the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 May 2021
Introduction to the ringing effect in satellite hyperspectral atmospheric spectrometry
Abstract. Atmospheric remote spectrometry from space has become in the last 20 years a key component of the Earth monitoring system: their large coverage and deci-kelvin stability have demonstrated their usefulness for weather prediction, atmospheric composition monitoring as well as climate monitoring. It is thus critical to investigate the possible sources of errors associated to this technique. One of them is the so-called "ringing error" that appears in Fourier transform spectrometers when the instrument transmission varies at the scale of the spectral resolution. This paper exposes the theoretical basis of this particular type of radiometric uncertainty. Its sensitivity to instrumental parameters as well as the impact on the radiometrically calibrated measurements is assessed in the context of atmospheric infrared sounding using Fourier transform spectrometers. It is shown that this error is an intrinsic feature of such instruments that could safely be ignored in early-generation instruments but will have to be taken into account in the new generation ones as it can yield a significant degradation of the radiometric error budget.
- Preprint
(1565 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on amt-2021-121', Anonymous Referee #1, 27 May 2021
The paper by Dussarrat et al. treats the ringing artefacts occuring in underresolved FTIR spectra. The treatment provided by the authors is correct and the added example is useful for demonstration of the effect. However, this effect and the methods for handling it when FTIR spectra are used for atmospheric trace gas analysis are known since decades and are well developed. Generally, the quantitative analysis of the measured spectra involves fitting simulated spectra to the measured atmospheric spectra. To my knowledge the modelling of the ringing effect is a natural component of any state-of-the-art processor dealing with FTIR spectra (so the expected ringing is included in the simulated spectrum). Exemplary sensors (/processors) are ATMOS, Mk IV, MIPAS-Envisat, ACE, and IASI. The community is also well aware of the computational costs imposed by this effect on the operational processing of spectra, so techniques as the application of numerical apodisation and different approaches for efficient convolution schemes are common knowledge.
Etaloning is another well-known effect, the presented joint discussion with ringing does not seem a very useful approach to me (instead, when it comes to etaloning, the effect of unresolved etaloning - which remains undetected in the calibration measurements - on atmospheric spectra would deserve some discussion).
As the paper does not treat any new or at least advanced aspects of the problem (note that the assumption of a wavenumber-independent apodisation function can already be a critical simplification and the modelling applied today is typically more refined in this respect). The submitted article would make a nice section in an undergraduate textbook on FTIR spectroscopy, but is far from being adequate for a scientific journal as AMT, as it fails to present "substantial new concepts, ideas, methods, or data".
Citation: https://doi.org/10.5194/amt-2021-121-RC1 -
AC1: 'Reply on RC1', Dorothée Coppens, 15 Jun 2021
1) In view of the referee’s comments, we become aware that the subject of our paper is confusingly introduced by frequent use of the generic term “ringing” (in the title, in the abstract, and in certain sub-sections) for the specific ringing phenomenon we are addressing.
Different kinds of ringing effects, related to instrument physics but also to processing artefacts, occurring in FTIR spectra are quoted as such in literature.
The most common “natural” ringing effect is due to the side lobes in the spectral response of weakly self-apodized FTS. This effect is not a measurement error, but constrains the simulation or inversion of the measurement (eg., handling of negative radiance, width of the spectral response to be accounted for). At the expense of spectral resolution loss, numerical apodisation is an efficient and commonly exploited technique in favour of reduced processing burden and of avoiding processing artefacts.
Artificial ringing effects can be caused by any FFT operation within the processing chain of FTIR measurements through improper smoothing of the band edges. Broadly, these artefacts can be assigned to Gibbs effects.
We agree with the referee that the above-mentioned types of “ringing” are well known and do not need to be introduced again as an article in a scientific journal.
Our paper addresses the residual effect of spectrally variable instrument radiometric transmission on the classically calibrated FTIR radiances. In first approximation, the classical linear two-point radiometric calibration (most commonly: deep space view and warm calibration target) removes the impact of instrument radiometric transmission. We intend to demonstrate the theoretical background of this approximation (sections 2.1 – 2.3), which consists in ignoring the irrecoverable mixing of under-resolved features of the source spectrum with local variations of the instrument transmission during the measurement acquisition. We further intend to illustrate semi-quantitatively the residual radiance errors after calibration for two common sources of instrument transmission variations:
- Low frequency variation (eg. transmission of optical elements), represented by a linear transmission gradient (sections 2.4, 3.1);
- High-frequency variation (eg. etaloning), represented by a modulation (sections 2.5, 3.2).
In practice, both low and high frequency variations will coexist, or dominate over the other in different domains of the spectral band of interest.
In reply to the corresponding reviewer’s remark, etaloning is a major source to the specific ringing effect we are addressing in our paper, and we know already about its criticality in the radiometric accuracy budget of future missions.
To clarify the purpose of our paper and to avoid any confusion, we propose to include the discussion of the ringing terminology as provided above in the introduction (section 1), and to clearly identify the effect we are addressing, throughout the entire paper (title, abstract included), with the original term “Calibration Ringing” in replacement of the admittedly misleading and ambiguous term “Ringing”.
2)
“Calibration Ringing” is, as mentioned, unavoidably linked to the interferogram acquisition. As such, it is not unknown since long ago, but its effect was generally negligible in past EO missions based on spaceborne IR interferometry.
We see two reasons why this is changing:
- Increasing radiometric requirements (in particular radiometric accuracy) and increasing NRT processing requirements.
- Advent of more and more sophisticated FT spectro-imager concepts, acquiring simultaneously up to thousands of interferograms with considerable spatial coverage.
While “Calibration Ringing” becomes a more and more critical contributor to the radiometric error budget, its handling, if uncorrected, requires the radiometric transmission to be explicitly considered by users of the calibrated L1 radiance product. In practice, this means that each detector of a spectro-imager has to be processed as a self-standing instrument (eg. for fast radiative transfer models), which is considered as unfeasible by the user community, all the more in view of NRT processing (and the potential need to update relevant instrument characteristics in case of temporal instability).
The user requirement to process FTIR measurements independently of any detector dependent instrument characteristics is thus further enhanced for multi-detector FT instrument concepts.
To our knowledge, the effect of “calibration ringing” was not considered in the processing of the (mono-pixel) solar occultation and passive limb sounding interferometer missions mentioned in the reviewer’s comment. In the case of IASI (vertical sounding, 2 by 2 detector arrays), “calibration ringing” is considered as irrelevant with respect to radiometric requirements; the instrument radiometric transmission is neither used in the spectral processing, nor it is communicated to the users.
To clarify this point, we propose to introduce a short discussion about the inclusion of SRF distortions in the L2 applications in parts 1 and 2.2.
3)
We intend to introduce a substantial effect that was negligible in past FTIR Earth observation missions.
The present paper is indeed meant to be pedagogical; we try providing a theoretical formulation and illustrating the propagation into the radiometric error budget, illustrated by simple examples (gradient and modulations) without explicit link to any particular FTIR mission.
In this context, we agree with the specific reviewer’s comment about the critical assumption of a wavenumber independent apodisation function. We also agree that this wavenumber dependence is well mastered in current and future operational FTIR processing schemes and therefore felt that a more detailed discussion of this issue would have a distracting effect.
Based on the generic description of the Calibration Ringing effect provided in the present paper, we are preparing a follow-up manuscript addressing the impact of Calibration Ringing, as well as correction schemes.
Citation: https://doi.org/10.5194/amt-2021-121-AC1 -
RC2: 'Reply on AC1', Anonymous Referee #1, 16 Jun 2021
In light of the authors' statement that they work on "a follow-up manuscript addressing the impact of Calibration Ringing, as well as correction schemes", my suggestion would be to merge the current material with the planned article (which obviously will be of a more substantial nature, incorporating new results). The current text could (in a condensed form) serve as introductory section for this extended resubmission.
Citation: https://doi.org/10.5194/amt-2021-121-RC2 -
AC4: 'Reply on RC2', Dorothée Coppens, 11 Aug 2021
The possibility to merge the current article with the planned article on the correction of the effect has been considered. However discussions with our scientific community still show that the described phenomenon is not well understood. No reference has been provided.
Citation: https://doi.org/10.5194/amt-2021-121-AC4
-
AC4: 'Reply on RC2', Dorothée Coppens, 11 Aug 2021
-
AC1: 'Reply on RC1', Dorothée Coppens, 15 Jun 2021
-
RC3: 'Comment on amt-2021-121', Anonymous Referee #2, 13 Jul 2021
The paper under discussion now at AMT, “Introduction to the ringing effect in satellite hyperspectral atmospheric spectrometry” by Dussarrat et al. is a novel and important contribution to the field of FTS.
The typical approach when specifying the Instrument LineShape (ILS) of an FTS sensor is to provide the Maximum Optical Path Difference (MOPD) value and the beginning and ending wavenumbers of the useful parts of the output (radiance) spectrum, and the calibration algorithm/process then removes various other artifacts, such as ILS distortions due to the off-axis effects, re-sampling to a standard output wavenumber scale and other effects specific to a given sensor. When simulating a spectrum (for example using a radiative transfer algorithm), the user therefore typically does not need to know anything other than the MOPD and spectral output range, and in the process of computing the spectrum an artificial function is typically used outside the wavenumber range of interest to smoothly scale the spectrum down to zero in order to limit in-band Gibbs effect ringing. For sensors which produce unapodized (or lightly apodized) spectra, this paper introduces an effect due to the non-flatness of the uncalibrated spectra which has an effect on the ILS which is typically not accounted for in the calibration process, and which is important to understand for sophisticated users of the data.
Individual comments, suggestions:
Paragraph beginning on line 25: Suggest that the authors should clarify what Revercomb meant by “true rining”, as my interpretation is that he did not represent this as a hardware error but rather the same ringing effect described in this paper and not accounted for (yet) in the calibration algorithm/processing of CrIS data.
The description of the effect starting on line number 75 is very good.
The paper includes two important cases of a linear gradient in R(v) (section 2.4) and an etalon effect (section 2.5). While not required, another important and common case is that where R(v) goes to zero at a spectral distance which is not far from the spectral output range. This is an important case, and also a case which may require additional information in the correction process, as there is missing information in the original observations.
In Section 3 (Simulation), the simulations are performed for a MOPD of 1 cm. Suggest also considering case where the MOPD is 0.8 cm, because this is closer to real world examples (CrIS and MTG-IRS). In the Longwave Band, and with the interferogram resonance at 0.64 cm due to the near constant line spacing of CO2 absorption lines, the interferogam amplitudes at MOPD are larger than at 1 cm, so the resulting ringing effects may be larger.
Line 220: Suggest confirming with Tobin and Taylor on the status of conclusions for a correction algorithm for CrIS. In addition, aside from a correction algorithm to produce spectra which have this ringing effect removed, it is also a consideration to include R(v) in the Line-by-line and Fast Model radiance calculations, such that calculated and observed spectra have the same ringing effect, and a correction algorithm may not be needed.
Citation: https://doi.org/10.5194/amt-2021-121-RC3 -
AC2: 'Reply on RC3', Dorothée Coppens, 11 Aug 2021
Thanks a lot for the constructive feedback.
Your comments and suggestions have been taken into account in the revised version we will submit.
Citation: https://doi.org/10.5194/amt-2021-121-AC2
-
AC2: 'Reply on RC3', Dorothée Coppens, 11 Aug 2021
-
RC4: 'Comment on amt-2021-121', Anonymous Referee #3, 15 Jul 2021
The manuscript Introduction tom the ringing effect in satellite hyperspectral atmospheric spectrometry by Dussarrat et al. addresses artefacts commonly and intrinsically occurring in Fourier transform spectrometers. The authors present their mathematical derivation and simulated examples of the effects caused by non-ideal spectral filter responses and the effects caused by etalon filter in the optical path of the instrument.
The authors did not adequately review the state of the art on both effects, which is immediately obvious to knowledgeable readers looking at the reference list and subsequently reading the manuscript. The first effect is a classical signal processing and filtering case-study commonly addressed in textbooks. The application of this case-study on Fourier transform spectrometry without a demonstration based on real data does not warrant a publication in itself. Specificities of this effect in the context presented by the author have not been thoroughly discussed. The derivation seems correct, but is at times unusual and awkward for the reader, perhaps because it is specifically related to processing methods not discussed in the manuscript. Some assumptions made in the derivation are not self-evident and their limits should be discussed or properly referenced.
The second effect, radiometric errors related to Fabry-Perrot filter, is very well known and thoroughly studied, modelled and observed. The authors did not present any compelling arguments for their derivation of this effect and why this radiometric error source requires revisiting. Moreover, the authors failed to acknowledge the considerable amount of prior work on this regard.
A recurring discussion point is the assumed lack of consideration for these effects in the past and a renewed need to address these effects for hyperspectral atmospheric spectrometry. It remains unclear to the reader on what both these assumptions are based and why it is different or more relevant in the context presented by the authors. A discussion on the limitations of the already existing mitigation methods and why (and when) those apply differently for atmospheric hyperspectral measurements could provide a better contribution to the journal. The provided example is well presented, but only considers a numerical apodization as a mitigation approach and does not address the state-of-the-art methods used in processing and data analysis methods. Processing cost is also argumentatively used, without any basis.
Some of the content could be suited as part for the future paper the authors mention in their concluding remarks. The derivation presented (and especially the assumption therein) should nevertheless be thoroughly revisited and presented with more rigor with regards to the assumptions.
Citation: https://doi.org/10.5194/amt-2021-121-RC4 -
AC3: 'Reply on RC4', Dorothée Coppens, 11 Aug 2021
Thanks a lot. Please find our answer:
1) As also discussed with referee 1 (section 1), different kinds of ringing effects, related to instrument physics but also to processing artefacts, occurring in FTIR spectra are quoted as such in literature. To clarify the purpose of our paper and to avoid any confusion, we propose to include the discussion of the ringing terminology and to clearly identify the effect we are addressing, throughout the entire paper (title, abstract included), with the original term “Calibration Ringing” in replacement of the admittedly misleading and ambiguous term “Ringing”.
2) We intended to write this paper because this topic has been discussed recently with several collaborators and we failed to find any pedagogical approach in the literature. Thus, we found it useful to first introduce theoretically the issue in preparation of its correction or mitigation. However, we may have missed the state of the art having, for example, not the right vocabulary, therefore we kindly ask the referee to point us toward a reference.
3) As also discussed with referee 1 (section 2), to our knowledge “Calibration Ringing” was not considered since long ago, as its effect was generally negligible in past EO missions based on spaceborne IR interferometry.
We see two reasons why this is changing:
- Increasing radiometric requirements (in particular radiometric accuracy) and increasing NRT processing requirements.
- Advent of more and more sophisticated FT spectro-imager concepts, acquiring simultaneously up to thousands of interferograms with considerable spatial coverage.
While “Calibration Ringing” becomes a more and more critical contributor to the radiometric error budget, its handling, if uncorrected, requires the radiometric transmission to be explicitly considered by users of the calibrated L1 radiance product. In practice, this means that each detector of a spectro-imager has to be processed as a self-standing instrument (eg. for fast radiative transfer models), which is considered as unfeasible by the user community, all the more in view of NRT processing (and the potential need to update relevant instrument characteristics in case of temporal instability).
The user requirement to process FTIR measurements independently of any detector dependent instrument characteristics is thus further enhanced for multi-detector FT instrument concepts.
To clarify this point, we propose to introduce a short discussion about the inclusion of SRF distortions in the L2 applications in parts 1 and 2.2.
Citation: https://doi.org/10.5194/amt-2021-121-AC3
-
AC3: 'Reply on RC4', Dorothée Coppens, 11 Aug 2021
-
EC1: 'Editor Comment on amt-2021-121', Gabriele Stiller, 18 Aug 2021
With this comment, my intention is to inform the authors about my concerns regarding a re-submission of their paper.
Two of three reviewers have found that the manuscript does not provide enough new material or insights to be published in AMT. While I am not an expert of the respective field, I have some knowledge on Fourier Transform spectroscopy, and I concur with the opinion of the two reviewers. In particular, the presentation in the manuscript made it hard to me to sort out what was basic textbook knowledge and what (if any) is the new contribution to the field. Any reference to previous work and the state of the art is actually not provided. The reference list is very short; it has only one peer-reviewed paper which is from 1988, while there are five references to conference presentations or private communications. While it is not the task of the reviewers to point to the relevant literature as asked by the authors, I will do so: (rather old) publications by J. Brault or L. Delbouille will provide the basics of FTS and data processing. Recent papers on FTS can be found when looking for papers on NDACC FTIR observations, or MIPAS, IASI, MkIV, GLORIA, ACE-FTS, TES, .... calibration or level-1b processing papers. The effect of ringing, either in the recorded scene spectra or the blackbody spectra taken for calibration, is a well-known phenomenon that has been taken into account (and not ignored as the authors state) in all recent systems.
A revised version of their manuscript needs to:
- make reference to the state of the art of Fourier Transform spectroscopy including the handling of the ringing effect;
- describe clearly where the presented approach goes beyond the state of the art, and what the theoretical basis for this approach is; it would be helpful for the reader if the description could be adapted to the usual "vocabulary";
- how this approach has been implemented in the processing of measured (if not available, then simulated) interferograms;
- and finally, quantify what the improvement of the new approach in contrast to the previously described state-of-the art is.Only if the authors feel that these requirements can be met, I recommend resubmission of a revised paper. An attractive alternative could be to use part of this manuscript as introductory section for the announced forthcoming paper, as suggested by two referees.
Citation: https://doi.org/10.5194/amt-2021-121-EC1
Status: closed
-
RC1: 'Comment on amt-2021-121', Anonymous Referee #1, 27 May 2021
The paper by Dussarrat et al. treats the ringing artefacts occuring in underresolved FTIR spectra. The treatment provided by the authors is correct and the added example is useful for demonstration of the effect. However, this effect and the methods for handling it when FTIR spectra are used for atmospheric trace gas analysis are known since decades and are well developed. Generally, the quantitative analysis of the measured spectra involves fitting simulated spectra to the measured atmospheric spectra. To my knowledge the modelling of the ringing effect is a natural component of any state-of-the-art processor dealing with FTIR spectra (so the expected ringing is included in the simulated spectrum). Exemplary sensors (/processors) are ATMOS, Mk IV, MIPAS-Envisat, ACE, and IASI. The community is also well aware of the computational costs imposed by this effect on the operational processing of spectra, so techniques as the application of numerical apodisation and different approaches for efficient convolution schemes are common knowledge.
Etaloning is another well-known effect, the presented joint discussion with ringing does not seem a very useful approach to me (instead, when it comes to etaloning, the effect of unresolved etaloning - which remains undetected in the calibration measurements - on atmospheric spectra would deserve some discussion).
As the paper does not treat any new or at least advanced aspects of the problem (note that the assumption of a wavenumber-independent apodisation function can already be a critical simplification and the modelling applied today is typically more refined in this respect). The submitted article would make a nice section in an undergraduate textbook on FTIR spectroscopy, but is far from being adequate for a scientific journal as AMT, as it fails to present "substantial new concepts, ideas, methods, or data".
Citation: https://doi.org/10.5194/amt-2021-121-RC1 -
AC1: 'Reply on RC1', Dorothée Coppens, 15 Jun 2021
1) In view of the referee’s comments, we become aware that the subject of our paper is confusingly introduced by frequent use of the generic term “ringing” (in the title, in the abstract, and in certain sub-sections) for the specific ringing phenomenon we are addressing.
Different kinds of ringing effects, related to instrument physics but also to processing artefacts, occurring in FTIR spectra are quoted as such in literature.
The most common “natural” ringing effect is due to the side lobes in the spectral response of weakly self-apodized FTS. This effect is not a measurement error, but constrains the simulation or inversion of the measurement (eg., handling of negative radiance, width of the spectral response to be accounted for). At the expense of spectral resolution loss, numerical apodisation is an efficient and commonly exploited technique in favour of reduced processing burden and of avoiding processing artefacts.
Artificial ringing effects can be caused by any FFT operation within the processing chain of FTIR measurements through improper smoothing of the band edges. Broadly, these artefacts can be assigned to Gibbs effects.
We agree with the referee that the above-mentioned types of “ringing” are well known and do not need to be introduced again as an article in a scientific journal.
Our paper addresses the residual effect of spectrally variable instrument radiometric transmission on the classically calibrated FTIR radiances. In first approximation, the classical linear two-point radiometric calibration (most commonly: deep space view and warm calibration target) removes the impact of instrument radiometric transmission. We intend to demonstrate the theoretical background of this approximation (sections 2.1 – 2.3), which consists in ignoring the irrecoverable mixing of under-resolved features of the source spectrum with local variations of the instrument transmission during the measurement acquisition. We further intend to illustrate semi-quantitatively the residual radiance errors after calibration for two common sources of instrument transmission variations:
- Low frequency variation (eg. transmission of optical elements), represented by a linear transmission gradient (sections 2.4, 3.1);
- High-frequency variation (eg. etaloning), represented by a modulation (sections 2.5, 3.2).
In practice, both low and high frequency variations will coexist, or dominate over the other in different domains of the spectral band of interest.
In reply to the corresponding reviewer’s remark, etaloning is a major source to the specific ringing effect we are addressing in our paper, and we know already about its criticality in the radiometric accuracy budget of future missions.
To clarify the purpose of our paper and to avoid any confusion, we propose to include the discussion of the ringing terminology as provided above in the introduction (section 1), and to clearly identify the effect we are addressing, throughout the entire paper (title, abstract included), with the original term “Calibration Ringing” in replacement of the admittedly misleading and ambiguous term “Ringing”.
2)
“Calibration Ringing” is, as mentioned, unavoidably linked to the interferogram acquisition. As such, it is not unknown since long ago, but its effect was generally negligible in past EO missions based on spaceborne IR interferometry.
We see two reasons why this is changing:
- Increasing radiometric requirements (in particular radiometric accuracy) and increasing NRT processing requirements.
- Advent of more and more sophisticated FT spectro-imager concepts, acquiring simultaneously up to thousands of interferograms with considerable spatial coverage.
While “Calibration Ringing” becomes a more and more critical contributor to the radiometric error budget, its handling, if uncorrected, requires the radiometric transmission to be explicitly considered by users of the calibrated L1 radiance product. In practice, this means that each detector of a spectro-imager has to be processed as a self-standing instrument (eg. for fast radiative transfer models), which is considered as unfeasible by the user community, all the more in view of NRT processing (and the potential need to update relevant instrument characteristics in case of temporal instability).
The user requirement to process FTIR measurements independently of any detector dependent instrument characteristics is thus further enhanced for multi-detector FT instrument concepts.
To our knowledge, the effect of “calibration ringing” was not considered in the processing of the (mono-pixel) solar occultation and passive limb sounding interferometer missions mentioned in the reviewer’s comment. In the case of IASI (vertical sounding, 2 by 2 detector arrays), “calibration ringing” is considered as irrelevant with respect to radiometric requirements; the instrument radiometric transmission is neither used in the spectral processing, nor it is communicated to the users.
To clarify this point, we propose to introduce a short discussion about the inclusion of SRF distortions in the L2 applications in parts 1 and 2.2.
3)
We intend to introduce a substantial effect that was negligible in past FTIR Earth observation missions.
The present paper is indeed meant to be pedagogical; we try providing a theoretical formulation and illustrating the propagation into the radiometric error budget, illustrated by simple examples (gradient and modulations) without explicit link to any particular FTIR mission.
In this context, we agree with the specific reviewer’s comment about the critical assumption of a wavenumber independent apodisation function. We also agree that this wavenumber dependence is well mastered in current and future operational FTIR processing schemes and therefore felt that a more detailed discussion of this issue would have a distracting effect.
Based on the generic description of the Calibration Ringing effect provided in the present paper, we are preparing a follow-up manuscript addressing the impact of Calibration Ringing, as well as correction schemes.
Citation: https://doi.org/10.5194/amt-2021-121-AC1 -
RC2: 'Reply on AC1', Anonymous Referee #1, 16 Jun 2021
In light of the authors' statement that they work on "a follow-up manuscript addressing the impact of Calibration Ringing, as well as correction schemes", my suggestion would be to merge the current material with the planned article (which obviously will be of a more substantial nature, incorporating new results). The current text could (in a condensed form) serve as introductory section for this extended resubmission.
Citation: https://doi.org/10.5194/amt-2021-121-RC2 -
AC4: 'Reply on RC2', Dorothée Coppens, 11 Aug 2021
The possibility to merge the current article with the planned article on the correction of the effect has been considered. However discussions with our scientific community still show that the described phenomenon is not well understood. No reference has been provided.
Citation: https://doi.org/10.5194/amt-2021-121-AC4
-
AC4: 'Reply on RC2', Dorothée Coppens, 11 Aug 2021
-
AC1: 'Reply on RC1', Dorothée Coppens, 15 Jun 2021
-
RC3: 'Comment on amt-2021-121', Anonymous Referee #2, 13 Jul 2021
The paper under discussion now at AMT, “Introduction to the ringing effect in satellite hyperspectral atmospheric spectrometry” by Dussarrat et al. is a novel and important contribution to the field of FTS.
The typical approach when specifying the Instrument LineShape (ILS) of an FTS sensor is to provide the Maximum Optical Path Difference (MOPD) value and the beginning and ending wavenumbers of the useful parts of the output (radiance) spectrum, and the calibration algorithm/process then removes various other artifacts, such as ILS distortions due to the off-axis effects, re-sampling to a standard output wavenumber scale and other effects specific to a given sensor. When simulating a spectrum (for example using a radiative transfer algorithm), the user therefore typically does not need to know anything other than the MOPD and spectral output range, and in the process of computing the spectrum an artificial function is typically used outside the wavenumber range of interest to smoothly scale the spectrum down to zero in order to limit in-band Gibbs effect ringing. For sensors which produce unapodized (or lightly apodized) spectra, this paper introduces an effect due to the non-flatness of the uncalibrated spectra which has an effect on the ILS which is typically not accounted for in the calibration process, and which is important to understand for sophisticated users of the data.
Individual comments, suggestions:
Paragraph beginning on line 25: Suggest that the authors should clarify what Revercomb meant by “true rining”, as my interpretation is that he did not represent this as a hardware error but rather the same ringing effect described in this paper and not accounted for (yet) in the calibration algorithm/processing of CrIS data.
The description of the effect starting on line number 75 is very good.
The paper includes two important cases of a linear gradient in R(v) (section 2.4) and an etalon effect (section 2.5). While not required, another important and common case is that where R(v) goes to zero at a spectral distance which is not far from the spectral output range. This is an important case, and also a case which may require additional information in the correction process, as there is missing information in the original observations.
In Section 3 (Simulation), the simulations are performed for a MOPD of 1 cm. Suggest also considering case where the MOPD is 0.8 cm, because this is closer to real world examples (CrIS and MTG-IRS). In the Longwave Band, and with the interferogram resonance at 0.64 cm due to the near constant line spacing of CO2 absorption lines, the interferogam amplitudes at MOPD are larger than at 1 cm, so the resulting ringing effects may be larger.
Line 220: Suggest confirming with Tobin and Taylor on the status of conclusions for a correction algorithm for CrIS. In addition, aside from a correction algorithm to produce spectra which have this ringing effect removed, it is also a consideration to include R(v) in the Line-by-line and Fast Model radiance calculations, such that calculated and observed spectra have the same ringing effect, and a correction algorithm may not be needed.
Citation: https://doi.org/10.5194/amt-2021-121-RC3 -
AC2: 'Reply on RC3', Dorothée Coppens, 11 Aug 2021
Thanks a lot for the constructive feedback.
Your comments and suggestions have been taken into account in the revised version we will submit.
Citation: https://doi.org/10.5194/amt-2021-121-AC2
-
AC2: 'Reply on RC3', Dorothée Coppens, 11 Aug 2021
-
RC4: 'Comment on amt-2021-121', Anonymous Referee #3, 15 Jul 2021
The manuscript Introduction tom the ringing effect in satellite hyperspectral atmospheric spectrometry by Dussarrat et al. addresses artefacts commonly and intrinsically occurring in Fourier transform spectrometers. The authors present their mathematical derivation and simulated examples of the effects caused by non-ideal spectral filter responses and the effects caused by etalon filter in the optical path of the instrument.
The authors did not adequately review the state of the art on both effects, which is immediately obvious to knowledgeable readers looking at the reference list and subsequently reading the manuscript. The first effect is a classical signal processing and filtering case-study commonly addressed in textbooks. The application of this case-study on Fourier transform spectrometry without a demonstration based on real data does not warrant a publication in itself. Specificities of this effect in the context presented by the author have not been thoroughly discussed. The derivation seems correct, but is at times unusual and awkward for the reader, perhaps because it is specifically related to processing methods not discussed in the manuscript. Some assumptions made in the derivation are not self-evident and their limits should be discussed or properly referenced.
The second effect, radiometric errors related to Fabry-Perrot filter, is very well known and thoroughly studied, modelled and observed. The authors did not present any compelling arguments for their derivation of this effect and why this radiometric error source requires revisiting. Moreover, the authors failed to acknowledge the considerable amount of prior work on this regard.
A recurring discussion point is the assumed lack of consideration for these effects in the past and a renewed need to address these effects for hyperspectral atmospheric spectrometry. It remains unclear to the reader on what both these assumptions are based and why it is different or more relevant in the context presented by the authors. A discussion on the limitations of the already existing mitigation methods and why (and when) those apply differently for atmospheric hyperspectral measurements could provide a better contribution to the journal. The provided example is well presented, but only considers a numerical apodization as a mitigation approach and does not address the state-of-the-art methods used in processing and data analysis methods. Processing cost is also argumentatively used, without any basis.
Some of the content could be suited as part for the future paper the authors mention in their concluding remarks. The derivation presented (and especially the assumption therein) should nevertheless be thoroughly revisited and presented with more rigor with regards to the assumptions.
Citation: https://doi.org/10.5194/amt-2021-121-RC4 -
AC3: 'Reply on RC4', Dorothée Coppens, 11 Aug 2021
Thanks a lot. Please find our answer:
1) As also discussed with referee 1 (section 1), different kinds of ringing effects, related to instrument physics but also to processing artefacts, occurring in FTIR spectra are quoted as such in literature. To clarify the purpose of our paper and to avoid any confusion, we propose to include the discussion of the ringing terminology and to clearly identify the effect we are addressing, throughout the entire paper (title, abstract included), with the original term “Calibration Ringing” in replacement of the admittedly misleading and ambiguous term “Ringing”.
2) We intended to write this paper because this topic has been discussed recently with several collaborators and we failed to find any pedagogical approach in the literature. Thus, we found it useful to first introduce theoretically the issue in preparation of its correction or mitigation. However, we may have missed the state of the art having, for example, not the right vocabulary, therefore we kindly ask the referee to point us toward a reference.
3) As also discussed with referee 1 (section 2), to our knowledge “Calibration Ringing” was not considered since long ago, as its effect was generally negligible in past EO missions based on spaceborne IR interferometry.
We see two reasons why this is changing:
- Increasing radiometric requirements (in particular radiometric accuracy) and increasing NRT processing requirements.
- Advent of more and more sophisticated FT spectro-imager concepts, acquiring simultaneously up to thousands of interferograms with considerable spatial coverage.
While “Calibration Ringing” becomes a more and more critical contributor to the radiometric error budget, its handling, if uncorrected, requires the radiometric transmission to be explicitly considered by users of the calibrated L1 radiance product. In practice, this means that each detector of a spectro-imager has to be processed as a self-standing instrument (eg. for fast radiative transfer models), which is considered as unfeasible by the user community, all the more in view of NRT processing (and the potential need to update relevant instrument characteristics in case of temporal instability).
The user requirement to process FTIR measurements independently of any detector dependent instrument characteristics is thus further enhanced for multi-detector FT instrument concepts.
To clarify this point, we propose to introduce a short discussion about the inclusion of SRF distortions in the L2 applications in parts 1 and 2.2.
Citation: https://doi.org/10.5194/amt-2021-121-AC3
-
AC3: 'Reply on RC4', Dorothée Coppens, 11 Aug 2021
-
EC1: 'Editor Comment on amt-2021-121', Gabriele Stiller, 18 Aug 2021
With this comment, my intention is to inform the authors about my concerns regarding a re-submission of their paper.
Two of three reviewers have found that the manuscript does not provide enough new material or insights to be published in AMT. While I am not an expert of the respective field, I have some knowledge on Fourier Transform spectroscopy, and I concur with the opinion of the two reviewers. In particular, the presentation in the manuscript made it hard to me to sort out what was basic textbook knowledge and what (if any) is the new contribution to the field. Any reference to previous work and the state of the art is actually not provided. The reference list is very short; it has only one peer-reviewed paper which is from 1988, while there are five references to conference presentations or private communications. While it is not the task of the reviewers to point to the relevant literature as asked by the authors, I will do so: (rather old) publications by J. Brault or L. Delbouille will provide the basics of FTS and data processing. Recent papers on FTS can be found when looking for papers on NDACC FTIR observations, or MIPAS, IASI, MkIV, GLORIA, ACE-FTS, TES, .... calibration or level-1b processing papers. The effect of ringing, either in the recorded scene spectra or the blackbody spectra taken for calibration, is a well-known phenomenon that has been taken into account (and not ignored as the authors state) in all recent systems.
A revised version of their manuscript needs to:
- make reference to the state of the art of Fourier Transform spectroscopy including the handling of the ringing effect;
- describe clearly where the presented approach goes beyond the state of the art, and what the theoretical basis for this approach is; it would be helpful for the reader if the description could be adapted to the usual "vocabulary";
- how this approach has been implemented in the processing of measured (if not available, then simulated) interferograms;
- and finally, quantify what the improvement of the new approach in contrast to the previously described state-of-the art is.Only if the authors feel that these requirements can be met, I recommend resubmission of a revised paper. An attractive alternative could be to use part of this manuscript as introductory section for the announced forthcoming paper, as suggested by two referees.
Citation: https://doi.org/10.5194/amt-2021-121-EC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 754 | 471 | 60 | 1,285 | 36 | 32 |
- HTML: 754
- PDF: 471
- XML: 60
- Total: 1,285
- BibTeX: 36
- EndNote: 32
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
