Dear authors of essd-2020-345,

by now the second round of reviews is complete. As referee #3 (report #2 in this round) has raised substantial concerns, please respond to these comments and also to the few remaining ones that were submitted by referee #2, and revise the manuscript accordingly.

Best regards,

Sibylle Hassler

Dear Dr. Cheng and co-authors,

as referee #3 is generally happy that you revised the manuscript according to most of the comments, but still raises two important points, I would encourage you to address these (and the other comments) thoroughly.
Major points:
- Please ensure that you cite the EC literature correctly and provide a comprehensible understanding of EC theory. This will help the readers to best understand the suitability of the methodology and resulting dataset.
- The "Comment 2" requires some more explanation and critical discussion of methodology on your part. If the referee is not convinced about the suitability of the approach, other readers might also be sceptical. You should address this by offering the limitations of the approach, clear description, possible uncertainties, and a critical discussion of the implications for the model validation.

Please also address the other comments raised by the referee.
I am looking forward to receive a revised version of the manuscript.

Best regards,
Sibylle Hassler

Dear authors of essd-2020-345,

thank you for replying to referee #3's comments. Regarding the paragraph you added as a response to comment #2, about the validation of your data product with limited EC tower availability, I'm afraid the paragraph you added is not easily understood and also does not completely address the concerns of the referee.

Could you please re-write the paragraph (and also, please consult a native speaker again for the English phrasing, some of it could use more clarity), so that you actually discuss the validation basis for your different ET classes, and give some sort of uncertainty estimation. Please take special care to discuss the implication of having only one ground truth EC tower for a class. This is relevant for people who want to use your dataset, it should be clear which parts of the classification have a good validation basis and which should be used with great caution.

Best regards,
Sibylle Hassler

Dear authors,

thank you for improving upon the paragraph about the validation with EC data.
I am happy to push the manuscript further along in the publication process now.

Best regards,
Sibylle Hassler