Air – sea exchanges of CO 2 in the world ’ s coastal seas

The air–sea exchanges of CO 2 in the world’s 165 estuaries and 87 continental shelves are evaluated. Generally and in all seasons, upper estuaries with salinities of less than two are strong sources of CO 2 (39± 56 mol C m−2 yr−1, positive flux indicates that the water is losing CO 2 to the atmosphere); mid-estuaries with salinities of between 2 and 25 are moderate sources (17.5 ± 34 mol C m−2 yr−1) and lower estuaries with salinities of more than 25 are weak sources (8.4± 14 mol C m−2 yr−1). With respect to latitude, estuaries between 23.5 and 50 ◦ N have the largest flux per unit area (63± 101 mmol C m−2 d−1); these are followed by lowerlatitude estuaries (23.5–0 ◦ S: 44± 29 mmol C m−2 d−1; 0–23.5 N: 39± 55 mmol C m−2 d−1), and then regions north of 50 N (36± 91 mmol C m−2 d−1). Estuaries south of 50 S have the smallest flux per unit area (9.5± 12 mmol C m−2 d−1). Mixing with low-pCO2 shelf waters, water temperature, residence time and the complexity of the biogeochemistry are major factors that govern the pCO2 in estuaries, but wind speed, seldom discussed, is critical to controlling the air–water exchanges of CO 2. The total annual release of CO 2 from the world’s estuaries is now estimated to be 0.10 Pg C yr −1, which is much lower than published values mainly because of the contribution of a considerable amount of heretofore unpublished or new data from Asia and the Arctic. The Asian data, although indicating highpCO2, are low in sea-to-air fluxes because of low wind speeds. Previously determined flux values rely heavily on data from Europe and North America, where pCO2 is lower but wind speeds are much higher, such that the CO2 fluxes are higher than in Asia. Newly emerged CO 2 flux data in the Arctic reveal that estuaries there mostly absorb rather than release CO 2. Most continental shelves, and especially those at high latitude, are undersaturated in terms of CO 2 and absorb CO2 from the atmosphere in all seasons. Shelves between 0 and 23.5 S are on average a weak source and have a small flux per unit area of CO2 to the atmosphere. Water temperature, the spreading of river plumes, upwelling, and biological production seem to be the main factors in determining pCO2 in the shelves. Wind speed, again, is critical because at high latitudes, the winds tend to be strong. Since the surface water pCO2 values are low, the air-to-sea fluxes are high in regions above 50 N and below 50 S. At low latitudes, the winds tend to be weak, so the sea-to-air CO 2 flux is small. Overall, the world’s continental shelves absorb 0.4 Pg C yr −1 f om the atmosphere.

Abstract.The air-sea exchanges of CO 2 in the world's 165 estuaries and 87 continental shelves are evaluated.Generally and in all seasons, upper estuaries with salinities of less than two are strong sources of CO 2 (39 ± 56 mol C m −2 yr −1 , positive flux indicates that the water is losing CO 2 to the atmosphere); mid-estuaries with salinities of between 2 and 25 are moderate sources (17.5 ± 34 mol C m −2 yr −1 ) and lower estuaries with salinities of more than 25 are weak sources (8.4 ± 14 mol C m −2 yr −1 ).With respect to latitude, estuaries between 23.5 and 50 • N have the largest flux per unit area (63 ± 101 mmol C m −2 d −1 ); these are followed by lowerlatitude estuaries (23.5-0 • S: 44 ± 29 mmol C m −2 d −1 ; 0-23.5 • N: 39 ± 55 mmol C m −2 d −1 ), and then regions north of 50 • N (36 ± 91 mmol C m −2 d −1 ).Estuaries south of 50 • S have the smallest flux per unit area (9.5 ± 12 mmol C m −2 d −1 ).Mixing with low-pCO 2 shelf waters, water temperature, residence time and the complexity of the biogeochemistry are major factors that govern the pCO 2 in estuaries, but wind speed, seldom discussed, is critical to controlling the air-water exchanges of CO 2 .The total annual release of CO 2 from the world's estuaries is now estimated to be 0.10 Pg C yr −1 , which is much lower than published values mainly because of the contribution of a considerable amount of heretofore unpublished or new data from Asia and the Arctic.The Asian data, although indicating high pCO 2 , are low in sea-to-air fluxes because of low wind speeds.Previously determined flux values rely heavily on data from Europe and North America, where pCO 2 is lower but wind speeds are much higher, such that the CO 2 fluxes are higher than in Asia.Newly emerged CO 2 flux data in the Arctic reveal that estuaries there mostly absorb rather than release CO 2 .
Most continental shelves, and especially those at high latitude, are undersaturated in terms of CO 2 and absorb CO 2 from the atmosphere in all seasons.Shelves between 0 and 23.5 • S are on average a weak source and have a small flux per unit area of CO 2 to the atmosphere.Water temperature, the spreading of river plumes, upwelling, and biological production seem to be the main factors in determining pCO 2 in the shelves.Wind speed, again, is critical because at high latitudes, the winds tend to be strong.Since the surface water pCO 2 values are low, the air-to-sea fluxes are high in regions above 50 • N and below 50 • S. At low latitudes, the winds tend to be weak, so the sea-to-air CO 2 flux is small.Overall, the world's continental shelves absorb 0.4 Pg C yr −1 from the atmosphere.

Introduction
Carbon is arguably one of the most important elements on earth, and understanding the global carbon cycle is fundamental to elucidating the effect of human activities in the Anthropocene era.The oceans are known to have an important role in regulating the climate on annual to millennial scales by absorbing CO 2 and exchanging carbon with various carbon-storing compartments, such as the atmosphere, the land, the biota and the fossil fuel carbon pool.Yet, despite the success of quantifying the air-sea CO 2 exchange and the uptake of anthropogenic CO 2 by the major oceans, the effect of the land on these processes is still poorly understood and little discussed (Khatiwala et al., 2013;Le Quéré et al., 2013;Schuster et al., 2013;Wanninkhof et al., 2013).
C.-T. A. Chen et al.: Air-sea exchanges of CO 2 in the world's coastal seas Coastal waters link the land, the oceans, the atmosphere, biota and sediments.Although they constitute only a little over 7 % of the surface area of oceans and less than 0.5 % of the volume of the oceans, coastal oceans have a disproportionately large role in primary and new production, remineralization and sedimentation of organic matter (Walsh et al., 1981;Walsh, 1988Walsh, , 1991;;Kempe and Pegler, 1991;Mackenzie et al., 1991Mackenzie et al., , 1998a, b;, b;Chen, 1993;Wollast, 1993Wollast, , 1998;;Gattuso et al., 1998;Carrillo and Karl, 1999;Liu et al., 2000;de Haas et al., 2002;Elliott and McLusky, 2002;Muller-Karger et al., 2005;Thomas, 2010).Coastal waters receive large inputs of terrestrial material, such as suspended sediments and nutrients in solution or in particulate matter, in organic or inorganic forms and through river and groundwater discharge, as well as by exchange with the atmosphere, the sediments and the open ocean.They therefore tend to show greater temporal and spatial variability than open oceans, and are more affected by human activities (Cameron and Pritchard, 1963;Alongi, 1998;Chen and Tsunogai, 1998;Rabouille et al., 2001;Chen, 2002Chen, , 2003Chen, , 2004;;Slomp and Van Cappellen, 2004;Beusen et al., 2005;Chavez et al., 2007;Doney et al., 2007;Radach and Patsch, 2007;Peng et al., 2008;Seitzinger et al., 2010;Dürr et al., 2011;Jiang et al., 2013).However, unlike the open oceans, in which millions of observations have been made and the air-sea exchanges of CO 2 have been valued using various developed models (such as by Khatiwala et al., 2013;Schuster et al., 2013;Wanninkhof et al., 2013), coastal waters have been relatively poorly examined.
Although estuaries are known to be generally sources of CO 2 (Frankignoulle et al., 1998;Cai et al., 1999Cai et al., , 2000;;Sarma et al., 2001Sarma et al., , 2011;;Abril et al., 2002;Borges et al., 2003;Dagg et al., 2005;Gao et al., 2005;Dai et al., 2008;Leinweber et al., 2009), only in the last few years have continental shelves been firmly established to absorb CO 2 from the atmosphere.(See, for example, Liu et al., 2000;Chen et al., 2003;Chen, 2004;Abril and Borges, 2005;Borges, 2005;Borges et al., 2005;Cai et al., 2006;Chen and Borges, 2009;Laruelle et al., 2010, and references therein.)Indeed, whether coastal seas are sources or sinks of CO 2 has remained an open question until only recently.The first report of the project on the Land Ocean Interaction in the Coastal Zone (LOICZ) under the International Geosphere Biosphere Programme (IGBP) is entitled, "Coastal seas: a net source or sink of atmosphere carbon dioxide" (Kempe, 1995).The first report of LOICZ did not provide any data concerning the airsea exchanges of carbon in the continental margins, although it concluded that net carbon oxidation in the coastal zone is around 7 × 10 12 mol yr −1 (Crossland et al., 2005), implying that the coastal zone is a source of CO 2 to the atmosphere.
Unfortunately, Fasham et al. (2001), summarizing the work of the Joint Global Ocean Flux Study (JGOFS, another IGBP project), concluded that there is a net sea-to-air CO 2 flux from continental margins of 0.5 Pg C yr −1 .They drew this conclusion despite the fact that, at the time, the joint JGOFS/LOICZ Continental Margins Task Team (Chen et al., 1994) had already gathered sufficient data to demonstrate that the margins, rather than being a source of CO 2 , are in fact a sink of CO 2 .Indeed, in the same year, Fasham published another paper that claimed that the continental shelves are actually a sink of CO 2 of the order of 0.6 Pg C yr −1 (Yool and Fasham, 2001).In 2003, the JGOFS also concluded that the shelves take up 0.3 Pg C yr −1 of atmospheric CO 2 (Chen et al., 2003).This view, however, was not universally accepted (Cai et al., 2003;Cai and Dai, 2004) until more data, especially data obtained in the winter, became available.Many shelves that had been thought to be sources of CO 2 are now known to be sinks of CO 2 when winter data reveal severe undersaturation of CO 2 (Thomas et al., 2004;Cai et al., 2006;Schiettecatte et al., 2007;Jiang et al., 2008b).
Strangely, despite the fact that coastal waters play a major role in the livelihood of humans, and are strongly affected by human activities, our understanding of these waters is mostly semi-quantitative.For example, such basic information as the area of the continental shelf is uncertain.The most recent work of Kang et al. (2013) yielded an area of 26.15 × 10 6 km 2 for waters shallower than 200 m.This value compares with 26.39 × 10 6 km 2 obtained by Laruelle et al. (2013), 24.72 × 10 6 km 2 presented by Laruelle et al. (2010), 30.16 × 10 6 km 2 obtained by Jahnke (2010), 26 × 10 6 km 2 presented by Chen and Borges (2009), 25.83 × 10 6 km 2 presented by Cai et al. (2006) and 36 × 10 6 km 2 presented by Liu et al. (2000), which may seem to be an outlier.Merely comparing the total flux across various studies may not be very useful, whereas comparing flux per unit area eliminates the problem of an uncertain global shelf area, which varies by as much as 50 % among studies.Even more strangely, despite the fact that rivers export approximately 1 Pg C yr −1 (Meybeck, 1982), or roughly half of the carbon that is absorbed by the open oceans each year, this value needs to be confirmed as it was based only on a few studies, and the well-regarded study of Meybeck was based on a database of only 27 rivers.
The export of carbon by rivers comprises 40 % organic carbon (0.22 Pg C yr −1 of dissolved organic carbon (DOC) and 0.18 Pg C yr −1 of particulate organic carbon (POC)) and 60 % inorganic carbon (0.43 Pg C yr −1 of dissolved inorganic carbon (DIC) and 0.17 Pg C yr −1 of particulate inorganic carbon (PIC)) (Meybeck, 1982;Richey, 2004;IPCC, 2007;Schlunz and Schneider, 2000;Dai et al., 2012;Huang et al., 2012).However, estuarine filtering prevents some of the carbon that reaches the estuaries from also entering the oceans (Keil et al., 1997;Kemp et al., 1997;Middelburg and Herman, 2007;Chen et al., 2012;Dai et al., 2012).Further, the exact extent of speciation changes between the organic and inorganic or dissolved and particulate carbon in the estuaries, and how much of each of these forms of carbon actually enters the oceans are still unknown (Woodwell et al., 1973;Raymond and Bauer, 2000;Wiegner and Seitzinger, 2001;Cai, 2011;Maher and Eyre, 2011).
Since the above complex and conflicting factors influence the pCO 2 of estuarine and shelf waters, the air-sea exchanges of CO 2 in these waters globally cannot yet be estimated by models although regional models have been attempted (Hofmann et al., 2011;Maher and Eyre, 2012;Wakelin et al., 2012).As a result, field data are still required.Determinations of the air-sea flux of CO 2 in the world's estuaries and continental shelves, based on direct measurements, are presented below.Data from the literature and some unpublished data from C. T. A. Chen are tabulated.Data for upper, mid-and lower estuaries are compared.Seasonal and latitudinal variations are discussed, and the global flux is presented.Data concerning continental shelves are also considered with reference to season and latitude before the global flux is determined.

Sea-to-air CO fluxes in estuaries
Rivers are the main sources of carbon to the estuaries.Riverine organic carbon is supplied primarily by the erosion of soil organic matter or plant detritus (allochthonous) and by phytoplankton in water (autochthonous).The inorganic carbon is derived mainly from soil and rock erosion, and by the oxidation of organic matter mostly through microbial processes (Odum and Hoskin, 1958;Odum and Wilson, 1962;Probst et al., 1994;Neal et al., 1998;Nelson et al., 1999;Pomeroy et al., 2000).These organic and inorganic forms of carbon in dissolved and particulate phases reach the estuaries, which are typically wider than river channels.Therefore, particles tend to settle down and decompose, releasing car-bon back into the water.Salt marshes, mangroves, and submarine groundwater discharge also export carbon to estuaries, increasing their pCO 2 .
Rivers are the main sources of nutrients to estuaries.However, high turbidity and limited light cause nutrients rarely to be fully utilized for biological production in rivers or estuaries.Hence, the biological drawdown of CO 2 does not suffice to reduce the estuarine water pCO 2 to below saturation.Consequently, almost all estuaries are sources of CO 2 to the atmosphere.The influence of freshwater from large rivers frequently extends hundreds of kilometers offshore.The enormous discharge of freshwater, sediments and the associated particulate and dissolved organic and inorganic carbon, nitrogen and phosphorus all greatly affect the biological and geochemical processes in the estuary, the plume and the adjacent continental shelf (Chen and Wang, 1999;Gong et al., 2000;Chen et al., 2003).Generally, net ecosystem production in estuaries tends to be net heterotrophic: respiration is larger than production (Battin et al., 2008).Various complex biogeochemical processes in estuaries are affected by the topography and river flow.As small deltas and large rivers' estuaries have short residence time (Dürr et al., 2011), physical mixing is the major factor affecting carbonate parameters.On the other hand, with longer residence time the transformation between inorganic and organic material becomes more active.This is because now suspended particles have more time to settle and aquatic organisms have more time to grow, and leach dissolved organic carbon, when light becomes more available in the nutrient-abundant estuaries.On the other hand, dissolved organic carbon decomposes more when the residence time is longer compared with physical force-dominant estuaries.A saline interface normally separates the plume water from the shelf water, with the width of the interface determined by interactions between river discharge and marine-driving forces (Shen, 2001;Shen et al., 2003;Chen et al., 2008a).Complex biophysical and geochemical processes govern the direction of CO 2 exchange between the plume-affected shelf area and the atmosphere (Kortzinger, 2003;Cooley et al., 2007), but in this investigation, river plumes outside of the estuaries are not considered.Tidal forcing on estuarine mixing affects submarine groundwater discharge, sediment burial and disturbance, the pCO 2 in the surface water as well as the air-to-sea CO 2 exchange.These, however, have not been evaluated in a quantitative way.
Numerical data are gathered for 165 estuaries (Fig. 1, Table 1), of which 99 are from literature.Unpublished data from 50 estuaries and 16 from data banks are also included, and the Wanninkhof (1992) quadratic equation is used to determine the flux.The methods used to calculate the flux, as well as sources of the gas exchange coefficient and wind speed, are listed in Table 2. Of note is that using different pCO 2 flux methods and gas transfer velocities causes disparity in flux estimations (Borges et al., 2004;Ferron et al., 2007;Jiang et al., 2008a;Zappa et al., 2007)       estuaries.Factors affecting gas exchange coefficients include wind speed, tidal current and bottom stress, whereas the wind speed is the most considered.It is important to point out that this paper deals mostly with published results.It is not possible to re-do the flux calculations, say, based on the same gas exchange coefficient, as the original data were not provided in the papers cited.Further, there is a lack of temporal coverage as previous studies (Bozec et al., 2011;Dai et al., 2009;Kitidis et al., 2012) have demonstrated short-term changes in pCO 2 at scales of days or less.Yet, typically data on such a scale are limited to only a few cruises.The lack of seasonality in the numerically averaged fluxes is almost certainly an artifact influenced by averaging all available data.
Figure 2 presents the pCO 2 and CO 2 fluxes per unit area in the upper, mid-and lower estuaries worldwide.Up-per, mid-, and lower estuaries are defined as those areas of estuaries with salinities below 2, between 2 and 25, and above 25, respectively, as salinity data are the most readily available.Otherwise, divisions are made based approximately on one-third of the distance from the point where the river starts to widen to the river mouth.Almost all estuaries outside of the Arctic region except for only a few release CO 2 to the atmosphere.Unsurprisingly, upper estuaries, where the riverine effect is the strongest (Kempe, 1979(Kempe, , 1982;;Chen et al., 2012), have the highest pCO 2 (numerical average = 5026 ± 6190 µatm) and the highest sea-to-air CO 2 flux (numerical average = 39.0 ± 55.7 mol C m −2 yr −1 , where the positive sign indicates that the seawater is losing CO 2 ); these are followed by the mid-estuaries (numerically averaged pCO      ).Lower estuaries have the lowest pCO 2 (numerical average = 723 ± 957 µatm) and CO 2 flux (numerical average = 8.4 ± 14.3 mol C m −2 yr −1 ).Except for those of the upper estuaries, these pCO 2 values compare favorably with those found by Chen et al. (2012), which were 3033, 2277, and 692 µatm for the upper, mid-and lower estuaries, respectively.This study yields much higher pCO 2 values for upper estuaries mainly because new data from Asia are associated with high pCO 2 values.The fluxes obtained by Chen et al. (2012), however, are higher.Their values are 68.5, 37.4 and 9.92 mol C m −2 yr −1 for the upper, mid-and lower estuaries, respectively.The seeming inconsistency among results is discussed below.
Figure 3 displays histograms of reported daily CO 2 fluxes per unit area in different seasons and the annual flux per unit area in the world's estuaries.Little seasonality is observed, except that the flux is lower in the winter when the pCO 2 is usually lower, perhaps because the temperature is lower than other seasons.The flux is only marginally higher in summer than in spring or autumn.The numerical average annual flux per unit area is 16.5 ± 27.7 mol C m −2 yr −1 , which is significantly lower than that, 23.9 ± 33.1 mmol C m −2 d −1 , obtained by Chen et al. (2012).The numerical average an- nual flux per unit area, however, is not used to calculate the global release of CO 2 because small estuaries dominate the numerical average, but they contribute relatively little to the total flux.Important to note is that there is a lack of temporal coverage in most of the data sets although previous studies (Bozec et al., 2011;Dai et al., 2009;Kitidis et al., 2012) have demonstrated short-term changes in pCO 2 at scales of days or less.Yet, typically data on such a scale are limited to only a few cruises.The lack of seasonality in the numerically averaged fluxes is almost certainly an artifact influenced by averaging all available data.
With respect to latitude, the numerically averaged flux per unit area is the highest, at 63.3 ± 100.7 mmol C m −2 d −1 between 23.5 and 50   N, however, significantly higher than those obtained herein.Notably, most other investigations have presented fluxes higher than those that were presented by Chen et al. (2012).
The fact that the annual average flux herein is lower than those reported previously, despite the fact the average pCO 2 is higher, warrants discussion.It follows mainly from the fact that many data from the low-latitude bands in Asia have been added, and these areas are mostly areas of low wind energy.Figure 5 plots the wind energy potential, which is, like the air-sea gas exchange rate, a quadratic function of wind speed.The areas of high wind energy at low latitudes are concentrated in the dry Middle East and northeastern, northern and northwestern Africa with few rivers, and therefore few estuaries.For example, the total area of the estuaries in the Red Sea region is almost zero (Table 3; Laruelle et al., 2013).Accordingly, the global average CO 2 flux herein is substantially affected by estuaries in areas of low wind energy, and therefore of low CO 2 flux.
The 50 newly considered estuaries in Taiwan, southern China and Southeast Asia, all at low latitudes, have lower fluxes than determined from previously obtained results (Table 1), which include many data for European rivers.For instance, only 2 of the 19 estuaries that were considered by Abril and Borges (2005), who published perhaps the first global study of CO 2 emissions from estuaries, are outside Europe and the eastern seaboard of the USA.Those authors found a global CO 2 flux per unit area of 35.7 mol C m −2 yr −1 , which is more than triple the value obtained in this study.This finding does not imply that Eu-ropean rivers have higher pCO 2 : they do not.Rather, Europe has more windy coasts than elsewhere in the world, and especially Asia.Parts of these higher fluxes may have resulted from higher wind speed.As mentioned above, the wind potential is a quadratic function of wind speed, as is the 1992 Wanninkhof air-sea CO 2 exchange equation.It is important to point out, however, that the water turbulence is an importance factor for gas transfer velocity in low wind speed regions, but little data is available.We have compared the Wanninkhof (1992) quadratic equation (k660 = 0.31 × U 10 2 ) with other equations such as those of Raymond and Cole (2001), Borges et al. (2004), Ho et al. (2011), andJiang et al. (2008a).Using Wanninkhof's (1992) quadratic equation may underestimate flux, although the value is similar to that of Ho et al. (2011) at low wind speed (< 5 m s −1 ).Note that there is no theoretical basis for the above equations as most are based on curve fitting techniques.Since we do not have data to show which equation is the best, we have chosen the Wanninkhof quadratic equation, which most references we cited used.Due to the fact that using different air-sea exchange equations results in large uncertainties, and that there is no universally accepted equation, the above conclusion can only be deemed preliminary.The mean pCO 2 of European estuaries is roughly 1600 µatm, whereas that of Asian estuaries is much higher, around 4000 µatm.Yet, the mean wind speed on European coasts is approximately 4 m s −1 , compared with about 1.6 m s −1 on Asian coasts.The resulting CO 2 fluxes for European estuaries average about 16.9 mol C m −2 yr −1 vs. a much lower 8.1 mol C m −2 yr −1 for Asian estuaries (Table 3; Fig. 6) despite their higher pCO 2 .
In the above calculation, the areas of groups of estuaries are taken from the most recent and comprehensive work of Laruelle et al. (2013), which divided the world into 45 regions and calculated a total estuarine area of 1.012 × 10 6 km 2 , slightly smaller than the value of 1.067 × 10 6 km 2 given in Laruelle et al. (2010).Table 3 lists the total surface area in each of the 45 regions and the numerically averaged CO 2 flux per unit area for each region.Our global flux calculation is based on the sum of regional fluxes for these 45 zones (area multiplied by zonal average CO 2 flux (mol C m −2 yr −1 )).These 165 estuaries are compartmentalized into 35 regions, and the numerically averaged CO 2 flux per unit area is calculated.For 10 regions without data, the mean flux for the same classification region is used (Table 3).The outgassing of CO 2 in global estuaries is 0.094 Pg C yr −1 , and is about 31 % of the global riverine organic carbon flux (Seitzinger et al., 2010).This compares with the 48 % of organic carbon released as CO 2 from estuaries and inland waters (Tranvik et al., 2009).
Estuaries in North America have the largest total area, but the lowest average flux per unit area among all continents, and therefore a low total flux of 10.8 Tg C yr −1 .That is, a continent with 41 % of the world's estuarine area accounts for only 12 % of the world's estuarine CO 2 release Table 3. Areas and air-sea fluxes of CO 2 in estuaries and continental shelves by biogeochemical provinces (Laruelle et al., 2013).(Fig. 6).African, European and South American estuaries have similarly high fluxes per unit area but the areas of the estuaries are only moderate so they are responsible for only 16 % (14.7 Tg C yr −1 ), 26 % (24.1 Tg C yr −1 ) and 12 % (11.6 Tg C yr −1 ), respectively, of the global release.The largest contributor is Asia, which has 31.5 % of the world's estuary area and releases almost the same percentage of the world's estuarine-released CO 2 (32 %, or 30.6 Tg C yr −1 ; Fig. 6).
Largely on account of the distribution of data, which include data from high wind regions on both sides of the North Atlantic and around the Arabian Sea in the Indian Ocean, as well as those generally in the low wind regions around the Pacific Ocean, the mean CO 2 flux per unit area is the lowest for estuaries that flow into the Pacific Ocean, with a value of 10.5 mol C m −2 yr −1 (Fig. 7).This value compares with 12.4 mol C m −2 yr −1 for estuaries that flow into the Atlantic Ocean and 13.9 mol C m −2 yr −1 for estuaries that flow into the Indian Ocean.Because the total area of estuaries that enter the Atlantic Ocean exceeds the sum of areas of estuaries that enter the Pacific and Indian oceans, the total flux of CO 2 released from the estuaries around the Atlantic  (54.1 Tg C yr −1 ) exceeds the total flux from estuaries around the Pacific (30.8 Tg C yr −1 ) and the Indian (13.3 Tg C yr −1 ) oceans.The total area of estuaries that enter the Arctic Ocean is substantial (324 × 10 3 km 2 ), equaling the total areas of the estuaries around the Atlantic and Indian oceans.Unfortunately, the relevant data are scarce, and the available data seem to reveal that the Arctic estuaries absorb rather than release CO 2 .The numerically averaged flux per unit area and total flux are −1.1 mol C m −2 yr −1 and −4.2 Tg C yr −1 , respectively.The global total release of 94 Tg C yr −1 is less than half of any previous estimates (Table 4).New data from low wind regions and the Arctic Ocean are responsible for this difference.

Air-to-sea CO 2 fluxes in continental shelves
Data are available from 87 continental shelves (Table 5 and Fig. 8).The method used to calculate the flux, and sources of the gas exchange coefficient and wind speed are listed in Table 6.Similar to the case for estuaries, different pCO 2 flux methods and gas transfer velocities also cause disparity in the flux estimations in coastal regions.For instance, Jiang      Figure 9 displays a histogram of the reported daily CO 2 fluxes in different seasons and the annual flux for the world's continental shelves.Respiration rates are higher in summer and autumn than in winter and spring (Hopkinson, 1985(Hopkinson, , 1988;;Griffith et al., 1990;Hopkinson and Smith, 2005;Jiang et al., 2010).However, as with estuaries, no seasonality of the numerically averaged flux per unit area on continental shelves is evident, and the values fall between −4.0 and −5.5 mmol C m −2 d −1 , except in autumn, when the flux is only −0.5 mmol C m −2 d −1 .A negative value indicates that the shelves absorb CO 2 .The numerically averaged annual mean air-to-sea flux is −1.09 ± 2.9 mol C m −2 yr −1 .Multi-plying this value by the total global area of the shelves yields a global flux of −0.40 Pg C yr −1 , which is slightly less than the published value (Table 7).
Figure 10 presents a histogram of the reported daily fluxes of CO 2 in different latitude bands.Most shelves absorb CO 2 from the atmosphere (negative fluxes) while shelves at low latitudes have a slight tendency to release CO 2 (positive fluxes).This finding is consistent with the work of Cai et al. (2006), who found that shelves at low latitudes between 30 • N and 30 • S are a source of CO 2 of the order of 0.11 Pg C yr −1 , whereas those in temperate and high-latitude regions are sinks of CO 2 of the order of 0.33 Pg C yr −1 .The CO 2 flux per unit area is highest on the South American shelves (−3.6 mol C m −2 yr −1 ), but since their total area is moderate, the South American shelves absorb the second largest amount of CO 2 from the atmosphere annually at −103.5 Tg C yr −1 , or 26 % of the global shelf absorption.Asian shelves have the highest total area, but their numerically averaged flux per unit area (−0.13 mol C m −2 yr −1 ) is the lowest of all, primarily because of the generally low wind speed and because some shelves release rather than absorb CO 2 .The total annual flux from Asian shelves is only −22 Tg C yr −1 , or 5 % of the global absorption by all shelves.North American shelves rank second in terms of both numerically averaged flux per unit area (−2.1 mol C m −2 yr −1 ) and shelf area.Accordingly, North American shelves absorb the most CO 2 from the atmosphere at −156 Tg C yr −1 , or 39 % of the global absorption by all shelves.Shelves around Antarctica have the third highest numerically averaged flux per unit area (−2.0 mol C m −2 yr −1 ) and the third largest total shelf area, resulting in the third highest total annual flux at −70 Tg C yr −1 , or 18 % of the global absorption (Fig. 11).Unfortunately, data are available for only two such shelves.
Figure 12 shows the total CO 2 flux per unit area and the total CO 2 flux from shelves in different oceans. of the global absorption), the third highest numerically averaged flux per unit area (−1.2 mol C m −2 yr −1 ) and the second highest total shelf area.The largest shelf area is that of the shelves around the Pacific Ocean, but since the numerically averaged flux per unit area is low (−0.4 mol C m −2 yr −1 ), their total absorption is only −49 Tg C yr −1 , or 12 % of global absorption.Shelves around the Indian Ocean have the least total area and the second lowest numerically averaged flux per unit area (−0.6 mol C m −2 yr −1 ), resulting in the lowest total flux (−18 Tg C yr −1 , or only 4 % of the global absorption).In total, the world's shelves absorb 396 Tg C yr −1 , or 0.396 Pg C yr −1 .
The shift from the sinking of CO 2 at higher latitudes to acting as a weak source at lower latitudes is explained by four major factors.The first is that waters on continental shelves   (Takahashi et al., 2002;Kaltin et al., 2002;Kaltin and Anderson, 2005;Chen et al., 2006aChen et al., , b, 2008a, b;, b;Ciais et al., 2008).Therefore, mixing with open ocean waters at high latitudes helps shelf waters become undersaturated, whereas mixing with open ocean waters at low latitudes frequently yields shelf waters that are still supersaturated (Hidalgo-Gonzalez et al., 1997;Ito et al., 2005;Cai et al., 2006;Chen et al., 2008aChen et al., , 2012)).
The second factor that contributes to the supersaturation of shelf waters at low latitudes is temperature because pCO 2 increases by 4.3 % for an increase of 1 • C (Bakker et al., 1999;Takahashi et al., 2002).Simply increasing by 15 • C, the temperature of the shelf water with a pCO 2 of 370 µatm in the example given above would result in a pCO 2 of above 600 µatm, if all other factors are held constant.Notably, the difference between the temperatures of shelves at highlatitude shelves and those at low latitude commonly exceeds 15 • C. Temperature similarly affects open ocean water, so not only do warm temperatures increase the pCO 2 of shelf waters but also these waters also mix with open ocean waters with higher pCO 2 , mainly because they are hotter than open oceans at high latitudes.The proximal shelves in waters with a depth of, say, less than 40 m, typically exhibit a greater seasonal range of water temperatures than the distal shelves with a water depth of between 40 and 200 m.The effect of temperature on the pCO 2 of proximal shelf waters is thus greater than that of distal shelf waters.Unfortunately, the readily available pCO 2 data do not suffice for a meaningful synthesis.As a matter of fact, for navigational or geopolitical reasons, pCO 2 is rarely measured along the coast, or the data are frequently not disclosed.
The third factor in affecting pCO 2 on shelves is the fact that the discharge of organic matter by rivers is higher at lower latitudes.As much as 60 % of the riverine organic carbon discharge to the shelves occurs between 30 • N and 30 • S (Walsh, 1988;Ludwig et al., 1996a, b;Borges et al., 2005).The total amount is approximately 0.3 Pg C yr −1 , most of which is decomposed in the continental margins (McKee, 2003;Cai et al., 2006).Importantly, however, the cited studies did not identify the recipient of the riverine export.As indicated above, a significant fraction of the export is decomposed in the estuaries and does not reach the shelves (Hofmann et al., 2011;Chen et al., 2012).
The fourth factor that is responsible for the higher pCO 2 in shelves at lower latitudes involves lower biological productivity.Shelves at mid-and high latitudes are generally highly productive, whereas those at low latitudes, especially the non-upwelling shelves, are typically oligotrophic.The more effective biological pumping in the shelves at mid-and high latitudes causes these shelves to have lower pCO 2 .changes in carbon fluxes (Friederich et al., 2002;Fransson et al., 2006;Macpherson et al., 2008;Ver et al., 1999b, a;Mackenzie et al., 2000Mackenzie et al., , 2004;;Thomas et al., 2007;Borges, 2011), whether the size of the estuarine source and the continental shelf sink for CO 2 changes with time has not been determined because too few data are available.As stated above, the fact that the shelves are a sink rather than a source of CO 2 has only been established in the last few years.The conclusion of LOICZ that the coastal seas are a source of CO 2 (Crossland et al., 2005) was not based on data.Rather, it was based on the reasoning that since rivers transport more organic carbon to the oceans than is buried in the sediments, the oceans must be releasing the remaining CO 2 back to the atmosphere (Smith and Mackenzie, 1987;Smith and Hollibaugh, 1993).However, the amount of organic carbon that is actually transported to the oceans after it passes through the estuaries is not clear, as mentioned above.Before industrialization, more organic carbon may well have reached the oceans than accumulated in the sediments, such that the oceans were overall a source of CO 2 .However, whether the CO 2 was released in the coastal oceans or in the open oceans remains unclear.Further, the increasing CO 2 in the atmosphere must have reduced the difference between the pCO 2 of the atmosphere and that of the oceans, even if the oceans used to be supersaturated with CO 2 before industrialization.In any case, the present data clearly demonstrate that the coastal oceans are a sink of CO 2 .Whether a threshold has been crossed or whether the metabolism of the ecosystem has been changed cannot yet be determined (Mackenzie et al., 2004).

Temporal changes
Interestingly, heterotrophic systems are commonly treated as CO 2 sources while autotrophic systems are considered to be CO 2 sinks (Walsh et al., 1981;Smith and Hollibaugh, 1993;Ducklow and McCallister, 2004).This assumption is usually true but must be applied with caution.For instance, if a shelf is regarded as a system, then it can be heterotrophic overall.However, its surface layer may still be undersaturated owing to cooling or biological production, and so it absorbs CO 2 from the atmosphere.Alternatively, the DOC or POC that is produced in the surface layer, or imported from rivers and submarine groundwater discharge, decomposes in the deep layer.As long as more CO 2 is generated in the deep layer than is taken up by the surface layer, the shelf as a whole will remain heterotrophic.

Future changes in carbon fluxes
Increasing air temperature (Belkin, 2009) tends to increase precipitation and continental runoff.These processes enhanced rock weathering during the last century (Probst et al., 1994).Intuitively, this fact suggests an increased export of carbon by rivers, but whether the global river runoff has increased is uncertain (Dai et al., 2011;Syed et al., 2010).The construction of dams around the world has caused a substan-tial fraction of exported sediment to be impounded in recent decades (Chen, 2002;Syvitski et al., 2005).A related issue is that global warming is warming the oceans as well.The global mean sea surface temperature has reportedly risen by 0.67 • C over the last century (IPCC, 2007;Trenberth et al., 2007).The most rapid warming, two to four times the global average of 0.177 • C per decade between 1981 and 2005, has been observed in the landlocked or semi-enclosed European and East Asian seas, including the Baltic, North, Black, Japan and East China seas as well as over the Newfoundland-Labrador Shelf (Belkin, 2009).The thermodynamics of seawater dictates that for each 1 • C rise in temperature, the pCO 2 increases by 4 %, or approximately 14 µatm.This fact would compensate for some of the increase in CO 2 in the atmosphere, which is of the order of 18 µatm per decade.With increasing atmospheric CO 2 concentration, the pCO 2 difference between the air and the shelf seawater will become larger.This is to the advantage of absorbing atmospheric CO 2 in coastal seas, and even some CO 2 -emitting regions may start to absorb CO 2 .A related issue is that the eutrophicated coastal area is growing due to human activities such as excessive nutrient inputs and enhanced soil erosion on land (Brush, 2009;Smith and Schindler, 2009).Values of pH in the coastal seawater will drop faster than in the open ocean because decomposition of terrestrial organic material increases the total alkalinity but reduces the buffering capacity (Chen et al., 1982;Cai et al., 2011).Further, certain species of phytoplankton may grow better in a high-CO 2 environment (Riebesell and Tortell, 2011), hence deterring the increasing trend of atmospheric CO 2 in general.These effects, however, are beyond the scope of this study.

Conclusions
Data from 165 estuaries and 87 continental shelves around the globe have been evaluated to show that the world's estuaries release 0.094 Pg C yr −1 to the atmosphere.This value is substantially lower than any published values mainly because of the newly available data in Asia, which has low wind speed in general, and hence low per unit area flux.In addition, new data in the polar regions indicate that estuaries there may absorb instead of release CO 2 .Overall, the world's continental shelves absorb 0.4 Pg C yr −1 from the atmosphere, which is in line with published data.

Fig. 3 .
Fig. 3. Histogram of reported daily CO 2 fluxes per unit area in different seasons and the annual flux of the world's estuaries.
Figure 4. Histogram of reported annual CO2 fluxes of world's estuaries in various latitude bands.

Fig. 4 .
Fig. 4. Histogram of reported annual CO 2 fluxes of the world's estuaries in various latitude bands.

Fig. 6 .
Fig. 6.Annual CO 2 flux (a), average CO 2 flux per unit area (b), total surface area (c), and percentage of total CO 2 flux (d) from estuaries in each continent.Numbers in parentheses indicate the number of estuaries studied.

Fig. 7 .
Figure 7. Annual CO2 flux (a), average CO2 flux per unit area (b), total surface area (c), and percentage of 2 total CO2 flux (d) from estuaries of each ocean.Numbers in parentheses indicate the number of 3 estuaries studied.4 5 6

Fig. 10 .
Fig. 10.Histogram of reported annual CO 2 fluxes of continental shelves in various latitude bands.

75
Figure 11.Annual CO2 flux (a), average CO2 flux per unit area (b), total surface area (c), and percentage of total CO2 flux (d) from continental shelves in different continents.Numbers in parentheses indicate the number of estuaries studied.

Fig. 11 .
Fig. 11.Annual CO 2 flux (a), average CO 2 flux per unit area (b), total surface area (c), and percentage of total CO 2 flux (d) from continental shelves in different continents.Numbers in parentheses indicate the number of estuaries studied.

Fig. 12 .
Fig. 12. Annual CO 2 flux (a), average CO 2 flux per unit area (b), total surface area (c), and percentage of total CO 2 flux (d) from continental shelves in different oceans.Numbers in parentheses indicate the number of estuaries studied.
Although efforts have been made to evaluate the parameters that affect the carbon cycle and to model past and future www

www.biogeosciences.net/10/6509/2013/ Biogeosciences, 10, 6509-6544, 2013Table 1 .
. However, there is still no consensus on the most suitable coefficient to use in Seasonal and annual sea-to-air fluxes of CO 2 in the world's estuaries.Summer flux Autumn flux Winter flux Annual flux References f

Table 1 .
Continued.Summer flux Autumn flux Winter flux Annual flux References f

Table 1 .
Continued.Spring flux c Summer flux Autumn flux Winter flux Annual flux References f

Table 2 .
The pCO 2 and flux method, the gas exchange coefficient and the wind speed in the world's estuaries.
Bold numbers are regions without data, and data from a similar region are given.EBC represents Eastern Boundary Current and WBC means Western Boundary Current.

Table 4 .
Summary of reported total sea-to-air fluxes of CO 2 in the world's estuaries.
c Austral seasons.d Not used in the calculation.e

Table 7 .
Summary of reported annual global air-sea CO 2 fluxes in the world's continental shelves.