Distribution and rates of nitrogen fixation in the western tropical South Pacific Ocean constrained by nitrogen isotope budgets

: Constraining the rates and spatial distribution of di-nitrogen (N2) fixation fluxes to the ocean informs our understanding of the environmental sensitivities of N2 fixation as well as the timescale over which the fluxes of nitrogen (N) to and from the ocean may respond to each other. Here we quantify rates of N2 fixation as well as its contribution to export production along a zonal transect in the Western Tropical South Pacific (WTSP) Ocean using N isotope (δ15N) budgets. Comparing measurements of water column nitrate + nitrite δ15N with the δ15N of sinking particulate N at a western, central, and eastern station, these δ15N budgets indicate high, modest, and low rates of N2 fixation at the respective stations. The results also imply that N2 fixation supports exceptionally high, i.e., > 50 %, of export production at the western and central stations, which are also proximal to the largest iron sources. These geochemically-based rates of N2 fixation are equal to or greater than those previously reported in the tropical North Atlantic, indicating that the WTSP Ocean has the capacity to support globally significant rates of N2 fixation, which may compensate for N removal in the oxygen deficient zones of the eastern tropical Pacific.


Introduction
The primary source of nitrogen (N) to the ocean is the biologically-mediated reduction of di-nitrogen (N 2 ) gas to ammonia, which is then assimilated into the biomass of the organisms carrying out this process, known as diazotrophs (Gruber, 2004). While the distribution and rates of this process in the ocean play a central role in regulating the fertility and community structure of marine ecosystems, these 5 first-order properties of marine N 2 fixation remain poorly constrained. Historically, the highest rates of N 2 fixation in the global ocean have been associated with the tropical North Atlantic (Mahaffey et al., 2005;Sohm et al., 2011). The high 15 N 2 incubation-based N 2 fixation rates observed in the tropical Atlantic (Luo et al., 2012) are consistent with both the preference of diazotrophs for warm waters (Breitbarth et al., 2007;Stal, 2009) as well as the high atmospheric dust flux to the region (Mahowald et 10 al., 2009;Prospero, 1996) that helps fulfil the high iron requirement of the enzyme, nitrogenase, carrying out N 2 fixation (Berman-Frank et al., 2001;Kustka et al., 2003). Additionally, the elevated ratio of nitrate (NO 3 -) to phosphate (PO 4 3-) concentrations (Gruber and Sarmiento, 1997) and low δ 15 N-NO 3 - (Knapp et al., 2008) in the upper thermocline of the North Atlantic are attributed to high regional N 2 fixation rates, and have supported the hypothesis that iron availability plays a key role in regulating 15 the spatial distribution of N 2 fixation in the ocean (Moore et al., 2009;Moore and Doney, 2007) ("δ 15  While the highest inputs of N to the ocean have traditionally been associated with the North Atlantic, it 20 has also been argued that this association results from the significant sampling bias in favor of the tropical Atlantic (Sohm et al., 2011), with large regions of the South Pacific and Indian Ocean undersampled with respect to direct N 2 fixation rate measurements (Luo et al., 2012). More recently, the Eastern Tropical South Pacific (ETSP) has seen increased sampling due to nutrient distribution-based modelling predictions that the highest global N 2 fixation rates would be found in surface waters above 25 and adjacent to oxygen deficient zones (ODZs), where significant phosphorus (P) would be available to support N 2 fixation (Deutsch et al., 2007). However, field campaigns have found exceedingly low rates of N 2 fixation in the ETSP gyre (Knapp et al., 2016a;Raimbault and Garcia, 2008) (Moutin et al.,4 2008), which have been attributed to limited iron availability (Dekaezemacker et al., 2013).
Consequently, existing measurements indicate that the dominant sinks for N in the ocean, benthic and water column denitrification and anaerobic ammonium oxidation, focused in the ODZs of the eastern tropical Pacific and Arabian Sea (Gruber and Galloway, 2008), are spatially segregated from the dominant N 2 fixation inputs in the tropical Atlantic. This spatial decoupling of N inputs and outputs 5 necessarily corresponds to a temporal decoupling, requiring the time scale of ocean circulation for N 2 fixation to respond to changes in rates of denitrification, and vice versa. In spite of the apparent spatial decoupling in the modern ocean, paleoceanographic evidence indicates that N fluxes to and from the ocean have been closely balanced over >20 kya, requiring feedbacks in the N cycle to operate on time scales shorter than ocean circulation, and thus implying a tighter spatial coupling of N sources and sinks 10 (Brandes and Devol, 2002;Deutsch et al., 2004). While N loss in the ocean is constrained to suboxic sediments and water column ODZs, similar constraints on the location of the largest N 2 fixation fluxes to the ocean are lacking, and thus the degree to which marine N sources and sinks have been coupled through time remains uncertain. 15 While prior modelling analyses emphasized the importance of iron or phosphorus in supporting N 2 fixation, the most recent modelling studies reflect the importance of elevated surface temperatures, adequate iron, and the potential for low surface ocean NO 3 -:PO 4 3concentration ratios to support a unique ecological niche for diazotrophs (Dutkiewicz et al., 2012;Monteiro et al., 2011;Weber and Deutsch, 2014). Attention has consequently shifted to the relatively undersampled Western Tropical 20 South Pacific (WTSP) Ocean, where atmospheric dust fluxes to warm surface waters are higher than in the central and eastern tropical South Pacific (Mahowald et al., 2009), and where surface ocean NO 3 and PO 4 3concentrations and ratios are relatively advantageous for diazotrophs (Moutin et al., 2005;Van Den Broeck et al., 2004 (Deutsch et al., 2007). Additionally, early remote sensing work detected significant and persistent 5 blooms of Trichodesmium spp. in the WTSP (Dupouy et al., 2000), consistent with more recent direct observations of elevated Trichodesmium spp. abundance and N 2 fixation rates observed near Melanesian islands (i.e., New Caledonia, Vanuatu, and Fiji) (Shiozaki et al., 2014;Yoshikawa et al., 2015) and in the Solomon Sea . These high Trichodesmium spp.
abundances and N 2 fixation rates have been attributed to sea surface temperatures >25 °C and 5 continuous nutrient inputs of terrigenous and volcanic origin (Labatut et al., 2014;Radic et al., 2011).
Prior molecular work has also shown higher rates of N 2 fixation in the WTSP at locations where surface ocean dissolved iron (DFe) concentrations were higher and where Trichodesmium spp. were less stressed for iron (Chappell et al., 2012). Together, these observations and modelling-based predictions highlight the potential for significant N 2 fixation rates in regions of the WTSP where diazotrophs can 10 meet their iron and phosphorus requirements.
Here we use geochemical tools to quantify rates of N 2 fixation along a zonal transect in the WTSP where surface waters are >25 °C, have favourable macronutrient concentrations and ratios, and where DFe concentrations are an order of magnitude higher than in the South Pacific Gyre, and are mainly 15 attributable to shallow hydrothermal input (Guieu et al., under review Hoering and Ford, 1960;Minagawa and Wada, 1986). In contrast, in the Pacific, NO 3 mixed up from the subsurface is impacted by water column denitrification and can have a NO 3 δ 15 N >20‰ (e.g., (Brandes et al., 1998;Casciotti et al., 2013;Rafter and Sigman, 2016)), although as upper thermocline waters move westward in the Pacific, the very high NO 3 δ 15 N signal is diluted and typical values are between 5 and 10‰ (Rafter et al., 2013). The relative importance of each source for 25 supporting export production can be determined using the two end-member mixing model described in Eqn. 1 ("δ 15 N budget") where the fractional importance of N 2 fixation for supporting export production (x) is defined as: Rearranging and solving for x yields: Multiplying the fraction of export production supported by N 2 fixation (x) by the PN sink mass flux provides a time-integrated N 2 fixation rate that can be compared with 15 N 2 incubation-based N 2 fixation rate measurements (Knapp et al., 2016a). Here it is hypothesized that both rates of N 2 fixation and its 10 importance for fuelling export production will be higher at stations in the western vs. central and eastern regions of the WTSP because of their closer proximity to iron sources (Guieu et al., under review).

Sample collection
Sampling for the Oligotrophic to UlTra-oligotrophic PACific Experiment ("OUTPACE") cruise was 15 conducted on the R/V L'Atalante, which left Noumea, New Caledonia on 18 February 2015 and arrived in Papeete, Tahiti, on 2 April 2015. This cruise followed a roughly zonal transect along 18 to 19 °S between 159 °E and 160 °W. Details of the cruise and experimental design are described comprehensively in , but briefly, sediment traps were deployed at three "Long Duration" (LD) stations A, B, and C (see Table 1 for station locations). Water column samples were 20 collected from Niskin bottles deployed on a CTD-rosette and water was stored at -20 ºC in HDPE bottles for analysis on land.

Sinking particulate N flux and δ 15 N measurements
Surface-tethered floating particle-interceptor traps (PPS5) were deployed on the OUTPACE cruise at 150, 330 and 520 m for ~5 days at Stations LDA and LDB, and at 150 and 330 m at LDC (Moutin et al., 5 2017). The mass flux ("PN sink flux") and δ 15 N of the PN sink flux was determined by combustion-GC interfaced to an isotope ratio mass spectrometer at the Mediterranean Institution of Oceanography with a lower detection limit of 2.2 µg N and precision of ± 0.3‰ for 80 µg samples, with a precision of ± 1.0‰ for 10 to 20 µg samples typical of what was collected in the sediment traps at the LD stations. and increased with depth, consistent with prior regional observations (Garcia et al., 2014) (Fig. 1). All nutrient concentration data are available at: http://www.obs-vlfr.fr/proof/php/outpace/outpace.php. 15 Water column profiles of thermocline NO 3 -+NO 2 δ 15 N show similar trends at the LD stations, with 650 m NO 3 -+NO 2 δ 15 N ~7‰, increasing to ~8.5‰ at 400 m ( Fig. 1), which fall within the range of previous regional measurements (Yoshikawa et al., 2015). The elevation of thermocline NO 3 -+NO 2 δ 15 N relative to the mean ocean NO 3 -+NO 2 δ 15 N of 5‰ is attributed to denitrification and/or anammox occurring in the ODZs of the ETSP, where thermocline NO 3 δ 15 N can exceed 20‰ (e.g., (Altabet et al., 2012;20 Casciotti et al., 2013)). The average, mass-weighted δ 15 N of the PN sink flux collected in the 150 m trap increased from the western to eastern stations, from 0.6 ± 1.0‰ at LDA, to 3.1 ± 1.0‰ at LDB, and to 7.7 ± 1.0‰ at LDC (Table 1) (Fig. 1).

Results of the δ 15 N budget: N 2 fixation rates and their contribution to export production
Estimates of N 2 fixation rates and their contribution to export production determined using δ 15 N budgets 25 include the quantitatively dominant fluxes of N into and out of the surface ocean. Here, the dominant 8 fluxes of N into the surface ocean include subsurface NO 3 and newly fixed N introduced from diazotrophs, and the dominant loss term is represented by the PN sink flux (Eq. 1). In the event that total  to 57 ±12%, and 0 to 8 ±11% of export production supported by N 2 fixation at stations LDA, LDB, and 25 LDC, respectively (Table 1). Multiplying the fractional importance of N 2 fixation by the PN sink mass flux yields a range of estimated N 2 fixation rates of 219 to 290, 11 to 21, and 0 to 9 µmol N m -2 d -1 at 9 stations LDA, LDB, and LDC, respectively (Table 1), where the range includes uncertainty in both the PN sink δ 15 N measurement as well as the NO 3 -+NO 2 δ 15 N end-member.
These geochemically-derived rates are lower than those measured by in situ 15 N 2 incubations at the same OUTPACE stations, with depth-integrated average N 2 fixation rates of 593 ± 51, 706 ± 302, and 5 59 ± 16 µmol N m -2 d -1 at LDA, LDB, and LDC, respectively (Caffin et al., 2017). Previous work has also found lower δ 15 N budget-derived N 2 fixation rates relative to 15 N 2 incubation-based N 2 fixation rates (Knapp et al., 2016a). To the extent that sediment traps under collect the export flux, the two different metrics of N 2 fixation may be reconciled by multiplying "x" from Eq. 2, the fractional importance of N 2 fixation for export production, by other metrics of new or export production such as 10 O 2 /Ar ratios, 234 Th deficits, or 14 C uptake rates (Knapp et al., 2016a). This explanation may reconcile the δ 15 N budget and 15 N 2 incubation-based N 2 fixation rate estimates at LDA, which differ by a factor of ~2.5, and potentially the rates at LDC as well, which, while they differ by a factor >6, both correspond to relatively low N 2 fixation rates. However, the δ 15 N budget and 15 N 2 incubation-based N 2 fixation rates observed at LDB, 11 to 21 and 706 µmol N m -2 d -1 , respectively, are more difficult to reconcile 15 based on sediment trap under-collection alone, and may be partially attributable to variability encountered while sampling at the end of a phytoplankton bloom at that station (de Verneil et al., 2017).
We note that the zonal trend in increasing PN sink δ 15 N to the east is similar to a zonal gradient in .036 mmol N m -2 d -1 , respectively, which is unexpected given the more typical mass flux attenuation with depth observed at LDA and LDC, as well as elsewhere in the ocean (Martin et al., 1987). This unusual trend in mass flux with depth suggests either non-steady state sinking flux conditions and/or a 10 problem with sample collection at LDB. Regardless, using the 14 C-uptake based estimate of net community production at LDB, 1.91 mmol N m -2 d -1 , instead of the PN sink mass flux to multiply "x" from Eq. 2 by yields an N 2 fixation rate of 2300 µmol N m -2 d -1 . These significant disparities in productivity metrics and resulting N 2 fixation rates at LDB suggests the potential for temporal decoupling of production and export and/or the underestimation of the export flux by the sediment trap, 5 and indicate that N 2 fixation rates are probably higher than those resulting from δ 15 N budget calculations based on the mass flux to the 150 m trap at LDB. Regardless, we take the zonal trend in PN sink δ 15 N to indicate a decreasing contribution from N 2 fixation to export from the west to the east to be robust as it is consistent with both the PN susp δ 15 N measurements as well as the broad trends in 15 N 2 incubation-based N 2 fixation rate estimates that decrease from the west to east. 10 Comparing the absolute magnitude of the δ 15 N budget-based N 2 fixation rates with previous measurements, we find that the 219 to 290 µmol N m -2 d -1 rate estimated for LDA represents a significant N 2 fixation rate relative to prior global measurements (Luo et al., 2012), in particular if it should be revised upwards to account for the under-collection of the export flux by the sediment trap. In 15 contrast, the estimated rate range at LDB, 11 to 21 µmol N m -2 d -1 , is quite low, as is the range of 0 to 9 µmol N m -2 d -1 at LDC, and both of these rates are broadly similar to the rates previously measured in the ETSP (Knapp et al., 2016a;Moutin et al., 2008;Raimbault and Garcia, 2008). Similarly, the δ 15 Nbudget based estimate of the contribution of N 2 fixation to export production at LDC is low and similar to previous δ 15 N-budget measurements in the North Pacific (Casciotti et al., 2008) and North Atlantic 20 (Altabet, 1988;Knapp et al., 2005). However, the fractional contribution of N 2 fixation to export production at LDA, 80 to 83%, is higher than all previous δ 15 N budget results. The contribution of N 2 fixation to export production at LDB, 50 to 57%, is also notably high. While the previous δ 15 N budgets of (Karl et al., 1997) and (Dore et al., 2002) found evidence for ~50% of export production supported by N 2 fixation near Hawaii, newer methods capable of measuring the NO 3 -+NO 2 δ 15 N at the lower NO 3 -25 +NO 2 concentrations found in the upper thermocline that represent a more realistic estimate of the endmember NO 3 source suggest that N 2 fixation may support closer to 25% of export during the summer in the North Pacific gyre (Bottjer et al., 2017;Casciotti et al., 2008). Consequently, the findings of 50 to 11 57% and 80 to 83% of export production being supported by N 2 fixation at stations LDB and LDA, respectively, indicates that N 2 fixation plays a significant role supporting carbon fixation and export production in this region of the WTSP, consistent with the high e-ratios (up to 9.7) reported by (Caffin et al., 2017). Direct export of diazotrophs has been reported by (Caffin et al., 2017), but most export is likely indirect, i.e., after the transfer of diazotroph-derived N to non-diazotrophic plankton, that is 5 subsequently exported (Caffin et al., This issue), as has been observed elsewhere in the WTSP (Bonnet et al., 2016;Knapp et al., 2016b).

Environmental sensitivities of N 2 fixation and the basin-scale coupling of N sources and sinks
The zonal gradient in both N 2 fixation rates as well as their contribution to export production in the 10 OUTPACE study supports emerging hypotheses regarding the controls on the distribution of marine N 2 fixation fluxes in the global ocean. Specifically, the low rates of N 2 fixation documented in this study at LDC and in the ETSP (Knapp et al., 2016a) indicate that low NO 3 -:PO 4 3concentration ratios in the absence of adequate iron (Blain et al., 2008;Fitzsimmons et al., 2014) are insufficient to support significant fluxes of new N to the ocean. Instead, the results presented here are consistent with recent 15 modelling work that has included both the high iron requirements of diazotrophs as well as the potential for low NO 3 -:PO 4 3concentration ratios to support elevated diazotroph abundance and N 2 fixation inputs to the ocean (Dutkiewicz et al., 2012;Monteiro et al., 2011;Weber and Deutsch, 2014 incubation-based N 2 fixation rates . 12 However, prior to the OUTPACE cruise, our knowledge of DFe concentrations and their sources in the WTSP was limited, especially in the western and central sectors. During OUTPACE, Guieu et al. (Under review) reported high DFe concentrations in the western sector of the WTSP (from 160 °E to 165 °W, average 1.7 nM within the photic layer), i.e., significantly (p<0.05) higher than those reported in the eastern sector (165 °W to 160 °W, average 0.3 nM within the photic layer). The high DFe 5 concentrations measured in the west were previously undocumented, and reveal several maxima (>50 nM), suggesting significant iron inputs to this region. (Guieu et al., Under review) found that atmospheric deposition in this region was too low to explain the observed DFe concentrations in the water column, and that the iron in the euphotic layer may instead derive from shallow (~500 m) hydrothermal sources associated with the Tonga-Kermadec subduction zone. 10 Recent studies performed in the western end of the WTSP in the Solomon, Bismarck Bonnet et al., 2009;Bonnet et al., 2015) and Arafura (Messer et al., 2015;Montoya et al., 2004) Seas also reveal extremely high N 2 fixation rates (>600 µmol N m -2 d -1 ), indicating that high N 2 fixation rates have been found over a significant region of the WTSP, extending west to east from Australia to 15 Tonga and north to south from the equator to 25 to 30 °S, or ~13 x 10 6 km 2 (i.e. ~20 % of the South Pacific Ocean area). These significant N inputs may offset the N loss occurring in the ODZs of the eastern tropical Pacific. The ability for marine N inputs and outputs to compensate for each other within the same ocean basin corresponds to a spatial and thus temporal coupling on the scale of years to decades, consistent with the paleoceanographic record (Brandes and Devol, 2002;Deutsch et al., 2004;20 Weber and Deutsch, 2014), and represents an intermediate view of the distribution of global marine N 2 fixation fluxes consistent with that proposed by (Weber and Deutsch, 2014) where iron availability controls local N 2 fixation rates but phosphorus availability regulates basin-scale N 2 fixation rates (Moutin et al., 2008 and this issue).

25
The goal of this study was to address the question: do regions other than the tropical Atlantic contribute significantly to global N 2 fixation fluxes? While our results should be taken as a "snapshot" view that cannot necessarily be scaled up to annual fluxes, at stations proximal to iron sources, geochemically-13 derived N 2 fixation rates of 219 to 290 µmol N m -2 d -1 were observed, and could potentially represent a lower bound of N 2 fixation rates due to the potential under-collection of the PN sink flux by sediment traps. Moreover, at stations LDA and LDB, separated by ~27º longitude, N 2 fixation was found to support >50% of export production, a finding that has not been replicated elsewhere with sensitive NO 3 -+NO 2 δ 15 N methods to our knowledge. Together with similar findings from 15 N 2 uptake experiments, 5 these results suggests that N 2 fixation can support a significant fraction of export production over a large region of the WTSP. At the eastern station most distant from iron sources, both rates and the contribution of N 2 fixation to export production were low, ~0 to 9 µmol N m -2 d -1 and 0 to 8%, respectively, similar to previous measurements in the ETSP where diazotrophs may also be challenged by iron availability (Dekaezemacker et al., 2013;Knapp et al., 2016a;Moutin et al., 2008). Significant 10 N 2 fixation fluxes in the WTSP may provide a means of balancing N loss occurring in the ODZs of the eastern tropical Pacific, and thus may help reconcile the paleoceanographic record requiring N inputs and losses to balance each other on time scales shorter than ocean circulation (Dutkiewicz et al., 2012;Monteiro et al., 2011;Weber and Deutsch, 2014). All data and metadata are available at the following web address: http://www.obs- 25 vlfr.fr/proof/php/outpace/outpace.php. The data supporting the conclusions of this paper may be obtained at the BCO-DMO database (# Pending), as well as the CCHDO site: