We estimate the global ocean N

Nitrous oxide (N

The large uncertainty in the oceanic emissions of N

Primary biological pathways of the oceanic nitrogen N cycle
represented in the model simulations, along with redox states of N.
Nitrification occurs in the oxic ocean (blue arrow). Denitrification yields
net N

Estimates of the contribution from suboxic regions of the ocean (about 3 %)
to the global N

Here, we estimate the global ocean N

In this section we provide an initial estimate of global marine N

Thus N

Apparent N

Published estimates of global ocean N

We used four databases to tune or optimise different aspects of the N cycle
in the PlankTOM10 ocean biogeochemistry model. The number of data points
reported for each database are after gridding to
1

To parameterise the model N cycle, we use a cost function to minimise the
difference between model and observations, following the methods of

This formulation gives equal weight to the relative correspondence between model and observations at small and large observational values. A value of 2 means that, on average, the model deviates from the observations by a factor 2 in either direction. To calculate the cost function (and also to calculate MSE in Eq. 6), the model was regridded to the same grid as the observations, and residuals were calculated at months and places where there are observations. The cost function results for the optimised simulations are summarised in Table 1.

Our initial biogeochemical model configuration is PlankTOM10

In order to represent nitrification rate, the state variable for dissolved
inorganic nitrogen was split into NO

Cost function (Eq. 3) for the optimisation simulations of Sect. 2.2–2.4, relative to the respective observational databases. The nitrification rate in bold was used in this study.

We used the calculation of the preferential uptake of NH

N

Surface NH

N

The prognostic submodel presented here is based on process understanding and
explicitly represents the primary N

N

In most of the simulations, atmospheric

Depth profiles of N

The PlankTOM10.2 biogeochemical model coupled with the two N

In previous versions of the PlankTOM model

In addition to the uncertainty that arises from the model-observations
mismatch, uncertainty is contributed by the uncertainties in the N

The global N

In order to find the optimal N

We used the surface

Depth (m) profile of average NO

High N

When we used observed atmospheric

MSE

MSE

Surface

Finally, we add the uncertainties in the solubility and the piston velocity
to the total N

Contributions of coastal (bottom depth

Models of the global marine C cycle have been in use for decades, and a lot
of the available information has been synthesised, cross-correlated and
interpreted in detail

This lack of knowledge also means that partitioning the global marine
N

Despite these shortcomings, the global marine N

Frequency distribution of

We also tested how much influence sampling biases of very high
supersaturation values might have on the estimated air–sea exchange. If the
40

Possible biases in ocean physical transport could in theory affect N

Global oceanic N

To improve the estimate of the ocean N

The four databases presented in this paper are
available as NetCDF files from

The authors declare that they have no conflict of interest.

This research was supported by the European Commission's Horizon 2020
programme through the CRESCENDO and EMBRACE projects (projects 641816 and
282672). We thank Martin Johnson for the database of NH