Global Biogeochemical Cycles

Fixed nitrogen loss from the eastern tropical North Pacific and Arabian Sea oxygen deficient zones determined from measurements of N2:Ar

Authors

  • Bonnie X. Chang,

    Corresponding author
    1. School of Oceanography, University of Washington, Seattle, Washington, USA
    2. Now at Department of Geosciences, Princeton University, Princeton, New Jersey, USA
      Corresponding author: B. X. Chang, Department of Geosciences, Princeton University, Guyot Hall, Washington Road, Princeton, NJ 08544, USA. (bonniec@princeton.edu)
    Search for more papers by this author
  • Allan H. Devol,

    1. School of Oceanography, University of Washington, Seattle, Washington, USA
    Search for more papers by this author
  • Steven R. Emerson

    1. School of Oceanography, University of Washington, Seattle, Washington, USA
    Search for more papers by this author

Corresponding author: B. X. Chang, Department of Geosciences, Princeton University, Guyot Hall, Washington Road, Princeton, NJ 08544, USA. (bonniec@princeton.edu)

Abstract

[1] Previous work estimating the N2excess above background due to denitrification has suggested that nitrate deficit-type methods may be an underestimate of fixed nitrogen (N) loss in the major oxygen deficient zones of the ocean. The N2excess approach has the advantage over nitrate deficit-type methods in that it does not depend on stoichiometric assumptions of fixed N to phosphate or oxygen utilization and avoids any uncertainly regarding the pathway of N loss. Measurements of N2:Ar from two stations within the eastern tropical North Pacific and from one station within the Arabian Sea oxygen deficient zones were used to determine the N2 excess due to denitrification. In both of the regions, the N2 excess was comparable in shape and magnitude to the concurrent fixed nitrogen deficit. In the eastern tropical North Pacific oxygen deficient zone, the N2 excess was near zero at the surface and rose to maxima of 13.7 ± 1.8 and 10.8 ± 1.9 μM N, compared to maximum fixed N deficits of 13.5 ± 1.5 and 12.3 ± 1.5 μM N, respectively. In the Arabian Sea oxygen deficient zone, the maximum N2 excess was 11.1 ± 1.5 μM N, compared to a maximum deficit of 12.5 ± 1.0 μM N. These results suggests that previous estimates of fixed N loss based on fixed N deficit calculations in these regions are likely reasonable, given the same considerations of volume and residence time of the water of the oxygen deficient zone.

1. Introduction

[2] Nitrogen (N) is a vital and often limiting nutrient to biological productivity in the ocean [Codispoti, 1989], yet fundamental questions regarding the oceanic N budget remain unresolved. For example, it is not known if the oceanic N budget is in balance [Middelburg et al., 1996; Gruber and Sarmiento, 1997; Brandes and Devol, 2002; Codispoti et al., 2001; Gruber, 2004; Codispoti, 2007]. Much of the uncertainty in the marine N budget hinges on uncertainty in the rates of fixed N loss from the ocean.

[3] The main process leading to the loss of oceanic fixed N is denitrification (defined as the conversion of fixed N to the gaseous end products: N2O and N2), which occurs in oxygen deficient environments in both the water column and sediments. The vast majority of oxygen deficient waters, and thus water column denitrification, are located in three major oxygen deficient zones (ODZs) of the ocean: the eastern tropical North and South Pacific (ETNP and ETSP, respectively), and the Arabian Sea (AS) [Codispoti and Christensen, 1985; Codispoti, 1989].

[4] The major ODZs form due to a combination of several factors. Wind-driven nutrient upwelling along the west coast of Central America, Peru, and Oman supports substantial productivity in these regions. This organic matter eventually sinks and fuels high respiration rates in the ETNP, ETSP, and AS ODZs. Additionally, these areas have a low O2 supply. The source water of the AS ODZ flows from the Indian Ocean thermocline and is fairly depleted in O2 by the time it reaches the AS from its outcrop at 40–45°S [Olson et al., 1993; Wyrtki, 1973]. The ETNP and ETSP ODZs are located in the “shadow zones” of the subtropical gyres, and also have distant outcrops [Fiedler and Talley, 2006].

[5] Measurements of the dissolved inorganic nitrogen (DIN) deficit, combined with estimates of the residence time and volume of denitrifying waters, has been the classic method of evaluating fixed N loss in the major ODZs of the ocean [Naqvi and Sen Gupta, 1985; Bange et al., 2000; Codispoti et al., 2001]. Based on this approach, the denitrification rate in the ETNP and ETSP ODZs are estimated to be roughly 25 Tg N/yr in each region [Codispoti and Richards, 1976; Codispoti and Packard, 1980; Deutsch et al., 2001] and approximately 30 Tg N/yr [Naqvi and Shailaja, 1993; Bange et al., 2000] in the Arabian Sea, giving a total water column denitrification rate of about 80 Tg N/yr for the three major ODZs combined. However, the accuracy of the DIN deficit to estimate the amount of fixed N lost depends upon the constancy of the elemental remineralization ratios of organic matter respired such that the amount DIN in a water parcel before the onset of denitrification can be predicted from either potential temperature (θ) or [PO43−]. Using measurements of N2 to determine N removal has the advantage of estimating the fixed N lost in the ODZs with virtually no assumptions of the composition of respired organic matter or pathways of N loss.

[6] Studies in two of the major ODZs of the ocean have found that the N2 excess due to denitrification varies from comparable to the DIN deficit, as in the ETSP ODZ [Chang et al., 2010], to nearly twice the DIN deficit, as was found previously in the AS ODZ [Codispoti et al., 2001; Devol et al., 2006]. These results indicate that the DIN deficit may not be an accurate estimate of the fixed N lost in the ODZs. Devol et al. [2006] hypothesized several reasons for the discrepancy between the DIN deficit and the N2excess in the AS ODZ: remineralization of N-rich organic matter such as organic matter produced by N-fixing organisms or the preferential degradation of amino acids, unconsidered reactions of fixed N with metals, and the contribution of sedimentary processes, all of which may produce non-Redfieldian distributions of N:P in the ODZ and thus violate the assumptions of the DIN deficit calculation. A different result was found in the ETSP ODZ [Chang et al., 2010], where the N2 excess roughly matched the DIN deficit implying that the N:P remineralization ratio derived outside of the ODZ was applicable to organic matter degradation within the ODZ. In this paper, we present measurements of N2:Ar ratios and N2 excesses in the ODZs of the ETNP and AS and compare those to the concurrent DIN deficits in order to help constrain estimates of global water column denitrification.

2. Materials and Methods

2.1. Study Sites and Sample Collection

[7] In the ETNP ODZ, nutrient and dissolved gas samples were taken at two stations in July 2008 aboard the R/V New Horizon (Table 1 and Figure 1). Dissolved gas samples representing waters not influenced by the ODZ were taken during an April 2005 cruise to the tropical Pacific aboard the R/V Ka'imimoana and an October/November 2007 WOCE P14 cruise aboard the R/V Mirai.

Table 1. Station Numbers, Locations, and Type
RegionStationLatitude (°N)Longitude (°W)Comments
ETNP ODZ4820.5106.5nutrient and gas
4921.5109.5profiles
Tropical Pacific Ocean0110gas profile
8110gas profile
18.5179gas profile
19179gas profile
Subtropical Pacific Ocean27179gas profile
27.5179gas profile
35.5179gas profile
36179gas profile
43.5179gas profile
44179gas profile
AS ODZ119.4−66.7nutrient and gas profiles
Tropical Indian Ocean8−63.6gas profile
−1.6−55gas profile
−10.2−52.5gas profile
Subtropical Indian Ocean−29−55gas profile
Figure 1.

Map of study sites. Circles denote stations within the ODZ; triangles are stations outside of the ODZ. Gas samples used for each ODZ are indicated by black (ETNP) and gray (AS) symbols. Dotted lines indicate regions from which WOCE nutrient data proximate to (N2:Ar) profiles outside the ODZs were taken. Contours show [O2] at σθ = 26.4 (World Ocean Atlas 2009 annual average).

[8] Samples for nutrients and dissolved gases in the AS ODZ were taken at one station in September 2007 aboard the R/V Roger Revelle (Table 1 and Figure 1). Dissolved gas samples outside of the ODZ were taken from a July/August 1995 WOCE I07N cruise aboard the R/V Knorr [Devol et al., 2006].

[9] Additional nutrient data from six 1995 JGOFS cruises to the Arabian Sea was used in calculations and comparisons (http://www1.whoi.edu/). The electronic WOCE atlas supplied further nutrient data from the Pacific and Indian Oceans (http://www.ewoce.org/) [Schlitzer, 2000]. These data were visualized and statistically analyzed using Ocean Data View v. 4.5.0 (R. Schlitzer, Ocean Data View, 2011, http://odv.awi.de/).

2.2. Nutrients and Hydrographic Data

[10] Oxygen, temperature, and salinity data in each of the ODZs were collected using Seabird CTDs equipped with dissolved oxygen sensors that were calibrated by discrete samples taken for either Winkler titrations or measurement on an isotope ratio mass spectrometer (method outlined in Section 2.3). In the ETNP, samples taken for nitrate (NO3), nitrite (NO2), ammonium (NH4+), and phosphate (PO43−) were frozen immediately after collection and subsequently analyzed on an Alpkem Flow Solution Autoanalyzer continuous flow system following the methods of Strickland and Parsons [1972]. Detection limits of NO3, NO2, NH4+, and PO43− were 0.1, 0.03, 0.3, and 0.02 μM, respectively. Nutrient samples taken in the AS were filtered through 0.45 μm Sterivex filters and frozen until analysis. Nitrate (NO3), nitrite (NO2), ammonium (NH4+), and phosphate (PO43−) were measured on a Technicon II Autoanalyzer according to the methods of Strickland and Parsons [1972]. Detection limits of NO3, NO2, NH4+, and PO43− were 0.08, 0.01, 0.07, and 0.03 μM, respectively.

2.3. Dissolved Gases

[11] Dissolved gas ratios were determined using the method described by Emerson et al. [1991] and Devol et al. [2006]. Briefly, approximately 150 mL of water were drawn directly from a Niskin bottle, taking care not to expose the sample to the atmosphere, into a 300 mL HgCl2-poisoned, pre-evacuated gas-tight glass flask of known volume. The flasks were weighed and the dissolved gases were equilibrated with the headspace of the flask at a constant temperature. The gases were then extracted into a stainless steel finger immersed in liquid He during which CO2 and water were removed cryogenically. Gas ratios N2:Ar and O2:Ar (when present) were measured using a Finnigan Delta XL. For all samples corrections were applied following Emerson et al. [1991] to account for the effect of O2 on the ionization efficiency of the other gases. Precision of duplicate N2:Ar was ±0.7‰.

3. Results

3.1. Hydrographic Conditions

[12] At both stations in the ETNP, the mixed layer depth (defined as a change in σθ > 0.125) was 10 m. At St. 48, the ODZ (defined as <5 μM) extended from approximately 130 to 770 m (σθ = 25.9–27.2) (Figure 2). At St. 49, the ODZ encompassed 200 to 770 m (σθ = 26.4–27.2). In the AS, the mixed layer depth was 40 m and the ODZ extended from 100 to 950 m (σθ = 25.3–27.4) (Figure 2). For both ODZs, potential density (σθ) is presented with [O2] in Figure 2.

Figure 2.

Depth profiles of oxygen concentrations ([O2], thick lines) with potential density (σθ, fine lines), and nutrients in the (top) ETNP (solid and dotted lines denote Sta. 48 and 49, respectively) and (bottom) AS ODZs.

3.2. Nutrients and Dissolved Inorganic Nitrogen Deficit

[13] Depth distributions of nitrate (NO3), nitrite (NO2), ammonium (NH4+), and phosphate (PO43−) in the ETNP and AS are presented in Figure 2. Dissolved inorganic nitrogen (DIN) is defined as the sum of the major dissolved inorganic nitrogen species ([DIN] = [NO3] + [NO2] + [NH4+]). In order to determine the quantity of DIN removed from the ODZs via denitrification (i.e., heterotrophic denitrification, anaerobic oxidation of ammonium [anammox], and/or reaction with metals), the amount of DIN expected (Nexp) in a water parcel before the onset of any denitrifying process must first be estimated. Though active N removal is known only to occur within the ODZ proper, the geochemical signal of denitrification is mixed outside of the ODZ such that upstream nutrient concentrations can contain evidence of denitrification. To account for this, a local Nexp specific to the source waters of each ODZ was estimated, rather than using an average global DIN:PO43−. In the past, two methods have been employed to determine Nexp: the conservative water mass tracer “NO” and its relationship to θ [Naqvi and Sen Gupta, 1985; Bange et al., 2000], and [PO43−] and its relationship to [DIN] outside of the ODZ [Codispoti et al., 2001; Devol et al., 2006].

[14] In this study, the relationship between Nexp and observed [PO43−] ([PO43−]obs) in each ODZ was derived using a linear regression of [PO43−] and [NO3] from WOCE outside and roughly upstream of the ODZs (Table 2) such that

display math

where m is the slope of the line and bis the y-intercept (Figure 3). Though WOCE does not report any measurements of NH4+ and relatively few of NO2, in the vast majority of the oxic ocean concentrations of NH4+ and NO2 are trivial and, thus, it is reasonable to assume all DIN outside of the ODZs is in the form of NO3.

Table 2. Density Intervals Used to Determine Nexp = m × [PO43−]obs + b
RegionDomain of DataaσθmbR2Number of ObservationsStandard Error (μM)
  • a

    “Gas” only considered WOCE stations proximate to the location of N2:Ar profiles outside of the ODZ. “All” used all WOCE stations in the domain selected (see Section 3.2 for details).

ETNPgassurface–27.814.4−1.10.993181.6
allsurface–26.414.3−2.20.9739501.5
26.4–27.213.01.80.9531561.5
27.2–27.810.510.80.8871840.9
ASgassurface–27.814.5−1.60.9932551.0
allsurface–25.314.4−2.50.9712770.9
25.3–27.014.6−1.60.9924571.0
27.0–27.410.57.70.9412260.8
27.4–27.89.311.60.9332840.4
Figure 3.

Dissolved inorganic nitrogen ([DIN]) versus phosphate ([PO43−]) in the (left) ETNP and (right) AS ODZs. DIN expected in the absence of denitrification (Nexp) determined from ‘gas’ and ‘all’ nutrient data represented by thick dashed and solid lines, respectively. Fine dashed lines indicate 1 SE of ‘gas’ regression, gray envelope shows 1 SE of ‘all’.

[15] The ODZs are poorly ventilated relative to other water masses at that depth and though source waters are not necessarily obvious or easily identifiable, in this study we assume the main water masses feeding the ETNP ODZ are the California Current from the north and the Northern Subsurface Countercurrent (NSCC) from the south. The NSCC is essentially modified 13°C Thermostad water (13DT). The 13DT flows north to the equator from its origin in the southwest Pacific, traverses the basin along the equator, mixing with waters from the north, and is transported north of the equator as the NSCC [Fiedler and Talley, 2006; Stramma et al., 2010]. The ETNP ODZ is underlain by the deeper, lower salinity North Pacific Intermediate Water [Fiedler and Talley, 2006]. Given these source waters, the domain of data used for this analysis ranged from 45°N to the equator and 180 to 115°W (further east would include waters influenced by the ETNP ODZ).

[16] The AS ODZ is fed to a small extent by the Red Sea and the Persian Gulf but the vast majority of the inflow originates in the south from the South Indian Ocean thermocline [Olson et al., 1993] of which Indian Central Water (ICW) comprises the bulk [You, 1997]. ICW is formed by convective overturning between 40 and 45°S [Wyrtki, 1973] and is transported north to the Arabian Sea via the western boundary current, the Somali Current [You, 1997]. To define the source waters of the AS ODZ, the domain of data used extended from 45°S to 11°N (any further north would include ODZ waters) and 50 to 77°E (Madagascar to the southern tip of India).

[17] Two approaches were used to derive the relationship of [DIN] to [PO43−] in the source waters of the ODZs. One method utilized the WOCE nutrient data at the same locations from which N2:Ar profiles representing waters unaltered by N removal processes were obtained (Table 2 “gas,” Figure 1, and further discussed in section 3.3). This ensured that any biases in the source waters, such as the presence of a denitrification signal (in either nutrients or N2:Ar), would be equivalently accounted for in both analyses. However, since availability of high precision N2:Ar data outside of the ODZs is extremely limited given the difficulties associated with making these measurements, the corresponding nutrient data is a relatively small subset of available data in the domain of the source waters.

[18] Another approach examined large-scale spatial gradients of nutrient concentrations along potential density surfaces in the applicable domain. The results of linear regressions of [DIN] to [PO43−] along several isopycnal intervals in the Pacific and Indian Ocean are presented in Table 2 and Figure 3. The decreasing slopes (m) with increasing density reflect the net accumulation of a denitrification signal (water column and sedimentary) in an aging water mass, which lowers the ratio of [DIN] to [PO43−], thereby decreasing the slope (m) and increasing the y – intercept (b). Nexp determined using these two different approaches are not appreciably different suggesting the relationship of [DIN] to [PO43−] is robust across the domains of data considered in each basin (Table 2 and Figure 3). Given this agreement, only Nexp calculated using regressions of [DIN] to [PO43−] along isopycnal intervals is considered for further analyses.

[19] The amount of DIN lost in each of the ODZs, or the DIN deficit (Ndef), can be estimated as the difference in the DIN expected in the absence of any nitrogen removal processes (Nexp) and the amount actually observed (Nobs):

display math

In the ETNP ODZ, Ndef was approximately 0.5 and 4.5 ± 1.5 μM at the surface of St. 48 and 49, respectively (Figure 4). It rose to a maximum of 13.5 ± 1.5 μM at 300 m at St. 48 and 12.3 ± 1.5 μM at 250 m at St. 49 (σθ = 26.5). Ndef decreased to near zero below 1500 m at both stations. At St. 1 in the AS ODZ, Ndef was −0.3 ± 0.9 μM at the surface, increasing to a maximum of 12.5 ± 1.0 μM at 230 m (σθ = 26.5), before decreasing to near zero by 1200 m (Figure 4).

Figure 4.

DIN deficit (Ndef, filled symbols) and N2 excess (Nxs, open symbols) relative to depth in the (left) ETNP ODZ and (right) AS ODZ. Error bars indicate 1 SD.

3.3. N2:Ar Ratios

[20] Biological processes like denitrification are not solely responsible for changes in the saturation state of dissolved gases; physical processes such as heating, cooling, and mixing may also cause super- and under-saturations in a given water parcel. Normalizing N2 to the biologically inert gas, Ar, accounts for some saturation changes that arise from physical processes. N2:Ar of each sample was further normalized to the atmospheric equilibrium N2:Ar at the θ and S [Hamme and Emerson, 2004] of a given sample [(N2:Ar)θ,S] to highlight departures from equilibrium [(N2:Ar)norm]:

display math

The waters outside of both ODZs exhibited similar trends in (N2:Ar)norm. The surface waters were approximately at atmospheric equilibrium (i.e., (N2:Ar)norm = ∼1.0) and then (N2:Ar)norm increased with depth (Figure 5) due to incomplete equilibration and bubble injection in regions where the deep waters were formed [Hamme and Emerson, 2002]. Within the ETNP and AS ODZs, additional supersaturation in (N2:Ar)norm due to denitrification in the ODZ is apparent. In the ETNP, (N2:Ar)norm rose to a maximum of 1.021 ± 0.7‰ at 300 m at St. 48 and 1.019 ± 0.7‰ at 350 m at St. 49 (σθ = 26.5 and 26.7, respectively). At St. One in the AS ODZ, (N2:Ar)norm increased to a maximum of 1.017 ± 0.7‰ at 200 m (σθ = 26.4). Below both ODZs, (N2:Ar)norm eventually reached values similar to the deep Pacific and Indian Oceans.

Figure 5.

Density profiles of measured N2:Ar ratios normalized to atmospheric equilibrium ratios [(N2:Ar)norm] both within (circles) and outside (triangles) of the (left) ETNP and (right) AS ODZs. The best fit line to (N2:Ar)norm outside of each ODZ is displayed [(N2:Ar)back]. Fine lines indicate 1 SE of (N2:Ar)back.

[21] Since there are no significant processes other than denitrification that either produce or consume N2 in the interior of the ocean, the amount of monatomic N added as gas by denitrification (Nxs) can be determined as the difference between the N2 in the ODZ versus in waters unaltered by denitrification:

display math

where (N2:Ar)norm is the observed normalized ratio, (N2:Ar)back is the normalized background ratio for waters outside of the ODZ, (N2)θ,S is the atmospheric equilibrium concentration of nitrogen gas at a given θ and S, and the factor of 2 converts to units of μM monatomic N.

[22] Background N2:Ar were determined using nonlinear least squares regressions of (N2:Ar)norm versus potential density (σθ) from outside and upstream of each ODZ (Figure 5). It is reasonable to use vertical profiles to characterize (N2:Ar)back as the horizontal gradients of N2:Ar across the Pacific are very small relative to the Nxs (Emerson, unpublished data). Gas data from the tropical and subtropical Pacific was used to represent waters entering the ETNP ODZ, prior to alteration by denitrification:

display math

(SE = 0.0017). To estimate the background for the AS ODZ, gas data from the WOCE I07N line was used [Devol et al., 2006]:

display math

(SE = 0.0015). In the ETNP ODZ, the Nxs was near zero at the surface of both St. 48 and 49, increasing to a maximum of 13.7 ± 1.8 μM at 300 m (σθ = 26.5) at St. 48 and of 10.8 ± 1.9 μM at 350 m (σθ = 26.7) at St. 49, before decreasing to near zero values in the deep waters (Figure 4). At St. One in the AS ODZ, Nxs in the surface waters was near zero (Figure 4), rising to a maximum of 11.1 ± 1.5 μM at 200 m (σθ = 26.4). Below this, Nxs decreased, reaching approximately zero in the deep waters.

4. Discussion

4.1. DIN Deficit Versus N2 Excess

[23] In light of the uncertainty in the ability of Ndef to predict the actual N denitrified in the ODZs, we compare Ndef and Nxs in the ETNP ODZ, the last of the three major ODZs of the ocean to be examined in this fashion (the ETSP and AS were examined by Chang et al. [2010] and Devol et al. [2006], respectively). Overall, Ndef and Nxs were comparable in both shape and magnitude. This result can be most easily observed in a plot of Nxs versus Ndef, aligned by potential density, where almost all of the values lie close to 1:1 (Figure 6). To compare these two methods of estimating N loss, the Bland – Altman test was used. In the ETNP ODZ, the mean of the differences between the two methods was approximately zero (−1.2 ± 1.9 μM) indicating that Ndef is a good estimate of the actual amount of DIN removed in the ETNP ODZ.

Figure 6.

A comparison of the N2 excess (Nxs) and DIN deficit (Ndef) in the (left) ETNP ODZ and the (right) AS ODZ shown with 1:1 line. Error bars are 1 SD.

[24] At St. 49, Ndef and Nxs were similarly correlated with the exception of the surface where Nxs was slightly negative compared to Ndef of 4.5 μM. Though the coarser sampling resolution of Nxs compared to Ndef does not allow a detailed comparison of the two parameters in the upper water column, a discrepancy between these two values at depths <60 m suggests the upwelling of Ndef-bearing waters while the accompanying Nxs was lost by equilibration with the atmosphere. Although the mismatch in Ndef and Nxs extended deeper than mixed layer observed in the study area (∼10 m), Nxs could have been lost by equilibration with the atmosphere at a time when the mixed layer was >60 m. The potential for the denitrification signal to be lost when Nxs-bearing waters are brought in contact with the atmosphere is a possible weakness in this method and would lead to an underestimate of the amount of N denitrified in the ODZ; however Nxs lost by atmospheric equilibration is small relative to entire Nxs.

[25] We also revisit the AS ODZ, which was previously examined by Devol et al. [2006], in order to evaluate any temporal variations in Nxs relative to Ndef. As in the other two major ODZs, Nxs and Ndef were comparable in both shape and magnitude, with the two values being close to 1:1 (Figure 6). Both Nxs and the Ndef were at a minimum at the surface. Below, Nxs increased to a maximum of 11.1 ± 1.5 μM N at σθ = 26.4 (200 m), whereas Ndef reached a maximum of 12.5 ± 1.0 μM N at σθ = 26.5 (230 m). The difference in the locations of the maxima are likely due to differences in the sampling interval and may not be the exact positions of the true maxima of these values in the water column. The Bland – Altman mean of the differences was −0.4 ± 1.2 μM, demonstrating that Ndef is a good measure of the actual amount of DIN lost in the AS ODZ.

[26] This result in the AS ODZ differs from that of Devol et al. [2006], who found Nxs was approximately twice Ndef. A possible source of disparity is a difference in Nexp and/or (N2:Ar)back employed in this study compared to previous work. To determine Nexp, Devol et al. [2006] used a linear regression of all USA JGOFS data with oxygen ≥65 μM and latitudes between 9.5 and 11°N. In this present study, Nexp was determined from the relationship of [DIN] to [PO43] along isopycnals in the Indian Ocean to the south of the Arabian Sea (see Section 3.2). Nexp between the two studies were similar, thus using the Nexp of Devol et al. [2006] with the ODZ nutrient data collected in the present study resulted in only a slightly smaller maximum Ndef (10.6 ± 0.6 μM). Also a different equation representing (N2:Ar)back was used in this study compared to Devol et al. [2006]. Though the same 1995 WOCE I07N N2:Ar data were used as background N2:Ar, updated solubility coefficients [Hamme and Emerson, 2004] were applied to determine N2:Ar at atmospheric equilibrium [(N2:Ar)θ,S], which affects (N2:Ar)norm (equation (3)). The nonlinear least squares regression of the updated (N2:Ar)norm background values resulted in equation (6). The difference due to using the (N2:Ar)back of Devol et al. [2006] with the (N2:Ar)norm collected in the present study yielded a negligibly larger Nxs (maximum 11.4 ± 1.5 μM N). Given that using Nexp and (N2:Ar)back of Devol et al. [2006] resulted in equivalent Ndef and Nxs within error and similar values to the present study, this cannot explain the disparity between this and prior comparisons of these values in the AS ODZ.

[27] Another possible explanation for this discrepancy is temporal variability in the size of Nxs relative to Ndef. The Arabian Sea is subject to strong physical forcings associated with seasonal monsoonal cycles which cause a reversal of the prevailing winds and water circulation and alter the patterns of productivity [Naqvi et al., 1990; You, 1997]. High precision N2:Ar data are relatively scarce; however, we may be able to assess the magnitude of the temporal variability in Nxs by examining fluctuations in the more ubiquitous nutrient distributions and Ndef.

[28] The 1994–1995 U.S. JGOFS Arabian Sea Process Study provides an ideal data set for exploring seasonal variations. High quality [DIN] and [PO43−] were measured over a complete annual cycle that included the SW and NE monsoons and intermonsoons. Station N7 is nearly identical to the location of this study in the AS ODZ. No significant seasonal variations in [DIN] and [PO43−] were observed in the depth interval of the ODZ [Morrison et al., 1999]. Consequently, seasonal variations in Ndef were also small (Figure 7). Nutrient data and Ndef from this study also show little deviation from the JGOFS data set. The constancy of nutrient distributions and Ndef over seasonal cycles, as well as interannually, does not suggest a significant difference in the actual amount of DIN lost from the ODZ, contrary to the twofold difference in Nxs reported by Devol et al. [2006] relative to this study.

Figure 7.

Density profiles of the DIN deficit (Ndef) at Sta. N7 for all JGOFS cruises. Solid line represents Ndef determined during this study (Sept '07). Error bars indicate 1 SD.

[29] Of the several factors Devol et al. [2006] invoked to explain Nxs being twice Ndef, the N:P of the organic matter remineralized in the ODZ has perhaps the most potential to be highly variable given the extremes of the monsoonal cycles and observations of both high productivity fueled by nutrient-rich upwelling [Naqvi, 1991] and N-fixingTrichodesmium blooms [Capone et al., 1998], which may have a relatively high N:P [Karl et al., 1992; White et al., 2006]. However, if temporal fluctuations in Nxs were more significant than in Ndef, the agreement between the Ndef and Nxsin all three major ODZs would be extremely fortuitous, especially in light of substantial hydrographic variability associated with the seasonal monsoons in the AS and the El Niño-Southern Oscillation in the ETSP. Nevertheless, the possibility that the difference between the findings ofDevol et al. [2006] and this study is due to natural variability cannot be excluded.

4.2. Implications for the Global N Budget

[30] Finding Nxs comparable to Ndef implies that a water column denitrification rate calculated using Ndef is a reasonable approximation of the DIN actually lost in both the ETNP and AS ODZs. Taken together with similar observations in the ETSP ODZ, where Nxs was also found to be comparable to Ndef, these results validate the major assumption of Ndef, i.e., the DIN denitrified in the three major ODZs of the ocean can be predicted by the Redfieldian N:P relationships established in the source waters of the ODZs. This conclusion also places some constraints on the DIN lost in the ODZ, namely that the major source of this denitrified DIN is deep-sea nitrate, which would be associated with a proportional amount of PO43−.

[31] It is worth noting, given the current debate regarding the pathways of N removal in the major ODZs of the ocean [Thamdrup et al., 2006; Nicholls et al., 2007; Hamersley et al., 2007; Farías et al., 2009; Lam et al., 2009; Ward et al., 2009; Canfield et al., 2010; Jensen et al., 2011], that although the results of this study indicate the major sources of the denitrified N, the exact pathway of N loss remains unresolved. There are several possible routes to N2 from the starting products. The observed vertical distributions of NO3, NO2, NH4+, and N2 can arise from varying contributions of dissimilatory reduction of nitrate to nitrite and ammonium (DNRN and DNRA, respectively), heterotrophic denitrification, and anammox, the sum of which are stoichiometrically indistinguishable [Koeve and Kähler, 2010].

5. Conclusions

[32] Estimates of the N2 excess above the background [N2] can quantify the amount of biologically mediated N removal from the major ODZs of the world without depending on assumptions of the N:P ratio of organic matter being degraded or the pathway of N2 production. In this study we find that, in contrast with previous findings in the AS ODZ, the N2 excess is comparable to the DIN deficit in both the ETNP and AS ODZs. Taken together with similar previous results in the ETSP ODZ, this indicates that estimates of fixed N loss based on deficit calculations in the three major ODZs of the world are likely reasonable, given the same considerations of volume and residence time of the water of the ODZ. This conclusion is based on a limited number of profiles from each ODZ; given the disparate findings in the AS ODZ, a more comprehensive survey should be undertaken to assess temporal and spatial variability in the relationship between the N2 excess and DIN deficit in order to accurately assess the rate of fixed N loss in the world's 3 major ODZs.

Acknowledgments

[33] We thank the Captains and crews of the R/Vs Ka'imimoana, Knorr, Mirai, New Horizon, and Roger Revelle for their assistance collecting the samples for this study. We are grateful to B. N. Popp and F. G. Prahl for collecting gas samples and providing nutrient data from the ETNP ODZ, and to K. L. Casciotti for providing nutrient samples from the AS ODZ. We thank C. Stump, J. L. Stutsman, and M. Haught for analytical support, and D. Nicholson and M. A. Altabet for their input interpreting data. We would also like to acknowledge N. Gruber and C. Deutsch, whose thoughtful comments improved the manuscript. Support for this work was provided by NSF (OCE-0647981 and OCE-1029316) and the Harry Hess Fellowship of Princeton University.