General comments

This work presents novel and interesting results that will be of interest to the readers of this journal and the manuscript is well written. The authors have presented a well-informed investigation and conducted a thorough examination of relevant datasets. I can find no major problems. There are a number of issues and corrections that should be addressed before publications though I consider these minor and that the work is substantially complete.

The work is an event attribution study of the 2022 West-Central European soil drought, focussing on types of drought relevant to ecology and agriculture.  The authors perform a detailed review of the state of the science including an assessment of impacts and the work is well targeted in this sense. The chosen methodology is that of the World Weather Attribution group, designed for rapid attribution and hence highly prescriptive but also by now applied in publication a number of times. This study is complex, providing analyses of a large number of different models of various types, different estimates of observed soil moisture, all over two different regions and with supporting analyses of temperature and precipitation. While clarification on points of methodology is sometimes required I judge that the application of the methodology has been appropriate, successful and useful. The authors are at pains to emphasise the sources of uncertainty in the analysis, not least of which arise from the observations or observations-based products.  This uncertainty is explored at length and conclusions have been well phrased in light of this. The authors provide a convincing argument that robust yet conservative conclusions about the change in soil drought can be made and highlight that with lower confidence stronger statements are possible.

Specific questions

1. L313 “three multi-model ensembles using different framings”. These models are all used as part of a single multi-model ensemble and (as far as I can see) are treated equally. As this involves a parametric fit to all of the residuals to the same covariate then the framing referred to appears to be the same for every model. If I’m right then the statement would better be along the lines of “three multi-model ensembles of different types of models that are combined into a single multi-model ensemble treated under the same framing.”
2. L316 Use of SSP5-8.5 scenario. Given that use of this scenario can be criticised for certain uses, it might be worth a comment that as the future scenario is based on a fixed warming level (2°C) not tied to timing of when that level is reached, then the use of this scenario is defensible.
3. L324 So is the AMIP model over-represented by 10x compared to the CMIP6 models (which use 1 member each) or is there an inverse ensemble size weight applied to each model?
4. L348 Use of GWI. Obviously other estimate of the change in GMST attributable to human influence are available, though I don’t expect this to make a substantial difference to the conclusions. Has GWI been used for this purpose in another study using the WWA protocol?
5. L355 Fit of Gaussian to soil moisture residuals to covariate. It is stated here that all the “variable of interest are reasonably Gaussian distributed” but I can’t find a goodness of fit assessment other than what is implicit in the return level figures. (Note this is a different matter to how well the fitted parameters agree between observations-based products and models.) Looking at the fit uncertainty lines displayed in, for e.g., Fig. 4 I’m not very concerned that the tails are badly represented, though a comment on the fit shown in the left hand panel of Fig. 5(a) (ERA5-Land, NHET root-zone) might be due as there is obvious non-trivial behaviour suggested in the residuals to the covariate.
6. L369 Here it is stated that the Gaussian for precipitation scales with GMST, while in the caption to Figure S3 states constant dispersion parameter (which it would be helpful to define). Meanwhile evaluation tables S1 – S4 state a single value of sigma for precipitation. Given that the large area mean precipitation is being modelled as a Gaussian I am not entirely sure what to expect but a simple shift with GMST, like the temperature, might be reasonable. Indeed Fig. S3 left panels seem to suggest a simple shift with GMST. Which of these three options (just scale, shift and scale or just shift) is being applied to the precipitation in this analysis? Is this a deviation from or addition to the WWA protocol?
7. L386 The phrase “natural variability” refers to internally generated variability here. I appreciate that this phrase is in conventional use (and used as such in the protocol paper, Philip at al., 2020) but always feel it is bound to cause confusion with “natural variability” from solar and volcanic external forcing, which is most certainly not internally generated. Maybe a parenthesis “(internally generated)” or some such would be helpful to readers familiar with other uses of the term “natural”.
8. L604 Only reference to important begged question: why is the choice of covariate GMST rather than NHET or WCE mean JJA mean temperatures? Is there a physical motivation the authors can make that global annual mean temperature is an appropriate covariate to choose? Without some defence of this choice we could imagine that a soil moisture response to a different temporal evolution in JJA could change the results. I have taken a look at NHET mean JJA mean temperatures (on Climate Explorer, using GISS 1200km monthly fields) and can see that the temporal evolution is so similar to annual mean GMST that I doubt using this as a covariate would make any difference to the results, so I am not concerned, but a brief defence should be made.
9. L629 This section is very brief and you have to strain your eyes a little to see the origin of the numbers “2-30 fold” for projections in Fig. S13 – S16. I can see combined model best estimates from around 1.5 to what may be around 30 in Fig. S16 if I read the scale right. A statement of the values in the text would make this more transparent. It might also be worth noting that, again, PR lower bounds include 1 in every case.
10. L629 §11.6.1.3 of AR6 (AR6 chapter 11) points out that “There is evidence [in ESMs] that surface soil moisture projections are substantially drier than total soil moisture projections, and may overestimate drying of relevance for most vegetation”, citing Berg et al., (2017). This contrasts with the state of affairs pointed to in L610 pertaining to historical times described as “largely consistent“. I note that the largest PR in the projections is for surface, NEHT, perhaps suggesting that the models used here concur with this expectation. Is this the case?
11. L634 Despite stressing that the study uses three different types of models (coupled, atmosphere-only and high resolution) there is no discussion of what these different model types have contributed and whether the evaluation and results appear to vary between model types. I appreciate the analysis is already complex but a comment on whether there is anything obvious going on would help. Referring back to my comment on L313, as these different models are all being treated under the same framing then perhaps this lessens the importance of combining different model types. It is also relevant to my comment to L629 concerning the separation of surface and soil moisture projections in ESMs.
12. L640 PR of “summer drought” in WCE is “about 5”. Is this a rounding down of the mean over synthesised best estimates of the 1850- and 1950- WCE root-zone analyses (0.5*(2.8 + 8.8) = 5.8)?
13. L643 PR inc. 1, “this also applies to the northern extratropics” – but Fig. 8 shows that NHET root-zone PR has both lower bounds > 1.
14. L668 Again the origin of the numbers requires a bit of work, given that brief section 5.5 simply stated a single PR range of 2 – 30 concerning four analyses. I think that the numbers ‘10 years’ and ‘every year’ have been arrived at by dividing 20 year return times for each region by the average of PR estimates over surface and root-zone best estimates (about 2, and about 20). Is this the case?

Technical corrections

L115 “crushed” – typo?

L240 “of” à “over” or “for” would be better?

L337 Statistical methods: this section could do with a sentence stating that the 2022 event is characterised by taking the return time of the event in the observations-based products and then querying the model distributions at the corresponding return level. Those already familiar with the WWA method will know already but I can’t find that this essential aspect of the method is explicitly stated anywhere. It may also then make more sense why later in L476 we are happy to take the two estimates for the observations based return times and simply average and round them.

L401 Typo: I think “Note that as to” should be “Note that so as to”.

L483 Type: I think “inform on” should probably be “concern”.

L497 I had to look “Mio” up, it being short hand for 1 million in German. Might best be just “million”?

L499 Type: “2002” should be “2022”.

L504 “excessive” seems a strange adjective. Suggest “very large”?

L548 repeated use of “fairly” is ambiguous language.

L583 “temperature surpluses” is a strange phase, something like “positive temperature anomalies” would be better.

L592 Apparent error in statement of uncertainty bounds for NHET change in intensity compared to Fig. S9.

L614 “Appendix” here I assume refers to the SI, Fig. S17 and S18.

L629 Typo: “0.8oC” should be “0.8°C”

L649 declining root-zone soil moisture in the 21^st century in “the entire northern extratropics” – we don’t have a figure showing this. Perhaps a reference to the literature?

Table S6 units for temperature intensity change: are the numbers really in % or are they °C? The numbers seem consonant with L489.

The quality of some figures is too low, e.g. Fig. S9 and S10, perhaps Figures 4 and 5 as the 2022 data point is hard to see.

Table S4 is for surface soil moisture but the column heading says root-zone.

References

AR6 chapter 11: Seneviratne, S.I., X. Zhang, M. Adnan, W. Badi, C. Dereczynski, A. Di Luca, S. Ghosh, I. Iskandar, J. Kossin, S. Lewis, F.  Otto, I.  Pinto, M. Satoh, S.M. Vicente-Serrano, M. Wehner, and B. Zhou, 2021: Weather and Climate Extreme Events in a Changing Climate. In Climate Change 2021: The Physical Science Basis. Contribution of Working Group I  to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R.  Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 1513–1766, doi:10.1017/9781009157896.013

Berg, A., J. Sheffield, and P.C.D. Milly, 2017: Divergent surface and total soil moisture projections under global warming. Geophysical Research Letters, 44(1), 236–244, doi:10.1002/2016gl071921.

Climate Explorer: https://climexp.knmi.nl/start.cgi

Philip, S., Kew, S., van Oldenborgh, G. J., Otto, F., et al.: A protocol for probabilistic extreme event attribution analyses, Adv. Stat. Clim. Meteorol. Oceanogr., 6, 177–203, doi:10.5194/ascmo-6-177-2020, 2020.

The author conducted attribution analysis of a soil moisture drought event in West-Central Europe in 2022 based on a large number of soil moisture estimates. It is found that the return period of the drought event is decreased substantially in current climate as compared with that in pre-industrial climate, although there are uncertainties in soil moisture estimates. The manuscript is well written and easy to follow, and there are only a few minor comments below.

1. GLDAS_ CLSM data includes v2.0 (1948-2014) and v2.1 (2000-2022). Is v2.0 and v2.1 used together in the article? How are they merged? Have corrections been made before merging? Only the June data was used in 2022? How about July and August?
2. Why does CMIP6 use SSP585 and historical scene splicing (usually SSP245), while CMIP5 uses RCP4.5 and historical splicing?
3. The uncertainty range for the surface and rootzone soil moisture anomalies (Figures 1a and 3a) could be shown in the supplement materials.
4. What is the definition of Intensity in this study?
5. Line 479: Does -9% (-13%.. -4%) refer to a change in intensity? Please explain in detail.
6. What is the physical meaning of representation error? Perhaps only when the minimum value of Representation error is significantly greater than 1 (less than 0) can the impact of anthropogenic climate change on Probability ratio (change in intensity) be considered significant?
7. There are a total of 25 models, but only 7 models in Figures 6-7 have passed the test. How about using all models? This may increase the reliability of the results.
8. In Figures 6 and 7, the attribution results of the probability ratio are not significantly greater than 1 (the minimum value of the confidence interval is less than 1), which does not seem to suggest that the impact of anthropogenic climate change on drought in 2022 is significant. Please explain.