Thank you for your comments. Including more information about precipitation rates in our simulations will significantly improve the paper, and we appreciate you pressing us on this point. We think, for reasons outlined below, that using a 300 K SST in our TC world simulations is probably not cause for concern (in brief, because the threshold for TC genesis in these simulations probably differs from the threshold on Earth today), but we will run TC world simulations at higher SSTs if a re-examination of precipitation rates shows that they are low relative to precipitation rates in the simulations described by Lackmann and Yablonsky (2004). Detailed responses to all of your comments are below, with your comments included in italics for clarity.

The results of the current paper are somewhat surprising given that precipitation rate increases would surely be significant in the warmer model simulations. That 25 mm of rainfall corresponds to 2.5 hPa of hydrostatic pressure mass is inescapable, which must mean that the mass loss is entirely compensated by horizontal convergence in these simulations.  At rather coarse model grid spacing, geostrophic adjustment is expected to prevent complete compensation at such large scales relative to the Rossby radius. There is a feedback involving moisture convergence, but it must evidently not be very effective, given the results.  But this leads to a suggestion:

●    Other than Fig. 4c (which is surprising in itself), there are not any other plots of precipitation, which is perhaps the easiest variable for readers to use in evaluating realism. Can you please add or show additional plots showing precipitation distribution with temperature for non-extreme events?

Your point about additional plots of precipitation being useful for evaluating realism is a good one. We plan to add a figure showing snapshots of midlatitude precipitation fields for the simulations currently shown in Fig. 2 (all treatments of the mass sink in the alpha = 1 and alpha = 6 simulations). The precipitation fields for simulations at alpha = 1 should be useful for evaluating realism in a climate state comparable to modern Earth, and the precipitation fields for simulations at alpha = 6 should provide some information about how the intensity and spatial structure of precipitation change in much warmer climates.

●    More could be said about why extreme precipitation decreases in the warmest simulations. And on line 164, do you mean lower percentiles?

The decrease in "extreme" precipitation rates in Fig. 4c appears because 99th percentile precipitation rates are calculated for all times and not just times when rain occurs (a detail we will clarify in the revised paper) and are not sufficiently far into the tail of the precipitation distribution to show the consistent increases with warming expected of the most extreme precipitation rates. Instead, they behave more like the mean midlatitude rain intensity (i.e., the average midlatitude precipitation rate conditioned on the occurrence of rain), which also decreases in warm climates, likely due to decreases in the vigor of midlatitude eddies (see response to your next comment). The text on lines 163-164 is correct: at higher percentiles, extreme precipitation rates increase across the entire temperature range. For the sake of a streamlined presentation, we plan to replace Fig. 4c with a plot of 99.9th percentile precipitation rates, which increase with warming across the entire temperature range.

●    The cyclones in Fig. 2 with alpha=6 look rather anemic, and I wonder if there are other changes (e.g., in the static stability) that are inhibiting their development?  Could you please plot mid-latitude potential temperature profiles for the runs? What might be causing this?

Midlatitude eddy kinetic energy decreases in warm climates due to reductions in the meridional temperature gradient and increases in the midlatitude dry static stability, both of which reduce mean available potential energy (see O'Gorman and Schneider 2008a, DOI: 10.1175/2008JCLI2099.1). We will add a more thorough summary of the results from O'Gorman and Schneider 2008a, and in particular will note that reductions in midlatitude eddy kinetic energy can be linked to changes to the midlatitude temperature structure. Because O'Gorman and Schneider 2008a already provides a detailed discussion of changes to the midlatitude temperature structure in similar experiments with an almost-identical model, we don't think that adding a plot of midlatitude potential temperature profiles is necessary to adequately address this comment.

For the 10Xms simulations, 25 mm of rainfall is 25 hPa pressure-mass reduction, which surely must be a powerful deepening mechanism for precipitating cyclones.  Thus it is very surprising that this doesn’t increase EKE, but again, that depends on the model producing realistic rain rates in such cyclones, which are not fully shown. That the rain rate increases at a given temperature in the 10X simulation indicates increased convergence and moisture convergence, but this evidently doesn’t produce appreciable additional vortex stretching.

We agree that it's surprising that including a 10x mass sink has little effect on EKE (though, as Fig. 3c and Fig. 4 show, it does have an effect on other fields), and we expect that the additional plots of precipitation fields will help to convince readers that this is not due to unrealistically-low precipitation rates in midlatitude cyclones. It's possible that large-scale constraints on EKE (imposed, for example, by the mean available potential energy) limit the extent to which it can change in response to the inclusion of strong mass sinks, and we plan to add some text speculating that this may help explain why it remains similar even in 10xMS simulations.

My other concern is related to the “TC-world” simulations, and this is why I recommend borderline major revisions.  These experiments should be re-done, because those presented in section 5 were run with a constant SST of 300K (26.85C), which is marginal for TC genesis even in today’s climate (traditionally, 26.5C is used as a cut-off value, though other studies have debunked this).  The experiments are run in a  marginal TC environment, so it is no wonder it was difficult to get them to develop at these grid lengths.  Large regions of the tropical ocean feature SST above this value even in present-day conditions (30C or higher), and the results will be sensitive to this especially at coarse grid spacing.  I suggest setting the SST to 305 and 310K, and you could see very different results.  Perhaps also reduce the Earth radius further, if computational resources allow.

The threshold SST for TC genesis is likely climate-dependent, and the fact that it is around 26.5 C on Earth (where, as your comment mentions, SSTs above 30 C are fairly common) does not necessarily mean that it is also a cut-off value for TC genesis in our TC world simulations (where the SST is 26.85 C everywhere). Past work (e.g., Merlis et al 2016, DOI: 10.1002/2016GL067730) shows that TCs form readily in TC world simulations at SSTs as low as 280 K, far below the SST we use. Merlis et al 2016 do find that higher SSTs produce more intense TCs, so re-running TC world simulations at higher SST may be worthwhile if we find that rain rates in our TC world simulations are low compared to the simulations of LY04. However, we don't think that the proximity of the SST in our TC world simulations to the threshold for TC genesis on Earth is reason for concern by itself. We will add text to the paper to explain this with a citation to Merlis et al 2016.

Further reducing the planetary radius of TC world simulations does meaningfully increase computational cost (it requires a shorter time step, which increases the model run time by about a factor of 2 for each factor-of-2 reduction in radius), and quarter-Earth-radius simulations already require about a week of wall time each. It's moreover not clear that simulated TCs would fit on an eighth-Earth-radius planet, since TC size is already comparable to the planetary surface area in quarter-Earth-radius simulations (see Fig. 8c). We agree that higher-resolution simulations would be helpful for evaluating robustness, but we think that quarter-Earth-radius simulations are as far as we can push this model and already provide a resolution comparable to the simulations of LY04, as discussed on lines 406-407.

Related questions and suggestions:

●    Again, what are the rain rates in these TCs? It is essential to see these values to assess whether these experiments are useful in reconciliation of the results with prior studies. Please share plots of rainfall for the TCs.

Again, this is a good point. We will add snapshots of precipitation fields to Fig. 8 and double-check that the simulations produce rain rates comparable to LY04's simulations. (The current version of the manuscript provides a comparison in the text on lines 407-410, but we agree that we should include a figure that supports this comparison.)

●    Is the increase in hyperdiffusion by a factor of 5-10 really necessary, and if not, can it be relaxed? Could this not be removing meaningful gradients and reducing the effect in question?

We're not sure if the increase in hyperdiffusion is necessary, and will run additional quarter-Earth-radius simulation without the hyperdiffusion increase. Depending on the results, we'll either add text near line 398 noting that the hyperdiffusion increase is required to produce TCs, or report on whether mass sinks affect TC intensity in quarter-Earth-radius simulations without the hyperdiffusion increase.

The question of spatial scale remains. The authors do mention frontal scales (such as in the studies by Qui et al.), and the authors mention this around line 533.  But for scales much smaller than the Rossby radius, we expect divergent flow to more readily remove perturbations resulting from the mass sink.

This is a good point. We will add text to the final paragraph noting that divergent flow can more easily remove perturbations from mass sinks at scales much smaller than a deformation radius.

In discussing planetary atmospheres around lines 495-500 or 510-515, please consider adding studies of the Martian general circulation, which find that sublimation and deposition of CO2 play a substantial role (see, e.g., Chow et al. 2019, JGR Planets, and references therein).

Chow, K.C., Xiao, J., Chan, K.L. and Wong, C.F., 2019. Flow associated with the condensation and sublimation of polar ice caps on Mars. Journal of Geophysical Research: Planets, 124(6), pp.1570-1580.

Some other useful references that present equations for moist atmospheres are Ooyama 2001, and Bott 2008, and references therein. It would be good to add these to the list and discuss how such processes may indeed become important at very small, non-hydrostatic grid lengths.

Bott, A., 2008. Theoretical considerations on the mass and energy consistent treatment of precipitation in cloudy atmospheres. Atmospheric research, 89(3), pp.262-269.

Ooyama, K.V., 2001. A dynamic and thermodynamic foundation for modeling the moist atmosphere with parameterized microphysics. Journal of the atmospheric sciences, 58(15), pp.2073-2102.

We will cite and discuss Chow et al 2019, Bott 2008, and Ooyama 2001 in the revised paper.

A few other specific questions:

Figure 2. For the alpha=6 simulations, the cyclones don’t look well.  How does the layer-average static stability change relative to the alpha=1 simulations?  Is this affecting the results? (As mentioned above)

The midlatitude layer-average static stability increases with warming, as discussed in O'Gorman and Schneider 2008a (DOI: 10.1175/2008JCLI2099.1). This reduces the mean available potential energy, and likely contributes to a reduction in eddy kinetic energy. We will discuss the effects of warming on mean available potential energy and eddy kinetic energy in the revised paper (relying on results from O'Gorman and Schneider 2008a, as discussed earlier in this response), and mention that reductions in MAPE and EKE may be why midlatitude cyclones look less "healthy" in warm climates.

Figure 8.  It is quite surprising that the near-surface wind speeds don’t seem to increase with decreasing grid length.  Can you include the numerical value of maximum wind-speed value in each simulation in the figure, caption, or in the associated discussion?  Why do the wind speeds not increase with resolution, or is that just peculiar to this particular snapshot?

Yes, we can add the numerical value of the maximum wind speed to each panel in Figure 8. We will also examine whether maximum wind speeds increase with resolution, and add text discussing our findings.

Figure 9. Panel (f) is mislabeled.

We will fix the mislabeled panel in Fig. 9.

Appendix A3.  The need for a mass fixer is explained here, but it is not obvious (to me).  If P ~ E,  why is this necessary, exactly? Where the reinstated mass is added within the model domain, or is it a sort of global distribution?  Are you adding dry air mass in precisely the locations where the precipitation sink is removing it? 

The mass fixer is required to maintain a fixed amount of dry (non-water) mass in the atmosphere. The dry atmospheric mass should be constant because there are no sources or sinks of dry air in our simulation, but numerical errors in our implementation of sources and sinks of moist mass produce unphysical reductions in dry mass. (This is shown by Eq. A12.) The mass fixer compensates for those unphysical reductions in dry mass and prevents a gradual downward drift of total atmospheric mass. The mass fixer adjusts the total atmospheric mass by multiplying surface pressure by a single number in all columns. Because our model uses sigma coordinates, this changes the mass thickness of each level by an amount proportional to the level's mass thickness. The model then divides specific humidity in each grid cell by the same (single) value, which ensures that the total water mass in the atmosphere does not change and that the mass fixer only adjusts the atmospheric dry mass. This procedure does not add dry mass in precisely the places where it is lost during the calculation of sources and sinks of water mass. We think that adding dry mass in precisely the locations where it is lost is unnecessary because changes in dry mass (proportional to dqv^2) are locally small compared to changes in moist mass (proportional to dqv), and the mass fixer is used primarily to allow stable long-term simulations that correctly conserve total dry mass. We will add text during revisions to clarify (1) where dry mass is added back to the atmosphere, and (2) why we don't add dry mass in precisely the locations where it is lost due to numerical errors.

Thank you for your comments. They highlight an important shortcoming in the original manuscript; namely, a lack of discussion of our choice to use a pseudoadiabatic model and of the possible consequences of that decision. For the reasons described below, we do not agree that the use of a pseudoadiabatic model means that this study provides no new insight into atmospheric dynamics---we think that it does---but we take responsibility for not adequately justifying our choice of model. We plan to make several changes (outlined below) during revisions to address the specific concerns brought up in your comments (reproduced in italics), and think that the changes will significantly improve how we frame and discuss our results.

Although this study has been conducted properly, I don’t think that it provides new insight into atmospheric dynamics. This conclusion is based on the following statements:

●    The authors based their investigations on the claim that a mass sink exist within the interior of the atmosphere. They derived the according mass sink term in Appendix A1. This term is added to the mass continuity equation and it is obvious that they assume with this approach that water mass is annihilated. Although the pseudoadiabatic scheme of the model give rise to this assumption, it is not possible. Instead, water remains in liquid or frozen form in the atmosphere. The only mass sink appears at the surface where water leaves the atmosphere via precipitation.

As you say, our derivation of the mass sink term in Appendix A1 relies on the pseudoadiabatic assumption, and assumes that water mass disappears immediately upon condensation. Strictly speaking, this is not possible, but the same result can be obtained by deriving a mass sink term without assuming that water disappears immediately upon condensation, and then assuming that time scales for precipitation formation and fallout are infinitely fast. Because our study focuses primarily in the dynamics of midlatitude and tropical cyclones (large-scale weather systems that evolve over time scales of days), we think that treating precipitation formation and fallout as infinitely fast is a reasonable simplifying assumption, and not one that should preclude our study from providing new physical insight. We will include the new derivation described above in the appendix of the revised manuscript, and add text discussing why we think that the pseudoadiabatic assumption is reasonable given the focus of our study.

●    The sedimenting hydrometeors exert a drag force on the air that is identical to their weight which is known a condensate loading. Therefore, Eq. A2 cannot be true since liquid or frozen hydrometeors do not contribute to the density in this equation. That this matters has been shown, e.g. by Xu and Emanuel (1989).

Equation A2 is simply a restatement of hydrostatic balance, and makes no assumptions about whether condensed water species contribute to the density and hydrostatic pressure fields. In Equation A1, the equality \rho = (1 + r_v) \rho_d does explicitly neglect condensate loading. However, as described above, the same mass sink term can be derived by using a version of Equation A1 that includes condensate loading, and later assuming that time scales for precipitation formation and fallout are infinitely fast. More generally, the neglect of condensate loading in our study follows from our use of the pseudoadiabatic approximation, which we think is justifiable. The neglect of condensate loading is likely to affect some aspects of our simulations (for example, the tropics may be nearly neutral to pseudoadiabatic ascent, rather than to adiabatic ascent as found in Xu and Emanuel 1989), but we do not think that there is a clear reason to assume that the neglect of condensate loading has a first-order impact on this study's results. Still, the neglect of condensate loading is a caveat that we should have highlighted, and we will clearly acknowledge and discuss it in the revised paper.

●    The potential vorticity calculations in sections 4.1-4.3 are based on the incorrect mass sink term and are, therefore, obsolete. The mechanical and thermodynamical interaction of sedimenting hydrometeors with air leads to sources in the potential vorticity equation but I don’t believe that they have the simple form as described in section 4.1.

The potential vorticity calculations in Section 4 are consistent with the mass sink term as it appears in our model, and so are credible to the extent that the mass sink term in our model is credible. As discussed above, we think that the version of the mass sink term used in our model, while approximate, is justifiable. It is true that the potential vorticity source from mass sinks takes a different form when prognostic hydrometeor fallout is considered, in which case it involves the divergence of the hydrometeor sedimentation mass flux (see e.g. Equation 20a in Schubert et al 2001, DOI: 10.1175/1520-0469(2001)058%3C3148:PVIAMA%3E2.0.CO;2). However, there is a connection between this version of the PV source term and the version used in our paper: if precipitation formation and fallout are assumed to be infinitely fast, then the divergence of the hydrometeor sedimentation mass flux can be replaced by the condensation rate, which gives the PV source term that appears in Equation 6. We will add an appendix showing that the the assumption of infinitely fast precipitation formation and fallout connects the pseudoadiabatic PV equation to the version of the PV equation in Schubert et al 2001, and will reference that appendix at the start of section 4 to ensure readers are aware of how the two versions of the PV equation are connected as well as how they differ.

To tackle the issue of the mass sink impact one should use a model that includes a cloud microphysical scheme and prognostic equations for liquid and frozen water. The simplified GFDL model used by the authors is likely not appropriate for this purpose. Consequently, I do not recommend publishing this article.

To summarize our response: it is certainly possible that effects our model neglects (e.g., prognostic hydrometeor fallout and condensate loading) could alter our results. We will clearly acknowledge this in the revised paper, and we would welcome follow-up studies examining whether these effects enhance the dynamical impact of precipitation mass sinks. In our view, however, it's far from clear that a full cloud microphysics scheme and prognostic equations for condensed water species are essential for our study, which focuses on the existence or absence of the mass sink itself. They could be details that have little impact on our results, and it seems premature to conclude that a model that neglects them can provide no useful insight into the effects of precipitation mass sinks.