This is an interesting and pragmatic approach to improving the parameterisation of dense water flows in NEMO. The evidence of a significant improvement is not definitive but there is sufficient to encourage further investigation. 

The premise of the paper is to use an analytical model of plume dynamics to derive a parameterisation. The assumptions underlying the analytical model are discussed but the suitability of this approach in a wider range of applications is not made clear. For example, at what model resolution do the assumptions breakdown even for the same IFR/FBC area?

The parameterisation adds momentum to the model and the authors should consider implications for energy balance diagnostics.

The implementation is unlikely to have impacted the computational performance of the model greatly but the authors should specify the cost of adding their parameterisation. The results sections appear to have been rushed and could be presented better to improve readability (see comments below). The code changes appear well-contained and using a recent enough version of NEMO that replicating this work in the latest releases should be straight-forward.

line 50 'only this' instead of 'this only'

line 74 'the conclusions of which' instead of 'and which conclusions'

line 83 Not sure turbulence is the best example here. Turbulence closures often include empirical relationships derived from observations and laboratory experiments and are not simply analytical studies.

line 114 'from which' instead of 'which yields'

Paragraph starting line 172. I find the discussion of u and v as alongslope and downslope velocities confusing in this section relating to the model grid. Why should the slope always be aligned such? Surely, any real world application could have the roles reversed or any orientation in between?

Section 3.1

 The sensitivity experiment is of insufficient length to evaluate the effect of the parameterisation in the target application of climate forecasting. Much longer tests in a global domain featuring all of the major overflows will be required before the universality of the approach and any impact can be assessed.

However, the tests do demonstrate some improvement and are a reasonable first step. The presentation needs to be improved however. Simple steps such as renaming DIFF as BD_DIFF (to make easier association with Beckman and Döscher) and ADV1 as CG_ADV1 (to make easier association with Campin and Goose) for example, would help. Figure 4 does not make for an easy comparison between the schemes and would benefit from addition panels showing DIFF-NOBBL, ADV1-NOBBL and ADV2-NOBBL.

Likewise, the comparisons of sections in figure 5 would benefit from a supplementary figure with columns NOBBL, DPLUME-FULL minus NOBBL, DIFF minus NOBBL, ADV1-NOBBL and ADV2 minus NOBBL

line 256 'a slight density increase' instead of 'a light density increase'

I see little value in appendix A as it stands. The only comparison is between NOBBL and BOM which achieves no new insights. It is already known that sigma coordinate models better preserve dense overflows (often too well). Unless equivalent comparisons with DPLUME-FULL are available, this appendix, and references to it, should be removed.

The Author are proposing a new parameterization of dense water plume in NEMO.  This parameterization is based on an additional momentum term which results in increasing bottom alongslope velocities. This approach as mentioned in the manuscript do not affect the tracer concentration but the effect on the energy budget of the model could be questioned.

While the introduction and implementation are clear and easy to follow, considering the results, the choice of showing only differences ( mainly Fig5) between then make difficult to really appreciate the results and the benefits of this new parameterization.   Although we understand that the objective is to increase the plume densities a reference simulation is missing. using BOM is a good idea but then it should be integrated in the results and not used as an appendix. Furthermore considering the comparison with BOM why is BOM  bathymetry represented with stair cases while it is supposed to be sigma model.

A thing which is not discussed and should be is the relevance of the parameterization in regards of the horizontal and vertical resolution of the model.