General Comments:

This study aims to develop a proxy for hail frequency, as determined by ESWD reports, based on high-resolution regional reanalysis atmospheric properties coincident with satellite OT detections. The authors demonstrate the proxy’s improved capability in matching with hail reports relative to previous similar studies performed with coarser global reanalysis, although, by their admission, determined with focus on qualitative aspects. The thorough explanation of SPHERA variables and description of robust quality analysis are appreciated.

The concept of the filter could be introduced earlier or more gradually – it felt as though we were suddenly discussing the filter as a given before it was properly explained. That said, it is not convincing that the goal of obtaining a new proxy for hail has been satisfied quantitively. That is, the filter approach may be overly conservative compared to a random forest or other machine learning approach. It is perhaps good at removing cases where hail is unlikely, but not capable of skillful positive predictions. This task of separating events by their hailstone sizes through environmental analysis seems better suited for perhaps a logistic regression or a Bayesian analysis approach – as the layering of independently determined statistical filters seems to be too broad of a solution. Furthermore, adding additional years seems necessary given the high variability of year-to-year hail reporting, as a means to alleviate overfitting concerns, and to help the environmental stratifications.

I recommend major revision with consideration of the specific comments listed below. Furthermore, the paper can benefit from grammatical revision – some examples of which are listed as technical corrections.

Specific Comments:

Lines 101-108: The method for coupling satellite, SPHERA, and hail reports should be previewed somewhere around here in the introduction. “optimally tuning the combination” is mentioned, but in what kind of approach? Furthermore, “optimally tuning” is stated, but that seems to extend as far as layering 5 statistical thresholds, which does not suggest tuning.

Fig 1/Line 137/Line 371: 1 h is said to be the cutoff for “high temporal accuracy,” but what is the justification for this designation, as it seems like a rather large uncertainty, especially compared to the significantly greater sampling at 15-min temporal accuracy? What is the expectation for how 15-min vs. 1-h temporal accuracy would impact the hail environment characterizations?

Line 139: “A substantial increase is noted for 2019 and, to a slightly lesser extent, for 2020.” – Is this natural, or owed to ESWD hail reporting reasons? Showing more years may reveal more about the expected year-to-year variability of reports.

Lines 193, 200, 237, 238: “… is included in the filter” – Again, reference to “the filter” and “full filter” is presented as a given without establishing what exactly is meant by the word. Reading further into the manuscript helps fill in and confirm the reader’s intuition, but it would be helpful if there was a more explicit introduction to the filter concept – perhaps as a paragraph at the start of Section 3.

Line 257: Please explain what is meant by “the limitation imposed by choice of the H₀ threshold.”

Lines 264-266: It isn’t very convincing that the regions highlighted as having lowest OT removal are distinct from “noise”. That is, the Gulf of Taranto and areas southwest of Sicily (among others) look comparably shaded to the ones the authors mention. Perhaps a discretized color bar in increments of 20% would help, but otherwise these discussion points should be removed or reframed in a way that’s better suited to the image.

Lines 271-272: There should be discussion of how the number of OT detections change over these same times of day and months, and how that may influence the interpretation of these fractions.

Lines 360-363: It’s not seen how “full independence of the two datasets” equates to no concern for overfitting. The authors are relating observed hail reports to coincident environmental analyses, that have spatial dependency, and then evaluating spatially dependent, environmental-analysis-filtered storm signatures. So it isn’t convincing that “full independence” applies. The claim should be tested by doing the method but with one evaluation year held out from consideration for the filter. Perhaps even doing that for each year and evaluating the average performance. Adding additional years would also help in addressing this concern.

Table 2: Was a similar Table for day vs. night and/or month-to-month performance considered?

Line 416: Please explain how it is decided which report with duplicate environmental conditions is kept.

Line 461: Please justify why “as expected.”

Lines 487-488: How can this conclusion be offered when the assessment used OTs to characterize the ΔT in missed cases? That is, if the OT are there and measured to be systematically warmer for missed cases, how can it be said that prominent OTs were not detected to characterize the events? This conclusion should be more clearly explained or removed. Furthermore, the preceding sentence requires grammatical revision, specifically “… are failed to be identified.”

Technical Corrections:

Line 24: Though understood, the phrase “… and so its water holding capacity;” reads awkwardly in this context.

Line 33: The phrase “… makes its observation still a major challenge…” probably can drop the word “still.”

Line 34: Similarly, the sentence with “… hail observing system is still missing” could benefit from a revision that does not include the word “still.”

Lines 35 and 86: Change “including” to “that include”, otherwise it reads awkwardly.

Line 43: The phrase “… constitute a precious way…” is distracting. Consider rephrasing and without use of the word "precious".

Line 73: “These are…” – What are? No specific parameter was mentioned yet. Again, this may just be a structure issue. Consider revising this sentence and the one preceding.

Line 88: Sentence requires grammatical revision.

Line 94 and 98: What does “description” mean in these contexts? As in “characterization” or “depiction”?

Lines 146 and 454: Suggest revision of the phrase, “…actually occurred OTs…”, e.g., “true OTs”.

Line 97: Consortium for Small-Scale Modelling?

Line 185: “Their different formulations… distinct parts of the numerical model equations…” – perhaps, though arguable. The sentence isn’t necessary – consider removing.

Line 223: It is unclear if “latter study” is referring to the approach of Punge et al. – check use of “latter” in this context.

Line 246: Perhaps “… filter mainly over certain areas of the sea…”, because largely the pattern across sea to land is rather continuous, especially for a) and b).

Lines 284 and 524: “Anyhow” should usually be avoided in formal writing. Furthermore, the structure of the Line 284 sentence, with too many qualifiers followed by an “empty this," makes it difficult to follow. Please revise.

Lines 465-466: The ending of this sentence does not read easily. Consider revising.

Line 470: Conclusions are typically separate from Discussion.

General comments:

The study addresses the challenges associated with reliably observing and simulating hazardous hailstorms. The authors propose an approach that combines information from different sources, including remote sensing instruments, observations, and numerical modeling, to enhance the understanding of the spatial and temporal patterns of severe hail occurrences in south-central Europe. The methodology involves developing a proxy for hail frequency by integrating overshooting cloud top (OT) detections from the Meteosat Second Generation (MSG) weather satellite with convection-permitting SPHERA reanalysis predictors describing hail-favorable environmental conditions.

While the paper is already quite robust, there are a few shortcomings which should be addressed by the authors to enhance the readability and importance of their work.

I recommend major revision with consideration of the specific comments listed below.

Specific comments:

Section 2

While the methodology is generally well-described, ensuring greater clarity would enhance the paper's accessibility. Providing, for example, a flowchart for data processing and analysis could be beneficial.

Section 2.2

I am wondering if you considered using the 5-minute rapid scan data from MSG, which may improve the quality of this paper, since you state in lines 538-544 that the low temporal resolution is a limitation of your work.

Section 2.3:

Why hasn’t the low-level moisture been added to the filter? You clearly and rightfully state that it is an important factor for hail formation, so I don’t really understand why the moisture wasn’t considered for the filter.

Moreover, a recent study showed that CAPE above the -10°C-isotherm stood out as the best predictor for Europe (Battaglioli et al. 2023). Another recent study (Nixon et al. 2023), for the US this time, showed that the depth of the storm (“maximum parcel level”) and storm-relative winds below the hail-growth-layer may play a key role in formation and size of hailstones. Hence, I would suggest rethinking the choice of ambient predictors.

Section 5:

I would suggest splitting the discussion and conclusion. The current content of this section is not well structured, and the take-home messages are not clearly outlined.

Providing context for the practical applications of the research would also highlight the significance of this paper (e.g., risk assessment, insurance).

Figure 1:

Large hail is defined with “≥ 2 cm” in the ESWD (and not “≥ 3 cm”). So, this error needs to be addressed throughout the manuscript to maintain coherence.

Technical corrections:

Line 203: “intrinsic”

Line 519: “large hail-producing storms”

References:

Battaglioli et al. 2023 (https://doi.org/10.1175/JAMC-D-22-0195.1)

Nixon et al. 2023 (https://doi.org/10.1175/WAF-D-23-0031.1)