Referee comments

"On the Capability of the Changing Atmosphere Infra-Red Tomography explorer 

(CAIRT) candidate mission to constrain O3 and H2O in the upper troposphere

and lower stratosphere" 

Quentin Errera et al.

AMT Discussion started 27Dec25

--------

SUMMARY 

The CAIRT instrument was a limb-viewing infrared radiometer proposed for the

ESA EE11 mission, one of two instruments which were considered at the final

selection. Although not selected ESA encouraged continued scientific studies.

This paper focuses on just one aspect of CAIRT which is to improve

measurements of O3 and H2O in the UTLS region, in this case via

incorporation of profiles into a data assimilation scheme. The

improvement is established not only with respect to a control but also

the long-established MLS instrument. For both instruments the

approach is to sample fields generated by an independent model at the

observation locations of the two instruments, apply the instrument

averaging kernels to simulate the measured profiles, and incorporate

these into a data assimilation scheme based on an different model.

The conclusions are that CAIRT shows improvement in the modelled

O3 and H2O fields, and is generally better than MLS, but, given the

set-up and CAIRT's much higher profile density, any other conclusion

would indeed have been surprising.

I have no major criticism with anything presented in the paper 

although overall I feel it includes too much unecessary technical

detail on model settings and geolocation procedures, better suited

to appendices or supplementary data. But that's just my preference.

GENERAL COMMENTS

I would like the authors to clarify the following points 

(see detailed comments below for further elaboration)

(a) There is mention of systematic errors but I would like a bit

more details on what these are, how they were assessed and how they

were handled in the data assimilation.

(b) The handling of cloud-contaminated CAIRT data is unclear, particularly

how it relates to "superobbing".

I also offer the following suggestions, for the authors consideration,

which I think would improve the paper

(c) There should be a plot of the limb spectrum as observed by CAIRT, 

identifying features associated with the main molecular species.

(d) I appreciate that there is more to data assimilation than simply

interpolating retrieved fields, but it would have been interesting to

see how a simple gridding/binning of the CAIRT and MLS profiles on the 

assimilation model grid compares with data assimilation approach (as

an extra test case).

(e) While the zonal mean plots can be interpreted in a number of different

ways according to the prejudices of the viewer, there should be a 

summary table quantifying the improvement with a single number or two,

eg mean bias and SD, or something fancier such as information content

added by each instrument. But, anyway, a hard number.

DETAILED COMMENTS

P3 L76 "review"

Table 1: While I understand that the O3 uncertainty requirements were

expressed as ppbv It would be useful to have some indication of actual

O3 concentrations in the UTLS region in order to put these values

into context (I eventually found this information by inspecting the

colour scales in Fig 5).

P4 L88: Is resolution 0.25cm-1 here specified before or after apodisation?

P4 L96: It would be interesting to know the minimum pixel size, ie before

        averaging spectra what is the FOV size of an individual spectrum? 

P4 L97: I don't think you can claim a "strong" heritage based on just a

        single predecessor.

P4 L99: This could be misunderstood. MIPAS actually had several

        detectors, covering different spectral ranges, all co-aligned to view

        the same 'pixel'.

P4 L100: Also liable to be misunderstood - while MIPAS (and MLS)

         produced profiles along a single track CAIRT would have produced

         profiles along multiple tracks so it's unclear whether these figures 

         refer to improvements in sampling along just a single orbital track 

         or, say, total number of profiles per day. 

P5 L121: "within which..."

P5 L127 & L138: Since there is quite a bit discussion in the main text based 

         on what these figures in the supplementary data show, they should 

         have been included in the main body of the paper.

P8 L216: Here it is 0.2cm-1 sampling but on P4 L88 a figure 0.25cm-1 is

         given for the instrument. 

         At the end of the paragraph it explains that this is after

         apodisation.  Knowing how this works, I assume it therefore

         represents a convolution of spectra calculated on a

         much-finer spacing with the apodised instrument lineshape,

         and sampled at 0.2cm-1, but just reading "computed with

         0.2cm-1 spectral sampling" a non-specialist might assume this

         was the original fine grid spacing.

P9 L223: "factorized random error covariance matrices". I don't understand 

         what "factorized" means in this context. If I had to guess I would say

         some sort of EOF/SVD decomposition of the L2 random error covariance. 

P9 L224: Non-LTE effects can be a problem for modelling stratospheric H2O

         lines. Were these included as part of the systematic error?

P9 L241: I didn't understand the statement setting y_a to zero to avoid

         any a priori bias. Isn't this going to bias all terms in the simulated

         L2 profile towards zero? 

P9 L245: I didn't understand this. How do clouds influence retrievals beyond

         the physical cloud boundary (unless you are just referring to partly

         cloud-filled pixels containing a cloud-contaminated component).

P9 L246: I assume the CAMS cloud fraction is the fraction in the CAMS

         model grid cell containing cloud of any sort, ie horizontal fraction 

         over some indeterminate but presumably large area compared with 

         CAIRT horizontal FOV extent.  It is unclear how this maps into a 

         limb-view.

 

P9 L247: If a cloud is assumed to have finite vertical extent below the

         cloud top, it is not obvious why 50km extent along the line-of-sight

         prevents any retrieval beneath cloudy regions. If the cloud were high 

         enough (eg PSCs or high-level cirrus) you would presumably still be 

         retrieving along the line-of-sight through the cloud-free paths

         below it?

Fig 2: It looks like this box spans the coast of East Africa. It would have

       been nice to add some geographical features and maybe a length scale to

       give a better impression of the spatial sampling of CAIRT.

P10 L254: With the reference to interferograms here I am now uncertain whether

          the assimilation is of CAIRT L2 profiles, L1C spectra, or L1A 

          interferograms?

P10 Eqs 2 & 3: This is confusing. You're using bold font for y, as in

       Eq 1, suggesting each y_i is a profile. Yet discussion of excluding

       cloud contaminated values suggests this averaging is performed

       level-by-level, ie y_i are the set of profile values at a single 

       altitude. If it is level-by-level does all the off-diagonal 

       information in the the L2 random error covariance get ignored in the 

       super-obbing? 

P11 L276: Missing Section reference ("??").

P11 L292: It is stated that the CAIRT data for the tropics is representative

          of other latitudes. But cloud heights and frequency in the tropics 

          are probably higher than other latitudes, and tropopause temperatures

          lower, so this seems pessimistic rather than representative.

P12 L296: I don't know how valid it is to compare systematic errors without

          a consistent definition of systematic errors for the two instruments.

P12 L303: Isn't Fig 6 the H2O equivalent of Fig 5? (rather than Fig S3)

Figs 5 and 6: Given the superficial similarity between CAMS and the two

         instruments, it would have been informative to add a 3rd column showing

         the differences Instrument-CAIRT. 

P14 L327: I was not aware that systematic errors could be included in data 

          assimilation, except as a fudge by simply inflating the random error.

          Can you explain a bit more how this was achieved?

P16 Section 8: Which of the four CAIRT assimilation runs listed in Table 2

          was being used for the results shown in this section? 

          From the text on P18 L354-360 I would guess AR_CAIRT.

P16 Section 8: It would be useful, at the start of this section, to have

          the equations which define the NMB, NSD and correl.coefficient.

          For example, I can think of two different ways of normalising the

          SD, based on instrument random error or atmospheric variability.

Typography: 

Note the distinction (available in LaTeX) between -, --, and --- for

different sorts of punctuation, as well as $-$ for a minus sign.  

Here they seem to have been used interchangably and randomly.

Also: undesirable paragraph indentation immediately following

equations (1) and (4). In LaTeX this can be avoided by not leaving

blank line after the equation.

----

General review

This paper demonstrates the capability of the recently proposed satellite instrument CAIRT to measure atmospheric trace gases at unprecedented scales. Although the instrument was designed for ESA’s Earth Explorer 11 programme and was ultimately not selected, its concept remains highly relevant for future missions and international collaborations. The core component of CAIRT is an imaging Fourier Transform Spectrometer that provides a two-dimensional tomographic view of atmospheric limb emissions. The maturity of this concept builds on the success of the MIPAS mission, which performed limb scanning along a single line-of-sight, as well as on the heritage of GLORIA, which has flown on balloon and aircraft platforms. In addition, CAIRT measurements are co-registered with those from the MetOp second-generation nadir-viewing instruments, such as IASI-NG, enabling them to observe the same air mass from complementary viewing geometries. The sampling density of CAIRT is approximately an order of magnitude higher than that of MLS and MIPAS. Owing to its fine horizontal resolution and high data density, CAIRT would represent a step change relative to existing limb-sounding instruments. It is particularly well suited for investigating fine-scale atmospheric structures, such as stratosphere-troposphere exchange in the UTLS region, thereby addressing a key observational need of the atmospheric research community.

To simulate and evaluate the CAIRT observations, the authors designed a processing chain comprising nature runs, control runs, and assimilation runs. State-of-the-art models were employed together with realistic instrument parameters, and the results were evaluated through comparison with MLS data. The methodology is well structured, logically organized, and clearly presented. The examples are representative, and the results are convincing, meeting the stated objectives.

Despite the extensive scope of the study, the manuscript is concise, follows a logical progression, and is easy to understand. I therefore recommend the manuscript for publication.

Minor corrections:

Line 18: Typo. “The results show that CAIRT (1) …” should be “The results show that (1) CAIRT …”.

Line 19: Typo. “than MLS –“ should be  “than MLS; “ .

Line 76: Typo. “reveiw” should be “review”.

Line 276: Typo. “see Sect. ??” should be “see Sect. 4.2”.

Line 280: Typo. “into account The …“ should be “into account. The …”.

Table 2 (between Lines 333 and 334): “BASCOE-CTM” may include the full expression “BASCOE-CTM (Chemistry Transport Model”. Similarly, “BASCOE-EnKF” may include the full expression “BASCOE-EnKF (Ensemble Kalman Filter)”.

Line 375: The acronym “NWP” appears here for the first time; the full expression “Numerical Weather Forecast (NWP)” should be provided.

Line 400: Use only the acronym “NWP”, as it has already been defined earlier.

Line 456: “Walker, and K.:” should be “Walker, K.:”.

Line 537: Typo. “reveiw” should be “review”.