Recent air-to-sea CO2 flux measurements at several major continental shelves suggest that shelves may act as a one-way pump and absorb atmospheric CO2 into the ocean. The U.S. South Atlantic Bight (SAB) contrasts these findings in that it acts as a source of CO2 to the atmosphere while simultaneously exporting dissolved inorganic carbon (DIC) to the open ocean. The shelf-wide heterotrophy and carbon exports in the SAB are subsidized by the export of organic carbon from the abundant intertidal marshes, which are a sink for atmospheric CO2. It is proposed here that the SAB represents a marsh-dominated heterotrophic ocean margin as opposed to river-dominated autotrophic margins. Based on this and other studies, DIC export flux from margins to the open ocean must be significant in the overall global ocean carbon budget.
 The coastal ocean plays a disproportionately important role in the ocean's biogeochemical cycle despite its small areal fraction [Gattuso et al., 1998]. However, global ocean and atmosphere carbon cycle models do not consider the coastal ocean in studies of CO2 exchange between the atmosphere and the ocean [Sarmiento and Gruber, 2002]. Recent measurements of net air-to-sea CO2 fluxes at several major continental shelves suggest that shelves may act as a one-way pump that absorbs atmospheric CO2 and exports it to the open ocean in forms of organic and inorganic carbon [Tsunogai et al., 1999; Wang et al., 2000; Frankignoulle and Borges, 2001; Boehme et al., 1998; DeGrandpre et al., 2002]. Tsunogai et al.  proposed a continental shelf pump concept for such a collective activity. These observations also favor the argument that continental shelves are autotrophic (i.e., net production of organic carbon). Such a global extrapolation must be validated since the suggested air-sea flux is significant in the oceanic CO2 budget. However, the issue whether continental shelves are a sink or a source of CO2 to the atmosphere, or, whether continental shelves are net autotrophic or net heterotrophic (i.e., net production of inorganic carbon), is far from certain [Smith and Mackenzie, 1987; Smith and Hollibaugh, 1993; Duart and Agusti, 1998].
 We evaluate the continental shelf pump hypothesis and coastal ocean trophic status by collecting inorganic carbon data along a cross-shelf transect off Wassaw Sound, Georgia to the shelf break at the central U.S. South Atlantic Bight (SAB) (Figure 1). Major features of the shoreline include extensive shoals and a series of barrier islands, behind which are lagoons with extensive salt marshes that may export materials to fuel biological activities in the shelf. The SAB is also bounded by the Gulf Stream along its shelf break which is only 70 m in depth. Extensive biological and physical oceanographic studies have been conducted in the SAB [Menzel, 1993; Pomeroy et al., 2000]. The shelf has no annual spring bloom and the episodic intrusion of the nutrient-rich Gulf Stream water is the driving force for biological production on the shelf.
 The survey cruises were conducted on Dec. 6–7, 2000, Apr. 29, 2002, Jun. 17, 2002, Oct. 2–3, 2002, and Dec. 12, 2002. Sea surface pCO2 was measured underway (i.e., pumping seawater at real-time) by a “combined laminate-flow and shower-head equilibrator plus an infrared analyzer (Li-Cor 6252)” system with a precision of 1 μatm and was calibrated against a standard gas traceable to NIST standard. All pCO2 data were corrected to 100% saturation of water vapor at 1 atmospheric pressure and in-situ temperature. The CO2 fluxes across the air-sea interface were calculated based on Wanninkhof relationship and daily wind speeds from an offshore tower [DeGrandpre et al., 2002]. Discrete sea surface water samples were measured for total dissolved inorganic carbon (DIC) by acid release of CO2 and subsequent quantification with a non-dispersive infrared CO2 analyzer with a precision of 0.1% [modified after Cai and Wang, 1998].
3. Results and Discussion
3.1. Surface Water pCO2 and Sea-to-Air CO2 Flux
 Surface water pCO2 has strong seasonal and cross-shelf variability (Figure 1). In December (2000 and 2002), with the exception of a small portion of nearshore water where over-saturation occurred, under-saturation dominated for the rest of the system with an area-weighted mean sea-to-air gradient of −70 μatm. In April (2002), sea surface pCO2 of the shelf became over-saturated with respect to the atmosphere with an area-weighted mean sea-to-air gradient of 140 μatm. This gradient climbed to 180 μatm in June, and declined to 80 μatm in October. The maxima of sea surface pCO2 always occurred near the shore; it declined quickly within the inner shelf. Spatial variation in the mid- and outer shelves was also noticeable.
 The seasonal pattern in the SAB, i.e., lower pCO2 during wintertime and higher pCO2 during warm months, is opposite to observations in the East China Sea (ECS) [Tsunogai et al., 1999] and the European Atlantic Shelves (EuAS) [Frankignoulle and Borges, 2001]. While this pattern is similar to that found in the U.S. Mid-Atlantic Bight (MAB) [DeGrandpre et al., 2002], it has higher absolute values. Based on a linear interpolation of sea surface pCO2 between cruises and daily wind speed from a nearby offshore tower, the sea-to-air CO2 fluxes at each section of the SAB shelf were calculated. The shelf-wide sea-to-air flux is summed as 2.7 MtC yr−1 (Million ton or 1012 gC yr−1) or 30 gC m−2yr−1. Annually, the SAB is a net atmospheric CO2 source, which is just the opposite of all the other shelves mentioned above. Therefore, continental shelf pump hypothesis cannot be generalized globally.
3.2. DIC Flux to the Open Ocean
 Spatial distributions of total dissolved inorganic carbon (DIC) also showed seasonality in the SAB (Figure 2). Relative to a conservative mixing line between the oceanic end-member and the river end-member, it is clear that DIC was added into the shelf year-round, particularly in the nearshore zone. More DIC was generated in April and June than in October and December.
 Given the different air-sea CO2 flux patterns in various shelves, one cannot quantify carbon transport to the open ocean from the coastal ocean with a shelf-wide air-sea CO2 flux alone. The DIC export must be determined. To estimate a shelf-wide rate of DIC export to the open ocean, excess normalized DIC (nDICex) relative to the offshore end-members is calculated. Weighted to the volumes of the shelf zones between neighboring stations, the total nDICex inventory on the SAB was estimated for each season. An annual average inventory divided by an average water residence time of 3 months [Menzel, 1993] allows an estimate of the shelf-wide DIC export rate of 2.6 MtC yr−1 [Wang, 2003].
 While the areal offshore DIC flux in the SAB (30 gC m−2yr−1) is only half of that estimated in the ECS (70 gC m−2yr−1, Tsunogai et al., 1997), this observation nevertheless supports the argument that shelf DIC export could be significant on a global scale. Using a 60% export efficiency as suggested by an ocean general circulation model [Yool and Fasham, 2001] and a global shelf area = 26 × 106 km2, the global continental shelf DIC export flux ranges from 0.6 (SAB) to 1.5 (ECS) GtC yr−1(i.e., 1015 gC yr−1) if the SAB and the ECS represent the two extremes. As a comparison, the world riverine DIC and DOC fluxes to the ocean are 0.4 and 0.2 GtC yr−1 respectively [Smith and Mackenzie, 1987], and the current estimate of net oceanic uptake of anthropogenic CO2 is 2 GtC yr−1 [Sarmiento and Gruber, 2002]. While large uncertainty may be involved in the global extrapolation, this result serves to show that the offshore DIC flux is neither insignificant nor unreasonably large. The issue warrants further studies.
3.3. Net Ecosystem Metabolism
 While the CO2 flux to the atmosphere and the DIC export to the open ocean are important in their own rights, the trophic status of the SAB must be evaluated with a mass balance analysis. For a given ecosystem (Figure 3), the net ecosystem metabolism is defined as R (total system respiration) - P (gross primary production). R-P should be balanced by the net input of organic carbon or net export of inorganic carbon, i.e, R-P = DICex + CO2a-s − DICin = OCin − OCex, where DICex and DICin are inorganic carbon export and input respectively; CO2a-s is air-sea CO2 exchange; OCin and OCex are organic carbon input and export respectively.
 DIC input from the intertidal marshes is estimated based on 20-month of monitoring DIC at six sites along the Duplin River, a blind tidal creek on Sapelo Island, and a longitudinal diffusive equation derived from the same area [Chalmers et al., 1985] as ∼0.7 MtC yr−1 [Wang, 2003]. Adding a riverine flux of 0.6 Mt yr−1 [Cai and Wang, 1998], the total DIC flux from the land side is about 1.3 MtC yr−1. Thus, the net ecosystem metabolism (= DICex + CO2a-s − DICin) of the SAB is 4.0 MtC yr−1. This shelf-wide net heterotrophy must be subsidized by organic carbon export from the abundant intertidal marshes [Hopkinson, 1988], and this argument is supported by a carbon mass balance analysis within the salt marsh (Figure 3).
 While export of organic carbon and inorganic carbon from salt marshes have been recognized based on estuarine studies, the degree and mode of export are highly debatable [Childers et al., 2000]. Our offshore-based research suggests that organic carbon export must have occurred to an extent exceeding what was observed to support the shelf-wide heterotrophy (Figure 3).
4. Marsh- vs. River-Dominated Margins
 The SAB is differentiated from other continental shelves that act as sinks of atmospheric CO2 in that it is a source of CO2 to the atmosphere while simultaneously exporting DIC and perhaps DOC to the open ocean. Two overarching physical and biological features that distinguish ocean margins are their modes of receiving organic matter and inorganic nutrients [Pomeroy et al., 2000]. Firstly, the SAB has no clear stratification and does not store nutrients due to its shallow depth nor does it receive significant amount of nutrients from rivers [Menzel, 1993]. In the ECS and the EuAS, strong spring and summer biological uptake of CO2 outweighs the increase of CO2 due to warming and result in a low pCO2 during warm months. Riverine nutrient supply during summer stratification appears critical to the drawn down of sea surface pCO2 [Frankignoulle and Borges, 2001; DeGrandpre et al., 2002]. In the MAB, spring and fall blooms drive CO2 below the atmosphere, and over-saturation occurs only during summer [DeGrandpre et al., 2002]. In the SAB, however, warming and high microbial respiration drive CO2 above the atmosphere during most of the year (March–November).
 Secondly, a striking feature of the SAB is its vast intertidal marshes. The area of salt marshes boarding the SAB accounts for 80% of all salt marshes along the U.S. Atlantic coast [Reimold, 1977]. Export of materials from the intertidal marshes has been hypothesized for supporting various biological activities in waters of estuaries and shelves. While the high marsh to shelf area ratio of the SAB is rare, it is not unique since salt marshes in the SAB area takes only a few percent of the world total inventory. We may characterize the SAB as a represent of marsh-dominated (heterotrophic) margins and the others as river-dominated (autotrophic) margins. More important than the amount of organic carbon supply is the fact that quality of organic carbon supplied by the rivers and by the marshes may be different in terms of its biological availability. Marsh-supplied DOC and POC may be younger and more readily available to microbes than those supplied by (large) rivers as maybe expected [Raymond and Bauer, 2001].
 This characterization of margins as marsh- or river-dominated does not mean to neglect ocean influences. In fact, ocean currents are either the principle nutrient supplier (in the SAB) or an important one (in other shelves). While the SAB, the ECS and the MAB are all western boundary current margins, their CO2 dynamics are quite different. Comparative studies of these margins will be highly instructive. Also, margins with abundant salt marshes can also be river-dominated such as part of the Louisiana-Texas shelf, which is driven by nutrient supply from the Mississippi River. Finally, other types of margins, such as those dominated by the upwelling of deep ocean waters, are not discussed here [Liu et al., 2000].
 It was proposed that “biological export of shelf [particulate organic] carbon [to the slope sediments] is a sink of the global CO2 cycle [Walsh et al., 1981].” Recent studies showed that DOC export could be significant [Vlahos et al., 2002]. We show that DIC is an important exporting term and the mechanism by which atmospheric CO2 is transferred into the ocean is different for different margins. We also suggest that ocean carbon sequestration can proceed effectively through the absorption of atmospheric CO2 by marsh grasses and the subsequent export to the open ocean.
 We thank L. R. Pomeroy and J. E. Bauer for discussions on the nature of ocean margins. Supports were provided by NOAA Luces program and NSF Georgia Coastal Ecosystems-LTER program. R. Jahnke and J. Nelson also provided ship times. Comments from two reviewers improved the presentation.