Major comments.
This manuscript has improved a great deal. Most of my concerns have been addressed. I have two remaining main comments and few detailed suggestions. I support publication of this important effort.
1. I thank that authors for clarifying most of the uncertainties about gap-filled data, especially the 10 sites that are completely gap-filled with no indication of which data are original vs. gap-filled (I believe - I have one clarifying question about this in the detailed comments). I remain uncertain, however, about how data from the remaining 40 sites are handled. Do the remaining 40 sites include some investigator gap-filled data? If so, are these identified in some way? I could not find this information in the text. Ideally I would suggest that the authors eliminate all investigator gap-filled data and use their own gap-filling to make the data product as consistent as possible. I have questions related to this topic in the detailed comments. I strongly recommend that the authors clarify this point. My suggestion that they replace all investigator gap-filling with their own method is that - a suggestion. If the authors do retain some investigator gap-filled data I would strongly suggest that they label these points separately from the original observations, and separate from their own gap-filling.
2. The description of methods has improved considerably. I have one remaining concern / suggestion. It can still be difficult at times to separate the workflow you used to test the gap filling and record prolongation algorithms from the workflow used for gap filling and record prolongation. These are mixed together in Figure 2. I believe it would help the clarity of the document and the long-term value of the manuscript if Figure 2 was used only to describe how the final data products are created. The authors could make a smaller “cut out” describing the work flow for algorithm testing. The results of the manuscript can then easily cite these two different figures and work flows and remove the remaining confusion.
My primary confusion remains in the AutoML algorithm vs. MDS, RF and XGBoost. The text says, if I understand it correctly, that AutoML selects from among gap-filling algorithms, suggesting that it uses one of these three methods for each site in its final gap filling. The text, however, compares AutoML to these algorithms. Please clarify this issue.
Detailed comments.
1. line 19-20.
“an automated machine learning approach (AutoML) of H2O framework”
The grammar needs work here. I don’t think the addition “of H2O framework” helps the text.
2. Line 65 begins an exceedingly long paragraph. Please break this up into paragraphs according to the flow of ideas within this text. Perhaps this is already broken into smaller paragraphs but I can’t tell due to the typesetting.
3. Line 195.
“for each flux towers”
each flux tower.
4. Thank you for pointing out the citations for the data sets in Table A2. This is traceable and acceptable.
5. Line 201-202.
“Among the final set of sites, 40 provide quality-controlled original LE observations, while the remaining 10 sites only provide continuous LE time series that have already been gap-filled within their observation periods.”
Thank you for addressing this important issue. I have two remaining questions that I suggest that you should clarify in the text. First, for the 10 sites that provide continuous time series, is it true that those sites do not distinguish the original observations from the gap-filled data points? If so, please make this clear. If not, you could do an independent gap filling. I believe this would improve the consistency of your data set.
Second, are there any investigator gap-filled data in the records from the other 40 sites? If so, do you retain those? Are they identified in the data sets? I would suggest that you replace all investigator gap-filled data with your gap-filling - if this is possible. I ask this because I note, for example, that sites 12-19 have less than 10% missing data. This is remarkably complete data if there was no gap filling.
I wonder if a flow chart would help to explain these issues of gap filled (investigator vs. yours) vs. original data, and data labelling in your data base?
6. Figure 1a. This background is helpful and I can readily identify the flux site vegetation types. I have two remaining suggestions. First, please define the vegetation cover categories and the source of that land cover data. Second, please delete the elevation color bar that remains in the figure.
7. Lines 275-277.
“Specifically, only observations flagged as high quality (QC/QA = 0) were retained, while data affected by instrument malfunction, low turbulence conditions (e.g., insufficient u*), or energy balance non-closure were excluded. Only reliable records were retained for model training.”
Please delete the second sentence. It is redundant.
More importantly, this again raises my earlier question about investigator gap-filled data and the “Number of LE Data” and “Missing Ratio” columns of Table A1. It would make the most sense to me if these columns reflected the number of “reliable records” used in this data product. Is that what is reported in Table A1? If not, I strongly suggest adding these additional columns somewhere in this document. I also strongly suggest clarifying, in your final data product, the origins of each data point - original observation, investigator gap filled, unknown (if you can’t distinguish original observations from investigator gap filling) and gap filled using your methods. If possible, I also suggest removing all investigator gap-filled data to provide the most internally consistent data product possible. If you choose to retain investigator gap filling, please make this clear in the text and in the final data product.
8. Figure 2 is a hybrid. It shows both the research the investigators used to identify (actually justify, since it appears that AutoML was selected at the onset of the study) their favored gap-filling method and the methods used to fill the final data set. I suggest that this should be broken into two diagrams. One can outline the testing of methods and a second could outline the final data gap filling only with the selected algorithms. I believe this would make the data processing much clearer to most readers.
9. Lines 287-289.
“AutoML of H2O framework (LeDell and Poirier, 2020; H2O.ai, 2023) was adopted as the core modeling tool for LE gapfilling in this study, which has been widely used for regression and prediction tasks in environmental and geoscientific applications (Guo et al., 2024; Li et al., 2025b; Zhao et al., 2026).”
The phrase “which has been widely used” is a dangling modifier. What has been widely used? I cannot tell.
10. Lines 300-311. Is one algorithm chosen for each site? Or can AutoML change algorithms (over time?) within the record of a single flux tower? This text suggests that only one algorithm is used for each tower - if I understand the text correctly.
11. Lines 300-311 seem to conflict with Step 2 of Figure 2 which suggests that AutoML is one of the possible models to be used. Instead, this text suggests that AutoML selects from among a set of algorithms. This is confusing. Please clarify and make Figure 2 and this text internally consistent.
12. Lines 316-317.
“…gap lengths of 30 min, 1 d, 7 d, and 30 d. These four scenarios were used only for performance evaluation and method comparison, not as four independent production gap-filling procedures.”
As with the algorithms, this causes confusion. I suggest that Figure 2 be split up. Please illustrate the testing of algorithms from the final data processing used to create your data set. I think that the “testing” diagram could be a simpler subset of the full path shown in Figure 2. Figure 2 could then be simplified by removing the algorithm testing elements (gap lengths?) that are not part of the final data processing.
13. Line 361.
“And RMSE was used…”
Please delete “And”.
14. Line 362-363.
“Except to these quantitative metrics, a qualitative assessment of diel-cycle reconstruction was conducted for representative sites.”
I’m sorry, but I don’t understand this sentence.
15. Line 405-406.
“The flag T denotes half-hourly LE values directly derived from original flux tower observations.”
Are these all observations for the 40 sites that were not gap-filled without labels for the gap-filled times? Or are investigator gap-filled data points included under the “T” designation? See my earlier suggestions.
16. Section 3.1.1. My understanding of AutoML, based on this manuscript, is that it selects from among the algorithms that it is being compared to here - MDS, RF and XGBoost. How can it be that it performs better than all of these, if in fact AutoML selects one of these for each site? I am afraid that I do not understand the text describing how AutoML works. Can you please clarify this text?
17. Figure 3. The caption is improved. But I’m sorry, I still don’t know what is being reported in the figure. The caption says, ‘for each flux tower’. But there is only one figure which I believe(?) describes all of the flux towers. There is also the mention of 20 repeated experiments - which I don’t understand. Please explain, in the caption, in good detail, the populations that are used to create the figure. Hopefully this is mirrored in the methods. After this you can say “as in Figure 3” for any additional descriptions of the same populations. My apologies if I am slow to follow your work, but this needs to be clear to typical readers.
18. I am confused by the caption of Figure 4. Figure 4 is a comparison of different gap length scenarios and gap filling methods. The caption, however, mentions “the final released dataset.” This figure does not deal in any way with the “final released dataset”, does it? I know that you are trying to state that this is methods evaluation and not a presentation of the final data set. I believe you have made this clear in the text already. I think this last sentence in the figure caption is confusing.
19. Line 719-720.
“Timestamps are provided in the format YYYYMMDD–HHMM (e.g., 20120101–1530), representing local date and time at each site.”
Does 1530 mark the beginning, middle or end of the 30 minute period of flux data? Please make sure this is clear in the metadata for the data product. |
General comments:
Thanks to the authors for their valuable contribution to the in-situ flux data collection for Latent heat flux data from various Chinese ecosystems for major climate zones.
The manuscript describes a continuous dataset based on half-hourly ET data from eddy covariance measurements complemented by site-specific ML approaches to fill data gaps and extend the time series beyond the observational periods.
The article is appropriate to support the publication of the data set, but could be shortened in some parts as exemplary indicated below.
The dataset is unique and useful in terms of duration and treatment for Chinese ET flux data. The public availability in open access repositories of such a dataset is relevant even if the creation of the gap-filled and prolonged dataset might be repeated in case observed data together with the software codes were available.
The aggregates for daily, monthly and yearly products could be omitted, as those time series could easily be derived by the users. In case those files are published, I’d recommend to change the naming of the single files within each archive: each file name should contain the time resolution as is done for the compressed archives.
Within a framework as established for this manuscript, I would expect some more information related to uncertainty of the final flux products that could e.g. be derived from random repetitions of the procedures.
data quality:
The 25 year long ET time series is of good quality, also input data for variables that drive ET from MODIS and ERA5 seem reasonable.
Specific comments:
What about uncertainty estimates of the gap-filled ET fluxes e.g. based on random repetitions of the procedures?
Several references in the text are missing, even though listed in the ‘References’ section. Please check.
Naming of the model: AutoML-H2O or H2O AutoML or only AutoML? Please use the abbreviation consistently.
Specific remarks related to the text:
Line 26: ‘…conventional methods for long-gap conditions of 7 or 30 days, respectively.’? You might reformulate the sentence as those gap conditions are artificially introduced. Under normal conditions, data gaps vary in duration.
Line 30: does the ‘strict quality control’ relate to the EC data from measurements or to the modelled ET data?
Line 65: ‘EC observations provide half-hourly measurements of latent heat flux (LE) in combination with its driving variables,…’. (EC itself delivers only fluxes)
Line 69: ‘Taking the FLUXNET2015 dataset as an example,…’
Lines 85ff: You claim that Chinese flux sites are underrepresented in integrative flux analysis. Might this also be due to the fact that data from Chinaflux are usually not accessible for non-Chinese researchers? Data sent to the FLUXNET portal should be accessible via the FLUXNET Shuttle (https://data.fluxnet.org/). Also see Papale, D.: Ideas and perspectives: enhancing the impact of the FLUXNET network of eddy covariance sites, Biogeosciences, 17, 5587–5598, https://doi.org/10.5194/bg-17-5587-2020, 2020.
Line 131 and later line 160: is the quality control related to EC data? What are the site selection criteria?
Lines 158ff: references for the processing and quality assurance steps should be added.
Lines 160 and 161ff: references and details for ChinaFlux and FLUXNET procedures as well as previous studies should be added.
Line 178: did you use the data from the 10 sites with pre-gap-filled time series in the same way as the data from the other 40 sites? So only artificial gaps introduced? Are the gap-filled data of those sites marked as such? If these gap-filled data are used for training, ‘no new information is generated’
Lines 182-183: ‘The half-hourly LE data form the foundation for subsequent gap-filling,…’ is that what you want to say? If yes, please re-formulate accordingly
Line 193: citation for ERA5-data missing in the text (e.g.: Copernicus Climate Change Service (2022): ERA5-Land hourly data from 1950 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS). DOI: 10.24381/cds.e2161bac (Accessed on 11‑11‑2025)
Line 203: I had no glue on the ERA5-Land data, but is Rn really net solar radiation, which would be Albedo? Instead, Rn is commonly used for net radiation including longwave components. Later in the text (line 571, line 577) Rn is used for net radiation.
Line 213: reference for ‘official documentation’?
Lines 210ff: citation for MODIS products
Line 216: what is the temporal resolution of MODIS products (‘relatively low’)? And is the assumption of constant NDVI and LAI ‘within each compositing period’ sufficient for fast growing plants, disturbances etc.?
Line 228: re-formulate: ‘Each flux tower site was treated…’ – it is not the tower that is of relevance.
Line 230: How was quality screening performed? According to standard FLUXNET and ChinaFlux procedures? See above
Lines 240ff: was the modelling tool adopted from somewhere else (reference!) and what is the contribution of the authors?
Line 249: Which models were selected by AutoML? Very different models? How many different models?
Lines 280ff: ERA5 data are used instead of onsite measured meteorological variables. How do those compare with locally measured data? What about the additional uncertainty?
Line 316: the repetitive sentence can be removed here (These prolonged datasets provide the basis for subsequent construction and analysis of multi-temporal-scale LE products.)
Lines 332-333: does that mean, in case there would be only one half-hour value missing within a 1 d aggregate, the 1 d value would get the flag F? Hardly any ET time series from EC measurements is complete due to unfavourable atmospheric conditions. As a result, any daily (7 day, monthly, respectively) aggregate is based on a mixture of measured and gap-filled data. How are the aggregates flagged then?
Line 432: suggestion: ‘…the 6-year training scenario represents more typical conditions and weather regimes at most sites.’
Chapters 3.3.2 and 3.3.3: this analysis and the accompanying figures seem to be redundant with very little additional information for a data paper, as the monthly values are just aggregated from the daily values which contain gap-filled data. In addition, figure 11 shows again some typical sites but with very different time periods, varying from only 2 years to 10 years. Why are these examples chosen so different if they are compared to check for seasonal variation?
Line 515: what exactly is ‘official ChinaFlux observations’? Please add reference!
Line 547: what is meant with ‘stratification’ in the context?
Lines 560ff: Especially for desert and shrubland, evaporation becomes more dominant compared to transpiration. So it is clear that vegetation-related variables explain less variability. This point might be considered here as well.
Lines 577ff: same as above, more soil is exposed, so more evaporation compared to dense canopy covers.
Figures: Readers should be able to interpret figures with the figure description. Most figures need more descriptive text.
Fig. 1b) more description needed,
Fig. 1d) length of observation periods (in years) for all sites
Figure 2: a bit overwhelming, but a good overview still. I don’t see QA/QC for EC-data (which contributes to gaps). The figures for comparison with MDS and ML methods and also the ones for performance do not add any value due to their small size. The text is not readable and legends are missing. Even if the content becomes clear for the results in the lower right corner, instead of T, F, and P for true, filled and prolonged you might use additional colour indication as T, F and P are not in the figures anyway.
Fig. 5 and 6: what is the measure for the significant bias marked by the blue boxes?
Fig. 7, a) and b): y-axis needs legend. More explanatory test in the figure description is needed, e.g. for ‘Relative density’
Fig. 8a): from the figure description it is not clear whether the bars or the lines relate to the left or right y-axis. Please provide more information in the figure description.
Fig. 9: might be removed or moved to the appendix. Instead give more details about statistics in the text in chap. 3.3.1
Fig. 10 and 11: same as for fig. 9, as daily and monthly values are just aggregates of half-hourly values in case less than 10% of missing data.
Fig. 12: ‘l’ is missing in word ‘Daily’ in c) and d)
Table A1: make sure that words in the table header are not wrapped (looks ugly)
Line 669: What is meant with: ‘..that this station provides only interpolated data.’Are these the same 10 sites mentioned above wwith gap-filled data? Are those data treated as measured data?