Dear authors,

I read the manuscript “How much water vapour does the Tibetan Plateau release into the atmosphere?” with great interest. The validation of many different ET products over these water towers of Asia has a lot of value. While the manuscript is generally well written and clear, I do have some specific comments and requests for clarification of the presented analyses.

Regarding the validation:

• Provide clear explanation on the temporal scale the analyses were conducted (monthly?), this is not always clear
• Provide clear explanation on the period used for the analyses (in some cases the overlap of the in situ data (either EC towers or water balance estimates) and products is rather short
• Basins used in the water balance estimation is not always clear, eg figure 1 doesn’t show the Heihe basin (is this the Hexi corridor and is the entire basin included in the map/analyses?). In figure 1 what does the stripped area refer to? A table with information would be useful with some additional information on the data used from the studies by Ma and Zhang and Wang et al. Also the basins are referred to as the Yangtze/ Yellow river basin, but as far as I understand these only cover the upper part of the basin. Please provide some additional information on the extent of each of the basins analyses (eg provide name of the gauging station where the basin was delineated). Also in figure 3, there is a reference to TP, which basin/ area does this refer to (the entire TP area shown in figure 1 or the area of all the basins combined, which are two different areas)?
• Color scheme of figure 3 is not fully intuitive, for example the r2 is deep red for high (=good) values)

Figure 5: what do the different colors of the bars mean?

Trend analyses (figure 7):

• The calculation of the trends could be affected by an exceptional year with high or low ET at the beginning or end of the time series (since there is quite some yearly variation and the trends are often relatively minor). Could you say something about the significance of these trends as well? Also for the SynthesisET both the first two years and the last two years seem to be outliers and related to the “temporal inconsistencies” of the product. Was this data properly vetted before including in the analyses?
• Why are many of the products with longer time series (eg ERA5Land, SynthesisET, BESS, MERRA2) not presented with their full timeseries?

Analyses of “ET components”

• As mentioned by the authors these different sub-components of ET are not validated and with the wide range of values derived from the different products, what conclusions can really be drawn? This is especially a question for the open water ET (maps in figure 9 shows large areas evaporating from water surfaces) and sublimation (which is validated how?)

Analyses related to the “response to different environmental factors”

• The purpose of these analyses are not entirely clear to me. First, the analyses are done for the median value of the correlation, whereas it was already very clear that there is a large variance between the different products. Also several products utilize these input data (Rn, LAI, P) for estimating ET, how is this kind of dependency considered in the analyses? Do different types of models have stronger or weaker correlation with these environmental factors? And what does that mean for the interpretation of the analyses?
• Did any of these factors also influence the partitioning of ET into ETc and ETs?

Discussion:

• General reflection of the validation methods employed, doesn’t really add much information. The incorporation of seasonal land cover conditions or lack thereof is only explained for 3 products, but then no reflection on how that has affected the results. Or how relevant negative latent heat fluxes are (does this happen often or only occasionally?). The reflection on the water balance estimations are also very general and could have been included in the introduction (there is no reflection based on this specific study). For example, the assumption of not incorporating meltwater could have been explained in the method but is not an outcome of this research.
• The discussion related to the different types of models comes a bit out of the blue, for example in table 2 the model type is not provided, which makes is difficult to validate a statement such as (first sentence) “ PM-type model demonstrated superior accuracy compared to other models”. Also “.. models that incorporate soil moisture to detect water stress…” can not be checked, which models do or do not incorporate soil moisture? Also to go in depth into the methodology of each product seems to go beyond the objective of this research, especially since it unclear why some models are singled out and others not (nor a statistical comparison between for example PM vs non-PM models is not done.
• The uncertainty of the SynthesisET product was already mentioned in the results section, is this really an important outcome of this research (important enough to single it out in the discussion?)

The article "How much water vapor does the Tibetan Plateau (TP) release into the atmosphere?" by Zheng et al. provides a comparative analysis of evapotranspiration (ET) on the Tibetan Plateau, an essential yet uncertain component of the water cycle. This comprehensive review examines various streams of ET data and compares them with in-situ flux measurements, aiming to address a significant research gap: Can ET estimates derived from satellite and land surface models accurately reflect in-situ ET observations?

While I appreciate the insights offered by this article, particularly its thorough incorporation of diverse data sources, there are concerns regarding the clarity and completeness of the methodology. Consequently, the obtained results lack sufficient substantiation. Therefore, before publication, these concerns need to be addressed thoroughly.

Major comments:

Regarding the Methodology:

1. The temporal coverage of the analysis is not clearly defined throughout the article. In line 197, it is written 2003 to 2015, in line 221, 2001 to 2018 while in line 312, it is written 2000 to 2020. These discrepancies need clarification to ensure consistency and accuracy in the reporting of the study period. I suggest keeping the results with consistent temporal coverage in the main section while any other information on supplementary information (SI).
2. Although it appears to be conducted at a monthly scale based on the information provided, it is unclear whether all datasets, such as ETMmonitor with daily resolution and MOD16 with 8-daily resolution, were aggregated to a monthly scale for comparison or were based on the native resolution of the dataset. Clarity is needed regarding the aggregation process of these datasets to ensure transparency and understanding of the methodology employed.
3. In line 135, it is mentioned that months with less than 50% valid daily ET values were excluded from the analysis. However, it remains unclear whether these excluded months were filled to maintain a continuous ET time series or if the comparison was limited to months with more than 50% valid ET values. Clarification on how the missing data was handled and its impact on the analysis is necessary for a comprehensive understanding of the methodology.
4. Providing information on the number of valid observations available for each dataset, either in the supplementary information or elsewhere, would be beneficial for assessing the comparability of sample sizes across datasets, especially if they are not analyzed for same temporal coverage. 

Regarding the results:

1. It appears that the regional-based formulations of ET, such as MOD16STM and PLMV2 ET Tibet, demonstrated the highest accuracy when compared to in-situ flux towers. However, it is crucial to ensure that the flux stations utilized in this study for comparison were not already included in the calibration of these datasets. If the same flux stations were used for calibration, the greater accuracy of these products may not be fully substantiated. Therefore, it is imperative to verify whether there is any overlap between the flux stations utilized in this study and those used for calibration to accurately assess the reliability of the results.
2. In Figure 3, it is unclear how the metrics were calculated for the entire Tibetan Plateau (TP). Does the metrics for TP represent averages or medians across the basins or was TP treated as a single basin?
3. In Figure 4, the color bar for ET standard deviation (ETsd) differs from the color bars used for other variables. This inconsistency can lead to confusion, particularly since the figures are presented together. Also, if possible, please keep the results in the order of datasets that appears in the Table 1.
4. Regarding Figure 7, it would be beneficial to highlight the trends observed specifically from data with long records to discern the presence of significant trends in ET, because the trend calculated with only some years of data would not add any conclusion to the overall trends in the ET. Additionally, it's essential to clarify how the trends were calculated—whether through linear regression or another method—and whether the significance of these trends was assessed.

On results specific to “Response of the ET to main governing factor.”

The author's intended message or purpose behind the analysis is not clearly conveyed. It seems to explore the relationship between annual ET and various water, energy, and vegetation variables. I will try to highlight my concerns in points here:

1. In my belief, the analysis of how annual ET responds to different water, energy, and vegetation variables could potentially be a separate study requiring a more comprehensive approach.
2. For instance, If Leaf Area Index (LAI) correlates well with both/or net radiation (Rn) and precipitation (P), which I believe will be the case, raises doubts about the conclusions drawn regarding the relative influence of these variables on evapotranspiration (ET). This is true especially when conclusion on  influence of these variables on ET is drawn simply from correlation of ET with these variables without controlling for the other confounding factors. To check whether this is the case or not, we can simply correlate LAI with Rn and P, as well as by correlating Rn with P.
3. Even if one were to accept the current analysis, which I personally disagree with for the reasons outlined in points 2, there remains a crucial need for clarification regarding the rationale behind correlating median ET from all datasets (if I understood it properly) with environmental variables (Figure 10). This need arises primarily from the significant variability observed among different ET datasets in terms of magnitude and hence I believe that the relative importance assessed from the simple correlation of ET with these variables will also vary. Consequently, any conclusions drawn from these correlations may lack robustness.
4. Again, in regions where Ec and Ei are the dominant modes of evapotranspiration (Figure 8), it would be valuable to investigate their correlation of ET with LAI compared to Rn and P, after removing the confounding effects.
5. Nevertheless, I still believe this could be separate research with robust approach. 

Additional technical comments: 

1. Before highlighting the monthly RMSE, it would be helpful to provide information on the magnitude of monthly ET observed at different flux stations based on in-situ observations. This would allow for a comparison of the magnitude of observed ET with the error represented by the RMSE.
2. It's advisable to maintain analysis with consistent spatial and temporal coverage in the main section, while keeping analyses involving datasets with inconsistent coverage to the supplementary section. This will enhance clarity of the manuscript.
3. In Figure 8, it is noted that while the total evapotranspiration (ET) may appear similar across different datasets, the partitioning of ET between datasets is not consistent. This observation is indeed a significant finding. However, the substantial explanation provided does not sufficiently clarify why the datasets differ so much, particularly for GLDAS and MERRA2.

Other comments:

Overall, there are numerous instances in the text which exhibits repetition and with typos, with numerous lines conveying similar information and occasionally out of context. Therefore, significant restructuring of the article's text is necessary.

For instances:

1. The passage from lines 60-65 highlights the significant uncertainty surrounding evapotranspiration (ET) estimation on the Tibetan Plateau (TP). However, the paragraph falls short in effectively conveying how the present research differs from existing literature. It is evident that this study introduces novelty to the field, particularly through its comprehensive comparison of various ET products with in-situ observations in TP. This contribution warrants greater emphasis in the introduction section.
2. The final three paragraphs in the introduction section needs to be rephrased for coherence, eliminating redundancy to convey the message clearly. For example, lines 81-86 present the research questions effectively. However, the same information is reiterated in the following paragraph (lines 88-91) within the main objectives, which essentially duplicates the content. This and other redundancy should be streamlined for clarity.
3. The classifications in the discussion sections (ET based on PM model, LST-based model, data-driven, and LSM type) seem abrupt as they haven't been introduced earlier in the text.Section 4.1 should be emphasized when analyzing the results, as much of the content there appears redundant in the manuscript, despite its scientific validity. This caveat should be highlighted without unnecessary repetition

    

   On the introduction section, it appears:

   These validations were generally based on either in-situ measurement by the eddy covariance system or the basin-scale ET estimated by water balance method, which represent the surface net water flux that integrates different processes (e.g., plant transpiration for the dense vegetation regions, snow sublimation for the dry snow cover periods for the eddy covariance system observations, even condensation when negative latent heat flux occurs),while these ET products mainly focus on the ET (positive upward latent heat flux), which attributes to the validation uncertainty.

    While in the section 4.1.1, it appears:

 The eddy covariance system observation represents the net water flux integrated across different processes (e.g., plant transpiration                    in the dense vegetation regions, snow sublimation during the dry snow cover periods, evaporation of canopy-intercepted water when                  the canopy is wet due to intercepted rainfall). The vaporization process observed by the eddy covariance system depends on the land                    surface condition, which may vary seasonally and yearly due to factors such as snow/ice, intercepted water, and vegetation.  Meanwhile, eddy covariance system observation includes condensation when negative latent heat flux occurs. Remote sensing-based ET products mainly focus on positive ET (positive upward latent heat flux) and omit processes such as condensation.

These two instances basically convey same information. I do agree this is important point to make reader aware about the validation. However, I think the author could be concise about it and avoid unnecessary repetitions.

4.  In line 100, it might be more appropriate to adhere to existing climatic regime classifications, such as those based on AI or other established frameworks. Because the term rather “monsoon” is kept here in between arid and humid climate types. So, how “different” is “monsoon” from the humid in these classifications? Or what does that monsoon mean when compared with “arid” and “humid”?

5. Figure 1: It’s not clear what does hashing represent. And for some “red” labels, they are not clear like names around “XG”.

6. Equation (1), please write equations of all metrics or skip even KGE. Please make it coherent.

Thank you! 

Goodluck!

When I read this paper, I found that the author may not comprehensively review the following papers:

Chen, X. et al., 2024. A doubled increasing trend of evapotranspiration on the Tibetan Plateau. Science Bulletin.

Yuan, L. et al., 2024. Long-term monthly 0.05° terrestrial evapotranspiration dataset (1982–2018) for the Tibetan Plateau. Earth Syst. Sci. Data, 16(2): 775-801.

Wang, B.*, Y. Ma*, Z. Su, Y. Wang and W. Ma. Quantifying the evaporation amounts of 75 high-elevation large dimictic lakes on the Tibetan Plateau. Science Advances, 2020, 6, eaay8558.

I agree to the author that they have collected more ET products in this study, but the generally conclusions are not really new compared with previous ET studies on the TP. Hereby, I suggest to focus more on ET components verification and their trends. This part has not been fully investigated by previous publications. The ET trends and annual ET estimation does not deserve more energy on it. This means that the title should be also changed. There are also some water balance ET studies. Hereby, this analysis is also not new. Introduction should really have a in depth review of previous work.

The large uncertainty of ET products over the TP has been reported by Chen et al. 2024 and Yuan et al. 2024. The abstract should more focus on the new scientific questions. Please revise the sentence: there is still significant uncertainty regarding the amount of water vapour released by the TP into the atmosphere, otherwise remove it. The abstract should emphasize the innovative results, not repeated information.

The response of annual ET to total precipitation, net radiation and leaf area index was explored to present their governing effect on ET, and the results indicated that precipitation effect mostly in the middle and northern TP and net radiation play significant role in the eastern TP.

There are many other factors which also influence ET. But they are not included in this paper. In addition, this conclusion is normal as other study. I suggest to remove this weak point from this paper.

TP has been indicated before line 60, hereby please replace “Tibetan Plateau” with “TP”.

Line 61, Chen et al. 2024 and Yuan et al. 2024 have listed the big differences of annual ET estimation for the TP. It is better to cite their results directly, since they have compared most ET product for the TP region.

Line 80, these specificities, are you talking about negative latent heat? If yes, please use negative latent heat directly.

Line 81, How accurate are these improved ET products, I understand that this question is already answered at least partly in Chen and Yuan`s publication. The snow/ice sublimation is new in this study. I suggest to revise the second question to: which processes play a significant role to the ET components trend. The third question, I did not find the author provide answers to which factor dominant different ET products. Hereby, the introduction should be rewritten and new scientific questions to be raised. Current formation is quite weak and not comprehensive.

The first aim of this paper is already investigated by Chen et al. Please change this point or further deep this aim. Actually, there are many attribution studies of TP ET trend. Please review their studies, then make a revision for the num 3 aim.

Line 94, I don`t really agree that pearson correlation analysis can provide us the response of ET to precipitation, Rn and LAI. Indeed, I don`t suggest to include this correlation analysis in this paper. These analysis weaken this paper, it does not benefit to this work.

Lines 122, These the sites, please correct this error.

Table 2, EB is a daily ET product, not monthly.

Figure 4, SEBS should be EB?

Please revise ‘in Tibetan Plateau’ to be ‘in the Tibetan Plateau’ or ‘in the TP’.

Figure 5, the figure caption should explain what is meaning for different colored bars.

Figure 7, it is quite difficulty to recognize which bar represent which product. Add the product name corresponding each would be more useful. All the trends are ended in 2020? Their curves in figure 7 do not exhibit the same end year.

Line 322, Among these products, there are nine that provide the main components of ET (Ec, Es, and Ei), it is better to directly say that ‘Nine products provide …’.

It is important to note that there is no independent reference available for the ET components.

I suggest to use the ensemble mean of ET components to check their differences with the ensemble mean. Nine products have provided the ET components. It’s a lot. Their ensemble may be close to the truth.

Figure 8, the blue color around TP lakes may not reflect the truth. Please check if this is caused by a wrong lake mask.

Figure 9, there are some reports about the annual ET amount for the TP lakes. Please cite these papers to verify Ew shown in the figure. I understand that Wang et al. Science Advance should also provide the Ew estimation for the TP. This study could benefit to verify the result in the figure.

Section 3.3, this part is not really persuasive. A simple correlation is not meaningful, in addition, other factors were not fully considered, such as air temperature, soil moisture, wind speed etc. In addition, the correlation of abnormal should be analyzed, not the original signal. I suggest to remove this section.

“the daily land cover inputted” please revise this.