Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 We report nitrate (NO3−) nitrogen isotope ratios for seawater samples collected in the Subantarctic Zone of the Southern Ocean during both winter and summer as part of the Australian Antarctic CRC Subantarctic Zone (SAZ) Project. The concentration and 15N/14N of the wintertime surface nitrate are very close to those of the subantarctic thermocline. The 15N/14N of nitrate in the surface increases sharply into the summer even though there is little seasonal change in nitrate concentration. There are two possible end-member explanations for this observation. First, there may be significant equatorward nitrate transport during the summer, including a supply from the Antarctic surface. Second, the isotope effect of algal nitrate assimilation may be higher than has been estimated elsewhere, for example, for the seasonal sea ice zone of the Antarctic. We use a simple geochemical box model of the SAZ surface mixed layer as it evolves over the course of the summer to simulate salinity, nitrate concentration, and the 15N/14N of nitrate and sinking N. Our results suggest that a significant portion (∼30%) of the summertime SAZ nitrate is supplied from south of the Subantarctic Front and that N export is ≥3.5 mmol N m−2 d−1. Our approach also identifies the necessity of an isotope effect for nitrate assimilation in the SAZ of ≥7‰ and probably 8–9‰. Comparison to laboratory results suggests that this relatively high isotope effect may result from light limitation of algal growth in the SAZ.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The Southern Ocean represents a region of intense communication between the ocean and atmosphere as well as a critical junction in the exchange of waters between the cold, deep ocean and the warmer, low-latitude surface and thermocline. The combined effect of wind-driven surface water transport and geostrophy as well as the densification associated with surface cooling and sea ice formation result in the surfacing of deep CO2-charged, nutrient-rich water in the Antarctic. Despite intense export production in some parts of the Antarctic, the consumption of the major nutrients (N and P) is incomplete, so that excess CO2 in upwelled deep water escapes to the atmosphere. Westerly winds advect surface water northward across the Polar frontal zone and into the Subantarctic Zone (SAZ), where biological uptake draws nutrient concentrations to lower values northward and generates a sinking flux of organic matter that reinjects part of the escaped CO2 back into the ocean interior [Takahashi et al., 1997]. However, the major nutrients are not completely consumed in much of the Subantarctic, which may be due to limitation of phytoplankton growth by iron, silica, and/or light [Boyd et al., 1999, 2001; Debaar et al., 1995; Hutchins et al., 2001; Martin et al., 1990; Mitchell et al., 1991; Sedwick et al., 1999, 1997].
 The degree of SAZ nutrient drawdown affects not only atmospheric CO2 but also the nutrient content of the thermocline and thus the fertility and biogeochemistry of the low-latitude ocean. Nutrients are constantly being lost from the low-latitude surface ocean and thermocline by the rain of organic detritus that survives into the abyss. These nutrients may be dominantly resupplied to the upper ocean from the new middepth waters that form in the Subantarctic [Sarmiento et al., 2004]. If so, the nutrient supply to the subtropical, tropical, and equatorial surface ocean depends largely on the incomplete consumption of nutrients by algae in the Southern Ocean surface; if nutrients in the Southern Ocean surface were more efficiently depleted, the supply of nutrients to the low-latitude thermocline and surface ocean would be reduced [Keir, 1988; Matsumoto et al., 2001; Robinson et al., 2005; Sigman et al., 2003]. Increased nutrient drawdown could occur in either the Antarctic or the Subantarctic, but the nutrient concentration in the Subantarctic is the final determinant of the nutrient supply to the low-latitude thermocline. Thus the development of a quantitative and mechanistic understanding of subantarctic biogeochemistry is central to an understanding of global ocean biogeochemistry.
 The integrated rates of net primary production and export production estimated from direct measures are plagued by the spatial and temporal variability of the surface ocean as well as artifacts associated with making these measurements (e.g., bottle effects and sediment trap artifacts). Dissolved geochemical tracers are useful because they integrate over heterogeneities and do not involve manipulation of the organisms responsible for the biogeochemical fluxes. Over the last forty years, there has been great progress in characterizing the spatial distribution of bioactive species in the ocean. The distributions of dissolved inorganic carbon, alkalinity, dissolved oxygen, and the nutrients nitrate, phosphate, and silicate are relatively well-defined, and the databases are being continuously improved [Conkright et al., 2002; Key et al., 2004]. The use of these distributions to quantify physical and biological fluxes is an area of active research. If the ocean circulation was known, then the nutrient fields could be overlain on the circulation field to quantify the uptake of nutrients and carbon throughout the surface ocean (“nutrient restoring” [Deutsch et al., 2001; Dunne et al., 2005; Jin and Gruber, 2003; Najjar et al., 1992; Schlitzer, 2002]). However, uncertainties in the model-derived circulation compromise these estimates. In the surface ocean, where transport and mixing occur in complex patterns, it can be difficult to derive rates and patterns of nutrient assimilation from nutrient concentration fields because both assimilation and water exchange affect the nutrient concentration. Thus additional dissolved tracers with different sensitivities to physical and biological processes would be of great complementary use.
 Because this anticorrelation between the degree of nitrate consumption and nitrate 15N/14N is propagated to the sinking flux out of the surface ocean and to deep sea sediments, the N isotopes have been used as an indicator of past changes in nutrient utilization (the ratio of nutrient uptake to gross nutrient supply) in Southern Ocean surface waters [Altabet and Francois, 1994; Francois et al., 1992, 1997; Robinson et al., 2004, 2005; Sigman et al., 1999b]. There are several processes and parameters that must be understood for the paleoceanographic application of the link between nitrate utilization and the N isotopes in the Southern Ocean. The δ15N of the nitrate supply and the amplitude of isotope discrimination associated with nitrate assimilation are two key parameters relating nitrate utilization to the isotopic composition of nitrate in oceanic surface waters and of the sinking flux to the seafloor. Both parameters require quantification in the modern ocean, assessments of their variability, and the development of a mechanistic understanding of the parameters that control them.
 As shown previously, the possible sources of nitrate to the subantarctic surface have distinct nitrate δ15N-to-[NO3−] relationships (hereafter “δ15N/[NO3−]”) [Sigman et al., 1999a, 2000]. From the perspective of paleoceanographic studies, this largely represents an unwanted complexity. However, for modern ocean studies, it may prove useful in defining and quantifying the routes of nitrate supply to the subantarctic surface. Moreover, mixing of waters that have previously undergone nitrate assimilation has an effect on the δ15N of nitrate that is very different from the effect of nitrate assimilation alone: while [NO3−] mixes conservatively, the δ15N of nitrate of a volumetric mixture of waters is weighted toward the end-member with the higher [NO3−]. Thus assimilation and mixing in the upper ocean should be distinguishable when the concentration and isotope constraints are coupled.
 Much as in paleoceanographic studies, the major limitation on nitrate δ15N as a tool for modern upper ocean studies is uncertainty in the amplitude of N isotope discrimination associated with nitrate assimilation. This discrimination is quantified as the isotope effect, ɛ, which is defined as (14k/15k − 1), where 14k and 15k are the rate coefficients of nitrate assimilation for the 14N- and 15N-labeled forms of nitrate, respectively. Currently, estimates of ɛ from the ocean range between roughly 4 and 10‰ [Altabet and Francois, 2001; Karsh et al., 2003; Lourey et al., 2003; Sigman et al., 1999a], while the range observed in batch culture experiments is much greater, 0 to 20‰ [Granger et al., 2004; Montoya and McCarthy, 1995; Needoba et al., 2003; Waser et al., 1997]. The first Southern Ocean data yielded isotope effect estimates in the range of 4 to 6‰ [Sigman et al., 1999a]; however, subsequent estimates from the Southern Ocean near the polar frontal zone have yielded higher values (7–10‰) [Altabet and Francois, 2001; Karsh et al., 2003; Lourey et al., 2003]. There may be coherent spatial (i.e., environmentally driven) variation within the Southern Ocean and among nutrient-rich regions in general. For the degree of nitrate consumption to be derived from nitrate isotope data, a better understanding of the isotope effect and its controls must be developed. At the same time, evidence has arisen that the magnitude of the isotope effect is affected by the factors that limit phytoplankton growth [Needoba and Harrison, 2004; Needoba et al., 2004], so that reliable estimates of the isotope effect in the subantarctic ocean may have implications for what controls phytoplankton growth in this region.
 Previous nitrate isotope measurements from subantarctic samples collected during the austral summer show clear signs of net northward transport of nitrate in the subantarctic surface layer, such that the underlying thermocline cannot be the sole source of nitrate to the surface layer [Sigman et al., 1999a]. The entire data set was consistent with nitrate being supplied from the Polar Frontal Zone and being assimilated with an isotope effect of ∼5‰. However, this study included no samples collected during winter conditions, under which nitrate supply from the thermocline is most likely to occur.
 Here, we report new data for 15N/14N of nitrate from samples collected during both summer and winter conditions across the Subantarctic and Polar Frontal Zones south of Australia. While seasonal changes in surface nitrate concentration are unremarkable, we observe a dramatic summertime increase in nitrate 15N/14N across the entire Subantarctic. This indicates that the meridional gradient in nitrate concentration is purely a circulation feature during the winter, whereas it is strongly modified by assimilation during the summer. As in previous work [Sigman et al., 1999a], there is evidence for equatorward transport of nitrate, especially close to the Subantarctic Front, the poleward border of the SAZ. However, the new seasonal information indicates that the high-nitrate 15N/14N in the SAZ surface is generated in the very short time interval associated with summertime stratification and nitrate drawdown. Over this short time interval, surface salinity has been used as a constraint on water transport [Lourey and Trull, 2001; Wang et al., 2001], as ocean/atmosphere exchanges of freshwater have been found to be insignificant during the spring/summer productive season [Rintoul and England, 2002]. The modest summertime decrease in subantarctic surface salinity essentially puts an upper bound on the quantity of Antarctic water imported into the Subantarctic during the summer. The data set reported here also benefits from sediment trap collections in this region and the constraint that they provide on the 15N/14N of sinking N [Lourey et al., 2003]. While uncertainties exist in the interpretation of these data, combining them with the seasonally resolved nitrate isotope data provides at least a rough constraint on the amplitude of the isotope effect of nitrate assimilation. We explore the constraints provided by the available data with a simple geochemical model of the SAZ surface mixed layer subsequent to the onset of warm season stratification, with emphasis placed on evaluating the magnitudes of Antarctic nitrate supply and N export during these periods and on developing a revised estimate of the isotope effect of nitrate assimilation in the SAZ.
2. Sample Sets
 Water column depth profiles (“hydrocasts”) were collected from two companion sediment trap deployment/retrieval cruises as a starting basis for comparison of winter and summer seasons in the Subantarctic Zone south of Tasmania (Figure 1 and Table 1). Meridional transects of hydrographic profiles were collected along 142°E during September of 1997 (AU9701) and between 141° and 144.5°E during March of 1998 (AU9706), both aboard the R/V Aurora Australis by the Antarctic Cooperative Research Centre (CRC). During AU9701, 7 profiles were collected for nitrate isotopes, augmented by 40 samples taken from ∼7 m depth by the ship's underway system; during AU9706, 10 depth profiles were collected. Both cruises traverse the Subtropical front into the Subantarctic Zone and extend across the Subantarctic Front into the Polar Frontal Zone at ∼52°S (Figure 1).
Table 1. Cruises and Sampling
R/V Aurora Australis is operated by the Australian Antarctic Division.
R/V Southern Surveyor is operated by the Australian Commonwealth Scientific and Industrial Research Organization (CSIRO).
 To improve our interseasonal comparison, our sample set was augmented by subsequent underway collections from other SAZ transect cruises AU9804, AU9901, and SS9902 (Table 1 and Figure 1b). AU9804 and AU9901 transects were completed on the R/V Aurora Australis during 29 October to 23 December 1998 and 16 July to 6 September 1999, respectively. AU9804 included 47 underway surface samples, AU9901 included 46 underway samples and 3 hydrocasts. SS9902 completed on the R/V Southern Surveyor of the Australian Commonwealth Scientific and Industrial Research Organisation (CSIRO) during 9–16 February 1999, consisted of 14 underway collections and 2 hydrocasts at 44°S and 54°S.
 Samples were collected in acid- and deionized water–washed 500 mL HDPE bottles. The bottles were rinsed twice with sample water before filling. The samples were preserved for nitrate concentration and 15N/14N analysis by acidification with the addition of 0.5 mL 50% HCl to each 500 mL seawater sample, bringing the sample pH to between 2 and 3.
 Nitrate concentration ([NO3−]) was analyzed by standard colorimetric methods at the University of Tasmania–Antarctic CRC [Eriksen, 1997; Strickland and Parsons, 1972]. Some samples were reanalyzed at Princeton University by reduction of NO3− (and NO2−) to NO using a V(III) reagent followed by measurement of NO by chemiluminescence [Braman and Hendrix, 1989]. NO2− was always scarce and is subsequently disregarded. The 15N/14N of nitrate was analyzed by the “denitrifier method” [Casciotti et al., 2002; Sigman et al., 2001]. Briefly, 10 or 20 nmol nitrate is quantitatively reduced to N2O by a strain of bacteria that lacks an active N2O reductase, and the product N2O is analyzed by continuous flow isotope ratio mass spectrometry. Referencing to the 15N/14N of N2 in air was through parallel measurement of the potassium nitrate reference material IAEA-N3, with a δ15N of +4.7‰ [Gonfiantini et al., 1995]. Replicate analyses were generally consistent with a reproducibility of 0.2‰ 1SD.
4.1. Physical Conditions of the Upper Water Column
 There is a large winter to summer change in the density structure of the study region (Figure 2). While the mixed layer is greater than 200 m in the Subantarctic and Polar Frontal Zones during the winter, summer mixed layer depth in these regions is less than 100 m, with a shallower mixed layer toward the North [Rintoul and Trull, 2001]. A climatology of seasonal mixed layer evolution [Kara et al., 2003] illustrates that the AU9701 cruise occurred during the time of year with the deepest mixed layer depth (although near its end), whereas AU9706 occurs near the end of the period of shallowest mixed layer (Figure 2c).
 The depth interval associated with Subantarctic Mode Water (SAMW) is clearly defined in the AU9706 (March 1998) profiles at a potential density of 26.8 (Figure 2b). The AU9701 (September 1997) stations poleward of 43°S have surface densities close to that of the summertime SAMW layer, consistent with development of SAMW from wintertime deep mixing in the Subantarctic [McCartney, 1977]. In detail, the existence of a slight density decrease toward the surface in the AU9701 profiles (Figure 2b) and the slightly lower density of the surface during AU9701 relative to other wintertime cruises (Figure 2a) suggest that AU9701 is just at the transition toward summertime conditions. There are also significant differences in surface water density between our two summertime samplings of subantarctic surface, on the order of ≤0.5 × 10−3 kg m−3 (Figure 2a).
4.2. Nitrate Isotopes
4.2.1. Hydrocast Profiles
 In the Subantarctic during both winter and summer, there is an upward decrease in [NO3−] and an associated increase in nitrate δ15N into the surface mixed layer (Figure 3). In the Subantarctic during winter, the upward decrease in [NO3−] and associated increase in δ15N of nitrate from the depth of Subantarctic Mode Water at 300–400 m into the surface mixed layer are modest and lack a clear depth of transition from thermocline to surface conditions (Figures 3a and 3b). This is particularly true for the more polar component of the SAZ (south of 45°S). The upward increase in δ15N of nitrate into the surface is much more clear and of greater amplitude in the summertime data than in the wintertime data (Figure 3d compared to Figure 3b), suggestive of a winter-to-summer increase in the ratio of nitrate assimilation relative to circulation-driven nitrate supply. As shown below, the summertime vertical gradient in the δ15N of nitrate is larger than would be expected from summertime vertical [NO3−] gradient (Figure 3c) if one were to assume nitrate supply from below and assimilation with an isotope effect of 5 to 10‰ (Figure 3d). This is suggested qualitatively by the winter-to-summer changes in the SAZ profiles: surface layer δ15N of nitrate increases remarkably from winter to summer (compare upper 100 m in Figures 3b and 3d) while surface layer [NO3−] decreases only slightly from winter to summer and actually increases at some latitudes (compare upper 100 m in Figures 3a and 3c).
 In contrast to the upward δ15N increase into the summertime SAZ surface, the δ15N of nitrate increases only slightly (<1‰) upward through the thermocline, despite a large (>12 μM) decrease in [NO3−] from ∼1000 m to the base of the mixed layer (compare Figures 3c and 3d). These thermocline observations are consistent with our previous understanding of the nitrate δ15N/[NO3−] relationship of Subantarctic Mode Water, in which mixing along isopycnals with the low-latitude thermocline leads to nitrate depletion without 15N enrichment of the nitrate [Sigman et al., 1999a, 2000]. Vertical mixing during the winter imposes the δ15N/[NO3−] relationship of the thermocline on the overlying surface layer, and nitrate assimilation modifies it only modestly. In contrast, the δ15N/[NO3−] relationship is greatly impacted by in situ nitrate assimilation in the summertime subantarctic surface. This is shown more directly in section 5.
4.2.2. Underway Collections of the Mixed Layer
 The underway data provide a more complete sampling of the wintertime surface layer, confirming that the wintertime subantarctic surface can only occasionally be distinguished from the δ15N, [NO3−] and δ15N/[NO3−] relationship of the subantarctic thermocline (Figure 4 and Figure 5a). In contrast to the comparison of summer depth profiles from AU9706 with the winter profiles of AU9701 (Figure 3), the underway samples do show a surface [NO3−] decrease into the summer, evident in the difference between the February 1999 samples and the rest of the underway collections, all of which occurred during the winter or spring (Figures 4a and 4b). Summertime surface [NO3−] deviates more noticeably from the winter trend when plotted versus salinity, an effect of the increased role of low-salinity, high-[NO3−] source water from the south during summer stratification (Figure 4b [Lourey and Trull, 2001]). As was described above for the hydrocast profiles, the surface nitrate δ15N in February 1999 is clearly different between seasons, as much as ∼9‰ higher in the summer than winter (Figures 4c and 4d).
 Consumption of nitrate by phytoplankton results in an increase in the δ15N of the ocean nitrate pool due to preferential uptake of the light isotope. If a pool of nitrate is consumed by algal assimilation with a constant isotope effect and without resupply during the consumption process, then the isotopic evolution of the remaining nitrate is given by the “substrate pool” equation in Rayleigh fractionation kinetics. Specifically, starting at some initial [NO3−] and δ15N, the nitrate pool will increase along an approximately straight line in the plotting space of δ15N versus ln([NO3−]), with the slope (or −1 times the slope) of that line approximating the isotope effect [Mariotti et al., 1981].
 Whether the Rayleigh model is appropriate for the interpretation of a given oceanic nitrate isotope data set depends on the specifics of the situation. For instance, the Rayleigh system can be appropriate when there are nonoverlapping events of nutrient supply (e.g., intense vertical mixing in the winter) and nutrient consumption (e.g., algal assimilation in a strongly stratified summer layer) or if an isolated parcel of surface water is being tracked during the consumption process. As will be described in detail below, the Rayleigh model is not adequate for the Subantarctic as a whole. Nevertheless, the “Rayleigh space” of nitrate δ15N versus ln([NO3−]) proves useful when interpreting data, including when this involves processes that clearly violate the Rayleigh model.
Figure 5 recasts our water column profile and underway data in Rayleigh space. In Figure 5a, we compare all winter versus summer data. In Figure 5b, focusing on the profile data, winter profiles are shown in solid symbols and summertime profiles in open symbols, with colors chosen to allow comparison of winter and summer data from a given latitude (grey indicates no matching winter station).
 Several key observations that have arisen in previous work are worth mentioning here. Rising from the deep Southern Ocean (Lower then Upper Circumpolar Deep Water, compressed at the lower right of Figures 5a–5c), nitrate δ15N increases as [NO3−] decreases upward through the thermocline (see water mass labels in Figure 5a). However, the magnitude of δ15N increase for a given [NO3−] decrease is small relative to what one observes for a given amount of nitrate consumption, assuming a reasonable value for the isotope effect of nitrate assimilation [Sigman et al., 1999a, 2000]. The cause of the thermocline trend (that is, the small magnitude of δ15N increase for a given amount of [NO3−] decrease upward through the thermocline) has been addressed previously [Sigman et al., 2000]. Regardless of its cause, it is an important background observation for what follows. In particular, for a given isotope effect for nitrate assimilation, surface waters that undergo nitrate consumption starting from SAMW composition will have a much lower δ15N for a given final [NO3−] than if they start from the composition of Antarctic or Polar Frontal Zone surface waters (see stations south of 51°S in Figure 5b).
 Our wintertime data indicate that, during this extensive portion of the year, the nitrate δ15N/[NO3−] relationship of the surface is similar to that of the thermocline (Figures 5a and 5b). While there is evidence for up to 2‰ 15N enrichment above the thermocline trend in the late winter (September and October) data, the data from July 1999 indicate that the surface and thermocline are completely indistinguishable during this deepest part of winter (blue solid circles of AU9901 in Figure 5a).
 There is an equatorward decrease in subantarctic surface [NO3−] during the winter without the level of enrichment in 15N expected from nitrate assimilation (Figure 4a). This lack of nitrate 15N enrichment indicates that the wintertime equatorward [NO3−] decrease in the SAZ surface is set solely by physical processes. The more equatorward subantarctic surface mixes vertically with underlying thermocline water that has lower density than the SAZ thermocline to the south. This lower-density thermocline water has more effective exchange with the subtropical surface waters and thus has a lower [NO3−]. In addition, the more equatorward SAZ is in closer contact with the nitrate-free surface waters of the subtropics, exchange with which will dilute SAZ nitrate without causing nitrate 15N enrichment.
 The nitrate isotope data indicate a clearly different situation during the summer, with nitrate assimilation greatly altering surface nitrate δ15N. As described above, the summertime increase in nitrate δ15N is large relative to the modest winter-to-summer decrease in surface [NO3−]. Plotting the data in Rayleigh space, two end-member options immediately arise to explain this observation. First, a high isotope effect may apply to nitrate assimilation in the Subantarctic (see trajectories for ɛ = 10‰ in Figure 5b). Second, nitrate in the summertime Subantarctic Zone may be fed from surface waters further south, such that a large amount of nitrate consumption is required to generate summertime SAZ [NO3−], in which case a lower isotope effect, similar to that estimated from the Antarctic, is appropriate.
 Comparing winter and summer data from the paired cruises of AU9701 (September 1997) and AU9706 (March 1998), it is not uncommon for the summer nitrate concentration to be similar to or even slightly higher than the wintertime nitrate concentration (Figure 5b). Yet, for all latitudes, summer δ15N of nitrate is higher than that for winter. In addition, for many of the summertime profiles, δ15N of nitrate increases markedly into the surface layer even though [NO3−] decreases very little or even increases (Figure 5b). As described previously [Sigman et al., 1999a], the cases where δ15N of nitrate increases into the surface even as [NO3−] is unchanged or increases represent strong evidence for equatorward transport of surface nitrate. Furthermore, temperature and salinity differences within the SAZ also indicate cross frontal supply of cold, fresh water from the south [Rintoul and Trull, 2001].
 However, the wintertime data reported here introduce a new constraint on the nitrate supply to the subantarctic surface, which requires alteration of the previous interpretation [Sigman et al., 1999a]. Specifically, the winter data require that the extreme 15N enrichment in summertime subantarctic surface waters is generated not over the timescale of years but rather within the much shorter time span of each annual warm season growth period (i.e., 4 months or less, Figure 2c). Over such a short time interval, salinity in the subantarctic surface is roughly conservative with regard to atmospheric fluxes [Kalnay et al., 1996] and has thus been taken as a measure of circulation [Lourey and Trull, 2001; Wang and Matear, 2001]. While there is robust evidence for equatorward supply of surface nitrate during the summer (see above), the equatorward increase in summertime subantarctic surface salinity precludes the possibility that all of the water (and thus all of the nitrate) in the summertime subantarctic surface is of direct Antarctic/Polar Frontal Zone origin (Figure 5c). Thus the data reported here indicate that neither a large isotope effect alone nor equatorward transport alone can explain the large seasonal change in δ15N of nitrate in the Subantarctic Zone south of Tasmania. Rather, both explanations must be involved.
5.1. Model of Subantarctic Zone Summertime Mixed Layer Evolution
 The data reported above are interpreted with a time-dependent geochemical box model that simulates the seasonal changes in salinity, [NO3−], and the N isotopes of nitrate and exported N in the Subantarctic Zone south of Australia (Figure 6 and Tables 2 and 3). The study region is modeled as one box from 42° to 50°S, with a depth of 73 m chosen according to the seasonal (December to February) averaged climatological mixed layer depth (138°–142°E and 42°–50°S) [Kara et al., 2003]. To the north and south, constant boundary conditions are imposed representing the summertime waters of the Subtropical and Antarctic Zones, respectively (abbreviated as STZ and AZ below; the Polar Frontal Zone (PFZ) is included here as part of the AZ).
The number of digits does not indicate significance; values shown are the exact values used in the model calculations.
summertime AZ end-member salinity
summertime AZ end-member nitrate concentration
isotopic composition of nitrate in the summertime AZ end-member
‰ vs. air
wintertime AZ end-member salinity
wintertime AZ end-member nitrate concentration
isotopic composition of nitrate in the wintertime AZ end-member
‰ vs. air
salinity in the STZ end-member
nitrate concentration in the summertime STZ end-member
salinity in the thermocline end-member
nitrate concentration in the thermocline end-member
isotopic composition of nitrate in the thermocline end-member
‰ vs. air
change in salinity in the SAZ from winter to summer
change in nitrate concentration in the SAZ from winter to summer
change in the isotopic composition of nitrate in the SAZ from winter to summer
‰ vs. air
climatological mixed layer depth averaged over 138°E to 142°E and 42°S to 50°S [Kara et al., 2003]
standard model season length
 Four parameters from the SAZ mixed layer are calculated in the model: salinity, S, nitrate concentration, [NO3−], the δ15N of the nitrate pool, and the δ15N of the accumulated export material, δ15Ntrap (Figure 6). Water fluxes are divided into two types: an Ekman transport term, Ek, and mixing components, ma and mst (the subscripts indicating mixing with the AZ and STZ, respectively). Since the AZ and STZ are treated as an infinite sources/sinks, we cannot distinguish between the advective and mixing-associated fluxes of AZ water into the SAZ; thus we are limited to using a sum of Ekman transport and southern boundary mixing, Ek + ma. Moreover, we have no record of the effect of Ek as it transfers SAZ water into the STZ (or into any other reservoir), so that mst is the only relevant term of water exchange with the STZ. This leaves four unknowns: the water fluxes (Ek + ma and mst), the isotope effect of nitrate assimilation (ɛ), and export production (P, assumed here to be equivalent to net nitrate consumption). The time rates of change of the model variables are:
where “circ” and “bio” are operators representing transport through ocean circulation and biogeochemical processes respectively. Epsilon, ɛ, is the isotope effect of nitrate assimilation. Cross frontal mixing and advection of a given tracer, X, is parameterized as:
The export production parameter, P, is the biological removal term, bio, integrated over the depth of the box, giving it units of mmol N m−2d−1:
The exported material collected to simulate the sinking flux is:
 Given the observational constraints (given below), the data cannot be fit by a steady state (i.e., time-independent) model. The summertime salinity of the SAZ is intermediate between the two inputs, requiring roughly equivalent exchange with the two potential sources. The STZ input dilutes the nitrate concentration in the SAZ so much that the export production required to fit the model is unrealistically low. To then fit the nitrate isotope data requires an extremely high isotope effect, such that the sinking flux δ15N is a large negative number, in this way indicating that P is unreasonably low. Thus the wintertime resupply from the thermocline is critically important to fuel the high level of production in the SAZ during the summertime transient state. In short, the observations dictate the use of a time-dependent model. This time-dependent, forward model is fit to match the observed winter-to-summer changes in salinity, [NO3−], and δ15N of NO3−.
 Data were sorted into 1° latitude bins, averaged (with [NO3−] weighting in the case of nitrate δ15N), then averaged to generate a single set of observational “mean” constraints for 42°–50°S latitude range (Table 2). While the definitions of Orsi et al.  indicate a Subantarctic Front latitude of ∼52°S in the SAZ study region, this front is somewhat further north (50°–51.5°) on the basis of slightly different criteria [Rintoul and Trull, 2001]. We chose to use 50°S as the polar limit of our SAZ box partially on the basis of these definitions and partially because [NO3−] weighting of the nitrate isotope data would have strongly biased our model toward conditions near the Polar Frontal Zone. It can be argued that this choice causes us to underestimate the flux of AZ/PFZ water into the SAZ, which often has its greatest signature near the Subantarctic Front (Figure 5c) [Rintoul and Trull, 2001; Sigman et al., 1999a]. Similarly, we chose a more northern extent for our simulated SAZ than suggested by Orsi et al.  to prevent Subtropical waters from being defined as having significant [NO3−]. The standard run length of the model is equivalent to 90 days, which is roughly appropriate for the evolution of the SAZ surface to midsummer conditions subsequent to shoaling of the mixed layer in October–November [Lourey and Trull, 2001; Wang and Matear, 2001] (Figure 2). In addition to the hydrological data, a model constraint is provided by the δ15N of sediment trap material, which was collected at three latitudes along 140°W for the time interval between the cruises AU9701 (trap deployment) and AU9706 (trap collection) [Lourey et al., 2003; Trull et al., 2001] (Figure 1a and Tables 1 and 4). For our comparison, we use the material collected during the summer period in “SAZ traps” at 47°S (center column in Table 4) [Lourey et al., 2003].
Table 4. Sediment Trap Data: δ15N of PN (‰ Versus Air) at 47°S (SAZ)a
 As described above, we assume that salinity is conservative over the time interval of the model experiment, that is, that ocean/atmosphere fluxes are not significant on this timescale [Kalnay et al., 1996; Lourey and Trull, 2001; Rintoul and England, 2002; Wang et al., 2001]. Given this assumption, the salinity constraint depends only on the circulation terms of the model. The input of water from the AZ, Ek + ma, decreases salinity (to the right in Figures 7a and 7e), while mixing with the STZ, mst, acts to increase it (upward in Figure 7a); thus the Ek + ma:mst relationship derived from the model is set by the observed summertime decrease in salinity (bold line in Figure 7a). [NO3−] is also affected by the circulation. Input of AZ/PFZ water through Ek + ma will increase the SAZ [NO3−] (to the right in Figures 7b and 7f), while mixing with low-[NO3−] STZ water (mst) will lower SAZ [NO3−] (upward in Figure 7b). However, [NO3−] is also affected by assimilation and export, P, which lowers SAZ [NO3−] (upward in Figure 7f). Because of these combined effects, for a given Ek + ma:mst relationship (set by the salinity data as shown in Figure 7a), a given [NO3−] can be reached at a combination of mst and P, with P at its maximum possible value when mst is zero. As described below, the N isotope constraints require as large as possible a value for P which forces mst to its minimum possible value, zero (bottom axis in Figures 7a–7d).
 In the context of constraints provided by [NO3−] and salinity observations (Figures 7a, 7b, 7e, and 7f), the δ15N of nitrate is used to set the model isotope effect, ɛ (Figures 7c and 7g). Finally, a test of a given set of parameters (Ek + ma, mst, P and ɛ) is provided by the δ15N of the sediment trap material (Figures 7d and 7h). For example, a model-derived δ15N for sinking N that is lower than the measured δ15N of the sediment trap material collected during the summer period at 47°S would suggest that the model-derived isotope effect is too high. This, in turn, would require that production was too low and that the input of Antarctic water has been underestimated. It is also through this process that the current model is forced to adopt a situation of no summertime mixing across the STF. Such mixing would dilute the nitrate concentration in the model without increasing the δ15N of nitrate. This would require lower export production and thus a higher model isotope effect to fit the summertime nitrate δ15N data, which would yield a lower sinking flux δ15N, diverging from the sediment trap measurements.
 There are four independent equations that are solved in the model (equations (1)–(3) and an equation combining (6) and (7) in terms of δ15N) and four observables that provide constraints (salinity, [NO3−], the δ15N of nitrate, and the δ15N of sinking N). Thus a unique solution should result. However, this solution falls in the parameter space of negative mst, which is physically impossible. Thus the solution we use (with mst = 0) must have some misfit with observations. We force the model misfit into the sinking flux δ15N, for two reasons. First, sinking flux δ15N is the least precisely constrained by measurements (see discussion of sediment trap data below). Second, and more important, focusing on the amplitude of model misfit of δ15Ntrap allows us to isolate the constraint on SAZ biogeochemistry that the N isotopes can provide. That is, we use the N isotopes to evaluate the solution that one gets without the N isotopes and to determine the likely directions in which this solution errs. In future work (and hopefully with more data), an inverse model could be used to yield a “best fit” solution that takes into account analytical and sampling uncertainties as well as model dynamics when distributing the misfit between the modeled and observed parameters (salinity, nitrate concentration, nitrate δ15N, and the δ15N of N export).
 There are a number of questions as to the reliability of sediment trap δ15N measurements as an indication of the δ15N of the N sinking out of the surface ocean. For example, while biochemical reactions typically favor the light isotope such that remineralization would cause sinking N δ15N to be constant or increase with depth, the moored traps at 47°S studied by Lourey et al.  indicate that sinking N δ15N decreases from 2.7‰ at 1060 m to 1.4‰ at 3850 m (Table 4). Indeed, there is an observed decrease in sediment trap δ15N with depth in many ocean regions, the cause of which is unknown [Altabet and Francois, 1994; Lourey et al., 2003; Thunell et al., 2004]. In this study, we take sediment trap δ15N to be only a rough (±1‰) indication of sinking N δ15N and use it as a “check” on our model-derived scenarios, using the other parameters as “perfect” constraints. Nevertheless, since we do interpret large (>2‰) deviations from the sediment trap measurements to indicate that a model solution is implausible, the coupling of isotope measurements of surface nitrate and sediment trap materials is central to our modeling exercise and the new constraint that arises from the N isotopes.
5.2. Model Results
 The data require a nitrate assimilation rate of ∼4.7 mmol N m−2 d−1, a rate for Ek + ma of 38 Sv, and an isotope effect for nitrate assimilation of 8.9‰ in the standard case model (bold lines Figure 7). The predicted sediment trap δ15N is +1‰ (bold dashed line in Figures 7d and 7h), which is close to but below the range of the sediment trap δ15N measurements at 47°S (∼1.4–2.7‰) [Lourey et al., 2003] (Table 4). Given the uncertainties about sediment trap δ15N mentioned above, this fit seems reasonably good. However, the generally low predicted sinking N δ15N suggests to us that the model overestimates the isotope effect of nitrate assimilation by ∼1‰.
 It is illustrative to consider what would occur if we were to force the model with a smaller isotope effect, so as to fit better fit the sediment trap δ15N data. With no other changes, this would lead to a decrease in the δ15N of the nitrate pool. To match the observed δ15N of nitrate would then require an increased ratio of N export to nitrate supply. If N export is then increased to match the observed summertime increase in the δ15N of nitrate (upward in Figure 7g), the summertime [NO3−] decrease will be overestimated (i.e., the summertime [NO3−] will be underestimated; upward in Figure 7f). To fix this nitrate concentration misfit would require an increase in the net influx of nitrate by increasing Ek + ma or decreasing mst. Given that mst is already at zero, the only option would be to increase Ek + ma (to right in Figures 7b and 7f). If this were done, the model SAZ would become too fresh over the course of the 90 day summer period (to right in Figures 7a and 7e). Therefore the observations could be made consistent with a lower value for ɛ (that is, more consistent with the sediment trap δ15N measurements) if one of the following were the case about our current estimates or assumptions for the winter-to-summer evolution of SAZ conditions: (1) the summertime [NO3−] drawdown was greater than observed; (2) the summertime salinity decrease was greater; or (3) there was enough evaporation during the summer to cause salinity to underestimate inputs of low-salinity water from the south.
5.3. Model Uncertainties
 A sensitivity test was performed in which the individual model constraints from our standard case (Table 3) were changed by ±25% (Table 5). While it is not possible to make conclusive statements in this first effort at simulating the N isotopes in the Subantarctic, we see this as a generous degree of error for most of the model parameters. In addition, two model simulations were performed to address uncertainties regarding summertime inputs to the SAZ mixed layer. In the first simulation, we assume wintertime AZ constraints rather than our standard case summertime AZ conditions; in the second, we allow for a continuous supply of 10 Sv of thermocline water into the SAZ mixed layer during the summer season (“AZinput” and “Thermo” in Table 5). The former simulation using wintertime AZ conditions yields a slightly worse fit to the data, yielding a greater misfit in δ15N of the sinking N (0.7 rather than 1‰ in the standard case). Its most significant effect on the solution is an increase in the estimated supply of AZ water to the SAZ mixed layer (40 rather than 38 Sv). The latter simulation including summertime thermocline supply requires slightly less AZ input (37 rather than 38 Sv). One of the greater sensitivities of our model-derived estimates for water flux and production (but not for isotope dynamics) is to the mixed layer that we assign for the summertime SAZ (Table 5). Our MLD of 73 m derives from the climatology of Kara et al. ; the climatology of Levitus and Boyer  suggests a ∼25% deeper mixed layer, which would have raised AZ input to 47 Sv and N export to 5.9 mmol N m−2 d−1 (Table 5). Finally, we also evaluated model uncertainty in a “Monte Carlo” analysis. Model constraint values were randomly generated with mean and standard deviations based on our available data, with a final summertime salinity of 34.53 ± 0.03‰, a [NO3−] of 8.43 ± 2 μM, a nitrate δ15N of 12.17 ± 1‰, a mixed layer depth of 73.1 ± 25 m and a season length of 90 ± 20 days. The model was solved with these constraints, and the solution set was parsed to include only those runs in which the δ15N of the trap material produced is within the range of 0–3‰. The estimates (Table 5) are consistent with the “standard case” solution presented above but also indicate the combined effect of the uncertainties.
Table 5. Sensitivity of Model Output to Error in Constraints From Data
Ek + ma, Sv
P, mmol N m−2 d−1
In this simulation, Antarctic end-member values were set to the wintertime values.
Model run included constant 10 Sv thermocline mixing over the entire summer simulation.
The Monte Carlo analysis of the model was performed using the following mean values and standard deviations for the modeled final (“summertime”) SAZ conditions: salinity = 34.53 ± 0.03‰, [NO3−] = 8.43 ± 2 μM, nitrate δ15N = 12.17 ± 1‰, mixed layer depth = 73.1 ± 25 m, and season length = 90 ± 20 days. Solutions with δ15Ntrap outside the range of 0 to 3‰ were discarded.
 All together, the model results lead to several conclusions, which are ranked here in terms of certainty. First of all, the isotope effect of nitrate assimilation is greater than 7‰ and likely closer to 8–9‰. Next, if one takes the sediment trap data as at least a rough (±1‰) recorder of the sinking flux δ15N, then the isotope effect cannot be greater than 11‰. If so, it becomes possible to draw inferences about the biogeochemical and physical rates in the SAZ. First, nitrate assimilation and export is 3.5 mmol N m−2 d−1 or greater (4.7 mmol N m−2 d−1 in the standard case, potentially as high as 6.3 mmol N m−2 d−1). Second and related, the net input rate of AZ/PFZ water into the SAZ south of Australia during the summer is 28 Sv or greater (38 Sv in the standard case, potentially as high as 51 Sv).
 To understand the implications of the water transport terms for the isotope dynamics of nitrate in the Subantarctic Zone, we compare the isotope effect derived from the model's standard case (8.9‰) with the isotope effect derived from a “naïve” use of the Rayleigh model, 11.4‰ (yielding a sinking flux δ15N of −1.5‰), or using a steady state model supplied from the subantarctic thermocline, 9.5‰ (yielding a sinking flux δ15N of 2.7‰), or from the Antarctic surface, 6.6‰ (yielding a sinking flux δ15N of 5.6‰). Thus not just for specific profiles but for the entire SAZ, the effect of summertime water transports on the N isotope dynamics is significant. The Rayleigh and Antarctic supply steady state models can only be fit simultaneously to the [NO3−] and nitrate δ15N data; they cannot simultaneously fit salinity or the δ15N of sinking N (Tables 3 and 4). The steady state model with thermocline nitrate supply can roughly fit observations for summertime [NO3−] and the δ15N of nitrate and of sinking N. However, it grossly underestimates N export, with 50 Sv of summertime SAZ thermocline/surface exchange supply yielding less than 1 mmol N m−2 d−1 (compare with Table 6); this quantity of exchange is physically unreasonable, given the stratification that is observed during the summer [Rintoul and Trull, 2001; Wang and Matear, 2001] and the fact that the region is a zone of net downwelling. As described above, thermocline input does not affect our estimate of the isotope effect of nitrate assimilation and has little to no effect on our AZ transport (Table 5).
Table 6. Previous Estimates of Export Production in the SAZ
 An isotope effect of ∼8–9‰ for subantarctic nitrate assimilation is higher than was estimated by Sigman et al. [1999a] and is the direct result of the wintertime measurements showing an extremely rapid surface δ15N of nitrate increase into the summertime. While this isotope effect is consistent with other recent measurements from the permanently open Antarctic near the PFZ [Altabet and Francois, 2001; Karsh et al., 2003; Lourey et al., 2003], it appears to be 2–3‰ greater than that measured in other polar regions, including the more polar Antarctic near the marginal ice zone [Sigman et al., 1999a; P. DiFiore et al., “Nitrate assimilation in the Antarctic marginal ice zone”, manuscript in preparation, 2006] and the subarctic North Pacific [Lehmann et al., 2005; Wu et al., 1997].
 Recent culture studies with the diatom Thalassiosira weissflogii indicate that algal nitrate assimilation under light limitation expresses a greater isotope effect than assimilation under conditions of iron limitation or maximal growth rate [Needoba and Harrison, 2004]. This appears to result from greater nitrate efflux back out of cell, allowing the relatively large isotope effect of the intracellular nitrate reductase (15–30‰) [Ledgard et al., 1985; Schmidtt and Medina, 1991] to be more fully expressed in the environment [Needoba et al., 2004]. This may indicate that algae under light limitation bring more nitrate into the cytoplasm than can be reduced, in anticipation of the possibility that light conditions will soon change (e.g., because of vertical mixing or the diurnal cycle).
 This finding of higher isotope effects under light limitation may explain the high isotope effect that we estimate for the SAZ. Sediment trap data and SeaWiFs observations of chlorophyll indicate that there is a large amount of early season production in the SAZ [Trull et al., 2001]. In the SAZ during the spring and summer, mixed layer depths are greater than in other regions of the Southern Ocean, especially the region of the Antarctic with seasonal ice cover (Figure 8) [Kara et al., 2003]. The permanently open Antarctic appears to be more similar to the SAZ in that it has relatively deep mixed layers (Figure 8), and a high isotope effect may also apply in this region [Altabet and Francois, 2001; Karsh et al., 2003; Lourey et al., 2003]. Thus there may be a coherent trend arising of high isotope effect estimates from the regions with deep spring/summer mixed layers and the potential for light limitation. However, even if the isotope data are an indication of light limitation during the spring-to-summer period of high algal growth in the SAZ, iron limitation may still be the most important ultimate limiter of growth, causing the eventual end of the summer period of high growth as iron becomes scarce. Silicate and iron colimitation may also be relevant if nondiatom algae prove to be ineffective consumers of nitrate [Boyd et al., 1999; Hutchins et al., 2001; Sedwick et al., 1999, 1997].
6.2. Cross-Frontal Transport of Nitrate
 Ekman transport has been estimated at 11 to 20 Sv from the mean buoyancy flux [Warren et al., 1996] and the curl of wind stress at 55°S [Karsten and Marshall, 2002; Tomczak and Godfrey, 2003]. Assuming these global estimates scale to our region, our estimate of 38 Sv or greater for Ek + ma would imply 18–27 Sv of mixing between the summertime SAZ surface south of Australia and the AZ/PFZ to the South. Using TOPEX sea surface height data in combination with an estimate of ∼20 Sv of Ekman transport from wind stress, Karsten and Marshall  estimate eddy induced transport of ∼30 Sv across the SAF, which is of similar magnitude to our estimates. Cross-frontal mixing represents an interesting term, in that it causes a large equatorward transport of nitrate without causing a net flow of water.
 Eddy mixing of Subtropical waters into the Subantarctic across the Subtropical Front does occur at some rate [McNeil et al., 2001; Speer et al., 2000], so our exclusion of this process (or rather, the exclusion of this process by the model's best fit) deserves some consideration. Our best guess is that Subtropical water does not mix very far south across the SAZ during the summer. If so, it has little effect on the high-[NO3−] waters of the more poleward SAZ, and thus is not evident in our calculation of mean SAZ conditions as a data constraint for the model.
 A simple exercise can be performed to estimate the relative contributions of each SAZ nutrient source to export production and to the standing nitrate pool throughout the summer season. Our numerical model of the SAZ was modified to have two tagged sources of nitrate: (1) the SAZ winter mixed layer, and (2) Antarctic/Polar Frontal Zone nitrate transported into the SAZ during the summer. We consider the accumulation and fate of these two tagged nitrate pools in the context of the standard model best fit scenario described above (Figure 9). The nitrate supply is initially dominated by the winter mixed layer (subantarctic thermocline) nitrate; over time, the relative contribution of surface Antarctic nitrate supply increases (Figure 9b). The Antarctic source represents 13% of the total spring/summer production (Figure 9a). At the end of the season, the Antarctic source water is 27% of the remaining nitrate pool. As described above, the N isotope constraints suggest that these percentages should be taken as lower bounds. The ultimate effect of this Antarctic source input of nitrate is to require roughly half the decrease in summertime nitrate that one would observe if there were no AZ/PFZ nitrate input (Figure 9b). This contribution is important from the standpoint of SAMW formation, which occurs in the SAZ during the winter and represents the dominant source of nutrients to the lower-latitude surface ocean [Sarmiento et al., 2004].
6.3. Export Production
 Our estimate for nitrate assimilation rate from our standard model fit (4.7 mmol N m−2 d−1) is similar to the estimate of ∼5.5 from Lourey and Trull , which is not surprising given that both studies attempt to correct the summertime nitrate drawdown for AZ/PFZ nitrate input using salinity (Table 6). In essence, the N isotope data indicate that the salinity normalization approach does not grossly overestimate N export. Again, we take our estimate of N export as a lower bound because our standard model fit predicts a δ15N for sinking N that may be ∼1‰ lower than measured in the sediment traps. This suggests to us that the isotope effect determined in the standard model fit may be up to 1‰ too high; with a lower isotope effect, greater nitrate consumption would be required to fit the nitrate isotope data.
 In our treatment above, we have assumed that, in the context of the full summer period, nitrate assimilation is equivalent to N export. Some decoupling may occur, leading to error in our estimate of N export, at least in the form of sinking N. Suspended particulate and ammonium are rarely high enough to suggest a significant lateral transport or downward export of these N forms; the role of DON in this regard remains unclear. The existing data on the δ15N of DON [Knapp, 2006; Knapp et al., 2005], while all from the subtropical ocean, do not indicate large differences between it and the δ15N of sinking N. If this is the case, regardless of whether assimilated N is lost from the SAZ mixed layer by transport as DON or as sinking N, the rate of N loss (“N export”) should not be biased greatly.
7. Summary and Perspective
Sigman et al. [1999a] noted that the δ15N of nitrate of the summertime subantarctic surface is >3‰ higher than in the underlying thermocline, even though nitrate concentrations are rarely 30% lower than in the thermocline. This observation in itself could be explained in two ways: (1) equatorward nitrate transport across the SAZ, including nitrate input from the Antarctic and Polar Frontal Zones to the south, and/or (2) a large isotope effect for nitrate assimilation in the SAZ. Sigman et al. [1999a] noted 15N enrichment in SAZ surface nitrate even where [NO3−] was the same or higher than in the subsurface, pointing toward the former explanation.
 The summertime nitrate isotope data reported here show the same indications of equatorward nitrate transport as noted previously. However, the wintertime data, which are the first of their kind, indicate that the 15N enrichment of SAZ nitrate observed during the summer is generated each summer starting from near-thermocline values. Over this short time period, it is impossible for enough AZ/PFZ nitrate to be transported into the SAZ surface layer to replace the wintertime surface nitrate pool across the entire latitude range of the SAZ, a point that is quantified here with salinity, which should provide an indication of the amount of water coming into the SAZ from the south. Thus, to explain the amplitude of the winter-to-summer increase in the δ15N of nitrate requires a higher isotope effect than estimated by Sigman et al. [1999a]. At the same time, the comparison of the change in the δ15N of nitrate with the δ15N of sediment trap material for this region of the SAZ constrains how large that isotope effect can be, in this way maintaining the need for AZ/PFZ-sourced nitrate in the summertime SAZ.
 The data and our numerical simulation of them yield several important conclusions regarding the SAZ south of Australia. First, following the shoaling of the SAZ mixed layer in spring, the supply of Antarctic nutrients across the PFZ accounts for 15% or more of nitrate assimilation and N export during the spring and summer blooms and 30% or more of the nitrate resident in the mixed layer by midsummer. Second, export production must be adequately high to produce the observed summertime nitrate drawdown in the face of a significant nitrate input from the AZ/PFZ through the warm season, which leads to higher N export than if this input were ignored (Table 6). Third, the isotope balance in our model suggests that the previous estimates of the isotope effect in the SAZ are too low, and we estimate an epsilon of ∼8–9‰.
 Without the N isotope constraints, our model simulation of the seasonal changes in salinity and nitrate concentration is in essence similar to the study of Lourey and Trull , who estimated summertime nitrate assimilation from the winter-to-spring decrease in surface [NO3−], using the summertime decrease in SAZ surface salinity to correct for the amount of nitrate imported from the AZ/PFZ. Because our model is fit to match these two parameters perfectly, our estimate of summertime SAZ nitrate assimilation is similar to theirs (Table 6).
 The unique aspect of this study is our use of the N isotope data to test this approach. While the δ15N of nitrate data can be fit simply by adjusting the assumed isotope effect for nitrate assimilation, comparison of these data with a sediment trap–based estimate of sinking N δ15N puts a constraint on the isotope effect, precluding arbitrarily high values for it. We did not carry out a model simulation that addressed this overconstrained system by trying to simultaneously fit all parameters. Instead, it proved most transparent to simply allow the model to fit the salinity, [NO3−], and δ15N of nitrate data perfectly and then compare the predicted sinking flux δ15N with the sediment trap observations, determining whether the fit to the sediment trap δ15N data was good and in which direction any misfit to the data tended to be.
 While the model simulation predicted a sinking N δ15N that was arguably within ∼1‰ of the sediment trap constraint, the sense of misfit was clearly for the model to underestimate sinking N δ15N. Far from being a detail or uncertainty, the observed sense of model misfit is a key warning that AZ/PFZ nitrate transport into the SAZ and nitrate assimilation within the SAZ may be underestimated by the approach of using seasonal nitrate concentration data and assuming that salinity is a conservative tracer of AZ water inputs to the SAZ. Two end-member alternative interpretations of the misfit are that: (1) salinity is not acting in a completely conservative manner, and there is somewhat more AZ/PFZ nitrate entering the summertime SAZ than we and Lourey and Trull  allow, or (2) the sediment trap δ15N is 1–2‰ higher than the true N export out of the surface ocean, so that the true isotope effect for nitrate assimilation can be somewhat greater than we accept on the basis of our data/model comparison for sinking flux δ15N.
 Coupled with recent progress in laboratory studies of the nitrate assimilation isotope effect, this study raises this isotope effect as a possible environmental tool to study the physiological state of phytoplankton in modern aquatic environments. That our estimate for the isotope effect in the SAZ is 2–3‰ higher than our estimates from the marginal ice zone of Antarctica [Sigman et al., 1999a; P. DiFiore et al., manuscript in preparation, 2006] suggests some difference in the constraints on algal growth between these two Southern Ocean environments. Given the available culture data and the characteristics of the SAZ, a role for light limitation in the SAZ is a natural hypothesis for our observations. Along with continued use of algal cultures to explore the mechanism of nitrate isotope discrimination, studies in the Southern Ocean that couple the isotope effect estimates extracted from regional tracer distributions (i.e., the approach taken here) with shipboard-board growth experiments should help to test this explanation and, more generally, illuminate the environmental controls on (and the physiological significance of) the isotope effect of nitrate assimilation.
 A fundamental goal of this study was to develop an understanding of the constraints on subantarctic circulation and biogeochemistry that can be extracted from the coupling of N isotope measurements with standard oceanographic measurements. Beyond the proof of concept that we believe was achieved, there are key lessons for the future. In particular, if the N isotope budgetary approach is to be pursued further, the uncertainties on the δ15N of exported N should be reduced. This includes not only improving our ability to measure the particulates sinking out of the surface ocean but also accounting for the N exported laterally, such as dissolved organic N.
 Sample collection was supported by the Australian Commonwealth Cooperative Research Centre Program and Australian Antarctic Science Award 1156 (T.W.T.). This work was funded by U.S. NSF grant OCE–0081686 (to D.M.S.) and by BP and Ford Motor Company through the Princeton Carbon Mitigation Initiative. We thank Michael Bender and Matt Reuer for discussions. This manuscript benefited from reviews by Niki Gruber and an anonymous reviewer.