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.
 A nitrogen stable isotopic model was constructed in order to constrain the Holocene marine-fixed nitrogen budget. The primary sources and sinks considered were riverine and atmospheric sources, nitrogen fixation, sedimentary and water column denitrification, and sediment burial. The source budget was found to be insensitive to changes in nitrogen fixation rates, and thus could not be used to constrain this term. However, the isotopic value of fixed nitrogen losses was very sensitive to the amount of sedimentary denitrification. If the isotopic value of marine-fixed nitrogen has not changed during the Holocene, as supported by sedimentary records, then in order to balance the isotopic value of sinks and sources, approximately 280 Tg N yr−1 of sedimentary denitrification is required. If such a high rate of denitrification has been sustained throughout the Holocene, it implies that present-day estimates of marine nitrogen fixation are grossly underestimated. It also implies that the marine nitrogen budget has a residence time of less than 2000 years.
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 marine nitrogen budget has been a subject of much debate over the last three decades. Initial budgets in the 1970s and early 1980s [Wada et al., 1975; Codispoti and Christensen, 1985] were linked to the idea that such budgets were balanced, i.e., sources equaled outputs. However, more recent studies have suggested that the fixed nitrogen budget has possibly swung between periods of deficit and excess, and that these swings might be linked to observed changes in atmospheric CO2 and N2O concentrations over the last 100,000–1,000,000 years [Altabet et al., 1995, 1999a; Ganeshram et al., 1995, 2000; Suthhof et al., 2001]. Even the most recent attempts at establishing a budget have been met with some controversy. At the heart of the debate are the sizes of sedimentary denitrification sinks and nitrogen fixation sources. As methods to measure sediment, denitrification rates have improved [Devol, 1991], and the aerial extent over which these measurements have been made has grown [Balzer et al., 1998; Devol and Christensen, 1993; Devol et al., 1997; Hammond et al., 1999; Hulth et al., 1997; Jahnke and Jahnke, 2000],while denitrification rate estimates have grown to greater than 100 Tg N yr−1. This has presented a difficulty in establishing a balanced fixed nitrogen budget, as total sinks (including water column denitrification and sediment burial) then exceed 200 Tg N yr−1. Some of these sinks are balanced by riverine and groundwater inputs, but the burden of replenishing marine-fixed nitrogen stocks falls upon nitrogen fixation, the other primary source of nitrogen to the oceans [Codispoti et al., 2001; Codispoti and Christensen, 1985]. However, direct measurements of nitrogen fixation have rarely supported rates in excess of 100 Tg N yr−1 [Capone, 2001]. It must be noted that neither fixation nor sediment denitrification rates are well constrained and they are subject to rapid episodic events that may dominate their overall rates [Gehlen et al., 1997; Lipschultz and Owens, 1996]. Furthermore, in most cases, global budgets are based upon a handful of values.
 There is a long history of using isotopes to constrain elemental budgets (e.g., 18O in the hydrologic cycle). There have only been a few attempts to use nitrogen isotopes to constrain all or parts of the fixed nitrogen budget [Altabet and Curry, 1989; Liu and Kaplan, 1988; Wada et al., 1975], and these attempts were limited by the lack of information on the isotopic values and fractionations associated with many of the major source and sink terms controlling the global oceanic nitrate budget. Work over the past decade has improved this situation, although care has to be taken to separate anthropogenic influences from baseline values. Given new information on the isotopic composition of the various marine nitrogen pools and fractionations associated with the processes that affect them, we have constructed a global isotopic budget to help constrain the Holocene marine-combined nitrogen cycle and understand the factors that control it.
2. Methodology: Important Terms in the Marine Nitrogen Budget
Table 1. Fluxes and Isotopic Values for Source and Sink Nitrogen Budgetsa
Flux, Tg N yr−1
Isotopic Value, ‰
Values in parentheses for isotopic values are estimated variances for each term. See text for references.
Atmosphere sources (DON + DIN)
Water column denitrification
 The argument over a balanced versus unbalanced fixed nitrogen budget has centered upon the relative sizes of the sediment denitrification versus nitrogen fixation terms. As more precise estimates of sedimentary denitrification have been made, the total loss rate for fixed nitrogen has grown markedly, from less than 30 Tg yr−1 [Wada et al., 1975] to well over 100 Tg yr−1 [Codispoti et al., 2001; Gruber and Sarmiento, 1997]. In a similar fashion, estimates of marine nitrogen fixation have grown from 15 Tg N yr−1 [Codispoti and Christensen, 1985] to 110 Tg N yr−1 [Gruber and Sarmiento, 1997], although the larger values have been justified primarily upon the need for a somewhat balanced budget and not upon field measurements.
 To help resolve both the issues of quantity and balance, we model a budget focusing upon the major terms, using the following terminology to describe the fractionations taking place. We will use the term “combined nitrogen” to refer to all forms of nitrogen except molecular nitrogen (N2). As the vast majority of this fixed nitrogen exists in the oceans in the form of NO3−, this is the species we will examine. The isotopic composition of nitrogen is defined as δ15N = (15N/14N)sample/(15N/14N)reference − 1) × 1000, with atmospheric N2 as the reference [Mariotti, 1983]. Isotopic fractionation from sources in the marine nitrogen cycle is represented as either α = 15R/14R or εdenit = (1 − α) × 1000, where 15R and 14R are the rates of gain (or loss) for 15NO3− and 14NO3−, respectively, for each process. Because the modern N budget is heavily impacted by anthropogenic inputs [Codispoti et al., 2001; Falkowski et al., 1998; Galloway et al., 1995; Nevison and Holland, 1997], careful examination of data from a variety of sources is necessary to examine Paleocene and Holocene nitrogen budgets. We have focused upon the available literature from pristine sites. In some cases, no specific data are available and inferences must be made from spatial or concentration patterns.
2.1.1. Atmospheric Precipitation
220.127.116.11. Mass Estimate
 Establishing an isotopic value for atmospheric inputs is difficult due to the large size and ubiquitous nature of anthropogenic inputs. These involve both inorganic NOx emissions from combustion engines and power plants,enhanced NH3 and NxO emissions from fertilized soils [Asner et al., 2001; Nevison and Holland, 1997; Perez et al., 2001], and possibly organic nitrogen emissions from combustion and soil releases [Cornell et al., 1995]. Duce et al.  and Galloway et al.  give a value of 15 Tg N yr−1 for preindustrial inorganic nitrogen inputs. The situation for organic nitrogen inputs is less clear. Cornell et al.  give a range of 24–84 Tg N yr−1, but assumed that most of this flux is of anthropogenic origin. A conservative estimate of 10 Tg N yr−1 for organic sources provides a total preindustrial atmospheric flux of 25 Tg N yr−1.
 Inputs from terrestrial sources include dissolved inorganic nitrogen (DIN) in rivers, streams, and groundwaters, as well as particulate and dissolved organic nitrogen (DON) carried by rivers. Codispoti and Christensen  give a value of 25 Tg N yr−1 for preindustrial riverine sources. More recently, Seitzinger and Kroeze  estimated DIN sources of 5 Tg yr−1 for the world's oceans. The particulate nitrogen loading to the oceans may be significantly more than this, roughly 20 Tg N yr−1 [Duce et al., 1991; Galloway et al., 1995]. Some of this may be lost by denitrification or sedimentation in estuaries [Nixon et al., 1996; Seitzinger and Nixon, 1985]. It has been speculated that groundwater fluxes of fresh water may equal riverine sources [Moore, 1996]. Little is known, however, about the fate of nitrogen carried into coastal marine systems and estuaries. There is evidence that groundwaters can be an important nitrogen source to estuaries [Holmes et al., 2000; McClelland and Valiela, 1998; McClelland et al., 1997; Valiela et al., 2000]. Some may be lost due to denitrification in suboxic sediments, and in terrestrial riparian systems suboxic regions have been shown to be very effective traps for groundwater nitrogen [Brandes et al., 1996]. However, groundwater studies suffer the same anthropogenic pollution problems as do riverine studies and to date no study has attempted to constrain this term. Given the uncertainties, we choose to use the 25 Tg yr−1 term for terrestrial fixed N sources, but we are cognizant that this term may need revision in the future.
Table 2. δ15N of Dissolved Nitrate in Amazon Main Stem and Tributary Riversa
δ15N Versus Air
Number of Determinations
Samples were acidified onsite and filtered (precombusted GFF) prior to shipment to the US. Nitrate isotopic values were determined using the method of Velinsky et al. .
Madre de Dios
3.05 ± 0.5
n = 3
4.76 ± 0.3
n = 2
Beni at Arriba
4.43 ± 0.1
n = 3
Beni at Riberalta
5.19 ± 0.1
n = 2
Beni at Sapecho
n = 1
Beni at Rurrenabaque
n = 1
2.88 ± 0.3
n = 2
1.16 ± 0.1
n = 2
Yara at Caranaui
n = 1
Solemoies at Marchaneria
2.2 ± 0.1
n = 2
Amazon at Manaus
4.1 ± 0.3
n = 6
 The particulate isotopic flux is more difficult to determine. Numerous studies have shown that particulate organic carbon carried by rivers is quickly remineralized before reaching the coastal oceans, and thus it is reasonable to assume that much of the organic nitrogen associated with this material will also be remineralized. However, in the absence of suboxic conditions this nitrogen will ultimately be released as nitrate, perhaps with little fractionation. Hedges et al. , summarizing data collected from a 1800-km-reach of the Amazon River, reported values between 2 and 5‰, with a mean of 3.5‰ for DON, fine particulate, and coarse particulate organic material. This value agrees well with our estimates for inorganic nitrogen in the same river system. Finally, groundwater nitrogen sources are poorly constrained. In general, estimates of “pristine” or “source” nitrate isotopic values in these systems are relatively enriched, from 2 to 10‰ [Burg and Heaton, 1998; Feast et al., 1998; Harrington et al., 1998; Kellman and Hillaire-Marcel, 1998; Wassenaar, 1995].
 Overall, we choose a moderately enriched value, 4 ± 4‰, for preanthropogenic terrestrial nitrogen sources. The bulk of the studies listed previously suggests such a value is reasonable. Terrestrial nitrogen sources include both N-fixation and nitrogen released from bedrock and soil weathering [Holloway and Dahlgren, 1999; Holloway et al., 1998]. The nitrogen released from rocks should have a value between 0 and 6‰, similar to marine sediments and other crustal rocks from which they are derived. Finally, soil denitrification and NH3 losses will tend to enrich groundwater nitrogen pools above these end-members.
2.1.3. Nitrogen Fixation
18.104.22.168. Mass Estimate
 The amount of nitrogen fixed by diazotrophs remains one of the more contentious subjects in the marine biogeochemistry community. Arguments based upon water column N:P relationships have been made [Gruber and Sarmiento, 1997] to support a nitrogen fixation term in the range of 110 Tg N yr−1, yet direct measurements of fixation have tended toward smaller numbers than this [Capone and Carpenter, 1982; Carpenter and Romans, 1991]. However, recent discoveries of additional species capable of nitrogen fixation [Carpenter et al., 1999; Zehr and Capone, 1996; Zehr et al., 2000, 2001] may increase overall marine nitrogen fixation rates. We will examine nitrogen fixation in light of the global isotopic nitrogen balance to be constructed below.
22.214.171.124. Isotopic Composition
 The process of nitrogen fixation is thought to produce organic nitrogen with relatively little fractionation. Direct measurements of Trichodesmium species in situ [Capone et al., 1997] and other diazotrophs in culture [Hoering and Ford, 1959; Minagawa and Wada, 1986] have given a range of −1.5 to 0‰. This represents a small offset from dissolved N2, which has a value of 0.6‰ [Emerson et al., 1991, 1999]. However, one published report of Trichodesmium species grown in culture has indicated much lower numbers (−3 to −2‰, [Carpenter et al., 1997]). One possibility is that field isotopic measurements may be influenced by other nitrogen sources, such as uptake of nitrate. Another unknown possibility is the possible production and release of dissolved organic substances by diazotrophs [Bronk and Ward, 2000], about which little is known isotopically. If this material is labile, it could represent an additional source of fixed nitrogen from these organisms. Given the available information, we will assume that nitrogen fixation represents a nitrogen source to the sea of −1‰, with a relatively small uncertainty of ±1‰.
Brandes et al. , Altabet et al. [1999b], and Voss et al.  have reported an average value for isotopic fractionation during water column denitrification, ε, of 25–27‰. This value was based upon isopycnal mixing models for both the Arabian Sea and ETNP suboxic regions. Given the similarity of data from the ETSP [Liu, 1979], it appears that water column denitrification exhibits a globally consistent fractionation factor. This value also agrees well with a recent laboratory estimate of fractionation by marine denitrifiers [Barford et al., 1999], but is slightly greater than older estimates [Delwiche and Steyn, 1970]. The isotopic value of nitrogen lost within the denitrifying water column is a function of the source nitrate isotopic value; in the Arabian Sea the apparent source waters are slightly lighter than those found in the Pacific (5‰ versus 6‰, [Brandes et al., 1998]). Although suboxic basins (e.g., Cariaco Trench, Black Sea) can be locally important to nitrogen budgets, the amount of nitrogen lost in these regions is small compared to the large marine suboxic regions. The integrated isotopic value of fixed nitrogen lost via water column denitrification can be calculated using a closed system (Rayleigh) assumption [Wada and Hattori, 1991] or by numerically integrating losses within advection-diffusion (A–D) models [Brandes et al., 1998]. Depending upon the method, the isotopic value of fixed nitrogen lost is from −19 to −21‰. Closed system calculations always result in slightly less apparent fractionation than those done for open (A–D) systems. An open system allows some “light” nitrate into the denitrification zone; thus in order to duplicate the isotopic enrichments observed in the data, the model fractionation factor must be larger. However, the differences are small in comparison to the uncertainty in calculating ε. Thus we adopt a value of −20 ± 3‰ for this process.
2.2.2. Sediment Denitrification
126.96.36.199. Mass Estimate
 As with nitrogen fixation, the rate of marine sediment denitrification remains one of the important unresolved issues in marine nitrogen budgets. Early estimates [Codispoti and Christensen, 1985] indicating rates less than 100 Tg N yr−1 have been used by recent investigators to support the notion of a balanced fixed nitrogen budget [Gruber and Sarmiento, 1997]. However, other researchers have revised these estimates up to nearly three times this level in recent years [Codispoti, 1995; Devol et al., 1997; Middelburg et al., 1996]. These higher rates may be partly due to anthropogenic influences [Codispoti et al., 2001], although much of the discrepancy in estimates may be due to denitrification in slope and deep-sea sediments, which should be less influenced by terrestrial and atmospheric anthropogenically influenced sources. We will use the isotopic budget in this paper to constrain this rate (see section 4 subsequently).
188.8.131.52. Isotopic Composition
 Although one report has been published on the isotopic fractionation during denitrification in sediments [Brandes and Devol, 1997], the data are from an estuarine location not necessarily typical of ocean margin sediments. To expand the database of sediment denitrification isotopic values, a series of measurements were made along the Washington and Mexican continental margins (Figure 1) using an in situ tripod incubation system [Brandes and Devol, 1997; Devol, 1987]. Briefly, 24–48 hour sediment incubations were conducted with the lander in “coring” mode, where a small-scale box core attachment was included on each incubation chamber to allow retrieval of overlying waters and intact sediments. Overlying waters were subsampled, frozen, and returned for later isotopic and compositional analysis. Isotopic analysis for δ15N-NO3− was performed using the method of Velinsky et al. . Further details of the methods used are described by Brandes  and Brandes and Devol . Sediments in this region are overlain by a strong oxygen minimum, with O2 concentrations off Mexico decreasing to below 10 μM l−1 in the depth interval between 150 and 800 m. The fraction of nitrate remaining versus change in isotopic concentration is shown in Figure 2. All data plot with fractionations between 0 and 3‰ fractionation, with an average value of 1.5‰. This is slightly larger than that measured for estuarine Puget Sound sediments [Brandes and Devol, 1997], but it is still much lower than that noted for water column denitrification. Given the average oceanic 15N-NO3− value of 5‰ [Liu and Kaplan, 1989; Liu et al., 1996; Sigman et al., 1999b, 1997b], sediment denitrification would therefore remove nitrogen at an isotopic value of 3.5 ± 2‰.
 One significant concern is that all the available data are from relatively organic-rich sediments with thin oxic zones. It may be possible that the fractionation factors for this process are greater in low carbon shelf sediments (as found in some US Atlantic Margin sediments, [Jahnke and Jahnke, 2000] or slope and deep-sea sediments. However, a recent study [Sigman et al., 2001] of the isotopic composition of pore water nitrate in a deep-sea sediment also indicated a small degree of fractionation during denitrification. We do not currently know the reasons for the lack of isotopic fractionation in the sediments studied so far. Possible explanations include the presence of steep concentration gradients that reduce the effective fractionation factor experienced by waters in contact with sediments [Bender, 1990; Brandes and Devol, 1997], abiotic or inorganic (e.g., Mn oxidation) reactions [Luther et al., 1997] with intrinsically small fractionations, or some other unknown process. However, it will be shown subsequently that sedimentary denitrification must proceed with little fractionation in order for the marine isotopic budget to approach a reasonable balance.
2.2.3. Sediment Burial
184.108.40.206. Mass Estimate
Gruber and Sarmiento  gave an estimate of 25 Tg N yr−1 for sediment nitrogen burial. This value agrees well with carbon budget estimates of Hedges and Keil , who estimated a global carbon burial rate of 160 Tg C yr−1. This translates into a putative nitrogen burial rate of 25 Tg N yr−1, given a C:N ratio of 7.
220.127.116.11. Isotopic Composition
 As no global database exists for sediment δ15N values, a precise estimate of this term is unavailable. In continental shelf and slope regions where mixed-layer nitrogen is assumed to be completely removed by primary producers, the δ15N of buried organic nitrogen is close to that of upwelled marine nitrate [Altabet et al., 1995, 1999a, 1999b; Ganeshram et al., 1995, 2000; Giraud et al., 2000]. Little diagenetic fractionation during burial has been found for shelf and slope sediments [Altabet et al., 1999b; Freudenthal et al., 2001]. As continental margin environments account for >90% of the total organic material permanently buried in the sea [Hedges and Keil, 1995], the average for all the organic material buried should be similar to average oceanic nitrate. The situation for deep-sea sediments is more complicated. There exists a significant uptake fractionation by photosynthetic organisms [Altabet, 2001; Sigman et al., 1999b; Waser et al., 1998; Wu et al., 1997]. This is not a concern if all upwelled nitrogen is eventually removed by phytoplankton, but there are exceptions to this, especially the southern Pacific HNLC and Antarctic regions [Altabet and Francois, 1994; Sigman et al., 1997b] (this may also be the case for some continental shelf and slope environments as well). Downwelling of waters containing appreciable amounts of nitrate will tend to remove isotopically enriched nitrate from the surface, leading to isotopically depleted nitrogen in sediments. Counteracting this possibility is the observation that low organic content marine sediments can exhibit significant diagenetic shifts (up to +4‰) from source-sinking organic material [Freudenthal et al., 2001; Holmes et al., 1996]. To balance this, overall material buried in deep-sea sediments may be slightly enriched over source nitrate. Thus the isotopic value for organic matter burial is set slightly higher than the average isotopic composition of source nitrate, 6 ± 2‰.
2.3. Existing Nitrogen Pools
 As discussed previously, the average isotopic composition of nitrate in the sea is about 5‰, although this value varies from about 4‰ in the North Atlantic to 6‰ in the North Pacific [Liu and Kaplan, 1989; Liu et al., 1996; Sigman et al., 1997b]. There are a few anomalous basins, such as the Mediterranean (−1‰, [Sachs and Repeta, 1999]), and the overall database of nitrate δ15N measurements is very limited; however, the main ocean basins are well represented. Dissolved organic forms, which represent 3–5% of total marine-fixed nitrogen, are not well represented. Benner et al.  report values ranging from 7.2 to 14.6‰ for ultrafiltered dissolved organic matter (DOM) in the equatorial Pacific. However, Abell et al.  reported very low values (0.5‰) for total DOM collected from the same region. There is some debate over how rapidly this material cycles in the oceans [Abell et al., 2000; McCarthy et al., 1997; McCarthy et al., 1998], but over the residence time of fixed nitrogen (∼1500 years) the DON pool may play a role. Given the inconsistencies in reported DON isotopic values, we choose to neglect this term in our isotopic budget, although this will need to be revised in the future, especially if the results of Benner et al.  are typical.
3. Results: A Marine Stable Isotopic Budget
 Our objective in modeling the marine-fixed nitrogen stable isotope budget is twofold. First, can the isotopic budget constrain overall source and sink rates to determine if marine-fixed nitrogen inventories have been at steady state during the Holocene? Failing that, can the isotopic budget constrain either the source or sink terms so that conclusions can be made about the overall rates of nitrogen cycling in the sea? In order to answer these questions flux-weighted source and sink model budgets were constructed, using the values given in Table 1. Because there are eight unknowns (source and output 15N/14N ratios, riverine, atmospheric and nitrogen fixation sources, sedimentary denitrification, water column denitrification, and organic burial sinks) and only two equations, six of the terms need to be fixed. The best known of these terms are sediment burial, water column denitrification, riverine sources, and atmospheric sources. With these values fixed, the flux-weighted budget model then examines the effects of varying the nitrogen fixation and sediment denitrification terms. The model equations are as follows:
 The effects of varying the rate of nitrogen fixation were examined by generating an artificial “data set” [Press et al., 1992] consisting of 10,000 isotopic values for each source term (e.g., (15N/14N)rivers), each with a standard deviation as given in Table 1. Standard deviations were assumed to be independent. Mass flux values for riverine and atmospheric sources were held constant. Using these values in equation (1), a corresponding set of estimates for the (15N/14N)(sources) was compiled, the average taken, and the standard deviation computed. This estimate was done for values of nitrogen fixation between 0 and 400 Tg N yr−1. The effect of varying nitrogen fixation on the isotopic value of marine-fixed nitrogen sources is shown in Figure 3. The average isotopic value remains nearly constant, between 0 and −1, across the range of fixation rates. Riverine and atmospheric sources tend to offset each other, giving a net 0‰ for that contribution, although the large amount of variability in each term is manifested in a high variance at low fixation rates. Because the isotopic value for nitrogen fixation is close to the average of river + atmospheric sources, the overall isotopic value for sources is insensitive to changes in this term. This prevents the use of the isotopic budget as a constraint upon nitrogen fixation, but it does have the benefit of establishing a relatively narrow range of possible source isotopic values. This is especially true for fixation rates greater than 100 Tg N yr−1, where the budget is dominated by the relatively well isotopically constrained nitrogen fixation term.
 The effect of varying estimates for sedimentary denitrification on the isotopic composition of fixed nitrogen sinks was investigated using the flux-weighted isotopic budget given by equation (2). As with the source model, 10,000 values for each sink term (e.g., (15N/14N)burial) each with a standard deviation as given in Table 1. Mass flux values for water column denitrification and sediment organic nitrogen burial were held constant. Using these values in equation (2), a corresponding set of estimates for the (15N/14N)(sinks) was compiled, the average taken, and the standard deviation computed as a function of sedimentary denitrification rate. Model results are shown in Figure 4. In contrast to the source model where little isotopic effect was seen in shifting the nitrogen fixation term, average isotopic values for sinks were very sensitive to changes in sedimentary denitrification. Model results ranged from −12.75‰ for the case of no sedimentary denitrification, to +0.25‰ for 400 Tg N yr−1 sedimentary denitrification (Figure 4). Because fixed N losses due to sedimentary denitrification are not strongly enriched in 15N, a large amount of sedimentary denitrification is necessary to balance water column denitrification and bring overall isotopic values of the losses into the range of sources.
 If one assumes a constant Holocene value of 5‰ for marine-dissolved NO3−, then the isotopic value of combined sinks must lie in the range covered by sources, 0 to −1‰. Given the best estimates of water column denitrification and sediment burial, this requires a minimum of 250–300 Tg N yr−1 in sedimentary denitrification, and a total marine-fixed nitrogen loss rate of nearly 400 Tg N yr−1 (Figure 5). This rate is in excess of most previously published estimates for global sedimentary denitrification rates [Codispoti and Christensen, 1985; Devol et al., 1997]; however, the database for sedimentary denitrification rates is relatively small and many estimates made prior to the 1990s are marred by measurement difficulties such as the neglect of coupled nitrification-denitrification reactions. Middelburg et al.  suggested a total rate of 220–280 Tg N yr−1 based upon a global sediment denitrification model. They suggested that much of the global rate was due to denitrification within slope and deep-sea sediments. What is critical to note about both Middelburg's estimate and ours is that these are long term, not anthropogenically influenced, rates.
 We have noted above that the evidence for a small fractionation factor for sedimentary denitrification is based upon a small data set from high carbon content shelf sediments. However, the isotopic budget provides some justification for this. If the fractionation factor for sedimentary denitrification were as much as one-fifth of the water column factor (i.e., 5‰), the source and output budgets could not be balanced without making very unreasonable assumptions (e.g., N2 fixation having a very large depleted isotopic value). Water column denitrification has a very large fractionation factor, removing relatively large amounts of isotopically depleted fixed N. This term must be counteracted by a loss of enriched N. Sediment burial removes a small amount, but sedimentary denitrification must act as a sink of enriched fixed N, and thus must have a small (<5‰) fractionation factor given average seawater fixed N isotopic values. In other words, if both sedimentary and water column denitrification exhibited the same fractionation factor, seawater nitrate would have to increase to greater than 20‰ in order for the isotopic value of sinks to equal that of sources (−0.5‰). The smaller the apparent sedimentary denitrification factor, the lesser is the sedimentary denitrification that takes place to balance the isotopic budget. If sedimentary denitrification were to have no isotopic fractionation (0‰), then our estimate of its global rate would fall to around 240 Tg N yr−1, still larger than traditional estimates of sedimentary denitrification.
 Another factor to consider is the possible enhancement/intensification of marine suboxic regions due to anthropogenic nitrogen sources [Codispoti et al., 2001; Naqvi and Jayakumar, 2000; Rabalais et al., 1996, 2001]. The result would be to inflate modern estimates relative to those in the recent past and lead to erroneously high sedimentary denitrification estimates. The first wide-scale measurements of Eastern Tropical Pacific (ETP) denitrification rates were done in the early 1970s [Cline and Kaplan, 1975; Codispoti and Richards, 1976]. It can be argued that this region is an unlikely candidate for anthropogenically driven intensification, yet these studies agreed upon a large value of 50 Tg N yr−1 for this region alone. This value has recently been upheld using more comprehensive measurements of circulation and water residence times [Deutsch et al., 2001]. We have chosen a value of 75 Tg N yr−1 for all marine suboxic regions, a value that includes the ETNP and the ETSP and Arabian Seas. Recent estimates for these regions have placed the global rate at greater than 80 Tg N yr−1 and there is some evidence for a rate as much as twice this if one includes nitrogen sources other than nitrate [Codispoti et al., 2001]. Sediment δ15N measurements from these regions [Altabet et al., 1995, 1999a; Ganeshram and Pedersen, 1998; Ganeshram et al., 1995, 2000] show no significant changes in Holocene denitrification intensities. Thus water column denitrification rates and their corresponding losses of depleted fixed nitrogen are unlikely to have been less during this time period (although short-term fluctuations may have occurred), and nor are the fractionation factors likely to have shifted. Brandes et al.  revisited the isotopic data of Cline and Kaplan  and found no difference in fractionation factors for the region over that time span. The fractionation factor appears to be constant over the range of denitrification rates found in the Arabian Sea and ETNP regions [Brandes et al., 1998]. Even if we postulate a water column denitrification rate of 50 Tg N yr−1, over 180 Tg N yr−1 of sedimentary denitrification is required for average fixed nitrogen sinks to match the value of sources (Figure 5). A doubling of sediment burial rates, which would remove isotopically enriched (relative to nitrogen fixation) nitrogen has an even smaller effect, with a requirement of at least 250 Tg N yr−1 (Figure 5). These estimates are predicated upon a nearly balanced fixed N budget, if nitrogen fixation rates are less than a larger sediment, denitrification rates are required to balance the isotopic budget.
4. Discussion: Consequences of Long-Term 400 Tg N yr−1 Marine-Fixed Nitrogen Losses
 A balanced marine nitrogen isotopic budget does not require a balanced fixed nitrogen budget. As the source isotopic budget is insensitive to changes in nitrogen fixation, it is possible to have a total marine nitrogen budget that is wildly out of balance and yet still maintain a balanced isotopic budget. However, there are several practical implications of a nearly 400-Tg yr−1 loss rate. Given our sediment denitrification estimate of 280 Tg N yr−1, if nitrogen fixation rates were only 100 Tg N yr−1, then the combined budget would be out of balance by over −200 Tg N yr−1. Such a loss rate would deplete the ocean of fixed nitrogen (at present inventory) within 3000 years. Thus the ocean at the glacial-interglacial transition would have had to contain three times the nitrate concentrations that it does today. Losses of this extent would also generate about 100 × 1012 mol yr−1 (0.1 Gt yr−1) CO2 from lost productivity (assuming near-complete surface nitrogen usage by primary producers during the Holocene). This amount of carbon release over a long time period poses difficulties as preindustrial Holocene CO2 levels did not rise significantly [Falkowski, 1997; Falkowski et al., 1998; Gruber and Sarmiento, 1997]. A possible alternative might be shifts in C:N ratios of primary producers, as has been speculated by some researchers [Archer et al., 2000]. If early Holocene phytoplankton sequestered more nitrogen-rich compounds (C:N ratios decreased) than their modern counterparts, then early Holocene CO2 fixation rates might have been comparable to modern values in spite of major changes in fixed nitrogen availability. It is possible for phytoplankton to vary their C:N ratios in the presence of excess nitrogen [Banse, 1994; Goldman and McCarthy, 1978; Goldman et al., 1979; Laws and Bannister, 1980; McCarthy and Goldman, 1979]. This process, which occurs primarily under low-light conditions, can lower C:N values to 5 from the Redfield 6.6 value, and under some exceptional circumstances C:N ratios can drop to near zero [Fraga et al., 1998; Rios et al., 1998]. It is an open question whether or not this high-nitrogen content material will survive remineralization to permanent burial, and there have been no reports of significant decreases in sediment C:N ratios during the Holocene. If C:N ratios are assumed to be relatively constant, then one has to assume either a decline in global nitrogen uptake percentages (not supported by sediment 15N values, [Sigman et al., 1997a] or shifts in carbon burial efficiencies (again not supported by the sediment record). Finally, significantly higher fixed nitrogen inventories are at odds with studies suggesting increased nitrogen uptake in surface oceans during the last interglacial [Sigman and Boyle, 2000], when presumably these inventories would have been at their highest [Ganeshram et al., 2000].
 The alternative position is to assume that fixation rates are much greater than those presently assumed and that the Holocene nitrogen budget is close to being balanced. This implies fixation rates 2–3 times the currently accepted values. In this case, the difficulties with a shifting marine nitrogen inventory are replaced with difficulties in justifying high fixation rates. The highest such estimates (110 Tg yr−1) are given by Gruber and Sarmiento  and even though direct observations have begun to support such a rate [Capone, 2001; Carpenter and Romans, 1991], a rate double than this is controversial. The discrepancy may be due to under- or insufficient sampling and methodological problems [Lipschultz and Owens, 1996], or very tight coupling between diazotrophs and other phytoplankton species [Carpenter et al., 1999; Villareal, 1994]. Although there are no definitive answers yet, there are some intriguing studies which suggest that existing fixation studies seriously underestimate total fixation. Sachs and Repeta  reported that the δ15N of nitrate in the deep eastern Mediterranean is about −1‰. They estimated that 40% of this pool came from fixation, but given the published δ15N values of diazotrophs [Capone et al., 1998], this number may be an underestimate. Several studies, including those by Montoya et al. , Sigman , and J. A. Brandes and A. H. Devol (unpublished data) have indicated that the north Atlantic is isotopically depleted compared to the southern ocean (<4‰ versus 4.8‰ [Sigman et al., 2000]. Given the small amount of deepwater exchange with the Mediterranean (0.2 Sv, [Kinder and Parrilla, 1987]), it is likely that most is from local diazotroph sources. If one assumes a value of −1‰ for this source, then nitrogen fixation is responsible for roughly one sixth of all nitrogen used by surface producers in the north Atlantic alone. This is not unreasonable as at times the amount of productivity fueled by fixation exceeds that from vertical diffusion of nitrate [Carpenter et al., 1999]. It is more difficult to assess the amount of nitrogen fixation in the other basins, as there is either a paucity of data (South Atlantic, South Pacific and South Indian Oceans) or the presence of suboxic regions adds a second, enriched isotopic signal. Enriched nitrogen from the ETNP suboxic region can be observed in the western Pacific [Liu et al., 1996] and as far north as California [Emmer and Thunell, 2000; Liu and Kaplan, 1989]. Thus the lack of depleted nitrogen in Pacific and Indian Ocean waters may not indicate a lower amount of nitrogen fixation, but instead indicate the presence of this additional denitrification signal that masks the fixation signal.
 Recent work has made it apparent that the marine-fixed nitrogen cycle is very dynamic, and has been so for some time. This work and that of Middelburg et al.  suggest a very high rate of sediment denitrification, in the order of 280 Tg N yr−1. It is important to note that these two estimates are based upon entirely different methodologies. Middelburg et al.  based their value upon a bottom-up mechanistic model of sedimentary denitrification which was scaled up using empirical relationships between global carbon fluxes and ocean bathymetry. Our estimate is a “top-down” estimate using isotopic and mass balances. Given that estimates for terrestrial and atmospheric nitrogen sources may have been overestimated due to anthropogenic inputs [Seitzinger and Kroeze, 1998], this leads to a nitrogen budget that is dominated by two terms: nitrogen fixation and denitrification. The exact level of nitrogen fixation is as yet unknown, but we argue that it must be far greater than the ∼100-Tg N yr−1 rates discussed in the literature. Our Holocene isotope budget implies a marine nitrogen residence time of 1600 years (sum of loss rates/reservoir size). Given this rapid turnover time, slight changes in source and sink rates can have serious effects upon global productivity and CO2 levels [Ganeshram et al., 2000]. Most of the fixed nitrogen budget terms are centered upon the upper ocean or at interfaces (land/sea, sediment/water, air/sea). Thus changes may take place even more quickly than implied by the overall residence time. Although short-term shifts are likely [Ganeshram et al., 2000; Pride et al., 1999], the existence of a long-term imbalance is unlikely given the lack of evidence in the sediment record. Finally, the disparate nature of the isotopic signatures implied by the three major terms (nitrogen fixation, sediment denitrification, and water column denitrification) suggests that more detailed studies of present and past marine 15N signatures could yield useful information on relative shifts in these three terms. For example, examining a detailed suite of nitrate isotopic values from the different ocean basins could provide an estimate of the relative contributions of N2 fixation and sediment denitrification on a basin-wide scale. This possibility is suggested by the observed differences in nitrate δ15N values between Atlantic, Indian and Pacific Ocean deep waters, as noted earlier. Furthermore, given previously reported shifts in water column denitrification rates [Altabet et al., 1995, 1999a; Ganeshram and Pedersen, 1998; Ganeshram et al., 1995, 2000] it is highly likely that Pleistocene marine δ15N nitrate values did not remain constant, as is assumed/implied by most researchers, but may have decreased during glacial interludes to reflect reduced losses of isotopically light fixed nitrogen in suboxic regions. Establishing a global fixed nitrogen isotopic signature during past era may well require examining sediment δ15N signatures from a wide range of locations (especially those away from likely suboxic regions), but could provide a strong constraint upon past global marine-fixed nitrogen budgets.
 We thank Paul Quay for the use of his stable isotope laboratory facilities. David Wilbur provided assistance in operating the mass spectrometer and in data analysis. Michael McClain provided the Amazon River and Amazon Tributary samples for δ15N analysis. Lou Codispoti and Jack J. Middelburg provided comments that greatly improved this submission. This work was supported by NSF grants OCE 91-16275 and OCE 94-16626 (A.H.D) and by the Department of Defense NDSEG fellowship program (J.A.B.).