The manuscript describes the extraction of stratospheric gravity-wave drag from ERA5 reanalysis data. Both zonal-mean results are presented and also drag over selected mountainous regions. The drag is decomposed into contributions from vertical and horizontal flux convergence. The contribution of the latter is quantified for the first time from such data. Probability density functions of the drag components are shown, and it is demonstrated that low-frequency waves with large horizontal wavelengths contribute essentially to the horizontal flux convergence. A seasonal cycle is determined and interpreted in terms of wave-mean-flow interactions, as well as long-time variability. Linear trends are presented and regression to various climate modes, e.g. NAO. ENSO, and QBO.

Major comments:

This is a very solid analysis with various innovative aspects. Much can be learned from it, both about seasonality of gravity-wave drag and about relations to leading climate variability modes, so that it should eventually be published.

1. However, it also appears to me that the role of the mean flow in shaping the drag and its seasonal cycle could be worked out more nicely. This mostly concerns the seasonality in the southern hemisphere over the Andes, where the southernmost region is not exhibiting most intensive fluxes in winter, but in summer. If I understand them correctly, the authors argue that this is because the mean zonal wind in this region does not reverse its vertical derivative in the lower stratosphere. They claim that this is in line with linear wave dynamics. This might well be true, but I think it should be demonstrated. This could be done by single-column calculations of the drag given the diagnosed winds, with a simple wave emitted in the troposphere. Analytical arguments would be very welcome as well.
2. The authors decide for a rhomboidal instead of a triangular horizontal spectral filter in order to extract the gravity-wave signal from the horizontal wind. It is not quite clear to me whether this is the more appropriate approach. After all it should be the total wave number that decides, not the meridional wave number. Zonally symmetric gravity waves are not excluded by definition. Whatever, the authors state that the triangular filter gives the same results. It would be good to demonstrate this with a single figure.

Minor comments:

l. 70 and Fig. 3: Would it not be more consistent to also filter the vertical wind? True, linear theory tells us that to leading order all vertical wind is gravity waves, but there is large-scale balanced dynamics for vertical flow (omega equation yielding the balanced ageostrophic flow) so that some large-scale vertical wind cannot really be attributed to gravity waves.

l. 77: I understand that the factor n was necessary in Prochazkova et al (2023) where WRF data had been analyzed, i.e. from a non-hydrostatic model. However, here IFS data are used, i.e. from a hydrostatic model where by definition n = 1. This should not be presented as an assumption but rather as a consequence of the model formulation. Perhaps one could even set n = 1 directly?

l. 167 – 168: Not quite sure whether the weakness of the zonal-mean meridional circulation is a good argument why weaker meridional drag is significant. To leading order, the time-mean residual circulation is determined by the zonal drag!

l. 174: Even without oblique propagation, in classic simple single-column gravity-wave parameterizations horizontal fluxes and their convergence are allowed. They are not considered, for simplicity, but they are there. Oblique propagation will horizontally redistribute vertical and horizontal fluxes.

l. 179 – 180: A diurnal signal in the gravity-wave drag could also be a signature of coupling with solar tides.

Fig. 5: Please mention in the caption that the spectra are for 70hPa.

Fig. 7 and 8: Replace ‘winter’ by ‘peak season’?

l. 231: Should it not be shorter (instead of longer) orographic waves that propagate mostly in the vertical direction?