The Southern Ocean with the largest uptake of anthropogenic nitrogen into the ocean interior

The oceanic external nitrogen (Nex) deposition to the global ocean is expected to rise significantly owing to human activities. The Southern Ocean (SO) is an important pathway, which brings external influences into the ocean interior. It touches the borders of several developing countries that emit a large amount of anthropogenic nitrogen. To comprehend the dynamics of Nex in the SO, we developed a new method to assess the change in the oceanic uptake of Nex (ΔNex) in the entire SO. We obtained the spatiotemporal distribution of ΔNex in the SO by applying this method to a high-resolution grid data constructed using ship-based observations. During the 1990s to the 2010s, Nex increased significantly by 67 ± 1 Tg-N year−1 in the SO. By comparing this value with the rate of Nex deposition to the ocean, the SO has received ~70% of Nex deposition to the global ocean, indicating that it is the largest uptake region of anthropogenic nitrogen into the ocean interior.

The reactive nitrogen (N r , i.e. NO x , NH y , and dissolved organic nitrogen) input to the open ocean has increased significantly since 1860, especially in the last two decades 1 . Such consistent increase in the reactive nitrogen input could lead to changes in the ocean nitrogen and carbon cycles apart from affecting the marine biological productivity. Anthropogenic nitrogen released by human activities such as industrial nitrogen fixation and combustion of fossil fuel has contributed the most towards this increase. Nearly 70% of oceanic external nitrogen (N ex ), which is defined as the input of fluvial and atmospheric N r in this study, is anthropogenic N r 2 . Considering that the turnover time of natural N r in the ocean is approximately 3,000 years 3,4 , the change in N ex (ΔN ex ) on the decadal timescale can closely reflect the change in the anthropogenic uptake in the ocean. The distribution of ΔN ex in the surface ocean has been reported by several studies 1,5,6 . However, the spatiotemporal distribution of ΔN ex in the ocean interior is yet to be revealed clearly; consequently, we lack the comprehension of the amount and storage of anthropogenic nitrogen received by the ocean as well as the variation in the oceanic uptake of anthropogenic nitrogen with time.
The Southern Ocean (SO, south of 30°S) covers approximately 30% of the global ocean surface area, and it is an important pathway that drives external influences such as anthropogenic impact into the global ocean interior owing to the strong movement of water masses (e.g. meridional overturning circulation) 7 . The SO is also very susceptible to anthropogenic materials because much of the sea surface water flowing into the SO touches the borders of several developing countries such as China, India, and South-East Asian countries. Therefore, clarifying the ocean dynamics of N ex in the SO is crucial for gaining a deep understanding of the human impact on the ocean.
However, there are two challenges in exploring the presence of N ex in the SO. One is the difficulty in acquiring ocean observations owing to the severe environmental condition of the SO. Ship-based observational data of the SO are considerably deficient compared with those of other oceans in the Northern Hemisphere. Recent studies on climate change in the SO have mainly focused on multiple repeated ship-based observations along the same lines every decade; the data collected is sparse owing to the difficulty in collecting data from the entire SO 8,9 . The other challenge is difficulty in separating N ex from the internal nitrogen (recycled nitrogen, N in ). Kim et al. (2014) reported the impact of anthropogenic nitrogen on the western North Pacific using N * and the water mass age 6 . Their approach could not remove the effect of nitrogen fixation and denitrification; consequently, it was difficult to estimate the anthropogenic nitrogen in the ocean accurately and apply it to the global ocean.
Recently, a new method capable of estimating the change in anthropogenic CO 2 impact on the ocean interior across decadal time intervals using parameterization techniques was proposed 10 , which makes it possible to Results and Discussion parameterization of reactive nitrogen. We use nitrate (N) to represent N r because nitrate accounts for more than 90% of N r and it is the most stable dissolved form of nitrogen in the interior ocean (where most of ammonium and organic nitrogen are already conversed into N through nitrification or remineralization) 3 . The parameterization technique allows us to reconstruct the nitrate concentration spatiotemporally in the SO by using other hydrographic properties 11 . We used the hydrographic data for dissolved oxygen (DO or O 2 ), water temperature (T), salinity (S), and pressure (Pr) along with the observed N (N obs ) to perform the parameterization of N in the SO. All the data we used were sourced from Global Ocean Data Analysis Project version 2 (GLODAP v2), Climate and Ocean: Variability, Predictability and Change (CLIVAR), and Carbon Hydrographic Data Office (CCHDO) (https://cchdo.ucsd.edu/; Table S1 and Fig. S1(a)) 12,13 . By giving several data constraints in obtaining an optimal parameterization (Table S2), we obtained the predicted concentration of N (N p ) in the SO, as follows:  Table S3. Several statistical tests and an independent dataset were used to confirm the accuracy of our parameterization method (see Supplementary Text S2 for details). Additionally, we compared the spatial distributions of N obs and N p in the SO of 30°S south at surface, 500 m, 1,500 m, 3,000 m and 5,000 m (Fig. S4); consequently, the distribution of N p was in good agreement with that of N obs , demonstrating that our parameterization has high accuracy and applicability to the reconstruction of N in the entire SO.

oceanic uptake of external nitrogen
Separation of n ex from oceanic n. N obs comprises an internal term (N in ) and an external term (N ex ) because the modern hydrographic data we used were already influenced by changes in the external matter. Heretofore, the separation of these two terms of N obs was challenging. A method to estimate the variation in the external term of the observed ocean carbon species across different arbitrary years was proposed recently 10 . This method could be extended to distinguish N in and N ex (see Supplementary Text S4). We assumed that N ex contained in N p is the average N ex between 2000 and 2016 (N ex 2008 ) and it remains constant with time due to the use of cruise data from 2000 to 2016 for constructing the parameterization of N p . We can estimate the variation in N in by considering the difference in N p across different years (ΔN p ) due to the difference in N ex as zero. The variation in N ex (ΔN ex ) can be obtained by subtracting ΔN p from the variation in the observed N (ΔN obs ) ( Fig. S6; Eqs. (S2-S5)). Here, N in includes the nitrate originating from the processes associated with DO, T, S, and Pr in the ocean, such as biological nitrogen fixation and remineralization; N ex represents only the effects of atmospheric deposition and riverine nitrogen.
Through this method, we noticed that we could estimate ΔN ex for a particular place by using ΔN p along with the data for ΔN obs of that place for different years ( Fig. S6(b)). In order to draw the cross sections of ΔN ex in the SO, we selected three repeated observations from the 1990s to the 2010s along the lines SR03, I08, and A12 as the representative data for the Pacific, Indian, and Atlantic basins (Fig. S7). Considering the uncertainty of the N p parameterization (RMSE = 0.80 µmol kg −1 ) and the propagation of uncertainty from the calculation (Eq. (S5)), ΔN ex has an uncertainty of 1.13 µmol kg −1 , which means that ΔN ex larger than this value must be significant. We estimated the meridional distributions of total water column inventory of ΔN ex along each section ( Fig. 1) by integrating ΔN ex from the surface to the sea floor. Both SR03 and A12 have high water column inventories of ΔN ex between the Antarctic Polar Front and the Subantarctic Front (50°S to 55°S), and both I08 and A12 near the Antarctic continent (60°S) also show high water column inventories of ΔN ex . Considering the low primary production on the surface of the SO 14 , the N ex deposited on the surface must mainly enter the ocean interior through the formation of intermediate and deep waters and the penetration of surface water mass in the SO 15 . The Antarctic Circumpolar Current has become more active due to the strengthening of the westerly winds caused by the Southern Annular Mode, which has been increasing in the past two decades 16 . This phenomenon has strengthened the vertical exchanges of water masses in the SO, which supports the inference that there were remarkable increases in N ex during the past 20 years in the Antarctic Intermediate Water and the Antarctic Bottom Water (Fig. S7).
Spatiotemporal distributions of ΔN p and ΔN ex over the Southern Ocean. Here, we used the same method as the previous sub-section to understand the distributions of ΔN p (variation in internal N) and ΔN ex over the entire SO. Considering the lack of observational data in the SO and the necessity for repeated observational data for the same location, we selected the observational data corresponding to the period 1990-1999 to represent the 1990s, 2000-2009 to represent the 2000s, and 2010-2017 to represent the 2010s. The data of each period were interpolated onto a common grid (see Supplementary Text S3). We used a grid with horizontal resolution of 1° × 1°, and 43 vertical layers with 50-m thickness from the surface to 500 m, 100-m thickness from 600 m to 1,500 m, and 200-m thickness from 1,700 m to the sea floor.
Seasonal differences between different cruises may affect our estimation. Owing to the severe environment of the SO, most of our observed data were collected in the warm period. In order to verify whether there was a significant difference between the data for cold period (for convenience, we call it wintertime) and warm period (for convenience, we call it summertime), we used the data of wintertime (April to October) and summertime www.nature.com/scientificreports www.nature.com/scientificreports/ (January to March) and calculated the average N obs and N p at each depth for these two durations (Fig. S8). We found that above the depth of 500 m the differences of both N obs and N p between the two seasons were ~3 μmol kg −1 as maximum; the corresponding differences at the depth of around 200 m became ~0.80 µmol kg −1 , which was equal to the RMSE of our parameterization. These two periods did not show an obvious difference below the depth of 500 m. Thus, we concluded that the seasonal difference in the observational data does not significantly affect the spatiotemporal distributions of ΔN ex along with the total water column inventory .
The spatiotemporal distributions of ΔN p and ΔN ex are shown in Figs. 2 and 3(a), respectively. The spatiotemporal distribution of ΔN p (Fig. 2) showed a large variation in the upper 1,000 m water column and it revolved around the Antarctic continent along with the Antarctic Circumpolar Current in the different time periods. This phenomenon may be due to the continuing enhanced nutrient-rich Circumpolar Deep Water upwelling derived from the strengthening of the Southern Hemisphere westerlies in recent decades 17,18 . Furthermore, ΔN p became  zero gradually with the increase of depth, implying that there is almost no nature-derived variation of N in the deeper water column. The distribution of ΔN p showed no obvious difference between the Pacific, the Indian and the Atlantic sector of the SO.
In Fig. 3(a), the spatial distribution of ΔN ex in the surface layer shows a tendency to diffuse along the continental coastal area to the open ocean (e.g. west coast of South America, southwest coast of South Africa, and south of Tasmania, Australia). The data for the continental shelf were removed to eliminate the uncertainty of river input in our parameterization construction based on the assumption that the riverine N ex has little effect on the open ocean 1 (see Supplementary Text S1). Jickells et al. (2017) found that approximately 75% of riverine N escapes beyond the shelf break and enters the open ocean, which may partly explain the significant rise in N ex in the coastal region in our study 5 .
In terms of the temporal distribution of ΔN ex , the Indian sector has shown a remarkable growth in N ex from the surface to the abyss during the period from the 2000s to the 2010s. By analyzing the spatiotemporal distribution, the reason for this can be attributed to the increase in anthropogenic nitrogen emission in developing countries such as India, China, and Southern Africa in the past decade 5,19,20 . According to the evaluation of the global meridional overturning circulation, the upwelling water in the surface North Pacific Ocean passes through the Strait of Malacca and reaches the northern Indian Ocean. Then, it goes south and flows into the Southern Ocean 7 . Meanwhile, the surface anthropogenic N is loaded on these waters along the coastal regions and brought to the Southern Ocean. Additionally, the enhancement of the Southern Annular Mode mentioned in the previous section can explain the increase in N ex in the ocean interior.
We also estimated the total water column inventory of ΔN ex from the surface to the sea floor in the SO (Fig. 3(b) and Table 1). During the 1990s to the 2010s, N ex in the Pacific, Indian, and Atlantic sectors grew at the rate of 24 ± 1, 42 ± 1, and 0.02 ± 0 Tg-N year −1 , respectively, and that for the entire SO grew at the rate of 67 ± 1 Tg-N year −1 . Uncertainties were given by the standard error of gridding estimation (Table S6). The ΔN ex in the Indian sector accounted for 63% of the increase in N ex in the SO. We also found that the Atlantic Sector, which has the most active vertical circulation in the world, did not show a high ΔN ex . This may be because of the following  www.nature.com/scientificreports www.nature.com/scientificreports/ two reasons: (1) the deviation caused by the seasonal differences in the surface data (Fig. S8); (2) the inflow of the deposition of N ex from the Atlantic sector into the Indian sector due to the Antarctic Circumpolar Current, which also explains why ΔN ex in the Indian sector is extremely high. In the Pacific Ocean, we mainly observed the accumulation of ΔN ex in the surface layer (Fig. 3) due to the upwelling area with relatively old water age in the deep Pacific 21 . These results can be considered reasonable compared with the previous model predictions 1,5 .
In an early study 1 , the deposition rate of N ex to the global ocean was predicted as 67 ± 30 Tg-N year −1 in the 2000s, the upper limit of which was 96 Tg-N year −1 considering the potential impact of riverine input. By comparing the deposition rate with our data, we found that the SO had received 69% of the global oceanic N ex input despite the SO covering only 29% of the global ocean surface area, which emphasizes the important role of the SO in integrating anthropogenic impacts in the global ocean.

conclusions
We presented the spatiotemporal distributions of ΔN p and ΔN ex in the SO from the 1990s to the 2010s using the simple parameterization of the predicted N along with the observed N (R 2 = 0.97; RMSE = 0.80 µmol kg −1 ). In the Indian sector, which borders several developing countries, N ex has grown at a rate of 42 ± 1Tg-N year −1 , accounting for approximately 63% of the overall rate of increase of the SO (67 ± 1 Tg-N year −1 ). y comparing our result with the global deposition rate reported by Duce et al. 1 , the SO was found to receive approximately 70% of the global oceanic input of N ex despite it covering only one-third of the global ocean area. In the future, a more detailed evaluation of N in the SO can be obtained by relying largely on ship-based observations and/or applying this parameterization method to autonomous biogeochemical Argo floats and CTD sensors [22][23][24][25][26][27][28][29][30] .