Modern seawater profiles of oxygen, nitrate deficit, and nitrogen isotopes reveal the spatial decoupling of summer monsoon-related productivity and denitrification maxima in the Arabian Sea (AS) and raise the possibility that winter monsoon and/or ventilation play a crucial role in modulating denitrification in the northeastern AS, both today and through the past. A new high-resolution 50-ka record of δ15N from the Pakistan margin is compared to five other denitrification records distributed across the AS. This regional comparison unveils the persistence of east-west heterogeneities in denitrification intensity across millennial-scale climate shifts and throughout the Holocene. The oxygen minimum zone (OMZ) experienced east-west swings across Termination I and throughout the Holocene. Probable causes are (1) changes in ventilation due to millennial-scale variations in Antarctic Intermediate Water formation and (2) postglacial reorganization of intermediate circulation in the northeastern AS following sea level rise. Whereas denitrification in the world's OMZs, including the western AS, gradually declined following the deglacial maximum (10−9 ka BP), the northeastern AS record clearly witnesses increasing denitrification from about 8 ka BP. This would have impacted the global Holocene climate through sustained N2O production and marine nitrogen loss.
 The Arabian Sea (AS) hosts one of the largest pools of denitrifying water in the ocean and accounts for at least one third of the loss of marine fixed N [Codispoti and Christensen, 1985]. In the near absence of O2, denitrification converts combined, bioavailable nitrogen to gaseous N2, strongly discriminating against the heavier isotope (15N). Monitoring denitrification intensity in this region is essential given its potential to alter surface fertility and in turn the marine carbon cycle [McElroy, 1983; Ganeshram et al., 1995; Altabet et al., 1995]. In addition, Nitrous oxide (N2O) is a by-product of denitrification and, once released to the atmosphere, contributes to the radiative forcing [Codispoti and Christensen, 1985].
 Given the regional decoupling between denitrification and productivity in the basin today, we propose to critically explore the impact of ventilation on the intensity and distribution of the denitrifying zone through millennial-scale climate shifts and across the Holocene. One crucial limitation in assigning clear causes to denitrification changes in the OMZs lies in the lack of unequivocal proxies for either oxidant demand or O2 supply. In an attempt to compensate for this limitation, we propose to compare the δ15N signals of six sediment cores distributed from the southwestern to the northeastern AS with recently documented changes in intermediate/thermocline circulation [Kuhnt et al., 2004; Pahnke and Zahn, 2005].
2. Nutrient and Oxygen Inventories in the Modern AS
2.1. Primary Productivity and Denitrification
 Latitudinal shifts in the position of the Inter-Tropical Convergence Zone over the Indian Ocean region results in drastic seasonal variation in atmospheric circulation and hence in upwelling intensity [Wyrtki, 1973]. Figure 1 shows the contrast in primary productivity between summer and winter in the surface AS. Primary productivity peaks off Somali and Oman during summer monsoon and in the northern AS during winter. Intermonsoon periods exhibit low productivity conditions. Strong seasonality in particle fluxes and export production result from these alternating atmospheric and surface ocean conditions [Nair et al., 1989; Haake et al., 1993; Ramaswamy and Nair, 1994]. Sediments traps record a bimodal distribution in biogenic particle fluxes with very high fluxes in summer and moderately high fluxes during winter time [Rixen et al., 1996; Honjo et al., 1999] (Figure 1). From 58 to 78% of the particle flux to the seafloor occur during the summer monsoon with the highest productivity observed in the southwest part of the basin [Haake et al., 1993].
Figure 2 shows the oxygen distribution in the AS. This figure highlights the clear west-east decrease in oxygen concentration, opposite to what might be expected given the distribution of primary productivity. Figure 2 also reveals that O2 deficiency at 100–200 m builds up during winter rather than during the highly productive summer monsoon, in agreement with de Sousa et al. . The OMZ is permanent in spite of drastic seasonal changes in productivity [Sarma, 2002]. This stability has been explained by invoking a balance between the biological pump (oxidant demand) and the physical pump (mixing and ventilation) of oxygen around the year [Olson et al., 1993; Sarma, 2002].
 Water column denitrification occurs when O2 concentrations fall below 0.15–0.2 mL/L [Naqvi and Jayakumar, 2000], a requirement which is not met within the entire OMZ but in areas restricted to the northeastern AS, mostly around 250 m water depth (Figure 2). Using hydrostation data from the World Ocean Atlas database (http://iridl.ldeo.columbia.edu/), we calculated nitrate deficit (N*) in the water column according to the formula published by Gruber and Sarmiento  where N* represents the nitrate lost through denitrification. Between 100 and 500 m depth N* is strongly negative in the northeastern AS, and roughly coincides with [O2] minima (Figure 2). However, oxygen content at intermediate depth in the western AS is not low enough to cause year-round denitrification at any depth. N* is less negative here than in the eastern AS and minimum values do not correspond to maxima in δ15N-NO3 (Figure 2). This could indicate that heavy nitrate found in the western AS is not fully produced in situ but partly brought from the northeastern AS in winter, when the thermocline circulation reverses [You, 1998]. Water column denitrification causes the δ15N of nitrate to be significantly higher than the mean ocean nitrate value of 4.5–6‰ [Sigman et al., 1997]. As a consequence δ15N-NO3 at 300 m depth reaches 18‰ in the northeastern AS and 12‰ in the western AS [Brandes et al., 1998; Naqvi et al., 1998; Altabet et al., 1999a], further highlighting the clear regional heterogeneity.
2.2. Circulation and Oxygen Supply
 The observed spatial segregation between productivity and denitrification maxima is caused by the circulation of thermocline and intermediate waters in the AS and the initially very low oxygen content of water entering the basin [Olson et al., 1993]. The main water sources at 200–1000 m originate both from the northwest (the highly saline Red Seawater (RSW) and the Persian Gulf Water (PGW)) and from the south (the Indian Central Water (ICW)) [Wyrtki, 1973; Swallow, 1984; You, 1998] (Figure 3). The latter is a mixture of aged Antarctic Intermediate Water (AAIW) and Banda Seawater (BSW) coming from the Indonesian throughflow [Sharma, 1972; Sharma et al., 1978; You, 1998], and enters the AS across the southwestern boundary during the summer monsoon with the Somali current [You, 1998]. Oxygen concentrations in the northern and southern water masses are equally low but the volume transported across the equator with the southern waters is 1 order of magnitude larger than that contained in the northern source inflow [Swallow, 1984]. Oxygen replenishment occurs mostly along the southern and western boundaries of the AS and hence the northeastern part of the basin is bathed by waters severely impoverished in O2.
3. Material and Method
 Core MD-04 2876 (24°50′57N; 064°00′49E; 33 m length) was retrieved at 828 m water depth on the Pakistan Margin within the present-day OMZ. The core shows alternating sequences of varved/laminated and bioturbated/massive sediments.
 Nitrogen content and isotopic composition in MD-04 2876 were analyzed with a Micromass Mass Spectrometer at the University of Bordeaux on bulk sediments. To ensure comparability with other AS records measured in the School of Geosciences, The University of Edinburgh [Ivanochko et al., 2005], we run a set of 10 sediment samples from MD-04 2876 as well as a an internal (sediment) standard on both Mass Spectrometers. Intercomparability was found to be within ±0.3‰. In total, 10 samples were additionally analyzed for δ13Corg using the same Mass Spectrometer, in order to investigate marine versus terrestrial origin of the organic matter.
 The calendar chronology for core MD04-2876 is based on 15 radiocarbon ages measured by accelerator mass spectrometry on samples composed of hand-picked shells of the planktonic foraminifera Globigerinoides ruber. Further details on the samples in relation to foraminiferal abundance variations, on the correction for reservoir age and on the conversion in terms of calendar ages will be presented elsewhere (E. Bard et al., manuscript in preparation, 2007). The sedimentation rates calculated between each dated interval are 53 ± 11 cm/ka on average.
 The new δ15N record was compared to 5 existing records from the AS (Figure 3): MD76 131 [Ganeshram et al., 2000], SO90-111KL [Suthhof et al., 2001], RC27-14 and RC27-23 [Altabet et al., 2002] (Figure 4e), and 905P from [Ivanochko et al., 2005]. Detailed comparison between the high-resolution records RC-27 14, RC-27 23, 905P and MD-04 2876 was achieved by tuning the existing records to the well-dated MD-04 2876 record. To do so, we used 28 tie points between the MD-04 2876 record and each of the published records. Linear interpolation of the δ15N record was undertaken at 100-year intervals between measured values. The Δδ15N curves, defined as the difference between our new δ15N records and each of the three existing records from western AS, were obtained by subtracting these interpolated records.
4. Millennial-Scale Changes in Denitrification
4.1. A New 50-ka Record of Denitrification Changes in the Northeastern AS
 High-resolution δ15N measurements from Core MD-04 2876 display large-amplitude, abrupt changes that faithfully mirror millennial-scale northern high-latitude climatic changes recorded in Greenland ice core δ18O and Monsoon fluctuations recorded in Hulu cave δ18O (Figure 4). The δ15N values range from around 3 to 8‰ (±0.3‰). The arithmetic mean δ15N value is 5.35‰. The δ13C of organic carbon is consistently around −19‰ which confirms that the organic matter is pristine planktonic material irrespective of the climatic period (Table 1). Sedimentary nitrogen is enriched in 15N during warm periods compared to today's mean ocean nitrate, testifying to the occurrence of intense denitrification. During Stadials and Heinrich events however, the signal systematically drops down to about 3–5‰, evidencing low surface/thermocline δ15N-NO3. These values are lower than the modern average isotopic signal of nitrate and reveal the occurrence of either incomplete N relative utilization or N fixation. Incomplete relative utilization of surface nitrate has been documented to have a very limited imprint on the δ15N signal in the AS [e.g., Schäfer and Ittekkot, 1993]. Furthermore, given that the δ15N signal is overall positively correlated to organic carbon fluxes at core site over the period studied (Figure 5) we assume that the δ15N record does not reflect changes in nitrate relative utilization off Pakistan.
Table 1. Organic Carbon Isotopic Signal in Core MD-04 2876
Depth in Core, cm
Calendar Age, Years BP
δ13C of Organic Matter, ‰
 Biological N fixation develops either in stratified, oligotrophic surface waters or in denitrifying zones owing to relative P enrichment in thermocline waters that favors N fixing biota over other producers [Karl et al., 2002]. Today, fixed N may account for a significant part of surface nitrate in the AS where denitrification is exceptionally intense [Brandes et al., 1998; Deutsch et al., 2007], hence partially cancelling out the isotopic enrichment caused by denitrification. It is, however, questionable whether N fixation was more active in the northeastern AS during Stadials and Heinrich events because (1) denitrification was reduced if not shut down [Reichart et al., 2004] and (2) surface waters were neither oligotrophic nor stratified during cold periods [Reichart et al., 2004]. In addition, Schenau et al.  documented reduced phosphorus regeneration in the northeastern AS during Heinrich 1 and the Younger Dryas, conditions that would hinder N fixation. Deutsch et al.  suggest that N fixation rates are bound to increase with enhanced denitrification. Simple mass balance calculation shows that today, fixed nitrate accounts for 30% of the total N exported to the seafloor in the AS [Brandes et al., 1998]. The same calculation also predicts that the δ15N signal of surface nitrate (to be used by the phytoplankton and exported to the sediment) always increases with the intensity of water column denitrification, even for low denitrification rates such as expected during cold periods. The only way to avoid this conclusion involves applying unreasonably high contributions of fixed N (>50%), i.e., increasing N fixation with decreasing denitrification rate. Hence we propose that although N fixation occurs in the AS and explains the occurrence of N isotopic values <4.5‰ during cold phases, the general relationship remains that the δ15N ratio of surface NO3 decreases with decreasing water column denitrification.
4.2. Regional Comparison of δ15N Records
 While the regional biogeochemical cycles are essentially controlled by the monsoon in the western AS, nitrogen and oxygen budgets in the northeastern AS are also prone to respond to slight ventilation changes at intermediate-thermocline depth, owing to the area being located at the end of the oxygenated water route. To document potential regional heterogeneity in the denitrification response within the AS through the past, we compare the δ15N signals of 6 cores (Figure 3) distributed across the AS.
 Although the δ15N records are comparable in terms of overall structure over the period studied (Figure 4), significant differences in arithmetic mean δ15N values are observed between core sites, with a maximal difference of 2.3‰ between MD76 131 (Indian Margin [Ganeshram et al., 2000]) and MD-04 2876 (Pakistan Margin). Gaye-Haake et al.  demonstrated that the diagenetic decay of organic compounds at the seafloor causes the δ15N of surface sediments to be 2–3‰ higher than the overlying deep trap values in the AS, with the exception of the northeastern AS (where Core MD 02-2876 has been retrieved) which shows no significant enrichment in 15N. This study further confirms the impact of export and burial rates on diagenetic N isotope fractionation [Lehmann et al., 2002; Brummer et al., 2002] and implicates that for a given initial δ15N signal carried to the seafloor, the recorded isotopic signal will be substantially higher (by 2–3‰) in areas characterized by low sedimentation rates than for those with rapidly deposited sediments. Accordingly, negligible isotopic effect during settling in the water column has been reported in high export fluxes, low oxygen environments [Altabet et al., 1999b; Thunell et al., 2004].
Figure 6 illustrates the negative correlation between arithmetic mean δ15N values and mean accumulation rates calculated for the six cores studied. The difference in the average δ15N values between cores depends on the rate with which sediments are deposited and buried at each specific site, and is independent of local denitrification intensity. Moreover, radiocarbon dating on cores MD-04 2876 (Figure 4), 905 [Ivanochko et al., 2005] and MD76 131 (T. Ivanochko, personal communication, 2006) reveal nearly constant sedimentation rates throughout these cores, including across Stadial/Interstadial transitions. We are therefore confident that millennial-scale δ15N variations from a given core are not caused by diagenetic effects attributable to temporal changes in sedimentation rate.
4.3. Spatial and Temporal Heterogeneity in Denitrification
 Superimposed on the clear pattern of basin-wide changes in denitrification rate over millennial-scale in the AS, we identify several differing trends of interest between the six records (Figure 4). The most conspicuous discrepancies between the records appear during Heinrich events 1, the Younger Dryas and the Holocene. The Holocene trend clearly decreases toward present in the Somalia Margin record, while it constantly increases in Core MD-04 2876. The Oman Margin cores exhibit an intermediate pattern. Further comparison is prevented by the absence of the late Holocene in cores MD76 131 and SO90-111KL. Sedimentary δ15N has also been found to increase over the Holocene until 2 ka BP in other records from the Indian margin [Banakar et al., 2005; R. S. Ganeshram, unpublished data, 2004], corroborating the observed intensification of denitrification in the north and eastern AS since the early Holocene.
 In order to better quantify the differences between western AS δ15N records (i.e., Somalia, Oman margin sites) and the northeastern AS record, we calculate the Δδ15N by subtracting the 905P, RC17-23, and RC17-14 records from the MD-04 2876 record (Figure 7). The three calculated curves show strongly consistent variations, both in timing and amplitude of changes. The overall amplitude of the Δδ15N variations is >4.5‰ in each case, comparable to the variation within the δ15N records themselves and much greater than both the long-term trends discussed above and the uncertainties associated with δ15N measurements. Over the past 50 ka, these changes occur at millennial timescales (Figure 7).
 Negative Δδ15N values indicate a less positive δ15N signal in the northeastern AS core relative to the western sites. More positive anomalies are found during Interstadials and the Late Holocene and more negative anomalies during Stadials and Heinrich events (and particularly the cold spells surrounding Termination 1). We interpret these variations as arising owing to east to west shifts of the OMZ from warm to cold periods, i.e., a collapse of the OMZ in the northeastern AS and the existence of small and probably seasonal denitrification areas confined to the western AS during Stadials (Figure 8). We propose that these swings were governed principally by ventilation changes in the AS.
5. Evidence for Circulation and Ventilation Changes
5.1. Millennial-Scale Changes in Circulation
 Recent studies document invigorated winter monsoon conditions during cold spells in the eastern AS [e.g., Reichart et al., 2004; Higginson et al., 2004]. Greater oxidant demand due to higher productivity related to the winter monsoon in the northeastern AS would result in shifting the denitrification maximum to the northeast in these periods. However, the Δδ15N index shows the opposite trend, pointing to another scenario.
Pahnke and Zahn  demonstrate that the ocean circulation was arguably as reactive as the atmospheric circulation to abrupt climatic changes and adjusted rapidly to perturbations in North Atlantic Deep Water (NADW) formation. According to the authors, the conversion of AAIW and Subantarctic Mode Water in the Southern Ocean increased substantially during Heinrich events (and perhaps more generally during Stadials) and in turn increased the O2 inventory in the ventilated thermocline and intermediate depth ocean. Such a northward expansion of the AAIW following perturbation of the Thermohaline Circulation (typical of Heinrich events) has been previously modeled and found to influence O2 inventory at low latitudes [Schmittner et al., 2007] and particularly in the AS [Schulte et al., 1999]. Likewise, the influence of the AAIW on the O2 budget in the modern AS is well known [e.g., Sharma, 1972; Olson et al., 1993].
 Comparison between the Δδ15N curves and the δ13C signal of benthic foraminifera from a Southern Ocean core [Pahnke and Zahn, 2005] reveals that northeast to southwest swings of the OMZ (negative Δδ15N anomalies) are concurrent with episodes of increased AAIW formation (Figure 7). Today, the AAIW spreads northward in the Indian Ocean up to 5°N where it feeds the ICW [You, 1998]. This meridional progression of oxygen-rich, southern water masses to the AS is hampered owing to (1) the low-salinity front created along the equator by the zonal BSW flow (<500 m water depth) and (2) the presence of very saline waters from the marginal seas north of the equator (Figure 8). During MIS 2 and probably MIS 3, however, both the inflows from the marginal seas and the BSW were strongly reduced owing to sea level change [Kuhnt et al., 2004], favoring northward circulation of southern waters in the AS and hence a greater contribution of the AAIW to the AS O2 inventory. The influence of changing AAIW flux on the AS is indeed clearly seen in the pre-Holocene Δδ15N curves whereas the observed correlation weakens after Termination I (Figure 7).
5.2. Holocene Trend
 The most prominent regional heterogeneity revealed by this multisite comparison occurs during the Holocene (Figures 4 and 9): Denitrification seems to intensify in the northeastern AS while it declines in the western AS. The continuous rise observed in our new δ15N record from around 8 ka BP contradicts evidence reported elsewhere [Deutsch et al., 2004] that globally, the world's OMZ gradually retreated during the Holocene following the last deglacial apex, around 10 ka ago [Deutsch et al., 2004, and references therein]. Such a retreat is attributed to an overall decline in productivity driven by the reduction of the marine nitrate inventory at the end of Termination I [Deutsch et al., 2004]. Our results reveal that N cycling in the AS did not uniformly respond to the denitrification feedback and that the influence of local/regional productivity and circulation changes has to be considered.
 Reconstructions of paleoproductivity in the western AS [Naidu, 2004; Ivanochko et al., 2005] and summer monsoon strength [Hong et al., 2003] show declining trends from 5 ka to present (Figure 9). On the other hand, winter monsoon and productivity in the eastern AS were reduced during the Holocene until 3 ka BP (A. Singh et al., manuscript in preparation, 2007) and hence cannot explain the Holocene increase in denitrification in the northeastern AS. We therefore invoke a causal role for ventilation changes.
 As proposed earlier (section 5.1), deglacial relative sea level rise caused a profound reorganization of the intermediate/thermocline circulation in the northern AS; i.e., (1) increased inflow of oxygen-poor, saline waters from the southwest at crucial depths (200−300 m) with respect to denitrification in combination with (2) the expansion of a strong frontal system along the equator preventing oxygen-rich southern water masses from reaching the northern Indian Ocean (Figure 7). We suggest that this reorganization lowered the postglacial O2 inventory in the northeastern AS. The Red Sea and the Persian Gulf are separated from the open ocean by shallow and narrow sills (137 and <100 m water depth, respectively). Because of this sill, the Red Sea outflow was reduced by 85% during the last glacial period [Rohling and Zachariasse, 1996] and by around 50% at 10 ka BP compared to today, and increased progressively with sea level rise until it reached its present state at around 6 ka BP, gradually altering the O2 budget in the AS. We propose that the very sluggish circulation of Red Sea and Persian Gulf Waters further prevented the reduced southern source water from ventilating the northern AS and enables a continuous built up of oxygen deficiency in the northeastern part of the basin since the beginning of the Holocene.
6. Implication for Climatic Changes and Nitrous Oxide Emissions
 The global ocean accounts for one half of the total global N2O source, with denitrifying regions such as the AS being the most important marine source [Suntharalingam et al., 2000]. Today, measurements of N2O flux from the surface ocean to the atmosphere show that most of the AS emissions originate from the north and eastern part of the basin [Lal et al., 1996; Bange et al., 2001].
 Atmospheric [N2O] inferred from ice core records spanning the last hundred of ka have experienced large changes at both orbital and suborbital scales [Flückiger et al., 1999, 2004; Sowers et al., 2003]. These changes occurred in phase with denitrification variations in the world's OMZs, including the AS, implying a tight causal link between denitrification intensity, N2O emissions and the Earth's climate through the past. During Dansgaard-Oeschger 9, contrary to preceding Dansgaard-Oeschger events, both denitrification intensity and atmospheric N2O concentration remain low while Greenland ice cores record a prominent temperature increase at the time (see Figure 4 and Flückiger et al. ). This decoupling emphasizes that the emission of greenhouse gas such as N2O and CH4 contributes only partly to suborbital climate changes [cf. Flückiger et al., 2004]. The nitrogen isotopic signature of atmospheric N2O reveals that in spite of these high-amplitude variations, the relative contributions of continental and oceanic sources have been fairly constant over the last glacial with the exception of the end of the YD, when continental source probably increased compared to marine N2O production [Sowers et al., 2003].
 Over the Holocene, denitrification rates in most of the suboxic ocean declined in response to deglacial reduction in marine productivity [Deutsch et al., 2004]. This global decline in the organic carbon rain rate is expected to cause a decrease in N2O production in the well-ventilated Ocean as well. Surprisingly, from 10 ka BP to present, [N2O] increases by 10 ppbv, while δ15N-N2O signal supports unchanged relative contribution of continental and marine sources or a slightly increased ocean source (Figure 9). These results suggest that the marine N2O source increased locally in order to compensate for the overall decline in marine denitrification, and to explain the relative increase in [N2O] from 8 ka BP. The clear denitrification increase observed in the northeastern AS since ∼8 ka BP may have had the potential to promote Nitrous oxide production in the context of global ocean productivity and denitrification slow down. This raises the possibility that the northeastern AS represents an important feedback on the Holocene climate through sustained increase in N2O production and marine N loss.
 The regional comparison of denitrification records from across the AS illustrates the occurrence of regional heterogeneity in denitrification intensity within the basin, both across millennial-scale climate shifts and during the Holocene. We show that the OMZ experienced east-west swings across the abrupt climate changes and through the Holocene epoch. Ventilation changes due to millennial-scale variations in AAIW formation as well as circulation reorganization in the north Indian Ocean following sea level rise are probable causes of the observed regional differences in denitrification.
 This study further strengthens the evidence that AAIW exerts an important control on denitrification in the AS and on the N cycle as a whole [Schulte et al., 1999; Galbraith et al., 2004; Meissner et al., 2005]. Our study is based on very high resolution records and demonstrates both the very tight coupling between the ocean and the atmosphere across climate transitions and between circulation changes and denitrification in the AS.
 Moreover, our new δ15N record clearly shows intensified denitrification from about 8 ka BP in the northeastern AS whereas denitrification in world's OMZs including the western AS generally decreased following the deglacial maximum. We suggest that the northeastern AS influenced the global Holocene climate through sustained N2O production and marine nitrogen loss, partly offsetting the global ocean trends.
 We would like to thank Frauke Rostek and Philipp Böning for technical assistance as well as Tara Ivanochko, Raja Ganeshram, Dick Kroon, and Arun Singh for thoughtful discussions. We thank Tom Russon for his care in editing the paper. The manuscript greatly benefited from the intellectual contributions and extremely constructive comments of two anonymous reviewers. We acknowledge the financial support from the College de France and the Gary Comer Science and Education Foundation.