The exchange of CO2 between the atmosphere and the global coastal ocean was evaluated from a compilation of air-water CO2 fluxes scaled using a spatially-explicit global typology of inner estuaries (excluding outer estuaries such as large river deltas) and continental shelves. The computed emission of CO2 to the atmosphere from estuaries (+0.27 ± 0.23 PgC yr−1) is ∼26% to ∼55% lower than previous estimates while the sink of atmospheric CO2 over continental shelf seas (−0.21 ± 0.36 PgC yr−1) is at the low end of the range of previous estimates (−0.22 to −1.00 PgC yr−1). The air-sea CO2 flux per surface area over continental shelf seas (−0.7 ± 1.2 molC m−2 yr −1) is the double of the value in the open ocean based on the most recent CO2 climatology. The largest uncertainty of scaling approaches remains in the availability of CO2 data to describe the spatial variability, and to capture relevant temporal scales of variability.
 Inner estuaries and other near-shore ecosystems are net sources of CO2 to the atmosphere [e.g., Frankignoulle et al., 1998; Borges et al., 2003] and may account for a global emission of CO2 of a similar order of magnitude as the CO2 sink from continental shelf seas, ranging between +0.4 PgC yr−1 and +0.6 PgC yr−1 [Abril and Borges, 2004; Borges, 2005; Borges et al., 2005; Chen and Borges, 2009]. This range of flux values reflects the heterogeneity and complexity of these highly active biogeochemical environments at the interface between the land and the ocean, but also demonstrates the insufficient data coverage both in time and space, and the lack of appropriate spatially-explicit numerical models for carbon cycling in the global coastal ocean.
 As an alternative, scaling approaches can be used where a reasonable flux value for a coastal system is multiplied by the respective surface area [Abril and Borges, 2004; Borges, 2005; Borges et al., 2005; Cai et al., 2006; Chen and Borges, 2009]. The success of such scaling approaches not only depends on the quality and quantity of the measurements and how representative they are for a given coastal environment, but also on the accurate determination of the respective surface area. In this study, we evaluate sources and sinks of CO2 in the global coastal ocean using a scaling approach, based on surface areas from a spatially-explicit coastal typology of both estuaries and continental shelf seas.
2. Budget Calculations
 We calculated the exchange of CO2 between inner estuaries and the atmosphere based on a compilation of 62 published annual air-water CO2 fluxes based on pCO2 measurements (Table S1 of the auxiliary material), and the surface areas of four estuarine types, based on morphological differences [Dürr et al., 2010]: I – small deltas and small estuaries, II – tidal systems and embayments, III – lagoons, IV – fjords and fjärds. Note that outer estuarine plumes protruding onto continental shelves were not considered as estuaries. This is the case for large-river deltaic estuaries (LDE) [Bianchi and Allison, 2009] such as the Amazon and the Changjiang. Average air-water CO2 fluxes were calculated for each estuarine type and extrapolated globally, based on the type-specific surface areas (Table 1). The air-water CO2 fluxes representative for each type are based on 19, 36, 6 and 1 estimates for Types I, II, III, IV, respectively.
Table 1. Air-Water CO2 Fluxes per Surface Area and Scaled Globally for Four Estuarine Typesa
Surface Area (106 km2)
Air-Water CO2 Flux (molC m−2 yr−1)
Air-Water CO2 Flux (PgC yr−1)
Air-water CO2 fluxes per surface area in molC m−2 yr−1 and scaled globally in PgC yr−1 based on averages of individual estimates given in Table S1, A positive value represents a source of CO2 to the atmosphere.
The standard deviation was estimated as ±80% based on the values of the other 3 types.
Small deltas and estuaries (Type I)
25.7 ± 15.8
0.026 ± 0.016
Tidal systems and embayments (Type II)
28.5 ± 24.9
0.094 ± 0.082
Lagoons (Type III)
17.3 ± 16.6
0.052 ± 0.050
Fjords and fjärds (Type IV)
17.5 ± 14.0b
0.096 ± 0.077
21.0 ± 17.6
0.268 ± 0.225
 The typology of continental shelf seas relies on 138 units with surface areas calculated using a geographical information system. The off-shore limit of the continental shelf is set to 200 m depth [Walsh, 1988; Wollast, 1998] and the related isobath was extracted from the 1' resolution global bathymetry of Smith and Sandwell . Each shelf unit was defined by extrapolating perpendicularly the limits of coastal segments from the shoreline [Meybeck et al., 2006]. These segments were designed by identifying homogeneous stretches of coast according to a set of parameters such as morphology, lithology, oceanic currents and climate, not biased by national or political boundaries. A type was attributed to each continental shelf sea unit: Type 1 corresponds to enclosed shelves; Type 2 includes Western and Eastern boundary currents characterized by coastal upwelling and are separated according to the oceanic basin (Pacific, Atlantic and Indian); Type 3 consists of all other open continental shelf areas, ranked by climatic zones (Type 3a (tropical): 0–30°, Type 3b (temperate): 30–60°, Type 3c (polar): 60–90°). A total of 37 published air-water CO2 fluxes were compiled (Table S2) and scaled by types based on their respective surface areas (Table 2).
Table 2. Air-Water CO2 Fluxes per Surface Area and Scaled Globally for Different Types of Continental Shelves Along Three Climatic Zonesa
Surface Area (106 km2)
Air-Water CO2 Flux (molC m−2 yr−1)
Air-Water CO2 Flux (PgC yr−1)
Air-water CO2 fluxes per surface area in molC m−2 yr−1 based on averages of individual estimates given in Table S2 and scaled globally in PgC yr−1. A positive value represents a source of CO2 to the atmosphere.
Standard deviation on the mean of seasonal fluxes at one site (Oman coast), while for the others the standard deviation is on the mean across systems of the same type.
 We excluded studies in estuaries and continental shelf seas that did not provide an adequate representation of the annual net CO2 flux. Since it not possible to evaluate the accuracy of the individual computed air-water CO2 fluxes given by the different studies, the standard deviation of the means for each continental shelf type and each estuarine type were propagated to provide an estimate of the uncertainty on the scaled fluxes.
 A detailed description of the estuarine typology we used is given by Dürr et al. . In brief, fjords and fjärds (Type IV) are dominant at latitudes north of 45°N and south of 45°S (Figure 1a) and are the most extensive of the four estuarine types (∼43% of the total surface area). Lagoons (Type III, ∼24% of the total surface area) are dominant in the tropics and subtropics of the Northern Hemisphere (0°–45°N). Small deltas (Type I, ∼8% of the total surface area) and tidal systems (Type II, ∼26% of the total surface area) show no clear latitudinal pattern. The total surface area of estuaries is 1.1 106 km2.
 The surface area of continental shelves totals 24.7 106 km2 with a contribution of 6% by enclosed shelves (Type 1), 9% by coastal upwelling systems (Type 2) and 82% by the open continental shelves (Types 3a,b,c) (Figure 1b). About 75% of the surface area of continental shelf seas is located in the Northern Hemisphere, and ∼45% is located north of 45°N.
 The emission of CO2 to the atmosphere from estuarine environments shows two maxima, one at the equator and another at ∼65°N (Figure 2a). These two maxima correspond to a small peak in surface area (associated to Types I, II and III with high air-water CO2 fluxes) and a large peak in surface area (associated to Type IV with a lower air-water CO2 flux), respectively (Figure 2c). The overall emission of CO2 to the atmosphere from estuarine environments is estimated at +0.27 ± 0.23 PgC yr−1 (Table 1). Tidal systems (Type II) and fjords and fjärds (Type IV) contribute equally (∼35%) to the global estuarine CO2 emission, while lagoons (Type III) and small deltas (Type I) contribute 20% and 10% to the global estuarine CO2 emission, respectively. About 79% of the total CO2 emission to the atmosphere from estuaries occurs in the Northern Hemisphere, comparable to the areal extent (81%). The contribution to the total CO2 emission by estuaries along climatic zones is relatively homogeneous: 32% for tropical systems, 31% for temperate systems, and 37% for high latitude systems, for 29, 31 and 40% of area, respectively.
 The exchange of CO2 between continental shelf seas and the atmosphere as a function of latitude shows a clear asymmetry with regions between 30°S and 30°N (Figure 2d) acting as sources of CO2 to atmosphere and temperate and high latitude regions (south of 30°S, north of 30°N) acting as sinks for atmospheric CO2 (Figure 2b). The continental shelf seas of the Northern Hemisphere are a net sink of CO2 of −0.24 PgC yr−1 and the continental shelf seas of the Southern Hemisphere are a weak source of CO2 of +0.03 PgC yr−1. Globally, continental shelf seas are a net sink of atmospheric CO2 of −0.21 ± 0.36 PgC yr−1.
 The integrated air-water CO2 flux in the global coastal ocean (estuaries and continental shelves) is close to neutral (+0.06 PgC yr−1). The latitudinal pattern of a CO2 source in low latitudes and sink of CO2 at temperate and high latitudes prevails when integrating both continental shelves and estuaries (Figure 2f).
 The general patterns of air-water CO2 fluxes in the coastal ocean in the present study are similar to those reported by previous studies [Borges, 2005; Borges et al., 2005; Cai et al., 2006; Chen and Borges, 2009]. Continental shelf seas in the tropics are sources of CO2 to the atmosphere, while temperate and high latitude continental shelf seas are sinks for atmospheric CO2. The overall emission of CO2 from estuarine environments is of the same order of magnitude as the sink of CO2 of continental shelf seas. Integrated CO2 fluxes from both continental shelf seas and estuarine environments are more intense in the Northern than in the Southern Hemisphere. An improvement in our study with respect to previous ones is that coastal upwelling systems are separated by ocean basins. Indeed, based on published data with reasonable or full annual coverage, coastal upwelling systems in the Pacific and Indian Oceans are sources of CO2 to the atmosphere, while coastal upwelling systems in the Atlantic Ocean are sinks of atmospheric CO2 (Table S2). This is related to the fact that oxygen minimum zones (OMZ) associated to coastal upwelling systems are shallow in the Pacific and Indian Oceans, and are deeper or absent in the Atlantic Ocean. The upwelling source waters in coastal upwelling areas associated to a shallow OMZ are sources of CO2 to the atmosphere as denitrification leads to excess of dissolved inorganic carbon relative to nitrogen [Friederich et al., 2008; Borges, 2010]. Due to the scarceness of data, we chose to keep the extrapolation scheme of Borges  and Borges et al.  by latitudinal bands of 30° irrespective of oceanic basins or biogeochemical provinces as applied by Cai et al. . Moreover, the air-sea CO2 fluxes in open continental shelf seas (Type 3) show a relatively regular pattern as a function of latitude (Figure S1).
 There are marked differences between the present and previous studies in the globally integrated air-water CO2 flux values for both continental shelf seas and estuarine environments. The sink of atmospheric CO2 over continental shelf seas (−0.21 ± 0.36 PgC yr−1) is at the low end of the range of previously published estimates (−0.22 to −0.45 PgC yr−1) based on compilations from different shelf systems [Borges, 2005; Borges et al., 2005; Cai et al., 2006; Chen and Borges, 2009], and distinctly lower than the estimate based on the global extrapolation of the air-sea CO2 flux from the East China Sea [−1.00 PgC yr−1, Tsunogai et al., 1999]. Note that the value of air-water CO2 flux in the East China Sea given by Tsunogai et al.  of −2.9 molC m−2 yr−1 is higher than the most recent evaluations in the East China Sea (−0.9 to −2.1 molC m−2 yr−1 (Table S1)). The total surface area of continental shelf seas used in the present study (24.7 106 km2) is lower than the one used by Borges , Borges et al.  and Cai et al.  (25.8 106 km2 based on the work by Walsh ) and than the one used by Chen and Borges  (30.0 106 km2). Furthermore, Borges  and Borges et al.  used a total surface area of continental shelf seas located between 30°N and 30°S (3 106 km2) that was under-estimated compared to the one of the present typology (10 106 km2). The use of skewed surface areas in these studies led to an overestimation of the sink of CO2, as the global air-water CO2 flux per surface area was −1.17 and −1.44 molC m−2 yr−1 for tropical and temperate shelf seas, respectively. The global air-water CO2 flux per surface area in the present study (−0.71 ± 1.23 molC m−2 yr−1) is identical to the one computed by Cai et al.  and close to the one by Chen and Borges  (−0.92 molC m−2 yr−1). The air-water CO2 flux per surface area over continental shelf seas is the double of the value in the open ocean based on the most recent CO2 climatology (−0.35 molC m−2 yr−1 [Takahashi et al. 2009]).
 The emission of CO2 from estuaries given by the present study (0.27 ± 0.23 PgC yr−1) is lower than previous estimates that range between +0.36 and +0.60 PgC yr−1 [Abril and Borges, 2004; Borges, 2005; Borges et al., 2005; Chen and Borges, 2009]. This is due to the fact that previous global scaling attempts of the CO2 emission from estuaries used the average of air-water CO2 fluxes across estuarine types, and due to smaller (older) data-sets possibly biased towards tidal European (often polluted) systems. Hence, the average air-water CO2 fluxes used for scaling ranged between +32.1 and +38.2 molC m−2 yr−1, which is higher than the global average value of +21.0 ± 17.6 molC m−2 yr−1 given in Table 1 that takes into account the relative surface area of different estuarine types. This is mainly due to the fact that a large fraction of the surface area of estuarine environments corresponds to fjords and fjärds that are characterized by lower air-water CO2 flux rates than Types I and II. Further, the global surface area of estuarine environments based on the typology of Dürr et al.  of ∼1.1 106 km2 is lower than the value of 1.4 106 km2 given by Woodwell et al.  used by Abril and Borges . The scaling of estuarine CO2 emissions by Borges , Borges et al.  and Chen and Borges  was based on a global estuarine surface area of 0.94 106 km2, also derived from the values given by Woodwell et al.  but excluding inter-tidal areas associated to marshes and mangroves.
 Our typology of continental shelf seas could be further improved by explicitly distinguishing between coastal upwelling systems with and without an OMZ. The estuarine typology could be improved by distinguishing between micro-tidal and macro-tidal systems, since the former are usually highly stratified and are lower sources of CO2 to the atmosphere than the latter that are usually permanently well-mixed [e.g., Borges, 2005; Koné et al., 2009]. However, the degree of detail in a typology depends on the availability of appropriate data for each type. At present, the lack of sufficient data is the major limitation in the quantification of the spatial and temporal variability of CO2 fluxes in coastal environments. In estuarine environments, there is a fair amount of data to characterize tidal systems (Type II) and small deltas (Type I). However, for fjords and fjärds (Type IV), that represent 43% of the total estuarine surface area, adequate air-water CO2 flux data are only available from one location. For lagoons (Type III), most of the available data were obtained from 5 contiguous systems located in Ivory Coast (∼5°N) although these estuarine ecosystems are ubiquitous at all latitudes (Figure 1).
 We did not attempt to explicitly scale CO2 fluxes in river plumes (or outer estuaries). Data with adequate spatial and temporal coverage in these systems to robustly evaluate air-sea CO2 fluxes are scarce. Some outer estuaries act as sources of CO2 to the atmosphere such as the Scheldt (+1.9 molC m−2 yr−1 [Borges and Frankignoulle, 2002]), the Loire (+10.5 molC m−2 yr−1 [de la Paz et al., 2010]), the Kennebec (+0.9 molC m−2 yr−1 [Salisbury et al., 2009]), while others act as sinks for atmospheric CO2 such as the Amazon (−0.5 molC m−2 yr−1 [Körtzinger, 2003]) and the Changjiang (−1.9 molC m−2 yr−1 [Zhai and Dai, 2009]). The direction of the annual net flux of CO2 in outer estuaries is, to a large extent, related to the presence or the absence of haline stratification that promotes export of organic matter across the pycnocline and enhances light availability for primary production [Borges, 2005]. Haline stratification generally occurs in high freshwater discharge systems, hence, LDE systems (Amazon, Changjiang) act as sinks of CO2, while smaller systems that are generally devoid of haline stratification (Scheldt, Loire, Kennebec) act as sources of CO2 to the atmosphere. Hence, a typological approach taking into account physical and biogeochemical characteristics that drive a net annual sink or source of CO2 is required to estimate the global surface of outer estuaries such as LDE and scale globally the CO2 fluxes from these environments, in addition to more observations in different systems.
 The data availability in continental shelf seas is strongly biased towards the temperate regions of the Northern Hemisphere, while coastlines of the Russian Arctic, eastern South America, eastern Africa, large sections of western Africa, and most of Antarctica are dramatically under-sampled. Finally, pCO2 temporal variability ranges from daily [Dai et al., 2009] to inter-annual [Friederich et al., 2002; Borges et al., 2008a, 2008b] scales. The (in)adequate representation of the full range of temporal variability can impact the evaluation of the overall net annual air-sea CO2 fluxes. For a more robust evaluation of CO2 fluxes in continental shelf seas, an intensive, integrated, international and interdisciplinary program of observational efforts is required.
 This work was funded by Utrecht University (High Potential Project G-NUX) and the Netherlands Organisation for Scientific Research (NWO Vidi grant 864.05.004). It is a contribution to EU IP CARBOOCEAN (511176), EU IP SESAME (GOCE-2006-036949), EU CSA COCOS (212196) and COST Action 735. AVB is a research associate at the FRS-FNRS.