A robust empirical model for deriving depth-integrated N2 fixation rates from surface measurements in the global open ocean
Abstract. Marine dinitrogen (N2) fixation (NF) is a critical nitrogen source supporting ocean productivity, yet its global quantification remains challenging because vertically resolved measurements are logistically demanding. To address this, we developed an empirical model to predict depth-integrated NF (INF, μmol N m−2 d−1) within the euphotic zone directly from surface NF (SNF, μmol N m−3 d−1) measurements. Using 690 measured NF profiles from the global open ocean, we established a robust relationship: log10INF = 1.85 + 0.84 × log10SNF (R2 = 0.82, RMSE = 0.31). The model performed well globally, particularly for typical profiles in which NF decreased monotonically with depth. It tended to underestimate log10INF for profiles with substantial subsurface NF maxima, especially in the subtropical South Pacific (bias = −0.20), where such maxima were likely sustained by deeper iron supply, and in the northwestern Pacific (bias = −0.33), where the maxima were likely driven by photoinhibition of surface unicellular diazotrophs. Overestimation of log10INF occurred in the eastern Atlantic (bias = +0.24), likely reflecting steep depth declines in NF driven by rapid decreases in light and iron availability. A small subset of profiles had uniformly low NF rates throughout the euphotic zone, associated with low iron and/or phosphorus concentrations. When applied to historical surface measurements lacking corresponding INF observations, model-derived INF estimates were statistically consistent with available INF measurements, supporting its preliminary application. Our model enables more efficient global surveys of N2 fixation, thereby improving our ability to map this critical process and refine understanding of ocean nutrient cycles.
(Disclaimer: I am a field oriented microbial ecologist. I do not have strong expertise in modelling or statistical or mathematical ecology).
Chen et al report development of an empirical model to help expand the range of useful observations for in situ nitrogen fixation in the oceans by predicting depth integrated rates from surface observations. The use of the large data set assembled by Shao et al. 2023, using criteria to assemble a uniform set of observations in terms of length of observation (i.e. 24h) and for the existing depth integrated values, specific depth and sampling resolution (at least 3 depths) criteria. They thus have available 690 depth integrated estimates based on observations from at least 3 depths down to 250m, and almost 3500 surface observations.
In their analysis, they note 3 types of profiles, 1) those in which nitrogen fixation rates decrease monotonically from the surface, 2) those with sub-surface maxima, and 3) those that are uniform with depth and consider how these patterns map to regional patterns of nitrogen fixation.
They then proceed to develop their predictive model which, while not perfect, seems quite robust in allowing extrapolation of the surface measurements. They undertake error analysis using both MLR and GAM to identify biases in the predictions. In general, extrapolations from those with monotonic distributions often suffered slight overestimation while the other profiles often yielded modest underestimations- all of which were potentially correctable when compared to the existing integrated profiles from each profile type. They also undertake an error analysis within specific basins and identify 3 regions where there may be some bias (e.g. the eastern Atlantic) and factors which may contribute to the bias (e.g. upwelling)
This is a very valuable contribution towards broadening our understanding of the distribution and quantitative importance of nitrogen fixation in the open ocean. The manuscript is very readable and logically presented. The error analysis seems to this modelling novice very reasonable and up front.
The authors also provide a useful and reasoned analysis of the external factors (temperature, light attenuation, Fe, P*) which may be most important in regulating the activity within and among the 3 profile types and at the sub-basin scale.
One interesting result (to this reviewer) is the comment that given the two methodological approaches to performing 15N2 uptake studies (the so called “bubble” method of Montoya et al 1996 and the “dissolved N2” method of Mohr et al. 2010 (for which there has been a modicum of hyperbolic discussion). They find little difference between the 2 approaches in their analysis. The authors should also be aware and reference the study of
Wannicke, N., M. Benavides, et al. (2018). "New perspectives on nitrogen fixation measurements using 15N2 gas." Frontiers in Marine Science 5: 120.
Which comes to a similar conclusion through direct empirical comparisons of field measurements.
It is unfortunate but perhaps understandable that the authors chose not to use some of the more quantitative studies using C2H2 reduction- which are often presented as daily rates (albeit often limited to an organism like Trichodesmium). (e.g. see Capone, D. G., J. A. Burns, et al. (2005) which depth integrated values for 154 stations from 4 research cruises in the tropical and sub-tropical N. Atlantic- see Figs 3 & 4). "Nitrogen fixation by Trichodesmium spp.: An important source of new nitrogen to the tropical and sub-tropical North Atlantic Ocean." Global Biogeochemical Cycles 19: 10.1029/2004GB002331.) This paper also provides a calibration of the C2H2 reduction method for Trichodesmium which approximates well the theoretical value of 4:1 (assuming the hydrogenase reaction). A separate paper (Mulholland, M. R. and D. G. Capone (2001). "Stoichiometry of nitrogen and carbon utilization in cultured populations of Trichodesmium IMS 101." Limnology and Oceanography 46: 436-443.) directly shows the correspondence of C2H2 reduction rates and the increase in culture N.
Finally, given the ability to develop these more robust data sets, it’s unfortunate Chen et al. do not try to provide an integrated estimate for global ocean nitrogen fixation, which would be useful to many biogeochemists. I would also point out that perhaps they should at least add this consideration to their discussion- as well as recognizing the efforts of
Weber, T. and C. Deutsch (2014). "Local versus basin-scale limitation of marine nitrogen fixation." Proceedings of the National Academy of Sciences 111(24): 8741-8746.
And
Wang, W.-L., J. K. Moore, et al. (2019). "Convergent estimates of marine nitrogen fixation." Nature 566(7743): 205.
which both attempt to resolve global scale oceanic inputs.
Doug Capone