Nitrification in the oligotrophic Atlantic Ocean

The recycling of scarce nutrient resources in the sunlit open ocean is crucial to ecosystem function. Ammonium oxidation, the first stage of the nitrification process, directs ammonium derived from organic matter decomposition towards the regeneration nitrate, an important resource for photosynthetic primary producers. However, the technical challenge of making nitrification rate measurements in oligotrophic conditions combined with the remote 10 nature of these marine systems means that data availability, and the understanding that provides, is limited. This study reports rate measurements of ammonium oxidation over a 13, 000 km transect within the photic zone of the Atlantic Ocean. These measurements, at relatively fine resolution (order 300 km), permit the examination of interactions with environmental conditions that may warrant explicit development and inclusion in model descriptions. At all locations we report measurable rates with significant variability between and within Atlantic provinces. This adds to evidence 15 that nitrification is an important component of pelagic nitrogen cycling which modifies the inorganic nitrogen inventory of the sunlit ocean. Particular features of interest included a significant hemispheric difference in ammonium oxidation rate and elevated rates associated with mesoscale eddy features. Statistical analysis of potential links between ammonium oxidation rate and routinely measured ecosystem variables indicated significant correlative structure, explaining ~65% of the data variability. Differences between sampling depths were of the same magnitude or greater 20 than horizontally resolved differences along the transect length, identifying distinct biogeochemical niches between depth horizons. Principle component analysis demonstrated that the best overall match between ammonium oxidation rate and environmental variables involved a combination of chlorophyll-a concentration, the duration of the light phase and silicate concentration (which we argue to be a short-term tracer of physical instability). Results allude to an association between ammonium oxidation and potentially short-term product(s) of photosynthetic activity and 25 subsequent degradation. Approximately 35% of data variability was not explained, which may include descriptions of DOM pool dynamics.


3
Luebbe AAIII autoanalyzer, using methods described previously (Woodward and Rees, 2001). The concentration of NH4 + was not determined during this study.

Analytical Flow Cytometry.
Microbial cells were enumerated based on their light scattering and auto-fluorescence properties using a BD FACSort™ flow cytometer. Photosynthetic cells were analysed live within 1-2 hours of collection. Non-photosynthetic cells were fixed with paraformaldehyde for 1 hour at 4 °C and then stained with SYBR Green I DNA stain (Life Technologies, Paisley, UK) in the dark at room temperature for 1 hour. Live samples were analysed at a flow rate of 100 180 μL min -1 for 4 minutes. Fixed/stained samples were analysed at either 50 or 12 µL min -1 for 1-2 minutes, depending on concentration. Data acquisition for photosynthetic cells was triggered on red (chlorophyll) fluorescence while the enumeration of SYBR Green I-stained bacteria and archaea was triggered on green fluorescence. Flow rates were calibrated daily using fluorospheres of known concentration.

105
Approximately 4 L of seawater collected from depths equivalent to 14 %, 1 % and 0.1 % of surface measured PAR (sPAR) was used for RNO2 studies. The isotope dilution approach  was used to derive the rate of ammonium oxidation, by measuring the rate of product formation (NO2regeneration). These 4 L sample volumes were amended with 15 NO2at < 10 % of the ambient concentration. Following amendment, triplicate 500 mL sub-samples were taken. Within these triplicate sub-samples, pre-incubation NO2 concentration and 15 N enrichment was 110 determined by synthesising Sudan-1 (see below) from sample NO2 -. Sudan-1 is synthesised for this purpose as it is readily formed and collected at room temperature, it is stable and amenable to both purification and isotopic analysis. From the remaining volume, triplicate 500 mL incubation vessels were filled and placed in deck incubation units covered with neutral density filters to simulate sPAR depths according to Joint Global Ocean Flux Study protocols (IOC, 1994). Incubation boxes were flushed with recirculating depth-appropriate temperature-controlled seawater.

115
Following approximately 10 hours of day-time incubation, experiments were terminated by GF/F filtration. Sudan-1 was synthesised from NO2within 500 mL samples, allowing post-incubation NO2concentration and 15 N enrichment to be measured. During these studies, a pre-incubation 15 NO2enrichment of 5.41±2.18 % was achieved, which was diluted by 0.92±0.83 % on average due to biological activity during incubation.
For the synthesis of Sudan-1, the first reagent (0.8 g of aniline sulphate in 200 mL 3 M HCl) was added to samples 120 in the proportion 0.5 mL per 100 mL sample volume. After mixing, samples were left for 5 minutes to homogenize, after which sample pH was verified to be approximately 2.0. Reagent 2 (24 g NaOH and 0.416 g 2-napthol in 200 mL Milli Q) was added in the proportion 0.5 mL per 100 mL sample volume. Samples were again mixed and left for 5 minutes before sample pH was verified to be approximately 8.0. The development of Sudan-1 was complete after 30 minutes of incubation at room temperature. Samples were acidified with citric acid (final pH 5.5) and an appropriate 125 volume of deuterated Sudan-1 internal standard (100 ng·µL -1 ) was added. Deuterated Sudan-1, which was used for sample quantification, was synthesized and purified according to methods described previously (Clark et al., 2006;2007). Samples were collected by solid phase extraction (SPE) using octadecyl (C18) 6 mL, 500 mg cartridges. SPE columns were stored at -20 °C until analysis.
At the land-based laboratory, samples eluted from SPE cartridges were purified by high performance liquid 130 chromatography (HPLC). Using a ramping methanol/Milli-Q water profile in combination with a Gemini-NX 5u C18 110A 250×4.6 mm column, sample peaks were resolved. The peak containing Sudan-1 internal standard was retained using a fraction collector while the remaining sample volume was discarded. Fraction-collected samples were dried and stored over anhydrous silica gel until further analysis.
Purified silylated samples were analysed by Gas Chromatography Mass Spectrometry using a HP-5ms column (30 135 m x 0.25 mm internal diameter), with the ramping profiles and extracted ions described in Clark et al. (2007). Using NO2concentration and enrichment measurements, the rate of NO2dilution was determined by applying the Blackburn-Caperon model (Blackburn, 1979;Caperon et al., 1979).

4
RNO2 rate data and associated explanatory variables were analysed using a range of multivariate statistical methods 140 (Clarke et al., 2014). Most variables were skewed, so prior to analysis all variables except phosphate excess (P*; calculated as P*=[PO4 3-]-([NO3 -]/16)) were log-transformed to reduce the influence of extreme values in subsequent analyses. Variables were normalized by subtracting the mean and dividing by the standard deviation to convert them all to a common dimensionless scale. Relationships among samples based on the explanatory variables were visualised using Principal Components Analysis (PCA), and hypothesis testing for differences among groups of samples was 145 undertaken using Analysis of Similarities (ANOSIM) with ordered factors (Somerfield et al., 2021). The generalised ANOSIM statistic is a scaled measure of effect size comparable across different tests.
To analyse relationships between RNO2 rates and explanatory variables the data for each sPAR band were analysed separately. To determine whether there was structure among variables that warranted further analysis, explanatory variables were subject to Type 2 Similarities Profiles (SIMPROF) analysis (Somerfield and Clarke, 2013). A Euclidean 150 distance matrix was calculated for RNO2 rates. A stepwise search algorithm (BVSTEP) with forward selection and backward elimination steps (Clarke and Warwick, 1998) was used to search for the smallest subset of explanatory variables that, in combination, 'best matched' the pattern of RNO2 rates across samples, the best match being defined as maximising the rank correlation between corresponding distance matrices. The algorithm was run with 20 restarts of 6 randomly selected variables. The significance of the maximum rank correlation was tested using an appropriate 155 permutation test (Clarke et al., 2008).

Contemporaneous physical oceanography data.
The ship, RRS James Cook made routine physical measurements whilst underway using the thermosalinograph to give temperature and salinity of the water at ~5m depth, and a depth profile of currents relative to the ship using the vessel-mounted 150 kHz Acoustic Doppler Current Profiler (ADCP). The latter were corrected using the ship's 160 navigation to give absolute instantaneous currents along the ship's track. To put these measurements in context, data were subsequently obtained from satellite altimeter and profiling floats that were in the area. These two sources of data were particularly useful in understanding anomalous observations at two of the CTD stations (28 °S, 39 °S).
The altimetry data were sourced from Copernicus Marine Environment Monitoring Service (CMEMS) who combine the sea surface height measurements from available observations and interpolated these data on to a regular 0.25º x 165 0.25º grid. Variations in sea level at the mesoscale (order 100-300 km) could be interpreted as eddies, with a local negative anomaly having cyclonic flow about it (clockwise in the southern hemisphere) and highs (anticyclones) having flow in the opposite direction. These flows were in geostrophic balance and thus represent stable current regimes, not incorporating tides or the effect of short wind bursts. Because of the long and near global altimetric record these data can be used to identify regions of persistent variability and to track the evolution of features over periods ranging from 170 weeks to years. However, satellites only give a near-surface view.
Profiles of temperature and salinity in the vicinity of the ship's track were provided by Argo floats, which are autonomous profilers typically recording 0-2000 m vertical sections every 10 days. As these instruments are freely drifting within the ocean, the location and timing of their measurements are not linked to that of the ship. Argo data for the required regions were obtained, and the profiles analysed to give a measure of the Mixed Layer Depth (MLD) as the 175 depth at which temperature was more than 3 ºC below that of the surface.

Temperature, salinity and PAR.
Characteristic physical features of oligotrophic oceans were demonstrated (Fig. 2a, b, c). The water column was thermally stratified, with sampling locations straddling a temperature range of approximately 12-29 °C. The deepest 180 mixed layer depths were associated with the central regions of the Atlantic gyres. Stratification weakened towards the northern and southern limits of the transect. In terms of salinity, sampling depths incorporated a range of 34.3-37.5. The most saline regions were associated with the ocean gyres and were highest in the Northern gyre. A region of relatively low salinity water was associated with the surface 30 m of the WTRA, defining the lower limit of the salinity range measured. In terms of the light field, 0.1% sPAR defined the deepest sampling point and equated to a depth within the spanned 2 orders of magnitude. Day light duration increased progressively south, associated with the seasonal transition from boreal autumn to austral spring.
The dissolved oxygen concentration was generally within the range 200-280 µmol L -1 , with the exception of the 0.1 % sPAR depth in the NATR/WTRA/northern SATL, where concentrations approaching 100 µmol L -1 were measured.
Plots of dissolved oxygen concentration against salinity (a), temperature (b) and NO3 -(c) are presented in Fig. 4.

200
Low dissolved oxygen concentrations measured at 0.1 % sPAR within NATR/WTRA/SATL provinces were associated with lower salinity, temperature, and relatively high NO3 -. RNO2 measurements associated with low-oxygen sampling depths were not anomalously elevated (Fig. 4d). In contrast, the RNO2 rate and physio-chemical characteristics of the eddy station identified as 39 °S distinguished it from the conditions associated with the majority of sampling stations within the transect.

Microbial cell distribution and abundance.
Profiles of chlorophyll-a concentration (Fig. 5a) indicated that within provinces of the North Atlantic, the subsurface maximum generally deepened progressively south. The highest chlorophyll-a concentration was measured within the WTRA, although there was considerable variability in the depth of the chlorophyll-a maximum within this province. Within the South Atlantic, although sampling stations were classified as SATL province, chlorophyll-a 210 profiles could be grouped into four illustrative sub-divisions, characterised by a general shallowing of the chlorophyll-a maximum progressively south.
Pico-eukaryotes (Fig. 5b) had a distribution that was similar to that of chlorophyll-a, in contrast to the profiles of the photosynthetic cyanobacteria Synechococcus and Prochlorococus (results not shown). Bacterial cell distribution (Fig. 5c) was comparable between gyres, though less defined within the WTRA, and indicated that the highest bacterial 215 abundance was measured within the relatively well-lit surface ocean.

Rates of RNO2
Within the physical, chemical and biological context described above, rates of ammonium oxidation were measured (Fig. 6). For all RNO2 data ( Fig. 7a-c), the range of volumetric rate measurements was 0.02-13.27 nmol-N L -1 d -1 , n=88 triplicated measurements. When integrated from the surface to the 0.1% sPAR, the range of RNO2 was 0.01-0.91 mmol 220 N m -2 d -1 . By extrapolating the vertically integrated results of each station (excluding 28°S and 39°S) from the equator into each hemisphere across a comparable distance (Fig. 8), the concentration of ammonium-N oxidised was calculated to be 0.36 mol N d -1 (1 x 4457 km transect) and 0.72 mol N d -1 (1 x 4637 km transect) for the NAG and SAG respectively. The box plot presented in Fig. 7(d) highlights variability within discrete sPAR bands of each hemisphere with mean RNO2 values generally skewed towards the lower region of the data range. The distribution of rate 225 measurements is presented in Fig. 7(e) while provincial variability, which averaged 25-fold, is presented in Fig. 7(f). Using a 2 sample t-test assuming unequal variance, RNO2 at 1% sPAR in the North Atlantic (M=0.76 nmol L -1 d -1 , SD=0.67 nmol L -1 d -1 ) was significantly lower than RNO2 at 1% sPAR in the South Atlantic (M=1.77 nmol L -1 d -1 , SD=1.52 nmol L -1 d -1 ); t(15)=-2.40, p=0.03 (no significant difference between hemispheres was found for 14% and 1%-50m sPAR depths). Volumetric rates of RNO2 equated to a NO2 turnover of 0.4-335 days (average 33 days).  (Table 1; Fig. 9) clearly demonstrated that differences between sPAR bands were of the same magnitude or greater than differences along the transect length. PC1, separating the sPAR bands, explained 42.7% of the total variation while PC2, representing latitudinal variation, explained 23.4%. This was supported by 2-way ANOSIM with ordered factors, 235 latitude and sPAR band, where the test for differences among latitudes removing differences over sPAR bands (R O =0.465, p<0.001) gave a lower value for the statistic than the corresponding test for differences among sPAR bands removing latitudinal trends (R O =1, the maximum possible, p<0.001). It was thus decided to analyse relationships among RNO2 rates and environmental variables within each sPAR band separately.
Type 2 similarity profiles (SIMPROF) showed evidence for significant correlation structure among variables in all 3 240 sPAR bands (14% sPAR band, =0.296, p<0.001; 1% sPAR band, =0.205, p<0.001; 0.1% sPAR band, =0.157, p<0.001). In the 14% sPAR band there was a significant relationship (=0.313, p=0.025) between patterns in RNO2 rates and a subset of 3 explanatory variables consisting of light hours, chlorophyll-a and silicate. In the 1% sPAR band the best subset, also of 3 variables, contained light hours and chlorophyll-a, but with NO2 instead of silicate. The overall relationship was not significant (=0.170, p=0.474). In the 0.1 % sPAR band the best subset (=0.415, p<0.001) 245 contained 6 variables, light hours and chlorophyll-a, temperature, NO2 -, PO4 3and P*. A 2-way search for the best overall match within sPAR bands removing any latitudinal effect selected light hours, chlorophyll-a and silicate, ave=0.267, p<0.001.

Eddy features 28 °S and 39 °S.
The majority of the AMT transect ran through quiescent waters free from rapidly varying mesoscale features. The 250 equatorial belt is a region of counter-flowing currents, but these did not lead to any unusual features during this particular AMT campaign. However, anomalous profiles were observed at 28 °S and 39 °S ( Fig.1), in the form of conspicuously prominent RNO2 rates ( Fig. 7a-c). The spatial separation of CTD stations (~300 km) was too great to resolve these physical features. Thus, altimeter and Argo data were obtained subsequent to the cruise in order to provide context to these observations.

255
Mapped fields of Sea Surface Height (SSH) from CMEMS were used to calculate the mean sea surface topography and the variations about that. A map of the root mean square variability (Fig. 10c) demonstrated that the typical changes north of 35 °S to be only ~0.1 m; however, even in this region there were a number of highs in the SSH anomaly (deviation from the mean) at the time of sampling the 28 °S station (Fig. 10a). The anticyclonic feature centred on station 28 ⁰S had anticlockwise geostrophic flow along the contours in SSHA, whilst the total near-surface velocity 260 recorded by the ADCP was mainly to the southwest. A series of SSHA composites a week apart showed a central anticyclonic feature to be moving westwards, whilst the contemporaneous Argo profiles showed that the thickest mixed layer was located close to the centre of this feature (Fig. 10b). A larger anticyclonic feature was seen 400 km further to the east. Examination of a time series of SSHA maps and Hovmöller diagrams (not shown) confirmed that these features had travelled from the Agulhas Retroflection, taking 2.5 years to get to this location.

265
The feature around CTD station 39 °S was very different. This is the region of the Brazil-Falklands Confluence, where two strong currents collide generating significant mesoscale activity and locally generated eddies. These features also move westwards, but their associated water masses were not from the Agulhas region. In this region the total velocities from the ADCP agreed with the contour directions governing the geostrophic flow. As it is a much more active region, Argo floats do not remain in the region for long, so the examples found ( Fig. 10d) were more dispersed 270 than in the previous case. Using the previously stated definition of MLD produced much larger values than at 28 °S, but again supported the concept that the surface layer was much deeper over the highs in SSH than over the lows.

Overview and objectives.
This study measured rates of ammonium oxidation within the photic zone across an extended spatial scale 275 (13,000 km), reporting significant variability between and within provinces of the oligotrophic Atlantic Ocean between ~49 °N and ~39 °S. Insights add to evidence that surface ocean nitrification is an important component of pelagic nitrogen cycling that modifies the inorganic nitrogen inventory. Particular features of interest included a significant 7 hemispheric difference in ammonium oxidation and elevated rates associated with mesoscale eddy features. Analysis of potential links between ammonium oxidation and routinely measured ecosystem variables indicated significant co-280 variability between ammonium oxidation, chlorophyll-a concentration, the duration of the light phase and silicate concentration, alluding to an association between ammonium oxidation and product(s) of photosynthetic activity. However, a significant fraction (~35 %) of variability in ammonium oxidation remained unexplained, indicating that additional factors were excluded from this analysis. We first consider the biogeochemical context to the study before presenting rate observations and insights from statistical analysis.

Physical context
A thermal range of 17 °C typically existed in vertical 250 m water column profiles. While temperature has a direct impact on rates of biological activity, the physical stability of the water column caused by thermal stratification establishes spatially distinct biogeochemical niches (Giovanoni and Vergin, 2012). Within the well-lit surface ocean that includes the 14 % sPAR, defined as the region between the sea-surface and the deep chlorophyll-a maximum 290 (DCM), turbulent mixing with nutrient rich deep water is supressed and biological activity maintains extremely low inorganic nutrient concentrations. Relatively sharp physio-chemical gradients at depths similar to the 1 % sPAR create a niche where low light-adapted phototrophs access relatively high inorganic nutrient concentrations, forming the DCM. At greater depths, incorporating the 0.1 % sPAR and typically referred to as the twilight zone (approximately 100-1000 m), water continues to cool. While there is relatively little information for the temperature sensitivity of 295 nitrification in the open ocean, Horak et al., (2013) demonstrated a relative insensitivity to a temperature range of 8-20 °C in their study of a fjord-like basin, although low pH at the study site may have constrained any temperature response (Beman et al., 2011).
The range in salinity was 3.2, with limits defined by notably fresher water in the sub-tropical region (WTRA), a feature potentially related to intense rainfall within the inter-tropical convergence zone, and relatively high salinity in 300 the central Atlantic gyres. While a contributory role for salinity in shaping nitrifying community structure has been identified in studies of estuarine sediments (Francis et al., 2003;Bernhard et al., 2010), such gradients are considerably more extreme than those of the open oceans, where little, if any information is available to describe its influence upon variability in nitrification rate.
The attenuation of light with depth leads to progressively decreasing daily light dose, ranging 2 orders of magnitude 305 across the three RNO2 sampling depths (Fig. 2c). Superimposed upon diel and depth-related light variability was a seasonal increase in day-light duration progressively south, associated with the transition from boreal autumn to austral spring. Seasonally variable characteristics are evident in the photic zone of sub-tropical Atlantic gyres driven by changes in solar insolation (McClain et al., 2004). Weaker solar insolation during winter supresses DCM productivity, allowing higher nutrient concentrations to pass this 'biological filter' into the well-lit surface ocean. This mechanism 310 supports higher chlorophyll-a concentrations in the surface ocean (Taylor et al., 1986) and would presumably stimulate broader biological activity, including the decomposition of recently produced organic matter and the regeneration of nutrients. Seasonality in the rate of nitrification was not evident in the study of Yool et al. (2007), although such a signal may exist (Newell et al., 2013).
Evidence demonstrates that light structures the distribution of nitrifying organisms, whose activity may be supressed 315 within the photic zone (Shiozaki et al., 2016;Olson et al., 1981). Further, by poorly competing with photosynthetic plankton for NH4 + resource, nitrifier distribution and activity is indirectly constrained by the light regime (Ward, 1985;Smith et al., 2014). While the light/dark cycle may lead to a diel variability in nitrifying activity (increasing at night with the recovery from light inhibition; Smith et al., 2014), resource competition with phytoplankton may persist due to their capacity for dark N-assimilation, especially of NH4 + (Clark et al., 2002).

Chemical context
In terms of the chemical context to the study, characteristically low concentrations of NO3 -, PO4 3and Si(OH)4 were evident within the well-lit surface ocean (14 % sPAR), progressively increasing at 1 % and 0.1 % sPAR ( Fig. 3.a, b, f). This distribution is consistent with biological draw-down, likely dominated by nutrient acquisition by phototrophs (Poulton et al., 2006;Marañon et al., 2000). Extremely high concentrations were a prominent feature of nutrient profiles In the mesopelagic ocean, nitrification maintains the accumulation of NO3 -, while the diapycnal nutrient flux to the surface open ocean is relatively weak, especially during periods of low wind energy input. NO3supplied via nitrification in the photic zone is either extremely low (Clark et al., 2008;Newell et al., 2013) or below detection (Lomas et al., 2009). Alternative NO3and PO4 3sources to the surface ocean includes introduction as a component of 330 atmospheric dust (Gruber and Sarmiento, 1997;Baker and Jickells, 2017), subsurface injections via the action of internal tides over mid-ocean ridges (Tuerena et al., 2019) or the passage of mesoscale eddies (McGillicudy et al., 1998;Oschilies and Garcon, 1998). No direct correlation between nitrification and NO3concentration has been reported. A relationship with PO4 3availability has been demonstrated in water treatment systems (de Vet et al., 2012), although there is no evidence to support this correlation in low nutrient environments.

335
Analytical capacity for nano-molar NH4 + measurements was unavailable for this study, creating a potentially significant shortfall in the data set, since NH4 + concentration has been demonstrated to influence the activity, distribution and composition of nitrifying communities (Bouskill et al., 2012;Smith, 2016). As the primary resource for nitrification, it is intuitive to expect a direct relationship between ammonium oxidation rate and NH4 + concentration (Smith et al., 2016), evidence for which supports the representation of this process as the specific rate (d -1 ; Yool et al., 340 2007). However, this correlation is not consistently demonstrated (Bouskill et al., 2012) and the process of NH4 + oxidation may exhibit saturation kinetics (Shiozaki et al., 2016;Martens-Habbena et al., 2009).
A phosphate excess similar to 0.2 µmol L -1 was evident at all depths in the WTRA and SATL (Fig. 3c). This feature is related to both source waters (Sarmiento et al., 2004) and the interhemispheric difference in N-fixation, supported by micro-nutrient inputs via atmospheric dust deposition (Baker and Jickells, 2017), leading to P* drawdown in the North

345
Atlantic (Moore et al., 2009). There is no evidence of a direct relationship between P* and nitrification rate. However, in Atlantic regions of enhanced N-fixing activity, evidence suggests that organic material produced by diazotrophs is recycled within the surface ocean (Mulholland, 2007). While nitrification has rarely been measured simultaneously with N-fixation (both were measured by Raes et al 2020 although no relationship was reported; it has been implied that Nfixation leads to an increase in the nitrate inventory; Mills et al., 2004), it likely has a role in redistributing newly fixed 350 nitrogen, which is directly released as NH4 + or regenerated as NH4 + from the decomposition of released dissolved organic nitrogen (Glibert and Bronk, 1994;Mulholland et al., 2004;Newell et al., 2011). During this study, the seasonal low in dust deposition to the North Atlantic may have shifted resource constraints towards phosphate-Fe co-limitation, with implications for the extent and rate of N-fixation, and speculatively extending to the subsequent processes of Nremineralisation.

355
A Primary Nitrite Maximum (PNM) was evident over the entire transect (Fig. 3e). A range of 0.05-0.50 µmol L -1 was measured, although sampling resolution may have limited the extent to which this feature was resolved (Beman, 2012;Meeder et al., 2012). Evidence suggests that nitrification has an important role in the formation and maintenance of the PNM against diffusive forces (Peng et al., 2018;Beman et al., 2012;Newell et al., 2013;Buchwald and Casciotti, 2013;Santoro et al., 2013;Meeder et al., 2012), with potential contributions from NO2release by phytoplankton 360 (Beman et al., 2012;Santoro et al 2013). This study is unable to inform the debate surrounding PNM formation as the isotope dilution approach adopted here does not discriminate between the processes contributing to NO2regeneration (ammonium oxidation, algal NO2release and potentially a photolytic route which produces NO2from NO3 - (Zafiriou and True, 1979) and humic substances (Kieber and Seaton, 1999)). The release of NO2by phytoplankton is most frequently associated with an intracellular resource imbalance linked to dynamic transitions (e.g. vertical mixing across 365 a light gradient; Santoro et al., 2013;Lomas and Lipschultz, 2006;Meeder et al., 2012) which indicates that ammonium oxidation likely dominated the contribution to RNO2. Across all depths, the oxidation of NH4 + lead to an average NO2turnover of 38.5 days (range 0.4-335 days) within the North Atlantic Gyre compared to an average of 37 days (range 2.9-160 days) for the South Atlantic Gyre. For the PNM, a range of 3-40 days was reported within the California current system (Santoro et al., 2013) and 33-178 days in the Arabian Sea (Buchwald and Casciotti, 2013).

370
The dissolved oxygen concentration was generally observed to be 200-300 µmol L -1 , with variability linked to seawater temperature (Fig. 4d). Dissolved oxygen concentration has been shown to structure the activity and distribution of nitrifying organisms (Ward, 2002, Beman et al., 2008, especially within low oxygen environments (Beman et al., 2008;Bouskill et al., 2012). A low oxygen feature of the 0.1 % sPAR equatorial upwelling water was distinguished in a series of correlation plots (Fig. 4 a-c), linked to relatively low temperature and salinity in 375 combination with a very high NO3concentration. This well documented oxygen minimum region is related to large scale circulation causing weak ventilation (Kartensen et al., 2008). This relative extreme in a combination of physico-9 chemical conditions did not result in distinct elevations of RNO2 (Fig. 4d). Similar results were noted by Shoizaki et al., (2016) in the North Pacific Ocean, where dissolved oxygen concentrations approaching 50 µmol L -1 were not associated with an enhanced rate of nitrification. Nitrification within low-oxygen environments is of interest as it links N-inputs to 380 marine environments via N-fixation to N-losses via denitrification and anaerobic ammonium oxidation (annamox), processes which are characteristically associated with marine regions of extremely low oxygen concentration (Francis et al., 2007;Ward et al., 1989).

Biological context
Profiles of biological parameters (chlorophyll-a, pico-eukaryote and bacterial abundance) demonstrated apparent 385 vertical structure. The chlorophyll-a distribution indicated that photosynthetic cells were present throughout the depth range used for RNO2 measurements. The presence of measurable chlorophyll-a implied the potential production and release of dissolved organic matter, in addition to organic matter released upon cell rupture (via grazing or viral attack). In contrast to the deep chlorophyll-a and picoplankton abundance maxima, bacterial cell profiles tended to describe higher abundance towards the well-lit region of the water column. Assuming that bacterial cell abundance reflected 390 resource availability, the observed distribution implied the availability of photosynthetically derived labile DOM throughout the photic zone (noting that the DCM represents only a small fraction of water column productivity and biomass; Marañon et al., 2000), and consequently the regeneration of NH4 + . While the subsequent fate of regenerated NH4 + would relate to environmental factors that shape the open ocean ecosystem, molecular evidence supports the potential for nitrifying activity throughout the surface ocean (Church et al., 2010;Santoro et al., 2010).

RNO2, measurements within the study.
Ammonium oxidation was consistently measured at three sPAR depths within the physical, chemical and biological context described above ( Fig. 6; Fig. 7a-c), indicating that NH4 + was available to support an oxidation rate of 0.02-13.27 nmol L -1 d -1 . When integrated between the surface and 0.1% sPAR, the range was 0.01-0.91 mmol m -2 d -1 . In their study of the tropical eastern Pacific and Gulf of California, Beman et al., (2012) report a range of 1.38-7.72 mmol m -2 400 d -1 in the upper 300 m although only 1.5-42 % of integrated ammonium oxidation was associated with depths shallower than 1 % sPAR. Assuming a range of 10-100 nmol L -1 NH4 + at sampling depths throughout the transect, a range consistent with historical measurements (Clark et al., 2008; British Oceanographic Data Centre data archives; www. BODC.ac.uk), a specific rate of 0.011-0.113 d -1 was derived. This basin-scale range is comparable to that typically applied in biogeochemical models, reported to be 0.02 to 0.1 d -1 (Moore et al., 2002;Wang et al., 2006;Christian et al., 405 2002;Jiang et al., 2003), although in their global model Yool et al. apply a value of 0.2 d -1 (Yool et al., 2007).
The vertical distribution and magnitude of RNO2 rates were consistent with previous observations, whereby low but detectable rates were measured within the well-lit ocean, a peak in ammonium oxidation rates was measured at 1 % sPAR, and a decline in rates was measured as depth increased below 1 % sPAR ( Fig. 6; 7d, e; Al-Qutob et al., 2002;Santoro et al., 2013;Shiozaki et al., 2016;Newell et al., 2013;Peng et al., 2018;Beman et al., 2012). The causes for 410 this distribution are a continuing subject of debate, but ultimately reflect a balance between multiple factors including the provision of labile organic matter supporting the regeneration of NH4 + , competition for this resource from other microbes including photosynthetic cells, and grazing pressure by mixotrophs and heterotrophic protists upon the nitrifying community.
A striking feature of the RNO2 data set was the 1 % sPAR rate range associated with south Atlantic stations (Fig. 7d), 415 the mean of which was significantly greater than that measured in the North Atlantic Gyre. Indeed, by extrapolating vertically integrated values across a comparable distance into each hemisphere from the equator, we derive that the amount of NH4 + -N oxidised to NO2was 0.36 mol N d -1 over a 1 m x 4457 km distance north of the equator and 0.72 mol N d -1 over a 1 m x 4637 km distance south of the equator. We are unable to conclusively explain this novel distinction between hemispheres. However, it is recognised that the Atlantic gyres have persistent biogeochemical 420 differences in dust deposition, the inorganic nutrient regime and biological features such as N-fixation (Baker and Jickells, 2017;Moore et al., 2009;Ussher et al., 2013) that likely shape and influence broader biological activity within the surface ocean of each Atlantic gyre. Seasonality may be an additional component that contributes towards this hemispheric distinction. In their analysis of multiple AMT transects, Poulton et al. (2006) noted that high rates of carbon fixation at the sub-surface chlorophyll-a maximum characterizing SATL stations during October were 425 potentially linked to the onset of spring conditions. At this time, nutrients supplied by winter mixing to the sub-surface https://doi.org/10.5194/bg-2021-184 Preprint. Discussion started: 9 August 2021 c Author(s) 2021. CC BY 4.0 License. chlorophyll-a maximum in combination with basin scale changes in irradiance, support the growth of larger eukaryotic phytoplankton (Letelier, 2004). The associated seasonal production of organic material would inevitably lead to an increase in decomposition and inorganic nutrient regeneration (including nitrification). Multivariate analysis, which identified an association between ammonium oxidation and chlorophyll-a, light duration and nutrients, provided support 430 for this association, as discussed below.
Multivariate approaches were used to investigate potential structure within the data and to examine relationships between ammonium oxidation and environmental variables. However, it is first worth considering limitations of the data set used for this analysis. While there is strong evidence to suggest that ammonium oxidation reflected ammonium oxidation activity rather than alternative sources of NO2 -, the measurement does not distinguish between archaeal or 435 bacterial nitrifiers, which likely dominate separate niches within depth profiles (Benman et al., 2012). Nor does it consider the influence of changes in nitrifier community composition along the extended transect. Microbial cell abundance is no indication of the activity of specific microbial groups, or their importance to biological processes.
Chlorophyll-a concentration does not reflect changes in species composition, or necessarily biomass, which are known to change in response to environmental forcing within the Atlantic Ocean (Poulton et al 2006). Within these limitations,

440
the PCA ordination of the entire data set (Fig. 9) demonstrated greater similarity within a sampling depth, equating to horizontal scales of hundreds of kilometres, than between sampling depths equating to vertical scales of tens of meters. A similar result was previously observed , implying that distinct niches exist between depth horizons.
Statistical analysis aimed to identify the subset of explanatory variables that, in combination, best matched the 445 observed pattern of ammonium oxidation rates. This match was provided by a combination of chlorophyll-a concentration, light duration and silicate concentration. A similar outcome was reported by Shiozaki et al., (2016) in their study of the oligotrophic North Pacific, in which a correlation between the activity of photosynthetic cells and ammonium oxidation was reported, although this relationship was not consistently observed at all sampling stations.
Chlorophyll-a and light duration as explanatory variables were common to all three sampling depths while depth-450 specific relationships incorporated additional variables. These results implied that, either directly or indirectly, ammonium oxidation responded to a product of actively photosynthesising cells (as opposed to other microbes identified by AFC). The production and release of labile dissolved organic matter during photosynthetic activity and its subsequent decomposition to regenerate NH4 + is an intuitive (and potentially spatially/temporally close) link between photosynthetic cells and nitrifying organisms that could be supported by this insight. However, a role for silicate has not 455 previously been reported, potentially highlighting a link to diatoms. Diatoms are relatively large, mineralising phytoplankton capable of rapid growth that contribute significantly to organic matter production and export in oligotrophic environments (Klass and Archer, 2006;Scharek et al., 1999). Alternatively, and perhaps more likely, silicate may be viewed as a tracer for sporadic nutrient inputs from below the nutricline (to include NO3 -, PO4 3-) that stimulate short-term growth of photosynthetic cells in oligotrophic environments (Poulton et al., 2006). In contrast to 460 other resources supplied by vertical mixing (e.g. NO3 -, PO4 3-), silicate is only utilised by diatoms which represent a fraction of the microbial community. It is reasonable to expect that following a localised mixing event, elevated silicate concentrations persist for longer than the simultaneously introduced NO3and PO4 3-, which are drawn down due to localised increases in primary productivity. Silicate effectively becomes a tracer for water that hosts the products of active or recently enhanced productivity, with subsequent decomposition and nutrient regeneration.

465
The current view of low nutrient environments sees a dominant role for the picoplankton community which supports the microbial food web, with the sporadic growth of large phytoplankton following episodic nutrient inputs due to localised short-term physical instability (Carpenter et al., 1999;Scharek et al., 1999;Dandonneau et al., 2003;Poulton et al., 2006). Increased nutrient availability favours larger and faster growing cells that outcompete the picoplankton to transiently dominate the microbial community in open ocean environments (Kiørbe, 1993). Organic matter released 470 during such transient events (directly from living cells or following cell lysis) would support a short-term localised increase in N-regenerating activity, including ammonium oxidation. Evidence indicates that archaea, which frequently dominate the euphotic zone nitrifying community, possess a high resource affinity and respond rapidly to NH4 + pulses in culture (Martens-Habbena et al., 2009;Beman et al., 2012). Whether the statistically significant link between rates of ammonium oxidation and a combination of chlorophyll-a, light duration and silicate relates to diatoms specifically 475 (which have an absolute requirement for silicate and are obligate phototrophs i.e. they do not express mixotrophy unlike most other photosynthetic plankton, a nutritional mode that affords flexibility in resource limited environments and decreases dependence upon photosynthetic CO2 fixation; Mitra et al., 2014) or more likely relates to a broader stimulation of transient micro-plankton growth remains to be tested. Should nitrifying activity be responding to a chain of events in the very recent past, instantaneous rates of nitrification may not be fully explained by routine measurements 480 of extant environmental conditions, such as those used in the present study. A detailed examination of labile DON production and decomposition may be required in combination with studies of nitrification.
Conspicuously prominent ammonium oxidation rates were associated with stations 28 °S and 39 °S, with measured ranges of 1.5-8.6 nmol L -1 d -1 and 1.3-13.3 nmol L -1 d -1 respectively ( Fig. 4d; 7a-c). Selected biogeochemical characteristics further distinguished station 39 °S (high P*, O2 concentration, lower salinity and temperature; Fig. 3; 4) 485 although those of station 28 °S were typical of the province. The use of Earth Observation data (sea surface temperature, ocean colour) revealed no anomalous features co-incident with these stations. However, Sea Surface Height Anomaly (SSHA) data evidenced that 28 °S and 39 °S were associated with mesoscale eddy features at the point of sampling (Fig. 9). Analysis revealed that no other stations of this study coincided with discernible mesoscale eddy features at the point of sampling.

490
Eddy 39 °S was locally generated. In contrast, eddy 28 °S likely originated within the Agulhas Retroflection, traversing and potentially influencing SAG biogeochemistry before dissipating on the Brazilian shelf some 3.5 years later. Nencioli et al., (2018) tracked similar features, showing that they provided a coherent transport of water masses, preserving some of the physical and biological signals over more than two years. The vertical flux of inorganic nutrients induced by the dynamics of traversing eddies has been argued to balance the Sargasso Sea nutrient budget (McGillicuddy et al., 1998). Similarly, the NO3deficit observed between new production requirements and diffusive NO3supplies in the Atlantic Gyres may be reconciled to some extent by eddy dynamics (Planas et al., 1999). While the nutrient regime of 28 °S was comparable to the provincial average, elevated ammonium oxidation (some 2-6 times higher than the provincial average) implied unusual biological activity; such activity is not sustained in isolation, but must couple (over appropriate scales of space/time) to the activity of other biological processes (e.g. nutrient 500 acquisition, photosynthetic growth, DON production and decomposition, NO2oxidation), presumably supported by new nutrients physically introduced into the photic zone by eddy dynamics. Unfortunately, given that only 3 ammonium oxidation rate observations were made at station 28 °S, located at the edge of a feature which measured approximately 300 km in diameter, little can be said about how representative of the feature this activity is, the relevance to broader nitrogen cycling activity or its significance as a source of atmospheric N2O (Dore and Karl, 1996).

Conclusions
This study reported relatively high-resolution measurements of ammonium oxidation rate at the oceanic basin scale. Ammonium oxidation took place at all locations investigated, adding to evidence that this process is widely distributed and active within the sunlit ocean. Considering all rate data, results implied that more active nitrogen cycling took place in the South compared to the North Atlantic Gyre. While seasonality may contribute towards this hemispheric 510 distinction, we speculate that this is a persistent feature. Historical AMT data sets demonstrate the broad inter-annual consistency of biogeochemical distributions and features to which ammonium oxidation rate significantly related. Should this testable hypothesis be substantiated, it would allude to a broader biogeochemical distinction between Atlantic gyres as the process of ammonium oxidation is intrinsically linked to other microbial processes that cycle resources. By extension, as chemolithoautotrophs, nitrifiers fix CO2 concurrently with nitrifying activity. Nitrifier 515 mediated C-fixation in the South Atlantic Gyre would also exceed that of the North. A statistical analysis of observational data indicated that ~65% of variability in ammonium oxidation rate was explained through co-variability with environmental variables. A previously observed link to photosynthetic cells (Shiozaki et al., 2016) was supported here, and summarised as a significant relationship between ammonium oxidation rate and a combination of chlorophyll-a concentration, light duration and silicate. We speculate that as an indication of recent physical instability, 520 silicate transiently marks seawater that hosts the products of active or recently enhanced photosynthetic productivity, with subsequent organic matter decomposition and nutrient regeneration. This leads us to conclude that a role for DOM, and specifically the short-term production, release and decomposition of labile material is amongst the factors contributing to the unexplained variability in ammonium oxidation.
Two mesoscale eddies supported very high rates of ammonium oxidation, an observation previously unreported.

525
Although the limited number of observations do not permit a meaningful analysis, results highlight the possibility that eddies which traverse the South Atlantic Gyre may disproportionately influence mixed layer biogeochemistry and ocean-atmosphere gas exchange.

Author contribution
DRCdesigned experiments, undertook sample and data analysis, written manuscript; APRconducted N-cycle 535 fieldwork investigations, contributed to manuscript preparation; CF -conducted N-cycle fieldwork investigations; CHconducted inorganic nutrient analysis of fieldwork samples; LAmass spectrometry sample analysis; PJS -statistical analysis, contributed to manuscript preparation; GDQ -satellite data analysis, contributed to manuscript preparation; SGsatellite data analysis; GL -data analysis, contributed to manuscript preparation; GT-conducted flow cytometry analysis of fieldwork samples, contributed to manuscript preparation.    Fig. 1). Open circles represent 0.1% sPAR data, highlighting distinct characteristics associated 810 with specific NATR/WTRA/SATL sampling points. To represent the distribution of rate data, the 88 RNO2 measurements were arranged in ascending order (d, insert has linear scale) with an indication of sampling depth, or the association with a specific feature (low oxygen, defined as < 150 µmol L -1 , or mesoscale eddy). Measurements within eddy features tend towards the high end of the scale, as do measurements at 1% sPAR. Low oxygen measurements show no clear distributional trend while 14% and 0.1% sPAR values tend towards the low end of the scale.   Fig. 1). Data are presented as a ratio of the maximum value measured for each parameter throughout the transect. Profiles frequently describe comparable distributions within a defined province.