General Comments:

1. This is an important description detailing the calibration strategy used to be able to assign meaningful and traceable mole fraction values to a global network of H₂ flask measurements and is exactly the type of manuscript that should be published in AMT. The difficulty in this “warts and all” description of how the calibration procedures have developed over time is that it makes for quite difficult reading due to the complex nature of the many different tank comparisons performed using many different instruments. I can’t recommend a better way of presenting the data, because ultimately all the useful information is contained within the manuscript and SI. The reader will just have to persevere, jumping between text, Figures, Tables and SI to find what is immensely useful information for setting up a calibration procedure for H₂
2. Section 3 of the manuscript describes the data quality assurance and quality control of the ~6000 glass flasks that have been collected at a global network of sampling sites between 2009-2021. This is an immensely impressive and useful dataset. I was a little surprise however, that this manuscript describing the analytical detail is being published after a paper whew the measurements have been used to assess the representation of the H2 atmospheric budget in the state-of-the-art GFDL-AM4.1 global atmospheric chemistry climate model (Paulot, F et al., 2023).
3. It is clear from the calibration work that has been undertaken by NOAA, that aluminium cylinders are not stable for H₂. This was recognised by NOAA many years ago and is why the primary calibration standards are filled into stainless steel electropolished Essex cylinders. However, even with this knowledge this hugely important global network for H₂ measurements has persisted in using aluminium cylinders for secondary and tertiary analysis and then tried to correct for the many different rates of calibration tank drifts. The paper details extensive problems using this approach (under-sampled cylinders, massively different rates of drift on a tank-to-tank basis), all of which propagates uncertainly into the measurements. Why has a different style of tank not been used, which does not suffer from these issues? I realise that Essex tanks (or a similar style of stainless-steel tank) are expensive but surely it is a requirement of a global H₂ network to reduce measurement and calibration uncertainties where practicable by using tanks that don’t drift?
4. I am a little unsure of the purpose of section 3.2.3, the text does not really indicate why the SPO measurements are given their own section (unless the point is to state that H2 stores well in glass flasks and SPO is uninfluenced by emission sources)?

Specific comments:

1. defined the calibration scale as a WMO scale, whilst L30 defines it a the MPI scale, this is a little confusing so early on in the paper.
2. Grey H₂ not Gray H₂
3. Novelli et al. [1991, 1992] not [1992, 1991]
4. 10-200ppb not 10s
5. Why are the Essex cylinders filled with dry air. I understand that Essex cylinders tend to be stable for H₂, however, is there any evidence that drying the air is a requirement for H₂ stability? In my experience Essex tanks filled with undried ambient air are also stable.
6. What is NOAA going to do with the pre-2009/10 ambient air record for H₂
7. L272-276. What caused the tail or noisy baseline? Do you think use of peak height might have caused a bias; what effect did the higher grade of helium have (removed the noise/tail)? Do you use peak height of peak area for the data using the higher-grade helium, are peak height and peak area data comparable now? Why did it take 4-years to decide to use cleaner helium?
8. The word “few” is not informative.
9. You state that typically H₂ tertiary standards lasted less than a year. However. Figure 3a & b show that many of these tanks lasted much longer than 1-year and most drifted quite appreciably.
10. Why only use 8 or the 11 standards
11. You now define the tanks as working tanks, not secondary or tertiary – why change the tank definition, it is confusing.
12. Why change at 250 psia, is there evidence that the tank drifts at pressures below this?
13. Do you know why H2 drifts in air filled aluminium cylinders? If a non-drifting tank is reused, is it still non-drifting and vice versa with a drifting tank?
14. If the tank shows signs of large initial growth in the first 0.5-2 years, why not fill then store a tank for this time before use?
15. L451-452. I assume the three tanks are aluminium filled with dry air? – this information is not detailed in the text or Figures.
16. SI L281. Figure 5 (a) is missing.
17. SI Figure 5. To understand the year in year comparisons it would be useful for the to have the error bars plotted.
18. SI Figure 5. The data in 5(b) are not that easy to understand. Why are the NOAA (2018) and MPI (2019) data carried out a year apart quite similar, but the NOAA (2021) and MPI (2022) a year apart quite different (~2 ppb). There are also very few NOAA data point to compare with MPI.
19. L 462-463. You state the MENSI program provides an important on-going check from MPI X2009 H2 calibration scale transfer in GML. What is not clear is how the results presented in SI Figure 5 are used?
20. Does the restating the information about the flask sampling systems need its own section (3.1), why can’t the information be contained in Section 3.
21. Is there any indication that H2 is stable/not stable in the glass flasks over time?
22. L677 to 678 and Figure 7. How are reliable results between S and P methods defined and tested? Visually from Figure 7, it looks like the S flask data are slightly below the P flask data (looking at the apex of the annual cycles in 2020 and 2021 for example)?
23. How do you define “reliable”, this is a bit non-specific.
24. What metrics have you used to determine that there are no biases?
25. L779-781. It is clear that the Mauna Loa data show more short-term variability that Samoa and South Pole, but not necessarily Barrow
26. L779 to 787. It is not clear how the maxima and minima for each site have been determined, and wouldn’t these vary year to year given that there is a growth trend in the data?
27. What does ASC stand for?
28. L 816 and Figure 10. Given all the sites are defined at the top of the plot, why do you need to use the x-axis to number the sites. Surely you should use it to illustrate the latitudinal gradient.
29. L 817 to 834. I can’t find any information in the manuscript of SI defining the site acronyms or detailing their lat/longs (useful). Can this be contained in a Table? Also, in the text you define some sites described in the text e.g., TPI site, on Taiping Island, but don’t define others e.g., TAP, AMY, LLN, CPT, KUM, WIS.
30. L828 to 834. Is this short description of moving sites required? There is no supporting evidence to explain why the mean level of H2 or seasonal cycle have changed since the move. Just the assumption that increased soil uptake is responsible – is the new location more inland? Can you use ozone deposition or radon measurements to confirm this?
31. L844. A large proportion of Africa (and fires) are in the NH. The flask sites at ASC, NMB & CPT look well located to sample SH fires from the Peoples Republic of Congo, Angola and Zambia?
32. Table 2. The time of use and Fill date time formats are different, less confusing if you use the same format.

This very thorough and detailed manuscript addresses the recalibration of the extensive 12-year NOAA global atmospheric hydrogen (H₂) measurement program to the MPI X2009 calibration scale.  It does this by adjusting for the significant and variable rates of growth of H₂ in the secondary and tertiary standard aluminum calibration tanks used for these measurements that were originally based on the NOAA X1996 calibration scale.  As the manuscript notes, this has become immensely important work as mankind moves increasingly toward H₂ as an energy storage medium that comes with a large potential to leak into the atmosphere and fundamentally alter the oxidation capacity of the atmosphere with respect to lengthening the atmospheric lifetimes the of methane and other anthropogenic greenhouse gases.

While acknowledging this importance, I had considerable difficulty reading the manuscript because of its distractingly minute detail.  One example is the inclusion of the serial numbers of the instruments used to make the measurements.  Another is the extensive comparisons of the performance of the three different “MAGICC” instrument systems used to make these measurements over time.  This is too much detail for the average reader’s attention span, even though it is important that it be recorded somewhere.

One solution would be for the main paper to outline the principles of what was done with a few illustrative examples, and to move the bulk of the details that should be recorded somewhere either to the supplementary information (SI) addendum together with this AMT paper, or to a separate project report published by NOAA and available through the NOAA GML website.  I recommend distilling the main paper to something like one third to one half of its present length, while still conveying the rigor of this important work, including a few examples, and providing the full details elsewhere.

A few more specific comments follow:

1)  Is NOAA GML now using stable stainless steel gas cylinders to propagate its calibrations going forward so that drift adjustments resulting from the use of aluminum cylinders will no longer be an issue?

2)  The dry air mole fraction ppb H₂ concentration units are not defined when they are first used.

3)  Please explain why calibration scales are necessary for atmospheric research, as compared to using individual calibration standards that are not related to a specific scale.  This important point is not widely appreciated, especially among national metrology institutes.

4)  The word “best”, used in line 170 to describe the post-2009 NOAA H₂ data, is subjective.  A better term might be “most precise”, or something to that effect.

5)  Following on the above comment, NOAA prepared their X1996 calibration scale well before the measurements that are recalibrated in this paper.  Are there pre-2009 NOAA H₂ data that could still benefit from being recalibrated to the MPI calibration scale despite their lower precisions?

6)  Should Paul Novelli, and perhaps Ed Dlugokencky, be coauthors of this work?  They are both retired from NOAA, but much of this work was done by them.

7)  The Teflon o-rings used in the flask stopcocks (line 568) are highly permeable to H₂ and other gases.  Since the flasks are pressurized and H₂ permeates much faster than N₂ and O₂, shouldn’t the H₂ mole fraction decrease with time between sampling and analysis? Has this been tested?

8)  Please use the widely accepted spelling of “gases” instead of “gasses”.  Please also change “data is” to “data are” since “data” is the Latin plural of “datum”. 

9) Given the range of coauthors and institutions involved in this work, and the number of years of data that are corrected, I assume that the acknowledgments and the financial support (lines 958-966) may be incomplete.

Summary:

This is important work that should be published, and AMT is an appropriate venue for this.  But it should first be distilled to a more readable structure that conveys the principles of what was done and summarizes the results, with the details presented either in the SIs or in a separate reference publication.