Seasonal Shifts in Diazotrophs Players: Patterns Observed Over a Two-Year Time Series in the New Caledonian Lagoon (Western Tropical South Pacific Ocean)

Coastal and open ocean regions of the Western Tropical South Pacific ocean have been identified as a hotspot of N2 fixation. However, the environmental factors driving the temporal variability of abundance, composition, and activity of diazotrophs are still poorly understood, especially during the winter season. To address this, we quantified N2 fixation rates and the abundance of seven diazotroph phylotypes (UCYN-A1 symbiosis, UCYN-B, UCYN-C, Trichodesmium, Het-1, Het-2, and Het-3) on a monthly basis during two full years (2012 to 2014) at four stations along a coast to open ocean transect in the New Caledonian lagoon. The total nifH gene concentration (sum of all nifH gene copies) clearly decreased from the barrier reef to the shore. Apart from UCYN-B, which peaked at very high abundances (106–108 nifH gene copies L–1) at two occasions at the coastal station, the UCYN-A1 symbiosis was the most abundant group at all stations, accounting for 79% of the total nifH gene copy counts along the transect (average abundance 4.2 ± 10.3 × 104 nifH gene copies L–1). The next most abundant groups were in order Trichodesmium (accounting for 14% of the total nifH gene copies), Het-groups (6% of the total) and UCYN-C (1% of the total). Statistical analyses reveal that the UCYN-A1 symbiosis and Het groups were associated with cold (<25°C) waters, high NOx and PO43–, weak winds from the south (occasionally southwest), while Trichodesmium and UCYN-C were associated with warmer (>25°C) waters, low NOx and PO43– concentrations, strong and (mostly) easterly winds. Average N2 fixation rates over the survey were 6.5 ± 6.7 nmol N L–1 d–1 and did not differ significantly among seasons. The year to year variability was more pronounced with average integrated rates significantly higher in the second year of the survey (162 ± 122 μmol N m–2 d–1) than the first year (66 ± 91 μmol N m–2 d–1). This dataset suggests that seasonality is less pronounced than previously thought, and that relatively high N2 fixation rates are maintained in the New Caledonian lagoon all year long, despite seasonal changes in the diazotroph community composition.


INTRODUCTION
Biological dinitrogen (N 2 ) fixation is a main source of fixed N to the oceans and is performed by a suite of specialized cyanobacteria, bacteria and archaea called "diazotrophs" (Zehr, 2011). Fixed N input fluxes on a global basis vary between 106-120 Tg N year −1 (Galloway et al., 2004;Gruber, 2008;Jickells et al., 2017). The accuracy of those estimates is inevitably affected by the temporal and spatial resolution of measurements. Only few studies have addressed how the paucity of data on spatial and temporal distribution of N 2 fixation may affect global fixed N input estimates. For instance, the global database compiled in Luo et al. (2012) contains 2480 data points of volumetric seawater N 2 fixation rates from the northern hemisphere, while samples from the southern hemisphere only amount to 502. The temporal distribution of these data points also suffers from a strong seasonal bias as >65% of the N 2 fixation measurements were made during the spring-summer season. Hence, most of our recapitulative knowledge of marine N 2 fixation stems from the northern hemisphere in spring-summer conditions, largely from the sunlit layer of open ocean regions of the North Atlantic and North Pacific Oceans (Benavides and Voss, 2015;Böttjer et al., 2017).
Although the number of N 2 fixation studies has increased in recent years, temporal and spatial resolution remains coarse, as oceanographic cruises commonly sample transects with stations separated by tens of nautical miles, and typically at monthly or seasonal time scales. For instance, N 2 fixation rates varied four-to nine-fold between summer and winter at Station ALOHA in the Pacific Ocean between 2005 and 2013 (Böttjer et al., 2017). If higher diazotrophic activity during summer prevails among the available datasets it could cause an overestimation of basin-wide estimates, which illustrates the need for conducting sampling programs at finer temporal scales and throughout all seasons of the year.
Recent intensive sampling of the western tropical South Pacific (WTSP) Ocean revealed this region as a hotspot of N 2 fixation with some of the highest rates ever recorded (Messer et al., 2015;Berthelot et al., 2017;Bonnet et al., 2017). However, most of our understanding of diazotrophic activity and diversity in the WTSP derives from open ocean cruises, the majority of which took place in austral summer. Little sampling effort has been devoted to coastal environments, despite their central role in the global cycling of nutrients (Bauer et al., 2013) and increasing evidence of their role as hotspots of diazotrophic activity (Mulholland et al., 2012(Mulholland et al., , 2019Tang and Cassar, 2019). In the Australian Great Barrier Reef, Hewson et al. (2007) reported nifH gene sequences corresponding to usual open-ocean diazotroph phylotypes such as the UCYN-A symbiosis, UCYN-B and the diatom-diazotroph association (DDA) Het-1 (Richelia associated with Rhizosolenia). Also in the Great Barrier Reef, Messer et al. (2017) found Trichodesmium erythraeum to be the dominant diazotroph in their nifH gene libraries and measured N 2 fixation rates from 3 to 68 nmol N L −1 d −1 .
The only other coastal N 2 fixation studies in the WTSP are those from the reef systems of New Caledonia, the second largest in the world after the Great Barrier Reef, enclosing a 24,000 km 2 lagoon. The New Caledonian lagoon waters are shallow (∼20 m on average) and mostly oligotrophic, although inorganic nutrients are detectable year-round in the waters affected by the river and anthropogenic discharges near Noumea city (Neveux et al., 2009;Ouillon et al., 2010). Previous studies have documented high N 2 fixation rates (1.9 to 29.3 nmol N L −1 d −1 , Garcia et al., 2007;Bonnet et al., 2016), occurrence of large Trichodesmium blooms (Rodier and Le Borgne, 2008;Tenório et al., 2018) and co-occurrence with other common oceanic diazotrophs such as DDAs, UCYN-A/haptophyte symbioses, Crocosphaera (UCYN-B; Biegala and Raimbault, 2008;Turk-Kubo et al., 2015;Henke et al., 2018), and UCYN-C (Turk-Kubo et al., 2015). Most of these studies were performed during austral summer conditions when diazotrophs are considered to be most abundant and N 2 fixation rates maximal. Yet the effect of seasonal variability and nutrient inputs from the coast on the distribution and activity of these diazotrophic groups is lacking. Here we examine N 2 fixation rates and diazotroph phylotype abundances at four stations along a coast to openocean transect in the New Caledonian lagoon during two full years (2012 to 2014) on a monthly basis in the framework of the "Grand Observatoire de l'environnement et de la biodiversité terrestre et marine du Pacifique Sud" (GOPS). Henke et al. (2018) examined the seasonal variability and diversity of UCYN-A/haptophyte symbioses relative to environmental conditions. In the current study we sought to expand our knowledge on the broader diazotroph community and on diazotrophic activity.

Sampling, Hydrographic and Meteorological Parameters
Between July 2012 and April 2014, seawater samples were collected monthly in the southwestern New Caledonian lagoon along a transect of four stations (L2, M09, M99, and D39; Figure 1) spanning from the reef barrier (Dumbea Pass) to the shore (Dumbea Bay), into which the Dumbea river flows. The sampling always started just before high tide to ensure that results would be comparable. Seawater was collected from three depths spanning the water column of the lagoon (1, 8, and 16 m at station L2, M09, D39, and 1, 3, and 6 m at station M99) using a trace metal clean Teflon pump connected to polyethylene tubing. Vertical hydrographic profiles were obtained at each station by casting a Seabird SBE19 CTD equipped with additional turbidity (Seapoint optical back scatter) and in situ fluorescence (Wet Lab Wetstar) sensors. Daily values of rainfall, wind direction and velocity were obtained from the Météo-France station at Faubourg Blanchot (Noumea, 22 • 16.30' S-166 • 27.06' E). Here we used averages of the data from the week preceding sampling in order to integrate meteorological variations.

N 2 Fixation Rates
Seawater samples were collected at each depth in transparent 4.5 L-polycarbonate bottles closed gas-tight with septum screw caps (Nalgene). Bottles were filled to overflow and amended with 5 mL of 98.9 atom% 15 N 2 (Cambridge isotopes) either by directly injecting the tracer as a bubble (Montoya et al., 1996) or as dissolved in a subset of seawater previously N 2 degassed (Mohr et al., 2010) according to the methodology fully described in Berthelot et al. (2015). A comparison between the two methods did not show any significant difference (p > 0.05) between N 2 fixation estimates (Supplementary Figure 1). To ensure accurate rate calculations, the 15 N/ 14 N ratio of the dissolved N 2 pool in the incubation bottles was measured regularly whatever the method used. For this purpose, after 24 h of incubation with the 15 N 2 tracer, 12 mL subsamples of each bottle were moved into Exetainers R fixed with HgCl 2 (final concentration 10 µg L −1 ), which were stored upside down in the dark at 4 • C until analyzed by membrane inlet mass spectrometry (MIMS) according to Kana et al. (1994). Lastly, time zero (not 15 N-enriched) samples were collected at each monthly survey and station to determine the natural N isotopic signature (EA-IRMS) of ambient particulate N (PN). The 15 N 2 gas batches used did not contain significant concentrations of other nitrogenous compounds labeled with 15 N, as previously measured (Dabundo et al., 2014;. The minimum quantifiable rates (quantification limit, QL) were 0.035 nmol N L −1 d −1 . Integrated rates over the water column were calculated by using the trapezoidal method.

DNA Extraction and Quantitative PCR
For DNA analyses, seawater samples were collected only from the surface (1 m) in 2.3 L polycarbonate bottles and 930-2,300 mL were filtered within 1 h through 0.2 µm Supor filters (Pall Gelman), that were stored at −80 • C in bead beater tubes containing a mixture of 0.1-mm and 0.5-mm diameter glass beads (BioSpec Products, Bartlesville, OK, United States). DNA was extracted using the Qiagen Plant kit with additional steps for enhanced DNA recovery from cyanobacteria (Moisander et al., 2008) before automated on-column purification and elution using a Qiacube (Qiagen). All DNA samples were quantified by NanoDrop ND-1000 Spectrophotometry (Thermo Scientific).
The abundance of diazotrophs was determined using TaqMan R qPCR assays on either an ABI 7500 or StepOnePlus Real-Time PCR System (Applied Biosystems), with primer-probe sets for Trichodesmium (Church et al., 2005), UCYN-B (Moisander et al., 2010), UCYN-C (Church et al., 2005), Het-1 (Richelia in Rhizosolenia), Het-2 (Richelia in Hemiaulus), and Het-3 (Calothrix in Chaetoceros; Foster et al., 2007), and UCYN-A1 and UCYN-A2 (Henke et al., 2018). For the purpose of this study, UCYN-A1 and UCYN-A2 have been summed, yielding to total UCYN-A abundances. Standard dilutions (10 7 -10 1 ) were run in duplicate, and samples and no-template controls (NTCs) were run in triplicate. NTCs were undetectable. The efficiency of the primer-probe sets was 82-102%. Inhibition tests were carried out on all samples and each primer-probe set by adding 2 µL of 10 5 standard to each sample. No inhibition was detected. The limit of detection (LOD) and detected but not quantifiable (DNQ) limits used were 1 and 8 gene copies per reaction, respectively, (Goebel et al., 2010). Samples that were below LOD were designated a value of 0 in the data set, whereas gene copies higher than LOD but less than DNQ were designated a conservative value of 1 nifH gene copy L −1 . Note that abundances are reported as nifH gene copies L −1 ; extrapolating cell numbers from nifH gene copies can be difficult due to the possibility that some taxa have multiple genome copies (polyploidy; Sargent et al., 2016;White et al., 2018).

Environmental Conditions
Seawater temperature ranged from 21.7 • C to 28.1 • C over the course of the survey, and was relatively homogeneous throughout the water column (Figure 2A). Maximum temperatures were measured between December and February, while minima were recorded between August and September, corresponding to the New Caledonian warm and cool seasons. The stations close to  Table 1). This freshening was particularly pronounced at coastal stations D39 and M99 located close to the Dumbea river ( Figure 2B). NO x and PO 4 3− concentrations (Figures 2D,E) ranged from undetectable to 0.31 µmol L −1 and to 0.16 µmol L −1 , respectively, (Figure 2), with no clear seasonal trend. The NO x :PO 4 3− was on average 3:1 over the study period, indicative of PO 4 3− excess relative to NOx. The Si(OH) 4 concentrations (range 0.8-15.0 µmol L −1 ) gradually increased from the reef (0.84-2.92 µmol L −1 ) to the coast (2.0-15.0 µmol L −1 ), where they were the highest at the surface ( Figure 2C). The Si(OH) 4 :NO x ratio was on average 45 over the study period, indicative of Si(OH) 4 excess relative to NOx. The biomass indicators (Chl a and PON concentrations) were higher at coastal stations (D39 and M99) compared to reef stations (M09 and L2), and generally higher during the warm season (November-April) compared to the cold one (May-October; Figures 2F,G). The average weekly wind direction indicated E-SE settled trade winds (6-10 m s −1 ) most of the year, with some sporadic S-SW episodes recorded in late winter-early summer 2012 and 2013 (Supplementary Table 1).

Rates of N 2 Fixation in New Caledonian Waters
N 2 fixation rates ranged from not quantifiable to 51.9 nmol N L −1 d −1 (mean over the survey: 6.5 ± 6.7 nmol N L −1 d −1 ). Consistent with hydrographic parameters, they were relatively homogeneous over the well-mixed water column and displayed similar temporal patterns among stations, although less variability was observed at station M99 compared to the three others (note that this station is shallower than the three others; Figure 3A). Generally, rates integrated over the water column ( Figure 3B) were highest during austral summer months (October-March), but not significantly different (p > 0.05), and highly variable from one monthly survey to another. The year to year variability was more pronounced than the seasonal one: the average annual rates were significantly (p < 0.05) lower during the first year of the survey (July 2012-May 2013, 66 ± 91 µmol N m −2 d −1 ) than during the second year (June 2013-April 2014, 162 ± 122 µmol N m −2 d −1 ). When averaged over the whole survey, rates were the highest at station M09 (168 ± 153 µmol N m −2 d −1 ), followed by stations D39 (168 ± 122 µmol N m −2 d −1 ), L2 (117 ± 110 µmol N m −2 d −1 ), and M99 (48 ± 36 µmol N m −2 d −1 , note that the depth of integration is 8 m instead of 16 m for the others).

Spatiotemporal Variability of Diazotroph Abundances
Total nifH counts L −1 (sum of all nifH gene copies L −1 of the five phylotypes targeted by qPCR assays) showed limited seasonal variability and were on average 2.9 ± 23 × 10 6 nifH gene copies L −1 along the entire survey ( Figure 4A). UCYN-B was the most abundant group, accounting for >99% of the total nifH gene copy counts on average over the survey all stations combined. This was due to very high abundances on two occasions at the coastal site D39 (8.4 × 10 6 and 2.0 × 10 8 nifH gene copies L −1 on Feb. 28 and May 2 2013; Figure 4B), but abundances were otherwise sporadic (below the limit of detection in 57 out of 71 samples). Not taking into account those two peaks, the total nifH clearly decreased along the transect from the barrier reef (Dumbea pass) to the shore (Dumbea river; Figure 4C). The UCYN-A symbiosis was the most abundant group, accounting for 79% of the total nifH gene copy counts along the transect (average abundance 4.2 ± 10.3 × 10 4 nifH gene copies L −1 ). The next most abundant groups were Trichodesmium (7.7 ± 28.2 × 10 3 nifH gene copies L −1 , accounting for 14% of the total nifH gene copies), Het-groups (3.0 ± 7.9 × 10 3 nifH gene copies L −1 , 6% of the FIGURE 4 | Temporal dynamics of total nifH gene counts L −1 over the survey in surface waters (A; the errors bars correspond to the errors propagation from triplicate analyses), relative abundance of each diazotroph phylotype targeted in this study and sea surface temperature (SST, red triangles; B), and average nifH counts L −1 over the survey per target group and per station (C).
While the total nifH gene count did not differ between seasons, the nifH gene abundance per target group showed temporal variability ( Figure 4B). Using temperature categories (< or >25 • C; Moisander et al., 2010;Bonnet et al., 2015), the UCYN-A symbiosis and Het groups were 4.8 and 4.6 times more abundant in <25 • C waters compared to >25 • C. The opposite trend was observed for Trichodesmium and UCYN-C, which were 1.6 and 1.2 more abundant in >25 • C waters.
The redundancy analysis (RDA) shows that the correlation between environmental factors and diazotroph phylotypes is mainly explained by the first two axes (75% of the total variance; Figure 5). The factorial axis 1 separates two phylotypes: (i) Trichodesmium and UCYN-C, associated with low NOx and PO 4 3− concentrations, warm waters, strong and (mostly) wind direction (dominant easterly winds), and high N 2 fixation rates, (ii) UCYN-A symbiosis and Het groups, associated with higher NOx and PO 4 3− , colder waters, weaker winds from the south (occasionally southwest), and lower N 2 fixation rates. The factorial axis 2 separates UCYN-B from all other nifH phylotypes. UCYN-B are mainly associated with rainfall, desalted and warm waters characterized by high biomass (POC, PON, and Chl a).

DISCUSSION
Here we present a spatiotemporal study of N 2 fixation rates and diazotrophic community composition over 2 years in the New Caledonian lagoon. N 2 fixation was detected and diazotrophs were present throughout the year with generally high total mean abundances (10 6 nifH gene copies L −1 ). Inorganic nutrients are supplied to the lagoon by the Dumbea river, however, bioavailable forms are generally low (∼0.05 µmol L −1 ) in most of the lagoon waters, and NOx are depleted relative to PO 4 3− , making these waters favorable for N 2 fixation. The total number of diazotrophs detected by qPCR increased toward the more oligotrophic reef stations ( Figure 4C). This pattern was mainly influenced by the UCYN-A symbiosis, which were more than two orders of magnitude more abundant at offshore stations receiving open-ocean influences (Fichez et al., 2010) compared to inner stations. This is in accordance with previous studies reporting surface UCYN-A symbiosis abundances of 10 3 -10 6 nifH gene copies L −1 in the Coral Sea off New Caledonia (Moisander et al., 2010;Bonnet et al., 2015;Messer et al., 2015), and at the entrance of the Noumea lagoon . This is also consistent with earlier reports of significant N 2 fixation and diazotroph abundances in the <3 µm size-fraction in New Caledonian waters (Garcia et al., 2007;Biegala and Raimbault, 2008). However, our results contrast with data from open ocean waters around Melanesian archipelagos, where Trichodesmium and UCYN-B outnumber other groups and the UCYN-A symbiosis is almost absent in surface waters and only develop deeper in the water column Stenegren et al., 2018), at least during austral summer conditions. However, Moisander et al. (2010) found high abundances of the UCYN-A symbiosis close to New Caledonia (>10 6 nifH gene copies L −1 ), coinciding with a local upwelling system characterized by colder temperatures (∼25 • C) compared with surrounding waters at the same latitude (27-30 • C). Bonnet et al. (2015) also reported high UCYN-A symbiosis abundances in the Coral Sea off New Caledonia but during austral winter conditions [Surface Sea Temperature (SST) ∼22 • C]. These studies support a niche partitioning between surface warm (>25 • C) waters where Trichodesmium and UCYN-B thrive, and cooler (<25 • C) waters preferred by the UCYN-A symbiosis. We observed the same pattern in the inner lagoon waters, where the UCYN-A symbiosis was ∼5 times more abundant during the winter season (SST < 25 • C), but to a lesser extent compared to open ocean waters. Indeed, the UCYN-A symbiosis still accounted for ∼60% of the diazotroph community during the summer season (SST > 25 • C), which is not the case in the nearby open ocean waters (Moisander et al., 2010;Stenegren et al., 2018). The dominance of the UCYN-A symbiosis in the New Caledonian lagoon also contrasts with recent reports of Trichodesmium dominance and low abundances of the UCYN-A symbiosis at Great Barrier Reef and in the adjacent Coral Sea waters off Australia, especially during the winter season (Messer et al., 2015(Messer et al., , 2017. The UCYN-A symbiosis dominated the diazotrophic community at the four stations over both the austral summer and winter seasons (Figure 4), suggesting the lagoon waters provide favorable conditions for their growth. Henke et al. (2018) found that SST and Chl a concentrations were negative predictors of UCYN-A symbiosis abundance, and NOx concentration was a positive predictor. This may be explained by the SE winds triggering an upwelling off the southwestern coast of New Caledonia (north of the Dumbea pass) that may decrease seawater temperature by 5 • C in summer and 1 • C in winter (Hénin and Cresswell, 2005;Ganachaud et al., 2010), and affects the lagoon SST by incoming tides. SE winds also drive cool, oligotrophic South Equatorial Current waters through the lagoon, arguing for a possible open ocean origin of the UCYN-A symbiosis detected in lagoon waters (Henke et al., 2018). The positive correlation between the UCYN-A symbiosis and S and SE winds (Henke et al., 2018) supports this hypothesis. NOx concentrations are an order of magnitude higher in the lagoon than in open-ocean waters (Moutin et al., 2018) and Great Barrier Reef (Messer et al., 2017), which may also explain why UCYN-A thrive in the New Caledonian lagoon compared to other more oligotrophic areas of the WTSP (open-ocean waters of the Melanesian archipelagos; Bonnet et al., 2015;Stenegren et al., 2018). This is also in accordance with Mills et al. (2020) who recently reported an enhanced activity of the UCYN-A symbiosis after NO 3 -additions in N-limited environments. Finally, due to the inability of UCYN-A to fix inorganic carbon (Tripp et al., 2010), it is likely that the high abundance of the symbiosis in lagoon waters is sustained by photosynthates of the haptophyte host and/or labile dissolved organic matter input from the Dumbea River. Indeed, a recent study reported an increase in N 2 fixation rates and UCYN-A nifH gene expression in the mouth of the Dumbea River compared to waters located close to the reef .
Surprisingly, Trichodesmium abundances were lower than those of the UCYN-A symbiosis (Figure 4). Over the past 20 years, Trichodesmium has been routinely studied in the New Caledonian lagoon and surrounding waters during summer conditions using satellite observations (Dupouy et al., 2000(Dupouy et al., , 2011Dupouy et al., 2018a,b) and direct field measurements (Rodier and Le Borgne, 2008;Turk-Kubo et al., 2015). These studies have revealed higher abundances in summer (∼0-250 trichomes L −1 and ∼10 4 nifH genes copies L −1 , respectively) than during our surveys (7.7 × 10 3 nifH genes copies L −1 ; note that microscope cell counts do not equal nifH counts; White et al., 2018). Trichodesmium abundances gradually increased from the coast to the open-ocean (D39 = 6.66 × 10 2 ; M99 = 1.4 × 10 3 ; M09 = 6.89 × 10 3 ; and L2 = 2.19 × 10 4 nifH gene copies L −1 ), suggesting that they could have an open-ocean origin and could be advected toward the coast with rising tide currents that flow over the reef (Jarrige et al., 1975;Rougerie, 1986). The abundance of Trichodesmium was higher during the summer and SST clustered positively with the abundance of Trichodesmium, in agreement with earlier findings (Tenório et al., 2018). This is consistent with previous studies showing that Trichodesmium is constrained by SST (20-30 • C; Capone et al., 1997) and thrives at temperatures of 25 • C or warmer (e.g., Breitbarth et al., 2007;Fu et al., 2014). The distribution pattern of Trichodesmium determined in this study is opposite to that of the UCYN-A symbiosis, whose abundances peaked during the winter. This points toward an ecological niche partitioning between the two phylotypes, a temporal segregation that mirrors the spatial segregation observed in open-ocean waters (Moisander et al., 2010;Bonnet et al., 2015;Stenegren et al., 2018).
Heterocystous cyanobacterial symbionts (Het-groups) were present almost throughout the survey (mean 3.0 ± 7.9 × 10 3 nifH gene copies L −1 ; Figure 4), and were clearly dominated by Het-1 targeting Richelia, in line with previous lagoon  and regional open-ocean studies (Stenegren et al., 2018). The abundance of Het-groups was maximum at the two inner stations (close to the Dumbea river mouth) impacted by nutrient input, and mostly in waters <25 • C, consistent with previous studies reporting high Hetabundances in the vicinity of the Amazon and Mekong river plumes (e.g., Carpenter et al., 1999;Subramaniam et al., 2008;Bombar et al., 2011).
Apart from the two peaks of UCYN-B at the coastal station (see below), the unicellular groups B and C were the least abundant in lagoon waters over the 2-years survey. They did not show any clear spatial or seasonal pattern, but were present at low abundances (10 2 to 10 3 nifH gene copies L −1 ) over space and time, which is consistent with our observations during a 3-week survey performed close to the reef . UCYN-B peaked on two occasions (8.4 × 10 6 and 2.0 × 10 8 nifH gene copies L −1 on February 28 and May 2 2013) coinciding with high rainfall and a peak of Si(OH) 4 on May, but not in February, making a linkage questionable. N 2 fixation rates (unquantifiable to 51.9 nmol N L −1 d −1 ; mean over the survey: 6.5 ± 6.7 nmol N L −1 d −1 ) are in line with those measured previously in the lagoon (8.0-27.7 nmol N L −1 d −1 ; Biegala and Raimbault, 2008;Berthelot et al., 2015;Bonnet et al., 2016) and open-ocean mixed layer waters around the Melanesian archipelagos (unquantifiable to 48 nmol N L −1 d −1 , mean 8.9 ± 10.0 nmol N L −1 d −1 , Bonnet et al., 2018), but lower than those off the tropical north-northeast coast of Australia (3 to 68 nmol N L −1 d −1 , mean 32.0 ± 24.0 nmol N L −1 d −1 ; Messer et al., 2017). When compared with Pacific surface water rates at similar latitudes, including the Eastern tropical South Pacific (∼0.9 nmol N L −1 d −1 ; Dekaezemacker et al., 2013) and the tropical South Pacific Gyre (∼0.5 nmol N L −1 d −1 ; Raimbault and Garcia, 2008;Bonnet et al., 2018), the rates reported here are high. Indeed, our observations, along with those previously reported in the WTSP (Montoya et al., 2004;Bonnet et al., 2009Bonnet et al., , 2015Bonnet et al., , 2018Messer et al., 2015Messer et al., , 2017Berthelot et al., 2017), reinforce the view of the WTSP region as a hotspot of diazotroph activity within the southern hemisphere . N 2 fixation rates were higher (although not statistically significant) during summer, but the year to year variability was more pronounced than the seasonal variability with significantly higher N 2 fixation rates in the second year of the survey (2013; Figure 3), consistent with low PO 4 3− concentrations. This pattern was mostly dictated by offshore stations and are synchronous with higher Trichodesmium and UCYN-C abundances (by a factor of 2.5) at station L2 in the second year. The explanation for such an increase is unclear and may be due to a combination of: (i) higher seawater temperatures (25.44 • C) in austral summer 2013 compared to 2012 (24.67 • C), which may have stimulated diazotroph activity (Stal, 2009), (ii) higher occurrence of S-SW winds in 2013 that may have advected more Trichodesmium from the open-ocean. Notably, the ENSO index and rainfall did not differ significantly between the 2 years and are, therefore, unlikely to explain the observed differences.
With a few exceptions (Garcia et al., 2007;Bonnet et al., 2015), previous N 2 fixation studies from the New Caledonian lagoon and the WTSP were conducted over the austral summer, which is considered the main N 2 fixation season in the region (Garcia et al., 2007). Our data suggest that seasonality in the New Caledonian lagoon is less pronounced than previously thought, and that relatively high N 2 fixation rates are maintained all year long, despite seasonal changes in the dominant diazotrophs. During the 2 years of study in the New Caledonian lagoon, the composition of the diazotrophic community was variable, but provided a constant N input.
Assuming that diazotrophs fix on average 115 ± 47 µmol N m −2 d −1 throughout the year (the average of the four stations) and considering a southwestern lagoon area of 2,066 km 2 , planktonic N 2 fixation would introduce 0.6-1.7 × 10 9 g N year −1 to the lagoon. For comparison, the inorganic N inputs from the three major rivers flowing into the southwestern lagoon (Dumbea, Pirogue and La Coulée rivers) calculated from mean annual river flows from Direction of Food and Rural Veterinary Affairs (Direction des Affaires Vétérinaires Alimentaires et Rurales, DAVAR, New Caledonia) and dissolved NOx concentrations data would introduce 0.03 × 10 9 g N year −1 , i.e., 22-56 times less than N 2 fixation. Consequently, planktonic N 2 fixation represents a significant source of N for biological communities of the southwestern lagoon. In comparison, Charpy et al. (2007) reported mean benthic N 2 fixation rates for the southwestern lagoon of 293 ± 96 µmol N m −2 d −1 , i.e., three times higher than planktonic N 2 fixation. Collectively, these results suggest that N 2 fixation plays a critical role in sustaining the productivity of the New Caledonian waters, which features an exceptional diversity and is listed as a UNESCO World Heritage Site.

CONCLUSION
Several studies have pinpointed the importance of N 2 fixation for sustaining both planktonic  and benthic (Benavides et al., 2016) productivity during the summer season in the New Caledonian lagoon, but here we document that high fixed N input takes place also during the cold season. Our analyses suggest that the meteorological and physicochemical characteristics of New Caledonian lagoon waters lead to the heterogeneity of diazotrophic populations both at seasonal and annual timescales. The UCYN-A symbiosis dominate the diazotrophic community, even during warmer, summer months, and likely provide a constant background of fixed N to surrounding waters, together with other UCYN groups (UCYN-B and UCYN-C). Trichodesmium is second most abundant group, peaking preferentially during summer months, and adding additional fixed N during the summer. However, the year to year variability was more pronounced, highlighting the need for longer time-series studies with fine temporal and geographical resolution to refine our understanding of the environmental factors regulating diazotrophs.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.