Thanks very much for your comments. We very much appreciate the time and effort the reviewer dedicated to revising our manuscript. We agree with most of the reviewer's feedback and will address them in detail in the revised manuscript. We will briefly address some of the reviewer's comments at this stage. 

2. Section 2.1 Page 3, lines 85+ ‘We consider only simulations...’

- There are some issues that can be considered related to the use of soil moisture in different models: The models have very different depths and therefore they will possibly produce different soil moisture statistics, even if depths only down to 0.7 m are considered. Using this set up of models is fair and possibly it contributes to inter-model differences, but also it is arguable that models with shallower depths may be less realistic. Note that some models like GISS only have a depth of 3m. This will produce potentially a different vertical distribution of moisture and different temporal variability. Perhaps it is something the authors would like to comment on.

Thanks very much for your point. We agree with the reviewer's comment regarding the need for an additional description of soil moisture for each model and a discussion on inter-model differences in our study. One point that we want to comment on is that soil depth below two meters is generally considered less important for atmospheric processes. Hence, the fact that GISS has only a three-meter depth would not significantly affect the atmospheric processes we focus on. However, vertical soil layers (up to two-meter depth) could be a more important factor that contributes to the observed model differences. We briefly discussed this issue in the discussion section (Section 5) relating to the discrepancy between MIROC-ESM and other models (lines 366–373), but we will extend this discussion in more detail in the revised manuscript.

3. Section 2.2.

Why are NOAH-LSM data used for soil moisture and ERA5 only for circulation? I mean what does NOAH-LSM offer that you would not get in this manuscript from ERA5 soil moisture?. There is probably good reasons for it. I would just suggest that a motivation for the use of these data, beyond the fact that soil moisture observations are scarce, is provided.

We used the soil moisture from NOAH-LSM instead of ERA5 because NOAH-LSM is forced with the observation-based dataset and the reanalysis data that the biases were corrected with respect to the observations (https://hydro1.gesdisc.eosdis.nasa.gov/data/GLDAS/GLDAS_NOAH025_3H.2.1/doc/README_GLDAS2.pdf). ERA5 does not directly assimilate any rain-gauge data except for the United States (Lavers et al., 2018). Therefore, we assume that NOAH-LSM could be more appropriate to show more realistic present-day soil moisture variability, which is mainly influenced by precipitation variability. 

-  Line 87 ‘... 12 ensemble members of CESM1’

 There are 13 available if I am not wrong. Nothing critical but maybe you want to state why that selection. Here it can also be stated this refers to the all forcing simulations, although it is quite clear in the context (see next).

Thanks very much for the observation. When we retrieved the CESM-LME dataset from https://www.earthsystemgrid.org, the first ensemble member (member 001) of the variable geopotential height (Z3) was missing for the period 850-1849. Hence, this member was not included in our analysis. 

- ‘We use annual mean anomalies in order to include winter conditions in the analysis, as it is an important season for the annual hydroclimate in the Mediterranean’. I think it is important to discuss this in the context of what is indicated in GC2.1.

Here we wanted to emphasize that we used the annual mean time series to account for the mean hydroclimate conditions from all seasons, including the wet seasons (referred to as only winter in our study, but we will correct this according to the reviewer's first comment.), unlike other studies in the region, which primarily used the summer mean time series. We will clarify this better in the revised manuscript. 

About Section 3 (Method)

As both reviewers commented, we will re-organize and correct the section to clarify the methodology in the revised version. 

Sincerely,

Thanks very much for your comments. We very much appreciate the time and effort the reviewer dedicated to revising our manuscript. We agree with most of the reviewer's feedback and will address them in detail in the revised manuscript. We will briefly address some of the reviewer's comments at this stage. 

Before I list specific points in the manuscript, I wanted to raise the issue to the authors (it is not a must, more a proposition): I personally found Fig. 2 and 3 very interesting but under- utilised (and to some extend Fig. 6 too). Clearly, some models are not very skilful at capturing the synoptic climate during historical droughts and this raises in itself an important issue: how can we confidently understand drivers using longer simulations? What setup does “work better” to capture what part of the signal (temporal vs. spatial)?

Thanks very much for the comment. There are several studies analyzing CMIP5 models' ability to capture principal modes of variability (e.g., Fasullo et al., 2020; Deser et al., 2018, for NAO), and biases (e.g., Davini and Cagnazzo, 2014) compared to the observational-based data. We will argue that if the models are able to represent well the present-day climate modes in an acceptable way, this would also be valid for the past period. Also, the same biases will be present in the past simulations. We will discuss this issue more extensively in the revised manuscript. 

Perhaps I am not familiar with the term, but it would be useful to clarify what you mean by “internal variability” (and that might be done by just explaining what is considered an external forcing in your study).

With internal variability, we refer to different modes of variability involved in droughts. By external forcing, we mean volcanic and solar forcings. However, volcanic forcing is an internal climate forcing and not an external forcing. We will clarify better what internal variability and forcings mean in our study in the revised manuscript. 

- Lines 286-288:I struggle with these sentences. Either too little or too much is said here. What is the role of models setup and skilfulness in this observation? I also do not understand what is meant by “counterfactual”?

By "counterfactual", we refer to the climate conditions without the anthropogenic trends from 1850 onward, so the detrended Historical simulations. We will modify this word and also rephrase the sentences for clarification. 

- Fig.4: Would it be possible to quantify the antiphase? (bi-plot, cross- correlation) 

Thanks for the suggestion. We will add more analysis in the revised manuscript. 

- Is there any way possible to test the observed seasonality pattern of climatic associations in the OWDA? Do you also observe an N-S antiphase (also mentioned in Markonis et al. 2018)? The E-W antiphase: is it stable on all timescales (Indeed the results you obtain are contradictory to Cook et al. 2016, this might be further discussed)?

We did not include the time series of OWDA, as our focus is on how the climate models represent droughts and associated circulation in the Mediterranean region. But we will add the analysis of OWDA in the revised version. About the N-S antiphase in Markonis et al. (2018), in my understanding, the N-S antiphase is between northern Europe and southern Europe. As our study covers only southern Europe, we would not be able to observe a similar result to Markonis et al. (2018).

About Section 3 (Method)

As both reviewers commented, we will re-organize and correct the section to clarify the methodology in the revised version. 

Sincerely, 

Dear Reviewer, 

Thanks very much again for your comments. Please find our final responses in the supplement. 

Sincerely, 

Woon Mi Kim on behalf of all co-authors. 

Dear Editor Dr. Goosse, 

Thanks very much for dealing with our manuscript and for allowing us time for the responses. We added more comprehensive responses to the reviewers' comments in AC3 and AC4. 

Sincerely, 

Woon Mi Kim, on behalf of all co-authors