the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 16 Apr 2025
Mapping seasonal nitric acid (HNO3) patterns in the extratropics with nadir-viewing infrared sounders – a retrieval perspective
Abstract. With this paper, we aim to shed light on the extent to which nadir-viewing hyperspectral infrared (IR) sounders can support the study of stratospheric chemical processes and ozone loss in the extratropics. We use CLIMCAPS (Community Long-term Infrared Microwave Combined Atmospheric Processing System) retrievals from JPSS-1 (Joint Polar Satellite System) CrIS (Cross-track Infrared Sounder) measurements as the baseline case. CLIMCAPS retrieves a suite of Earth system variables that includes atmospheric temperature (Tair), water vapor (H2Ovap), carbon dioxide (CO2), carbon monoxide (CO), methane (CH4), ozone (O3), and nitric acid (HNO3). Unlike the Rodgers (2000) Optimal Estimation (OE) retrieval approach, CLIMCAPS regularizes its Bayesian inverse solution dynamically using singular value decomposition (SVD) at run-time to separate measurement signal from noise. We illustrate how the CLIMCAPS approach enables stable retrievals of stratospheric HNO3 under highly variable conditions, allowing characterization of seasonal patterns. Nitric acid is typically used as an indicator species for heterogeneous chemical processing inside the winter stratospheric polar vortices. This paper summarizes our diagnostic evaluation of CLIMCAPS observing capability during the Northern Hemisphere winter of 2019/2020. We contrast CLIMCAPS HNO3 retrievals with those from the limb-viewing MLS (Microwave Limb Sounder) to illustrate the capability of this retrieval approach and lay the foundation for in-depth validation studies in future.
- Preprint
(6080 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-1569', Anonymous Referee #1, 30 Jul 2025
The authors present a progress report on the development of the CLIMCAPS HNO3 retrieval, building on their own preceding work. Background to these efforts is the foreseeable “limb gap” caused by the imminent discontinuation of current limb sounders operation. The aim of the development is to obtain a data product that represents conditions in the stratosphere, specifically temporal and spatial patterns and developments. This goal is attempted to be achieved through the appropriate selection of spectral ranges and dynamic regularization adapted to local atmospheric conditions.
The technical part of the paper, i.e., the investigation into the selection of a suitable parameter set, is carefully presented. However, there are some weaknesses in terms of clarity and structure in the general presentation and also in the presentation and critical discussion of the comparisons with MLS data. Some aspects deserve better or more detailed presentation and discussion.
My general comments are as follows:
The introduction should be better structured, possibly with the help of subsections, since some of the information it contains seems scattered and disjointed. Some later, and more generic, passages could be inserted here (example see below).
The use of the term averaging kernel (AK) is inconsistent and in some cases does not follow standard terminology. Specifically:
The curves in Figure 2 do not represent the AK profiles as stated, but rather the profiles of the diagonal values of the averaging kernel matrix (AKM). More accurately, the rows of this matrix would be called averaging kernels, as it is done, e.g., in Figure 4 of Smith & Barnet (2020). There, the profiles of the diagonal values in Figure 5 are referred to as “averaging kernel diagonal vectors.”Figure 2: Instead of the profiles of the diagonal values of the AKM, it would be desirable and instructive to see the AKs (i.e., the rows of the matrix). By their shapes, one can get an impression of the altitude range which contributes to the HNO3 profile at a given pressure level.
p.7ff, l.202ff Please do not present the mathematical relations between the quantities in the running text, but rather use equations with numbering. This also facilitates references in the text. A more detailed description of the retrieval quantities and their mathematical relations, or at least references to where the respective relations are documented, would be beneficial. Perhaps a brief recapitulation of the retrieval approach and the related quantities in a separate section earlier in the text would be appropriate.
I have the most problems with the presentation of the comparison between VLIMCAPS-x and MLS HNO3 in Figure 4. Only for November and February do the spatial patterns seem to match well, while the same cannot be said for the other three months. There, the spatial patterns are clearly different. Based solely on this initial comparison, one would conclude that the desired/attempted reproduction of spatial gradients (see p.17, l.446) is not working reliably. A critical discussion of this comparison is virtually completely missing, and therefore has to be supplied in the revised manuscript. In particular, the question arises as to how future data users can recognize when the data adequately reflects stratospheric conditions and when it does not.
There are the following comments regarding the technical implementation of the comparison shown in Figure 4: First of all, it is not clear whether and how the MLS data belong to one specific fixed level or whether they have been averaged over several levels. In general, for the eye it is difficult to compare a map of values, such as those used for CLIMCAPS-x data, with the sampled MLS data. A purely visual comparison is easier to do when the CLIMCAPS-x HNO3 data is sampled, and shown, at the locations of the MLS measurements.
Of course, I understand that this presentation is mainly concerned with the development of the retrieval and not meant to anticipate a validation that is still to be carried out in future. On the other hand, the two HNO3 datasets are so different in terms of altitude resolution that I would strongly advise to transform the MLS data (regarded as "the truth"), using AKM and a priori of the CLIMCAPS-x retrieval, to that profile that would be seen by the nadir instrument/retrieval combination. I think there is no other reasonable way for a sound and robust comparison.
And finally, there are these specific points:
p.1, ll.1-2 Shouldn't the title already contain, or refer to, “stratospheric”?
p.3 l.83 "MLS-like" I think this can be stated more general (e.g. "limb-viewing instrument")
p.4 l.102 Strictly speaking, contrary to FORLI, MLS is no retrieval system.
p.5 l.147 Parentheses seem to be unnecessary.
Table 1 Please give the relation between lambdac and Bmax and their definitions before the quantities are used.
p.7, l.192 & p.8, l.215 How can a SNR be destabilized?
p.7, l.193 RTAERR=0 is not an underestimation? That means that destabilization cannot occur?
p.8, l.205 I think that i is meant to be an index for the height level. Is it counted from top to bottom, as implied by the order of the pressure values in Table 1, or from bottom to top?
p.9, l.246 Why do the DOFS represent the sum of all Rfac_i values? Is there a reference?
p.12, l.333 This would be a good place to refer to the (newly introduced) equation number from Section 3.
p.15, ll.421,422 mix-up of East & West?
p16,17 ll454-451 This would fit nicely into the introduction as part of a general motivation subsection.
In Figure 3, the individual maps would benefit from clearer visibility of the longitude and latitude grid.
Why is Greenland so clearly visible in Figure 3 for R3-R5, both in dof and in stddev(dof)? And (roughly) the Pyrenees and the Bering Sea?
p.19, l83 "that the resulting HNO3 product generally reflects real atmospheric variations" is not supported by Figure 4: It is not generally true, but only under certain conditions (see my general comments).Citation: https://doi.org/10.5194/egusphere-2025-1569-RC1 -
RC2: 'Comment on egusphere-2025-1569', Anonymous Referee #2, 04 Aug 2025
Review of "Mapping seasonal nitric acid (HNO3) patterns in the extratopics with nadir-viewing infrared sounders - a retrieval perspective", Smith et al., submitted to Atmos. Meas. Tech.
This manuscript explores information content and seasonal patterns of HNO3 in the upper Arctic atmosphere, as retrieved by several experimental CLIMCAPS versions applied to data collected from CrIS and ATMIS on the JPSS-1 satellite. While HNO3 is part of the operational CLIMCAPS retrieval, that portion of the retrieval has not been optimized as well as other trace gas species, and this paper outlines potential algorithm parameter updates for improved HNO3 retrieval. Due to the nearing end of life for the NASA Aura mission, other sources of stratospheric composition retrieval are needed and this study helps illustrate how IR sounders could partially fill this future information gap for HNO3.Overall, this study is relevant and appropriate for AMT. The subject is timely given the status of NASA Aura.
I find the overall message of the paper unclear, and several aspects need improved explanation. I recommend major revisions, as I think the unclear aspects are crucial to the message of the paper.
Major revisions:The manuscript title suggests a general discussion of nadir IR sounders applied to HNO3 retrieval, and in the initial sections the description is aimed toward comparing nadir IR sounders to a microwave limb sounder (specifically, Aura-MLS). However, in the main body of the paper, most of the comparison is done between FORLI and CLIMCAPS, two retrieval systems applied to nadir IR sounders (IASI and CrIS, respectively). Thus the title and abstract do not seem to match the bulk of the paper, which I find primarily focused on CLIMCAPS HNO3. I would suggest changing the title and abstract to reflect that message.
In much of the descriptions contrasting the FORLI and CLIMCAPS approach, it is stated that the FORLI system only retrieves a total column HNO3. However, the Ronsmans 2016 and 2018 clearly state that FORLI retrieves a profile for HNO3. It was simply a choice made in Ronsmans 2018 to post process the FORLI data to total columns for the analysis done in that paper. Thus, it seems misleading to characterize FORLI as is done in the manuscript; other researchers could simply choose to post process FORLI to extract partial columns in the stratosphere, which would be similar to the CLIMCAPS approach proposed here.
Throughout the manuscript, the dynamic regularization approach of CLIMCAPS is claimed to more effectively separate the stratospheric and tropospheric portions of the HNO3 profile. It is unclear why this is necessariily true, and it is simply stated in the manuscript without any direct analysis or evidence to demonstrate improved separation. I think a more detailed comparison of FORLI and CLIMCAPS retrieval output, compared to MLS, would be needed to demonstrate that CLIMCAPS is an improvement relative to FORLI in this way. Such a comparison is argued to be "beyond scope" for the present paper (line 104) - which is fine, but without such analysis I don't think the strat/troposphere separation is demonstrated.
Similarly, there are some statements made about the R3, R4 configurations being the most optimal. This argument is based purely on the DOFS and AK plots. I am not familiar with the CLIMCAPS dynamic regularization approach, but within the OE framework, there is a similar trend in the strength of regularization vs information content. Increasing the prior variance, or reducing measurement variance in OE, would have a similar effect as increasing Bmax in the CLIMCAPS system: the DFS would increase. The tradeoff that occurs is that we cannot increase the DFS too far as the retrieval can become unstable, and overly sensitive to measurement noise or forward model error; however, that could only be detected by examining the actual retrieved profiles, or via comparison to "truth" or validation data. In other words, the DFS would increase but the RMSE (compared to validation data) would also increase. In the present manuscript, there is no analysis of the quality of the retrieval in any way, all arguments seem to be made solely from the perspective of the information side (DFS). Thus it is unclear as to why R3 or R4 are the most optimal configurations.
To state this another way, we could view the main aim of the R2 - R5 tests is to "tune" the Bmax parameter. As Bmax increases, the DOFS increase - but there is no demonstration of any metric that is getting worse - only that "we argue that it is preferable for HNO3 DOFS to approximate 1.0 " (Line 402), but it is unclear WHY that is preferable.
Minor suggestions:One point that stood out was the relative difference between the prior (0.1 - 1 ppb, line 174) and the retrievals (on average - often > 14 ppb, from Figure 4) - this is more than an order of magnitude. It would be interesting to show Jacobians for these two cases: is the total optical depth of HNO3 small enough, that the Jacobian has a similar shape across this entire concentration range? Similarly, does the DOFS also change drastically across that concentration range?
In Table 1, the right column that lists the pressure grid seems to suggest the trapezoid functions were different in the R1, R2 experiments, versus the R3, R4, and R5. Or, maybe this listing two different ways to describe the 8 trapezoid functions. If it is the former (that there was a change in the trapezoids as part of the experimental versions), that needs more explanation. If the latter, I would suggest splitting out the pressure level information into a separate table. Also, a figure showing the trapezoids would be helpful.
Some minor technical correction:
In line 184 "... seven of the eight broad retrieval layers" - which one was omitted, and why?
Line 220: each term here is written as "\delta x_n x_n^T", is this a typo? I think it should always be "\delta x_n \delta x_n^T"?
Line 235: On the fact that O3 has such a large eigenvalue, I assume the CLIMCAPS retrieval uses a targeted retrieval spectral window on the O3 band (1000-1100 1/cm)? I think that must play a role in "accentuating" the eigenvalues.
Figure 3 & 4: the color maps here seem like they might be saturated, meaning many of the data points are beyond the max value in the color bar. (particularly for upper right panel of Fig 3, and the CLIMCAPS HNO3 maps in Figure 4). Can these be replotted with the max color bar values increased?
Figure 4: Is this the R4 CLIMCAPS test version? I don't see that listed anywhere - would help to state that directly in the figure caption.
What pressure range is sampled with the MLS retrievals?Citation: https://doi.org/10.5194/egusphere-2025-1569-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 397 | 34 | 14 | 445 | 13 | 33 |
- HTML: 397
- PDF: 34
- XML: 14
- Total: 445
- BibTeX: 13
- EndNote: 33
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1