Comment 1: It would be very helpful to include stream order, COMID and NextDownID based on the MERIT hydrography. The thinking here is that many catchments are nested because the dataset is gauge-based. Including the suggested columns/fields could enable the users to incorporate upstream-downstream relationship in their analyses, for instance, where scaling behavior is key.

Thank you for your suggestion. We added another file “GSHA_MERITinfo.csv” in the “/Global_files/” directory with COMID, order, and NextDownID fields from the MERIT database. We also included the uparea field from MERIT as a comparison of our watershed boundaries.

Comment 2: Minor comment: the writing mixed present and past tenses, causing confusion on which actions are taken by the authors and which are taken in previous studies. For instance, the sentence from L288 to L292 is grammatically ill and hard to understand.

Thanks for your comment. We checked the verbs in the article and modified the tense of the wrong sentences. The description in 3.4.1 Meteorology datasets section involves information of many dataset names and might be confusing. Therefore, we modified the entire section to shorter sentences to clarify the descriptions. We also modified some other grammar mistakes in the text. We are sorry for the inconvenience.

Comment 3: The dataset involves merging several datasets – taking weighted average. I expect a brief introduction of the motivation/justification of applying certain method, for instance, Lu et al.,2021. This is important because it actually matters whether to merge several data sources or to simply provide their individual values. To me, merging is only meaningful when estimation error from individual merged component is informed. This is why the authors need to better justify/explain the merging of multiple datasets.  

Thank you for your comment. We did not use the weighted average to merge the datasets, which is different from the purpose of Lu et al., 2021. The area-weighted approach was used to extract grid data to watersheds. Because the grids are fully or partially contained by the watershed boundaries, it is not accurate to simply calculate the arithmetic mean of all the grids intersecting with the watershed boundary. Therefore, we calculate a weight for each grid representing the proportion of grid contained by the watershed boundary. This follows the method adopted by Addor et al., 2017, which is the foundation of CAMELS datasets. For dataset merging, we agree that simply merging the datasets by their mean values without further estimating errors is irresponsible. Therefore, we treat each dataset as an independent estimation of the variable and did not merge them for an averaged value. We added “based on the proportion of the grid area contained in the basin boundary” in line 348 to clarify the spatial aspect of this issue.

We thank the reviewer for offering constructive comments to our manuscript. Attached please find our point-to-point response.