First, I don’t need to be anonymous: Dario Papale.

This is a paper that a lot of people were waiting for, and I think it is an incredible contribution toward the FLUXNET growth and the inclusion of JapanFlux in the international collaboration already happening. I appreciate the effort and the willingness, and the dataset is of very high interest, no doubts that it should be published in this journal. Asian contribution in FLUXNET was very limited until now and this dataset really move Japan and its very nice community upfront in the idea of an open and accessible FLUXNET. Congratulations and thanks for this.

That said, my comments are more related to the formats and procedures. I have no specific comments on the way the authors decided to process the data, there are multiple options of processing and they selected one (REddyProc, specific site management and exceptions etc.). However, what can really cause misinterpretation, confusion and misuse of the data by the users is the consistency of products.

In particular, the authors decided (really nice in principle and appreciated) to follow the FLUXNET2015 standard, that is obtained with the ONEFlux pipeline (Pastorello et al. 2020), that is the same used also by AmeriFlux, ICOS, the European Database and recently OzFlux. The problem is that following a standard means that format, content and processing are the same (Interoperability). I checked carefully the products for a number of sites and the main issue I found is that the meaning of the variables (with the same name), but also the file content and the list of variables are different between this collection and the standard FLUXNET/ONEFlux. Some examples:

• The way USTAR filtering is managed is completely different. In the ONEFlux two methods are used and a joint population of threshold is used to estimate uncertainty, here this is not happening and one method is used and 3 percentiles extracted.
• The meaning of the NEE_REF is completely different, since in the ONEFlux is extracted among 40 realizations and based on the NEE timeseries, here is based on one threshold of USTAR
• The energy balance corrected version of the H and LE is not produced in this colelction
• There are variables that do not exist in the FLUXNET standard (neither in the FULLSET or SUBSET), like all the variables with the three position indexes (e.g. TA_1_1_1, that are very important but not included in the standard FLUXNET files), or variables such “RH_multiple” and “SW_IN_SLOPE_PI_1_1_1”, or some QC that is used with a different meaning (e.g. to select fluxes based on wind sectors)
• The content and split of variables between the FULLSET and the SUBSET is not the same as the FLUXNET/ONEFlux and follows a different logic
• The processing of sites where either the USTAR filtering or the partitioning failed is also fundamentally different respect to what is done in the FLUXNET/ONEFLux pipeline
• The inclusion of CH4 is also something not present currently in the FLUXNET/ONEFlux pipeline

All this makes the statement “The JapanFlux2024 dataset is compatible with the datasets provided by FLUXNET (e.g., FLUXNET2015).” at line 116 not true and valid.

Now, I need to be very clear on this: I’m not criticizing the approach of the authors or promoting what is done in FLUXNET/ONEFlux or suggesting to use ONEFlux. My only concern is that different things should be named in different ways to avoid confusion in the users.

For this reason, my suggestion and request would be to not use the official FLUXNET naming of files and also of variables that are not exactly produced in the same way; just use a different naming structure that you define in the paper and metadata associated to the data, the contribution would be still extremely valuable and a ONEFlux version of the data, if you want, can be still created later.

There are two additional important points to consider, again to ensure a full compatibility and participation to the international effort, and obtain what the authors state in the conclusions: “JapanFlux2024 dataset will bridge collaborations between researchers from Asia and FLUXNET.”

• The metadata and BADM: it is an incredible effort and crucial information, these are very important data, thanks for collecting and sharing them. However, checking some of them I found that the data are not always following the controlled vocabularies and format requirements of the BADM system and this would create problems in case users want to mix data across networks. In this case I see two alternatives: either use a different metadata scheme or make the BADM compatible with the standard (see below)
• The FLUXNET site codes: the authors assigned FLUXNET codes to the sites, however I see two main issues here, again that can affect users: 1) some of the sites already participated in the past to data collections (LaThuile 2008, FLUXNET2015, FLUXNETCH4) and they already had a FLUXNET code. These codes are supposed to the persistent and unique so in this case the same code should be used. Examples: JP-Tky, already shared as JP-Tak, JP-Tmk (JP-Tom), JP-Tse (JP-Tef), RU-Spl (RU-SkP) and possibly others (CN-Qhb-CN-Ha2, ID-Puf-ID-Pag…). It should also be noted that the FLUXNET codes (agreed across regional networks) are case sensitive (so for example JP-Smf was already JP-SMF and this last should be used). 2) in the general FLUXNET code standard system, the small “L” and the capital “i” letters and not any more used, for the confusion between them in some fonts (I and l). Examples: JP-Ksl, JP-Swl, MN-Udl...

On these two points, as already discussed in the past in the context of the Regional Networks collaborations between ICOS/European Database and the AsiaFlux and JapanFlux offices, I can provide assistance to check the BADM and automatically identify inconsistencies with the standard and to support the FLUXNET codes assignment and check against already used codes. All this clearly keeping myself external to the paper due my reviewer role.

Additional points:

Did the authors verified all the geographic coordinates? This is a very important point in particular for the remote sensing link

Line 220: not clear the meaning of the sentence “Calculation of GPP and RECO by the daytime partitioning method was based on the parameterized model, which did not directly use observed NEE.”. The NEE are used for the parameterization of the model, like for the nighttime method.

Thank you for your great contribution to FLUXNET

The AmeriFlux, EuroFlux, and OzFlux networks have been able to ensure the accessibility and usability of high-quality, long-term ecosystem flux measurement data, which are essential for regional ecosystem studies, modeling, and multi-site synthesis. This has been made possible through the support of the AmeriFlux Management Project (AMP), Integrated Carbon Observation System (ICOS), and Terrestrial Ecosystem Research Network programs (TERN), which collaborate with the principal investigators of each flux observation site.

However, unlike other regional networks, AsiaFlux lacks a dedicated support program, resulting in its database being limited in both scale and update frequency. This presents one of the challenges in using eddy covariance flux data for a bottom-up approach to estimating the global greenhouse gas (GHG) inventory, despite the advantage of the eddy covariance method in directly measuring GHG fluxes rather than concentrations.

This data paper by Ueyama et al. is expected to be a breakthrough in overcoming the limitations of flux data sharing in the Asian region. JapanFlux2024 will serve as a role model for flux observation networks in other Asian countries, and I hope that flux datasets from various Asian countries will continue to be released in a series. Ultimately, I look forward to the rapid launch of a new AsiaFlux database that integrates these national flux databases.

I sincerely appreciate the authors’ efforts and provide the following minor comments for further improvement of the manuscript:

1. L140-141: Most of the sites were established for long-term monitoring of CO₂ fluxes, but intensive observations for about a week in the 1990s were also included in the dataset.

--> Short-term data can still hold significant value depending on the purpose. Data from short-term experiments can be utilized in various ways. In particular, datasets collected before the widespread adoption of the eddy covariance system may be of great importance.

2. L186-187: If only the rainfall was measured, the correction ratio was determined using liquid precipitation, which was defined as precipitation when air temperature was greater than 0 ℃.

--> When distinguishing precipitation as either rain or snow, I recommend considering relative humidity (RH) in addition to air temperature (Matsuo et al., 1981).

Matsuo, T., Sasyo, Y., & Sato, Y. (1981). Relationship between types of precipitation on the ground and surface meteorological elements. Journal of the Meteorological Society of Japan. Ser. II, 59(4), 462-476.

3. L194-195: In this dataset, we determined the u* threshold each year to consider its potential shift over the years, which is termed as a Variable u* Threshold (VUT) in FLUXNET2015 (Pastorello et al., 2020).

--> Doesn't the u* threshold vary seasonally? Are there no influences from seasonally varied meteorological conditions or phenological changes of the canopy?

4. L199-200: For urban sites, the threshold was generally not used (e.g., Liu et al., 2012; Ueyama and Ando, 2016); thus, the u* filtering was not applied for highly urbanized sites (JP-Sac and JP-Kgh).

--> Is u* filtering generally not applied to urban sites? It would be helpful to provide a brief explanation for readers who may not be familiar with urban flux observations.

5. L208-209: H, LE, and NEE were filled with the four different u* thresholds (reference, 5th, 50th, and 90th percentile values) using MDS.

--> Does this apply only to nighttime data? Is it typical practice to apply u* thresholds to H and LE as well? Given that turbulent mixing is closely related to aerodynamic and radiative coupling, isn't applying u* thresholds more complicated for H and LE compared to CO₂ flux? Were the u* thresholds used here derived from nighttime CO₂ flux? More details would be helpful.

6. L305-306: A quality information flag was provided for gap-filled variables, where 0 is the original data, 1 is a gap-filled value of the most reliable quality, 2 is a gap-filled value of medium quality, and 3 is a gap-filled value of the least reliable quality.

--> It would be useful to include a brief explanation of how the quality levels for gap-filled data are determined.

7. L368-369: The maximum CO₂ sink (i.e., negative NEE) with each temperature range appeared to increase with temperature up to the annual mean temperature of approximately 15°C.

--> Where can this result be found?

8. What does the circle in Figure 8(b) represent?

9. L447-449: As in the data policy of FLUXNET2015, in case of a synthesis using both CC BY 4.0 and other private data, all data should be treated as Tier Two of the FLUXNET data policy (data producers must have opportunities to collaborate and consult with data users).

--> Besides the JapanFlux2024 dataset, which is publicly available under the CC BY 4.0 copyright policy, are there other datasets uploaded on the ADS website?

10. https://ads.nipr.ac.jp/japan-flux2024/

--> It would be beneficial to include IGBP classifications in the site list table, as this information is highly relevant for data users.

JapanFlux2024 is a highly significant contribution to FLUXNET, greatly expanding the amount of available eddy covariance data in a part of the world that has historically been under-represented.  Thus, I thank and applaud the authors for their valuable contribution to the community.

Since the two other reviewers have also posted detailed comments regarding the compatibility between this product and other FLUXNET products (thank you Dario) and the data processing itself, I will focus my review more on the readability and interpretation of the manuscript.  I found the manuscript very well-written, logical, and the database summary was useful.  My comments are mostly related to clarification:

Figure 2: I find the red circles hard to see on the green background, and it would be particularly difficult for people with red/green colorblindness.  I suggest using a different marker color, and potentially filling in the markers.

Figure 4: To improve clarity I suggest renaming the x-axis to “years of data” and the y-axis to “Number of sites”.

Line 176: How were data aggregated if there were multiple measurements of the same parameter?  Were the measurements averaged?

Line 181: If you calculate vapor pressure from RH, how can you fill gaps in RH with vapor pressure?

Line 182: “If all meteorological variables were missing in some years, the bias was corrected using a regression for the entire data record” – please explain this in more detail, it isn’t clear to me what you mean.

Line 193: Provide a citation for your u-star threshold determination.

Line 229:  It would seem more logical to also filter nighttime fluxes by wind direction so filtering is consistent across the data product, even if that means a greater loss in data.

Line 285 – this is a tiny thing, but it should be [FIRST_YEAR]-[LAST_YEAR] instead of [FIRST_YEAR]_[LAST_YEAR] (dash instead of underscore)

Line 399 – if you artificially set GPP, RECO, and NEE to zero to produce mean annual values, how do you define the start/end of winter?  I suggest putting an asterisk next to the MN-Skt and MN-Kbu values in question and include a more  in-depth explanation of how you determined when to set data to zero.

Line 407-408 If the continually-mowed site significantly changes your box plots, you could consider separating that site out into it’s own category.

Line 424 – evaporation is also coupled with air temperature, not just transpiration.

Figure 8. Please clarify what the box plots and whiskers represent (are whiskers set at 1.5 times the interquartile range?)

Table 2 – why are many of the values N/A? Please explain.