This paper performs hindcast CAM simulations for data taken during the EURECA field campaign to attempt to improve the representation of the boundary layer in trade-wind regimes by testing experimental versions of the CLUBB parameterization that include upgradient fluxes for momentum.  Overall, I found this to be an interesting paper that addresses an issue that typically receives little attention.  Additionally, I found this manuscript to be exceptionally well written, which always makes the job of a reviewer much easier and should not go unnoticed.  I do feel this is worthy of publication, after the authors consider some of the following points which I feel would improve the presentation of results and perhaps increase the impact. 

• Unlike the Larson et al. (2019) paper, the authors here find that the boundary layer thermodynamic structure is improved with their experimental configurations. I feel it is important to show what impact this has on the simulation of clouds.  If the impact is negligible then that could be displayed in a figure or two and would be a worthy reference point for other studies.  If the change is more significant then perhaps this should be highlighted a bit more (but I understand this was never the intended focus of the paper) with a discussion of potential implications for climate length simulations and other modeling centers wishing to improve momentum transport.
• I am not a fan of the naming convention for the simulations (x001,x101, etc,). Why not refer to “x001” simply as “CNTL” and devise similarly brief but easily distinguishable names for the other sets of simulations?
• The authors present results for control CAM, CAM with prognostic momentum fluxes, and CAM with prognostic momentum fluxes but revised length scale definition. I feel it is also very important to present results from CAM with only the revised length scale modification (essentially the Guo et al. 2021 configuration).  Having this data point is essential to help the reader tease out the relative contribution of improvement stemming from this modification alone.
• I am confused why the authors chose to perform tuning experiments for the configurations that include the revised length scale definition. While I understand it is important to exploit model sensitivity to tunable parameters, it is odd that they chose to do it for this configuration but not experiment x101 (which also introduces new tunable parameters) and this makes the article feel a bit disjointed.  In addition, presenting results for all four x20 simulations is a bit redundant and cumbersome.  If the authors feel it pertinent to include these tuning simulations perhaps they could just show the average result (and summarize the sensitivity/spread in a short section with a figure or two)?  In this instance I feel the authors would also need to add a strong justification why they felt it necessary to do a tuning suite for this particular configuration while neglecting to do so for the others (or perhaps add a tuning suite for the other configurations).   Overall, though (unless justification is provided) the main purpose of this paper didn’t seem to be about finding optimal tuning parameters nor exploiting CLUBB’s sensitivity to them.  Therefore, I’m inclined to suggest the authors just present results from one realization of x201 (with whatever the default parameters are from Guo et al. 2021) and briefly summarize that the sensitivity to tunable parameters was explored and highlight the most important points of that analysis.

By conducting a series of CAM hindcast simulations during the EUREC4A/ATOMIC field campaign in the Tropical Atlantic in early 2020, this study points out that the prognostic treatment of momentum flux improves the simulation of tropical trade-wind to a greater extent, compared with the default CAM-CLUBB. This study improves the CLUBB scheme by further implementing a generalized calculation of the turbulent length scale, which reduces model bias and RMSE relative to observation. Recently, more and more studies have noticed that momentum fluxes can be upgradient in shallow convection regimes, and it is challenging to parameterize them. In my opinion, this study is interesting. This paper also gives us some insights on how to further improve the prognostic treatment of momentum flux and simulation of trade-wind regimes in global models.

In addition, this paper is well-written. I therefore recommend the publication of this paper in GMD after some revisions.

1 I noticed that Guo et al. (2021) applied additional damping of non-cloudy layers stable stratification to scaler flux and also w’3 in the stable layers, because of gravity-wave dispersion under cloudy conditions (please refer to eq23 and 24 in Guo et al. 2021). This might be of some help in the simulation of shallow convections and stratocumulus. The case of momentum flux may be similar to the scaler flux, should an additional clear sky damping be considered in Eq 6 as well? And, will it further improve the results?

2 C6 is used in scale flux and momentum flux, which are very important for the return-to-isotropy terms. But, I also noted that C6=C6b= 1 in Guo et al. (2021). So, the Taus could have full control over the return-to-isotropy in the momentum flux equation. This study uses the default CAM setting of C6=4 (and I assume C6b=6), which would involve the skewness function in the calculation of the pressure term (Larson 2022), which overlaps with the role of the Tau scheme. It creates some difficulties in understanding the performance and improvement of parameterization. I suggest that the authors set C6=C6b=1 for the experiments to simplify the problem.

3 The difference between the X101 and X204 horizontal momentum fluxes budgets is remarkable. The budgets of horizontal momentum fluxes in X204 qualitatively resemble those of BOMEX and RICO in LES (e.g. Figure 7 and 9 in Larson et al., 2019), with the buoyant term predominantly balancing the turbulence production term. I'm curious what causes this big change, is it due to the tuning listed in Table 1 or the model structure changes? This study does give us some explanations, but it would be better to add some more discussions and show more turbulent profiles that help the reader to better understand the improvement, such as w² , u² ,  v² and scaler fluxes.  u² ,  v² would also benefit directly from the prognostic treatment of momentum flux.

4 Also, how were the parameters in x201-x204 determined (Table 1)?

5 The names of the experiments in the main text make it a bit difficult for me to read, could you please give them more visual names?

6 Figure 8, legend explanation?

7 If possible, it would be better to add some LES results to the plots, like in Figures 1, 5, and 12?

8 Line 410, page 18, “between 200 and 2 km”, do you mean 200 meters?

The study seeks to assess improvements to the prediction of tropical shallow cumulus regimes by modifying CLUBB to allow for counter-gradient momentum fluxes. Overall, it is important work and I found the paper well written and  straightforward to follow. I have a few comments that require revisions before the article should be published, but I consider them minor. I have attached specific comments in a PDF.