• The authors examine convective outflows in ICON convection permitting simulations and with parametrized convection over Germany. Overall, the authors have made a substantial effort in applying and explaining their interesting method and also made good efforts in statistically evaluating the significance of their work. However, at the end, especially concerning the 2nd half of the manuscript, the reader is still uncertain of what is the main outcome of the paper, are the results robust (small sample, noisy divergence fields, need to apply to situations with larger systems), what is the main difference between explicit and parametrized convection and why should momentum transport effect divergent outflow. Therfore I would suggest some major revisions along the following lines:
• pleae make sure from the beginning, already in the Abstract that you are dealing with Ensembles (otherwise reader, seeing just one case would think this is not significant)
• your sample size is still very small. From the modelling point of view it should not be much work to add more days - but dont know how much additonal work it is for your classification method. If you can do so, please add at least a few days with larger systems, this would be easily achievable for US systems or African squall lines with large and long-lived systems, but you can also add more days over Europe with prefrontal convection. You can still keep your main figures for Germany but for the scatter plots this would be extremely helpful. As it is I dont think you can quntitively talk about cMT or divergence in unorganized vs 2D squall line or Derecho type convection
• instead of using precipitation, could it be more useful to use low values of OLR or BT to charcaterize outflow, this might produce smoother results and better identify the upper level circulation?
• I am unsure it is helpful to normalize momentum transport by precipitation, yes it is natural to normalize divergence by heating but moemntum transport should be normalized by shear
• I found section 6 of your manuscript not convincing and not well written, please redo
• also in Figure 6a the divergence is so scattered, in the explicit run, corresponding to small scale systems="noise" in Figure 5b that I dont think it makes sense to discuss linear regressions (couloured lines) as a function of lead time or system evolution. In that sense it would even make more sense to discuss better the parametrized runs with larger outflow

Specific points:

• l12-13 "as opposed to expectations, CMT". I think this part also in result section should be revised or could be dropped in absence of sufficient cases/samples with shear
• l19 "initial condition flow variability". In you manuscript you actually use physics perturbation ensembles but never talk about it in the introduction
• l76-86 you formulate here 3 hypotheses. you might drop some of these in absence of larger samples and systems, as the main (also expected outcome) is a broadly linear relationship of divergent outflow and precipitation or better mass flux  which should the also hold for laregr are (opposed to individual system)
• l188 "satellite systems" avoid this expression as "satellitee" misleading here
• l195 "vertical integral of CMT over all levels below". Not sure this is intended as over the whole troposphere cMT integrates to zero so in that case if you decide to keep CMT you might need to consider upper and lower troposphere separately
• l195-200 normalize CMT with precipiattion, see main comments
• l227 you migth add reference to https://doi.org/10.1002/qj.4185    here as actually the organisation is not so simple and sometimes too spotty in explicit compared to parametrized convection
• Also you might add refernce in the text to studies that explained that the convective outflow only has an effect to the circulation and subsequent predictability if the outflow occurs in the vicinity of the jet
• l297 "increase over time" this looks like the diurnal cycle evolution
• Figure 5: add (a), use same legend bar for (b) as for (c) and (d). Here and in Figure use everywhere units of [1/s] for divergence (not multiplied by density in some places, eg Figure 6) and scale legend bars by 1.e5
• Figure 6a, coloured slopes not convincing. Figure 6 b more convincing, but why is in Figure 6b shallow (magenta) so different form no convection run (organge)?
• l439-440 Why do you say "larger-scale flow feedbeck in PAR likely not accurately represented"? Isn't supported by anything in manuscript and how do you know that PER effect is not overestimated etc?
• As said above best would be to add more samples (days in Europe US or Africa) and also to rewite section 6 and adapt all Figure from Figure 6