...see supplement

The role of radiation in the Northern Hemisphere troposphere-to-stratosphere transport

Summary

This paper presents a climatology of troposphere-to-stratosphere (TST) exchange in the Northern Hemisphere midlatitudes. They find TST from Lagrangian trajectories calculated from ERA5 that increase from 1 to 2 PVU. From these TST trajectories they investigate the physical processes involved in TST, with a particular focus on long-wave radiation. They show that most trajectories (84%) have a positive PV contribution from long-wave radiation and a small fraction have a very strong (>0.5 PVU, 8%) positive PV contribution from long-wave radiation. This contribution can range from a slow accumulation over a long time to more rapid PV production a the tops of sharp humidity or cloud gradients.

This is a well written paper on a interesting subject with novel methods and results, and a good fit for WCD. I would also like to credit the authors on the quality of their figures. They appear to have put a lot of thought into conveying a lot of information in a clear and concise way, and it has worked well.

I have added some minor/technical corrections below. I have also added some discussion that does not need extra work for publishing the paper, but would be good to include in the discussion and consider for future work. In particular, I think that the results are limited by only considering individual trajectories rather than using ensembles of trajectories to look at a full air mass undergoing TST and the total PV changes within it.

Corrections

Figure 6 caption mentions red circles, but the circles are black.

Line 493 – Delete “The way of”

The data availability citation for ERA5 says single-level data, but the DOI links to the pressure-level data. However, in the data and methods section you mention the model level data and diabatic tendencies which are available from the MARS archive, not CDS. Can you clarify what data is used to calculate the trajectories and PV tendencies. Is it model level or pressure level data or a mix of both?

Discussion

It is very noticeable from your case examples that the spikes in PVR_rad do not map well to the increases in PV (figures 1 and 5), so there must be some strong cancellations in the PV tendencies along individual trajectories. This has a large effect on what you might consider “radiatively dominated”. e.g. in figure 1 more that half of the PV increase from 1 to 2 PVU has happened before there is any significant PV tendency from radiation. I think this is probably because you only consider individual trajectories. I would suggest replacing the word “cases” with “trajectories”. e.g. in the abstract you state “Radiation is present in 84% of TST cases”. To me, “cases” implies individual events that would be composed of many trajectories. If you were to look at the total contribution of radiation to the complete air mass undergoing TST, you could get a different result for the percentage of events in which radiation has a positive contribution. (My guess is it would be higher). My thought would be that radiation is creating a local sharp PV tendency and turbulent mixing or diffusion is acting to spread in out. If you were to look at a full air mass you might see a strong radiative PV tendency in the middle, cancelled out by negative tendencies from turbulent mixing/diffusion, surrounded by positive tendencies from turbulent mixing/diffusion. Considering the full air mass, you would get a weaker but overall positive contribution from radiation and a roughly zero contribution from the mixing (although it would depend on how it is cut by the 2PVU limit).

“The distribution of radiative PV changes along trajectories is skewed because of the selection bias related to our choice of focusing on TST trajectories. If a similar analysis was performed along randomly selected trajectories in the upper troposphere, the radiative PV changes are expected to be symmetrically centred at zero.” – The statement about the increased skew due to selection bias makes sense. However, I would still expect upper-tropospheric PV tendencies from long-wave radiation to have some positive bias because you will more often be looking at cloud tops or the tops of humidity gradients.

Your case in figure 5a and 6g,h reminded me of the work of Rodwell et al (2013) (and following papers) on forecasts busts. The location matches up to the region where they highlighted that ascent in MCSs was too weak leading to an underamplified Rossby wave. The timeseries shows a sharp drop in PV which appears to coincide with the 12-hour assimilation cycle. This would be consistent with the assimilation acting to shift the tropopause higher if it was underestimated by the forecast and therefore lowering PV in that region

I was actually fairly surprised how infrequently the data assimilation is leading to spurious TST, from the spikes every 12 hours in figures 4b and 5a.