General Comments

This study investigates how cloud seeding influences hail via the microphysical pathways and processes responsible for hail growth.  This is done using a modeling framework developed and published in a prior study and expanded upon here for a larger variety of events to also test the dependence on synoptic environment.  Despite ice and graupel showing evidence of the beneficial competition theory, increases in hail diameter were seen, however a reduction in hailfall area was also seen.  The paper is well-written, with a comprehensive introduction and methods sections.  The analysis is detailed at walking the reader through the relevant microphysical processes and relating changes in these to their figures.  Overall, there are no major reservations about the paper, its methods, or conclusions.  Several minor corrections are suggested for clarity.  One request is made to provide more in-depth analysis on the convective core to see if this provides better clarity in the results.  

Specific Comments

1. Was there analysis (akin to Figures 4-7) on changes in the hail growth zone specifically? If not, paired with a higher updraft threshold (see comment 2), I’m wondering if this would illuminate some of the divergence seen in the hail results specifically regarding the increase in diameter.  I believe it may be beneficial to at least see if changes in the convective core differ from those seen when looking at the whole of the storm.     
2. Analysis of hail was conducted using the 1 m s^-1, and while Figure 12 shows that changing thresholds don’t alter the main conclusion (i.e., changes to hail remain inconclusive), using a 1 m s^-1 threshold doesn’t seem appropriate for hail given that updrafts on the order of 10 m s^-1 are generally considered the minimum needed for hail growth (e.g., Knight and Knight 2001). As such, basing the conclusions primarily on this— again even though it doesn’t seem to make a difference in this study— is somewhat misleading, given that when looking at larger hail, the processes discussed take place in stronger updrafts (convective core).  That said, the analysis with the 1 m s^-1 threshold is still incredibly valuable, providing insights on the graupel to hail transition and tracking the result of smaller, numerous graupel on hail (and as seen in Table 1, the storms used in this study have weaker maximum updraft speeds).  I think acknowledging that this updraft speed is generally insufficient to support hail growth and referencing section 3.5 where this is discussed in more detail would be fine.
   • Knight, C. A., &Knight, N. C. (2001). Hailstorms. Severe Convective Storms, Meteorological Monographs, 28, 223-248.
3. Lines 156-160 imply that differing terrain will be evaluated in addition to environmental conditions however this is not the case. Would suggest re-wording.
4. Section 3.2: the inclusion of this section comparing modeled and observed rain rate is unclear and does not seem necessary, especially given the study is evaluating changes in hailstones. I could only see this section be somewhat relevant if changes in the rainfall rate/accumulations as a result of seeding (and to see if the melting of smaller, rimed hydrometeors potentially led to increases) were also investigated.
5. Section 3.3 & 3.4: All analysis is implied to be conducted in the bounding boxes seen in Fig. 1 (constructed using the tracking threshold of 1 m s^-1 and AgI concentration of 150 L^-1) however I think this needs to be emphasized better.
6. Line 200: changes in updraft strength were analyzed (as seen in Fig. 4) but were changes in the updraft area (in the context of Dennis and Kumjian 2017) investigated at all?
   • Dennis, E. J., and M. R. Kumjian, 2017: The Impact of Vertical Wind Shear on Hail Growth in Simulated Supercells. J. Atmos. Sci., 74, 641–663, https://doi.org/10.1175/JAS-D-16-0066.1.
7. Figure S1: a bounding box outlining the study domain (seen in Figure 1) would be beneficial to include.
8. Is there a potential that storms down-wind of those seeded could be influenced?  

Technical Corrections

1. Figure 1: caption does not detail what the grayscale shading represents (assumed terrain).
2. Line 261: a comma and period appear at the end of the sentence.

This manuscript presents a numerical study of cloud-seeding impacts on hail suppression in convective storms. The authors use the COSMO model to simulate eight convective storms over the Alpine region, with ten perturbation ensemble members per storm. Early-stage AgI seeding and threshold-based storm tracking are applied to evaluate microphysical responses and their dependence on CAPE, wind shear, vertical wind, and AgI thresholds. The manuscript concludes that seeding generally increases hail size while reducing hail-covered area, and reports no consistent relationship between seeding impact and environmental conditions (CAPE/wind shear).

Overall, this study provides interesting scientific implications that could benefit the operations of convective cloud seeding. The multi-case ensemble design is a valuable framework for identifying model sensitivity, and the reported median behavior is broadly consistent across cases despite large inter-member spread. However, there are a few methodological descriptions is a little bit hard to follow or lack technical details, especially for the seeding implementation and averaging/threshold workflow.

Major comments

1. The description of the seeding procedure is not sufficiently clear and needs substantial clarification (lines 130-138).
   
   1. There are a few ambiguous terminologies:
   
   >Please define "target cell" (used for maximum-updraft identification). How is it different from the seeding box? And also clarify whether the target cell and the seeding location move over time as the storm evolves.
   
   >What is defined as "cumulus stage"? Accordingly, when is the actual seeding period for each storm?
   
   >The temporal release process of AgI seeding is unclear and needs clarification. How is AgI released during the 10-minute window (constant rate or time-varying)? What is the release rate? The manuscript states that t"he seeding area was taken to include all three boxes from the different time steps"; does this imply discrete release at output times rather than continuous release? Correspondingly, the manuscript reports a peak AgI concentration at the end of seeding, but additional release-process details are needed to interpret this number physically (is it due to accumulation over time or increasing release rate, etc).
   
   >The statement that seeding was "mirrored" from the 2025 study appears inconsistent with the current wording. In the current paper, s"eeding was applied for 10 min to five model levels immediately below and at the cloud base", while in their 2025 paper, a"t 1200 UTC, seeding is carried out across five distinct model levels underneath the cloud. ... Considering the observed reduction in cloud-base height between 1200 and 1210 UTC, seeding after 1200 UTC is conducted in model levels at or above the cloud base").
   
   2. The seeding box is a 5x5x5 grid boxes (5.5km x 5.5km horizontal with 5 vertical model levels), which is very large. Operationally, aircraft release is close to a point-source. What is the justification of using this big box instead of a single grid box? This set up may over-broaden the AgI dispersion. The manuscript acknowledges this limitation, but this could be improved (though not perfectly) by using one grid box instead. Have the authors conduct any sensitivity of the seeding impact to the box size? Given the chaotic nature of the convective storms, this treatment may substantially affect the spatial distribution of the AgI and thus the seeding signal.
   
   3. Actual time-window of the simulation and seeding start/end time are missing. The paper stated all simulations start at 0 UTC to cover early day storms, but no information of the actual storm time was given: how many early day storms are selected? What's the simulation length for each storm: is it same for all storms or ending differently based on the actual storm time? What are the identified start/end seeding time for each storm?
2. Figure 2 does not show good model-observation agreement for precipitation evolution, what can it tell us, to what degree can we trust the results if it does not represent the precipitation trend and magnitude in reality? How would that potentially affect the interpretation of the simulated hail intensity and therefore seeding impacts? This type of discussion would be critical before jumping into the simulated seeding impact interpretation.
3. In Figure 6, most hail does not reach the surface in CTRL (very low number concentration << 1/m^3), although hail was observed in these events. So are these simulated number physically realistic and expected for the selected cases? In other words, is there a typical number concentration or threshold to define a hail event? Since the purpose is to suppress hail, it makes sense to realistically simulate hail reaching the surface. I would also suggest to use log scale in the x-axis to better illustrate the magnitude since hail at the surface is the most important metrics.
4. Figure 8 caption requires clearer metric definition:
   1. Doe values reflect ensemble mean of the domain-time average for each case? Given the nature of convective storm, that would average out too many physically meaningful structures and stage-dependent signals. Domain average will smear out the local signals, calculate the values based on sub regions (such as your target cells) will help identify seeding signals reponding to local CAPE/wind shear.
   2. Averaging over the entire simulated period mask out the different cloud seeding response to storm at different stage (cumulus, development stage, mature stage, as well as dissipating stage), and for hail formation, usually the mature stage will be the focus.
   3. Ensemble mean is not a great matrix to look for relationship, given members exhibit a large spread. Please consider showing all members. Also an average over the entire domain would smear out the local signal.
   4. This also gives you more sample points with physically meaningful representations.
5. Figure 10: Please clarify how time-average values are computed? Did you first compute the time-average and then apply thresholds, and then average over these grid points, or first apply thresholds at each time, and then then take time and grid average? These can produce different results as vertical wind and AgI number vary over time and location. Correspondingly, Figure S3, please clarify how AgI thresholds are identified for CTRL since it does not have AgI? If locations are identified in SEED but values are sampled in CTRL, what is the physical meaning of these classification? (showing SEED results will have physical meaning but CTRL value based on the ghost AgI concentrations does not make physical meaning)
6. Current analysis emphasize mostly on mean/medians/total. The manuscript would benefit from more from showing some spatial analysis, as a core strength of modeling over observations is resolved spatial structure. For example, for supporting key claims such as seeding reduces hail-impacted area, it would be helpful to show the hail impacted region (e.g., accumulated hail mass/number on ground) overlaid with terrain map before and after the seeding. This straight-forward comparison could also help understand how terrains play a role over the Alpine region.

Minor comments

1. Line 182: "icrease" should read "increase".
2. Line 218: "aproximately" should read "approximately".
3. Lines 70-71: remove duplicated "however" in the sentence.
4. Line 184: "2 of the 8 cases are no longer tracked".
5. Line 177 (Table 1 caption): "~5.5 km^2 box" is inconsistent with the main text (5.5kmx5.5km).
6. Figure 9 caption: "Same as Figure 5" likely should reference Figure 8?
7. Figure S1: specify whether pressure is surface pressure or sea-level pressure. The colorbar tick-label is very small (and blurry when zoomed in), please make the font larger for readability.