I’d like to thank the authors for uploading the revised version of their manuscript entitled „Global hydrological droughts in the 21st century under changing hydrological regime“. The revisions are based on the reviewers’ comments and the authors addressed these comments by providing a response letter.

With regards to the specific comment of Referee #1 on the selection of the GHM and its performance I would like to strengthen the importance of the first innovative aspect and discuss the model more prominently in the paper as recommended by Referee #1. The PCR-GLOBWB model is described in Section 2.1 and the advantages of using this model has been justified by its performance as described in earlier work (van Beek et al. 2011, Wada et al. 2012, 2013, 2014). There is no doubt about the model, however, there are two shortcomings related to the earlier work in combination with the application in this study, which I think require more detailed information:

First, the model has been validated and tested against 3600 GRDC observations but without any information on the number of stations where these observations were recorded. According to the references given in the text, the model has been evaluated in 23 major river basins (in case of drought deficit volume) and 26 major river basins (in case of monthly and annual mean flow). The authors should add this information to the text in order to further discuss this source of uncertainty.

Second, the performance measure R² is not a good one to evaluate the performance of a model to calculate deficit volumes based on daily time series as it is insensitive to systematic over- or underestimation. As a minimum, not only R² should be provided but also the slope of the regression line. In addition, Wada et al. 2013 have made a comparison of simulated deficit volumes per drought event against observed river discharge from selected GRDC stations in 23 river basins (R² = 0.75). However, this comparison is based on monthly rather than daily streamflow . Based on the previous work, for this study it would be of benefit if the authors showed the model performance based on daily streamflow for the Q90.

As recommend by Referee #1, I invite the authors to further discuss their model results in the context of uncertainties related to the model structure (Sect. 4.3), i.e. the use of one GHM vs. an ensemble, as well as the uncertainties inherent to the model as provided in the author response.

Regarding Section 4.2, the authors should avoid to compare the outcomes of their model simulations driven by RCP2.6 (RCP4.5) climate with the work of Wanders and van Lanen (2013) and Forzieri et al. (2014) who have used climate input based on SRES A2 and A1B, respectively.

Dear Authors,

many thanks for your response to my comments made on your manuscript. I carefully went through your replies and agree with almost all of your changes made in the manuscript and explained in your reply to the editor comments. I still have two minor comments which you have not addressed so far:

1) You use "observation" and "station" synonymously. I think this is confusing.
2) Uncertainties inherent to the model are not yet described as recommended by Referee #1 and myself. Your argumentation that the model is comparable with other models and that a single model approach will not have a significant impact on the obtained results does not comply with the request. And further, I do not agree with your statement that "a single model approach will not have a significant impact on the obtained results".

After these two points have been addressed the manuscript will be sent out for re-review.