Foraminiferal Isotope Evidence of Reduced Nitrogen Fixation in the Ice Age Atlantic Ocean

See allHide authors and affiliations

Science  09 Jan 2009:
Vol. 323, Issue 5911, pp. 244-248
DOI: 10.1126/science.1165787


Fixed nitrogen (N) is a limiting nutrient for algae in the low-latitude ocean, and its oceanic inventory may have been higher during ice ages, thus helping to lower atmospheric CO2 during those intervals. In organic matter within planktonic foraminifera shells in Caribbean Sea sediments, we found that the 15N/14N ratio from the last ice age is higher than that from the current interglacial, indicating a higher nitrate 15N/14N ratio in the Caribbean thermocline. This change and other species-specific differences are best explained by less N fixation in the Atlantic during the last ice age. The fixation decrease was most likely a response to a known ice age reduction in ocean N loss, and it would have worked to balance the ocean N budget and to curb ice age–interglacial change in the N inventory.

The sources of fixed N to the ocean are terrestrial runoff, atmospheric deposition, and, most important, marine N fixation. The main sinks are sedimentary denitrification, mostly in continental shelf sediments, and water column denitrification in the eastern tropical Pacific and the Arabian Sea. Sediment records from modern denitrification zones show clear N isotopic evidence of reduced water column denitrification during the Last Glacial Maximum (LGM) relative to the current interglacial (Holocene) (1, 2). The history of other processes, especially N fixation, has proven more difficult to reconstruct. Decreased denitrification and/or increased N fixation would have raised the N inventory of the ocean during the LGM, thereby strengthening the ocean's biological pump and contributing to the observed reduction in atmospheric CO2 during the ice age (15).

N fixation produces oceanic fixed N with δ15N values between –2 and 0 per mil (‰), close to that of atmospheric N2 (6, 7). Sedimentary denitrification removes nitrate (NO3) from the ocean with minimal isotope discrimination (8). In contrast, water column denitrification leaves residual nitrate enriched in 15N that raises the δ15N of mean ocean nitrate above that of newly fixed N (9). In the context of mean ocean nitrate with a δ15N of ∼5‰, N fixation causes a regional decrease in the δ15N of thermocline nitrate, which is observed throughout the western subtropical and tropical North Atlantic (1012).

Thus, the δ15N of organic N sinking out of the surface ocean, if preserved in the sediments, should record N fixation by two related mechanisms. First, most newly fixed N is remineralized in the thermocline, producing low-δ15N nitrate, which is then provided repeatedly to the euphotic zone. Second, N fixation can contribute low-δ15N N directly to the plankton in the surface ocean and the sinking flux that they produce. However, in the low-productivity, low-latitude open ocean where N fixation is thought to be focused, seafloor processing increases the δ15N of bulk sedimentary N from that of sinking N (13), reducing confidence in sedimentary N as a recorder of sinking-flux δ15N. In these environments, where little marine organic N is preserved and buried, foreign (e.g., terrestrial) N can also contaminate the regional ocean signal.

In the subtropical and tropical ocean, planktonic foraminifera are an important component of the sinking flux to the sediments. The organic matrix laid down in foraminiferal tests is physically protected by the test during sedimentary diagenesis (14, 15). Thus, foraminifera-bound N is a promising paleoceanographic archive in sites where sedimentary processes may complicate bulk sediment N isotope records (16, 17). Planktonic foraminifera are heterotrophic zooplankton, and many species also have dinoflagellate symbionts. They obtain N mostly from particulate organic matter (POM), including phyto- and zooplankton (18). The δ15N of zooplankton, including foraminifera, tends to be ∼3‰ higher than that of their food source because of preferential excretion of 14N-rich ammonia (19, 20). The ammonia excretion, in turn, lowers the δ15N of POM relative to the N supply to the euphotic zone (21). In the modern Sargasso Sea, the δ15N of subsurface nitrate supplied to the euphotic zone is 2 to 3‰ (11), the δ15N of suspended POM is closer to 0‰ (21, 22), and the δ15N of zooplankton appears to center around 2 to 3‰ (21), similar to the δ15N of nitrate supply and the N export (22). Thus, we expect the δ15N of foraminifera to be similar to and correlated with the integrated δ15N of the new N supply to the euphotic zone, including nitrate from below and N fixation.

We measured the foraminifera-bound δ15N in surface sediments from several regions (23) (fig. S3) and compared them with the δ15N of subsurface nitrate. The foraminifera-bound δ15N of most species from these locations is close to (rarely more than 1‰ different from) the δ15N of the shallow thermocline nitrate available for upward transport into the euphotic zone (Fig. 1). At some sites, deeper-dwelling species (in particular, Globorotalia truncatulinoides in the southwest Pacific samples) have a higher δ15N than other species, which is best explained by the broadly observed increase in POM δ15N with depth below the euphotic zone (24). One weakness of the comparison in Fig. 1 is that it does not take into account N fixation as a low-δ15N N source within the euphotic zone that augments the nitrate supply, but this simplification appears reasonable (23).

Fig. 1.

Foraminifera-bound N δ15N in surface sediment and core-top samples, compared with measured or estimated subeuphotic zone nitrate. Data are from the North Atlantic near Barbuda-Antigua (5 to 10 cm in gravity core EN18, >250-μm size fraction), Little Bahama Banks (grouped box cores, 0 to 5 cm, >500 μm), Great Bahama Banks (grouped box cores, 0 to 5 cm, from 75- to 125-μm as well as >500-μm size fractions), the Makassar-Bali Basin east of Indonesia (1 to 2 cm depth in multicore, >150 μm), and the western South Pacific near New Zealand (0 to 1 cm in multicores, >150 μm) (23). Here and in subsequent figures, solid error bars indicate SD for replication of the full foraminifera cleaning and analysis protocol; dashed error bars are for different surface sediment samplings in the same region. The three black horizontal lines indicate the estimated δ15N of the subsurface nitrate supply in each of these regions (i.e., the nitrate directly below the euphotic zone): 200 m at the Bermuda Atlantic Time-series Study (BATS) site (11), 100 m in the Makassar-Bali Basin, and, in the case of the western South Pacific samples, Subantarctic Mode Water as measured in the Subantarctic Zone south of Australia (40), which forms nearby and should be relatively unaltered at this site.

Foraminifera-bound δ15N and bulk sedimentary δ15N were measured in sediments representing the past 30,000 years at Ocean Drilling Program (ODP) Site 999A in the Columbian Basin of the Caribbean Sea. In individual foraminifera species (Fig. 2A) as well as in mixed foraminifera samples of varying sizes (Fig. 2B), the LGM δ15N is higher than the interglacial values. Focusing on the picked species data, the interglacial δ15N is 2.7 to 3.2‰, similar to the thermocline nitrate δ15N observed today in the tropical and subtropical western North Atlantic (11, 12). The glacial δ15N is 4.3 to 5.9‰, varying with species. This is generally close to the δ15N of modern deep Atlantic nitrate measured just below the thermocline, 5.0 to 5.5‰ (11, 12). The higher foraminifera-bound δ15N in the LGM could be due to higher mean ocean nitrate δ15N at that time. However, in relevant (i.e., nonpolar) open-ocean records, which derive from various oceanographic environments, there is as yet no evidence for a 2‰ decrease in mean ocean nitrate δ15N upon deglaciation (25).

Fig. 2.

Foraminifera-bound and bulk sedimentary δ15N in the Caribbean Sea from ODP 999A (12°45′N, 78°44′W; 2827 m; sedimentation rate ∼4 cm per 1000 years) during the past 30,000 years (ka, thousands of years ago). (A) Planktonic foraminifera-bound δ15N of individual species. Blue, O. universa; red, G. sacculifer; green, G. ruber. O. universa and G. sacculifer were picked from the >355-μm size fraction; G. ruber was picked from the >250-μm size fraction. (B) Planktonic foraminifera-bound δ15N of different foraminifera size fractions. Blue, >355 μm; red, 250 to 355 μm; green, 125 to 250 μm. Neogloboquadrina dutertrei, O. universa, G. sacculifer, and G. ruber are abundant in the >355-μm size fraction. O. universa, G. sacculifer, and G. ruber are abundant in the 250- to 355-μm size fraction. G. ruber and G. bulloides are abundant in the 125- to 250-μm size fraction. Black and gray arrows indicate present-day subeuphotic zone and subthermocline nitrate δ15N in the western North Atlantic, respectively (12). (C) Bulk sedimentary δ15N, with error bars indicating SD calculated from replicates. (D) δ18O of G. ruber (pink) calcite [‰ versus Vienna Pee Dee belemnite (VPDB) standard] from (41). The age model is from (41), based on 14C dating and δ18O correlation.

The low δ15N of Holocene and surface sediment foraminifera from the North Atlantic tracks the low δ15N of thermocline nitrate in the region, which in turn derives from the remineralization of newly fixed N. Thus, we interpret the high LGM foraminiferal δ15N to reflect a weakening in the δ15N decrease that occurs upward through the thermocline in the modern North Atlantic, such that the euphotic zone was supplied with nitrate with a δ15N much more similar to the nitrate currently found at the base of the thermocline (Fig. 3). This most likely requires that N fixation was much reduced in the ice age Atlantic; a simple estimation, assuming no glacial-to-interglacial changes in thermocline circulation, suggests that it was ∼20% of the Holocene rate (23) (fig. S5).

Fig. 3.

Depth profile of δ15N of suspended POM and nitrate in western North Atlantic during the present interglacial and the last ice age. Solid red circles denote the δ15N of suspended particles in the surface 100 m near the Bermuda Atlantic Time-series Site (BATS) averaged over 1985 to 1988 [data from (31)]; dotted lines indicate 1 SD. Note that the suspended POM δ15N increase accompanying the decreasing POM concentration below the euphotic zone (24), mentioned in the text, occurs below the depth range shown here. Orange circles denote average δ15N of nitrate in the upper 200 to 1000 m at BATS (11); dotted lines indicate the measured range. Solid red and green squares denote foraminifera-bound δ15N during the Holocene and the LGM, respectively. Foraminiferal depth maxima are taken from (30). Green and blue lines are proposed LGM depth profiles of the δ15N of suspended POM in the photic zone and of nitrate in the thermocline (POM calculated from LGM-to-Holocene changes, using the depth ranges indicated for G. ruber and O. universa and assuming a constant foraminifera-POM δ15N relationship for each of these two species; nitrate calculated from the G. ruber change).

In addition to the clear glacial-to-interglacial change, there is a small (0.5 to 1.5‰) deglacial maximum in δ15N apparent in each of the individual species records and in at least the >355-μm size fraction record. This feature, which is common in the deglacial section of sediment records from all basins (25), most likely reflects a peak in the δ15N of mean ocean nitrate. It may result from a transient deglacial peak in water column denitrification relative to sediment denitrification and/or from a transient N loss from the ocean (16), the former process better matching the available data (26). Our record at Site 999 suggests that the deglacial δ15N maximum is relatively weak in comparison to observations from denitrification-influenced records, but we suspect that it was overprinted by an increase in Atlantic N fixation at that time (Fig. 4, dashed interval).

Fig. 4.

Comparison of the ODP 999 species-specific foraminifera-bound δ15N records for the past 30,000 years with bulk sediment δ15N records from each of the major water column denitrification zones and from the nearby Cariaco Basin (27). Denitrification zone records are from the eastern Pacific off Chile (42), the eastern Pacific off California (Santa Barbara Basin) (43), and the Arabian Sea off Oman (44) (fig. S3).

The Caribbean foraminiferal δ15N record is similar to the bulk sediment δ15N record from Cariaco Basin (27, 28) (Fig. 4). The Cariaco Basin record is from anoxic waters, and seafloor isotopic alteration and allochthonous N are demonstrably unimportant there today (29). However, the basin's barriers to circulation have complicated the interpretation of its δ15N record, potentially affecting its changes. Nonetheless, the published interpretation of the deglacial δ15N decrease in the Cariaco record—enhanced Atlantic N fixation upon deglaciation (27, 28)—is supported by our open Caribbean foraminiferal data.

The foraminiferal species Orbulina universa, Globigerinoides sacculifer, and Globigerinoides ruber have similar δ15N during the deglaciation and interglacial; however, they are coherently different during the LGM, with O. universa the highest and G. ruber the lowest (Fig. 2A). In the modern ocean, G. ruber is most abundant in the mixed layer, roughly the upper 30 m in this region. G. sacculifer inhabits a broader depth interval down to the deep chlorophyll maximum (DCM), at ∼80 m in this region, and O. universa tends to have its abundance maximum near the DCM (30) (Fig. 3). Thus, the order of decreasing δ15N during the last ice age coincides with habitats of progressively shallower depth.

In the modern tropical and subtropical western Atlantic, across the ∼100-m-deep euphotic zone, suspended POM δ15N is uniformly low relative to the thermocline nitrate supply, as a result of some combination of N fixation and N recycling (Fig. 3) (21, 22). This rough uniformity is consistent with the similarity in δ15N of O. universa, G. sacculifer, and G. ruber. The δ15N divergence of these species during the last ice age probably arose from an increase with depth in euphotic zone suspended POM δ15N. This might have been the result of a larger δ15N difference between N fixation and the subsurface nitrate supply, which, according to our data, had a higher δ15N during the LGM. However, the species converge in δ15N during the deglaciation, before the δ15N of the nitrate supply had decreased, which suggests that this is not the sole explanation for the interspecific differences during the LGM.

In the modern Sargasso Sea, preferential 14N recycling should work to lower the δ15N of the mixed layer relative to the deeper euphotic zone (31), but this gradient is apparently diluted by N inputs to the entire euphotic zone, including N fixation. With less N fixation during the LGM, the recycling-driven δ15N gradient may have been unobscured and thus stronger, leading to the clear δ15N differences among species. It is also possible that, with less N fixation, the remaining fixation and its isotopic signal [as well as that of atmospheric deposition (32)] was focused in the warm, nutrient-poor surface mixed layer, isolated from the nitrate supply from below. In this scenario, the species' convergence in δ15N at the δ15N maximum marks the deglacial acceleration in Atlantic N fixation (Fig. 4).

Our bulk sediment δ15N record from ODP Site 999 has limited correspondence to our foraminiferal records, possibly showing a deglacial maximum but only a weak (<0.5‰) glacial-to-interglacial decrease that is insignificant in a Student t test (Fig. 2C). We suspect that changing inputs of shelf material to Site 999 have altered the relationship of sinking to sedimentary N across the deglaciation. At Site 999, Holocene sediment has ∼35% terrigenous material by weight, with ∼60% during the LGM (33). Associated with this change, the δ13C of organic matter is ∼1‰ lower during the LGM (fig. S6), whereas marine POM δ13C should have been higher (34). This is consistent with a greater proportion of terrestrial organic matter during the LGM. The sediment organic carbon–total nitrogen ratio is lower in the LGM interval (fig. S6), and distinct trends in this ratio during the LGM and Holocene intervals imply a greater contribution of clay-bound N during the LGM (fig. S7). Both terrestrial organic matter and clay-bound N tend to be low in δ15N (35) and would have worked to lower bulk sediment δ15N during the last ice age, and thus would have destructively interfered with the oceanic change apparent in the foraminiferal δ15N data.

The Site 999 data indicate that N fixation in the Atlantic was low during the LGM, when water column denitrification in the eastern Pacific and Arabian Sea was reduced (Fig. 4) and when sedimentary denitrification was probably also much lower (36). In this context, the LGM-to-Holocene increase in N fixation is consistent with a proposed negative feedback on ocean N content, due to N fixation: Denitrification-driven deficits in N relative to phosphorus (P) stimulate N fixation because N fixers are favored in N-depleted, P-bearing surface waters (37), causing N fixers to restore the ocean N:P ratio toward the ∼16:1 “Redfield” ratio of plankton (38, 39).

It has been argued that the micronutrient iron controls or modulates N fixation rate (3, 4). The higher dust flux during the last ice age has been proposed to have caused higher N fixation at that time (3, 4), a hypothesis apparently disproved by our data for the Atlantic. However, we have not constrained the history of N fixation in the Indo-Pacific, where iron is typically more scarce. Moreover, although our results demonstrate a response from N fixation that works to stabilize the N budget, they do not preclude a glacial-to-interglacial decrease in the N inventory. Reconstruction of N fixation in the other basins and better information about the timing of changes should help to quantify the strength of the N fixation feedback and the limits that it places on the ocean N inventory. Foraminifera-bound N will facilitate this effort.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S7

Table S1


References and Notes

View Abstract

Navigate This Article