This manuscript by Shao and Xu et al. describes an updated version 2 of the global oceanic diazotroph database. It build upon the previous version by adding additional measurements of marine diazotrophic abundance, N2 fixation rates, microscopic and qPCR-based diazotropic abundance. The spatial coverage significantly improved most notably in the Indian Ocean. The newly revised estimate for global N2 fixation rate is significantly higher (+123 Tg N yr-1, almost doubled) when calculating using a standard arithmetic mean, although surprisingly the geometric mean did not significantly change. A brief analysis and discussion of the 15N2 bubble vs. dissolution indicated a potential general underestimation from the bubble method particularly at high rates, however noting the comparison of samples were from different times so it is not a formal error analysis (which the authors acknowledge). The database is available to download from the provided link in the abstract.

Overall I find this to be an important update to the database mainly due to the significant increase in included measurements and spatial coverage. The database is transparent and mostly well described. The analysis and first preliminary quantification of the 15N2 bubble vs. dissolution is also an important contribution. Perhaps some additional details/analysis could be provided (see comments below), but additional analyses can also be performed independently by users who download the data for their specific interest. There is one important aspect that needs additional clarification in my view before I would endorse this manuscript for publication (global N2 fixation rate calculation, see below).

-Christopher Somes

GEOMAR Helmholtz Centre for Ocean Research Kiel

Major Comment: Global N2 fixation calculation description

Since this paper will likely often be cited for revising the global N2 fixation rate significantly upwards, the description of this calculation should be more transparent and comprehensive:

line 266 (Table 5 caption): “Data are first binned to 3x3 grids…”

This needs to be better described. For example, was there any type of interpolation method used or simple averaging of all measurements in each bin? It would be interesting to know what percentage of bins in each ocean basin has data coverage. How do you define the Southern Ocean region and is that area removed from the other southern regions?

How was the vertical coordinate handled?  Is it evenly spaced or according to the depths ranges in Figure 7?

It is not clear to me how the “Areal sum” calculation was made based on the “Mean N2 fixation rate” (Table 5). Does the “Mean N2 fixation” rate include all measurements or only the “Depth-integrated N2 rates”, which requires 3 measurements in the vertical? If the vertical coordinate is uneven, do measurements that get binned into a larger volume in larger deeper layers have more weight on the depth-integrated rate than shallower layers?

When calculating the “Areal Sum”, do you assume that the “Mean N2 fixation rate” extrapolates across the entire region or do you only consider the area of the bins that have data coverage? For example, the Indian Ocean has about 36% of the bins compared to the South Pacific. Therefore I was expecting a much larger decrease when calculating the Areal Sum relative to the Mean N2 fixation rate for the Indian Ocean compared to the South Pacific. However this relative decrease is quite subtle in Table 5 between these regions. I acknowledge there is no truly perfect way to estimate a global ocean N2 fixation rate with the current coverage, but all of the assumptions and details that go into the calculation should be specifically stated and described.

The authors do not give much context on interpreting the geometric vs. arithmetic mean despite that it is mentioned multiple times throughout the manuscript and gives a significantly different result. From what I understand, geometric mean is less sensitive to the high-end rates compared to arithmetic mean. Does this mean that most of the increase in the arithmetic mean is driven by newly included high-end rates? It would be valuable to know how much of the large increase in the arithmetic areal sum is driven by additional spatial coverage versus generally higher rate values. I would suggest to include a histogram of the previous version in one of the supplementary figures for comparison. If newly included rate values tend to be significantly higher, it would be interesting to know how much of that may be attributable to growing numbers of the dissolution method compared to bubble method (i.e. based on Figure 10).

Minor Comments:

line 84 and data file: Metadata

In the data file, the meta data are titled “Surface …”, yet they are associated with a specific depth, so are they really surface? I am used to seeing chlorophyll expressed by volume not area.

lines 127-129: daily vs. daytime vs. nighttime normalization

I am still a little confused about the time normalization with this brief description. If the incubation is only performed during the day, you convert hours to day by 12 hr/day which assumes no rates at night? I see that incubation hours vary a lot and in some cases not a multiple of 12 hours or 1 day. Perhaps you can describe how individual studies typically convert to a daily rate depending on the incubation period. Would it make more sense to multiply by the daytime of each location during the time of sampling instead of assuming 12 hours?

Table 5.: “n” is missing in Indian Ocean

Figure 7:

Why do you choose geometric mean over the more commonly used arithmetic mean in this figure? Does it look significantly different if you use arithmetic means?

I would be interested to see a euphotic vs. aphotic depth-integrated rate. I am curious how much the generally low to moderate rates occurring below 100 meters contribute to the total depth-integrated rate since they can occupy more volume. Perhaps adding a <100m and >100m panel would be useful? At what depths are the deepest N2 fixation measurements?

Section 4.1/Figure 10:

As mentioned above, I think is a useful first investigation into methodological uncertainties on N2 fixation rates. Is there enough data coverage to do a similar analysis for acetylene reduction?