Water column denitrification rates in the oxygen minimum layer of the Pacific Ocean along 32°S



[1] To date, estimates of water column denitrification rates in global- and basin-scale nitrogen budgets do not include any values obtained in the open ocean outside the oxygen minimum zones (OMZs). The high-quality, 32°S Pacific repeat hydrographic surveys conducted in 1992, 2003, and 2010 offer an opportunity to estimate denitrification rates on basin scales in the open ocean away from the Eastern Tropical Pacific (ETP) oxygen minimum zones. Here, extended optimum multiparameter analysis and N* are used to estimate water column denitrification (ΔNdeni) in the oxygen minimum layer (OML; 1100–2000 dbar). Water column denitrification rates for 2003 and 2010 are estimated from the regression analysis using ΔNdeni and transit time distribution ages based on CFC12. Estimated mean denitrification rates in the OML at 32°S range from 3.1 to 7.5 µmol N m−3 yr−1 (=2.6–6.8 mmol N m−2 yr−1 with a thickness of 900 m). The OML denitrification rates are 2–3 orders of magnitudes lower than those estimated in the OMZs of ETP. However, their significance to Pacific nitrogen budget needs to be investigated in the future. Although uncertainties remain, our results may ultimately contribute to an improved understanding of marine nitrogen cycles.

1 Introduction

[2] The marine nitrogen cycle is composed of a variety of biogeochemical processes and pathways (e.g., nitrogen fixation, denitrification, dissimilatory nitrate reduction to ammonium, and anaerobic ammonia oxidation) in the ocean [Brandes et al., 2007]. Study of the global nitrogen balance between sources (i.e., nitrogen fixation) and sinks (i.e., denitrification) provides important insights into both glacial-interglacial climate change and the global carbon cycle through which the availability of nitrogen controls the efficiency of the biological pump absorbing atmospheric CO2 [Gruber and Sarmiento, 1997; Karl et al., 2002; Gruber, 2008; Gruber and Galloway, 2008]. For these reasons, establishing the nitrogen mass balance of the global oceans in terms of sources and sinks has been a long-term scientific goal [Gruber and Sarmiento, 1997; Deutsch et al., 2001; Karl et al., 2002; Codispoti, 2007].

[3] Denitrification is a dissimilatory process in which bacteria use nitrate as an electron acceptor instead of oxygen under low oxygen conditions. Nitrate (NO3) is typically reduced to nitrite (NO2) and then to nitrous oxide (N2O) and dinitrogen (N2) during the process. Among the sink processes in the marine nitrogen cycle, denitrification (NO3− → NO2− → N2O/N2) generally plays the dominant role in removing nitrogen from the ocean environment, and owing to widely nitrogen-limited conditions within the oceans, it has a direct influence on biological productivity [Falkowski, 1997].

[4] Direct measurements have been substantially limited in space and time so at large scales in the open ocean, the N* (=N − 16P + 2.9 µmol kg−1; N* > 0: nitrogen fixation, N* < 0: denitrification) technique has been a convenient method for estimating global water column denitrification rates [Gruber and Sarmiento, 1997; Deutsch et al., 2001]. Although useful for obtaining broad estimates, this technique relies upon fixed ratios and does not account for mixing. Generally, however, observed measurements of ocean parameters are determined by physical mixing combined with biogeochemical processes [Anderson and Sarmiento, 1994]. Extended optimum multiparameter (eOMP) analysis provides one way to consider physical mixing and biogeochemical changes together. Hupe and Karstensen [2000] applied the eOMP analysis to estimate physical mixing ratios and biogeochemical changes in the Arabian Sea and demonstrated this technique to be a viable alternative to the N* method for the estimation of denitrification using hydrographic data on basin scales.

[5] At present, estimates of water column denitrification rates in global- and basin-scale nitrogen budgets do not include any open ocean values for the full water column, partly because denitrification rates are expected to be significantly lower or nonexistent in the open ocean water column outside the oxygen minimum zones (OMZs) where dissolved oxygen is most depleted, and also because high-quality hydrographic data and time-dependent tracers (e.g., Chlorofluorocarbons and SF6) are required for the analysis. Macroscopic aggregates of detritus (i.e., “marine snow”) are ubiquitous in the pelagic water column in the world's oceans [Alldredge and Silver, 1988], forming pelagic microzones. They result in high microbial activity and consequently deplete oxygen within and around detrital aggregates [Alldredge and Cohen, 1987; Shanks and Reeder, 1993; Wolgast et al., 1998], providing “reducing microenvironments.” These conditions may allow denitrification to occur in the pelagic microzones despite aerobic conditions in ambient waters [Li and Peng, 2002]. Yamagishi et al. [2005] recently reported that denitrification at Station KNOT (44°N, 155°W) in the North Pacific OMZ may be driven by the microzones of aggregates despite aerobic conditions. OMZs develop in specific regions (e.g., the eastern tropical Pacific and the Arabian Sea) experiencing extremely high primary production, whereas oxygen minimum layers (OMLs) are widely distributed throughout the global open ocean, lying at various depths beneath thermocline [Karstensen et al., 2008]. Here, it is suggested that within these broadscale OMLs, active microbial respiration in the water column is possible, and therefore, OMLs could represent favorable environments for forming aggregate microzones in which denitrification could occur. Although the magnitude of such denitrification in OMLs would not be expected to as large as what is seen in the OMZs, the area involved is substantially larger, suggesting that the magnitude of the integrated effect could be substantial. Readily available high-quality long-line transect observations across open oceans make testing of the hypothesis feasible.

[6] Gruber and Sarmiento [1997] described N* minima in the Eastern Tropical North and South Pacific (ETNP and ETSP) where there are OMZs. They showed that variation of N* in the Pacific Ocean outside of the OMZs is small and suggested that either there is little N2 fixation or denitrification in anoxic microzones occurs on a large enough scale to maintain a balance between the two processes. Deutsch et al. [2001] estimated water column denitrification rates in the ETNP (110°W–coast and 10°S–20°N, 22 Tg N yr−1) and the ETSP (110°W–coast and 10°S–25°S, 26 Tg N yr−1), using the N* method with CFC12 tracer ages. Deutsch et al. [2007] estimated nitrogen fixation rates with an ocean general circulation model applied at global and basin scales and suggested that nitrogen fixation rates are higher in the nutrient depleted subtropical Pacific Ocean than elsewhere, except for the OMZs of ETNP and ETSP. Recently, DeVries et al. [2012] estimated water column denitrification rates in the ETNP (23–29 Tg N yr−1) and the ETSP (21–33 Tg N yr−1), using a data-constrained ocean circulation model. These estimates are quite similar to those of Deutsch et al. [2001].

[7] To improve the accuracy of the oceanic nitrogen inventory mass balance, it is crucial to estimate denitrification rates in unexplored regions. At present in the Pacific Ocean, the model-based estimate of the nitrogen fixation rate (93–108 Tg N yr−1; 40°S–65°N [Deutsch et al., 2007]) is larger than the total estimated denitrification rates (water column denitrification rate in OMZs + benthic denitrification rate = 48 + 15 = 63 Tg N yr−1; 30°S–60°N [Deutsch et al., 2001]). This result is opposite to the thinking that in a global oceanic nitrogen mass balance, nitrogen sinks are greater than nitrogen sources [Codispoti, 2007]. Denitrification has not been measured directly in open oceans to date, so it is valuable to investigate water column denitrification rates in the OML outside the OMZs to determine whether these estimated OML denitrification rates represent a significant portion in the Pacific nitrogen budget.

[8] The technique of CFC12 tracer ages that Deutsch et al. [2001] used for their analysis is based on the assumption that a water parcel moves from the surface mixed layer into the interior through advection alone, without mixing with surrounding waters. However, in reality, physical mixing does occur in ocean environment, and even though shallow thermocline ventilation might experience a minimum bias in tracer age from mixing, the effect of mixing on the estimation of water parcel age [Karstensen and Tomczak, 1998; Hall et al., 2004] is just as important as the magnitude of denitrification itself to the calculation of water column denitrification rates. The transit time distribution (TTD) technique used for the present analysis accounts for the effects of mixing in the calculation of water parcel age (see section 3.2).

[9] The hydrographic survey at ~32°S, known as the P06 line, was occupied approximately decadally (1992, 2003, and 2010) in the subtropical South Pacific Ocean as part of the U.S. World Ocean Circulation Experiment (WOCE) and CO2/Climate Variability and Predictability Research (CLIVAR) Repeat Hydrography programs. These high-quality observations provide an opportunity to estimate open ocean denitrification rates away from the South Pacific OMZ.

[10] To investigate the hypothesis that the open ocean water column denitrification in the OML may be a significant component of the marine nitrogen cycle, this study (1) estimates water column denitrification in the South Pacific with the P06 (~32°S) hydrographic data set using the N* method as well as eOMP analysis, (2) compares the magnitude of the differences between the two methods, and (3) estimates mean water column denitrification rates in the OML of P06 (~32°S) using the TTD method.

2 Data Sets

[11] The P06, 32°S repeat hydrographic transects were conducted in 1992, 2003, and 2010 (Figure 1). These data as well as the other data sets described below and in Figure 1 are available from the Climate Variability and Predictability (CLIVAR) and Carbon Hydrographic Data Office (CCHDO; http://cchdo.ucsd.edu/).

Figure 1.

Map showing the P06 (~32°S) hydrographic section thick white line and bottom topography of the Pacific Ocean. Hydrographic data from the WOCE P16N–17N, P03, P21, P15S, and P19C (CCHDO; http://cchdo.ucsd.edu/) sections illustrated here (thick black lines) were used to define the physicochemical characteristics of source water types (i.e., AAIW, PDW, U/LCDW, and AABW). GLODAP (Global Ocean Data Analysis Project) and CARINA (Carbon in the North Atlantic) data extracted between 40°S and 20°S (thin white lines) (http://cdiac3.ornl.gov/waves/discrete/) (≥600 dbar) were used to estimate the regional Redfield ratios for the P06 line along with the observed P06 data.

[12] The parameters used for the eOMP analysis are latitude, longitude, pressure, temperature (T), salinity (S), dissolved oxygen (DO), nitrate (N), phosphate (P), silicate (Si), total alkalinity (Talk), and dissolved inorganic carbon (DIC). Missing Talk and DIC values were estimated from paired DIC—total pH data and paired Talk—total pH data, respectively, through the CO2SYS program [Van Heuven et al., 2009]. Note that only bottle data labeled with WOCE/CLIVAR “acceptable” quality flags [Joyce and Corry, 1994] were used. To reduce bias from seasonal variation at the surface layer, only data deeper than 600 dbar were included in the eOMP analysis.

3 Methods

3.1 Extended OMP Analysis

[13] The eOMP analysis is an inverse method based on an overdetermined system [Hupe and Karstensen, 2000]; that is, the number of equations is greater than number of unknown variables. Such systems have multiple solutions. Solutions are found using nonnegative least squares with the constraint of mass conservation inline image [Tomczak and Large, 1989]. Here, modified silicate and total alkalinity equations are included (supporting information), and the matrix form for the eOMP analysis is given as

display math(1)

where the matrix A is defined as the physicochemical characteristics of the source water types and the vector X is composed of mixing ratios (xi) for the source water types, the amount of remineralized phosphate (ΔPremi), denitrification (ΔNdeni), inorganic carbonate dissolution (ΔCinorg), and inorganic silicate dissolution (ΔSiinorg). The vector b represents observed concentrations, and the vector R constraint residuals. The last row consisting of 1's enables mass conservation (∑xi = 1). The ratios rCorg:Si:N:P:-O2 indicate Redfield ratios [Redfield et al., 1963], and rD is the amount of phosphate produced by denitrification given as 1/104 [Gruber and Sarmiento, 1997; Hupe and Karstensen, 2000]. More details concerning the silicate and total alkalinity equations in this eOMP analysis are provided in the supporting information.

3.1.1 Physicochemical Characteristics of Source Water Types for the 32°S Line

[14] The South Pacific Ocean is biogeochemically important as it provides the only source of oxygen to the deep North Pacific basin through the deep western boundary current system [Tomczak and Godfrey, 1994]. Major water masses present at 32°S below 600 dbar include Antarctic Intermediate Water (AAIW), Pacific Deep Water (PDW), Upper/Lower Circumpolar Deep Water (U/LCDW), and Antarctic Bottom Water (AABW) [Tomczak and Godfrey, 1994; Wijffels et al., 2001; Murata et al., 2007; Talley et al., 2011]. All these waters include admixtures of waters ventilated in the Southern Ocean. As eOMP analysis requires the physicochemical characteristics of source water types, five different source water types were defined to explain the physical mixing for the eOMP analysis at 32°S below 600 dbar (Table 1 and Figure 2).

Table 1. The Physicochemical Characteristics of Source Water Types (SWT) and Weights for Each Parameter Used for the eOMP Analysis on the P06 Line (~32°S)a
SWTPTb (°C)S (psu)DO (µmol kg−1)Si (µmol kg−1)NO3 (µmol kg−1)PO4 (µmol kg−1)TALK (µmol kg−1)DIC (µmol kg−1)
  1. aAAIW: Antarctic Intermediate Water, U(L)CDW: Upper (Lower) Circumpolar Deep Water, PDW: Pacific Deep Water, AABW: Antarctic Bottom Water.
  2. bPT: potential temperature.
  3. cPDW was defined at the P03 (~24°N) line in the North Pacific Ocean (Figure 1).
Group 1
Group 2
Figure 2.

Temperature-salinity diagram for P06 (~32°S) based on the 1992, 2003, and 2010 data sets. The physical characteristics of source water types (AAIW: circle, UCDW: square, PDW: diamond, LCDW: triangle, and AABW: star) are also indicated.

[15] Because PDW is formed within the deep overturn through mixing throughout the North Pacific Ocean, the physicochemical source water characteristics of PDW are defined at three different sites using the P16N/17N line (at 152°W and 135°W, north of 40°N), P03 (24°N), and P21 (18°S) (Figure 1), with the criteria of silicate and alkalinity maxima for comparison (Table 3). The effect of spatially varying PDW's definitions on the results of eOMP analysis is discussed in section 4.2. Here, unless otherwise stated, results based on the eOMP analysis using PDW defined at 24°N are quoted.

[16] Due to a limitation in the of number of source water types that can be used in the eOMP analysis at any one time (i.e., x1x4 in the X vector of equation (1)), we divided the data into two groups; Group 1 consists of AAIW, UCDW, PDW, and LCDW, and Group 2 with UCDW, PDW, LCDW, and AABW, with substantial amount of overlap between the two groups. The effects of the different depth ranges for Groups 1 and 2 are discussed in section 3.1.4. Detailed information on physicochemical characteristics of source water types is described in the supporting information.

3.1.2 Assignment of Redfield Ratios

[17] The eOMP analysis represents the biogeochemical changes according to stoichiometric ratios (rC:Si:N:P:-O2) (matrix A, equation (1)). The traditional Redfield ratios (rC:N:P:-O2 = 106:16:1:138) [Redfield et al., 1963], revised ratios, e.g., rC:N:P:-O2 = 117 ± 14:16 ± 1:1:170 [Anderson and Sarmiento, 1994], and rC:N:P:-O2 = 106:16:1:150 ± 10 [Anderson, 1995] have all widely been used to account for the biogeochemical changes in the ocean. To consider the sensitivity according to the use of different Redfield ratios in the eOMP analysis, we also estimated the regionally representative Redfield ratios for the Pacific 32°S line through regression analysis using observed DO and nutrient (N, P, and Si) data from the 1992/2003/2010 P06 hydrographic lines, GLODAP (Global Ocean Data Analysis Project), and CARINA (Carbon in the North Atlantic) below 600 dbar between 40°S and 20°S (available at http://cdiac3.ornl.gov/waves/discrete/) (Figure 1). Assuming rC:P = 106, the regional Redfield ratios for the P06 line were estimated as rC:Si:N:P:-O2 = 106:18.5 ± 0.5:13.6 ± 0.1:1:126 ± 0.5. Human-induced anthropogenic perturbations are altering the global carbon and nitrogen cycles [Gruber and Galloway, 2008; Deutsch and Weber, 2012]. These ongoing spatial and temporal changes imply that it might be difficult to determine a single set of representative Redfield ratios for the full Pacific line over the almost 20 years the observations span. In addition, the Redfield ratios are not constant with depth [Anderson and Sarmiento, 1994]. Therefore, the four different cases of Redfield ratios mentioned above were considered (Table 2): case 1 rC:Si:N:P:-O2 = 106:18.5:16:1:138 [Redfield et al., 1963], case 2 rC:Si:N:P:-O2 = 117:18.5:16:1:170 [Anderson and Sarmiento, 1994], case 3 rC:Si:N:P:-O2 = 106:18.5:16:1:150 [Anderson, 1995], and case 4 rC:Si:N:P:-O2 = 106:18.5:13.6:1:126 (regional ratios). The Si:P ratio (rSi:P = 18.5) derived from the regionally observed Redfield ratios for the P06 line was collectively assigned to cases 1–3. We then used the results averaged from the four cases for the final quoted values.

Table 2. Four Different Cases of Redfield Ratios Applied to the eOMP Analysis
113811618.5106Redfield et al. [1963]
217011618.5117Anderson and Sarmiento [1994]
315011618.5106Anderson [1995]
4126113.618.5106This study (2012)

3.1.3 Parameter Weights

[18] Among the hydrographic parameters used, only temperature and salinity are determined solely by physical mixing. As they are independent from biological effects, we consider them to be conservative parameters. On the other hand, the DO, nutrients, and components of the carbonate systems are nonconservative parameters because they are influenced not only by physical mixing but also by biological activity.

[19] The eOMP analysis uses conservative and nonconservative parameters together. The former have high accuracy, whereas the latter have relatively lower accuracy due to biological activity. To resolve this systematic problem, Tomczak and Large [1989] assigned weights to each parameter:

display math(2)

where σj2 is the variance of parameter j calculated from the physicochemical characteristics of source water types in the region specified in section 3.2 and δjmax is the largest observed variance in parameter j in the source regions at the time of water mass formation. Due to the difficulty in determining δjmax, a modified weight equation [Kim and Lee, 2004] is used here.

display math(3)

where σj is the standard deviation of parameter j calculated from the physicochemical characteristics of source water types and accuracyj is the measurement error of parameter j. The weights for each parameter are summarized in Table 1.

3.1.4 Depth Range Validation

[20] Five different source water types are assumed to be involved in the physical mixing that produces the water masses along the Pacific 32°S line below 600 dbar. For ease of use within the eOMP method, these source waters are grouped as AAIW, UCDW, PDW, and LCDW (Group 1) and UCDW, PDW, LCDW, and AABW (Group 2).

[21] Mass conservation residuals (RMC) are examined to validate solutions [Tomczak and Large, 1989; Kim and Lee, 2004].

display math(4)

[22] The RMC (%) for Group 1 is less than 5% between 1100 and 2800 dbar but is larger between 600 and 1100 dbar and below 2800 dbar (Figure 3a). The higher Group1 RMC (%) between 600 and1100 dbar is due to the influence of South Pacific Central Water (SPCW) and Subantarctic Mode Water (SAMW) [see Talley et al., 2011, Table S10.4], while the higher residuals below 2800 dbar represent the influence of AABW, which is not included in Group 1. The RMC (%) for Group 2 is less than 5% below 2400 dbar (Figure 3b). In summary, seawater properties between 1100 and 2800 dbar are well described by Group 1, while those below 2400 dbar are best illustrated by Group 2. We use a cut off at 2600 dbar (i.e., between the Group 1 lower limit of 2800 dbar and the Group 2 upper limit of 2400 dbar based on a residual threshold of less than 5%), so that the eOMP analysis results for Group 1 covers the depths range from 1100 to 2600 dbar and those for Group 2 applies to depths below 2600 dbar (Figure 3c).

Figure 3.

Residuals of mass conservation (%) versus pressure (≥600 dbar) for (a) Group 1 (i.e., AAIW, UCDW, PDW, and LCDW), (b) Group 2 (i.e., UCDW, PDW, LCDW, and AABW), and (c) Group 1 for 1100–2600 dbar and Group 2 for >2600 dbar with the criterion of residuals (<5%).

3.2 Transit Time Distributions (TTD) Method

[23] The TTD technique is mathematically based on Green's functions (G), which translate the surface concentration of transient tracers into the ocean interior by advective and diffusive processes [Holzer and Hall, 2000; Waugh et al., 2003; Waugh et al., 2006]. The general form of the TTD method is expressed as

display math(5)

where C(xinterior,t) is the concentration of a passive tracer at time (t) and the interior position (xinterior), C0(xsurface,t) is the surface boundary condition representing the history of surface concentration, t-t′ is the elapsed time since the water parcel was last in contact with the atmosphere, and G(x,t′) is the Green's function (i.e., TTD) that propagates the surface concentration of transient tracer into the ocean interior. It is assumed that a TTD at its interior position is determined by the inverse Gaussian function as follows [Waugh et al., 2003]:

display math(6)

where Γ is the mean age and Δ represents width.

[24] Waugh et al. [2004] reported that TTDs calculated such that the mean age (Γ) equals the width of the TTD (Δ) reproduce the transient tracer (e.g., CFCs, tritium, and helium) features in the subpolar North Atlantic Ocean well. It is, however, important to find an appropriate Δ/Γ ratio for a specific study area to reduce the uncertainty of the TTD method [Tanhua et al., 2008]. A higher Δ/Γ ratio means that the effect of physical mixing is large and water parcel ages are older, whereas a lower Δ/Γ ratio indicates that advection processes are dominant and water parcel ages are younger. Here, the combination of three transient tracers, CFC11, CFC12, and SF6, are used to find an appropriate Δ/Γ ratio for the 32°S line with the data observed in 2010. Ideally, the TTDs from different tracers should be identical. However, since there are some potential errors (e.g., measurement error, different nonlinear input history of tracers into the interior of the ocean, and a lack of initial surface water saturation conditions [Tanhua et al., 2008]), some data deviate from the 1:1 regression line, Here, the best estimate Δ/Γ ratio for the 32°S line was defined as the one with the highest correlation coefficient (r) between tracer TTDs and was determined to range between 0.4 and 0.8 depending on the tracer pair combination used. The mean Δ/Γ = 0.6 (the best estimate for P06) was used along with available CFC12 measurements (2003 and 2010) to estimate TTD ages (Figure 4). The estimated TTDCFC12 ages indicating mean ventilation time scales for when water parcels were transferred from the surface to their interior locations by advection-diffusion processes range between ~100 and ~350 years for the water column of 27.2–27.84 σθ (~1100–6100 dbar). Using the TTDCFC12 age information, mean water denitrification rates were estimated. As so few CFC12 measurements were made in 1992, these data were not included in the estimation of water column denitrification rates.

Figure 4.

Vertical profile of P06 TTDCFC12 age (year) versus potential density (kg m−3) (2003, circle; 2010, square). Error bars represent standard deviation.

4 Results and Discussion

4.1 Vertical Distribution of Denitrification: eOMP Analysis versus N*

[25] The mean ΔNdeni estimated by the eOMP analysis gradually increases from ~0–0.3 µmol kg−1 at 27.2–27.3 σθ (~1100–1200 dbar) to ~0.7–1.0 µmol kg−1 at 27.6–27.7 σθ (~1600–2900 dbar), and it remains constant within ~0.5–1.0 µmol kg−1 below 27.7 σθ (Figure 5, right panel). On the other hand, the mean ΔNdeni estimated by the N* method (Figure 5, left panel) rapidly increases from −0.8–0 µmol kg−1 at 27.2 σθ (~1100 dbar) to −1.9 to −1.5 µmol kg−1 at 27.5–27.6 σθ (~1250–1750 dbar) layer (note: the greater the negative N* value, the higher the mean ΔNdeni). Below 27.6 σθ, it gradually decreases with increasing density (Figure 5a). The mean ΔNdeni estimated with N* is higher by ~0.6 µmol kg−1 on average than that computed with the eOMP analysis, but overall, the vertical pattern suggested by the two methods is similar.

Figure 5.

Vertical profiles of potential density (kg m−3) versus denitrification (µmol kg−1) for the 1992, 2003, and 2010 P06 transects. Circle (1992), square (2003), and star (2010) indicate eOMP-based denitrification. Triangle (1992), diamond (2003), and cross (2010) indicate N*-based denitrification. Error bars represent standard deviation. Note that although both methods compute denitrification, their definitions result in solutions of the opposite sign.

4.2 The Effect of PDW Definition on eOMP Results

[26] The eOMP analysis requires the physicochemical characteristics of source water types to be defined at their source regions [Tomczak and Large, 1989]. However, PDW is formed internally by mixing among intermediate, deep, and bottom waters in the North Pacific [Tomczak and Godfrey, 1994]. Here, PDW source water characteristics are defined at three different sections of P16N–17N (>40°N), P03 (~24°N), and P21 (~18°S) (Figure 1 and Table 3). The 2010 ΔNdeni results for these differing PDW definitions were compared (Figure 6). The differences in mean ΔNdeni are small (less than ~0.1 µmol kg−1) between 27.2 and 27.7 σθ (~1100–2900 dbar). Although the ΔNdeni estimates based on different PDW definitions are not significantly different from one another below 1100 dbar, there is a tendency for those that use the southern P21 (18°S) source water characteristics to produce lower ΔNdeni estimates and for those that use the northern P16/17 source water characteristics to produce higher ΔNdeni estimates. This pattern is particularly apparent below 27.7 σθ but becomes less obvious in the bottom waters (see the convergence between 27.82 and 27.84 σθ (~3400–6100 dbar) in Figure 6). The largest difference in mean ΔNdeni is ~0.4 µmol kg−1 within the 27.74–27.76 σθ (~2200–4400 dbar) layer, and overall, the averaged difference based on the various PDW definitions is 0.1 ± 0.2 µmol kg−1. Thus, the effect of spatially varying PDW's definitions on the results of eOMP analysis can be considered small.

Table 3. Physicochemical Characteristics of Pacific Deep Water (PDW) Defined at the P16N–17N (>40°N), P03 (~24°N), and P21 (~18°S) Lines of Pacific Ocean (Figure 1)
PDWPT (°C)S (psu)DO (µmol kg−1)Si (µmol kg−1)NO3 (µmol kg−1)PO4 (µmol kg−1)TALK. (µmol kg−1)DIC (µmol kg−1)
Figure 6.

Vertical profiles of P06 (2010) denitrification (µmol kg−1) estimated by eOMP analysis versus potential density (kg m−3) according to the PDW definitions defined at the P16N–17N (circle), P03 (square), and P21 (star), respectively. Note that the values (ΔNdeni and xi) averaged from cases 1–4 are used. Error bars represent standard deviation.

4.3 Estimating Water Column Denitrification Rates in 2003 and 2010 at ~32°S

[27] The main goal of this research is to estimate water column denitrification rates on open ocean scales. The water column denitrification rates (µmol N m−3 yr−1) are estimated from the slope of the regression line (i.e., mean ΔNdeni versus TTDCFC12 ages). Since we expect very low water column denitrification rates in open ocean, we primarily focus on the OML that lies approximately between 1100 and 2000 dbar along the P06 transect (Figure 7a) and investigate the spatial variation of water column denitrification rates between the west (~ 150°E–140°W) and the east (140°W to ~ 70°W) on the transect based on the vertical distribution of dissolved oxygen (Figure 7b). Vertical distributions of mean ΔNdeni and TTDCFC12 ages in the OML are shown in Figure 8.

Figure 7.

The P06 line is divided into two regions at 140°W based on the distribution of oxygen minimum layer (Western: 150°E–140°W circle, Eastern: 140°W–70°W square). (a) Vertical profile of dissolved oxygen (µmol kg−1) versus pressure (dbar) for the two regions and (b) map illustrating the east-west division along P06.

Figure 8.

Vertical distributions of (a) denitrification estimated by eOMP analysis, (b) N*, and (c) TTDCFC12 age (year) within the oxygen minimum layer (1100–2000 dbar) for P06 line (~32°S) for 2010. Note that the negative N* values were multiplied by −1 to facilitate comparison between the results of two techniques.

[28] The eOMP-based water column denitrification rates within the open ocean OML for the full 2010 P06 transect are estimated to be 3.1 ± 0.4 µmol N m−3 yr−1 (Figure 9a), about half the rate estimated according to the N* method, 7.5 ± 0.7 µmol N m−3 yr−1 (Figure 9b). Note that the negative N*-based values have been multiplied by −1 to facilitate comparison between the results of two techniques (Figures 9a and 9b). The N*-based rate was based on only the N/P ratio of 16, while the eOMP analysis considered mean values averaged from various cases of Redfield ratios including regionally observed N/P ratio (13.6) as well as the traditional N/P ratio (16) (Table 2) and also included the effects of physical mixing. It is likely that the different denitrification rates estimated from the two methods are at least partially the result of these different assumed conditions.

Figure 9.

Mean water column denitrification (µmol m−3) versus TTDCFC12 age (year) in the oxygen minimum layer (1100–2000 dbar) of P06 line estimated by regression analysis for the eOMP-based and N*-based denitrification rates. (a, c, e, and g) eOMP analysis and (b, d, f, and h) N* results. 2010 (Figures 9a and 9b), western 2010 (Figures 9c and 9d), eastern 2010 (Figures 9e and 9f), and eastern 2003 (Figures 9g ang 9h). Dotted lines represent standard deviation.

[29] The eOMP analysis of the 2010 P06 data suggests lower water column denitrification rates in the west, 3.2 ± 0.6 µmol N m−3 yr−1 (Figure 9c) compared to the east, 4.9 ± 0.6 µmol N m−3 yr−1 (Figure 9e). The N* method again suggests rates nearly twice as large as those from the eOMP analysis but also indicates a similar west versus east pattern with 4.8 ± 0.6 µmol N m−3 yr−1 (Figure 9d) in the west and 8.8 ± 1.2 µmol N m−3 yr−1 (Figure 9f) in the east. Comparing these to the 2003 estimates in the eastern basin, both the eOMP analysis (Figure 9g) and the N* technique (Figure 9h) suggest lower rates in the earlier observations, 3.8 ± 0.8 µmol N m−3 yr−1 (eOMP) and 6.6 ± 1.7 µmol N m−3 yr−1 (N*).

[30] The spatial gradient of water column denitrification rates in western versus eastern regions would result from physical and biogeochemical differences. As a result of strong upwelling of nutrient rich waters, the eastern region including the Chilean coasts shows higher chlorophyll concentrations than the western region (http://oceancolor.gsfc.nasa.gov/). This pattern may result in relatively high particle concentrations in eastern region [Sheldon et al., 1972], which may stimulate more favorable condition for denitrification. The spatial gradients of dissolved oxygen and nitrate concentrations in the hydrographic sections (http://www.ewoce.org/gallery/Map_Pacific.html) also reflect the greater nutrient supply and greater oxygen consumption in the coastal upwelling regions of the eastern basin. Lack of sufficient data in the western portion of the 2003 data set precludes the same comparison for eastern versus western regions. By multiplying the denitrification rate by the thickness of the OML (ΔH ≈ 900 m in Figure 9b), the water column denitrification flux in the P06 line is calculated to be 2.8–6.8 mmol N m−2 yr−1 (Table 4) with the lower bound given by the eOMP estimate and the upper bound by the N* estimate. Denitrification flux estimates in the western basin tend to be lower (2.9–4.3 mmol N m−2 yr−1) than those in the east (4.4–7.9 mmol N m−2 yr−1) regardless of the calculation method employed.

Table 4. Estimated Mean P06 Water Column Denitrification Rates in the Oxygen Minimum Layer (1100–2000 dbar, ΔH ≈ 900 m)a
YearMethodData (Number)Mean ΔNdeni Rate (µmol N m−3 yr−1)Mean ΔNdeni Flux (mmol N m−2 yr−1)Location
  1. aNote that for ease of comparison, the negative numbers of N* slopes are changed to positive numbers. Uncertainty of 1 standard deviation is represented.
  2. bEastern is 140°W–70°W (Figure 7b).
  3. cWestern is 150°E–140°W.
2003eOMP633.8 ± 0.83.4 ± 0.7easternb
2003N*636.6 ± 1.75.9 ± 1.5eastern
2010eOMP3883.2 ± 0.62.9 ± 0.5westernc
2010eOMP3474.9 ± 0.64.4 ± 0.5eastern
2010N*3884.8 ± 0.64.3 ± 0.5western
2010N*3478.8 ± 1.27.9 ± 1.1eastern
2010eOMP7353.1 ± 0.42.8 ± 0.4full transect
2010N*7357.5 ± 0.76.8 ± 0.6full transect

[31] The effect of variation in PDW definition on the estimation of denitrification rates was also investigated (Table 5). The estimated denitrification rates based on PDW defined north of 40°N (P16N–17N) and at ~18°S (P21) were similar, 3.0 ± 0.4 µmol N m−3 yr−1 and 2.9 ± 0.4 µmol N m−3 yr−1, respectively, and the corresponding denitrification fluxes with ΔH ≈ 900 m were calculated to be 2.7 ± 0.4 mmol N m−2 yr−1 and 2.6 ± 0.4 mmol N m−2 yr−1, respectively (Table 5). The lack of variation in the solutions indicates that the effect of spatially varying PDW definitions on the results of extended OMP analysis is small and not detectable (Tables 4 and 5).

Table 5. Mean P06 OML Water Column Denitrification Rates Based on eOMP Analysis According to PDW Defined Using the P16N–17N (>40°N) and P21 (~18°S) (Figure 1)a
YearPDW Defined atMean ΔNdeni Rate (µmol N m−3 yr−1)Mean ΔNdeni Flux (mmol N m−2 yr−1)
  1. aUncertainty of 1 standard deviation is represented. To estimate mean water column denitrification flux, a ΔH = 900 m is used.
2010P16N–17N3.0 ± 0.42.7 ± 0.4
2010P212.9 ± 0.42.6 ± 0.4

[32] Considering the effect of all potential uncertainties on the results, such as the variation of PDW definitions and the different techniques (eOMP versus the N*), this analysis suggests the overall range of denitrification rates estimated along 32°S in the Pacific Ocean by the eOMP analysis and N* using the TTD method is 3.1–7.5 µmol N m−3 yr−1 (≈3–7 × 10−3 µmol N kg−1 yr−1, or 2.6–6.8 mmol N m−2 yr−1 with ΔH = 900 m) (Table 5).

[33] Deutsch et al. [2001] estimated the water column denitrification rates using N* in the OMZs of ETNP and ETSP at 0.4 ± 0.1 µmol N kg−1 yr−1 (25.08–26.75 σθ) and 1.2 ± 0.7 µmol N kg−1 yr−1 (25.8–26.6 σθ), respectively. Although the water column denitrification rates estimated in the OML of the 32°S line (27.2–27.84 σθ) are two orders of magnitude lower than those estimated in the ETP, their significance to Pacific nitrogen budget needs to be explored in the future.

4.4 Uncertainty and Future Study

[34] Here we suggest that denitrification signals are derived from the microenvironments provided by marine snow, despite high oxygen conditions. The results presented here support this hypothesis, but direct evidence is not yet available and a qualitative and quantitative study on how reducing microenvironments are distributed in open oceans is needed. Both the N* technique and eOMP analysis are based on the difference between expected nitrate and observed nitrate concentrations, assuming all the nitrate deficit is from denitrification. Neither method can distinguish between denitrification signals derived from the water column itself and those derived from sedimentary interaction in the bottom layers. Thus, we confined our results to the OML to minimize the influences of other denitrification signals arising from sediments and OMZs. However, since additional tracers such as nitrogen and oxygen isotopes and N2 gas were not available to exactly determine where denitrification signals come from, the possible influence of these sources cannot be ruled out. Surveys that include direct measurement in open ocean OMLs are needed for the future study of marine nitrogen cycles.

5 Summary

[35] The amount of denitrification in the water column below 1100 dbar (27.2–27.84 σθ) was estimated along the P06 (32°S) line of South Pacific Ocean with the hydrographic data observed in 1992, 2003, and 2010 using eOMP analysis and the N* technique. The water column denitrification rates in the OML of the P06 line were estimated for 2003 and 2010 using the relationship between the amount of denitrification (ΔNdeni) and TTDCFC12 ages through linear regression analysis. The mean water column denitrification rate estimated by eOMP analysis (3.1 ± 0.4 µmol N m−3 yr−1) was found to be about half that estimated using the N* technique (7.5 ± 0.4 µmol N m−3 yr−1). Spatial variation of the water column denitrification rates showed that the water column denitrification rates in the eastern P06 (~140°W–70°W) are about two times higher than those in the western P06 (~150°E–140°W) regardless of the calculation method used. The mean water column denitrification rates estimated in the OML of the P06 line between 1100 and 2000 dbar (3.1–7.5 µmol N m−3 yr−1 = 2.6–6.8 mmol N m−2 yr−1 with ΔH = 900 m) are 2–3 orders of magnitudes lower than those estimated in the OMZs of ETP. However, to improve our quantitative understanding of Pacific nitrogen budget, future study is needed to examine potential influence of OMZs to open ocean OMLs using direct measurements of the full water column. Although uncertainties remain, the present results may ultimately contribute to an improved understanding of marine nitrogen cycles.


[36] We sincerely thank all the scientists, technical staff, ship's officers, and crew who contributed to the WOCE (1992 P06), the BEAGLE (2003 P06), and the CLIVAR (2010 P06) programs. In particular, we thank the crew of R/V Melville for their efforts on 2010 P06 cruise on which I.-N.K. and A.M.M. participated. We also thank T. Tanhua for help with the TTD calculations. I.-N.K. is partially supported by the NSF-CO2/CLIVAR Repeat Hydrography program for 2010 P06 participation. A.M.M. was funded through NSF grants OCE-0223869 and OCE-0926651, and NOAA grant NA11OAR4310063. This is the University of Texas Marine Science Institute contribution 1674.