Minimal Influence of AAN Deposition on the North Atlantic.

The average CS-δ¹⁵N in the Hog Reef coral core is 4.17‰, with a SD of 0.26‰ (Fig. 2A). This is 1.6–1.7‰ higher than the thermocline nitrate δ¹⁵N of the surrounding ocean (Fig. 1B), consistent with both prior work on Bermuda (29) and the global correlation between the δ¹⁵N of the N supply to surface waters and CS-δ¹⁵N (19). Despite the low-δ¹⁵N of AAN deposition at Bermuda (12, 13, 15, 22), the CS-δ¹⁵N record is remarkably stable. There appears to be a long-term decline in CS-δ¹⁵N of 0.6‰ from the early 20th century to the 1980s, with several superimposed decadal oscillations of 0.5–0.6‰ (see smoothed record in Fig. 3). One might argue that the long-term δ¹⁵N decline, although weak, was caused by increasing AAN deposition. However, since the 1980s, CS-δ¹⁵N has increased back to the earlier 20th century values, a change that has not been observed in terrestrial or ice core δ¹⁵N records (Fig. 2C). This suggests that factors other than AAN have dominated variations in Bermuda CS-δ¹⁵N over this time period. The decadal variability aside, the stability of CS-δ¹⁵N across the entire record suggests that the AAN can only have been a minor contributor to the N supply to the remote North Atlantic surface ocean over the 20th century.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

The Bermuda CS-δ¹⁵N record and its comparison with other records. (A) The Bermuda CS-δ¹⁵N record (red circles and black line, the latter being a 10-y running mean), with the CS-δ¹⁵N record from Dongsha Atoll in the South China Sea shown for comparison (gray line) (10). The Bermuda CS-δ¹⁵N record was compiled from two coral cores: the continuous coral core from Hog Reef (AD 1880–2014; solid red circles) and the fossil coral core from the John Smith’s Bay (AD 1780–1840; open red circles). Due to a problem in core archiving, only short intervals in the bottom part of the John Smith’s Bay coral were available for CS-δ¹⁵N analysis. (B) Comparison of CS-δ¹⁵N changes to modeled surface ocean δ¹⁵N changes (28) in the North Atlantic and South China Sea over the 20th century. (C) Greenland Summit ice core nitrate δ¹⁵N record (green line) (25) and power functions (black lines) fitted to δ¹⁵N records from 24 remote North American watersheds (23). The watershed δ¹⁵N records were normalized to each record’s preindustrial value (23). In both A and C, δ¹⁵N decreases upward, so that interpreted greater AAN importance is upward. (D) US N fertilizer consumption and NO_x emissions (red and orange lines), compared with China fertilizer consumption and coal consumption, the latter intended as a rough indicator of combustion N sources (black and gray lines). Data in D were compiled by the Earth Policy Institute.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Decadal variation in the Bermuda CS-δ¹⁵N record and the history of the North Atlantic Oscillation. (A) The Bermuda CS-δ¹⁵N data (black circles) and its 10-y running mean (red line). (B) The station-based North Atlantic Oscillation (NAO) index (light gray line) and its 10-y running mean (red line) (57). The CS-δ¹⁵N record shows maximum positive correlation (r = 0.70) with NAO when CS-δ¹⁵N lags NAO by roughly a decade (SI Appendix, Fig. S3).

AAN deposition can lower the δ¹⁵N of ocean ecosystems and thus corals in two related ways. First, in each year, AAN deposition represents a N input to the euphotic zone (the photosynthetically active upper layer) that is lower in δ¹⁵N than the nitrate supply from below and in situ N₂ fixation, thus working to lower the δ¹⁵N of organic N produced by phytoplankton (Fig. 4A). Second, operating over multiple years, the lowered δ¹⁵N of organic N sinks into the shallow subsurface and is remineralized, gradually lowering the δ¹⁵N of the subsurface nitrate (Fig. 4A). This subsurface nitrate is the dominant annual supply of fixed N to the surface ocean, and thus, its δ¹⁵N decrease is communicated back to the surface ocean ecosystem. We first address the δ¹⁵N of subsurface nitrate in the Sargasso Sea and then consider the annual N isotope budget of its euphotic zone (the sunlit surface layer).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

(A) The euphotic zone fixed nitrogen budget in the modern Sargasso Sea and the depth profiles of temperature, nitrate concentration, and nitrate δ¹⁵N near Bermuda during phases of (B) positive NAO and (C) negative NAO. On an annual basis, AAN deposition (with a δ¹⁵N of −5‰) represents less than 5% of N supply to the euphotic zone in the modern Sargasso Sea (SI Appendix, Fig. S2), with the vertical nitrate supply dominating the total fixed N supply and N₂ fixation (with a δ¹⁵N of −1‰) contributing a small fraction of N supply. The thermocline waters from which nitrate is supplied, in turn, exchange water with the deeper and higher-latitude waters, which is simplified in A as “mixing” with underlying deep water. During a positive phase of NAO (B), a thin STMW layer is generated in the subsurface, increasing the nitrate concentration and nitrate δ¹⁵N in the thermocline. During a negative phase of NAO (C), a thick STMW layer is generated in the subsurface, decreasing the nitrate concentration and nitrate δ¹⁵N in the thermocline. The temperature and nitrate concentration depth profiles are from ref. 51. The nitrate δ¹⁵N depth profile under positive NAO state is from Fig. 1, whereas the nitrate δ¹⁵N depth profile under negative NAO state (dashed yellow line) is based on a calculation (SI Appendix, Figs. S4 and S5).

In the modern Sargasso Sea, the subsurface nitrate δ¹⁵N at 200 m depth is 2.5‰ (Fig. 1) (16, 18), lower than the mean ocean nitrate δ¹⁵N of 5‰ (32). It has been suggested that N₂ fixation in the euphotic zone and subsequent sinking and remineralization of the resulting organic N is dominantly responsible for the deviation of thermocline nitrate δ¹⁵N from the mean ocean value (33, 34). However, because AAN at Bermuda has an even lower δ¹⁵N than newly fixed N, it could not be ruled out that AAN is an important contributor to the low δ¹⁵N of the nitrate in the Sargasso Sea thermocline (11). Similarly, it has been suggested that the low nitrate δ¹⁵N (of ∼3‰) in the Mediterranean Sea (35) is dominantly due to anthropogenic inputs (36).

However, the modern CS-δ¹⁵N is similar to the CS-δ¹⁵N in the late 19th century. Moreover, the fossil coral from the south shore of Bermuda (John Smith’s Bay; Fig. 1A) going back to approximately AD 1780 confirms this low δ¹⁵N (Fig. 2A), suggesting that the low nitrate δ¹⁵N in the thermocline water has been a persistent feature of the Sargasso Sea since before the Industrial Revolution. Because anthropogenic perturbation of the atmospheric reactive N cycle was very limited before the Industrial Revolution (7, 25), the stability of Bermuda CS-δ¹⁵N argues that natural processes such as N₂ fixation in the surface ocean, not AAN deposition, are the dominant drivers of the low δ¹⁵N in the nitrate of the modern Sargasso Sea thermocline. Qualitative expectations are that relatively rapid AAN deposition would be needed over many years to significantly change the δ¹⁵N of thermocline nitrate, due to the substantial burden of fixed N in the thermocline and its continuous exchange with the rest of the ocean’s vast nitrate reservoir (“mixing” in Fig. 4A). Thus, this finding is not particularly surprising.

In comparison with the δ¹⁵N of thermocline nitrate, the annual N isotope budget of the euphotic zone is more directly affected by a given rate of AAN deposition, and yet the stability of CS-δ¹⁵N argues that it also changed little over the 20th century. A simple isotope mixing calculation of this budget indicates that the 0.2‰ decrease in Bermuda CS-δ¹⁵N from the early 20th century to the year 2000 translates to only a 2% increase in the annual fixed N supply to the Sargasso Sea euphotic zone due to rising AAN deposition (SI Appendix, Fig. S2). One potential caveat in this calculation is that the rate of N₂ fixation in the North Atlantic might have decreased over the 20th century in response to increasing AAN input. In our isotope mixing model, when including suppression of N₂ fixation that perfectly compensates for AAN deposition, AAN deposition represents a maximum of 5% of total fixed N supply to the euphotic zone for the year 2000 (SI Appendix, Fig. S2). Thus, it appears that compensation by N₂ fixation is not dominantly responsible for the stability of the Bermuda CS-δ¹⁵N record; rather, AAN deposition increases must have been small relative to the nitrate supply from below (Fig. 4A).

The lack of a clear decline in CS-δ¹⁵N at Bermuda was not expected. A model study of the effect of the recent history of AAN deposition rate on the δ¹⁵N of the organic N produced in the upper ocean predicts a decline of 0.8‰ from the early 20th century to the year 2000 in the Sargasso Sea (28), greater than the observed 0.2‰ decline in the Bermuda CS-δ¹⁵N record over this time period (Fig. 2B). The 0.8‰ decline in the model occurred despite the inclusion of N₂ fixation suppression by AAN; a larger δ¹⁵N decline would have been simulated without it (28). The Bermuda record also contrasts with a CS-δ¹⁵N record from Dongsha Atoll in the South China Sea (10), which shows a δ¹⁵N decline of 0.7‰ by the year 2000 (Fig. 2A); the South China Sea CS-δ¹⁵N change is consistent with the model-based prediction for that region (Fig. 2B). In the South China Sea, the same isotope mixing model as applied to Bermuda suggests that by the year 2000, AAN deposition had increased the total N supply to the surface ocean by 17–23% (10) (SI Appendix, Fig. S2).

The difference of the Bermuda CS-δ¹⁵N record from ocean model predictions raises the possibility that ocean models have overestimated open-ocean AAN deposition. In general, the variation in atmospheric N deposition across models, and thus the reliance of the community on model ensembles (37), points to the substantial uncertainties in the model estimates. In terms of systematic bias, overestimation of AAN deposition has been suggested and related to ocean/atmosphere cycling of both ammonium/ammonia and dissolved organic N (27), for which there is isotopic and chemical evidence at Bermuda (13, 14). The Bermuda CS-δ¹⁵N record thus provides impetus for further investigation of the processes responsible for AAN delivery to the open ocean (27).

An alternative set of explanations for the discrepancy relates to oceanic transports of fixed N. Specifically, the global ocean models used for the simulations may underestimate the rate of nitrate supply to the euphotic zone from below, thus artificially raising the proportional importance of AAN deposition in the N budget of the euphotic zone. The existence of such a bias toward underestimating the subsurface nitrate supply is supported by a documented discrepancy between ocean models and observations for net primary productivity in the subtropical gyres of both the North Atlantic (at BATS) and the North Pacific (38). Such a bias might apply in the subtropical gyres but not in the South China Sea because the gyres do not experience the wind-driven upwelling that occurs in the South China Sea, so that more complex mechanisms of nutrient supply are proportionally more important in the gyres. Global ocean models are unlikely to adequately simulate these more complex mechanisms (17, 39).

The lack of clear CS-δ¹⁵N decline in the Bermuda coral core also contrasts with the strong δ¹⁵N declines observed in remote North American terrestrial records (>3‰ on average) and Greenland ice cores (>10‰) (23, 25, 26) (Fig. 2C). On an annual basis, both terrestrial and marine ecosystems receive most of their fixed N from internal recycling (40, 41). In the ocean, circulation and mixing supply subsurface nitrate that itself is the product of remineralization of sinking organic N. On land, soil organic N is decomposed and remineralized, making the N available to plants (40, 42); the freshwater bodies from which the terrestrial δ¹⁵N records were generated also receive N from soil organic matter remineralization (23). The rate of this recycled N supply appears to be comparable between land and ocean (40, 43), so a land-to-ocean difference in this term is probably not the primary driver of the distinction in the AAN δ¹⁵N signal between the terrestrial δ¹⁵N records and the coral CS-δ¹⁵N record (Fig. 2).

One cause for the greater response of the terrestrial (lake) δ¹⁵N records is probably simply their greater proximity to the anthropogenic N sources and thus their receipt of more AAN deposition (44). However, given the starkness of the distinction, it may have additional causes. For example, as discussed above, the δ¹⁵N of thermocline nitrate is buffered by continuous exchange with the global ocean nitrate reservoir and thus is expected to change only very gradually due to anthropogenic N accumulation in the subsurface (28). In soils, however, the source of the remineralized N flux from soil organic matter may be restricted to a small fraction of the total soil N, with the rest cycling so slowly as to be considered virtually unavailable (45). If so, the remineralized N flux to plants and runoff may decline in δ¹⁵N relatively rapidly as AAN accumulates in this small, actively cycled fraction of soil N. Putting aside this specific proposal, the contrast between the land and coral δ¹⁵N records promises to advance our understanding of the differences in N dynamics between land and ocean.

Influence of the North Atlantic Oscillation on the Sargasso Sea N Cycle.

The clearest variation in the Bermuda CS-δ¹⁵N record is the 0.5–0.6‰ oscillation on roughly decadal time scales, which appears to be related to the North Atlantic Oscillation (NAO) (Fig. 3), as supported by the magnitude-squared coherence between the two records (SI Appendix, Fig. S3). The biogeochemistry of the North Atlantic has been shown to be strongly affected by the NAO (46⇓⇓–49). There is evidence that Bermuda reef metabolism is sensitive to the NAO through open ocean productivity and organic matter transport to the reef (50). However, the proposed mechanism for this connection involves changes in mixing-driven nitrate supply to the sunlit surface waters around Bermuda, which would involve an immediate response to the NAO driver (50). In contrast, the strongest correlation of the CS-δ¹⁵N record with the NAO index (with a correlation coefficient of 0.7) is achieved with a lag in CS-δ¹⁵N of roughly a decade (SI Appendix, Fig. S3).

The correlation and lag between the Bermuda CS-δ¹⁵N record and the NAO are best explained by the formation Subtropical Mode Water (STMW) north of Bermuda, which has been shown to be affected by the NAO and in turn to affect the nutrient dynamics of the subtropical North Atlantic through thermocline ventilation (51) (Fig. 4 B and C). In a negative NAO state, a thick layer of STMW forms north of Bermuda and is transported southward. This relatively thick STMW layer is characterized by a temperature of 18 °C and a low nitrate concentration of 2–3 μM from 100 m down to 600 m (Fig. 4C). Its low nitrate concentration allows the remineralization of a given amount of low-δ¹⁵N, recently fixed N from the euphotic zone to cause a greater decrease in the δ¹⁵N of the thermocline nitrate. As a result, the negative NAO state should encourage a decline in the δ¹⁵N of the thermocline nitrate supply to the euphotic zone. In contrast, in a positive NAO state, with less STMW formed, the thermocline should occupy a narrower depth range (Fig. 4B). As a result, deeper, denser water with high nitrate δ¹⁵N and concentration should approach the euphotic zone, increasing the δ¹⁵N of the nitrate supply and thus causing the recent rise in CS-δ¹⁵N (Fig. 3A). The lag between NAO state and CS-δ¹⁵N can be explained by the decadal time scales of (i) the gradual thickening of the thermocline and (ii) the ingrowth of nitrate from the remineralization of newly fixed N. A strong positive relationship between the nitrate concentration and the nitrate δ¹⁵N is observed spatially in the modern North Atlantic thermocline (34), supporting the plausibility of the proposed mechanism. Regardless of whether this specific explanation is incorrect, the correlation of the decadal variation in the Bermuda CS-δ¹⁵N with NAO argues that the former is driven by natural processes, rather than by AAN changes.

Implications for the Future North Atlantic.

Our findings have implications for the future of the effect of AAN on the North Atlantic. The primary anthropogenic atmospheric N source to the North Atlantic is the United States, in which anthropogenic N emission rates have leveled off (Fig. 2D), starting in the 1980s with the Clean Air Act. If N emissions from the United States have indeed plateaued or peaked, given their undetectable role over the 130-y period of the Bermuda CS-δ¹⁵N record, the open North Atlantic N cycle will remain minimally affected by AAN deposition for the foreseeable future.