Corrigendum: Both respiration and photosynthesis determine the scaling of plankton metabolism in the oligotrophic ocean

Nature Communications 6:6961 doi: (2015); Published 24 April 2015; Updated 15 June 2016 The original version of this Article failed to fully credit the use of the Ocean Data View software in figure 3, which appears below: Schlitzer, R., Ocean Data View, http://odv.awi.de, 2016.

T he microbial biota of the surface ocean is a prime determinant of the ocean's productivity and biogeochemical functioning. Marine phytoplankton is responsible for half of the Earth's oxygenic photosynthesis, which contributes to the redox state of the planet and the global cycles of C and O 2 in the ocean and atmosphere, affecting the global climate and creating habitable conditions for many organisms 1,2 . Part of the photosynthetically produced organic matter is respired in the surface ocean, either to support the maintenance costs of the phytoplankton or by the heterotrophic activity of other microbes. This oxidation counterbalances the fluxes of C and O 2 from photosynthesis, and reduces the net amount of organic matter available for consumption by metazoans. The difference between plankton photosynthesis (P) and auto-plus heterotrophic respiration (R) is net community production (NCP), the amount of photosynthetically fixed C available for sequestration to the deep ocean or for transfer up the marine food web 3 . Over significant scales, NCP summarizes the energy flow and metabolic state of a planktonic ecosystem, setting upper bounds to its contribution to O 2 and CO 2 fluxes, and hence to global climate.
The oligotrophic ocean is the most extensive biome of the world, occupying about half of the Earth's surface. Its areal productivity is limited because of the strong vertical stratification of the water column that reduces nutrient inputs to the illuminated zone. However, its extent makes its contribution to global productivity, C export and biologically driven oceanatmosphere fluxes significant [4][5][6] . In addition, climate change is increasing both the area 7 and stratification of oligotrophic gyres 7,8 , with feedbacks to the biological rates and the metabolic state. Despite its importance, the metabolic state of the oligotrophic ocean has remained controversial 3 , since del Giorgio et al. 9 published a seminal paper on the relationship between primary production and bacterial respiration in the ocean. Direct measurements of community metabolism in lowproduction ecosystems show a prevalence of heterotrophy (PoR) 10 , that leads to the 'intriguing, if not disquieting' idea that the extensive and remote oligotrophic ocean is heterotrophic 3 . These results are based on the in vitro incubation of small seawater samples (B125 ml) during short periods of time (B24 h), which are susceptible to biases from the bottle confinement of natural plankton communities 11 . An alternative to incubation-based methods is to study the indirect impact of plankton metabolism on inventories of biogenic or limiting compounds in large water bodies (km 2 ) during long periods of time (weeks to months) [12][13][14] . Contrary to direct measurements of planktonic activity, results from this approach show a prevalence of autotrophy (P4R) in the oligotrophic ocean 11 . These calculations involve a large uncertainty derived from the estimations of water mass mixing and air-sea gas exchange 10 , and assume a constancy in gradients of stocks and water transport that conflicts with non-steady ocean dynamics 15 .
Beyond methodological biases that have played a major role in the debate 3,10,11 , these two approaches measure different processes over completely different time and spatial scalessnapshots of the NCP of single communities versus the integrative metabolism of a succession of communities in a water mass. Comparison hence requires scaling-up in vitro data to the large and long spatial and temporal extent of in situ estimates. To this aim, empirical P:R relationships have been used to predict NCP from the extensive global data set of P (mainly 14 C derived) [16][17][18][19] . All the extrapolations carried out during the last 15 years have found heterotrophy to prevail whenever P is low, leading some authors to conclude that the oligotrophic ocean is heterotrophic and the (autotrophic) in situ data are biased 10 . According to these authors, the input of organic matter to the open ocean is much higher than the observations of stocks and export from the euphotic zone, and non-oxygenic autotrophic processes (not included in P estimations) are significantly more relevant than currently believed 10 . On the contrary, others have concluded that the oligotrophic ocean is autotrophic, and that all the in vitro measurements of plankton production-including tens of thousands of 14 C measurements-could be biased because of the perturbation of the light conditions 11 . Such a widespread and constant state of autotrophy would require a supply of inorganic nutrients to the surface oligotrophic ocean at rates much higher than observations of stocks and fluxes indicate 15 . Both positions hence ultimately imply unresolved explanatory hypotheses that would force a reconsideration of our knowledge about the magnitude, variability and control of organic matter production and cycling in the open ocean.
Despite the lengthy discussion on methodological biases, very little attention has been paid to the scaling procedure that is essential for data comparisons. In all cases, the empirical P:R [16][17][18][19] or P/R:P 17,20 relationships are assumed universal [16][17][18]20 or constant for a certain latitudinal band (for example, 10-40°) 19 , which relies on some important and untested assumptions. The idea that NCP (that is, P minus R) may be predicted in the oligotrophic ocean from P alone is based on the hypothesis that changes in P but not in R control the metabolic balance either globally 10,[16][17][18]20 or at low latitudes 11,19 . This may occur because the variability of R is irrelevant or minor [20][21][22] , or because it is tightly coupled to the variability of P [16][17][18][19] . In either case, the use of a single P:R relation implies that the influence of P-independent processes on heterotrophic R (for example, magnitude and quality of allochthonous organic matter inputs, consumption of organic matter accumulated during productive phases, composition, structure and activity of the heterotrophic communities) are uniform throughout at least the low latitude band of the global ocean. This is a very tenuous assumption, as even in the most extensive data sets 18,19 used both to calculate balances and to derive P:R relationships 10,11,18,19 , the great majority of data from the oligotrophic ocean come from one province, the Eastern Atlantic gyre (NAST-E). However, a reduced number of observations in the South Atlantic gyre have suggested differences in community metabolism that would reflect distinct P:R relationships 23,24 .
To test the assumption that no regional differences exist in either the P:R relationship or the metabolic balance throughout the oligotrophic ocean, that is, to assess the basis of the scaling procedure sustaining the debate, we have compiled 194 vertical profiles (median five depths) in the epipelagic zone (defined here as the layer between the surface and the 1% of photosynthetically active radiation-PAR-depth) of in vitro derived P, R and NCP measurements (DO 2 after 24-h light and dark incubations) made across the Atlantic (50°N-50°S) during 10 Atlantic Meridional Transect (AMT11-18, AMT21-22) cruises from 2000 to 2012. All the cruises except AMT17 and AMT18 included both the North and South gyres (NAST and SATL). This is a very consistent data set in terms of sampling, incubation and analytical methodology and precision, it is equivalent to B76% of non-AMT Atlantic data in the latest published global NCP data set 18 , and includes 46 profiles from the very undersampled SATL. The analysis of this data set confirms the prevalence of net heterotrophy in the NAST, but shows that the plankton metabolism in the similarly oligotrophic SATL was predominantly autotrophic. Our data reveal significant differences in R, P:R relationship and NCP between the two Atlantic ocean gyres, indicating that the oligotrophic ocean is functionally diverse. This would bias any regional to global prediction of NCP based on generalized P:R relationships, constraining the scaling of plankton metabolism to functionally coherent ecosystems. Results Plankton metabolic rates in the Atlantic gyres. The AMT cruises traversed several biogeographic provinces of the Atlantic Ocean (Fig. 1). Gyre stations were identified when the depth of the thermocline and chlorophyll a maxima were 4100 m (B20-38°N for the NAST; B10-32°S for the SATL).
Integrated NCP was clearly different between the NAST and SATL (t-test (n ¼ 91) ¼ À 7.14, P ¼ 0.0001; Fig. 2a, Table 1). Net heterotrophy prevailed in the NAST, agreeing with the literature 10,[17][18][19][20]23,24 ; 85% of our integrated NCP data were negative, the mean ( ± s.e.m.) NCP was À 14.55 ± 2.42 mmolO 2 m À 2 per day and the 95% confidence interval was À 19.31 to À 9.79 mmolO 2 m À 2 per day. However, in the SATL, a balanced or net autotrophic metabolism prevailed (78% of the data), with a mean NCP of 9.78 ± 2.34 mmolO 2 m À 2 per day and a 95% confidence interval of 4.96 to 14.59 mmolO 2 m À 2 per day. The composite vertical and latitudinal variability of volumetric rates during AMT11 to AMT22 shows that the significant differences in NCP between the NAST and SATL were consistent throughout the entire epipelagic zone (t-test (n ¼ 433) ¼ À 6.63, P ¼ 0.0001; Fig. 3). Sixty-two per cent of the 202 volumetric data from the NAST were net heterotrophic, with a mean (± s.e.m.) of À 0.12 ± 0.02 mmolO 2 m À 3 per day, while 63% of the 233 data from the SATL were net autotrophic, with a mean (± s.e.m.) of 0.06 ± 0.02 mmolO 2 m À 3 per day (Table 1). Although these percentages in the NAST concur with the literature (57% (ref. 10)), conversion of volumetric data into discrete categories with a cutoff NCP of zero (that is, auto-or heterotrophic) misrepresents the prevailing metabolism because of the relatively high variability of these very low rates (Fig. 4), especially in the upper 25 m (B33% light depth) and at the 1% light depth, where the mean NCP rates were very close to zero at both the NAST and SATL (Fig. 5). This explains the lower percentage of the volumetric data that were auto-or heterotrophic in the SATL and NAST (63 and 62%, respectively), compared with the depth integrated data (78-85%). Vertical profiles of the mean and median NCP through the epipelagic zone provide a more realistic account of the clear differences in the prevailing metabolism of the two gyres (Fig. 5).
Most interestingly, the regional differences observed in NCP were not caused only by differences in P, but fundamentally by the variation of R. The assumed relative constancy of R [16][17][18][19][20][21][22] is only found when we analyse the entire latitudinal range of the transect, that is, when the highly productive equatorial and temperate provinces are included together with the oligotrophic gyres (Figs 2b and 3). However, within the two gyres, we found similar ranges of variability for both volumetric and integrated P and R rates (Figs 2b and 3 and Table 1), and this variability of R has consequences for the differences in metabolic balance between the gyres. The mean integrated P was lower in the heterotrophic NAST than in the autotrophic SATL (57.51±3.50 versus 68.32 ± 3.13 mmolO 2 m À 2 per day, respectively), consistent with the hypothesis that P alone controls NCP [10][11][12][13][14][15][16][17][18][19][20] . However, the mean integrated R was lower in the autotrophic SATL than in the heterotrophic NAST (58.53±3.85 versus 72.09 ± 3.47 mmolO 2 m À 2 per day), where very few data lower than B35 mmolO 2 m À 2 per day were found (Fig. 2b). Volumetric rates emphasize the importance of R to the significant differences in NCP between the NAST and SATL. R was significantly higher in the heterotrophic NAST (0.68±0.03 mmolO 2 m À 3 d À 1 ) than in the autotrophic SATL (0.48 ± 0.02 mmolO 2 m À 3 per day; t-test (n ¼ 439) ¼ 6.08, P ¼ 0.0001), while P rates were not significantly different (0.56 ± 0.02 and 0.55 ± 0.02 mmol O 2 m À 3 per day in the NAST and SATL, respectively; t-test (n ¼ 433) ¼ 0.56, P ¼ 0.57; Table 1). Volumetric rates of P were mostly lower in both gyres than the latest published threshold P for net heterotrophy (1.26 mmolO 2 m À 3 per day) 18 (Fig. 3). However, the corresponding prediction of net heterotrophy was only fulfilled in the NAST, precisely the oligotrophic province that contributes the most to the global data set whose generalized P:R relationship sustains both this threshold 18 and the hypothesis of relative R constancy 10,18 .
Scaling plankton metabolism in the Atlantic gyres. The difference in the relative importance of R, and its consequences for scaling the metabolic balance are clearly shown in a plot of the P:R relationships in the two gyres ( Fig. 6). We focus on the integrated rates because they provide a more accurate representation of the interaction of auto-and heterotrophic processes in the epipelagic ocean 16,23 . The NAST P:R relationship ARTICLE resembles the generalized relationships that have been used to scale-up in vitro data for the last 15 years [16][17][18][19] . The degree of heterotrophy is inversely related to the magnitude of P, so that below a certain threshold P (B120 mmolO 2 m À 2 per day), the net heterotrophy would always be predicted. As P in the oligotrophic ocean is typically below calculated thresholds, using this type of relationship to extrapolate NCP, together with the assumption of trophic invariability, has led to the conclusions that either the oligotrophic ocean is net heterotrophic and requires substantial allochthonous inputs of organic matter 10 , or in vitro data are biased 11 . However, the SATL P:R relationship is essentially a 1:1 relationship (t-test, Po0.01), indicating a tight coupling of local auto and heterotrophic processes. This relationship would never predict net heterotrophy from P alone. Actually, neither the NAST nor SATL relationship is able to correctly predict the metabolic balance in the other oligotrophic gyre from P alone. Consequently, differences in R between the two gyres, and not only in P are critical for the prediction of metabolic balances in the oligotrophic Atlantic, which should therefore be based on distinct P:R relationships for (at least) each of the two gyres.
Potential seasonal bias. Most AMT cruises took place in boreal autumn, with only AMT12, 14 and 16 in spring. This implies that the seasonality of plankton metabolism in the gyres might bias our conclusions on regional variability. Seasonal differences in P, R and NCP in the NAST and SATL (Table 2) concur with a previous analysis of a partially overlapping data set 24 (note that gyre boundaries were identified in Gist et al. 24 from salinity/ density fronts and surface chlorophyll rather than the depth of the thermocline and DCM). To assess a potential seasonal bias, we have separated the entire AMT11-22 data set into astronomical season and repeated the analysis. Geographic patterns of P, R and NCP remain unaltered irrespective of the season, except for the fact that autumn NCP in the S gyre is not different from metabolic balance, which nonetheless sustains the significant difference between the N and S gyres (Table 2). Moreover, the slope of the P:R relationship in the SATL remains 1 in both spring and autumn, differing from the o1 slopes of the N gyre relationships (Fig. 7). Although further empirical evidence is necessary to define the seasonality of plankton metabolism in the Atlantic gyres, our generalized analysis of regional differences appears unbiased.

Discussion
The debate about the metabolic state of the oligotrophic ocean has evolved into a polarized dispute between two contrasting positions that are very difficult to reconcile (that is, the oligotrophic ocean is heterotrophic or it is autotrophic) and require substantial unaccounted for allochthonous inputs of either organic or inorganic nutrients 10,11 . Both positions arise from the same fundamental view that 'the oligotrophic ocean' is a  single steady-state ecosystem whose balance of trophic processes is meaningful across scales. This perception is evident in the very discussion about its 'metabolic state', a property of either organisms or ecosystems whose calculation requires knowledge of the spatial and temporal scales of key trophic processes and connections in the system (production, accumulation, transport, consumption and oxidation of organic matter) 25 . However, the definition of the oligotrophic ocean is solely based on the level of productivity, controlled by the degree of water column stratification that determines nutrient limitation for the phytoplankton. Moreover, the oligotrophic ocean includes distant regions whose trophic connections are difficult to determine. Hence, estimating a single metabolic state for the global oligotrophic ocean from scattered measurements of P and R assumes scale independency in its trophic dynamics, that is, rests on the important assumptions of functional unity and steady state.
Our findings indicate that both this logic and the empirical evidence sustaining the debate are flawed. The systematic prevalence of autotrophy in the oligotrophic SATL challenges ARTICLE the generally accepted view that net heterotrophy prevails in in vitro NCP measurements in the oligotrophic ocean 10,11 , and the conclusion that such a prevalence is due to systematic methodological biases 11,19 . Moreover, the difference in NCP between the NAST and SATL refutes the hypothesis that a single type of balance prevails throughout the oligotrophic ocean 10,11 . The oligotrophic ocean is neither auto-nor heterotrophic, but functionally diverse. The debate has partially resulted from the use of universal P:R relationships where the heterotrophic NAST-E is over-represented. The prevalence of R over P in the NAST implies carbon and energy deficits that need to be compensated through anoxygenic primary production, inputs of allochthonous organic matter or a higher efficiency in the use of metabolic energy. The paucity of reducing substrates and feeble redox gradients in the epipelagic zone of the oligotrophic ocean constrain the efficiency of chemoautotrophy and anoxygenic photosynthesis, whose relative contributions to euphotic zone carbon fluxes are o1% compared to oxygenic photosynthesis 26,27 , that is, too small to explain the metabolic differences between the NAST and SATL. Photoheterotrophic prokaryotes can use both light and organic matter for energy but require organic molecules as sources of carbon and electrons. Although proteorhodopsin-based and aerobic anoxygenic phototrophic bacteria are abundant and active in the oligotrophic ocean 28 , the implications of photoheterotrophy for the metabolic balance in the upper ocean are unclear. On one hand, phototrophic energy production reduces the need for aerobic respiration to sustain the maintenance costs of heterotrophs. On the other hand, photoheterotrophy increases the survival, and on occasions the growth, of prokaryotes under very limiting conditions 29 , possibly by supporting the energetic costs of processes such as active transmembrane transport, production of ectoenzymes, breakdown of complex organic molecules and cell motility 30 , but also by light induced anaplerotic CO 2 fixation, which may provide up to 18% of cellular carbon demand 29 . This augmented survival, biomass production and the ability for organic matter utilization would increase the potential for heterotrophic respiration in the oligotrophic gyres. The consequences of photoheterotrophy for the metabolic cycling of carbon and the competitive advantage of planktonic bacteria [28][29][30] suggest an important, and yet unknown, role in energy and carbon balances in the oligotrophic ocean. However, it is difficult to infer the relationship between photoheterotrophy and the regional differences in R and NCP, and photoheterotrophy would not reduce the estimated deficit of organic carbon in the NAST per se.
Calculations of dissolved organic carbon import to the surface of the NAST are one to two orders of magnitude too small to support previous estimations of net heterotrophy 11 . However, the mean mixed layer (upper 50 m) in vitro NCP deficit in the NAST used in these calculations ( À 1 mmolC m À 3 per day) 11 is B10 times higher than the À 0.12 mmolO 2 m À 3 per day mean epipelagic NCP (average 112 m depth) estimated from our data set (Table 1; see also Fig. 5). In addition, our data show a depthdependent distribution of NCP through the epipelagic layer, with lower heterotrophy near the 1% light depth (B100 m depth) and in the upper 25 m (B33% incident PAR; . Although we are reluctant to calculate annual balances from our data set, assuming a respiratory quotient of 0.8 11 and a mean mixed layer of 50 m 11 , 365 times our mean NCP in the upper 50 m of the NAST ( À 0.07±0.03 mmolO 2 m À 3 per day) amounts to À 1.1 ± 0.4 molC m À 2 per year, that is, less than twice the 0.7 molC m À 2 per year dissolved organic carbon input estimated 'with high uncertainty' because of the ill constrained inputs from the margins 14 . These same lateral inputs are also needed to balance phosphorous budgets in the NAST. The phytoplankton of the North and South Atlantic gyres is limited primarily by nitrogen availability 31 ; however, a distinctive characteristic of the NAST is that the concentration of bioavailable phosphorous in the euphotic layer is the lowest of all the ocean gyres and limits phytoplankton production 13 . The similar rates of phytoplankton carbon fixation 32 and P (Table 1) in the NAST and SATL hence concur with the similar euphotic zone concentrations of bioavailable nitrogen, but are at odds with the chronic phosphorous limitation in the NAST. The lateral input of dissolved organic phosphorous (DOP) from the shelf region of NW Africa into the gyre interior helps to balance this discrepancy 33 . DOP imported by a combination of gyre and eddy circulations is estimated to support up to 70% of the particle export over much of the gyre 33 , and should therefore impact on the balance between P and R. DOP may be directly utilized by the phytoplankton or recycled by heterotrophic bacteria, but the latter have been found to easily outcompete Prochlorococcus (the most abundant phytoplankter in the oligotrophic ocean) for DOP (ATP) uptake 34 . Estimating the metabolic impact of DOP import into the NAST requires resolving the variability of group-specific differences in DOP uptake and utilization, which might also interact with the photoheterotrophic enhancement of dissolved organic matter bioavailability, and with an anticipated increase in the open ocean phosphorous limitation with global warming 35 . A mechanistic understanding of the regional differences in community metabolism requires not only solving the spatial and temporal variation of NCP and relative inputs of allochthonous inorganic and organic nutrients within the gyres, but also the study of the multiple biochemical strategies for nutrient stress 36 and energy and carbon acquisition across spatial and temporal scales 37 .
Our results show that R heterogeneity in the oligotrophic ocean is important for NCP variability and prediction, and that global metabolic balances in the oligotrophic ocean represent the average of different metabolic states, rather than the metabolic state of a single global biome. While the potential of P:R relationships for NCP scaling remains, extrapolation should be specific to provinces showing a coherent trophic functioning 38 . The different geographic extent of these provinces requires spatially explicit approaches, that is, a trophic biogeographic partitioning of the ocean based on the spatial and temporal variability not only of P but also of R 23,38 . This is a difficult task because only the substantial P data set and reliable predictive algorithms 39 allow for comprehensive regional and seasonal depictions [39][40][41] , which is not enough to reveal the trophic diversity of the ocean. Besides, the scale dependency of ecosystem metabolism along the temporal axis is arguably as important for the highly dynamic, non-steady-state planktonic ecosystem, and particularly relevant in the context of anthropogenic change 42 . Although R is a slow variable compared with P, its temporal variability could be important for prediction of trophic dynamics over scales significant to environmental change 43 .
The estimation of the metabolic balance of the oligotrophic ocean continues to be an important and urgent challenge for ocean biogeochemistry. Integrative in situ approaches are invaluable for constraining metabolic balances over long and large scales. However, only in vitro or other instantaneous methods may resolve the spatial heterogeneity and temporal dynamics of ecosystem processes, which is a prerequisite for prediction. Our findings call for an effort to improve the scope  ARTICLE and resolution of R and NCP measurements in the ocean. While we are far from a mechanistic understanding of the variability of plankton trophic functioning, the scale difference between in situ and in vitro methods provides unique opportunities to derive and test system-dependent empirical models for NCP extrapolation. A better appreciation of the scaling of metabolic processes to biogeochemical fluxes is a priority to improve our prediction of the ocean's interaction with the changing climate.

Methods
Sampling. Ten latitudinal (B50°N-45°S) transects of B70 stations across the Atlantic Ocean were conducted from 2000 to 2012 on RRS James Clark Ross, RRS Discovery and RRS James Cook between the United Kingdom and South America (Fig. 1). All the cruises except AMT17 and AMT18 included measurements of metabolism in both the North and South gyres (NAST and SATL). AMT cruises 11, 13, 15 and 17-22 took place in boreal autumn, while AMT12, 14 and 16 were in boreal spring. Details of sampling, incubation and analytical procedures for each cruise, including the complete cruise reports, may be found at http://www. amt-uk.org/ and Robinson et al. 44 Chlorophyll a concentration. During each cruise, vertical profiles of temperature and fluorescence were obtained twice daily (at B2 h before dawn and at solar noon) to a depth of B300 m with SeaBird and WETLabs Wet star sensors fitted to a SeaBird 9/11 plus CTD system. Chlorophyll a concentration was derived from calibrated fluorescence readings. A rosette of 24 Â 20 l Niskin bottles fitted with the CTD system was used to collect seawater samples from six to nine depths (including light depths sampled for photosynthesis and respiration incubationssee below) to calibrate the CTD fluorometer following Welschmeyer 45 . Samples of 250 ml were filtered through 47-mm 0.2-mm polycarbonate filters. The filters were placed in a vial with 10 ml 90% acetone and left in a freezer for 24 h. The samples were then analysed on a precalibrated Turner Designs Trilogy fluorometer with a non-acidified chl module (CHL NA #046) fitted.
Plankton metabolism. Gross photosynthesis (P), dark community respiration (R) and NCP were determined from in vitro changes in dissolved O 2 after 24-h light and dark bottle incubations. Although in situ incubations would be preferred to deck incubations, they were prevented by cruise logistics (AMT cruises are cruises of opportunity where stations last B2 h and are separated by B350 km (B13,500 km transect sampled over 42 days). Seawater was sampled daily, B2 h before dawn, from three to six depths (median 5) using the rosette of 24 Â 20 l Niskin bottles. The total number of suitable stations ranges from 8 (AMT17) to 37 (AMT22) (mean of 19). Water was collected into opaque polypropylene aspirators from 4 (AMT12), 5 (AMT11, 14,15,16,17) or 6 (AMT13, 18,21,22) light depths covering the PAR range from B97 to 1% of surface irradiance. Actual light depths changed between stations and cruises, but typically included 97, 33, 14, 7 and 1% of surface irradiance. Irradiance levels were determined from light measurements made the previous day, and the assumption that the deep fluorescence maximum approximated the depth to which 1% of surface irradiance reached. The depth of the 1% light in the next noon PAR profile was always within 7% of our estimated depth. Twelve to 18 120-cm 3 , gravimetrically calibrated, borosilicate glass bottles were carefully filled from each aspirator bottle using silicon tubing, overflowing by 4250 ml. From each depth, four to six replicate 'zero-time' bottles were fixed immediately with Winkler reagents (1 ml of 3 M MnCl or MnSO 4 and 1 ml of (8 M KOH þ 4 M KI) solutions) added separately with an automatic multipipette. Two further sets of four to six replicates were incubated for 24 h in surface water-cooled deck incubators or in temperature controlled water baths at in situ temperatures. These numbers of replicates are enough to obtain average coefficients of variation of the O 2 concentration measurements in the zero, dark and light bottles o0.1%, while allowing the daily completion by one person of all the analyses corresponding to five to six depths. One set of replicates was incubated in the dark, the other set in the equivalent irradiance to that found at the in situ sampling depth using various combinations of neutral density and blue plastic filters. Owing to limitations of sample water volume and analysis time during AMT12 and 13, we assumed that plankton community structure was homogeneous within the surface 15 m and so incubated a set of replicates sampled from the 55% light depth at 97% of surface irradiance. Flow cytometry data from AMT13 confirmed that this was a reasonable assumption, since picoplankton cell abundance varied by less than 5% between the two light depths throughout the transect, and nanoplankton varied by 11% (ref. 24). Incubations always started at dawn, and during the hours of darkness the incubators were covered with opaque screens to prevent interference from the ship's deck lights. After the 24-h incubation period, the dark and light bottles were fixed with Winkler reagents. Dissolved oxygen concentration was determined with automated Winkler titration systems using potentiometric (Metrohm 716 DMS Titrino) 46 or photometric 47 end-point detection. For the potentiometric method, aliquots of fixed samples were delivered with a 50-ml overflow pipette. Fixing, storage and standardization procedures followed the recommendations by Grasshoff et al. 48 Production and respiration rates were calculated from the difference between the means of the replicate light and dark incubated bottles and zero-time analyses: NCP ¼ DO 2 in light bottles (mean of [O 2 ] in 24-h light-mean zero time [O 2 ]); R ¼ DO 2 in dark bottles (mean zero time [O 2 ]-mean [O 2 ] in 24-h dark); P ¼ NCP þ R. The average s.e.m. of both the NCP and R rate measurements (which includes both experimental and biological variability) was 0.18 (n ¼ 875 and 876, respectively) mmolO 2 m À 3 per day. Integrated values were obtained by trapezoidal integration of the volumetric data down to the depth of 1% surface incident photosynthetically active irradiance. The s.d. of integrated P, R and NCP was calculated through propagation of the random error in the volumetric measurements. Samples were excluded (16 stations) when significant negative values of either P or R were measured, or when the variance of replicated measurements was anomalously high. The NAST and SATL P, R and NCP data are normality distributed (Lilliefors test, a ¼ 0.05).
The complete volumetric data set of P, R and NCP is available at http://www.uea.ac.uk/environmental-sciences/people/People/Faculty þ and þ Research þ Fellow/robinsonc#research maintained by Carol Robinson. The complete data set of all the physical, chemical and biological variables from each cruise is available at BODC (http://www.bodc.ac.uk/projects/uk/amt). The conditions under which the data may be used are in line with the NERC Data Policy (http://www.nerc.ac.uk/research/sites/data/policy/), and will be explained following a request, before the delivery of data.