Prevalence of strong vertical CO2 and O2 variability in the top meters of the ocean


  • Maria Ll. Calleja,

    Corresponding author
    1. Department of Global Change Research, Instituto Mediterráneo de Estudios Avanzados, CSIC-UIB, Esporles, Spain
    2. Instituto Andaluz de Ciencias de la Tierra, CSIC-UGR, Armilla, Spain
    • Corresponding author: M. Ll. Calleja, Instituto Andaluz de Ciencias de la Tierra, CSIC-UGR, Avda. de las Palmeras 4, 18100 Armilla, Granada, Spain. (

    Search for more papers by this author
  • Carlos M. Duarte,

    1. Department of Global Change Research, Instituto Mediterráneo de Estudios Avanzados, CSIC-UIB, Esporles, Spain
    2. The UWA Oceans Institute, University of Western Australia, Crawley, Western Australia, Australia
    Search for more papers by this author
  • Marta Álvarez,

    1. Instituto Español de Oceanografía, Centro de A Coruña, Coruña, Spain
    Search for more papers by this author
  • Raquel Vaquer-Sunyer,

    1. Department of Geology, Lund University, Lund, Sweden
    Search for more papers by this author
  • Susana Agustí,

    1. Department of Global Change Research, Instituto Mediterráneo de Estudios Avanzados, CSIC-UIB, Esporles, Spain
    2. The UWA Oceans Institute, University of Western Australia, Crawley, Western Australia, Australia
    Search for more papers by this author
  • Gerhard J. Herndl

    1. Department of Marine Biology, University of Vienna, Vienna, Austria
    2. Department of Biological Oceanography, Royal Netherlands Institute for Sea Research, Den Burg, Netherlands
    Search for more papers by this author


[1] The gradient in the partial pressure of carbon dioxide (pCO2) across the air-sea boundary layer is the main driving force for the air-sea CO2 flux. Global data bases for surface seawater pCO2 are actually based on pCO2 measurements from several meters below the sea surface, assuming a homogeneous distribution between the diffusive boundary layer and the upper top meters of the ocean. Compiling vertical profiles of pCO2, temperature, and dissolved oxygen in the upper 5–8 m of the ocean from different biogeographical areas, we detected a mean difference between the boundary layer and 5 m pCO2 of 13 ± 1 µatm. Temperature gradients accounted for only 11% of this pCO2 gradient in the top meters of the ocean; thus, pointing to a heterogeneous biological activity underneath the air-sea boundary layer as the main factor controlling the top meters pCO2 variability. Observations of pCO2 just beneath the air-sea boundary layer should be further investigated in order to estimate possible biases in calculating global air-sea CO2 fluxes.

1 Introduction

[2] The quantification and understanding of processes controlling the air-sea exchange of carbon dioxide (CO2) is critical on determining the global carbon budget and its future evolution on a global climate change scenario [Fung et al., 2005; Gruber et al., 2009; Le Quéré et al., 2010; McKinley et al., 2011]. The air-sea CO2 flux is mainly determined by changes in surface oceanic partial pressure of CO2 (pCO2), which varies spatially and temporally between 150 and 550 µatm [Takahashi et al., 2009], about 60% below and 30% above the mean atmospheric pCO2 of 394 µatm (annual mean for 2012, from Mauna Loa records).

[3] Over the last decade, major concerted efforts have been made to assemble a global surface water pCO2 data set. Currently, two data sets capturing seasonal and geographical variability have been recently released to the public: the LDEO database [Takahashi et al., 2012] with about 6.4 million measurements from 1957 to 2011 and the Surface Ocean CO2 Atlas (SOCAT) database [Pfeil et al., 2013] with 6.3 million measurements from 1968 to 2007. Both data sets are largely derived from automated underway systems onboard research vessels and ships of opportunity that are continuously recording the pCO2 in the overlying marine air and subsurface waters. These systems commonly sample water from 3 to 7 m below the surface, the depth of the intake on the vessel haul [e.g., Corbière et al., 2007; Fung and Takahashi, 2000; Inoue et al., 1996; Lüger et al., 2004; Metzl et al., 1995; Murphy et al., 2001; Santana-Casiano et al., 2009], and consider these measurements to be representative of that at the seawater surface by assuming a homogeneous distribution of gases below the air-sea diffusive boundary layer. Thus, the air-sea pCO2 gradient (∆pCO2), which drives the air-sea CO2 flux, is calculated assuming that pCO2 is vertically homogeneous within the upper meters of the ocean. However, there is published but limited evidence of pCO2 vertical variability between the surface microlayer and the bulk subsurface mixed layer in both lakes [Hari et al., 2008] and oceans [Gong et al., 2007; Liu et al. 2008].

[4] Previous examinations of near-surface vertical gradients of CO2 partial pressure have focused on temperature gradients [Robertson and Watson, 1992; Ward et al., 2004] and air-sea heat fluxes [Phillips, 2004], revealing thermal stratification within the upper meters of the ocean. The presence of a “warm layer skin” [Fairall et al., 1996] is commonly observed at low latitudes and low wind speed and mixing regimes. Also a “cool layer skin” [Donlon et al., 2002] is persistent on a global basis at wind speeds exceeding 6 m s−1. This small-scale temperature stratification in the upper meters of the ocean already complicates the direct measurements of surface water pCO2, as the solubility of CO2 in seawater is strongly temperature dependent [Weiss, 1974; Takahashi et al., 1993]. Accordingly, Ward et al. [2004] demonstrated that anomalies in air-sea CO2 fluxes result, at low wind conditions, from variation in the thermal structure of the upper few meters of the ocean. Other studies have also reported significant near-surface temperature gradients even under high wind speed [Gemmrich and Farmer, 1999], challenging the homogeneity expected from vertical mixing of the boundary layer. Takahashi and co-workers [Takahashi et al., 2009] recently considered the thermal skin effect on pCO2 to be negligibly small. However, they also recognized the need to examine this and other surface layer effects in order to reduce systematic errors in air-sea pCO2 differences and consequently in global air-sea CO2 flux estimates.

[5] In addition to physical processes, seawater pCO2 in the upper top meters of the ocean has been recently shown to be affected by nonconservative biological processes [Calleja et al., 2005; Ducklow and McCallister, 2004], and it is well known that the key biological process affecting seawater pCO2 is the net community production, the net balance between photosynthetic organic carbon fixation, which reduces the partial pressure of CO2, and respiration of organic matter, which releases CO2. Hence, pCO2 vertical variability in the upper meters of the surface layer could also be driven by vertical variability in the net planktonic community metabolism.

[6] During the past decade, significant evidence has been provided pointing to different biochemical and microbial activity and composition in the surface and subsurface layer of different oceanic regions [Cunliffe et al., 2013, and references therein]. For example, differences in bacterial diversity (North Sea) [Franklin et al., 2005], microbial strains (NW Mediterranean) [Agogué et al., 2005], and autotrophic and heterotrophic nanoflagellate abundance (NW Mediterranean) [Joux et al., 2006] between surface microlayer and subsurface waters have been reported. These differences in biological diversity and abundance may be related to particulate and dissolved fractions of macronutrients or other biochemically active compounds. For example, Obernosterer et al. [2005] identified consistently greater concentrations of particulate organic carbon and nitrogen and of dissolved organic carbon in South Pacific surface microlayer water compared with surface water (5 m). Furthermore, other studies conducted during the 2003 and 2004 cruises where this work was carried out reported significantly higher microbial metabolic rates in the surface microlayer than in the subsurface waters [Calleja et al., 2005; Reinthaler et al., 2008], which lead to significant differences in CO2 partial pressure between the two layers and exert a control over the air-sea CO2 flux [Calleja et al., 2005].

[7] Despite evidences, and due to logistics problems, the analysis of the relationship between the evidenced vertical biological heterogeneity and pCO2 gradients within the upper meters of the water column are scarce, particularly for the open ocean comprising the vast majority of the oceans' surface.

[8] We believe that as the accuracy and precision of surface pCO2 measurements is increasing [Körtzinger et al., 2000], the thermodynamic and biological effects on pCO2 variability within the upper meters of the surface ocean become even greater potential sources of uncertainty, which can no longer be ignored to accurately assess global air-sea CO2 fluxes.

[9] In this work, we examine the vertical variability of pCO2, partial pressure of oxygen (pO2), and temperature in the upper 5 to 8 m of the ocean from 83 profiles collected between 2003 and 2007 and covering different oceanic biomes. The general aim is to study the vertical homogeneity of pCO2 in the upper surface meters and also separating thermodynamic effects from biological effects. While changes in temperature lead to parallel changes in pCO2 and pO2, since the solubility of both gases is a strong inverse function of seawater temperature, metabolic processes lead to reverse changes in these gases, as a positive net production causes a pCO2 decrease and a pO2 increase, while a net negative production (net respiration) leads to higher pCO2 and lower pO2.

2 Sampling and Methods

2.1 Location

[10] Our measurements were conducted at 83 stations in four different regions (Figure 1): along the NE subtropical Atlantic Ocean (between 14°W and 32°W, and 19°N and 28°N) during May–June 2003 (cruise COCA-2 on board the R/V Hesperides), September–October 2004 (cruise BADE-2 on board the R/V Pelagia), August–September 2006 (cruise RODA-1 on board the R/V Hesperides), and February 2007 (cruise RODA-2 on board the R/V Hesperides); the Southern Ocean (between 55°W and 70°W, and 62°S and 67°S) during February 2005 (cruise ICEPOS-2 on board of the R/V Hesperides); the Mediterranean Sea and Black Sea (between 7°E and 31°E, and 35°N and 42°N), during June–July 2006 (cruise THRESHOLDS on board the R/V García del Cid); and the Arctic Ocean (between 20°E and 14°W, and 68°N and 81°N) during June–July 2007 (cruise ATOS on board the R/V Hesperides).

Figure 1.

Map showing the location of the stations conducted in different cruises where profiles for pCO2, O2 concentration, and temperature were examined within the upper 5 m of the ocean.

2.2 Instrumentation

[11] Vertical profile measurements of CO2 and O2 between the top centimeters and 5–8 m depth were performed at 83 open ocean stations from a small pneumatic boat drifting away from any possible contamination source from the research vessel, which could be safely deployed at wind velocities up to 15.7 m s−1.

[12] Determination of water pCO2 was performed using a high-precision (± 1 ppm) nondispersive infrared gas analyzer (EGM-4, PP-systems) fitted with an electrochemical cell Oxygen probe (OP-2, PP-systems, precision ± 0.02%), averaging measurements at 1 min recording interval. The closed gas stream flowing through the gas (CO2 and O2) analyzer was previously equilibrated with the sampled seawater by the use of a peristaltic pump which introduced the seawater into a gas exchange column (MiniModule 1.25 × 9 Membrane Contactor, Celgard) with an effective surface area of 0.5 m2, total volume of 52 ml, and water flow of about 300 ml min−1, resulting in a residence seawater time of 10 s. Temperature was measured (RTD probe, Fluke, with accuracy and precision of 0.01°C), before (in situ seawater Temperature) and after flowing through the equilibrator, and no temperature difference was detected. Before entering the gas analyzer, the gas stream was circulated through a Calcium Sulfate column to avoid interferences from water vapor.

[13] The gas analyzer was calibrated, in all the cruises, using two dry standards: pure nitrogen (0.0 ppm CO2) and a gas mixture of CO2 and N2 containing a CO2 molar fraction of 541 ppm, from Carburos Metalicos (Barcelona, Spain), which revealed an accuracy of ± 2 ppm in the determinations of pCO2 measurements. Pure nitrogen gas (0.0% O2) and pure synthetic air (21.00% O2) were used for calibration of the Oxygen probe revealing an accuracy of 0.2% in the determinations of pO2 measurements. Final calculations for pCO2 and pO2 were obtained at 1 atmospheric pressure with 100% saturation of water vapor and in situ temperature [Weiss, 1974]. Oxygen concentrations (µmol O2 Kg−1) were calculated from pO2 (%), temperature, and salinity according to Benson and Krause [1984]. O2 concentrations obtained strongly and significantly correlate (R2 = 0.8, P < 0.05) with those measured by the use of the traditional Winkler method (12 parallel measurements, using both methods, were performed during the THRESHOLDS cruise). The results obtained with our method tended to overestimate oxygen concentrations (by, on average, 9 ± 3 µmol O2 Kg−1) than those obtained by the Winkler method.

[14] The intake of the peristaltic pump was submerged and held in position by a floating device, to measure gases in the top centimeters of the surface. The pump inlet was then progressively lowered allowing gases to equilibrate at each discrete depth where measurements were taken.

[15] Water temperature and salinity profiles from the upper meters were obtained using an Eureka Manta® multiprobe data-logger recording depth and temperature with resolutions of 0.01 m and 0.01°C, respectively, at THRESHOLDS, RODA-1, and RODA-2 cruises. During COCA-2 and ICEPOS-2, temperature profiles could not be obtained, and only discrete measurements at the top centimeters and 5 m depth were available, which were performed within a minute in 1 L flasks with water pumped from these two depths. During BADE-2, temperature profiles, with a resolution of 1 m, were obtained from conductivity, temperature, depth (CTD) casts performed from the research vessel at the same time as gas profiles were obtained from the small boat. During ATOS, a Seabird 19 CTD was manually deployed from the small boat and temperature data averaged at 0.5 m intervals.

[16] Wind speed was measured at 1 min intervals using the research vessel equipment: an Aanderaa meteorological station at COCA-2, RODA-1, RODA-2, ICEPOS-2, and ATOS cruises; a Davis Vantage Pro meteorological station at THRESHOLDS cruise; and a Royal Netherlands Meteorological Institute meteo system at BADE-2. Wind speed data were averaged over 60 min prior to the measurements until the end of the sampling period.

[17] Pitch, roll, and heading of the research vessel were also recorded at 1 min intervals and used in a routine embedded in the software integrating navigation and meteorological data to correct wind speed for the ship movement and flow distortion. The corrected wind velocities were then converted to wind at 10 m (U10) using the logarithmic correction U10 = Uz [0.097 ln (z/10) + 1]−1 where z is the height of the wind sensor position [Hartman and Hammond, 1985].

[18] Additional water samples for Chlorophyll a (Chl a) were collected, for all cruises, using a Rosette sampler system at 5 m depth. Water samples of 150 mL were filtered through 25 mm Whatmann GF/F filters from each station. Collected filters were placed in tubes with a 90% acetone solution for 24 h and extracted Chl a was analyzed spectrofluorimetrically on a Shimadzu RF-5301 PC spectrofluorimeter calibrated with pure Chl a [Parsons et al., 1984]. Surface Chl a concentrations were used as a proxy to define the productivity of the different studied areas.

3 Results and Discussion

3.1 Geographical Distribution

[19] Our measurements encompassed a wide range of oceanographic conditions, extending from high-latitude polar regions to lower latitude subtropical waters (see Figure 1). From nutrient-rich and highly productive waters in the Southern and Arctic Oceans (4.7 and 5.6 µg Chl a L−1, respectively) to subtropical NE Atlantic Ocean, where conditions ranged from very oligotrophic gyre waters (< 0.1 µg Chl a L−1) to productive NW African coast upwelling waters (3.5 µg Chl a L−1). Samples from the Mediterranean Sea ranged from 0.1 µg Chl a L−1 to 4.2 µg Chl a L−1. Overall, sampled surface waters ranged from strongly undersaturated in CO2 (134 µatm) in the Arctic to highly supersaturated (495 µatm) in the eastern Mediterranean (Black Sea), relative to atmospheric CO2 concentrations (Table 1). Oxygen concentrations ranged from 221.7 to 343.6 µmol O2 Kg–1, both observed in the Subtropical Atlantic (Table 1). Water surface Temperature ranged between −1.24°C in polar waters and 27.08°C in subtropical waters (Table 1). Wind speeds, at the time of sampling, ranged from very gentle conditions (0.8 m s−1) to high winds (15.7 m s−1) across the study. Higher wind velocities were reported at the Southern Ocean (mean ± SE, 7.5 ± 1.6 m s−1) and at the NE Atlantic (6.3 ± 0.4 m s−1), whereas the Arctic Ocean and the Mediterranean Sea presented intermediate (5.7 ± 0.5 m s−1) and low (3.6 ± 0.7 m s−1) wind regimes, respectively.

Table 1. Number (N) of Observations and Sampling Stations for Each Cruise and Mean ± Standard Error (SE), Minimum (Min) and Maximum (Max) Values Measured for Temperature (T,°C), pCO2 (µatm), and O2 Concentration (µmol Kg−1) Within the Top 5 m of the Ocean Surface for Each Cruise
 NE Subtropical AtlanticSouthern OceanMediterranean SeaArctic Ocean
N observations7410227820130116
N stations85413102320
Mean T (°C)26.4622.5324.0020.631.5221.171.76
Mean pCO2 (µatm)386347412348327410181
Mean O2 (µmol Kg−1)242.2299.1261.1254.2241.9263.5261.3

[20] Temperature differences observed between the top centimeters and 5 meters depth were relatively small but above the instrumental error (0.22 ± 0.05°C). However, it was particularly variable in the cruises conducted where lower wind regimes were reported; in the Arctic, where a thin layer of ice melt waters was confined within a near-surface (1–2 m) pycnocline leading to observations with a cold top centimeters layer, and in the Mediterranean Sea, where the upper meters where stratified in the opposite direction (Figure 2a). Overall, in 60% of the data collected, the top centimeters of the ocean were warmer than the waters a few meters below.

Figure 2.

Box plots showing the distribution of the differences between the top centimeters and 5–8 m depth of temperature (°C), pCO2 (µatm), and oxygen concentration (µmol Kg−1) observed in all different studied regions: North East Subtropical Atlantic (NE Atl), Mediterranean Sea (Med), Southern Ocean (SO), and Arctic Ocean (Arctic). All data from all regions are included. The vertical line in the boxes represents the median, the boxes extend from the 25% to the 75% quartile of the data distribution, and the horizontal lines indicate the 5% and 95% quartiles.

[21] The range of variability of the difference between the pCO2 at the top centimeters and at 5 m was quite substantial for pCO2 (12.6 ± 1.4 µatm) and the maximum vertical difference reached 63 µatm (Figure 2b). Oxygen concentration within this layer also showed considerable vertical variability within stations (6.7 ± 1.0 µmol O2 Kg−1), and the maximum difference observed at any one station was −67.3 µmol O2 Kg−1 (Figure 2c). In polar waters, the most productive areas during spring-summer, when sampling was performed, a clear relationship between vertical changes in O2 and pCO2 was observed (Figures 2b and 2c). Southern waters presented generally lower pCO2 values at the top centimeters than at 5–8 m below. Accordingly, the upper centimeters of the water column were generally higher in O2 evoking community metabolism to be exerting a strong control on vertical CO2 variability among those meters and suggesting a net enrichment in microautotrophs in the very top of the water column fixing CO2 at higher rates than those from a few meters below in Antarctic waters. On the contrary, we observed the Arctic Ocean to show a general increase in pCO2 and decrease in O2 concentrations in the top centimeters when compared to those from a few meters below, suggesting a heterotrophic enrichment activity in the top centimeters in those waters.

[22] Among the overall data set, pCO2 and oxygen concentration values differed systematically, but not consistently, between the top centimeters and 5 m depth, showing almost all possible profiles (see Figure 3 for some examples, and Figure S1 in the supporting information for all profiles) revealing either metabolic effects driving pCO2 vertical variability with nonsignificant changes in temperature in the NE Atlantic and Southern Ocean (Figures 3a, 3b and 3e, 3f, respectively), or the biological and thermodynamic effect driving pCO2 changes towards the same direction in the Mediterranean Sea and the Arctic Ocean (Figures 3c, 3d and 3g, 3h respectively). Note that those are just examples and do not represent general patterns of any of the studied areas.

Figure 3.

Examples of observed pCO2 (solid circles), O2 concentration (open circles), and temperature (solid triangles) vertical profiles within the top meters of the ocean. Wind speed at the time of sampling is shown in insert. n.d. stands for no data.

3.2 Controls of pCO2 Vertical Variability in the Upper Meters of the Ocean

[23] On a global scale, and considering all data collected, pCO2 correlated strongly with water temperature. The dependency of pCO2 with water temperature (pCO2 T) is described by the following equation: pCO2 T (µatm) = 190 (±4) + 8.77 (±0.23) T (°C) [R2 = 0.77, P < 0.0001, see Figure 4], with pCO2 values doubling for every 21°C temperature increase (∂ln pCO2/∂T = 0.0324°C−1). This relationship exhibits less increase than that established by Takahashi et al. [1993] for isochemical seawater conditions, where pCO2 values double every 16°C (∂ln pCO2/∂T = 0.0423°C−1), meaning that the sum of other effects is counteracting the pCO2 increase expected by T changes alone. In contrast, oxygen concentrations were independent (P > 0.05) of temperature changes. We observed that pCO2 strongly correlated with absolute latitude (R2 = 0.70, P < 0.0001) presenting lower values at polar productive cold waters and higher values at subtropical oligotrophic warm waters.

Figure 4.

Relationship between seawater pCO2 (µatm) and temperature (°C), for the whole data set: Southern Ocean (+), Arctic Ocean (x), North East Atlantic Ocean (○), and Mediterranean Sea (◊). The solid line represents the fitted linear regression equation pCO2,T (µatm) = 190 (±4) + 8.77 (±0.23) T (°C), R2 = 0.77, P < 0.0001. The dashed line represents de mean value of atmospheric pCO2 for the year 2005 (from Mauna Loa records), the middle year for our data set range (2003–2007).

[24] To evaluate the biological effect on changes in pCO2, we removed the temperature effect by normalizing the pCO2 observed values (pCO2 obs) to a constant temperature of 15.9°C, the mean surface temperature for our data set (Tmean), using the equation proposed by Takahashi et al. [1993, 2002]: pCO2 (Tmean) = pCO2obs e0.0423 (Tmean-Tobs), where pCO2 (Tmean) is the pCO2 normalized to the mean surface Temperature (Tmean), and Tobs is the temperature observed or measured in situ. The dependency of pCO2 (Tmean) with changes in O2 concentration was then analyzed for the different studied areas. We found pCO2 (Tmean) to be significantly and negatively correlated with changes in O2 concentration when considering all data from polar latitudes (from the Arctic and Southern Oceans) and the Mediterranean Sea [pCO2 (Tmean) (µatm) = 1771 (±131) − 5.5 (±0.5) O2 (µmol Kg−1), R2 = 0.31, P < 0.0001, N = 266] (Figure 5). The changes in oxygen concentration explained more than 30% of the temperature-normalized pCO2. When considering only polar data (exceeding 60°N and 60°S), a similar and even stronger dependency is found [pCO2 (Tmean) (µatm) = 2374 (±190) −7.8 (±0.7) O2 (µmol Kg−1), R2 = 0.46, P < 0.0001, N = 136]. As expected, higher oxygen values were associated with lower pCO2 values than those expected by temperature, due to higher photosynthetic rates, while lower oxygen values are associated with increased pCO2, reflecting higher net community respiration rates. The magnitude of that effect is amplified at polar latitudes, where an increase in oxygen of 20 µmol Kg−1 was associated with a drawdown of 150 µatm in surface pCO2, whereas in the Mediterranean Sea, the same oxygen change was associated with a biological pCO2 drawdown of 14 µatm, hence suggesting a lower capacity for biological CO2 drawdown at warm subtropical latitudes than at high polar latitudes. Similar effects were observed by Takahashi and co-workers [1993, 2002], who postulated biological effects on oceanic pCO2 at high latitudes to exceed 140 µatm and to be around 50 µatm in subtropical and tropical latitudes; however, data from the Mediterranean Sea was lacking in their data set.

Figure 5.

Relationship between seawater pCO2 normalized to the mean temperature of 15.9°C (pCO2 (Tmean) (µatm)) and oxygen concentration (µmol Kg−1), considering all data from the Southern Ocean (+), Arctic Ocean (x), and Mediterranean Sea (◊). The solid line represents the fitted linear regression equation pCO2 (Tmean) (µatm) = 1771 (±131) − 5.5 (±0.5) O2 (µmol Kg−1), R2 = 0.31, P < 0.0001.

[25] On the other hand, data collected in the North East Subtropical Atlantic presented lower pCO2 (Tmean) and they do not significantly correlate with oxygen concentration, probably due to low primary production in these nutrient-depleted waters. However, pCO2 (Tmean) did positively and significantly correlate with distance to the coast, represented by longitude (°W), [R2 = 0.62, P < 0.0001, N = 158] exhibiting an increase in sea surface pCO2 magnitude and variability towards the North West African coast due to the monsoon-induced upwelling of deep waters in this coastal area.

[26] The role of temperature variability on the observed vertical variability in pCO2 and O2 concentration was further investigated by taking pCO2 at 5 m depth (or the highest depth sampled, between 3 and 8 m, when 5 m depth data was not recorded) and calculating the corresponding values expected at other depths from the observed Tobs. That was done using the corresponding temperature dependence equation for pCO2 (δ ln pCO2/δT = 0.0423°C, pCO2 (expected at Tobs) = pCO2 (at 5 m) e0.0423 (Tobs-T5m)) [Takahashi et al., 1993]. Vertical variability in O2 concentration expected by the observed temperature was also investigated by using the corresponding temperature dependence equation for O2 [Benson and Krause, 1984]. The observed variability of temperature-corrected pCO2 and temperature-corrected O2 concentration within each profile averaged 10 ± 1 µatm of CO2 and 5.1 ± 0.8 µmol O2 Kg−1, both significantly greater than 0 (t test, P < 0.001), confirming temperature-independent variability in pCO2 and O2 concentration within the upper meters of the ocean.

[27] Temperature variability within the upper 5 m accounted for (mean ± SE) 11 ± 3% and 9 ± 2% of the variance in pCO2 and O2 across stations, respectively, indicating that most of the vertical gas variability was caused by changes other than those in temperature. There was no significant relationship between the changes in wind velocity and the pCO2 variability within the upper 5 m across stations (P > 0.05). Moreover, substantial variability in pCO2 and O2 within the top 5 m was observed even at moderate to high wind velocities (see Figures 3a, 3b, 3e, and 3f).

[28] To confirm that temperature is not the main process controlling pCO2 vertical variability within the upper meters of the ocean, the relationship between the vertical pCO2 and O2 concentration anomalies, as the deviations from the values expected from temperature differences, were examined as a proxy to test the possible metabolic basis of pCO2 vertical variability within the top meters. That was done by analyzing the differences between the observed pCO2 and O2 concentration changes with depth [ΔpCO2 observed/Δdepth and ΔO2 observed/Δdepth], and the temperature predicted pCO2 and O2 concentration changes with depth [ΔpCO2 T/Δdepth and ΔO2 T/Δdepth], where pCO2 T and ΔO2 T are calculated by using Takahashi et al. [1993] and Benson and Krause [1984] equations, respectively. Vertical pCO2 anomalies (in µatm m−1) are then calculated as [ΔpCO2 observed/Δdepth] − [ΔpCO2 T/Δdepth] and O2 anomalies (in µmol Kg−1 m−1) are calculated as [ΔO2 observed/Δdepth] − [ΔO2 T/Δdepth]. The pCO2 and O2 anomalies negatively and significantly correlated (R2 = 0.21, P < 0.001) (see solid line in Figure 6) when analyzing all profiles together. We found this correlation between pCO2 and O2 anomalies to be similar for polar waters (R2 = 0.17, P < 0.05) and much stronger in the Mediterranean Sea, when excluding the Black Sea, (R2 = 0.52, P < 0.001) (see dashed line in Figure 6). However, the vertical O2 concentration range in the measured profiles of the Mediterranean Sea shows smaller values (3.4 ± 0.5 µmol O2 Kg−1) than those observed at high-latitude polar waters (7.3 ± 1.5 µmol O2 Kg−1). Altogether, our data display evidence of metabolic changes occurring in the top meters of the ocean that significantly control the vertical anomalies in CO2 observed within this surface layer, explaining more than 50% of the pCO2 anomalies observed in the Mediterranean Sea. On the other hand, data collected on the North East subtropical Atlantic presents lower pCO2 vertical anomalies than those in the Mediterranean and polar waters and they do not significantly correlate with oxygen anomalies.

Figure 6.

Relationship between vertical pCO2 and O2 concentration anomalies (in µatm m−1 and µmol Kg−1 m−1, respectively) relative to temperature changes observed within the upper 5 m of the surface ocean. The solid line represents the fitted linear regression equation pCO2 anomalies (µatm m−1) = −0.02 (±0.51) − 1.05 (±0.24) O2 anomalies (µmol Kg−1 m−1), R2 = 0.21, P < 0.0001, corresponding to all data set correlation. The dashed line represents the fitted linear regression equation pCO2 anomalies (µatm m−1) = 1.29 (±0.65) − 2.9 (±0.66) × O2 anomalies (µmol Kg−1 m−1), R2 = 0.52, P < 0.0001, corresponding to data from the Mediterranean Sea.

[29] Overall, the small temperature vertical changes observed within this study were comparable to those reported in the past [Donlon et al., 2002; Fairall et al., 1996; Gemmrich and Farmer, 1999; Ward et al., 2004] and accounted for less than 20% of the observed vertical pCO2 variability. This suggests the strong influence of biological activity contributing to changes in pCO2 within the upper meters of the surface ocean, which is supported by the biological and chemical data available from three of the same cruises where this study was conducted (COCA-2 [Calleja et al., 2005, 2009], BADE-2 [Reinthaler et al., 2008], and ICEPOS [Calleja et al., 2009]). This data, shown in Table 2, revealed that the surface microlayer was significantly enhanced in microbial respiration rates (by a factor of 3 to 9) and dissolved organic carbon (DOC) concentrations (by 25–45%). A consistent and pronounced enrichment was also observed on the concentration of Dissolved Organic Nitrogen (of a factor of more than 3) and of total hydrolizable amino acids (THAA), a major carbon and energy sources for marine microbes [Keil and Kirchman, 1992] that showed differences of one order of magnitude within 1 m of the water surface [Reinthaler et al., 2008].

Table 2. Summary of Selected Biological and Chemical Parameters (Average ± SE): Bacterial Respiration (BR) in µmol O2 L−1 d−1, Dissolved Organic Nitrogen (DON) in µmol L−1, Total Hydrolizable Amino Acid Concentration (THAA) in µmol L−1, and Dissolved Organic Carbon (DOC) in µmol L−1, in the Surface Microlayer (SML) and the Underwater Layer (UWL), for two of the Studied Geographical Regionsa
Geographical RegionBR(µmol O2 L−1 d−1)DON(µmol L−1)THAA(µmol L−1)DOC(µmol L−1)
  1. aSML in Reinthaler et al. [2008] represents the top 40–80 µm surface layer, while SML in Calleja et al. [2005, 2009] represents the top 2 cm layer. UWL in Reinthaler et al. [2008] represents water sampled at around 30–50 m depth, while ULW in Calleja et al. [2005, 2009] represents water sampled from 5 m depth.
NE Subtropical Atlantic        
COCA-2 [Calleja et al., 2005, 2009]6.7 ± 1.92.2 ± 0.9    110 ± 576 ± 6
BADE-2 [Reinthaler et al., 2008]7.5 ± 1.60.8 ± 0.818.3 ± 1.95.5 0.25.45 ± 1.430.45 ± 0.02161 ± 24111 ± 17
Southern Ocean        
ICEPOS [Calleja et al., 2009]      70 ± 455 ± 2

[30] The observed prevalence of negative relationships between the deviations of pCO2 and O2 concentration from the values expected from temperature variability further suggest biological processes as key drivers of the observed pCO2 heterogeneity. These results are in accordance with the mentioned data showing that the top centimeters of the ocean support metabolic rates far greater than the waters at 5 m depth [Calleja et al., 2005; Reinthaler et al., 2008] and directly affect air-sea pCO2 gradients. The evidence of strong biologically driven pCO2 and O2 concentration changes within the upper meters of the ocean suggest that metabolic effects on pCO2 changes could be often faster than mixing time scales, reported to vary between 0.5 h and hundreds of hours [Denman and Gargett, 1983], hence allowing these heterogeneities to persist in the presence of mixing.

4 Concluding Remarks

[31] Our results show the prevalence of substantial variability in pCO2 (averaging 12.6 ± 1.4 µatm and reaching a maximum value of 63 µatm) within the upper several meters of the open ocean in all the areas sampled. This contrasts with the assumed homogeneity in pCO2 within the top meters of the ocean, implicit in conventional survey programs [Fung and Takahashi, 2000; Takahashi et al., 2002, 2009] and inherent to most applications, as routine and conventional sampling instruments (e.g., CTD-Rosette sampling systems) cannot properly resolve the top meter of the ocean, where sensors are affected by ship motion. The assumed homogeneity of pCO2 within the upper several meters of the open ocean has led international efforts to improve CO2 flux estimates to target improvement in the precision and accuracy of analytical systems [Körtzinger et al., 2000], quality control of the integrated data bases [Pfeil et al., 2013], and parameterization of the gas exchange coefficient [Wanninkhof et al., 2009] as key actions. However, our results provide ample evidence for the prevalence of substantial heterogeneity in pCO2 within the top meters of the ocean. The evidence presented here should prompt efforts to resolve and understand departures from the assumption of pCO2 vertical homogeneity within that surface layer, as it can constitute a significant source of error in both the magnitude and the direction of air-sea CO2 flux estimates. By adding the mean difference between the surface water pCO2 at 0 and 5 m from our measurements to the Takahashi's climatology air-sea CO2 gradient, referred to the year 2000 [Takahashi et al., 2009], we have calculated the new net air-sea flux at the overlapping grid points and months (see Table S1 in the supporting information for a detailed explanation of the calculations). The results estimate that the Arctic Ocean would be a stronger CO2 sink (up to 3.5 times higher) and the Southern Ocean would be a more moderate CO2 sink than that predicted by Takahashi. On the other hand, the North East subtropical Atlantic does not seem to show a clear trend (see Table S1).

[32] In summary, the data presented here clearly indicates that the top meters of the ocean is an area far more dynamic and heterogeneous in metabolic gases than hitherto considered. This heterogeneity and the combination of physical and biological processes driving it should be further investigated to adequately improve our understanding of air-sea CO2 exchange and accurately assess the global oceanic uptake of atmospheric CO2.


[33] This research was funded by the projects COCA (REN-2000-1471-C02), RODA (CTM-2004-06842-CO3-O2), BADE (REN-2002-12284E/MAR, NWO-ALW 812.03.001), ICEPOS (REN2002-04165-CO3-O2), THRESHOLDS (CTM2005-24238-E/MAR), ATOS (POL2006-00550/CTM), the MALASPINA 2010 Expedition (CSD2008-00077), R + D National Plan of the Spanish Ministry of Science and Innovation, and the Dutch Earth and Life Sciences. We thank the crew of the R/V Hespérides, R/V Pelagia, R/V García del Cid, and the UTM personnel for their assistance. M.Ll.C was supported by the Spanish Research council [CSIC, grant JAEDOC030, co-funded by the Fondo Social Europeo (FSO)] and the Agency for Administration of University and Research Grants (AGAUR, grant 2007 BP-A 00156). M.A. was funded by grant ORCASEX (RYC-2006-001836). R.V.-S. was funded by THRESHOLDS integrated project (003933–2).