Sample collection was conducted as part of the GEOTRACES TONGA (shallow hydroThermal sOurces of trace elemeNts: potential impacts on biological productivity and the bioloGicAl carbon pump) research cruises (doi.org/10.17600/18000884) aboard the R/V L’Atalante in November 2019 and R/V Alis in October 2020, with both cruises leaving from and returning to Nouméa, New Caledonia. The 2019 primary cruise collected samples at 13 stations along a roughly zonal transect at ~20° S, sampling Melanesian waters (MW), the Lau Basin (LB), and crossing the Tonga-Kermadec Arc into the deeper South Pacific Gyre (SPG). In 2020, samples were collected at four stations in MW and the LB ( Figure 1 ). On both cruises, water column samples (n=200) for nitrate + nitrite ([NO₃ ^-+NO₂ ^-]) concentration and δ¹⁵N analysis were collected from Niskin or GoFlo bottles deployed on conductivity, temperature, and depth (CTD), TOW (small, classical CTD with 12 Niskin bottles), or trace metal clean (TMC) rosettes equipped with sensors. At discrete depths from each of the casts, 60 mL of 0.2 μm filtered seawater were collected in duplicate and stored in acid and deionized water-washed, sample-rinsed (three times) high-density polyethylene bottles. These samples were then immediately frozen at -20° C and subsequently sent to Florida State University for post cruise analysis.

Short-term Particle Interceptor Traps [PIT, collecting area of 0.0085 m², aspect ratio of 6.7, and filled with 0.2 μm filtered seawater with added formaldehyde brine (5% formaldehyde, final concentration) buffered with sodium tetraborate (pH 8)] were deployed on a drifting mooring in close proximity to the hydrothermal vents at 170 m and 270 m at station 5a-2019 for five days and at 270 m at station 10a-2019 for four days during the 2019 cruise, collecting sinking particulate N (“PN_sink”). Additionally, a long-term Technicap PPS5 trap (1 m² collecting area, aspect ratio of 5.3) was deployed (containing the same buffered formaldehyde brine solution described above) at station 12-2019/4-2020 at 1000 m, collecting samples every 14 days (bimonthly) for 12 months from November 2019 to October 2020. Although sediment traps are a standard tool used to capture sinking particles, uncertainties remain in their collection efficiency within the water column and between trap designs (Buesseler et al., 2007; Baker et al., 2020; Tilliette et al., 2023).

For the 2019 samples, [NO₃ ^-+NO₂ ^-] was determined using colorimetric analysis (Aminot and Kerouel, 2007) with a detection limit of 0.05 µM and a standard deviation (S.D.) of ± 0.1 μM. Furthermore, the [NO₃ ^-+NO₂ ^-] for the 2020 samples was measured by chemiluminescence (Braman and Hendrix, 1989) using a Thermo 42i NO_x analyzer at Florida State University. Briefly, samples were injected into a heated, acidic vanadium (III) solution that reduces NO₃ ^-+NO₂ ^- quantitatively to nitric oxide gas (NO_(g)). The NO_(g) then reacts with ozone inside the analyzer to produce light, the intensity of which is quantitatively related to the amount of NO_(g) in the sample and thus the original [NO₃ ^-+NO₂ ^-]. The range of detection of the instrument was adjusted according to the concentrations of the samples. Sample [NO₃ ^-+NO₂ ^-] was calibrated using a standard curve that bracketed the range of samples with a lower reporting limit of 0.1 µM and an average S.D. of ± 0.1 μM.

The nitrogen (N) isotopic composition of NO₃ ^-+NO₂ ^- (δ¹⁵N_NO3+NO2) was determined using the “denitrifier” method (Sigman et al., 2001; Casciotti et al., 2002; McIlvin and Casciotti, 2011; Weigand et al., 2016). This analysis was performed when sample [NO₃ ^-+NO₂ ^-] ≥ 0.3 μM. The δ¹⁵N_NO3+NO2 values were reported when the standard deviation of replicate analyses was <0.5‰. Samples were calibrated with IAEA N3 and USGS 34 as described in McIlvin and Casciotti (2011).

The bulk PN_sink mass flux, and its associated isotopic composition, “δ¹⁵N_PNsink”, collected by the sediment traps, was measured using an Elementar Analyser - Isotope Ratio Mass Spectrometer (EA-IRMS) at the Mediterranean Institute of Oceanography (SERCON INTEGRA 2). The Quantification Limit was 7 µg N and the precision was between ± 0.3‰ for highest masses and ± 3.5‰ for masses close to Quantification Limit (k = 2).

A two end-member mixing model was used to evaluate the δ¹⁵N budgets. This model assumes two quantitatively important “new” N sources to surface waters, subsurface NO₃ ^-+NO₂ ^- and biological N₂ fixation, as well as one loss term, PN_sink. The isotopic composition of N₂ fixation inputs, “δ¹⁵N_{N2 fix}” was assumed to be -1‰ (Hoering and Ford, 1960; Minagawa and Wada, 1986; Carpenter et al., 1997), while the δ¹⁵N of subsurface NO₃ ^-+NO₂ ^- and PN_sink were measured. The relative contribution of N₂ fixation to export production, “f_nfix”, is calculated by the following (Knapp et al., 2018):

which can be rearranged to solve for f_nfix:

The depth from which subsurface NO₃ ^-+NO₂ ^- is sourced likely varies and is difficult to constrain; therefore, the short-term δ¹⁵N budgets are evaluated over a range of subsurface δ¹⁵N_NO3+NO2 source values. These values include the shallowest δ¹⁵N_NO3+NO2 minima and the δ¹⁵N_NO3+NO2 of the sample collected immediately below the minima (Knapp et al., 2021). The long-term δ¹⁵N budgets are evaluated using the shallowest average δ¹⁵N_NO3+NO2 minima at station 12-2019/station 4-2020 as well as the average δ¹⁵N_NO3+NO2 of South Pacific Sub-tropical Under Water (SPSTUW, 150m – 250 m depth) at station 4-2020. Note, sampling resolution at station 12-2019 did not include SPSTUW. These δ¹⁵N_NO3+NO2 end-members for the moored trap δ¹⁵N bugdet were chosen to encompass the range of NO₃ ^-+NO₂ ^- likely entrained to surface waters over an annual cycle (Moutin et al., 2018). This seasonality is also observed at the station ALOHA, Hawaii (Casciotti et al., 2008; Böttjer et al., 2017). Once the fraction of export supported by N₂ fixation is calculated, an N₂ fixation rate, “R_nfix” can be calculated by multiplying f_nfix by the PN_sink mass flux, yielding a geochemically-derived N₂ fixation rate.

Bathymetric data from the sampling region were considered to better understand the bathymetric-induced current steering as well as the proximity of the stations to the shallow hydrothermal vents. These bathymetric data were downloaded from NOAA’s National Centers for Environmental Information page (https://www.ngdc.noaa.gov/mgg/global/). The ETOPO1 Global Relief Model was used with the grid version ETOPO1 Bedrock.

The [NO₃ ^-+NO₂ ^-] in the upper 100 m of the WTSP was ≤0.1 μM except at stations influenced by the hydrothermal vents, i.e., stations 5-2019 and 10-2019, where the [NO₃ ^-+NO₂ ^-] was 0.2 ± 0.0 μM to 0.7 ± 0.0 μM (average ±1 S.D.) between 60 m and 100 m ( Figures 3A, C ). At 400 m, the [NO₃ ^-+NO₂ ^-] increased to between 10 and 15 μM, corresponding to δ¹⁵N_NO3+NO2 ranging from 6 to 8‰ ( Figures 3B, D ). This is consistent with the presence of WSPCW (Tilliette et al., 2022) and more particularly, SPSTMW (Lehmann et al., 2018) and aligns with other studies in the region (Knapp et al., 2018). At station 8-2019, the [NO₃ ^-+NO₂ ^-] increased to 26.7 μM at 630 m with a corresponding δ¹⁵N_NO3+NO2 of 6.9 ± 0.1‰ ( Figures 3A, C ), consistent with AAIW (Talley et al., 2011; Lehmann et al., 2018; Tilliette et al., 2022). The majority of stations within MW and the LB (i.e., stations 1-2019 to 5-2019, 10-2019, and 12-2019) had δ¹⁵N_NO3+NO2 ranging from 4 to 6‰ at 200 m, decreasing shallower in the water column to 2 to 4‰ at 150 m associated with the transition between SPSTUW and surface waters ( Figures 3B, D ). The SPG stations (stations 6-2019, 7-2019 and 8-2019) had an average δ¹⁵N_NO3+NO2 of 6.8 ± 0.0‰ at ~200 m characteristic of SPSTUW. This δ¹⁵N_NO3+NO2 was significantly higher than the δ¹⁵N_NO3+NO2 at 200 m in MW and LB samples, where the average δ¹⁵N_NO3+NO2 at 200 m was 5.0 ± 0.6‰ and 5.1 ± 0.6‰, respectively (p<0.5 and p<0.01, respectively, evaluated with the Kruskal-Wallis test; Kruskal and Wallis, 1952). Furthermore, the elevated δ¹⁵N_NO3+NO2 in SPG SPSTUW corresponded to significantly lower [NO₃ ^-+NO₂ ^-] at 200 m (2.4 ± 0.8 μM) compared to the LB, where the average [NO₃ ^-+NO₂ ^-] at 200 m was 3.3 ± 0.6 μM (p<0.01, Kruskal and Wallis, 1952) ( Figures 3A, C ). Within the surface waters (upper 150 m), the δ¹⁵N_NO3+NO2 at hydrothermal station 5-2019 was significantly lower between the four casts (5a – d), 1.8 ± 0.9‰, compared to the average δ¹⁵N_NO3+NO2 in the upper 150 m at hydrothermal station 10 (10 a, b), 4.4 ± 0.6‰ (p<0.01, Kruskal and Wallis, 1952) ( Figures 3B, D ). The lowest δ¹⁵N_NO3+NO2, 0.7 ± 0.1‰, was observed at station 5c-2019 at 100 m, above which δ¹⁵N_NO3+NO2 increased to 1.0 ± 0.0‰ at 95 m ( Figures 3A, C ). The δ¹⁵N_NO3+NO2 reported here for the TONGA study are publicly available (Knapp and Forrer, 2023).

The PN_sink flux collected in the shallow short-term drifting traps was calculated to be 350 µmol N m^-2 d^-1 (170 m, station 5a-2019), 436 μmol N m^-2 d^-1 (270 m, station 5a-2019), and 693 μmol N m^-2 d^-1 (270 m, station 10a-2019) ( Table 2 ). Further, the average (± 1 S.D.) δ¹⁵N_PNsink was -0.5 ± 3.5‰ and -0.2 ± 1.9‰ at 170 m and 270 m at station 5a-2019, respectively, and -0.6 ± 2.3‰ at 270 m at station 10a-2019 ( Table 2 ) ( Figure 4 ). In comparison, the annual average PN_sink flux collected in the long-term 1000 m PPS5 moored trap at station 12-2019/4-2020 was an order of magnitude lower, 16.5 ± 14.3 μmol N m^-2 d^-1, and had a higher annual mass-weighted average (± 1 S.D.) δ¹⁵N_PNsink, 3.4 ± 1.9‰, compared to the shallower, short-term traps ( Table 2 ) ( Figure 4 ).

The seasonal average (± 1 S.D.) PN_sink fluxes collected in the deep, moored trap were higher in the austral summer and autumn, at 30.9 ± 11.0 μmol N m^-2 d^-1 and 17.0 ± 15.4 μmol N m^-2 d^-1, respectively, compared to the average austral winter and spring PN_sink fluxes of 8.0 ± 9.0 μmol N m^-2 d^-1 and 8.5 ± 8.5 μmol N m^-2 d^-1, respectively ( Table 2 ) ( Figure 4 ). The average, mass-weighted (± 1 S.D.) summer δ¹⁵N_PNsink from the 1000 m PPS5 moored trap, 1.5 ± 0.7‰, was lower than the wintertime average of 5.9 ± 1.1‰, while the average (± 1 S.D.) spring and autumn mass-weighted δ¹⁵N_PNsink, 2.9 ± 0.5‰ and 3.3 ± 1.7‰, respectively, were intermediate between summer and winter values.

The δ¹⁵N budget described above compares the δ¹⁵N of the primary form of N exported from the euphotic zone, δ¹⁵N_PNsink, with the δ¹⁵N of the two input terms, subsurface NO₃ ^-+NO₂ ^- and N₂ fixation. This provides a geochemically-derived estimate of the fractional contribution of N₂ fixation to export production (f_nfix) as well as rate of N₂ fixation (R_nfix). Given that the subsurface δ¹⁵N_NO3+NO2 source is difficult to constrain, we evaluate the shallow δ¹⁵N budgets using the shallowest subsurface δ¹⁵N_NO3+NO2 minima as well as the sample immediately below the minima (Knapp et al., 2021), while the δ¹⁵N budget using the deep trap PN_sink flux uses the average subsurface δ¹⁵N_NO3+NO2 minima at station 12-2019/4-2020 and the average SPSTUW δ¹⁵N_NO3+NO2 at station 4-2020. At the shallow trap stations in close proximity to the hydrothermal vents, the subsurface δ¹⁵N_NO3+NO2 ranged from 1.2 to 2.2‰ and 3.6 to 4.7‰ at stations 5-2019 and 10-2019, respectively, while at the deep mooring (station 12-2019/4-2020) the average subsurface δ¹⁵N_NO3+NO2 minima and SPSTUW δ¹⁵N_NO3+NO2 were 2.4‰ and 4.6‰, respectively ( Table 2 ). Additional uncertainty in the δ¹⁵N budgets includes the standard deviation of the δ¹⁵N_PNsink analysis ( Table 2 ). The δ¹⁵N value for N₂ fixation, (δ¹⁵N_Nfix) of -1‰ is based on literature reports of diazotrophic biomass δ¹⁵N (Hoering and Ford, 1960; Minagawa and Wada, 1986; Carpenter et al., 1997) ( Table 2 ). Comparing these values with the δ¹⁵N_PNsink, we calculate that at station 5-2019, the f_nfix ranged from 77 to 84 ± 159% for the 170 m trap, while the f_nfix ranged from 64 to 76 ± 86% at the 270 m trap ( Table 2 ). At station 10-2019, the f_nfix at the 270 m trap ranged from 90 to 92 ± 50% ( Table 2 ). We further calculate a geochemically derived R_nfix of 282 ± 550 µmol N m^-2 d^-1 and 331 ± 375 µmol N m^-2 d^-1 for the 170 m and 270 m traps deployed at station 5-2019, respectively, and 638 ± 347 µmol N m^-2 d^-1 for the 270 m trap deployed at station 10-2019 ( Table 2 ). The δ¹⁵N budget calculated for the moored trap at station 12-2019/4-2020 (note that the trap is deeper, i.e. 1000 m instead of 170 – 270 m for the drifting ones) yielded a mass-weighted f_nfix of 12 to 64 ± 29% (average = 42 ± 13%) for the summer, while the spring and autumn had similar but lower mass-weighted f_nfix of 0 to 37 ± 21% (average = 0 ± 29%) and 0 to 43 ± 69% (average = 12 ± 36%), respectively, and winter had a mass-weighted f_nfix of 0 ± 18%. These f_nfix values indicate that the highest contribution of N₂ fixation to export production collected at 1000 m was during the summer, followed by autumn and spring. Due to significant PN_sink flux attenuation with depth (Martin et al., 1987), we do not calculate a R_nfix for the moored trap data.

The results of the short-term δ¹⁵N budgets suggest that N₂ fixation rates were high (282 to 638 μmol N m^-2 d^-1) near the hydrothermal vents at the time of the 2019 cruise. While there are notable differences in the magnitude of these geochemically-derived N₂ fixation rates and the bottle-based ¹⁵N₂ uptake rates reported by Lory et al., 2023, we consider these results to be broadly consistent with one another, as well as consistent with estimates of diazotroph abundance measured contemporaneously (Bonnet et al., 2023b). Specifically, bottle-based ¹⁵N₂ uptake rates were 3 to 7 times higher than the R_nfix estimated from the δ¹⁵N budget ( Table 2 ); however, both rate estimates are at the upper-end typically reported by each technique (Gruber, 2004; Capone et al., 2005; Casciotti et al., 2008; Luo et al., 2012; Knapp et al., 2016a) and align with previous work from the region (Montoya et al., 2004; Berthelot et al., 2017; Bonnet et al., 2017; Knapp et al., 2018; Shao et al., 2023). Prior work investigating N₂ fixation’s contribution to export production has attributed discrepancies between these two metrics to potential sediment trap under collection of exported material (Knapp et al., 2016a; Böttjer et al., 2017, Knapp et al., 2018), alternate sources of fixed N to the euphotic zone including horizontal advection (Böttjer et al., 2017), phytoplankton bloom stage as well as temporal delay between organic matter formation and capture by the sediment trap (de Verneil et al., 2017; Caffin et al., 2018; Knapp et al., 2018), and bottle-based ¹⁵N₂ incubations with their associated methodological considerations (i.e., bottle effects) (White et al., 2020). Additionally, our δ¹⁵N budgets should be considered a lower bound for estimated N₂ fixation rates because of the mixing-model’s inherent assumption that the only fate of newly fixed N is to be balanced by the sinking flux, and that no newly fixed N is released to the dissolved pool, which is likley unrealistic (Capone et al., 1994; Glibert & Bronk, 1994; Mulholland and Capone, 2004; Bonnet et al., 2016; Knapp et al., 2016b). Further, the trap depth used here, 170 m, is below the base of the mixed layer, and thus underestimates new/export production. While euphotic zone nitrification is a source of low-δ¹⁵N_NO3 that could lead to an overestimation of N₂ fixation supported export production, rates of euphotic zone nitrification from this and other similar oligotrophic regions are low (<10 nmol L^-1 d^-1) (Smith et al., 2014; Raes et al., 2020) compared to the very high rates of N₂ fixation using multiple metrics in this study. This therefore suggests that N₂ fixation is the dominant mechanism generating the low-δ¹⁵N_PNSink signal observed.

Regardless of the mechanism(s) driving these discrepancies, both the δ¹⁵N budget and ¹⁵N₂ uptake rate estimates are also consistent with the elevated diazotroph abundances (Bonnet et al., 2023a; Lory et al., 2023) observed on the 2019 TONGA cruise ( Figure 5 ). Here we compare averages of biological and geochemical metrics associated with N₂ fixation for the three hydrographic regions, MW, the LB, and the SPG (defined in Section 2.1). While results from the qPCR analysis targeting nifH genes indicate that Trichodesmium spp. and UCYN-A dominated the diazotroph assemblage in the upper 50 m across the TONGA transect, Trichodesmium spp. were most abundant in the LB near the hydrothermal vents (average 1.0x10⁷ gene copies L^-1), followed by UCYN-A (average 3.0x10⁶ gene copies L^-1) ( Figures 5A, B ). These exceptionally high abundances of Trichodesmium spp. and UCYN-A were on the order of one to four times higher than previous studies (e.g., Zehr & Turner, 2001; Moisander et al., 2010; Turk-Kubo et al., 2012; Moisander et al., 2014; Benavides et al., 2018; Benavides et al., 2020; Confesor et al., 2022). Notably, the highest regionally-averaged, trapezoidally-integrated upper 50 m ¹⁵N₂ uptake rates were found in the LB (1038 ± 600 µmol N m^-2 d^-1), where Trichodesmium spp. were most abundant and where the lowest regionally-averaged δ¹⁵N_NO3+NO2 subsurface minima of 2.8 ± 1.5‰ ( Figure 5F ) was observed between 100 and 200 m. This underscores the potential significance of Trichodesmium spp. supporting N₂ fixation near the hydrothermal vents. UCYN-A was the dominant diazotroph in the upper 50 m in MW and the SPG, with regional averages of 2.5x10⁸ gene copies L^-1 and 3.1x10⁸ gene copies L^-1, respectively, followed by Trichodesmium spp., with regional averages of 3.0x10⁵ and 3.7x10⁵ gene copies L^-1, respectively. Average abundances of UCYN-B (2.6x10⁴ to 3.2x10⁵ gene copies L^-1), UCYN-C (4.3x10¹ to 1.7x10² gene copies L^-1) and Gamma proteobacteria (5.4x10³ to 9.9x10³ gene copies L^-1) remained comparatively low across the transect ( Figures 5C–E ), but were similar in magnitude to previous studies in the region (Moisander et al., 2014; Benavides et al., 2018; Benavides et al., 2020). Although UCYN-A have extremely high gene abundances and dominate the upper 50 m in MW and the SPG, the trapezoidally-integrated upper 50 m ¹⁵N₂ uptake rates were lower than in the LB, 713 ± 691 and 537 ± 629 µmol N m^-2 d^-1, respectively, and were associated with a higher regional average δ¹⁵N_NO3+NO2 subsurface minima of 4.4 ± 1.2‰ and 5.5 ± 1.9‰, respectively, observed between 100 to 200 m ( Figure 5F ). While care should be taken when relating nifH gene copies to diazotroph biomass, these gene copy abundances broadly correspond to elevated diazotroph abundances (Meiler et al., 2022) and confirm the significance of diazotrophy in the region.

The agreement between the geochemically-derived N₂ fixation rates, ¹⁵N₂ uptake rates and diazotroph abundances together indicate that export production in the WTSP at the time of this study was driven by N₂ fixation (Bonnet et al., 2023b). This is in contrast to prior work in other oligotrophic regions where the majority of export was supported by subsurface NO₃ ^-, even when N₂ fixation inputs were high (Casciotti et al., 2008; Bourbonnais et al., 2009; White et al., 2013; Böttjer et al., 2017). The shallow sediment traps deployed for the TONGA project indicate that N₂ fixation supports a majority of export production (f_nfix = 64 to 92%) near the hydrothermal vents, at least in the late spring ( Table 2 ; Figures 6A, B ). These f_nfix values are similar to those calculated from the previous OUTPACE campaign during the late summer/early autumn in MW (station A, 80 to 83 ± 13%), and near the Tonga-Kermadec Arc (station B, 50 to 56 ± 12%) ( Table 2 ; Figure 6C ) (Knapp et al., 2018). The high f_nfix values reported here from the short-term traps deployed at TONGA Stations 5 and 10 replicate a geochemical signature that has not been observed outside of the WTSP, underscoring the significance of N₂ fixation regionally (Bonnet et al., 2023b). We also emphasize the potential for the f_nfix from our δ¹⁵N budgets to underestimate the importance of N₂ fixation to export due to the mixing-model’s inherent assumption that the only fate of newly fixed N is the PN_sink flux captured by the sediment traps, as opposed to being released and persisting as dissolved organic nitrogen (Capone et al., 1994; Glibert and Bronk, 1994).

While the results of the short-term δ¹⁵N budgets from both the TONGA and OUTPACE campaigns found that N₂ fixation supports the majority of export production in the late spring, late summer, and early autumn in the LB and MW, we also consider the δ¹⁵N_PNsink collected in the deep, moored trap at station 12-2019/4-2020 in the LB to evaluate this trend over an annual cycle ( Figure 4 ). Unsurprisingly, the mass flux collected in the shallower, short-term PIT traps was higher and the δ¹⁵N_PNsink was lower than that collected in the deeper, moored trap. However, we note that the moored trap was deployed ~200 km west of the shallow traps and hydrothermal vents ( Table 2 ) ( Figures 1 , 4 ), and the conical shape of the PPS5 has been observed to undersample (Baker et al., 2020; Tilliette et al., 2023). The δ¹⁵N_PNsink in the shallow PIT traps collected in the late spring at stations 5-2019 and 10-2019 ranged from -0.2 to -0.6‰ compared to a seasonal average δ¹⁵N_PNsink of 2.9 ± 0.5‰ and 1.5 ± 0.7‰ observed in the moored PPS5 trap during the spring and summer, respectively. We expect that the higher δ¹⁵N_PNsink found in the deeper moored trap likely resulted from the collection of PN_sink from a larger surface area than the shallow short-term traps, where export production may have been supported by a mixture of N sources with higher δ¹⁵N (Siegel and Deuser, 1997; Siegel et al., 2008). Additionally, horizontal advection of particles generated at locations not impacted by the shallow hydrothermal vents potentially decouples the euphotic zone diazotrophic abundance and/or importance from the PN collected in the moored trap (Waniek et al., 2000). Indeed, the exported material in the deep trap was observed to be compositionally different from that captured in the shallow, short-term traps, where the hydrothermal signature of the particles was less evident due to organic matter remineralization while being transported to depth (Tilliette et al., 2023). The associated distance and time components of sinking PN to the deep trap potentially underestimates the importance of N₂ fixation to export production. Similar flux and isotopic composition offsets have been observed between shallow and deep sediment traps in the ETSP (Berelson et al., 2015; Knapp et al., 2016a; Tilliette et al., 2023).

Since aerosol inputs to this region are minimal (Guieu et al., 2018), we expect that the seasonal variability of δ¹⁵N_PNsink in the deep, moored trap reflects shifts in the importance of N₂ fixation and subsurface NO₃ ^- for supporting export production over seasonal timescales. In the deep trap, annual PN_sink fluxes peaked in the summer (30.9 ± 11.0 μmol N m^-2 d^-1), coinciding with the lowest average mass-weighted δ¹⁵N_PNsink, 1.5 ± 0.7‰, while the lowest PN_sink fluxes were observed in the winter (8.0 ± 9.0 μmol N m^-2 d^-1) and coincided with the highest average mass-weighted δ¹⁵N_PNsink, 5.9 ± 1.1‰, indicating that isotopically distinct N sources support export seasonally. Since there are steep gradients in both [NO₃ ^-+NO₂ ^-] and δ¹⁵N_NO3+NO2 with depth ( Figures 3 , 6 ), a higher winter δ¹⁵N_PNsink may reflect entrainment of a deeper NO₃ ^-+NO₂ ^- source (likely SPSTUW) with a higher δ¹⁵N_NO3+NO2 due to winter mixing (Moutin et al., 2018). The net effect of a higher subsurface δ¹⁵N_NO3+NO2 end-member would be to raise the estimated f_nfix ( Table 2 ) (Böttjer et al., 2017). As a result, the mass-weighted seasonal f_nfix values for the deep trap range from 12 to 64 ± 29% in the summer and 0 ± 18% in the winter, describing a largely N₂ fixation supported export system in the summer. Further, since the majority of annual export is focused in the summer, and was supported by low-δ¹⁵N N sources, we attribute an important fraction of annual export production and deep (>1000 m) long-term C sequestration to N₂ fixation at station 12-2019/4-2020 ( Figure 4 ; Table 2 ).

Considering the elevated chlorophyll a concentrations observed over a large area in this region ranging up to 360,000 km² (Bonnet et al., 2023b), and given the high R_nfix and f_nfix values estimated at station 5a-2019 and 10a-2019, along with the large fraction of N₂ fixation supported export production at station 12-2019/4-2020 over an annual timescale, the otherwise oligotrophic WTSP appears to be biogeochemically unique where N₂ fixation supports a large fraction of annual export production as a result of the influence of shallow hydrothermal vents. The significance of these regional N₂ fixation inputs in the WTSP are further pronounced in the gradients of water column δ¹⁵N_NO3+NO2 both zonally as well as with depth ( Figures 3 , 6 ). In particular, δ¹⁵N_NO3+NO2 between 150 and 400 m decreases from east to west (SPG to MW) across the zonal transect, and also decreases from ~400 m to shallower depths. These isotopic gradients reflect the accumulation of low-δ¹⁵N N inputs in the upper thermocline to the west along this transect that are presumably associated with the remineralization of diazotrophic inputs (Casciotti et al., 2008; Knapp et al., 2008). This accumulation of low-δ¹⁵N_NO3+NO2 in the upper 400 m of the WTSP erodes the elevated δ¹⁵N_NO3+NO2 originating from dissimilatory NO₃ ^- reduction occurring in the oxygen deficient zones of the ETSP (Bourbonnais et al., 2015; Peters et al., 2018; Casciotti et al., 2013), a geochemical signature that reflects basin-scale compensation of N losses in the east with N inputs in the west that is consistent with peloceanographic records (Brandes and Devol, 2002; Deutsch et al., 2004; Weber and Deutsch, 2014; Knapp et al., 2018).