General comments

The authors present a study detailing updates to the AROME turbulence, shallow convection, cloud and microphysics schemes in an effort to improve the representation of shallow clouds and turbulent transport in the model. As tools, they choose large eddy simulations of four canonical shallow cloud cases to provide a reference, and test their model changes in a single column version of AROME. The High Tune Explorer is used to optimize the parameter space of the updated model.

How to adequately model turbulence and shallow convection at kilometer (and increasingly sub-kilometer) scales is an unresolved question that urgently needs answering as operational regional forecasting takes place at these higher resolution, fully entering the turbulent gray zone. I therefore find this study, which aims to address this topic in a framework that treats turbulence and shallow convection in a consistent manner, to be timely and appropriate for publication in ACP.

The model changes described consist of several incremental upgrades to existing schemes, rather than entirely new parameterizations, but this is generally the strategy pursued in operational NWP and does not, in my opinion, detract from the study’s relevance.

While I think that all the pieces are there, I would like to see the authors restructure and clarify their discussion (see specific comments below).

Specific comments

One weakness of the authors’ chosen methodology – comparing individual “golden day” LES and SCM simulations – is that model improvements seen for these idealized cases do not always translate into model improvements in the real world where conditions are rarely “ideal”. Including the SANDU transition case is a start. I don’t mind the authors’ strategy of focusing on a single case to demonstrate the individual parameterization upgrades, but I would like to see more emphasis put on demonstrating that the updates discussed in the context of one case are of benefit to all four cases (or not). A few places where this is not obvious are:

1. The authors state as a main motivating factor that AROME has large radiative biases associated with low clouds. Together with cloud fraction, the next most important cloud property for the cloud radiative effect is likely the condensate amount. The authors show improved liquid water content for the RICO case (Fig. 7d) from the new cloud scheme, but we don’t see the impact on LWP/cloud condensate for all four cases after the full implementation and parameter optimization (Sec. 4). While doing a radiation evaluation may be outside the scope of this paper, it would still be good to get a clearer picture how CF and LWP combine to potentially improve the radiation bias for the final version of the model. This would close the circle from motivation to conclusion regarding the radiation bias.

2. Why is the SANDU case not included in Fig. 13? To me, this case is of special interest as it includes both the well-mixed Sc case, as well as an increasingly decoupled Cu under Sc boundary layer towards the end of the simulation. Many models struggle exactly with this “in between” state, and a good performance here would demonstrate promise that the model changes will also lead to improvements in less idealized settings. The SANDU case is not included in any of the in-depth discussions in section 3, so I would like to see more emphasis put on this case at least in the discussion of the final model configuration after the parameter optimization. The total TKE in Fig. 13 would be great, a decomposition as in Fig B1 would be even better, to see if/how the balance of TKE contribution terms shifts as the BL decouples.

Additional discussion points:

• Sec. 3.2.1 What is the impact of removing the small updraft fraction assumption?
• It appears that the greatest impact on the diurnal cycle of stratocumulus was due to the limitation on the entrainment (not to exceed detrainment), which was removed as a preliminary step (Eq. 7, Sec. 3.1). The improved diurnal cycle is brought up in a few places as a major outcome of the model upgrades (L12 abstract, L620 conclusion), but receives hardly any discussion throughout the paper. If this fix is considered to be a major improvement, then I think the authors should discuss in a bit more detail how the old limiter impacted the EDMF scheme negatively, and what the MF transport looks like without the limiter. So far, there’s exactly half a sentence of explanation on L248. Alternatively, if the authors consider this to be on the level of a bug fix or algorithmic fix then this change should be de-emphasized as a main outcome of this study (though the impact is large enough I’d go with the first option).

Overall, section 5 (Discussion) does not seem to flow as well as the text in other sections, and needs reworking.

• It starts out with the statement that the original AROME did not match LES energy diagnostics well, and specifically that TKE was too weak for the ED component. The text (L567) suggests that this is a known, previously established problem in AROME. While Sec. 3 shows that the TKE budget is improved by the addition of the anisotropic MF contributions, the systematic TKE underestimate is not previously brought up as a starting point (either in the introduction, or in Sec. 3), so the statement comes as a bit of a surprise in Section 5. I think it would be worth bringing up this point earlier in the paper to motivate the work.
• In L 574, it is not entirely clear which bias the statement “It is not clear what causes the TKE bias in Fig. 13” refers to – which experiment? Which specific aspect of the bias? Or do they mean “the remaining bias”?
• L578: I thought the BL89 length scale was replaced by the adaptive Rodier et al. (2017) length scale (3.2.2)? Why is BL89 mentioned here? Also, the impact of changing the length scale isn’t discussed anywhere (and should be).
• L581 “We did not use local lengths so.” appears to be an incomplete sentence.

L626: „The model can accurately reproduce cloud fractions, cloud water content and turbulence according to LES conditional sampling diagnostics.“ I find this too strong a statement, after just pointing out in the previous section that there are still some rather large and unresolved errors in the TKE, for example, and little is shown on the improvements in water content (see above comment). A more appropriate statement might be that „the improved model more accurately reproduces cloud fractions etc. …. according to LES conditional sampling diagnostics.“  There is clear improvement, but still some error.

Technical corrections:

Abstract, first sentence: “… for the parameterization of the Atmospheric Boundary Layer (ABL)”, or leave out “the” and use plural (Atmospheric Boundary Layers).

Typo L25: “One” should be capitalized

L42: there’s a “;” that doesn’t belong here before Tan et al.

L47: mis-spelling of “entrainement”

L47: It may be advantageous to use the word “lateral” at least once at the beginning of this discussion of entrainment/detrainment to make clearer what type of entrainment is meant (given that top-entrainment for stratocumulus is another place where entrainment is uncertain).

L59: maybe better computationally “expensive” rather than “intensive”?

L67: “limited-area” sounds better than “area-limited”

L116: might be good to add here an approximate model layer thickness in the BL

Figure 1, and following figures: I find the y-axis (height) on the time-height cross section plots a bit confusing: Sometimes units of metres are used, other times kilometers. For the ARM Cu case, it appears the vertical axis refers to height above mean sea level (or a reference geoid), rather than to height above ground. I would suggest the authors choose consistent units, and show „height above ground“ on the y-axis.

Sec. 2.2.5 Liquid water content is referred to as r_c in this section, later on, it is referred to as q_l (caption Fig. 7). Please use a consistent naming convention for the variables.

L316: What is meant by a “nearing environment”?

Eqn. 16: The second term in the MAX function has a minus sign here, but doesn’t in Eqn 4. Is that a typo?

L320: The sentence “The wet part is further complicated to model.” sounds incomplete. Do the authors mean “The wet part is more complicated to model.”?

L332: You probably mean “upward part”, rather than “upper part”?

L514: The acronym MUSC isn’t introduced anywhere.

The section numbering in section 4 is confusing. Sec. 4.3 contains only a single sentence. Should 4.4 be a sub-section of 4.3?

Attached.