Extensive hydrogen supersaturations in the western South Atlantic Ocean suggest substantial underestimation of nitrogen fixation

Abstract

The nitrogen cycle is fundamental to Earth's biogeochemistry. Yet major uncertainties of quantification remain, particularly regarding the global oceanic nitrogen fixation rate. Hydrogen is produced during nitrogen fixation and will become supersaturated in surface waters if there is net release from diazotrophs. Ocean surveys of hydrogen supersaturation thus have the potential to illustrate the spatial and temporal distribution of nitrogen fixation and to guide the far more onerous but quantitative methods for measuring it. Here we present the first transect of high resolution measurements of hydrogen supersaturations in surface waters along a meridional 10,000 km cruise track through the Atlantic. We compare measured saturations with published measurements of nitrogen fixation rates and also with model-derived values. If the primary source of excess hydrogen is nitrogen fixation and has a hydrogen release ratio similar to Trichodesmium, our hydrogen measurements would point to similar rates of fixation in the North and South Atlantic, roughly consistent with modeled fixation rates but not with measured rates, which are lower in the south. Possible explanations would include any substantial nitrogen fixation by newly discovered diazotrophs, particularly any having a hydrogen release ratio similar to or exceeding that of Trichodesmium; undersampling of nitrogen fixation south of the equator related to excessive focus on Trichodesmium; and methodological shortcomings of nitrogen fixation techniques that cause a bias toward colonial diazotrophs relative to unicellular forms. Alternatively, our data are affected by an unknown hydrogen source that is greater in the southern half of the cruise track than the northern.

Key Points

• Widespread H₂ supersaturations reported in Atlantic Ocean between 30°N and 40°S
• Supersaturations are similar N and S, suggesting N₂ fixation S is similar to N of equator
• Measured N₂ fixation is low S of equator, suggesting gaps in sampling or a novel source

1 Introduction

Hydrogen is rather uniformly distributed in the atmosphere at mixing ratios around 0.5 ppm in the northern hemisphere and 0.52 in the southern [Simmonds et al., 2000]. Its sources include biomass burning and atmospheric photolysis of formaldehyde, as part of the oxidation sequence of methane, and its primary sink is uptake by soils (indicative of bacterial consumption) [Rhee et al., 2006]. Were there no substantial marine sources or sinks, hydrogen concentrations in the surface waters would be around 0.3 nmol/L. But nitrogen fixation, a key process in the ocean that over various time scales brings into balance losses of combined nitrogen from denitrification [Karl et al., 2002] and anammox, has long been known to produce hydrogen [e.g., Hadfield and Bulen, 1969]. If hydrogen were released stoichiometrically with fixation (equation 1), the instantaneous production of 1 nM ammonia would result in a 147% supersaturation of hydrogen at 25°C (i.e., [H₂] about 2.5 times atmospheric equilibrium).

(1)

where Pi is inorganic phosphate [Wilson et al., 2010a]. It follows that surface waters in oceans regions where nitrogen fixation is occurring would be expected to be supersaturated in hydrogen unless there are consumption processes in the ocean that exactly and instantaneously balance the hydrogen produced from nitrogen fixation. Currently there is no means of rapidly detecting nitrogen fixation in the oceans but hydrogen is easily analyzed at subnanomolar concentrations, so it would seem that surveys of saturation levels might offer a way of at least qualitatively revealing the pattern of nitrogen fixation in the ocean.

Few studies have yet been made of the net release of hydrogen by marine diazotrophs and those that have indicated a range of release rates varying widely from a high of about 0.3 mol of net H₂ production per mol of N₂ fixed for the relatively well-studied diazotroph, Trichodesmium, to yields 1 and 2 orders of magnitude less for the unicellular Cyanothece sp. and Crocosphaera watsonii, possibly due to efficient recycling of hydrogen as a source of energy [Wilson et al., 2010a]. While there are no reports of hydrogen release by endosymbiotic diazotrophs such as Richelia intracellularis, such diazotrophs are less likely to contribute to net hydrogen release to the host's environment as there are two organisms that might utilize it as an energy source. No fully quantitative relationship can be expected between nitrogen fixation rates and supersaturation on account of such variations in release rates driven by hydrogen recycling by diazotrophs, as well as utilization by other bacteria as an energy source, and the dependence of hydrogen concentrations on the ventilation rate of the mixed layer, a process itself dependent on mixed layer depth and wind speed. Nonetheless, it seems reasonable to suggest that nitrogen fixation will lead to supersaturation of hydrogen and that observed supersaturations of hydrogen might be used to calculate lower limits on underlying rates of nitrogen fixation.

Other marine sources of hydrogen that must be considered in accounting for any observed supersaturations are fermentation, ocean vents, photochemical production, microbial oxidation of phosphite, and contamination. Of this list, fermentation can be dismissed in oxygenated surface waters of the deep ocean. The influence of deep sea vents does not extend high into an oxygenated water column on account of microbial oxidation of hydrogen as demonstrated by many deep water profiles [e.g., Schropp et al., 1987; Moore et al., 2009b] that show extremely low concentration beneath a depth of a few hundred meters. Karl [2014] refers to the possible release of hydrogen during microbial oxidation of phosphite [Yang and Metcalf, 2004], but points out that there are no estimates of phosphite concentrations or fluxes in the ocean, so at present this potential source is very speculative. While hydrogen can be produced by cyanobacterial bidirectional hydrogenase, Bothe et al. [2010a] report that the enzyme appears not to be present in open ocean cyanobacteria. We will discuss the other possible hydrogen sources: nitrogen fixation, photochemistry, and contamination.

A series of measurements of dissolved hydrogen in the equatorial Pacific [R. M. Moore et al., 2009b] supported a relationship between observed supersaturations and nitrogen fixation activity measured by ¹⁵N₂-uptake incubations. That study has motivated both a streamlining of the hydrogen analysis to permit semicontinuous measurements and a longer, more detailed survey across 10,000 km in the Atlantic Ocean as part of the Atlantic Meridional Transect programme, the results of which are presented here.

2 Methods

Dissolved hydrogen was measured between Southampton, UK, and Punta Arenas, Chile during the AMT20 cruise, 12 October to 25 November 2010 (Figure 1). Seawater from the ship's continuously pumped supply was passed at a controlled rate as a bubble-segmented stream through a glass coil equilibrator [Xie et al., 2001] held in a water jacket, the temperature of which was recorded. The gas phase in the equilibrator (supplied as ultrahigh purity air) was separated from the water stream and fed to the sample loop (1 mL) of an analyzer comprising a molecular sieve column, a heated mercuric oxide bed, and a UV detector (Peak Laboratories, USA). Using the hydrogen solubility at the appropriate temperature, pressure, and salinity [Wiesenburg and Guinasso, 1979] and the measured concentration of hydrogen in the gas stream leaving the equilibrator, the equilibrium concentration of hydrogen in the aqueous phase was calculated. The calculated concentration of hydrogen in the seawater entering the equilibrator is the sum of the concentration in the aqueous phase leaving the equilibrator and the amount transferred into the gas stream per unit water volume. A correction factor (100/E) was then applied to account for the efficiency (E) of the equilibrator being less than 100%. The efficiency was measured daily by analyzing a water sample having a known hydrogen concentration, prepared by purging seawater with a 5 ppm hydrogen standard. The average efficiency was 78% (standard deviation 12%, n = 111). For samples collected over 2 days south of 43°S, efficiencies were anomalously low (average 56%)—a result that could be attributable to incomplete equilibration during the preparation of the seawater standard. This could result in saturations being overestimated by a factor of 1.4 for the southernmost samples; applying such a correction would reduce saturations at this end of the transect from around 250% to 180%. The supersaturation in the surface ocean is the measured hydrogen concentration in seawater minus that calculated to be in equilibrium with the atmosphere (C_eq) at ambient temperature and atmospheric pressure expressed as a percentage of C_eq. Measurements of equilibrated air were interspersed with scrubbed air and gas standard that had been made up by gravimetric dilution of a measured volume of hydrogen in zero air. Blanks (0.1 nmol/L, standard deviation 0.02, n = 30) were determined by running seawater purged with zero air that had been scrubbed of hydrogen.

Figure 1

AMT20 cruise track of RRS James Cook from Southampton to Punta Arenas, 12 October to 25 November 2010. Breaks occur where sampling stopped as the ship traversed the territorial waters of the Azores and Ascension Island. Colored circles show saturation of dissolved hydrogen (%) along track, obtained by resampling the full data set, shown in Figure 2, approximately every 15 nautical miles.

A major potential source of hydrogen contamination in waters pumped into the vessel's supply is corrosion of zinc anodes present on the hull. Prior to the AMT20 cruise, anodes located in front of the ship's water intake were removed. Nevertheless, as hydrogen released from other anodes can result in a spreading cloud of contamination around the hull while the ship is stationary, no samples are included in the data set that were collected while the ship was on station or within 30 minutes of approaching and leaving.

Because it had earlier been found [R. M. Moore et al., 2009b] that H₂ could be lost from the ship's water supply (an effect presumed to result from the presence of organisms colonizing the ship's plumbing), the plumbing was treated with bactericide (Decon) initially and twice during the cruise. Occasional extremely high measurements (23 of 5987) that could not be accounted for were identified and removed by determining a running median (in blocks of 11 measurements) for all underway measurements and identifying as anomalous all measurements that were 3.6 standard scores above or below the median; these were excluded from the analysis. As the analytical system does not respond instantaneously to a change in hydrogen concentrations, a number of peaks immediately following these 23 identified outliers were also excluded, giving a total of 42 points excluded in 5987 surface measurements.

Estimates of nitrogen fixation rates along the cruise track were obtained from three models; two are ecosystem models (“MIT” and “UVic-Kiel”) [Dutkiewicz et al., 2012; Somes et al., 2013] that simulate the distribution and activity of diazotrophs based on resource availability, and one [Deutsch et al., 2007] from a model that uses nutrient data to infer N₂ fixation rates needed to account for observed excess P consumption.

The UVic-Kiel results are from a coupled ocean-atmosphere-sea ice-biogeochemical Earth System Climate model of intermediate complexity [Weaver et al., 2001]. The ocean biogeochemical-ecosystem component is based on Somes et al. [2013] and incorporates the ecosystem improvements outlined in Keller et al. [2012]. The state variables included are two nutrients (nitrate and phosphate), two phytoplankton classes (N₂-fixing diazotrophs and other phytoplankton), zooplankton, sinking detritus, dissolved organic matter as well as carbon and oxygen. Iron limitation is calculated using monthly surface dissolved iron concentrations from the BLING model [Galbraith et al., 2010]. Diazotrophs grow more slowly relative to other phytoplankton, but they are not nitrogen limited. Denitrification, the consumption of nitrate in suboxic zones (O₂ < 5 µM) in the water column and seafloor sediments, decreases the nitrate:phosphate ratio in deep source waters and thus increases nitrogen limitation in the euphotic zone, thereby creating an ecological niche for diazotrophs in the model. Diazotrophs are also capable of consuming dissolved organic phosphorus after phosphate is depleted, which can stimulate additional N₂ fixation in the oligotrophic ocean where iron is sufficiently available, most notably in the tropical/subtropical North Atlantic. Nitrogen fixation has a strong seasonal cycle with highest N₂ fixation rates occurring in summer/early fall when stratification is strong and temperatures are high. The global rates for nitrogen fixation, water column denitrification, and benthic denitrification are 175, 61.7, and 113 Tg N yr⁻¹, respectively, in the model of Somes et al. [2013].

The MIT results are from simulations of the MIT general circulation model (MITgcm) coupled to a biogeochemical-ecosystem model. The model [Dutkiewicz et al., 2012] includes the cycling of carbon, phosphorus, nitrogen, silica, iron, and oxygen through inorganic, living, dissolved, and particulate organic phases. It resolves several nondiazotrophic autotrophs, two diazotrophs (analogs of Trichodesmium and unicellular), and two grazers. The diazotrophs are parameterized as growing more slowly and needing more iron than nondiazotrophs, but are not restricted in distribution by temperature [Monteiro et al., 2011]. Two simulations come from two different physical model resolutions (1° × 1°, 23 levels and 2° × 2.5°, 22 levels) [Dutkiewicz et al., 2012, 2014]. Within these two physical resolutions are also included simulations from sensitivity studies involving initial conditions and key parameters that are difficult to constrain with certainty (such as remineralization rates, and zooplankton grazing rates).

The approach of the Deutsch et al. [2007] model is entirely different in that it has no dependence on the growth dynamics of diazotrophs, but rather it identifies those regions of the ocean in which mass balance for dissolved phosphorus can only be achieved by a process that removes it independently of nitrate. In these regions the loss of excess phosphorus is attributed to diazotrophy, and the nitrogen fixation is calculated using a ratio of N/P consistent with the composition of diazotrophs. The main sources of uncertainty in the model are deficiencies in the underlying ocean circulation model and uncertainties in the seasonal distributions of nitrate and phosphate in both inorganic and organic forms.

3 Results and Discussion

Hydrogen saturations are shown in Figure 2a as a function of latitude along the ship's track. Breaks occur where sampling stopped as the ship traversed the territorial waters of the Azores and Ascension Island. Prominent features are high saturations (up to ca. 800%) in two bands lying between 25°N and 35°S, separated by a section with saturations around 500% between 3°S and 20°S. Southern hemisphere saturations are markedly more variable on a scale of tens of km. Nitrate concentrations are uniformly low between 40°N and 37°S (Figure 3), but phosphate concentrations are significantly higher (0.05–0.2 µmol/L) between the equator and 40°S. Saturations of 500% and higher occur over a temperature range of around 26°C north of the equator to around 17°C to the south.

Figure 2

(a) Saturation (%) of dissolved hydrogen in surface waters, SST, and H₂ flux to the atmosphere mmol m⁻² yr⁻¹, which can be used to estimate a minimum N₂ fixation rate assuming a 1:4 ratio of H₂ released to N₂ fixed; (b) measured nitrogen fixation rates (mmol N m⁻² yr⁻¹) on the same cruise track of AMT17 to 20°S [C. M. Moore et al., 2009a], shown as joined red circles; fixation rates from the PANGAEA database binned to 3° grid squares, measurements within 3° of the cruise track shown as circled blue points, and all other N. Atlantic measurements shown as blue dots. (c) Annual mean nitrogen fixation rates integrated to 120 m predicted by three models UVic-Kiel/Deutsch/MIT.

Figure 3

Plot of surface and from AMT20 (μmol/L) against latitude.

It is difficult to fully assess the possibility of contamination effects and their magnitude on our underway samples. Samples collected by Niskin bottles when the vessel was stationary showed that levels of hydrogen in the surface layer were highly variable and frequently in the range between 10 and 100 nmol/L consistent with contamination from sacrificial zinc anodes on the ship's hull (but, as noted in section 2 none was forward of the seawater intake). Inspection of Figure 2a shows that some scatter occurs above a narrow, dense band of values, but minimal scatter beneath. Contamination could contribute to positive scatter, but is not consistent with the tightly packed values forming the broad bimodal pattern with minimum values still showing high supersaturations of hydrogen.

Photochemistry has been shown to yield hydrogen from freshwater high in humic materials, and to a lesser extent from coastal seawater [Punshon and Moore, 2008]. This process, together with possible biological effects, was offered by Herr et al. [1984] as possible explanations of a relationship between hydrogen concentrations and solar radiation in waters of the tropical S. Atlantic. But extrapolation of measured production rates in colored water to the published absorption coefficients at 350 nm of oligotrophic waters led Punshon and Moore [2008] to conclude that only a fraction of the increase in hydrogen concentration reported by Herr et al. [1984] could be attributable to photochemical production. We are not aware of any thorough study involving direct measurements of photochemical production of H₂ in open ocean waters, but H₂ data [R. M. Moore et al., 2009b] from a transect between Fiji (18°S) and Hawaii (19°N) (Figure 4) were tightly correlated with measures of nitrogen fixation rates, so there was no need to invoke any significant photochemical production.

Figure 4

H₂ supersaturations along the AMT20 transect and a transect between Fiji and Hawaii [R. M. Moore et al., 2009b].

The production of H₂ from N₂ fixation has been known for decades [e.g., Hadfield and Bulen, 1969; Scranton et al., 1987] even though details of the production mechanism have been slow to emerge. Ogo et al. [2004] proposed that H₂ is evolved by reductive elimination from a dihydride species in the reactive centre of the nitrogenase that leaves the centre in an activated state, primed for binding of N₂ and its subsequent reduction. This mechanism is supported by Hoffman et al. [2014]. While the primary stoichiometry of the mechanism is 1 mol H₂ produced:1 mol of N₂ fixed (equation 1), much of the reducing power in H₂ is reutilized intracellularly through the action of uptake hydrogenases [Wilson et al., 2010b, 2012]. It can be speculated that if the diazotroph itself does not consume some of the waste H₂, it is likely to be consumed by other bacteria in oxygenated water. It is the net release rates from diazotrophs to solution that is relevant here. Laboratory measured ratios of net H₂ production to N₂ fixation range from 0.2 to 0.3 for Trichodesmium, which fixes N₂ during the day [Wilson et al., 2012] to 0.05 for the unicellular cyanobacterium Cyanothece, and 0.03 for the unicellular Crocosphaera [Wilson et al., 2010a]. Both of these unicellular diazotrophs fix N₂ at night, thereby separating photosynthetic production of oxygen from oxygen-sensitive nitrogen fixation. In turn, lower nocturnal availability of reductant would be expected to limit rates of hydrogen release [Wilson et al., 2010b, 2012]. It should be noted that Cyanothece has a coastal and benthic distribution rather than oceanic.

What is the minimum rate of nitrogen fixation that would be consistent with our observed H₂ concentrations? Estimating that rate is done in two steps, both of which have major uncertainties. The first is converting the observed supersaturations to fluxes, and the second is choosing a conversion factor. The net production of H₂ in the water column is set equal to its flux to the atmosphere (Figure 2a), that being the product of the excess H₂ and an exchange velocity calculated from measured wind speed using the relationship of Wanninkhof [1992]. Our calculated rate of N₂ fixation will be an underestimate on account of N₂ fixation beneath the mixed layer that will not contribute significantly to the H₂ flux to the atmosphere, and some unknown fraction of the H₂ produced throughout the water column would be consumed by microbes. Most flux values lie in the range 10 ± 5 mmol H₂ m⁻² yr⁻¹. As we are estimating the minimum nitrogen fixation rate capable of accounting for the estimated hydrogen production, we select the release ratio of H₂ to N₂ fixed reported for Trichodesmium, 1:4 [Wilson et al., 2012]. This leads to an estimated 80 mmol N m⁻² yr⁻¹ (range 40–120 mmol N m⁻² yr⁻¹), close to a weighted grand average of N fixation by Trichodesmium for 154 stations in the tropical North Atlantic, 87.2 ± 13.9 (s.e.) mmol N m⁻² yr⁻¹ [Capone et al., 2005], and thus consistent with expectations if Trichodesmium dominates N-fixation in the tropical North Atlantic, as has been suggested by Goebel et al. [2010], based on quantification of nifH genes from different taxa. The calculated H₂ fluxes in the southern hemisphere show higher mesoscale variability (Figure 2) but require N₂ fixation of a similar magnitude, or more if H₂:N₂ release rates are lower than assumed.

Measured nitrogen fixation rates for the Atlantic Ocean binned to 3° grid squares extracted from the PANGAEA database assembled by Luo et al. [2012] are shown in Figure 2b. Measurements within 3° of the cruise track are identified and also some along the same cruise track (as far as 20°S) from AMT17 [C. M. Moore et al., 2009a] when nitrogen fixation was strongly biased to the northern hemisphere and was reported to closely match the distribution of Trichodesmium, the abundance of which was at, or close to, zero south of the equator, a fact attributed to deeper mixed layers (from ∼3°N to ∼10°S) and, in particular, to iron limitation. It should be noted that the track of AMT17 diverged from that of AMT20 at 20°S and proceeded in a southeasterly direction. C. M. Moore et al. [2009a, Figures 2b and 2c] show elevated N* and decreased P*, both indicators of nitrogen fixation, in the western South Atlantic from ∼25°S to 35°S in an area traversed by our cruise track.

It is apparent that an estimate of nitrogen fixation in the range 40–120 mmol N m⁻² yr⁻¹ is consistent with the data between 30°N and the equator, but in the southern hemisphere the very sparse measurements available are much below this range. The implications of the comparison between hydrogen production estimates and measured nitrogen fixation measurements are (i) that the observed hydrogen supersaturations in the northern hemisphere Atlantic could be accounted for by nitrogen fixation if the molar ratio of H₂ released to N₂ fixed is near 0.25—a value reported for Trichodesmium in culture, and (ii) that N₂ fixation is markedly higher in the southern hemisphere western Atlantic than indicated by the PANGAEA database, and/or the source of H₂ in the southern hemisphere is different from that in the northern.

A deficiency of published nitrogen fixation rates is that some unknown fraction of them are likely to be underestimates on account of methodological shortcomings, namely, the incomplete equilibration of ¹⁵N₂ gas added to the incubated seawater [Großkopf et al., 2012]. These authors report evidence that the underestimation can be up to 6 times greater for unicellular diazotrophs than for Trichodesmium. Any general underestimation of N₂ fixation rates would strengthen the case for attributing hydrogen supersaturations to diazotrophy, certainly in the case of a Trichodesmium source, and for unicellular diazotrophs that have a greater release ratio than the two species that have so far been studied (Cyanothece and Crocosphaera), which fix N₂ at night and hence derive greater benefit from recycling H₂. It is interesting that a recently discovered and so far uncultivated unicellular diazotroph UCYN-A seems to have adapted to daytime N₂ fixation, when proximity to photosynthetic oxygen production is a problem, by lacking PSII and being symbiotically associated with a Prymnesiophyte [Bothe et al., 2010b; Thompson et al., 2012]. These traits could have consequences for net H₂ production. Besides reducing exposure to high concentrations of photosynthetically produced oxygen, a loose, surface association of UCYN-A with the alga could permit ready loss of H₂ to the surrounding water. Fixation of nitrogen during the day diminishes the need for energy production from hydrogen recycling, suggesting that the ratio of net H₂ production to N₂ fixation might be higher than for other coccoid cyanobacteria that fix N₂ at night [Wilson et al., 2010a]. Evaluation of these suggestions, and related predictions for diatom-diazotroph symbioses, might have to await the successful culturing of UCYN-A and further studies of the associations of cyanobacteria with symbiotic algae [cf. Goebel et al., 2010].

In view of the sparse measurements of nitrogen fixation in the South Atlantic and questions regarding their reliability, we have considered three model-based estimates of N₂ fixation rates along the cruise track (Figure 2c). In the case of the MIT model a measure of uncertainty is provided from a range of simulations that yield variations in the ratio of supply rates of iron and phosphate relative to dissolved inorganic nitrogen. These ratios are crucial in dictating where diazotrophs can flourish [Ward et al., 2013], so their variation leads to the spread of results seen in Figure 5. The envelope of uncertainty from the Deutsch model is provided by the maximum and minimum for each month. Despite the differences in assumptions and approach, in all these models the mean values of N₂ fixation along the cruise track (integration depth 120 m) support the occurrence of widespread N₂ fixation between latitudes 30°N and 30°S (Figure 2c). On an annual basis, none of the models predicts significantly higher N₂ fixation in the northern hemisphere than southern along our cruise track (Figure 1). Unless there is a large-scale difference in the hydrogen release ratios for diazotrophs between the two regions, the diazotrophic contribution to hydrogen supersaturations would be expected to be similar. This would be broadly consistent with our observed pattern of hydrogen supersaturations and suggests that nitrogen fixation has been underestimated in the southern hemisphere Atlantic, resulting perhaps from too restricted sampling, or use of methods that tend to underestimate nitrogen fixation from important classes of diazotrophs, such as unicellular forms. Any such underestimation would need to be associated with diazotrophs that have a hydrogen release ratio similar to (or higher than) that of Trichodesmium. Alternatively, it would seem necessary to argue that the three models all have shortcomings that cause them to overestimate nitrogen fixation in the southern hemisphere Atlantic. Certainly, the models are not consistent with each other in their more detailed latitudinal simulations of nitrogen fixation, but they do combine our current knowledge of the controls on diazotrophs and their ecology, and our knowledge of the distribution of excess phosphate (P*) in the Atlantic [Deutsch et al., 2007].

Figure 5

(top) The modeled mean nitrogen fixation rate (mmol N m⁻² yr⁻¹) along the AMT20 cruise track and range of uncertainties resulting from seasonal variations (Deutsch model); (bottom) the modeled mean and range along the same track from an ensemble of MIT model runs.

Other modeling studies also suggest nitrogen fixation south of the equator: the models of Hood et al. [2004] and Stukel et al. [2013] indicated Trichodesmium in the S. Atlantic, mainly in the eastern basin, but also around 10°S in the longitude range of our models and cruise track, and the models of Moore and Doney [2007] and Krishnamurthy et al. [2009] indicate very similar nitrogen fixation as our models and along much of the cruise track. The compilation of data on Trichodesmium distribution in LaRoche and Breitbarth [2005] also shows it to have been observed in the latitude range 20°S–35°S close to the AMT20 cruise track. However, while Tyrrell et al. [2003] report a moderate abundance of Trichodesmium at ca. 35°S, also close to the AMT20 track, they found generally low abundances south of the equator in contrast to high abundances between 0°N and 15°N around 15°W–25°W.

Can it be the case that nitrogen fixation plays no significant role at all in the large-scale distribution of hydrogen supersaturations? This proposal is inconsistent with field results reported from a lower resolution study of hydrogen supersaturations and nitrogen fixation in the equatorial Pacific [C. M. Moore et al., 2009a], and also with laboratory studies that show Trichodesmium to release H₂ at a ratio to N₂ fixed of around 0.25.

Could the database of nitrogen fixation measurements in the Atlantic be biased such as to emphasize diazotrophy in some areas and underrepresent it in others? Historically there has been a bias toward making nitrogen fixation measurements in warm waters and a focus on Trichodesmium spp., the colonies of which are conspicuous and undoubtedly a major contributor to marine nitrogen fixation. Trichodesmium spp. distribution are reported [Carpenter and Capone, 2008] to be roughly limited by the 20°C isotherm, and while the organisms may be found in cooler water their activity is low and their growth slow. While laboratory studies [Staal et al., 2003] have shown acetylene reduction in a Trichodesmium culture (strain IMS 101) increasing right up to 35°C, Breitbarth [2005] reports maximum growth rates between 24 and 30°C. Karl et al. [2002] comment that most reported Trichodesmium blooms occur in waters at or warmer than 25°C. But there is now evidence for the presence of unicellular diazotrophs over a wider temperature range. For example, while Langlois et al. [2005] found filamentous cyanobacterial nifH sequences, with one exception, in waters having temperatures between 26.5 and 30°C and undetectable nitrate, they found nifH sequences of nonfilamentous diazotrophs in waters from 15 to 30°C and sometimes containing detectable nitrate. Unicellular diazotrophs have also been reported to be active in the temperature range 19–25°C, though typically showing lower N₂ fixation rates than Trichodesmium [Needoba et al., 2007]. The modeling study of Monteiro et al. [2011] also suggested a wider range and almost equal importance of unicellular diazotrophs to Trichodesmium.

If it is the case that unicellular diazotrophs have been underrepresented in studies of nitrogen fixation in the Atlantic, or their activity underestimated, it can begin to account for our reported distribution of hydrogen only if the ratio of H₂ released to N₂ fixed is higher than that of the only two species so far examined. This in turn strongly suggests that for any missing unicellular diazotroph to be significant in this respect, it would likely be a daytime fixer of nitrogen, so having a reduced need to recycle the hydrogen that it produces.

4 Conclusions

Supersaturations of hydrogen measured in the Atlantic Ocean north of the equator can be accounted for by typical published nitrogen fixation rates when a hydrogen:nitrogen release ratio matching that for Trichodesmium is used.

Our data show similar hydrogen supersaturations in the western Atlantic south of the equator. These are not consistent with published nitrogen fixation rates, even using release ratios around 0.25, typical of Trichodesmium. Possible explanations could include (i) an unknown source of hydrogen that does not play a major role in the North Atlantic, (ii) underestimation of nitrogen fixation south of the equator associated with a diazotroph that has a low efficiency of recycling hydrogen—this suggests a daytime fixer of N₂. Possible reasons for nitrogen fixation having been underestimated include a bias toward sampling in waters warm enough to support Trichodesmium growth, and the use of methods for measuring fixation rates that discriminate against unicellular diazotrophs.

Our hydrogen distribution in the Atlantic is consistent with modeled nitrogen fixation in that the distribution of each is approximately symmetrical about the equator—but with a requirement that the hydrogen:nitrogen release ratio cannot substantially differ between the north and south.

Interpretation of our data is hampered by the absence of information of hydrogen release rates of any unicellular diazotrophs that fix nitrogen during the day. Of particular interest is a UCYN-A that lacks photosystem II, thereby reducing the constraint imposed by photosynthetic oxygen production on the activity of oxygen-sensitive nitrogenase. Measurement of hydrogen release by this bacterium would be a great interest, but the organism has yet to be successfully cultured.

Acknowledgments

This study is a contribution to the international IMBER project and was supported by the UK Natural Environment Research Council National Capability funding to Plymouth Marine Laboratory and the National Oceanography Centre, Southampton. This is contribution number 251 of the AMT programme. The authors wish to acknowledge the contributions made by the Chief Scientist of AMT20, A. Rees, the Natural Science and Engineering Research Council of Canada, and DFG support for project SFB 754 (C.J.S). M.F. and S.D. are grateful for support from the Gordon and Betty Moore Foundation. We appreciate the constructive comments of reviewers.