Dear Dr. Y. Song

Thanks for responding to all of the comments by the reviewers and myself. Please modify your manuscript accordingly. Pay attention to fully address the comments and make sure that the revised version is consistent regarding figure captions etc.

I would like to add one more remark regarding the use of the M values for splitting your data set. As mentioned also by a reviewer, your selection is rather arbitrary. This was the reasons why I suggested to consider a multivariate regression that avoids the grouping of M values into classes. The point of doing so is NOT to improve the fit – the purpose is to get rid of the arbitrary classification. Your observation that the resulting model does not outperform the regression according to groups is not surprising. It actually may be considered as an overfitting of the data. You may consider to test the model performance by applying the Akaike or Bayesian information criteria to compare whether or not your set of models actually performs better if one considers that you need 16 model parameters while a single model may only need 4(?) parameters.


Sincerely

Dr. Christian Stamm
Editor HESS

Dear Dr. Song

thanks for your Revision. There a just two technical details:

My comment to L. 1624: I was probably not clear want I exptected. You mentioned in your first response that you provide the regressions for Fig. 3 etc. without points that can be considered outliers. I had these metrics in mind. It would be nice to add a table in the SI listing These regressions (including R2, p values) and refering to that table in the figure captions.

L. 328 in the latest Version: I suggest to correct to "... CDOM essentially originates from ...".

Upon these modifications I can accept the manuscript.

Sincerely

Christian Stamm