemporary aged primary production, moderately-aged riverine POC (e.g., this study), and more highly aged sedimentary POC brought in by intrusions and upwellings from the eastern and southeastern MAB slope [Anderson et al., 1994; Bauer et al., 2001, 2002; DeMaster et al., 2002].

[24] In contrast, the mean riverine Δ14C-DOC signature of this study was much more enriched than the bulk DOC fractions reported for the MAB (e.g., mean riverine Δ14C-DOC of 17‰ versus mean MAB Δ14C-DOC of − 251‰; Figure 2b) [Bauer et al., 2001, 2002]. Similar to POC, however, previous studies suggest that the MAB DOC reflects a predominant contribution from DOC of inflowing northern waters, with further DOC inputs from degradation of MAB POC, primary production, riverine DOC (e.g., this study), and highly aged MAB slope water DOC [Peterson et al., 1994; Raymond and Bauer, 2000, 2001a; Bauer et al., 2002; Raymond and Hopkinson, 2003].

[25] The highly-aged riverine DIC of this study also contrasted with the modern-aged DIC reported for the MAB (Figure 2c) [Bauer et al., 2001, 2002]. As demonstrated in earlier studies, however, most of the MAB DIC derives from waters near atmospheric equilibrium (i.e., modern-aged) such as the Gulf of Maine and Georges Bank waters to the north (Δ14C ∼− 50‰ to 50‰; [Sherwood et al., 2008]) and MAB slope waters to the south and southwest (Δ14C ∼− 50‰ to 80‰; [Bauer et al., 2002]). Additional atmospheric exchange of CO2 can also be expected during the approximately 100 days residence time on the MAB shelf [Mountain, 1991].

3.2 Total, Allochthonous, and Aged C Exports

3.2.1 Total C Exports

[26] Mean annual total (i.e., allochthonous, autochtho-nous, aged, and modern) C export from the eight rivers combined was estimated at 100 Gg·yr − 1 as POC, 300 Gg·yr − 1 as DOC, and 1330 Gg·yr − 1 as DIC (Table 2). Approximately one third of the total river C exports to the MAB were attributable to the Susquehanna River alone, the largest of the study rivers. Areal yields (i.e., export normalized to drainage area; not shown) ranged from 0.2 Mg·km − 2·yr − 1 to 0.8 Mg·km − 2·yr − 1 of C for POC, and from 1.0 Mg·km − 2·yr − 1 to 2.0 Mg·km − 2·yr − 1 of C for DOC, which compared favorably to the broad range of estimates for similar regions and bioclimates (0.04 Mg·km − 2·yr − 1 to 4.7 Mg·km − 2·yr − 1 of C for POC and 0.3 Mg·km − 2·yr − 1 to 41.7 Mg·km − 2·yr − 1 of C for DOC) [Meybeck, 1993; Hope et al., 1994, 1997b; Ludwig et al., 1996; Worrall et al., 2007; Ogrinc et al., 2008; Lauerwald et al., 2012]. For DIC, areal yields were more variable, ranging from 1.3 Mg·km − 2·yr − 1 to 13.0 Mg·km − 2·yr − 1 of C, but were within the range of estimates reported by Moosdorf et al. [2011] for North America of 0.05 Mg·km − 2·yr − 1 to 25.8 Mg·km − 2·yr − 1 of C (note, the Moosdorf et al. [2011] estimates were actually for bicarbonate, which was typically ∼80% of the DIC pool in this study).

Table 2. Mean Annual Exports of POC, DOC, and DIC From the Eight Study Rivers to the Nearshore MABa
 Mean Annual Export (Gg·yr − 1)
POCDOCDIC
  1. aFor each river, estimates for mean annual exports of POC, DOC, and DIC based on the discharge-weighted mean concentrations (Table 1) and mean annual discharge (Table 1 in Hossler and Bauer [2013] are provided. The 95% CI (obtained by bootstrap with 10,000 resamplings) is also indicated for each estimate (“range” row, with values in parentheses).
 Connecticut River
Mean1354120
95% CI(8, 20)(36, 75)(89, 160)
 Hudson River
Mean1357250
95% CI(9, 20)(42, 79)(200, 300)
 Delaware River
Mean6.631150
95% CI(4.0, 11)(21, 46)(120, 210)
 Schuylkill River
Mean1.37.763
95% CI(0.9, 1.9)(5.1, 12)(49, 91)
 Susquehanna River
Mean2674460
95% CI(12, 46)(48, 120)(330, 680)
 Potomac River
Mean2642230
95% CI(8, 60)(22, 88)(140, 470)
 Pamunkey River
Mean2.36.23.9
95% CI(0.2, 5.7)(1.6, 14)(1.9, 10)
 Roanoke River
Mean8.12858
95% CI(2.4, 21)(12, 62)(29, 120)
 All Eight Rivers Combined
Mean1003001330
95% CI(60, 160)(210, 460)(1050, 1940)

3.2.2 Allochthonous C Contributions to River Export Fluxes

[27] Of particular interest in the present study were the contributions of allochthonous and aged materials (for definitions, see section 2.3.3) to the regional C exports from the study rivers. Contributions were estimated for six of the eight study rivers (i.e., the Connecticut, Hudson, Delaware, Susquehanna, Potomac, and Roanoke), but were unattainable for two of the rivers (i.e., Schuylkill and Pamunkey) because obvious radiocarbon contamination precluded application of the isotope mixing models used to differentiate C source contributions (see Hossler and Bauer [2012] and section S5 in Hossler and Bauer [2013]).

[28] Estimated allochthonous contributions to exported POC ranged from 11% to 67% (95% CI) in the Delaware River up to 61% to 100% (95% CI) in the Roanoke River (Figure 3a and Table S2). POC exports from the six rivers combined were derived 46% to 91% (95% CI) from allochthonous sources. The predominance of allochthonous POC in these rivers agreed with previous studies that have qualitatively identified terrestrial source materials mainly through isotope analysis, C:N ratios, and lignin biomarkers [Hedges et al., 1986a; Pocklington and Tan, 1987; Tan and Edmond, 1993; Hope et al., 1997a; Onstad et al., 2000; Kanduč et al., 2007; Wu et al., 2007; Ogrinc et al., 2008]. (A more in-depth discussion of allochthonous sources of POC (as well as of DOC and DIC) can be found in Hossler and Bauer [2012].)

Figure 3.

Percent contributions of (a–c) allochthonous C and (d–f) aged C to river POC, DOC, and DIC exports from the six rivers to the MAB. Mean estimates for each river are indicated by the smaller horizontal black bars, while the shaded vertical bars span the 95% CI. The dashed horizontal lines indicate the mean estimates across all six rivers, with the dotted horizontal lines indicating the approximate 95% CI. (See section 2.3.3 for estimation details. Note that estimates could only be obtained for six of the eight study rivers because radiocarbon contamination in the Schuylkill and Pamunkey rivers precluded application of the isotope mixing models used to differentiate C source contributions (see text for additional details; see also Hossler and Bauer [2012] and section S5 in Hossler and Bauer [2013]).)

[29] DOC tended to be composed of allochthonous materials to a much greater extent than POC in the six rivers (Figure 3b and Table S2). Allochthonous C represented over three quarters of exported DOC in four of the six rivers. The Susquehanna River was the primary exception, with estimated allochthonous contributions as low as 8% (95% CI) to exported DOC (Figure 3b and Table S2). This reduced allochthonous contribution may be attributable in part to both greater presence of finely textured soils and fewer wetlands within the Susquehanna watershed. (Factors controlling allochthonous contributions to DOC (as well as to POC and DIC) are addressed more thoroughly in the companion paper [see Hossler and Bauer, 2013].) For the six rivers combined, the estimated allochthonous contribution to DOC was 67% to 96% (95% CI), the majority of which derived from either recently-fixed plant material or young SOC [Hossler and Bauer, 2012]. Allochthonous materials have also been found to constitute the majority of DOC in various other streams and rivers through correlations between DOC flux and extent of peat cover, isotope analysis, C:N ratios, and UV adsorption [Hope et al., 1997a; Schiff et al., 1997; Palmer et al., 2001; Hélie and Hillaire-Marcel, 2006; Sanderman et al., 2009].

[30] Allochthonous contributions to the DIC pool of the six rivers were lower compared to POC and DOC (Figure 3c and Table S2). The notable exception was the Connecticut River, in which allochthonous C comprised 38% to 99% (95% CI) of the exported DIC. Factors leading to greater allochthonous DIC in the Connecticut River watershed likely relate to apportionment between surface and subsurface runoff, such as soil texture (e.g., Connecticut watershed soils tend to be more coarsely textured) and land use cover (e.g., the Connecticut watershed is more forested and less developed than the other five watersheds) [for additional details, see Hossler and Bauer, 2013]. For the other five rivers, estimated contributions tended to be in the range of 30% to 60%, which was similar to the estimated combined allochthonous contribution to exported DIC (95% CI: 37% to 52%) . Both carbonate dissolution and OC remineralization have been identified as major allochthonous sources of DIC in previous studies, with the balance between the two sources depending on factors such as lithology, hydrologic flow path, and OC bioavailability [e.g., Telmer and Veizer, 1999; Finlay, 2003; Raymond and Hopkinson, 2003; Raymond et al., 2004; Zhang et al., 2009; Zeng et al., 2011]. (Factors controlling allochthonous contributions to DIC (as well as to POC and DOC) are addressed more thoroughly in the companion paper [see Hossler and Bauer, 2013].) For the six rivers in the present study, allochthonous contributions to DIC were approximately one third carbonate dissolution (note that this proportion includes IC from both the carbonate rock and the weathering agent carbonic acid) and two thirds remineralization of allochthonous OC [Hossler and Bauer, 2012]. The remaining 40% to 70% of exported DIC (95% CI) derived from atmospheric CO2 exchange (i.e., was in isotopic equilibrium with atmospheric CO2), which has also been established as an important process affecting DIC composition in other riverine systems based on 13C analyses [Pawellek and Veizer, 1994; Hélie et al., 2002; Finlay, 2003; Brunet et al., 2005; Wachniew, 2006; Zhang et al., 2009].

3.2.3 Aged C Contributions to River Export Fluxes

[31] Aged C typically represented less than 30% of the different C fluxes for the six rivers in the present study (Figure 3d–3f and Table S2). For POC, aged C comprised 7% to 18% (95% CI) of the combined river exports. The aged POC fraction was lowest in the Potomac River (95% CI: 1% to 8%) and highest in the Susquehanna River (95% CI: 9% to 34%; Figure 3d). On average, 74% of the aged POC fraction was estimated to derive from fossil OC sources, with the remainder from passive-turnover SOC (turnover time 5000 years; [Hossler and Bauer, 2012]). The proportions of aged—and particularly fossil—POC were much lower for these east coast rivers than has been reported for small mountainous rivers, such as the Eel River in the northern California, USA, (∼50%; [Blair et al., 2003]) and the Lanyang Hsi River in Taiwan ( > 70%; [Kao and Liu, 1996]). This is consistent with the hypothesis of Blair et al. [2003] that aged contributions would be lower for rivers draining “passive” continental margins (e.g., U.S. east coast). An aged POC component has also been identified in numerous other rivers, with sources including soil erosion, rock weathering, and fossil-fuel combustion [Hedges et al., 1986b; Masiello and Druffel, 2001; Raymond and Bauer, 2001b; Raymond and Hopkinson, 2003; Raymond et al., 2004; Longworth et al., 2007].

[32] Of the three major river C pools (i.e., POC, DOC, and DIC), aged C sources contributed least to the exported DOC (Figure 3e and Table S2). Per river, the estimated aged DOC fraction ranged from 0% (the lower range for four of the six rivers) to as much as 23% in the Delaware River (95% CI). For the six rivers combined, the estimated aged C contribution was 3% to 11% (95% CI) of the exported DOC. Of this aged DOC export, only 18% was derived from fossil OC sources [Hossler and Bauer, 2012]. In other riverine studies measuring both DOC and POC Δ14C, the DOC component has been observed to be more modern in origin [Masiello and Druffel, 2001; Raymond and Bauer, 2001b; Raymond and Hopkinson, 2003; Longworth et al., 2007]. This age disparity suggests significant differences in the sources and processes controlling catchment POC and DOC contributions to rivers [Raymond and Bauer, 2001b; Raymond and Hopkinson, 2003]. In the present study, for example, while fossil OC contributed the majority of aged POC, passive SOC comprised the majority of aged DOC [see also Hossler and Bauer, 2012]. An aged DOC fraction originating primarily from soils has also been suggested in other studies [Schiff et al., 1997, Longworth et al., 2007, Sanderman et al., 2009, Sickman et al., 2010]. Possible controlling factors for these differences in source contributions are explored in Hossler and Bauer [2013].

[33] Aged C contributions to river DIC comprised 11% to 16% (95% CI) of the combined six river export to the MAB (Figure 3f and Table S2). Per river, the aged DIC fraction was lowest in the Roanoke River (95% CI: 0% to 9%) and highest in the Delaware River (95% CI: 14% to 24%). Of the aged DIC fraction, approximately 60% derived from fossil IC (e.g., carbonates), with the remaining 40% contributed by remineralization of aged POC and DOC fractions (24% originating from fossil OC and 16% originating from passive-turnover SOC) [Hossler and Bauer, 2012]. Moosdorf et al. [2011] estimated that ∼36% of riverine bicarbonate flux stemmed from carbonate weathering in North America. Among the rivers of the present study, 80% of the DIC flux was in the form of bicarbonate on average, of which an estimated 10% to 30% resulted from carbonate weathering [see also Hossler and Bauer, 2012]. This estimate is less than that of Moosdorf et al. [2011]; however, the northeastern U.S. region is not particularly rich in carbonate-bearing rocks compared to other North American regions such as the midwestern U.S. (e.g., see lithologic map presented in Jansen et al., [2010]).

3.3 Seasonal Patterns in Total, Allochthonous, and Aged C Exports

[34] Riverine C flux dynamics in this study were dominated by discharge and tended to be highest in the late fall and winter months during peak flows (Figures 4a–4c and S2), although there was considerable site-to-site variability (Figures S1 and S3). The dominating discharge effect also accounted for the synchronous temporal behavior of the various C pools (i.e., POC, DOC, and DIC, and total, allochthonous, and aged). The control of discharge on C flux dynamics has been well established in the literature [e.g., Meybeck, 1982; Bluth and Kump, 1994; Ludwig et al., 1996; Moosdorf et al., 2011; Alvarez-Cobelas et al., 2012; Lauerwald et al., 2012] and is primarily due both to its direct effect on flux (i.e., C flux is the product of discharge and C concentration, hence as discharge increases, C flux should also increase) and its larger coefficient of variation over that of C concentration [see e.g., Benson, 1965; Kenney, 1982].

Figure 4.

Time series plots showing (a–c) total C exports (kg · s − 1) and percent contributions to total C exports from (d–f) allochthonous C and (g–i) aged C for the rivers of the present study. Exports (i.e., Figures 4a–4c) are plotted against the leftmost y axes and scales vary by parameter; percent contributions (i.e., Figures 4d–4i) are plotted against the rightmost y axes. Regions of 95% CI are indicated by shading and estimates are indicated by thick black lines. Vertical shaded bars indicate winter months (i.e., December (D) through March (M)); spring, summer, and fall months are unshaded (e.g., June (J) and September (S)). (Note: in Figures 4a–4c, data have been compiled from all eight study rivers, whereas in Figures 4d–4i, data have been compiled from only six of the eight study rivers; allochthonous and aged C estimates could not be obtained for the Schuylkill and Pamunkey rivers because of model violations (see text for additional details; see also [Hossler and Bauer, 2012]).)

[35] The discharge effect on flux can also be mediated through differences in C concentration. For POC, the discharge-concentration relationship is typically direct [Pocklington and Tan, 1987; Correll et al., 2001]. For DOC, the observed relationship with discharge has been found to be direct [Correll et al., 2001; Wilson and Xenopoulos, 2008; Sanderman et al., 2009; Raymond and Saiers, 2010], inverse [Wilson and Xenopoulos, 2008], or not significant [Pocklington and Tan, 1987; Raymond and Hopkinson, 2003; Wilson and Xenopoulos, 2008]. In the present study, discharge correlated positively with POC concentration for three of the rivers (Connecticut: r = 0.75, p = 0.05; Potomac: r = 0.89, p = 0.01; and Pamunkey: r = 0.96, p = 0.02; not shown), and with DOC concentration for only one river (Pamunkey: r = 0.83, p = 0.04; not shown). In contrast, the discharge-concentration relationship has generally been found to be inverse for DIC [Bluth and Kump, 1994; Finlay, 2003; Brunet et al., 2005; Baker et al., 2008; Barnes and Raymond, 2009; Zeng and Masiello, 2010]. However, only two of the rivers in the present study demonstrated negative correlations between discharge and DIC concentration (Connecticut: r = − 0.93, p = 0.003; and Delaware: r = − 0.87, p = 0.01; not shown).

[36] Across all C pools, allochthonous inputs to the six rivers assessed in this study tended to be higher in the winter (Figures 4d–4f and S4). Various riverine studies have also observed allochthonous contributions to be higher in the winter, while autochthonous contributions peak in the summer when conditions are more favorable for primary production [Pocklington and Tan, 1987; Atekwana and Krishnamurthy, 1998; Hellings et al., 1999; Kendall et al., 2001; Hélie et al., 2002; Hélie and Hillaire-Marcel, 2006; Wachniew, 2006].

[37] Aged C inputs were more variable than total (i.e., aged plus modern) allochthonous C inputs, although they tended to increase in late winter/early spring (Figures 4g–4i and S5). For POC, this pattern may be attributed to increased soil erosion and rock weathering during high flow [Kao and Liu, 1996; Blair et al., 2003]. For DOC and DIC, however, an opposite pattern might be expected: i.e., more aged contributions during low flow because of greater contributions from groundwater (as opposed to surface water) which would be in contact with deeper, older soils in the case of DOC [Schiff et al., 1997, 1998; Duan et al., 2007; Sanderman et al., 2009] and with carbonate-rich bedrock in the case of DIC [Atekwana and Krishnamurthy, 1998; Finlay, 2003; Kanduč et al., 2007] (see also Table 2 in Hossler and Bauer [2013]). Instead, the positive relationship between discharge and aged DOC and DIC observed in the present study (Figures 4h and 4i) suggests that similar mechanisms controlling aged POC may also operate for aged DOC and DIC in certain watersheds.

[38] One aspect of the observed seasonality in riverine OC fluxes is that the timing may be offset from peak primary productivity in coastal ocean margins, such that while on an annual basis fluvial OC inputs provide only minor subsidy to coastal ocean heterotrophy (section 3.4.3), on a seasonal basis fluvial input becomes significant. For the MAB region, NPP on the shelf tends to peak in spring [Rowe et al., 1986], whereas fluvial OC inputs (particularly from allochthonous sources) tend to peak in the late fall and winter (this study). As a result, fall and winter fluvial OC inputs may subsidize the net heterotrophy observed on the shelf also during these times of year [Rowe et al., 1986]. A potential consequence of this scenario is that much of the fluvial OC reaching the shelf becomes remineralized because it is delivered primarily in the off-seasons (i.e., fall or winter), even though it is considered less labile compared to autochthonous OC.

3.4 Extrapolations to the Middle Atlantic Bight Region

[39] Using mean discharge data compiled for coastal systems in the contiguous United States [Coastal Assessment and Data Synthesis System, 1999], we estimated that the eight rivers included in this study comprised approximately 80% of the freshwater discharged directly to the MAB annually. Under the assumption that these rivers are representative of riverine discharge and C export to the MAB in general, the combined eight river exports were scaled-up to the freshwater discharge for the entire region (section S2) yielding estimated total C exports to the nearshore MAB of 80 Gg·yr − 1 to 200 Gg·yr − 1 of C as POC (95% CI); 260 Gg·yr − 1 to 570 Gg·yr − 1 of C as DOC (95% CI); and 1300 Gg·yr − 1 to 2400 Gg·yr − 1 of C as DIC (95% CI; Table 3). The upper DOC export estimate compares favorably to the 566 Gg·yr − 1 of C as DOC value calculated in a previous study by Raymond and Bauer [2000].

Table 3. Exports, Losses, and Inputs of POC, DOC, and DIC in the MAB Region.a
 POCDOCDIC
  1. aEstimates from this study of mean annual C exports, proportions of allochthonous C contributions, and proportions of aged C contributions (Tables 2 and S2) were extrapolated to the entire MAB drainage region (section S2). The extrapolations include total, allochthonous, and aged C exports to the nearshore MAB coastal margin, estimated terrestrial losses from the MAB drainage region, and inputs to the MAB shelf (see also Figures S6 and S7). For the estimated terrestrial losses, the values reported refer to the form of C (i.e., POC, DOC or DIC) that is exported to the river. For DIC, a portion of the exported material will have originated from mineralization of terrestrial OC (section 3.4.1). (Note, for conversion to areal yields, the MAB drainage region is approximately 3.7 × 105 km2 in area based on data available through the National Oceanic and Atmospheric Administration's Coastal Assessment and Data Synthesis System (http://coastalgeospatial.noaa.gov)).
 Total C Exports to the Nearshore MAB (Gg·yr − 1)
Mean1203701650
95% CI(80, 200)(260, 570)(1300, 2400)
 Allochthonous C Exports to the Nearshore MAB (Gg·yr − 1)
Mean90290780
95% CI(40, 150)(200, 490)(540, 1120)
 Aged C Exports to the Nearshore MAB (Gg·yr − 1)
Mean1020220
95% CI(10, 20)(10, 40)(150, 340)
 Estimated Terrestrial Losses (Gg·yr − 1)
Mean1203803960
95% CI(60, 210)(260, 650)(2750, 5690)
 Net Fluvial Total C Inputs to the MAB Shelf (Gg·yr − 1)
Mean70280800
95% CI(10, 200)(130, 570)(150, 2400)
 Net Fluvial Allochthonous C Inputs to the MAB Shelf (Gg·yr − 1)
Mean50220390
95% CI(0, 150)(100, 490)(70, 1120)
 Net Fluvial Aged C Inputs to the MAB Shelf (Gg·yr − 1)
Mean1020100
95% CI(0, 20)(0, 40)(20, 340)

[40] The amounts of allochthonous and aged riverine C exported to the nearshore MAB were determined in a manner similar to that for total POC, DOC, and DIC exports (Table 3). These calculations were based additionally on the allochthonous and aged C contributions that had been estimated for six of the eight study rivers (Table S2) and scaled up by a factor 1.29 (i.e., the six study rivers for which allochthonous and aged C contributions could be determined comprised approximately 78% of the freshwater discharged directly to the MAB annually) rather than by 1.24 as was the case for the total C exports which were based on estimates from all eight study rivers.

3.4.1 Terrestrial C Losses via Rivers in the Middle Atlantic Bight Region

[41] The allochthonous riverine POC, DOC, and DIC exports to the nearshore MAB (Table 3) represent conservative estimates of terrestrial C losses because some fraction of the terrestrially-derived materials would have been lost (e.g., microbial and photochemical remineralization of POC and DOC; degassing and photosynthetic uptake of DIC) during downstream transport (i.e., upstream terrestrial loss of C is greater than the C export we measure near the mouth of the river; see section S2 and Figure S6). Lauerwald et al. [2012], for example, estimated that ∼23% of DOC was lost through in-stream processing based on geospatial models of North American rivers. Similar estimates of 20% to 30% in-stream allochthonous OC respiration have been obtained specifically for the Hudson River [Cole and Caraco, 2001; Maranger et al., 2004].

[42] For the region draining to the MAB (Figure 1), we estimated annual terrestrial C losses based on our estimates of allochthonous C exports scaled up for the entire MAB region and compensating for in-stream losses during transport (section S2). Based on studies by Cole and Caraco [2001], Maranger et al. [2004], and Lauerwald et al. [2012], we assumed a 25% loss rate for terrestrially derived POC and DOC. For DIC, we used a 7% loss-rate from photosynthetic consumption based on flux-weighted isotope mixing model estimates [Hossler and Bauer, 2012], and a 74% loss-rate from degassing based on data presented in Butman and Raymond [2011] (see also section S2). Additionally for DIC, because the allochthonous export estimate included both external (i.e., on land) and internal (i.e., in stream) contributions from POC and DOC remineralization, the estimated amounts of POC and DOC remineralized in-stream (i.e., 25% of the total terrestrial POC and DOC inputs) were subtracted from the DIC estimate. With these assumptions, terrestrial losses from the MAB drainage region were estimated to be 320 Gg·yr − 1 to 860 Gg·yr − 1 in the form of OC (i.e., POC+DOC; 95% CI); and 2750 Gg·yr − 1 to 5690 Gg·yr − 1 in the form of DIC (95% CI; Table 3).

[43] Of the terrestrial DIC losses, approximately 83% derived from terrestrial OC remineralization (and the remaining 17% from carbonate dissolution, based on source contribution estimates from Hossler and Bauer [2012]), bringing the annual OC terrestrial loss (independent of export form) to 3800 Gg·yr − 1 on average. Based on annual terrestrial NPP data available from Imhoff et al. [2004b], the regional catchments draining to the MAB produce 250,000 Gg of net OC annually. The estimated annual terrestrial OC loss would then represent less than 2% of the annual NPP in the MAB region.

[44] These estimates of terrestrial NPP exported by rivers compare favorably to previous studies. Hope et al. [1994], in their review of 117 temperate and boreal catchment studies, estimated that 1% of NPP was exported as OC, while Moeller et al. [1979], in their study of four regions across the U.S., estimated 0.1% to 0.4% of NPP was exported as DOC alone. In their review of coastal ocean OC dynamics, Smith and Hollibaugh [1993] estimated that 0.7% of terrestrial NPP was exported as riverine OC on a global basis. Other global estimates of terrestrial OC losses to inland waters range from 1.9 Pg·yr − 1 to 2.9 Pg·yr − 1 [Cole, 2007; Battin et al., 2009; Tranvik, 2009], which would represent 3% to 6% of the estimated global terrestrial NPP of 50 Pg·yr − 1 to 60 Pg·yr − 1 of C [Randerson et al., 2002; Imhoff et al., 2004a; Ito, 2011].

[45] However, what must also be considered is the proportion of exported OC derived from modern inputs (see also section S2). In the MAB region, for example, modern aged sources comprise just 64% of the terrestrial OC loss, with the remaining 36% deriving from OC sources cycling over time periods ranging from decades (e.g., slow-turnover SOC) to millennia (e.g., passive-turnover SOC) to over tens of thousands of years (e.g., fossil OC) [Hossler and Bauer, 2012]. Adjusting for the proportion of annual terrestrial OC loss that is modern in origin thus reduces the terrestrial OC export estimate for annual NPP to around 1%, or 2400 Gg·yr − 1 of C on average, for the MAB region.

[46] Similar adjustments for modern versus aged OC contributions will likely need to be made for other regions and globally when relating fluvial C exports to terrestrial losses of NPP. Across numerous riverine systems, it has become apparent that a significant fraction of the OC load is comprised of aged materials such as fossil OC and aged SOC [Spiker and Rubin, 1975; Meybeck, 1993; Kao and Liu, 1996; Cole and Caraco, 2001; Leithold and Blair, 2001; Masiello and Druffel, 2001; Raymond and Bauer, 2001a,2001b; Blair et al., 2003; Raymond et al., 2004; Longworth et al., 2007; Sickman et al., 2010]. The relative proportion of aged OC will vary from system to system depending on such factors as lithology, hydrogeomorphology, and anthropogenic impact [Hossler and Bauer, 2013].

3.4.2 Net Fluvial C Inputs to the Middle Atlantic Bight Shelf

[47] The riverine OC and IC exported to the nearshore MAB (Table 3) is expected to be substantially modified during transport through nearshore (including estuarine) waters before reaching the MAB shelf (Figure S7). While nearshore transit will both remove fluvial C from and add nearshore C to the riverine OC and IC export [e.g., Kemp et al., 1997; Raymond and Bauer, 2000; Abril et al., 2002; Cai, 2011], we focus only on the removal processes in order to quantify the actual fluvial C export reaching the MAB shelf.

[48] Through processes such as remineralization (microbial and photochemical), sedimentation, and flocculation [Fontugne and Jouanneau, 1987; Lucotte et al., 1991; Keil et al., 1997; Fisher et al., 1998; Moran et al., 1999, 2000; Raymond and Bauer, 2000, 2001a; Abril et al., 2002; Bauer and Bianchi, 2011; Bianchi and Bauer, 2011; Dai et al., 2012], up to 90% of riverine POC and 50% of riverine DOC may be lost in nearshore transit. Actual OC loss across the nearshore will depend on such factors as residence time [Raymond and Bauer, 2000; Abril et al., 2002], temperature [Raymond and Bauer, 2000], OC concentration [Abril et al., 2002], and chemical composition [Sun et al., 1997; Hopkinson et al., 1998]. Residence times in the nearshore MAB, for example, can range from a few days in the smaller estuaries to half a year in the larger embayments [Pilson, 1985; Nixon, 1996; Sin et al., 1999; Hagy et al., 2000; Howarth et al., 2000; Abdelrhman, 2005; Gay and O'Donnell, 2009].

[49] For DIC, previous river studies suggest up to 90% loss through degassing during nearshore transit [Cai and Wang, 1998; Abril et al., 2000; Hunt et al., 2011]. Net loss of riverine DIC across the nearshore will depend on the amount of riverine OC remineralization and p CO2 water-to-air differential, in addition to residence time [Cai and Wang, 1998; Abril et al., 2000].

[50] For simplicity, we assumed that during nearshore transit, the major fate for POC was a balance of sedimentation and remineralization to DIC; the major fate for DOC was remineralization to DIC; and the major fate for DIC was degassing (Figure S7; see also section S2). Using the full range of loss estimates from the literature (i.e., 0 to 0.9 for POC; 0 to 0.5 for DOC; and 0 to 0.9 for DIC), the expected net fluvial inputs to the MAB shelf were 10 Gg·yr − 1 to 200 Gg·yr − 1 of C as POC; 130 Gg·yr − 1 to 570 Gg·yr − 1 of C as DOC (95% CI); and 150 Gg·yr − 1 to 2400 Gg·yr − 1 of C as DIC (Table 3).

[51] Vlahos et al. [2002], in their DOC budget for the MAB, estimated that the total nearshore DOC input (i.e., riverine plus estuarine) ranged from 400 Gg·yr − 1 to 770 Gg·yr − 1 of C, with a mean input of 590 Gg·yr − 1 of C. Assuming a 40% to 60% fluvial contribution to total nearshore input [Raymond and Hopkinson, 2003], the total nearshore DOC input to the MAB for this study would be 220 Gg·yr − 1 to 1420 Gg·yr − 1 of C, with a mean input of 560 Gg·yr − 1 of C, which is consistent with the estimate of Vlahos et al. [2002].

[52] Also of note is the apportionment of total C exports between allochthonous versus autochthonous C, and between aged versus modern C. Using the same approach and assumptions as for total (i.e., allochthonous, autochthonous, aged, and modern) fluvial OC and DIC MAB inputs, the estimated net fluvial DIC input apportionments were 70 Gg·yr − 1 to 1120 Gg·yr − 1 as allochthonous DIC and 20 Gg·yr − 1 to 340 Gg·yr − 1 as aged DIC (Table 3). Net fluvial OC inputs to the MAB shelf ranged from 100 Gg·yr − 1 to 640 Gg·yr − 1 as allochthonous OC and 10 Gg·yr − 1 to 60 Gg·yr − 1 as aged OC (Table 3). However, given the greater presumed recalcitrance of terrestrial OM in comparison to autochthonous OM [e.g., Enríquez et al., 1993; Kemp et al., 1997; Sun et al., 1997; McCallister et al., 2004, 2006; Khodse and Bhosle, 2011], a disproportionate amount of the terrigenous OC (including aged OC) may escape remineralization across the nearshore. Such preservation may be particularly true for aged OC based on recent studies indicating preferential degradation of younger and presumably more labile OC [e.g., Raymond and Bauer, 2001b; Raymond and Hopkinson, 2003; but see also McCallister et al., 2004].

3.4.3 Net Fluvial OC Input Relative to NPP in the Middle Atlantic BightShelf

[53] For perspective, we can compare the riverine OC contributions to the MAB to estimates of annual NPP in the MAB. Across the MAB shelf, estimates of annual areal NPP range from 100 Mg·km − 2·yr − 1 to 600 Mg·km − 2·yr − 1 of C with a mean of 350 Mg·km − 2·yr − 1 of C [Malone et al., 1983]; Liu et al., 2000; Lohrenz et al., 2002; Mouw and Yoder, 2005; Fennel et al., 2006; Druon et al., 2010; Filippino et al., 2011]. With a total shelf area of approximately 1.5 × 105 km 2, the annual total NPP on the MAB becomes 15,000 Gg·yr − 1 to 93,000 Gg·yr − 1 of C. Based on this estimate, riverine OC contributions to the MAB shelf (Table 3) range from <1% to 5% of the annual NPP. However, we emphasize again that the timing offset of riverine C inputs relative to peak shelf productivity could augment the importance of fluvial OC contributions to the MAB shelf at different times of the year (although rough estimates suggest that the fluvial OC contribution relative to MAB shelf NPP is at most 25% even during the winter when the fluvial OC contribution is near maximum (this study) and the shelf NPP is near minimum [Malone et al., 1983; Rowe et al., 1986; Mouw and Yoder, 2005; Druon et al., 2010]; see also section 3.3). Indeed, seasonal significance of fluvial OC inputs to heterotrophic activity has been demonstrated for nearshore systems [see e.g., Hopkinson, 1985; Kemp et al., 1997; Breed et al., 2004] and may extend to coastal shelf systems as well [e.g., Serret et al., 1999].

4 Summary

[54] For the eight rivers of this study, concentrations and areal yields of POC, DOC, and DIC compared favorably with previous observations from similar regions and bioclimates. Allochthonous-derived C (as opposed to autochthonous C) comprised the majority of river exported POC and DOC and half of the total POC, DOC, and DIC flux. A smaller but still significant fraction (<25%) of river exported C was further composed of aged C materials deriving from fossil C (a particularly important source for aged POC and DIC) and aged SOC (a particularly important source for aged DOC).

[55] For the MAB region, it was estimated that, on average, 3800 Gg (or 10 Mg·km − 2) of terrestrial OC was removed annually by fluvial transport. Of this terrestrial OC loss, an estimated ∼64% was contemporary C, or approximately 1% of the annual terrestrial NPP. The estimate of fluvial removal of terrestrial NPP was lower than previous global estimates. However, earlier studies had not taken into consideration the apportionment between contemporary (i.e., NPP) and aged OC sources.

[56] Of the riverine C exported to the nearshore MAB, an estimated 0 to 80% was removed during transit through nearshore waters (based on the literature), leaving 140 Gg·yr − 1 to 770 Gg·yr − 1 of OC (∼76% as DOC) and 150 Gg·yr − 1 to 2400 Gg·yr − 1 of DIC to be imported to the MAB shelf. This fluvial OC input is relatively small (<1% to 5%) with respect to annual MAB NPP; however, the timing of riverine C fluxes, which peaked in the late fall and winter, appeared to be offset from peak NPP (at least on the MAB shelf). Hence, fluvial OC inputs to the coastal (and perhaps open) ocean may be important on a seasonal basis, although seemingly minor on an annual scale.

Acknowledgments

[57] This research was supported by the National Science Foundation's Integrated Carbon Cycle Research Program Grant No. EAR-0403949 and Chemical Oceanography Program Grant Nos. OCE-0327423 and OCE-0961860. For assistance with sample collection and processing, we thank Ed Keesee, David Perkey, and Andrew Wozniak. We also wish to thank Dr. James Sickman and an anonymous reviewer for their insightful and comprehensive comments and suggestions for improvement of this manuscript.