The main achievement of the analysis performed by Petrovic et al. consists of providing insights into the utility of improving spatial resolution and customizing the model setup of RCMs aimed at reproducing (and, possibly, projecting) drought characteristics. Through a simple yet straightforward experiment, they give clear answers that can drive further development of RCMs aimed at reproducing and projecting drought characteristics. I only have two main suggestions for the authors:

1) I wonder if SPEI behaviour for other aggregation scales would be the same. Would it be possible to give any information (even as supplementary material) also for SPEI-6 and SPEI-12?

2) In the Introduction, the authors declare the objective of gaining insights into drought development for Germany etc., but I can’t find specific indications about that in the paper. A detailed section could help; otherwise, I suggest avoiding emphasizing this objective in the introduction.

Please find below some minor comments and typos.

L84: ERA-Interim, capital I.

L87: “concluded”. In the line below: “saw”.

Table 1: are those listed the only models available? Are there others (e.g. HIRHAM)? Please specify in the text. If some models are neglected, please explain why.

L222: to make the paper more self-consistent, I suggest providing more information about how the SPAEF metric works. On the other hand, less room can be devoted to the Mann-Kendall test, which is older and more widely known.

Section 4.1: Taylor diagrams don’t provide information about possible bias. I suggest adding this piece of information.

L296: from 50, I guess.

L314: I suggest removing Table 4 and introducing the mean correlation coefficient as an inset in the respective map (or near the title).

Section 4.3: why not use WRF@15 km here, at least in Table 5? It would further highlight the benefits of model settings.

Table 5: are the Mean SPEI values and the other statistics averaged over the German territory or the whole domain?

L347: … of this RCM …

L355: in the authors’ opinion, why do we see these results with the Tiedtke scheme? However, from Table 2, I observe that also RegCM uses Tiedtke.

Figs. 4 and 5: I would merge them. In general, I suggest always considering as two different configurations WRF@5km and WRF@15km (see my comment to Section 4.3).

Section 4.5.1: E-OBS drought frequency looks too high for some areas (up to 22/24 times over 30 years). What is the reference period on which the index is calculated? 1980-2009?

L414: so, it is mean absolute error (MAE).

Sections 4.4 and 4.5: discussion, specifically in terms of comparison to existing literature, is mostly missing.

Conclusions: maybe, the main achievements of the paper can be highlighted with bullet points (e.g., increased resolution and setup are not useful for drought characteristics; they are helpful for correlations; they are useful for trends, etc.).

The manuscript describes an evaluation study of EURO-CORDEX hindcast simulations and additional WRF simulations with focus on the SPEI drought index. The methods are clearly described and mostly sound. In particular, the authors discuss the potential benefits of higher spatial resolution and regionally-tuned model setup, which may help to design efficient yet suitable model setups in future studies. The manuscript is suggested for publication after minor revision as described below.

Minor Revisions suggested:

Please provide your motivation for the model selection you applied (only 6 simulations). E.g, I’m pretty sure that EURO-11 simulations with several WRF configurations are available since many years. Why are they not used in this study?

In most practical applications, RCM simulations are only used after bias correction. Did you analyze the effect of bias correction (of temperature and precipitation) on your results? Please comment.

I would like to see a critical discussion on the practical relevance of the SPEI in climate change (i.e. trend) studies. Since NHESS has a focus on hazards and the hazard is rather the agricultural drought than the meteorological drought, the question arises to which degree the SPEI is able to describe drought hazard and what the of the SPEI to serve as proxy for agricultural droughts are. E.g., SPEI does not cover the effect of increasing surface runoff during heavy precipitation events, which, however, is a loss in the soil moisture budget. Similarly, SPEI does not regard increased transpiration due to longer vegetation periods in a warmer climate. I respect that this is not the main topic of the manuscript and therefore cannot be treated in a quantitative manner, but it is a very important boundary information and should be discussed in the conclusions and/or introduction sections.

L295: “Therefore, it can be assumed that if existent, the benefits of a resolution increase from 5 to 12.5 km are less distinct,” A single simulation does not allow drawing such general conclusions. This statement refers also to other conclusions (e.g. line 321, 322).

L329: “Because of the results in the previous section, here we focus on the values from the WRF@5 km run in direct comparison to the reference values from OBS (Figure 3).“ In the previous section, you argue that the 15km WRF simulation outperforms the 5km simulation, at least for temporal correlation. How does this lead to the decision to use only the 5km simulation in the subsequent section?

More generally speaking: By (partly) removing the 15km WRF simulation from the analysis, you lose the option to directly compare it to the WRF 5km simulation and therefore you lose the most direct indicator for the effect of the spatial model resolution. On the other hand, major results of the study are statements like “computation resources could therefore be saved, since a coarser resolution can provide similar results” (Abstract) or “WRF’s increased resolution and setup is turned out to be beneficial in the analysis of the monthly values of the meteorological variables and the correlations of the SPEI time series”. Therefore, I suggest keeping both the 5km and 15km WRF simulations in each part of the analysis in order to better support your conclusions. (Side benefit: The additional Fig.5 could be avoided, if you had the 15km simulation in the general analysis (Figure 4).)