This paper presents a straightforward analysis of tropical stratospheric upwelling derived from MLS water vapor (H2O) observations of the ‘tape recorder’, including analysis of variability and comparisons to reanalysis and another tracer derived dataset. The upwelling derived from MLS H2O was introduced many years ago but has received relatively little attention, and the study here can be useful to update results and serve as a reference for critically evaluating the derived upwelling. Regression analyses of the time series show important variability linked to the QBO throughout the lower stratosphere and ENSO close to the tropopause, which are well-known behaviors from other data sets. Anti-correlations are observed between upwelling and ozone in the tropical lower stratosphere, which is expected behavior but good to confirm with these data. Comparisons of the MLS H2O-derived upwelling to the other data sets show reasonable agreements along with some differences, and given the limitations of each of the data sets (including low vertical resolution and limited effective time sampling for MLS H2O) it is unclear which are the better estimates. Overall, the paper is a useful contribution that might help spur further uses of these data, but a somewhat more critical evaluation of the new data would be helpful for potential users. I recommend a minor revision, including addressing the following comments.

Specific comments

I suggest using ‘Tropical stratospheric upwelling…’ in the title.

It would be good to know more about the details and uncertainties of the MLS H2O upwelling calculations. The authors choose to use correlations between vertical levels ~ 4.5 km apart, so that the results represent mean upwelling over these broad layers. This detail is not well emphasized, and vertical profile results are shown with a 1-km grid (e.g. Fig. 1). What do the upwelling calculations show if narrower vertical layer differences are used (even between adjacent levels)?  How sharply peaked are the lag correlations in time, and is there any corresponding information on uncertainties in derived upwelling? The 6-month window calculations look reasonable in the reanalysis upwelling comparisons, and why aren’t these used throughout the paper?

I think including the SWOOSH results prior to 2004 is interesting, but the results look problematic to me given the data gaps and much noisier character of the time series seen in Fig. 2. Figure 9 highlights poor coherence between adjacent pressure levels and poor agreement with reanalyses for the early period. What are the causes of the data gaps? Why is additional time smoothing (3rd order polynomial fits) needed for these data? A ‘spike’ near 2001 in the derived upwelling is discussed but this looks more like noise in the calculations to me. In my opinion the authors should be more critical of these issues in the results derived from SWOOSH.

Figure 1 left axis should be Pressure, not Altitude (also Figs. 7 and 10). Do the error bars in Fig. 1 represent the standard deviation of the calculated means, or the standard deviation of the monthly time series?

The regression results regarding QBO and ENSO impacts on upwelling are very similar to previous results of Abalos et al 2015 (doi:10.1002/2015JD023182)

I’m curious about the detailed results in Fig. 4 – how can regression results with R2 less than 0.1 be statistically significant? How are the degrees of freedom evaluated for these low-frequency variations?

I like the comparisons of upwelling and ozone time series in Fig. 5, but shouldn’t you use the ozone level closer to the midpoint of the upwelling calculation for the most meaningful comparison? 

The residual vertical velocity from reanalyses is given in terms of pressure vertical velocity – does this include the eddy heat flux term, or is it just the zonal mean pressure velocity? (there are small differences in the deep tropics, but this should be clarified). For more direct comparisons to the H2O results, I recommend calculating and plotting the reanalyses upwelling in terms of mm/s.  

The ANCISTRUS upwelling results (Figs. 10-11) seem overly smooth in terms of vertical structure and time evolution. While these estimates are derived from numerous chemical tracers, there are very small vertical gradients or seasonal variations in most of the tracers in the tropical lower stratosphere, and I frankly wonder about the information content in this region. Why is there no seasonal cycle in upwelling in the 100-46 hPa ANCISTRUS results in Fig. 11c? Also, the 6-month window MLS H2O results in Fig. 11 don’t seem to match the corresponding results in Fig. 8 in terms of seasonal cycle variations. As with the other parts of the paper, I think it would be useful to be somewhat more critical of the ANCISTRUS comparison results.

This paper investigates tropical upwelling velocities in the ascending branch of the Brewer-Dobson circulation in the lower stratosphere. A new estimate of effective upwelling velocities is calculated from stratospheric water vapour satellite observations (MLS, SWOOSH), based on the ascending tape recorder signal and lag times between different levels. This estimate is further compared to residual circulation velocities from reanalyses (ERA5, MERRA2, JRA3Q) and observationally-based effective upwelling calculated alternatively with the ANCISTRUS model. It is found that the water vapour-based upwelling velocities are largely consistent with current knowledge on the stratospheric circulation, showing mean upwelling of reasonable magnitude, a seasonal cycle with maximum in boreal winter, and strongest inter-annual variability related to the QBO. An interesting result is the reduced variability in tropical upwelling after about 2015. A relation to the QBO-disruptions in winters 2015/16 and 2019/20 is discussed.

I think that this paper presents valuable results and addresses a subject which is relevant to a large readership of ACP and falls well into the scope of the journal. I have one major more general and a few specific comments below, which hopefully help to further improve the paper and which I'd ask the authors to consider before I'd recommend publication.

Major comment:

As explained in the paper, the upwelling velocity estimate from the water vapour tape recorder presented here is an effective transport velocity which includes both advection by the residual mean mass circulation and mixing. I don't question that this is a valuable quantity to calculate, but I'm wondering about its comparability to the residual upwelling velocity w*. What is urgently needed for model and reanalysis validation is an observational constraint for w*. Whether the effective velocity presented here serves as a meaningful constraint is not clear to me and should be worked out in more detail.

Qualitatively, the effective upwelling appears to be very consistent with residual circulation upwelling. However, there are quantitative differences, e.g. regarding the mean profiles (Fig. 7) or the seasonal cycle (Figs. 8, 9). Are these quantitative differences significant given the differences in the methods (effective upwelling vs. residual circulation), methodological uncertainties (e.g. measurement errors in H2O, lag time calculation over broad layers), and the variability around the means (e.g. there is large variability around mean profiles in Fig. 7, also see specific comment below)? In other words, can a quantitative recommendation be given on the reanalyses considered here (e.g. MERRA2 upwelling being too slow)? Or are all uncertainties together so large, that only qualitative consistency remains as a result?

In my opinion, the "golden way" would be to carry out a "proof-of-concept" study within a controllable model environment first, where all quantities are well-known. This would imply calculating effective velocities from model water vapour tape recorder and comparing with the model w* velocity. If such an analysis is beyond the scope of this paper, at least a thorough discussion should be included, as to what degree the effective upwelling can be considered a quantitative constraint for the residual circulation, and hence how to interpret the differences to reanalysis w* presented here.

Minor comments:

1. I was wondering how much the assumption of a fixed lag in the ENSO regression would influence the results (e.g. L195ff). There are other studies, as e.g. cited the Diallo et al. (2018, 2022), which use variable lag time for the MEI index. I'd find it helpful if potential influences of the fixed lag in the regression were discussed. A full MLR with lag optimization is not needed here in my opinion, but perhaps some sensitivity test of the shown results regarding different lag times.

2. For relating upwelling and water vapour variability it would be particularly useful to consider upwelling at a level lower down in the tropical stratosphere, say 100-80hPa, as this should be most closely linked to cold point temperature variability. The Randel et al. (2006) study shows a persistent strengthened upwelling at 100hPa during the years after 2000 (e.g. their Fig. 9) correlated with anomalously low cold point temperatures. Perhaps, the fact that this is not seen here is just related to the broad vertical region over which the upwelling estimate is averaged, and because of that a missing link to tropopause temperatures? It would be good to discuss these points in more depths here, and perhaps even try the upwelling velocity calculation over a narrower vertical layer 100-80hPa.  

Specific comments:

L20: Anti-correlation holds only for long-lived tracers with stratospheric sources. For long-lived tropospheric tracers, e.g. N2O, correlation holds. Please clarify. 

L94: As far as I know, SWOOSH v02.7 in addition includes ACE-FTS and SAGE III. Please clarify (also at later places in the paper).

L127: According to VonClarmann et al. (2016, 10.5194/acp-16-14563-2016), the ANCISTRUS model is based on an effective transport formulation (not a TEM residual circulation framework) and therefore provides effective transport velocities also if mixing is explicitly taken into account. If so, this should be clearly explained here (and text later needs to be adjusted accordingly).

L140: Please give an approximate level here, e.g. "up to about 10hPa".

L142: "water vapour altitude" --> "pressure altitude"

L154: When first reading I got confused here as to whether 12 or 6-month intervally are used in the study (became clear later that both are used). Perhaps make this directly clear by writing something like: "In addition to the calculation based on 12 months intervalls, we also ..."

L173: I think MLS is not the only dataset contributing to SWOOSH v02.7 (see my comment above).

Fig. 3 and L203ff: Why are different instruments shown for the different layers? If there is some reason, please explain. If not, I'd suggest to either show SWOOSH or MLS for both layers.

L212: I was wondering why the decline in regression performance was not discussed here at all, unless I saw that it is then in detail discussed later. Perhaps mention that already briefly here and refer to the later sections.

L256: "... such as found for ..:"

L263ff (throughout Sect. 3.2): I miss some discussion of the effects of mixing in ozone-rich extratropical air into the tropical lower stratosphere. This process hace been shown to substantially influence the tropical lower stratospheric ozone budget just above the tropopause (e.g. Abalos et al., 2013, 10.5194/acp-13-10787-2013), strongest in the NH tropics (Stolarski et al., 2014, 10.1002/2013JD021294).

L295: "... reduced variability after 2015 ..."

Fig. 7, L327: There is remarkably larger variability in the reanalysis profiles compared to the observational estimate in Fig. 7. Has the seasonal cycle variability filtered in the reanalysis data before? If the variability in reanalysis is about 5-times as large as in the observational estimate, this needs to be discussed.

Fig. 8: It's hard to see the lines behind the symbols in the seasonal cycle plots and therefore almost not possible to distinguish between ERA5 and MERRA2. It would be helpful to include a legend for the symbols.

L347: Is the agreement in seasonal cycles really "good"? To me it seems that the seasonal cycle in reanalyses in Figs. 8, 9 is substatially larger than from observations. Perhaps a comparison figure or table with seasonal cycle amplitudes could make this point clearer.

L348: I find the relation to the shallow branch here somewhat confusing. As far as I know the deep branch of the stratospheric circulation shows a clear seasonal cycle with maximum upwelling during boreal winter, while the seasonality is not so clear for the shallow branch (c.f. Lin and Fu, 2013, 10.1029/2012JD018813; Baikhadzhaev et al., 2025, 10.5194/acp-25-12753-2025). In this context, it is important to note that the upwelling across a given level in the lower stratosphere, as is presented and discussed here, includes the mass flux related to both the shallow and deep branch. Please clarify.

L377: As said before, isn't this just a cause of the ANCISTRUS formulation and not only due to the neglection of mixing.

Fig. 10: Why are there only 2 error bars vor ANCISTRUS?

L439: "... with long-lived stratospheric tracers ..." (see comment above).

L458: "... about six measuerement days ..."?