Dear authors,

Thank you for your detailed responses to the referee comments, which were very supportive, but also suggested a few improvements and pointed out potential deficiencies that needed to be addressed. I think that you addressed most of the points very well and that your proposed revisions will make the paper a very valuable addition to the scientific literature. In addition to your proposed edits, I would also like you to consider the following points:

- Referee #3 was missing a clear statement "in what ways the current paper adds up to the currently reported literature on uncertainties in ET products (e.g. Mueller et al 2014) or in other words what is new here that we didn’t know before?". In your response to the referee's comment about Line 555-560, you do highlight the novel aspects of your analysis, but I believe that it would be good to discuss more systematically in the introduction already what has been done to date (e.g. Mueller and Seneviratne 2014 and citations therein) and what gaps the present study intends to fill, which, as I understand it, goes beyond the application of old methods for bias detection to the new CMIP6 data, as mentioned in Line 130. In particular, the comparison with the catchment water balance and the associated uncertainty bounds are an important addition in my opinion. The paragraph starting in Line 123 could be a good place for this.

- Data and code availability: Thank you for providing links to the datasets used, a summary of the extracted data on zenodo and the scripts for raw data processing on github. Could you provide the full link to the zenodo repository, instead of just the doi? This would make it easier to find: https://doi.org/10.5281/zenodo.4271331
Since github.com is not a suitable permanent repository as outlined in the HESS data policy, would you be able to publish the relevant version of your github repo on zenodo as well? Zenodo offers a very convenient way to link up to a github repo. The data provided in zenodo only contains basin-mean monthly estimates, but Fig. 3 and Fig. 6 show sub-basins. The uncertainty bounds in Fig. 6 are also not included in the zenodo file and it is not clear how the numbers in zenodo actually connect to the original datasets. Would you be able to describe how this table can be recreated from the original datasets using the scripts provided, and similarly, how the sub-basin data and error bounds can be recreated? This would help tremendously, should a reader wish to do a similar analysis for different datasets, as they would be able to verify that they are actually doing the same thing. You stated that the python scripts used for data analysis are available from the authors upon request. Is there a reason not to add them to the scripts used for raw data processing? I thank you already for your additional efforts to make your data analysis FAIR, as described in the HESS data policy (https://www.hydrology-and-earth-system-sciences.net/policies/data_policy.html).

Dear authors,

Thank you very much for addressing all comments very thoroughly and
further improving the manuscript. A special thanks for making available all
the scripts used to process the original data. Could you add the link and/or doi
of the HESSD manuscript to each of the zenodo records?

Except for a couple of technical corrections listed below I see no more issues
to be addressed before publication. I am confident that the readership of HESS
will find your work very helpful.

Best regards,
Stan

TECHNICAL CORRECTIONS:

l324: ...due to there being...
l658: ...such as sap-flow.

Table S2: Could you explain the variables in the table caption, i.e. units and
references to the equations used to compute them?