Global Biogeochemical Cycles

Nitrate isotopic composition between Bermuda and Puerto Rico: Implications for N2 fixation in the Atlantic Ocean

Authors


Abstract

[1] N and O isotope analyses of water column nitrate between Bermuda and Puerto Rico document a bolus of low-δ15N nitrate throughout the Sargasso Sea thermocline, which we attribute primarily to the input of recently fixed N. Although previous work suggests southward increases in N2 fixation and ventilation age, no meridional trend in nitrate δ15N is apparent. In the upper 200 m, the algal uptake-driven increase in nitrate δ18O is greater than in δ15N, because of (1) a higher fraction of nitrate from N2 fixation at shallower depths and/or (2) cycling of N between nitrate assimilation and nitrification. A mean depth profile of newly fixed nitrate estimated from the nitrate isotope data is compared with results from an ocean circulation model forced with different Atlantic fields of N2 fixation. The nitrate from N2 fixation is communicated between the model's North and South Atlantic and suggests a whole Atlantic N2 fixation rate between 15 and 24 Tg N a−1. One important caveat is that fixed N in atmospheric deposition may contribute a significant proportion of the low-δ15N N in the Sargasso Sea thermocline, in which case the relatively low rate we estimate for N2 fixation would still be too high.

1. Introduction

[2] Di-nitrogen (N2) fixation by marine diazotrophs is the dominant means by which the ocean receives biologically available nitrogen (N), a critical nutrient for phytoplankton growth. Quantifying this N input has been an objective of biological oceanographers for decades [Dugdale et al., 1964; Carpenter and McCarthy, 1975; Carpenter and Price, 1977; Carpenter and Romans, 1991; Capone et al., 1998; Montoya et al., 2002]. More recently, this work has been motivated by the suggestion that rates of denitrification, the main sink for oceanic fixed N, greatly exceed N2 fixation rates [Codispoti, 1995; Codispoti et al., 2001; Brandes and Devol, 2002; Codispoti, 2007]. Developing robust estimates of the global rate and distribution of N2 fixation from “direct” shipboard measurements of N2 fixing activity is complicated by the inherent spatial and temporal variability of this biologically mediated flux. Thus, geochemical approaches for estimating N2 fixation inputs are becoming an increasingly important complement to shipboard assays.

[3] One region that has received attention from researchers making both in situ biological as well as geochemical estimates of N2 fixation is the North Atlantic Ocean, and the Sargasso Sea in particular [Dugdale et al., 1964; Carpenter and McCarthy, 1975; Carpenter and Price, 1977; Altabet, 1988; Michaels et al., 1996; Gruber and Sarmiento, 1997; Carpenter et al., 1999; Orcutt et al., 2001; Carpenter et al., 2004; Hansell et al., 2004; Knapp et al., 2005; K. M. Achilles, 2004, Bioavailability of iron to Trichodesmium colonies and other ecologically important cyanobacterial cultures: Implications for global carbon and nitrogen cycling. Ph.D. thesis, University of Delaware, unpublished]. Several geochemical estimates have used deviations from the ∼16:1 concentration ratio of nitrate to phosphate ([NO3]/[PO43−]) expected from the remineralization of marine biomass to identify and quantify N2 fixation in this region of the North Atlantic (e.g., N*, which is defined as [NO3] − 16 × [PO43−] + 2.9 μmol/kg [Michaels et al., 1996; Gruber and Sarmiento, 1997; Hansell et al., 2004, 2007]. In the North Atlantic basin, N* has been observed to increase with ventilation age on isopycnals within the thermocline, yielding a rate of N2 fixation that, when scaled to the global ocean, led to a significant upward revision of the whole ocean N2 fixation rate [Gruber and Sarmiento, 1997]. While the nutrient-ratio approach is a critical new tool that has transformed the study of the marine N cycle, it does have significant uncertainties, motivating the development of complementary geochemical approaches.

[4] The isotopic analysis of nitrogenous species provides such a complementary tool. Newly fixed N has a δ15N close to that of atmospheric N2 (∼−2–0‰ [Hoering and Ford, 1960; Minagawa and Wada, 1986; Carpenter et al., 1997] δ15N (‰) = {[(15N/14N)sample/(15N/14N)reference] − 1} × 1000, where the reference is atmospheric N2). In contrast, mean ocean NO3 has a δ15N close to 5‰ [Sigman et al., 2000]. Thus, on a local or regional basis, the collective δ15N of a system will decrease as N2 fixation increases relative to NO3 supply from the mean ocean pool. Other processes can redistribute the N isotopes so as to lead to local 15N depletion in some N pools (e.g., N recycling and its effect on the δ15N of suspended particles in the surface ocean [Checkley and Miller, 1989]), so the occurrence of low 15N/14N in any given N pool should not immediately be assumed to result from N2 fixation inputs [Altabet, 1988; Karl et al., 2002; Knapp et al., 2005]. Nevertheless, the recent demonstration that NO3δ15N decreases upward from the deep ocean into the Sargasso Sea thermocline to a minimum of 2 to 3‰ is hard to explain without the input of low-δ15N N [Karl et al., 2002; Knapp et al., 2005], and this interpretation is consistent with the high N* of the Sargasso Sea thermocline [Michaels et al., 1996; Gruber and Sarmiento, 1997].

[5] The fixed N in atmospheric deposition to the Atlantic, while inadequately studied, appears to have a δ15N that is similar to or lower than that of oceanic N2 fixation (at Bermuda, −4.6‰ for NO3 and −2.3‰ for total fixed N [Hastings et al., 2003; Knapp et al., 2008] such that it will also lower thermocline NO3δ15N [Knapp et al., 2008] while raising its [NO3]:[PO43−] ratio [Gruber and Sarmiento, 1997; Hansell et al., 2007]. Previous estimates of the rate of atmospheric N deposition are low [e.g., Knap et al., 1986] relative to geochemical estimates of N2 fixation rates by Gruber and Sarmiento [1997] and Hansell et al. [2004] on the basis of N*. Thus, N2 fixation has been taken as the dominant cause for high N* and low NO3δ15N in the Sargasso Sea thermocline [Gruber and Sarmiento, 1997; Karl et al., 2002; Knapp et al., 2005], although it remains possible that future work will alter this view.

[6] The N isotope data for NO3 could be used to quantify newly fixed N in the subtropical NO3 pool and thus the rate of N2 fixation in the overlying surface waters. One can imagine doing so in two ways. First, one could take an approach analogous to that of Gruber and Sarmiento [1997] and quantify changes in NO3δ15N on specific isopycnal surfaces as a function of time since the water left the surface. Second, one could use the isotope data to estimate the quantity of newly fixed N in the Sargasso Sea thermocline. This quantity, for a given exchange time between the thermocline with the deep ocean NO3 reservoir, would yield an area-normalized rate of N2 fixation. The latter approach is employed in this study.

[7] One might also hope to gain insight into the spatial distribution of N2 fixation from the NO3δ15N of the Sargasso Sea thermocline. Between Bermuda and Puerto Rico, coherent gradients exist in maximum winter mixed layer depth [Levitus and Boyer, 1994; Kara et al., 2003], modeled and measured atmospheric dust deposition [Gao et al., 2001; Sedwick et al., 2005], and measured in situ N2 fixation rates [Carpenter and Price, 1977; Orcutt et al., 2001; Capone et al., 2005; Achilles, 2004]. These gradients, along with a southward increase in the CFC-derived ventilation age of the thermocline (∼2 years between Bermuda and Puerto Rico [Hansell et al., 2004, and references therein]), raise the possibility of a southward decrease in NO3δ15N in this region.

[8] We report measurements of NO3δ15N and δ18O for water column samples collected along a meridional transect between Bermuda (32°N) and Puerto Rico (19°N) (Figure 1) in October 2002, during or immediately following the late summer peak in N2 fixation typically observed in this region ([Orcutt et al., 2001]) δ18O (‰) = {[(18O/16Osample)/(18O/16Oreference)] − 1} × 1000, where the reference is Vienna Standard Mean Ocean Water, or VSMOW). The NO3 isotope profiles from the transect, as well as an additional profile from 10° west of the transect, indicate minimal lateral variation in thermocline NO3δ15N within the region investigated. On the one hand, the coherent 15N depletion of NO3 in the shallow thermocline across the entire transect suggests that there is a large bolus of recently fixed N in the Sargasso Sea thermocline. On the other hand, the lack of a clear north-south trend in the NO3 isotope data implies that N2 fixation rates are slow relative to the homogenization of the NO3 pool associated with the physical circulation of the Sargasso Sea. Following the arguments outlined above, the NO3δ15N data are used to estimate a regional inventory of N2 fixation-derived NO3 in the thermocline. In this context, we use NO3δ18O to determine when local NO3 assimilation has affected the NO3δ15N in the upper water column. We then compare our inventory of “recently fixed NO3” with an ocean global circulation model forced with several alternative Atlantic basin-wide distributions of N2 fixation.

Figure 1.

Mercator projection of the western North Atlantic. Cruise stations are shown as white circles. The dash-dotted line indicates the two adjacent model boxes, together yielding a swath centered on 65.6°W spanning 22.2°N to 31.1°N, that are compared with the observations. Gray scale corresponds to the March mixed layer depth from the Kara et al. [2003] climatology and indicates the decreasing importance of subsurface NO3 as a source of new N to surface waters with increasing stratification to the south. Higher N2 fixation rates, measured with bottle incubations, have also been observed to the south along this transect. Moreover, measured and modeled atmospheric dust input and measured thermocline, ventilation age are greater to the south. Thus, this transect would appear to be well situated to sample gradients in the importance of N2 fixation as a source of NO3 to the Sargasso Sea thermocline.

2. Sampling Locations and Protocol

[9] Samples were collected aboard the R/V Weatherbird II during BATS validation cruise number 32 (“BVAL 32”) in October 2002 at stations one degree apart between 31°N and 19°N (30°40′ N, 29°40′ N, etc.) between roughly 64°W and 65°W (Figure 1, white filled circles). The depths of water column sampling typically included 0, 20, 40, 60, 80, 100, 120, 150, 200, 300, 350, 400, 500, 600, 750, 800, 1000, and 1200 m, with different samplings among profiles mostly reflecting the BVAL effort to sample the core of Subtropical Mode Water (STMW), which is most often found between 300 and 400 m. Acid-washed 60 mL HDPE bottles were rinsed with sample water three times before filling and were frozen at −20°C until analysis. Additional water column samples were collected at the BATS site (31.4°N, 64.1°W), also aboard the R/V Weatherbird II, in October 2000 as part of the BATS core program (samples from this station are subsequently identified as “BATS”), and aboard the R/V Cape Henlopen in July 2001 at 32°N, 76°W (subsequently identified as “32°N, 76°W”), using similar sample collection protocols.

3. Methods

3.1. NO3 Concentration

[10] The sum of the NO3 and nitrite (NO2) concentration in samples was measured by reduction to nitric oxide (NO) using heated acidic V(III), followed by chemiluminescent detection of NO [Braman and Hendrix, 1989]. The detection limit was approximately 0.05 μM for NO3 + NO2 concentration. NO2 concentration in these waters rarely exceeds the detection limit for that analysis (0.001 μM [Lipschultz et al., 1996; Lipschultz, 2001]), so values are reported only as [NO3]. The standard deviation for these measurements is ≤2% for concentrations of 1 μM and higher.

3.2. NO3δ15N and δ18O

[11] For all water samples that have a [NO3] ≥ 0.5 μM, the δ15N of NO3 was determined by the “denitrifier” method described by Sigman et al. [2001] and Casciotti et al. [2002]. On the basis of sample replicates, the standard deviation for NO3δ15N analysis was 0.1–0.2‰. Samples were analyzed for NO3δ18O following the procedure and exchange and blank correction schemes described by Casciotti et al. [2002]. The standard deviation for this analysis was ∼0.5‰, with better precision for higher [NO3] samples. The reported δ18O values were adjusted from their calibration to a previously reported IAEA-N3 δ18O of 22.7‰ relative to VSMOW [Bohlke et al., 1997] to an updated value of 25.6‰ [Bohlke et al., 2003]. In addition, a 0.6‰ decrease was also applied to account for a change in the O isotope exchange correction, as described by D. M. Sigman et al. (The dual isotopes of deep nitrate as a constraint on the cycle and budget of oceanic fixed nitrogen, submitted to Deep Sea Research, Part I, 2008). Taken together, these explain the ∼2.2‰ increase in deep NO3δ18O reported here relative to previous work in our lab [e.g., Sigman et al., 2005].

4. Results

[12] NO3 is typically below the detection limit above 150 m, although individual water samples between 40 and 80 m depth occasionally had measurable [NO3], ranging from 0.1 to 1.0 μM. Water column profiles show [NO3] increases with depth from ∼1 μM at 150 m to a maximum of ∼25 μM at 800 m (Figures 2a, 3a, S1, and S2). The vertical structure of NO3δ15N is similar to that observed at the BATS site [Knapp et al., 2005]. NO3δ15N decreases from ∼5.3‰ at ∼1000 m to an average minimum of ∼2.6‰ at 200 m (Figures 2b, 3b, and 4) . Above 200 m, NO3δ15N increases to as much as 7.3‰. NO3δ18O also increases above 200 m to as much as 10.0‰, which represents a greater increase than in NO3δ15N over the same depth range (Figures 2c, 3c, and 4); this is discussed below in section 5.1.

Figure 2.

Depth and isopycnal profiles of [NO3] (a and d), NO3δ15N (b and e), and NO3δ18O (c and f). Individual station data are shown as plus symbols, with color indicating station location, and the meridional transect average profile is shown with open black circles. Transect means were calculated by interpolation of individual station profiles onto a uniform depth grid and then averaging. For all stations, there is a decrease in NO3δ15N from 650 m to a subsurface minimum of ∼2‰ at ∼200 m, indicating the accumulation of low-15N NO3, most likely from the remineralization of recently fixed N. The additional station from 32°N, 76°W is also plotted (pink filled triangles), with its NO3δ15N profile showing remarkable similarity to the BATS and BVAL stations, especially when plotted versus density.

Figure 3.

Sections of [NO3] (a), δ15N of NO3 (b), δ18Osal of NO3 (c) and Δ(15, 18) (d) from 32°N to 19°N versus depth at ∼65°W. There is no clear trend in NO3δ15N across the transect in the depth range of the upper thermocline (600 m and shallower). The observed variability occurs on a smaller spatial scale, apparently driven by patchiness in the isotopic composition of NO3 in the shallow subsurface (i.e., upper 200 m), which is largely due to fractionation during NO3 assimilation near the top of the nitracline at ∼150 m (see Figure S1). At ∼1000 m depth, there is a southward increase in NO3δ15N, potentially the signature of southern sourced intermediate water. Plotting the transect data versus density yields a similar picture (Figure S2).

Figure 4.

Depth profiles of cruise average NO3δ15N (pluses), NO3δ18O (open circles), and Δ(15, 18) (open triangles). There is a decrease in Δ(15, 18) from ∼800 m to 300 m as NO3δ18O remains nearly constant and NO3δ15N decreases. The salinity-corrected version of NO3δ18O (filled circles, NO3δ18Osal) indicates the increase in measured NO3δ18O from 800 m to 300 m can be fully explained by an increase in the δ18O of ambient seawater incorporated during nitrification. Removing this effect, the decrease in Δ(15, 18) observed from ∼650 m to 100 m (filled triangles) is due either to an increase in the fraction of recently fixed NO3 upward in the water column and/or to a partial NO3 assimilation-remineralization cycle occurring in the shallow subsurface (see section 5.1).

[13] There is no clear trend in NO3δ15N or δ18O along the transect in the upper thermocline (600 m and shallower; Figures 3 and S2). NO3δ15N in particular, when plotted versus density, is remarkably uniform across the Sargasso Sea (Figures 2e and S2b). The observed variability is of smaller scale, mostly associated with patchiness in the isotopic composition of NO3 in the shallow subsurface (i.e., upper 200 m), which is largely due to fractionation during NO3 assimilation near the top of the nitracline at 150 m (see section 5.1 below).

[14] CFC-12 ventilation ages for this region suggest a 2-year increase in the age of the thermocline to the south between 32°N and 19°N [Hansell et al., 2004, and references therein], so an isotopic gradient might be expected if the processes affecting thermocline NO3δ15N (e.g., N2 fixation) were adequately rapid. However, calculations for the 26.4 sigma theta surface based on the accumulation of NO3 from N2 fixation added to the upper 600 m at the rate of 0.072 mol N m−2 a−1 [Gruber and Sarmiento, 1997] would translate into only a ∼0.1‰ decrease of NO3δ15N toward the south along this transect. Given the uncertainty in the isotope measurements and other natural sources of variability, this change would be difficult to detect. Still, it remained possible that a strong meridional gradient in N2 fixation rate would have been apparent in the data.

[15] The additional profile of NO3 isotopes from the Gulf Stream region to the west of the BVAL transect (Figure 1, white circle at 32°N, 76°W) displays remarkable uniformity with the BVAL and BATS stations, especially for NO3δ15N (Figure 2; compare pink triangles to black circles). This suggests that there is little zonal variability in NO3 isotopes within the Sargasso Sea, mirroring the absence of meridional variability indicated by the BVAL section, although additional data are needed.

[16] As previously observed, NO3δ18O in the ocean subsurface is close to that of ambient water ([Casciotti et al., 2002; Sigman et al., 2005; Lehmann et al., 2005] ∼2‰ versus VSWOW using the most recent NO3 reference information). However, a slight vertical gradient is apparent in our data, with NO3δ18O increasing from ∼1.8‰ at 1000 m to ∼2.7‰ at 300 m. Over the same depth range, salinity increases by 1.4 practical salinity units (psu), which should correspond to an upward increase in the δ18O of water of 0.7‰ [Bigg and Rohling, 2000; G. A. Schmidt et al., 1999, Global Seawater Oxygen-18 Database, available at http://data.giss.nasa.gov/o18data/]. Since water appears to be the dominant source of O atoms for NO3 [Casciotti et al., 2002], nitrification in ambient water should cause an increase in the δ18O of NO3 over the same depth range. Since more than 50% of the NO3 at all thermocline depths along this transect is regenerated [Palter et al., 2005], the nitrification of the majority of the NO3 occurred at in situ δ18O. Moreover, even the preformed NO3 was likely nitrified in waters with a δ18O that is similar to the water in which it currently resides. Thus, we attempt to correct for the effect of salinity-driven depth variations in the δ18O of water on the δ18O of NO3. We refer to this as “salinity-corrected” NO3δ18O:

equation image

where “NO3δ18O” is the measured oxygen isotopic composition of a NO3 sample, “sal” is the salinity of a sample (in psu), and “salm” is the mean salinity at 1000 m along the cruise transect (35.05 psu). The 0.5 factor applied to the difference in salinity between shallower samples and 1000 m derives from the approximate slope of the upper ocean relationship between seawater δ18O and salinity [Bigg and Rohling, 2000; Schmidt et al., 1999]. The difference between NO3δ18O and NO3δ18Osal increases up through the water column from 1000 m since salinity also increases up through the water column; NO3δ18Osal is not significantly higher at 300 m than at 1000 m (Figure 4, black filled circles), with a weak (∼0.3‰) δ18O maximum between 600 and 800 m.

[17] Culture studies indicate that algal NO3 assimilation causes both N and O isotope fractionation [Granger et al., 2004]. Thus, the increase in both the δ15N and δ18O of NO3 into the upper 200 m provides strong support for the interpretation that this heavy isotope enrichment results from NO3 assimilation. While this assimilation may have occurred in the higher-latitude surface waters that were subducted into the thermocline, it appears that the sharp increase in NO3δ15N and δ18O above 300 m can be explained locally by assimilation at the base of the euphotic zone. The fluorescence maximum along BVAL 32 is at ∼125 m, so that the base of the chlorophyll peak seems to overlap with our upward increase in NO3δ15N and δ18O at 150–200 m depth (Figure S1). Slow NO3 assimilation in these waters would drive an increase in the NO3δ15N and δ18O as observed.

5. Interpretation and Discussion

5.1. Impact of NO3 Assimilation on NO3δ15N and δ18O

[18] In sections 5.2 and 5.3, we use the depression of Sargasso Sea thermocline NO3δ15N below that of mean ocean NO3δ15N to estimate the rate of N2 fixation in the Atlantic Ocean. However, before we can proceed, we must address other N cycle processes that might cause NO3 in the Sargasso Sea to diverge from the mean ocean δ15N of ∼5‰.

[19] Water column denitrification is a major cause of NO3δ15N enrichment in the ocean interior, and it appears to account for most of the elevation of global ocean NO3δ15N above that of oceanic N2 fixation, the dominant N input to the ocean [Brandes and Devol, 2002, Deutsch et al., 2004]. However, water column denitrification within the Atlantic is thought to be trivial, because of the near-complete lack of water with <5 μM [O2], and thus is neglected in calculations in sections 5.2 and 5.3.

[20] The isotopic imprint of NO3 assimilation, in contrast, is likely to be important in our data. Within our depth profile, NO3δ15N increases above 200 m. If the upward NO3δ15N increase is due to NO3 assimilation, then it will artifactually lower our estimate of the input of low δ15N-N from N2 fixation in the shallowest samples. Since culture studies indicate a 1:1 increase in NO3δ15N and δ18O during algal NO3 assimilation and respiratory denitrification [Granger et al., 2004, 2008], one might attempt to use the O isotopes to correct for the effect of NO3 assimilation on the δ15N of shallow NO3. Sigman et al. [2005] used an analogous approach to remove the component of NO3δ15N and δ18O variation in the eastern North Pacific subsurface that results from water column denitrification, in order to quantify an apparent low-δ15N (or high-δ18O) residual. The derived parameter in that study, Δ(15, 18), quantifies the difference in NO3δ15N from the expected covariation with NO3δ18O:

equation image

where δ15Nm and δ18Om were defined by the δ15N and δ18O of NO3 in eastern North Pacific deep water (“m” for mean deep water), δ15N and δ18O are the measured isotopic composition of NO3 in a given sample, and ɛ, or the “isotope effect,” is defined in the case of the N isotope effect (15ɛ) as (14k/15k − 1) × 1000‰, where 14k and 15k are the rate coefficients of the reactions for the 14N- and 15N-bearing forms of NO3, respectively. In the work of Sigman et al. [2005], the 15ɛ/18ɛ is the N-to-O isotope effect ratio for denitrification, estimated from culture studies to be ∼1 [Granger et al., 2008]. Δ(15, 18) can also be used in the present study, with the modifications that δ15Nm and δ18Om are the mean δ15N and δ18O of NO3 in the deep Atlantic, and 15ɛ/18ɛ is the isotope effect ratio for NO3 assimilation, which culture studies also indicate to be ∼1 [Granger et al., 2004]. In this case, Δ(15, 18) quantifies the deviation of the relationship between the N and O isotopes from the behavior expected from NO3 assimilation alone, focusing on the Atlantic thermocline and surface and thus taking the underlying deep Atlantic water as a reasonable reference point. Given the large salinity gradients in this region, we favor the use of the salinity-corrected NO3δ18O profile when calculating Δ(15, 18) (see section 4); the corrected version is shown in Figure 3c (and S2c), while both are shown in Figure 4. δ15Nm and δ18Om are assigned as 5.36‰ and 1.80‰, taken from 875 m in our transect average profile, the depth with the highest δ15N, excepting the shallowest data. The significance of this assignment is clarified below, but one can think of it as simply defining Δ(15, 18) relative to a reference depth of 875 m.

[21] The upward decrease in Δ(15, 18) starting at 650 m and extending to 300 m is essentially entirely due to the upward decrease in NO3δ15N, at least in the salinity-corrected Δ(15, 18) profile (Figure 4). Above 300 m, depending on the station, both NO3δ15N and δ18O increase sharply, but NO3δ18O increases by ∼40% more than NO3δ15N (Figure S3). This greater increase in NO3δ18O causes Δ(15, 18) to continue decreasing above 300 m even though NO3δ15N increases toward the surface (Figure 4).

[22] If the increase in NO3δ18O in the upper 200 m results from incomplete NO3 assimilation, and the (15ɛ/18ɛ) of NO3 assimilation is known to be 1, one might argue that the Δ(15, 18) profile provides a better indication of N2 fixation inputs than does the NO3δ15N profile alone. However, we cannot currently argue that our data unambiguously support this interpretation, for reasons described below.

[23] For our purposes, there is a potential complication in the use of the NO3δ18O data to remove the effect of NO3 assimilation on the NO3δ15N data. Remineralization and nitrification of N assimilated from a given NO3 pool, by mass balance, must produce NO3 with a δ15N similar to that initially consumed. Thus, a perfect balance between NO3 assimilation and remineralization/nitrification at the base of the euphotic zone would have no effect on the δ15N of NO3 in this water. However, the δ18O of NO3 will drift upward with the onset of such an internal cycle because the δ18O of the NO3 consumed will initially be lower than the δ18O of the NO3 produced by nitrification (preliminarily estimated at ∼1.4‰ by D. M. Sigman et al. (submitted, 2008)). Thus, a cycle of partial NO3 assimilation and nitrification in the upper ocean can cause the δ18O of NO3 to increase relative to NO3δ15N [Granger et al., 2004; Sigman et al., 2005], an effect that has been documented in the upwelling zone off Monterey Bay and considered in detail by Wankel et al. [2007]. One strong suggestion that a cycle of partial NO3 assimilation and nitrification is at work in the Sargasso Sea is that the Δ(15, 18) minima coincide with NO3δ15N maxima (Figures 3b and 3d). If low Δ(15, 18) was solely a signal of N2 fixation, then there is no reason to expect lower Δ(15, 18) where especially large isotopic signals from NO3 assimilation are observed.

[24] In short, the upward decrease in Δ(15, 18) above 300 m may be due either to an increase in the fraction of NO3 from newly fixed N or to the existence of an internal cycle of NO3 assimilation and remineralization in the shallow subsurface, with the likelihood that both contribute. For this reason, we leave the explicit use of Δ(15, 18) to estimate N2 fixation fluxes for future work. We do, however, allow the shallow NO3δ18O data to influence one aspect of our use of the NO3δ15N data: the NO3δ18O data confirm that the δ15N increase into the upper 200 m is due to NO3 assimilation. With this support, we extend the minimum in NO3δ15N measured at 200 m up to shallower samples (i.e., we remove the increase in NO3δ15N above 200 m) when calculating the depth profile of the fraction of NO3 from recently fixed N (Figure 5). However, we do not use this correction in the depth integrations of newly fixed N, because the low δ15N NO3 removed from the upper water column by assimilation is added back to the deeper water column when the resulting export production sinks into the interior and is remineralized. In any case, this is not quantitatively significant for the water column integral of newly fixed N, because the waters in which NO3δ15N and δ18O elevation occur account for a minimal fraction of the NO3 in the Sargasso Sea thermocline.

Figure 5.

Estimation of newly fixed NO3 from NO3δ15N. (a) Cruise average NO3δ15N (plus symbols) is ∼5‰ from 1200 m to 800 m and decreases from 650 m to 200 m before increasing from 200 m into the surface. NO3δ15N corrected for assimilation using NO3δ18O data (filled circles) indicates this increase above 200 is fully explained by NO3 assimilation and is shown as filled circles (see text). Model assumptions for the δ15N of NO3 imported from outside the Sargasso Sea thermocline include 5.4‰, corresponding to the NO3δ15N of Upper Circumpolar Deep Water (UCDW), which is also the highest observed NO3δ15N in the transect average profile (open circles); 5.7‰, corresponding to the NO3δ15N of Antarctic Intermediate water (AAIW; cross symbols); and variable NO3δ15N corresponding to the δ15N on sigma theta surfaces from the Subantarctic (Subantarctic profile; open diamonds, see Text S1). (b) The imported NO3δ15N are used in a two end-member mixing model to generate a profile of the average fraction of recently fixed NO3 (assigned a δ15N of −1‰) (symbols follow from a). The fraction of recently fixed to total NO3 is greatest in the upper thermocline, reaching as high as ∼55%. (c) Multiplying this fraction by the cruise average total [NO3] gives the average recently fixed [NO3]. The concentration of recently fixed NO3 peaks at 2 to 4 μM from 300 m to 600 m.

[25] A more critical unknown in estimating the burden of recently fixed N is the δ15N of the NO3 imported into the entire system being studied. While mean ocean NO3δ15N is ∼5‰, the exact δ15N of NO3 being supplied to a system will depend on the water masses involved. For example, NO3δ15N measurements in the Subantarctic Zone of the Southern Ocean, which borders the Atlantic to the south, indicate a NO3δ15N of 4.7‰ in Lower Circumpolar Deep Water, 5.4‰ in Upper Circumpolar Deep Water, 5.7‰ in Antarctic Intermediate Water, and 6.5‰ in Subantarctic Mode Water (keeping in mind that these estimates are from the eastern Indian and Pacific sectors of the Southern Ocean [DiFiore et al., 2006; Sigman et al., 2000]). With regard to the potential use of the NO3 O isotopes to reconstruct the δ15N of the imported NO3, see Text S1.

5.2. Quantifying “Recently Fixed NO3” With the N Isotopes

[26] The spatial uniformity of NO3δ15N profiles that we observe in this study suggests that we have enough data coverage to compile a robust mean profile for NO3δ15N in the Sargasso Sea (Figure 5a). From this mean profile, we then estimate the amount of NO3 in the Sargasso Sea that is from “recently fixed N” (with a δ15N of ∼−1‰ [Hoering and Ford, 1960; Minagawa and Wada, 1986; Carpenter et al., 1997]), to be distinguished from NO3 that is imported by circulation from outside the tropical and subtropical Atlantic. The δ15N of this NO3 mixture can be expressed as follows:

equation image

where δ15Nmeas represents the measured NO3δ15N of a given sample (Figure 5a, solid circles), δ15NN2fix represents the δ15N of NO3 from the remineralization of recently fixed N (chosen as −1‰), δ15Nimported represents the δ15N of the NO3 that is imported by circulation into the Atlantic at its northern and southern boundaries (Figure 5a, open symbols and X's see below), and X represents the fraction of NO3 in a sample that originates from recently fixed N (Figure 5b). We can then calculate the [NO3] of a sample that results from recently fixed N by multiplying X by the measured [NO3] at each depth (Figure 5c). In the depth plot of the fraction and concentration of recently fixed NO3 (Figure 5b and 5c), we remove the increase in NO3δ15N that occurs in the upper 200 m (Figure 5a), as all evidence indicates this is a result of fractionation during NO3 assimilation.

[27] Since all NO3 N in the ocean was fixed at some time in the past, it is worth considering further what “recently fixed NO3” signifies. The proportion of the NO3 pool that the N isotopes associate with “recent” N2 fixation is the proportion that has not had the opportunity to be homogenized with the large global ocean NO3 reservoir, so that it does not yet bear the 15N enrichment associated with water column denitrification, most of which occurs in the Indo-Pacific. In this context, the loss process for this recently fixed NO3 pool is its gross flux out of the Atlantic basin. In the literature, “newly fixed” N typically refers to N that is supplied to the euphotic zone from N2 fixation, to be distinguished from the circulation-driven supply of NO3 from the subsurface, which in our case would be partially “recently fixed NO3,” or N fixed within the Atlantic that has not yet left the basin, and partially “imported” NO3, or NO3 in the model Atlantic that was brought in across the southern or northern boundary of the basin. We place the boundaries at 50°N and S, rather than having the southern boundary at the equator, in order to assume reasonable values for the NO3δ15N imported at the boundaries, as will be clarified by the model simulations described in section 5.3.

[28] We calculate profiles of recently fixed [NO3] on the basis of three different assumptions for the δ15N of imported NO3. In one case, we assume that the upper water column is supplied with NO3 solely from Upper Circumpolar Deep Water (UCDW) in the Southern Ocean, which borders the Atlantic to the south. The measured δ15N of UCDW is 5.4‰ [Sigman et al., 2000], which is the highest δ15N of NO3 observed in our mean Sargasso Sea profile, at 875 m, a depth at which water also has the same density as that of the UCDW (sigma theta ∼27.4). In a second case, we assume that the upper water column is supplied with NO3 solely by Antarctic Intermediate Water (with a δ15N of 5.7‰ [Sigman et al., 2000]), the density of which would have it entering our section at ∼750 m (sigma theta ∼27.2), with subsequent diapycnal processes mixing its NO3 vertically. In a third case, we assume that the δ15N of imported NO3 at any given density level matches the density- δ15N relationship of the Subantarctic interior [DiFiore et al., 2006] (see Text S1), which, as noted in section 5.1, sits at the southern boundary of the Atlantic basin. In this case, the shallow part of the Sargasso Sea section with densities less than those found in the interior of the Subantarctic are assumed to have an imported NO3δ15N of the lowest-density interior Subantarctic waters (i.e., that of Subantarctic Mode Water, 6.9‰). Two obvious weaknesses in these simplistic cases must be mentioned. First, the available Southern Ocean measurements used in the above cases are not from the Atlantic sector of the Southern Ocean, which would be most relevant to the circulation fluxes into the Atlantic. Second, our assumptions regarding imported NO3δ15N are derived solely from Southern Ocean data, because there is not yet an analogous data set for the high-latitude North Atlantic.

[29] Regardless of which assumption is made regarding imported NO3δ15N, the two end-member model predicts the largest fraction of recently fixed N in the shallow thermocline, where the lowest NO3δ15N is observed (Figure 5b). Within the STMW (i.e., “18°C water,” with a sigma-theta of 26.4 [Worthington, 1959]), we calculate that 40 to 50% of the NO3 originates from the nitrification of recently fixed N. Combining these results with the transect-average [NO3] measurements, we calculate a maximum in the [NO3] originating from recently fixed N at and just below the depth of STMW (∼300 to 600 m) of 2–4 μM (Figure 5c). From this approach, we estimate that recently fixed N has added 1.0–1.8 mol NO3 m−2 (Table 1) to the upper 1200 m of the Sargasso Seawater column.

Table 1. Inventories and Fluxes of Recently Fixed Nitrate in the Atlantic Ocean
 Recently Fixed [NO3] (mol N m−2)Total [NO3] (mol N m−2)RatioN2 Fixation Rates (Tg N a−1)a,b
  • a

    N2 fixation rate estimates for data-based inventories are calculated from the model N2 fixation rates, using least squares regression between prescribed fixation rate and model ratio of recently fixed to total [NO3].

  • b

    Rates from Hansell et al. [2007], Gruber and Sarmiento [1997], and Deutsch et al. [2007].

  • c

    Imported NO3δ15N is 5.4‰, typical for Upper Circumpolar Deep Water [Sigman et al., 2000].

  • d

    Imported NO3δ15N is 5.7‰, typical for Antarctic Intermediate Water [Sigman et al., 2000].

  • e

    Imported NO3δ15N values are set by the potential density to δ15N relationship in a profile from the Subantarctic Mode Water formation region of the Southern Ocean (P. DiFiore, unpublished data, 2008).

  • f

    Fixed N input rate from Hansell et al. [2007] is applied to the North Atlantic (10–35°N).

  • g

    Fixed N input rate from Hansell et al. [2007] is applied to the North Atlantic (10–35°N) and also extended to the South Atlantic (10–35°S).

  • h

    N2 fixation field from Deutsch et al. [2007].

  • i

    N2 fixation flux from Gruber and Sarmiento [1997] is applied to the North Atlantic only (0–45°N).

  • j

    N2 fixation flux from Gruber and Sarmiento [1997] is extended onto the southern Atlantic basin.

Data
UCDWc1.1616.270.0715.0
AAIWd1.8416.270.1123.7
Subantarctic Profilee1.6716.270.1021.5
 
Models
Hansell If0.8117.800.0510.92
Hansell IIg1.0318.040.0621.84
Deutschh2.3018.190.1327.85
GS Ii3.2619.200.1728.10
GS IIj5.6321.710.2656.20

5.3. Estimating the Rate of N2 Fixation

[30] The calculations described in section 5.2 allow us to estimate an inventory of recently fixed NO3 in the Sargasso Sea thermocline. In order to convert that inventory into a N2 fixation rate, we must divide that inventory by the appropriate timescale over which this recently fixed NO3 has accumulated. We incorporate this timescale implicitly through a suite of general circulation model experiments in which various fields of N2 fixation were prescribed on the basis of previous N2 fixation rate estimates for the Atlantic Ocean [Gruber and Sarmiento, 1997; Hansell et al., 2007; Deutsch et al., 2007]. Each of these experiments yielded an Atlantic distribution of recently fixed NO3 which was then compared with that derived from the transect-average NO3δ15N profile. We used the third release of the Princeton/GFDL modular ocean model (MOM 3.0 – P2A configuration [Pacanowski and Griffies, 1999; Gnanadesikan et al., 2004; Matsumoto et al., 2004]) with modifications described below. The model has 3.75° horizontal resolution, with 24 depth levels. The focus of this study was solely the N2 fixation input to the Atlantic Ocean. To that end, the model domain was limited to the Atlantic basin by imposing “sponge walls” for NO3 at 50°N and 50°S restored at all depths to the Levitus and Boyer [1994] ocean atlas [NO3] values. Productivity was parameterized by nutrient restoring [Najjar et al., 1992] with a meridionally graduated (hyperbolic tangent) restoring timescale that was tuned to fit the preformed [NO3] profile at the grid cells corresponding to the transect observations. Modeled “recently fixed NO3” and total NO3 were tracked as separate tracers that were allowed to evolve within the model circulation, each reaching steady state within the 1200-year run time. For the two model grid cells representing our Bermuda-to-Puerto Rico transect (Figure 1, dashed box), profiles of [NO3], recently fixed [NO3], and the fraction of recently fixed to total NO3 were generated and depth-integrated. These profiles and integrated values are compared to the transect data (Figure 6 and Table 1). Both the concepts and model-specific considerations of this approach will be described in more detail in a separate manuscript (C. Deutsch et al., in preparation, 2008).

Figure 6.

Depth profiles of (a) recently fixed [NO3], (b) nonrecently fixed [NO3] (i.e., total [NO3] minus recently fixed [NO3]), and (c) the ratio of recently fixed to total [NO3] from our N isotopic data along the BVAL transect (gray lines; dashing as in Figure 5) and the different model runs (connected black symbols). “GS I” and “GS II” refer to forcing of the model with an N2 fixation rate taken from the isopycnal N* based estimate for the North Atlantic of Gruber and Sarmiento [1997]; in GS I, an N2 fixation rate of 28.1 Tg N a−1 is applied evenly over 45°N to 0°N (open diamonds), while GS II extends the North Atlantic rate to the South Atlantic from 0°N to 45°S, yielding a total Atlantic rate of 56.2 Tg N a−1 (open squares). “Hansell I” and “Hansell II” refer analogously to a fixed N input rate taken from Hansell et al. [2007]; in the first case, 11 Tg N a−1 is applied between 35°N and 10°N, while the second case applies this and the same rate from 35°S to 10°S. Table 1 quantifies this data-model comparison through the ratio of water column integrals for recently fixed and total NO3 in each of the data- and model-based estimates.

[31] Model N2 fixation forcing schemes are based on three previous geochemical estimates for the North Atlantic. Gruber and Sarmiento [1997] calculated a mean North Atlantic N2 fixation rate of 0.072 mol N m−2 a−1 based on changes in N* along isopycnals between 10°N to 50°N, yielding a total flux of 28.1 Tg N a−1. Model “GS I” applies the Gruber and Sarmiento [1997] bulk rate estimate for the northern Atlantic, 28.1 Tg N a−1, homogenously across the model North Atlantic from 0°N to 45°N. The 45°N maximum latitude was set to limit numerical issues that occur when forcing at the sponge walls. Model “GS II” extends the GS I North Atlantic rate to the South Atlantic from 0° to 45°S, yielding 56.2 Tg N a−1 for the whole Atlantic. Model “Deutsch” applies the 50°N to 50°S distribution of N2 fixation estimated by Deutsch et al. [2007] on the basis of the upper ocean convergence of [NO3] and [PO43−], 27.85 Tg N a−1, equivalent to a rate of 0.032 mol N m−2 a−1, to the model Atlantic from 45°N to 45°S (again, the more equatorward limit is to avoid numerical issues at the sponge wall). This rate is somewhat higher than that reported by the authors (20 Tg N a−1 in Table 1 of Deutsch et al. [2007]), as they integrated from 40°S to 65°N. Hansell et al. [2007] also generated an estimate of N2 fixation in the North Atlantic based on [NO3] and [PO43−] and some additional considerations. Hansell et al. [2007] estimated a rate of excess NO3 accumulation due to both N2 fixation and atmospheric deposition of 5.62 × 1011 mol N m−2 a−1 occurring over the region of 10–35°N, where atmospheric N inputs and N2 fixation were of roughly equivalent importance. As described in section 1, Hastings et al. [2003] and Knapp et al. [2008] have found that atmospheric N inputs by wet deposition at Bermuda have a δ15N that is lower than or similar to that of newly fixed N. On this basis, we have combined Hansell's estimate of N2 fixation and atmospheric N deposition, as these two different sources of new N have similar impacts on the N isotopes. Similar to our treatment of the Gruber and Sarmiento [1997] estimate, we have carried out an additional simulation (Hansell II) in which the Hansell et al. [2007] estimate for the North Atlantic is accompanied by an equivalent input in the South Atlantic, at the same latitudes of that hemisphere.

[32] We compare the data and simulations by plotting with depth the concentration of recently fixed NO3 (Figure 6a) and the ratio of recently fixed NO3 to total NO3 (Figure 6c). Finally, while the vertical structure of the recently fixed NO3 pool is certainly of interest, our main focus is the column-integrated burden of recently fixed NO3. Thus, we compare these column integrals for the data and models, reporting both the quantities of recently fixed NO3 (in mol N m−2) and the ratio of recently fixed to total NO3 (Table 1). The latter is a better measure for model/data comparison, as it corrects for any data-model mismatch in nutrient depth distribution due to the model's nutrient restoring. The Deutsch et al. output falls between those of Hansell et al. [2007] and Gruber and Sarmiento [1997], so it is not plotted in Figure 6, but it is included in Table 1.

[33] While the model profiles of recently fixed NO3, like the observations, have a mid-depth maximum, the model's maximum is significantly deeper and not so sharply bounded at its base (Figure 6a). As described in section 5.2, there are uncertainties in the isotope-derived estimate of recently fixed NO3 due to end-member assumptions. Nevertheless, given that the base of elevated N* in this region is 700 m [Lipschultz et al., 2002; Hansell et al., 2004], indistinguishable from the depth at which NO3δ15N begins to decrease upward (Figures 2b and 3b), we suspect that the model's deep maximum in recently fixed NO3 is unrealistic. Possible explanations for this apparent model error include (1) too much vertical exchange in the model's tropical and subtropical thermocline, allowing remineralization to have a proportionally greater impact on NO3 throughout the water column [Gnanadesikan and Toggweiler, 1999]; (2) too great of a length scale for the remineralization of sinking N [Yamanaka and Tajika, 1996]; (3) too much N2 fixation in higher-latitude regions (or during winter months), causing recently fixed N to sink into dense subsurface waters, just below the surface, which then ventilate the deep ocean; (4) inadequate flow across the model's sponge walls at 50° N/S at ∼800 m, allowing a deep pool of recently fixed NO3 to accumulate. Only in the case of explanation (4) would the model's basin integrated excess burden of recently fixed N be due to something other than the imposed N2 fixation field, and this last explanation seems less likely than the others, since the model's meridional overturning circulation fits transient tracer constraints [Matsumoto et al., 2004]. We compare data- and model-derived column integrals of recently fixed NO3 down to a depth of 1200 m, the bottom depth of most of our profiles and a reasonable base for the integration, below which one enters the more slowly circulating deep ocean.

[34] The data indicate that the Gruber and Sarmiento [1997] N2 fixation rate, when extrapolated to the entire Atlantic (45°S to 45°N, GS II), yields too large a burden of recently fixed NO3 at all depths (Figure 6a; compare open black squares to the dashed, dotted, and solid gray lines). Additional model runs using Gruber and Sarmiento's areal N2 fixation rate of 0.072 mol N m−2 a−1 yield greater total fluxes of N2 fixation, build in more recently fixed NO3, and would produce even larger misfits (results not shown). In contrast, the lower North Atlantic N2 fixation rate of Hansell et al. [2007], with or without equal rates in the South Atlantic, yields a burden of recently fixed NO3 that is generally lower than the isotope-derived estimates (Figure 6a; compare black plus and black cross symbols with the dashed, dotted, and solid gray lines). Comparing our three isotope-based estimates of recently fixed [NO3] in the Sargasso Sea, which derive from our three cases for the δ15N of NO3 imported into the Atlantic, with a linear regression of imposed N2 fixation against the burden of recently fixed NO3 generated by all of the model cases, we derive a range of estimated Atlantic N2 fixation rates of 15 and 24 Tg N a−1 (Table 1).

[35] To better understand the link between N2 fixation and the accumulation of recently fixed NO3 in the model, we compare results from model experiments in which the area-normalized N2 fixation rate from Gruber and Sarmiento [1997] was applied variously to the North Atlantic alone, the South Atlantic alone, and across both (Table 2). When N2 fixation is imposed exclusively in the Northern and Southern Hemisphere of the Atlantic basin, respectively, 63% and 44% of the recently fixed NO3 accumulates in the northern basin (Table 2). This model result indicates that there is a large amount of north-south exchange of recently fixed NO3 in the thermocline. Such high exchange in the model suggests that the thermocline NO3δ15N signal is more or less a basin-wide signal, indicating the basin-average (not the regional or local) N2 fixation flux. This provides support for the exercise conducted above, in which data from the Sargasso Sea is used to infer N2 fixation for the entire tropical and subtropical Atlantic. Reflexively, it indicates that, even without N2 fixation in the South Atlantic, the δ15N of NO3 of its thermocline should be lowered by the flux of low δ15N from the North Atlantic. This would make it difficult to partition NO3 into recently fixed and imported fractions if we were to consider the North Atlantic only (i.e., if we placed our southern boundary at the equator). With N2 fixation only in the Northern Hemisphere, only 37% of the recently fixed NO3 from the North Atlantic accumulates in the South Atlantic (Table 2), suggesting ∼7% more transport northward than southward. Thus, the model displays a slight tendency for N2 fixation throughout the Atlantic to lead to the accumulation of recently fixed NO3 in the Sargasso Sea thermocline. This bias may be due to the net northward flow of warm water through the Atlantic, which balances the southward export of North Atlantic Deep Water from the basin.

Table 2. North-South Atlantic Exchange Experiments
BasinPercent Recently Fixed NO3 in North AtlanticaPercent Recently Fixed NO3 in South Atlantic
  • a

    Surface forced with the per square meter N2 fixation flux of Gruber and Sarmiento [1997], 0.072 mol N m−2 a−1. All values reported have been normalized to account for the area difference between basins. Here, the percentage refers to the recently fixed NO3 in the North or South Atlantic relative to that pool in the entire Atlantic between 50°S and 50°N, such that each row sums to 100%.

  • b

    Flux applied to the model North Atlantic only (0° to 45°N).

  • c

    Flux applied to the whole model Atlantic (45°S to 45°N).

  • d

    Flux applied to model southern Atlantic only (45°S to 0°).

North Atlantic onlyb63.136.9
North and South Atlanticc54.046.0
South Atlantic onlyd44.155.9

6. Conclusions and Directions for Future Work

[36] We have established here and in previous work at the BATS site [Knapp et al., 2005] that the δ15N of NO3 in the Sargasso Sea thermocline is ∼3‰ lower than that of deep NO3, suggesting an important input of low-δ15N N, the most obvious candidate for which is N2 fixation. This finding qualitatively supports previous studies based on elevated [NO3]/[PO43−] ratios indicating significant inputs of newly fixed N to the Atlantic thermocline [Michaels et al., 1996; Gruber and Sarmiento, 1997; Hansell et al., 2004]. Here, we have used our first spatial view of Sargasso Sea NO3δ15N, with some insight from NO3δ18O, to provide quantitative constraints on the Atlantic rate of N2 fixation.

[37] Our transect of depth profiles between Bermuda (∼32°N) and Puerto Rico (∼19°N) shows no clear meridional trends in NO3δ15N or δ18O. The constancy of NO3δ15N along the transect is consistent with in situ and geochemical estimates of regional N2 fixation relative to the rate of thermocline ventilation and circulation, which would predict only small gradients in these parameters that would be difficult to detect. In the future, with appropriate high-resolution models, data such as these could potentially be used to provide an upper limit constraint on the meridional gradient in N2 fixation within the Sargasso Sea. In spite of the lack of meridional gradients, large vertical (i.e., diapycnal) gradients in NO3δ15N indicate that recently fixed N is an important source of NO3 to the Sargasso Sea thermocline, providing 2 to 4 μM (40 to 50%) of the total NO3 in the STMW between ∼300 and 500 m.

[38] A comparison with ocean circulation model simulations reveals that our observations are consistent with a range of N2 fixation flux estimates of ∼15 to 24 Tg N a−1 to the whole Atlantic Ocean, with the range arising from the range in assumptions made for the δ15N of NO3 imported into the Sargasso Sea thermocline. This range is consistent with the flux estimates of Hansell et al. [2007] when applied to both the North and South Atlantic (22 Tg N a−1) and with that of Deutsch et al. [2007] (28 Tg N a−1 heterogeneously distributed across 45°N to 45°S). Moreover, given the additional assumptions that go into our estimates, we would not argue that they are inconsistent with the fixed N input estimate of Hansell et al. [2007] for the North Atlantic alone (11 Tg N a−1 evenly distributed between 10°N and 35°N) or with the N2 fixation field of Gruber and Sarmiento [1997] for the North Atlantic alone (28 Tg N a−1 evenly distributed between 0°N to 45°N). More significantly, extension of the area-normalized rate of N2 fixation estimated by Gruber and Sarmiento [1997] to the South Atlantic, which those authors employed in their estimation of the global ocean rate of N2 fixation, yields an Atlantic-wide N2 fixation field that clearly conflicts with the data. This last finding relates to important dynamics that arise in the model: circulation distributes the NO3δ15N signal of N2 fixation, making it a basin-wide thermocline signal, and the net northward flow in the upper water column of the Atlantic creates north/south asymmetry, leading to slightly preferential accumulation of recently fixed NO3 in the Sargasso Sea.

[39] There are various limitations to the present study, which have been discussed throughout section 5 but deserve reiteration. The δ15N of NO3 being fed into the Atlantic is a critical input to the calculation of recently fixed NO3, for which we have assumed several alternative cases in section 5.2. Whole ocean model simulations of the NO3δ15N field could be used to improve our knowledge of this term. However, there is also a need for NO3δ15N measurements from the subpolar regions that represent the sponge boundaries in our model simulations.

[40] One might also identify the isotopic impact of NO3 assimilation within the boundaries defined by the model as an uncertainty in our data/model comparison. This process can affect the vertical structure of NO3δ15N within the Atlantic water column. However, it will not impact our water column inventories of recently fixed NO3 unless, for instance, NO3 that becomes enriched in 15N within the Atlantic flows out of the basin across the defined model boundaries. In any case, future model simulations should attempt to explicitly include the isotopic impacts of NO3 assimilation, and new data from the margins of the subtropical gyres would aid in validation.

[41] Culture studies indicate that assimilation by algae elevates the δ18O and δ15N of NO3 similarly. Thus, one promise of NO3δ18O is to record NO3− 15N enrichment that has been countered by the input of low-δ15N N. Above 200 m, there is an abrupt increase in the δ18O of NO3 associated with NO3 assimilation, and NO3δ18O increases by roughly 40% more than does the δ15N of NO3 over the same depth range. The greater upward increase in NO3δ18O relative to δ15N in the upper 200 m could be due to an upward increase in the fraction of NO3 from N2 fixation. However, as described in section 5.1, NO3δ18O and δ15N may be decoupled when nitrification proceeds simultaneously with NO3 assimilation. If the uncertainties associated with this latter process could be overcome, NO3δ18O would become a more powerful tool to remove the impacts of NO3 assimilation on NO3δ15N.

[42] Another obvious weakness of our study is the limited spatial coverage of the data being compared to the model. Work by our group and others will continuously improve this situation, and we hope in the future to reapply our model analysis to a more exhaustive Atlantic Ocean NO3 isotope data set. The model's tendency to exchange recently fixed NO3 across the North and South Atlantic suggests that the thermocline NO3δ15N data we report for the Sargasso Sea reflects N2 fixation across the entire Atlantic, such that these data should provide a fair indication of the basin-averaged N2 fixation rate. At the same time, this cross-equatorial exchange is not complete, such that thermocline NO3 isotope data from the North Atlantic are more sensitive to N2 fixation occurring in the North Atlantic than that occurring in the South Atlantic.

[43] Another approach for estimating N2 fixation rates that will arise as the NO3 isotope data set grows is the isopycnal analysis used by Gruber and Sarmiento [1997] with N* data. In this case, the estimate of N2 fixation would derive from the decrease in δ15N along an isopycnal as it evolves from its preformed conditions at the ocean surface. Toward this end, transects sampling larger variations in thermocline ventilation age would prove most valuable. Such a sampling campaign, beyond improving spatial coverage, would provide a greater dynamic range in thermocline ventilation age with which to evaluate the development of the low NO3δ15N in the North Atlantic thermocline as it proceeds from its ventilation outcrop. In this vein, samples collected further to the north of this transect (i.e., ∼37 to 40°N) during the late winter/early spring may capture the STMW formation, allowing the preformed isotopic signature of this water mass to be characterized. Here again, the δ18O of NO3 promises to provide a constraint on the role of NO3 assimilation in isotopic variations, if the current unknowns in NO3 O isotope systematics can be addressed.

[44] We have generally assumed that N2 fixation accounts for all of the low-δ15N N input to the Atlantic required to produce the observed depletion in 15N within the Sargasso Sea thermocline. However, in our model simulations of the Hansell et al. [2007] fixed N flux estimate, we included those authors' estimate of the fixed N contribution from atmospheric deposition, on the basis that its δ15N is similar to that from N2 fixation [Knapp et al., 2008]. In that case, we thus implicitly assumed that a portion of the fixed N input described as N2 fixation is in fact atmospheric N deposition. Given our range of estimates for fixed N input to the Atlantic, 15 to 24 Tg N a−1 between 50°N and 50°S, our results leave open the possibility that a significant fraction of that input derives from atmospheric N deposition, which, for example, has been estimated at 5.0 ± 1.5 Tg N a−1 between 0° and 40°N in the Atlantic [Dentener et al., 2006]. That is, our estimates of N2 fixation alone may be too high because they do not explicitly account for the fraction of low-δ15N fixed N input that derives from atmospheric deposition. To progress from this situation, we require more work on the atmospheric flux of fixed N to the ocean and its isotopic composition.

[45] Finally, while we have focused on the isotopic composition of the NO3 pool, there are significant concentrations of dissolved organic N both in the surface ocean and in the thermocline [Hansell and Carlson, 2001; Knapp et al., 2005; Bronk et al., 1994; A. N. Knapp, 2006, The stable isotopic composition of dissolved organic nitrogen and nitrate in the subtropical ocean. Ph.D. thesis, Princeton University, unpublished]. This pool must be included if we are to develop a complete view of thermocline N fluxes and the constraints provided by the stable isotopes.

Acknowledgments

[46] We thank the crew and scientists aboard the R/V Weatherbird II for assistance with sample collection. Additionally, we thank D. McCorkle, C. Hintz, and J. M. Bernhard for collection of the 32°N, 76°W profile. Funding for this work was provided by a ASEE/NDSEG/DOD graduate fellowship and NOAA Climate and Global Change postdoctoral fellowship to A.N.K., by U.S. NSF Biocomplexity grants OCE-9981479 (to D.M.S., through the MANTRA project) and DEB-0083566 (to Simon Levin and Lars Hedin), by NSF CAREER grant OCE-0447570 to D.M.S., by NASA I.D.S. grant NNG04G091 to D.M.S., by NSF grant OCE-9911654 to J. M. Bernhard and D. McCorkle, and by BP and Ford Motor Company through the Carbon Mitigation Initiative at Princeton University. Figures 3 and S2 were generated with the application “Ocean Data View” provided by R. Schlitzer of AWI, Bremerhaven, Germany. This is BBSR/BIOS contribution 1704.

Ancillary