This paper explores the impact of global warming on the TTL cold trap, which is expected to weaken due to the projected acceleration of the BDC and deceleration of the Walker circulation (WC). These expected changes in the future climate were proposed by Held and Soden (2006) or Vecchi and Soden (2007) and have been critically discussed in recent high-level publications (Lee et al., 2022, Chung at al., 2019, or Heede et al., 2021).

My main criticism of the paper is that it does not consider the anticipated changes in the temperature structure of the TTL in the future, which according to current knowledge (Brewer, 1949; Randel and Park, 2019), exerts a primary influence on the entry values of stratospheric water vapor (SWV). As indicated in the aforementioned papers, a future climate, resembling "El-Nino", is expected with a weaker cold trap over the Maritime Continent and a stronger cold trap over the Eastern Pacific. This is mainly due to the significant changes in the relative position and strength of the cold traps, i.e., their temperature minima, during El Ninos compared to La Ninas or neutral configurations, as our current understanding of stratospheric moistening due to strong El Ninos suggests (Scaife et al., 2003; Randel et al., 2009; Konopka et al., 2016). While I agree that the strength of the horizontal and vertical winds may play a role as a second-order effect, discussing the second-order effect without considering the first-order effect seems inappropriate to me. I am curious why the authors did not use the meteorology provided by the ACCMIP. Using the trajectory-based method described in this paper for the 2000-09 meteorology and comparing it with the 2100-2109 data would be fantastic. First, separate the temperature effect, and then discuss the wind effect. Additionally, I recommend not including microphysics, which is only qualitatively discussed in this paper and, in my opinion, a third-order effect in the projected changes.

The authors' abstract claims to present a mechanism for a stratospheric water vapor feedback that can exist without changes to TTL temperatures. However, according to our current understanding, TTL temperatures are a result of wave forcing, which must change if an "El-Nino" like future is expected. Therefore, it is difficult to imagine changes in BDC and WC without changes in TTL temperatures. Overall, I find this study to be very idealized and recommend a "very major revision" or even withdrawal of the paper to explore the ideas discussed above. The only interesting statement in this paper is that "changes in the zonal wind are more important while changes to vertical winds have a negligible impact on water vapor concentrations." However, this statement in relation to vertical winds is even slightly different from the findings of Fueglistaler et al. (2014).

Lower stratospheric water vapor is expected to increase with global warming, leading to a positive feedback on surface temperature.  While much of this effect has been expected due to warming of the cold trap (the coldest temperatures in the TTL region, which set the amount of water vapor that can get through), the authors here explore the potential for changes in circulation to contribute to this positive feedback.  In particular, they show that even if temperature structure remains fixed, the expected slowing of the Walker Circulation (and hence zonal winds in the TTL region) will also lead to a moistening of the stratosphere.  This conclusion is obtained with some elegantly simple perturbation experiments, where zonal and vertical velocities obtained from ERA5 reanalysis are modified to capture the expected effects of greenhouse gas increase on the circulation.  The key is that reduced zonal winds associated with a slowing of the Walker Circulation reduce the fraction of particles that reach the cold trap, effectively warming the Lagrangian cold point.  

The authors also show that changes in circulation affect microphysical processes (sublimation and sedimentation), and ultimately moderate the potential drying of the air by the cold trap.  While acknowledging the limitations of their experimental set up to fully capture all effects, they convincingly show that the full response cannot be properly simulated without a careful treatment of the circulation (in particular, its 3D spatial structure and short term temporal variability).  This supports further investigation with more complex/high resolution models that capture both the dynamics and microphysics. 

I recommend publication of this manuscript pending consideration of the minor concerns and suggestions listed below.  I found the manuscript particularly well written and the figures well done, and commend the authors for the clarity of their exposition!  It was a pleasure to read.  

-Ed Gerber

Minor comments

(1) My only significant question about the manuscript concerns the divergent nature of the horizontal winds in the TTL.  I was initially concerned that the resulting flow field will be inconsistent, with divergence in the zonal flow not properly balanced by meridional or vertical flow.  Could this then lead particles to arbitrarily bunch up / disappear from regions with effective convergence/divergence?  If so, could this bias your results if the coldest region is associated with divergence/convergence?

I know that the authors are aware of this (lines 110-115), as they explain this inconsistency is a problem because it can lead to artificial dispersion, particularly in the case of stronger winds. If dispersion leads to under sampling of the temperature field, I assume it would have a net warming effect on the Lagrangian cold point, and thus if anything weaken the effects they see (where increased zonal wind cools the Lagrangian cold point).  

To be constructive, I would have appreciated more discussion on this point.  If previous studies have shown an inconsistency in the circulation to be minor when considering trajectories, or you have other reasons to believe that the effects are minor [say, that the distribution of particles stays uniform and my concerns above were unfounded], it would bolster confidence in the conclusions of the manuscript.

(2) As purely a comment (not criticism), I am perplexed by the relative insensitivity of the Lagrangian cold point to vertical velocity omega, as opposed zonal wind u.  Naively, I would have thought that omega would set the residence time in the TTL region, so that if you increase it, particles would have less time to sample the flow, and it would have the same impact as slowing the wind.  I suspect my intuition is wrong because I am assuming the flow to be zonally uniform.  The authors explain that variability in omega means that it barely changes in regions of maximum ascent/descent, even though the overall change is significant (20 or 50%).  If most of the ascent is localized, changing the overall velocity may have little impact.  (Or perhaps more subtly, its about the location of the ascent relative to the cold point.)  

To be constructive (and also stressing that these suggestions are minor: at the authors’ discretion), could one provide more insight by considering how the parameter changes effect the total lifetime of particles in the TTL and/or their expected transit length.  My naive expectation would have been that increasing/decreasing (in an absolute sense) omega would decrease/increase the time spent in the TTL, and so give particulars less/more time to roam and hence find the cold trap.  But perhaps the localized nature of upwelling means this insight is wrong?

In terms of the zonal wind, naively I would expect a larger value of u to increase the mean distance particles travel (and hence increasing the temperatures they cover).  Would this metric be easy to compute and confirm/dispell that intuition?  It could be the case that even though the lifetime does change with omega, this has minimal effect on the distance particles travel due to zonal asymmetries.

Finally, perhaps even if my intuition was correct that changes in omega do change the lifetime of particles which does effect how far they can travel (faster BDC = less time to travel = less distance covered), the spatial structure of the flow doesn’t really limit particles ability to see the cold trap in this case.  In this limit, the careful trajectory calculations here are then the simplest way to capture the effects of omega and u on the Lagrangian cold point.  

(3) There is little discussion of statistical significance in the body of the paper.  I appreciate that the authors have computed these calculations from three different winters, but these figures are in the supplement.  Could you include key results from these integrations to give a sense of the sampling uncertainty in the main text?  For instance, at line 145 the impact on the 50% change of wind scenario is shown to increase the cold point from 183.4 to 186.7.  Is this overall change identical in the other years — if so, you have high confidence.  Or you could give the range of changes to give a rough sense of the sampling.

Another option (which I think would also help sum up your results, independent of my concern about statistics) would be to include a summary figure that shows how the mean Lagrangian cold point varies with the wind perturbation.  Say, x axis = perturbation, -50% -20, 0 20 50, y axis Lagrangian cold point temperature.  Three lines could show how this varied across the three winters, with a different color to highlight what happened when the omega was changed or held fixed.  

Very minor suggestions by line number

Is WC a standard acronym for the Walker Circulation?  Currently living in a country where that is more uniformly associated with a vital (albeit unrelated) facility, I found myself chuckling at times.  The text might flow more easily with Walker Circulation spelled out.  (This said, I am quite used to BDC for a shortening of the Brewer Dobson Circulation. This perhaps bias in the literature I know.)

19-20. Consider shifting “in 1949” to the start of the sentence, as I felt it broke up the connection between the water vapor observations and Alan’s remarkable insight.

Figure 2. Consider adding headings above each column of panels with the model names, which would make the figure easier to follow quickly.

147 I think you don’t have to “note” this fact.  Rather, just state it, without ( ).  “The trajectories with …”.  In support of my comment (1) above, is it possible to quantify/show this dispersion effect?

Figure 3 (like all the others) is well made, but I think more clear captions would make it easier to parse.  

The top headings (+\omega, -U) could be put in words: “stronger BDC, decreased zonal wind” “strong BDC, increased zonal wind”

The inner captions could read a) \omega +20%, U -20%, b) \omega +20%, U +20\%, etc..

You could include information about the mean Lagrangian cold point with vertical bars that correspond with the colors of the different integrations.  (These values are referred to in the text.)

Finally, would be be reasonable to include the basic Classius-Clapeyron scaling on the WV changes in the other figures too?

158 As noted in comment (3) above, if this tiny warming significant at all?

Figure 6: I was initially confused how these curves appeared to be piecewise linear, at a temporal resolution longer than that of your model (1 hour).  The text did eventually explain that these discrete jumps are a result of the highly nonlinear nature of ice nucleation and sublimation (starting at line 236).   For my understanding, and perhaps for the readers, I assume this jumps would vary considerably for slightly different trajectories, so that if you were to do this analysis along the temperature structure of each individual trajectory, and then average, you would likely get a much smoother field.  I trust this would be very nonlinear: do you expect that if you took into account variability, would the changes be amplified (more dehydration in the runs with weaker U) or weakened?

260 There’s a stray space: ‘) .’

Table 2 Could provide another opportunity to assess the statistical certainty, by contrasting results based on the other years?  (Or do I misunderstand that you completed the full analysis on the other years.  I do appreciate the immense effort it takes to do this analysis (hundreds of thousands of trajectories per experiment).  

Alternatively, could you assess statistics from 2008 alone by sub-sampling trajectories and/or bootstrapping?  I want to emphasize that my concerns about the statistics are minor, in the sense that I feel this paper is more about identifying mechanisms and processes than trying to quantify trends vs. noise, where such quantification is essential.  I do think the paper would be stronger, however, with more attention to uncertainties associated with sampling.