This paper explores the effects of SF6 loss on the mean age values and trends derived from SF6 model output and observations.  The conclusion, that SF6 loss can reverse the sign of the mean age trend over recent decades compared to mean age from a tracer without loss, is important in helping us understand the long running discrepancy between modeled and measured trends.  The explanation is actually incredibly straightforward, that the higher the concentration of a species in the atmosphere, the more absolute loss will occur and therefore the bias will grow over time.  It’s kind of hard to believe none of us studying this topic hadn’t thought of this possibility before, but there you go.  Nice work by the authors of this paper to recognize the importance of SF6 loss and quantify the effect. 

There are still outstanding issues, the trend in CO2 derived mean age is one of them and the SF6 lifetime derived in the EMAC model is another since it seems to be higher than other recent estimates.  Nevertheless, this paper is an important step forward in our understanding of mean ages derived from observations and how models can be used to help put them in context.  My main comments described below revolve around how best to use this information to help us make the observationally derived SF6 mean ages more accurate.  It doesn’t make sense to cast the measurements aside when it comes to mean age estimates so we need to be careful to frame the current understanding in a more inclusive way. 

I do recommend this paper be published with consideration of the comments below. 

Main comments

• It would be really nice to see a plot of latitude vs. altitude trend differences between REF(WS, SF6) and REF(NS, SF6).  I realize you’ve focused the trend analysis on the locations of the Engel et al. papers in the 4-6 year mean age range, and that’s important to see.  But at what mean age does the trend discrepancy emerge from the uncertainty to become significant?  How does this vary with location?  You hint at it in Figure 1 where the NS and WS profiles diverge but it would be good to see more.  I think this is important for the community to know what age ranges we can reliably use SF6 as a mean age tracer.  For instance, in my 2014 paper I showed mean age trends at four altitude ranges and in the bottom layer, where mean ages were 2-3 years, the SF6 measurement trends agreed within uncertainties with the modeled mean age trends.  Is this expected based on your work? 

• As a corollary to your findings, couldn’t a correction to SF6 mixing ratios be made in the calculation of mean age to account for loss? This would be similar to the way CO2 is adjusted for production from CH4 oxidation before calculating mean age.  This correction would have to vary with time and location, which you could derive from the differences between your REF(NS, SF6) run and the MIPAS data. 

Specific comments

Line 6:  I would add ‘in the NH extratropical middle stratosphere’ after ‘positive trend’.  Or something similar to give some reference to the Ploeger et al. (2015) and Stiller et al. (2017) hemispheric shift effect on the age trends that you mention in the introduction.  Also, in Ray et al. (2014) I showed that the balloon age trends are negative in the lower extratropical stratosphere.

Figure 1:  It seems like you could do a better job of averaging model data to match the March 2000 balloon profile location.  This balloon flight was clearly in the vortex as shown in Figs. 2 and 3 in Ray et al., 2017 with equivalent latitudes from 68-75N.  By averaging model locations at all longitudes and latitudes as low as 60N you are mixing locations in and out of the vortex.  It’s relevant to the lifetime estimates to see how the polar vortex model and measurement profiles compare since this is the region most impacted by the SF6 loss. 

Line 130:  ‘approximately’ is misspelled, change ‘turned out to be’ to ‘are’

Line 199:  ‘definition of Braesicke…’

Figure S4:  No y-axis labeling.

Figure S5:  In the caption you state ‘Values averaged over the region 30-50N and 2007-2010’.  I think the 30-50N part is not needed here since these are latitude vs altitude plots.

The paper presents a modelling study of the Age-Of-Air and its trends derived from the mixing ratios of SF6 in the atmosphere.

The authors performed multi-decade simulations of the stratospheric content of several SF6-type tracers to reveal the influence of the mesospheric sink of SF6 on the Age-Of-Air (AoA) derived form the SF6 observations.  It is shown that the mesospheric sink introduces a strong positive trend in the apparent SF6 AoA, that overrides other possible reasons for trends.

Unfortunately, the paper lacks several important details,  does not properly describe the current state-of-the-art, and has several methodological issues.  Therefore I request a major revision to address the points below.

General comments:

1. The introduction suggests that main objective of the study is to reveal the cause of discrepancy between modelled and observed AoA and their trends. If it is the case, it should be stated explicitly.  The lack of clearly formulated objectives and research questions makes it difficult to understand e.g. the model experiment design and justification for the choice of specific setups for the model experiments.

2. Further, the introduction states that "a comprehensive explanation for the trend differences between models and observations is still missing".  The effect of the SF6 destruction on the apparent AoA  has been already pointed by Waugh and Hall (2002, Sec 3.2) and addressed by Kouznetsov et al. (2020), who has simulated the effects of the mesospheric sink of SF6, and concluded that  "The apparent over-ageing introduced by the sink is large and variable in space and time. Moreover, the over-ageing due to the sink increases as the atmospheric burden of SF6 grows". (For more details, please refer to Sec.6.3 of the latter paper.) Therefore, it should be clearly stated what makes the existing explanations non-comprehensive.

3. The conclusions are formulated quite vaguely. It should be clearly stated what are the findings of the present paper, and how they  agree/disagree with earlier results, and what of the findings are new. It might make sense to have separate "Discussion/summary" (all references to earlier results etc.) and "Conclusions" (the concise statements that the authors are ready to defend).

4. Modelling studies of long-term evolution of SF6 distribution in the atmosphere have been reported by e.g. Reddmann et al (2001), Kovacs et al.  (2017), Kouznetsov et al. (2020).  The need for the present study and its similarities and differences from earlier ones should be clearly indicated.

Specific comments:

Sec 2.1: A brief characteristic of the model is missing., e.g. "online spectral chemistry-climate model with hybrid sigma-pressure vertical layers".

Sec 2.2: The description of the SF6 sub-model is very unclear. Probably, most of the reactive species from Table 1 were not implemented as actual tracers in the model. One has to indicate which species were taken as climatology, which were forced from other models, and which were actual tracers.  Was the submodel implemented as prescribed destruction rate as a function of altitude, latitude and season, or was it something more sophisticated?  The description should be sufficiently detailed to allow for an independent reproduction of the experiment with another model.

Contrary to stated in ll. 94-95, Fig.S1 does not show the relative importance of various reactions for SF6 destruction, but rather shows same reactions as in Table 1, but in a graphical form.

Sec 2.3. This is by far not the first simulations of this kind. What are the similarities and differences from the setups used in earlier modelling studies?

What is justification for specific model experiments, i.e. research questions to be addressed with each of the setups?

Sec 3.1: The section has one comparison against 3-year-mean MIPAS profile for a latitude belt of 30N-50N, and one in-situ profile. Those are nice for illustrations, but are insufficient to judge on the model performance in reproducing SF6 distribution in the stratosphere.

Fig.1a: It is not clear why this specific latitude belt, and these specific years were selected.  The MIPAS error bars show standard deviation of individual MIPAS profiles. How those are related to the uncertainty of the  average (of millions?) profiles that are shown? The MIPAS averaging kernel and spatio-temporal collocation notably affect the comparison (Kouznetsov et al. 2020).  This issue has to be at least discussed.

Fig 2a. Interestingly, Kouznetsov et al. (2020, Fig 5 there) shows very similar offset of the model profiles with respect to the in-situ one. I wonder if it is a coincidence, or an indication of a similar issue with both model setups.

Sec 3.2: The methodology for the lifetime estimate is quite unclear. Instead of explaining the method used, the authors put a reference to a 600-page report (Braesicke et al, 2019). The method, probably, refers to the equation in p. 1.20 of the report.  The equation assumes well-mixed atmosphere and implies that the destruction of SF6 is proportional to its burden, which is not the case: the destruction does not depend on the tropospheric content of SF6, but rather on its content in mesosphere.  In a situation when the change of SF6 burden is substantial at a time scale of ~10 years (AoA in the stratosphere) the difference leads to "surprising" results like reduction of SF6 lifetimes by 25% over 100 years.

Given the slow destruction of SF6 one could still define the lifetime in terms of well-mixed assumption, but that would require a long-term simulation without emissions, to let the mixing ratio relax to its equilibrium distribution and get to the exponential decline of the total burden.  Alternatively, as it was done by Kouznetsov et al. (2020), one could use a total burden that corresponds to the mixing ratio next to depletion layers.

Sec 3.4 -- 3.5: Same behaviour of trends in the apparent SF6 AoA has been pointed out by Waugh and Hall (2002), Waugh et al.(2003) and demonstrated  with extensive model simulations (Kouznetsov et al., 2020) for various latitudes and altitudes. Please specify what is a new finding here with respect to those studies.

l.329-365: The fact that SF6 destruction causes the positive trend in the apparent AoA follows from the simple fact that the SF6-AoA is proportional to a difference between stratospheric and tropospheric SF6 mixing ratios. Since the destruction is proportional to the SF6 mixing ratio, the difference increases together with the increase of the atmospheric SF6 burden. There is no need to involve any equations or advanced concepts (like Green functions etc.) to explain that.