Review of 'Influence of radiosonde observations on the sharpness and altitude of the midlatitude tropopause in the ECMWF IFS' by Konstantin Krüger et al.

MS No.: egusphere-2023-2094

General comments

================

I found the manuscript to be useful and interesting. 

It should be published after revision. 

It is generally well written, but a little long in places. 

The description of the biases round line 475 needs improvement, see 

detailed comments. 

In the data assimilation literature 'representation error' 

(or representativeness errors) is a useful concept relevant to this 

manuscript (see eg Hodyss and Nichols, 2015; Janjic et al, 2018). 

For wind I find it useful to think of the spectra of rotational and divergent 

wind (eg Boer, 1983). Any finite numerical model truncates the spectra and 

hence reduces the wind speed (only slightly in general, but more where there 

is a lot of variability on small scales, eg near a jet stream). 

On a related note it would be interesting to define a set of 300 m vertical layers (similar resolution to the IFS near the tropopause) average high resolution observed profiles over these layes and see how the averaged profiles compare with the original ones. 

Some properties such as 'Profiles with sharper tropopause exhibit stronger background biases' are more-or-less inevitable when averaging onto a coarser grid. 

Another possible (hopefully minor) addition: is there a relation between 

the sharpness of the temperature inversion and the maximum wind speed? 

(Perhaps one should look at this separately for mid-latitudes and sub-tropics?) 

Boer, G. J. 1983 Homogeneous and isotropic turbulence on the sphere. J. Atmos. Sci., 40, 154-163

Daniel Hodyss, Nancy Nichols, The error of representation: basic understanding, Tellus A, 2015, 67, 0

Janjic, T., N. Bohrmann, M. Bocquet, J.A. Carton, S.E. Cohn, S.L. Dance, 

S.N. Losa, N.K. Nichols, R. Potthast, J.A. Waller, and P. Weston, 2018: 

On the representation error in data assimilation.  

Q.J.R. Meteorol. Soc. 144, 1257-1278, https://doi.org/10.1002/qj.3130.

Spefific/technical comments

===========================

Abstract - a bit long

lines 35-36 'Above the tropopause ...' make it a bit clearer that this is 

a description of average or typical conditions. 

39 'sharp distributions' - 'sharp gradients'

48 'an accurate representation of ... sharp gradients is of high importance 

for NWP'  - a bit too strong 'may be of high importance' 

(other models with coarser vertical resolution still perform well)

61 'underestimated UT wind maxima ... in the ERA-15 reanalysis' 

ERA-15 is quite old now, and coarse resolution by current standards, 

I think this should be mentioned. I think that the discrepancy has 

reduced (but not disappeared) in more recent versions. 

64 'satellite observations ... that DA smears out' 

- 'that satellite DA smears out' 

It is well known that satellite soundings have limited vertical resolution 

(broad weigthting functions, especially broad for microwave) and they are 

very numerous so this is not a surprise. 

75-76 'Hence, no definitive conclusion can be drawn as to whether DA 

sharpens or smooths the tropopause.' 

My guess would be that it smooths slightly overall because the numbers of 

satellite soundings are so large. Also, if DA of RO and radiosonde data 

does sharpen the tropopause it raises the question of whether the DA has 

added detail that is inconsistent with the model dynamics at the current resolution. I suggest that the sentence be rewritten. 

97 '[extra radiosondes] ... launched and applied in an OSE' 

replace 'applied', perhaps with 'used in an OSE' but 'their impact was 

studied in an OSE' would be more precise. 

110-113 'about 9200 radiosonde profiles' 

Are these all high-resolution reports? A small proportion of reports in this 

area are only available at lower resolution (as for alphanumeric reports). 

The ship BUFR reports are at lower resolution than most of the land stations, 

but after the ECMWF vertical thinning there probably isn't much difference. 

115 'on demand' to 'on-demand'

119 'aircrafts' to 'aircraft' (yes, 'aircraft' is its own plural!)

130 'With the aim' delete? Start with 'To investigate ..'

137 'atmospherics state' - 'atmospheric state'

147 'using pressure used' - delete 'used'

216 'provides a high data coverage' - 'has a good data coverage'

246 'temperature decreases' - should be 'increases'?

247 'above tropopause' - 'above the tropopause'

260 'a wind speed increase in the analysis' - this is what I would expect, 

see general comments.

Figure 6. The legend, especially the subscripts, is too small to read. 

308 'an uni-modal' - 'a uni-modal' (sounds right to me)

Figure 8. I am not sure that this figure adds much. 

To me a more interesting question (general comments) is the link, if any, 

between the tropopause sharpness and the maximum wind. 

370 'changed the interval' delete 'the'

370 'exhibits' - 'exhibit'

383-384 'For the interval of smallest innovations, ... deteriorated tropopause 

altitude.' This is just a sampling effect. If you have an innovation of ~0, 

then the only way it can change in the analysis is to get larger in magnitude. 

416 'sharpen' - 'sharpens'

438 'other observations to influence' - 'other observations also influence'

440 'e.g., GPS radio occultation or dropsonde observations' I suggest 

'aircraft or GPS radio occultation observations'. Dropsondes are too sparse 

(and sometimes dropped too low) to have much effect. 

474 'The remaining LS cold bias in the analysis (0.2 K) corresponds to previous assessments ...' 

Add that the main cause appears to be excessive humidity in the analyses at those levels - giving radiative cooling. This is mentioned in two of the references (Shepherd et al and Bland et al). 

Perhaps mention recent changes at ECMWF that have reduced, but not eliminated the cold bias:  https://www.ecmwf.int/sites/default/files/elibrary/2021/19875-stratospheric-modelling-and-assimilation.pdf

475-476 'The warm bias at the tropopause (1.2 K) is in line with Ingleby et al (2016). ... compared to Bland et al (2021).' - Needs rewriting. 

I'm not sure where the 1.2 K comes from. 

Ingleby et al (2016) has a statement "direct use of the tropopause significant

level may result in a local bias (observation cooler than background)" but 

doesn't give a value for the bias. 

I am also confused by the comparison to Bland et al which seems to say that they found a significantly larger bias. 

One factor is aircraft temperature bias and the many aircraft reports at 

200 and 250 hPa. The following is from p 10 of Ingleby (2017): 

"At 200 hPa the O-B difference is more negative than at adjacent levels - this is due largely to a warm bias in flight level aircraft temperatures feeding through to the background fields.  Figure 3.2 shows that at 200 hPa the background values at radiosonde locations are about 0.2° higher without aircraft assimilation" 

Ingleby B. 2017: An assessment of different radiosonde types 2015/2016. 

ECMWF Tech. Memo. 807 (ECMWF website)

484 'positive shear in the up to' - 'positive shear below'

502-504 'In case the LRT latitude of background and observation is comparable 

... resolution of the IFS.' 

As mentioned above (see 383-384) this is a sampling issue and I recommend that 

it is deleted from this section. 

521 'routinely radiosondes' - 'routine radiosonde or aircaft data' 

521 'at a close-by location' - 'nearby' (one word will do)

524-525 'the B-matrix ... spreads information ... horizontally and vertically' 

This is true, but the vertical spreading is less important when assimilating 

a high-resolution profile from a radiosonde. 

530 'strongly rely' delete 'strongly' (similar to comment on line 48). 

542 'These increments are corresponding to' - 'These increments correspond to'

548-549 'sharper ... tropopauses, that are typically associated with ridge 

situations (high tropopause)' Any background/evidence for this (I don't 

think it was mentioned earlier in the text). 

580-581 'The feedback files analysed ...' 

These are not stored in MARS, I think someone (Gabor?) must have supplied them directly. 

The raw BUFR data (without feedback) are available from 

https://www.ncei.noaa.gov/data/ecmwf-global-upper-air-bufr/archive/

670-671 'Lavers ... Accepted' - now published online

Review of 'Influence of radiosonde observations on the sharpness and altitude of the midlatitude tropopause in the ECMWF IFS' by Konstantin Krüger et al.

The paper addresses the representation of tropopause sharpness in forecast data and studies the effect of radio sonde observations on the assimilated tropopause structure.

For this purpose the authors use more than 9700 radiosonde profiles in autumn 2016. Out of these 500 sondes were released as additional soundings in the frame of the NAWDEX experiment. These are used for an IFS observing system experiment with and without these additional soundings. For the full data set the authors analyze the emerging increments, innovations and residuals of temperature, wind (as well as shear) and static stability. Importantly they do this in tropopause relative coordinates to extract the effect of the assimilation of additional soundings on the tropopause thermal structure and winds.

In general their analysis clearly shows that the sondes lead to a sharpening of the tropopause in the assimilation. They further split the data according to Brunt Väisälä frequency in sharp, smooth and medium gradient tropopause cases and show the that the sharpest tropopauses require the strongest increments, similar for the winds. 

Overall they found a sharpening of the tropopause with increased N²_max and increased shear values from positive at the wind maximum to negative above the tropopause. In particular they infer from a comparison with and without the additional sondes a substantial contribution of the additional sondes to the assimilation. They also show that the analysis tropopause altitude is shifted towards the sounding observations. The comparison of the OSE runs highlights that the main contribution to the tropopause sharpening can be attributed to the radiosondes.

The only point which could be discussed by the authors is the role of humidity as possible reason for the temperature deviations at the tropopause (see below) , though the humidity is not assimilated, it might explain at least partly the discrepancies of tropopause sharpness compared to the observations.

Overall the paper is very clear, well structured and each analysis step is clearly motivated. The methods are are well documented and appropriate, the emerging conclusions are are scientifically sound - it was a pleasure to read.

The paper clearly merits publication in WCD and I see only minor points.

l.100: Although moisture is not assimilated the incorrect representation in the IFS, it may lead to larger temperature differences above e.g. cirrus clouds compared to clear sky observations. Cirrus occurrence in observational data might be misrepresented or missing in the IFS data, particularly for the N²_max cases (i.e. ridge regions). Humidity is not assimilated and therefore not analyzed by the authors. Nonetheless it could be discussed (maybe in the final discussion) as possible cause for the misrepresentation of temperature at the tropopause. Would it be possible to relate the temperature increment at the sounding location to the observed humidity compared to the background humidity? A larger increment for different saturation conditions for IFS versus sounding would provide a potential explanation of temperature increments at higher tropopauses. 

Fig.2 and related discussion: Fig.2 How does the altitude distribution of the 500 additional sondes compare to the rest? Could you add the PDF for those additional 500 soundings as separate contour?

Since the subtropical tropopause and the extratropical tropopause have partly different drivers, how do the results change when only considering extratropical tropopauses with altitudes less than 14000m? Wouldn't one expect different effects of the assimilation of the mainly extratropical 500 soundings for the extratropical tropopause compared to the subtropical tropopause?

l.43: Please also refer to the work of Kaluza et al. 2021 (WCD) who showed the existence of a shear layer in tropopause relative coordinates in ERA5.

l.484/485: Sentence reads strange, please rephrase.

References: Kaluza, T., Kunkel, D., and Hoor, P.: On the occurrence of strong vertical wind shear in the tropopause region: a 10-year ERA5 northern hemispheric study, Weather Clim. Dynam., 2, 631–651, https://doi.org/10.5194/wcd-2-631-2021, 2021.