Investigating long-term changes in polar stratospheric clouds above Antarctica: A temperature-based approach using spaceborne lidar detections

Leroux, Mathilde; Noel, Vincent

doi:https://doi.org/10.5194/egusphere-2024-131

Preprints

https://doi.org/10.5194/egusphere-2024-131

Preprints

19 Jan 2024

| 19 Jan 2024

Investigating long-term changes in polar stratospheric clouds above Antarctica: A temperature-based approach using spaceborne lidar detections

Mathilde Leroux and Vincent Noel

Abstract. Polar stratospheric clouds play a significant role in the seasonal thinning of the ozone layer by facilitating the activation of stable chlorine and bromine reservoirs into reactive radicals, as well as prolonging the ozone depletion by removing HNO₃ and H₂O of the stratosphere by sedimentation. In a context of climate change, the cooling of the lower polar stratosphere could enhance the PSC formation and by consequence cause more ozone depletion. There is thus a need to document the evolution of the PSC cover to better understand its impact on the ozone layer. In this article we present a statistical model based on the analysis of the CALIPSO PSC product from 2006 to 2020. The model predicts the daily regionally-averaged PSC density by pressure level derived from stratospheric temperatures. Applying our model to stratospheric temperatures from the CALIPSO PSC product over the 2006–2020 period shows it is robust in the stratosphere between 10 and 150 hPa, reproducing well PSC variations over daily timescales and seasonal differences (2006–2020). The model reproduces well the PSC seasonal progression, even during disruptive events like stratospheric sudden warmings, except for years characterized by volcanic eruptions. We apply our model to gridded stratospheric temperatures from reanalyses over the complete south pole domain to evaluate changes in PSC seasons over the 1980–2021 period. We find two distinct periods in the evolution of the PSC season duration. Between 1980 and 2000, the PSC season increased by 15 days at 10–20 hPa with an increasing lengthening as we descend in altitudes to reach 30 days at 100–150 hPa. This lengthening is in possible relation with major eruptions occurring over this period. After 2000, a temporary drop mostly visible at high (10–20 hPa) and lower altitude (100–150 hPa) is followed by a progressive increase in PSC season duration. Over the 1980–2020 period, the PSC season increased by 20 days between 30–100 hPa. These changes are altitude-dependent and statistically significant. We discuss the impact of non-temperature stratospheric changes on the variations of PSC seasons.

Received: 16 Jan 2024 – Discussion started: 19 Jan 2024

Download & links

Mathilde Leroux and Vincent Noel

Status: closed

RC1:
'Comment on egusphere-2024-131', Anonymous Referee #1, 05 Feb 2024
The authors present a method for estimating Antarctic PSC cover from modelled temperatures based on a simple temperature threshold and the verification of PSC occurrence from spaceborne CALIPSO lidar measurements. The work is of interest to the readers of ACP. However, the presentation style gives the impression that a thesis was transformed into a publication without properly accounting for the rigorous trimming that is usually advisable for such a process. This is reflected in a somewhat unfocussed and repetitive presentation of findings, redundant text, the inclusion of results that don't necessarily advance the reader's understanding, and an extensive appendix. Unfortunately, the work does not include a prognosis of the likely development of future PSC occurrence. While it's debatable that this issue is implied by the title, it would certainly increase the importance of this work. Therefore, major revisions are needed to improve the quality of the paper.
General comments:
I suggest to omit the redundant text related to what will be presented in each section. This also holds for the abundant references to other sections and figure elements that should be described in the figure captions (red dots, solid line, etc)

Section 2.1 contains lots of information that isn’t really relevant for this study. I suggest shortening to just some basic information on CALIPSO and the PSC mask v2.

The authors need to be more precise regarding their wording. They refer to CALIPSO measurements as points. However, it is not always clear if this means profiles or height bins. This is particularly important in Section 3.1 where it is unclear if the constructed grids include profiles per grid cell (N?), PSC bins per pressure level (n?), or a mixture of both. I also disagree with the term simulating PSC densities. The developed statistical model has no prognostic capabilities but rather uses a threshold to ESTIMATE the expected PSC densities.

The authors should elaborate on the transformation of CALIOP data from their native height resolution to that imposed by MERRA-2 pressure levels which are not equidistant. It would be interesting to learn about the number of CALIOP height bins within the different pressure levels and how this might affect PSC detection rate. Is a PSC detected as long as there’s at least one PSC height bin in the PSC mask v2 product? Or is a fractional threshold used for PSC detection, i.e. a certain percentage of height bins in a layer has to feature PSCs?

The presentation of the development of the statistical model in Section 3 should be improved. The current mix of presenting findings for a single day (Figure 1), a month (Figure 2), a year (Figure 4), and finally the entire time series is rather confusing and not well motivated. I suggest to include a flow chart that outlines the steps in the development of the statistical model and where iterative loops are involved. Right now, it is not entirely clear if the authors developed the model first and adapted the temperature threshold subsequently or vice versa. The text in Section 3.1 indicates that the analysis goes back and forth between CALIOP’s native height spacing and the pressure levels. I suggest to stick to a fixed height grid in the presentation of results and to cover the transformation between grids in more detail in the methods section.

I suggest to omit Section 4. While the focus in individual years is useful to demonstrate the skill of the statistical model, the authors are going in circles by inferring fit parameters and temperature thresholds for a single year and subsequently applying it to the same year. A truly independent assessment would either need to split the available observations into data for training and verification or apply fit parameters and temperature thresholds derived for another year or the entire time series. I suggest to go with the latter, i.e, skip directly to Section 5.

The discussion in Section 5 nicely presents the findings and the capability of the statistical model. However, the presentation is rather repetitive going from one pressure level to the next. I suggest to further condense the results into a presentation that covers all height levels (as in Figure 12) and years (as in Figure 7) with PSC density variations as colour map. The same applies to Section 6. Another thing that remains entirely unclear to me are the PSC density thresholds of 10%, 25%, and 40%. How have these values been derived? Should they be the same at all height level? I suggest to add some text that motivates their selection.

Section 6 presents the application of the inferred statistical model to a time period no covered by the CALIPSO time series. This is the core advancement of the work as it expands available knowledge to climate time scales. However, by including all model data south of 60 degree S, the authors have skipped a step that would have allowed for a fairer assessment of the long-term development of PSC densities. Why not consider the same region covered by CALIOP observations first and extending southward afterwards? This might even provide you with an quantitative explanation or a correction for the discrepancies in Figures F3 and G3. Simply accounting this to the difference in area without having a closer look doesn’t seem right to me. In that context, I would also have expected a more conclusive discussion (with references) regarding the step in the trend line at around 1999 in Figures 13 and 15. Is there a physical explanation for those? There must be time series of stratospheric temperature or large-scale circulation that could be considered? If there have been events such as SSW, it’s worthwhile to mark them in the plots of time series. Also relating to the trends: the authors refer to a significant increase. If they have performed significance tests, I suggest to provide the results (e.g., kind of test, p-value) to support their statements.

I strongly advise the authors to add a section in which the statistical model is applied to the output of a climate model to estimate PSC coverage and season length until 2100. This would add a major novelty to the study and strengthen its overall scope. I can understand that the authors might want to focus on this during subsequent work. However, using a single climate model that best resemble stratospheric temperature of MERRA-2 during 1980 to 2020 to demonstrate the feasibility of their approach wouldn’t impede more detailed follow-up studies that might involve a set of climate models.
Citation: https://doi.org/10.5194/egusphere-2024-131-RC1
- AC1: 'Reply on RC1', Mathilde Leroux, 05 Apr 2024
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-131/egusphere-2024-131-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-131-AC1
RC2:
'Comment on egusphere-2024-131', Anonymous Referee #2, 09 Feb 2024

In this work, the authors present a statistical model to evaluate the existence period of Polar Stratospheric Clouds (PSCs) from global gridded stratospheric temperature datasets. The model parameterization is derived from PSC-observations performed by the CALIOP lidar on the CALIPSO satellite between 2006 and 2020. Subsequently this model is used to analyse the trend of the PSC season length over Antarctica at different stratospheric pressure levels over an extended period from 1980 to 2020 based on the MERRA2 reanalysis dataset.
In general, the manuscript is well written albeit a bit lengthy in some places where the material could be presented more compressed with less repetitions. I would suggest to really concentrate only on the really necessary parts and move some of the less significant items in the appendix. Content wise the manuscript fits into the scope of ACP and I would support its publication after consideration/implementation of the following comments.
Specific comments
Chapter 2.1:
Seems too long with information not needed in the following. Please concentrate on describing the dataset which is used within the study and put references for further reading.
Chapter 2.2:
Here (or later in the discussion section) a discussion on the reliability of MERRA2 would be valuable to be able to judge on the conclusions drawn on the pre-CALIPSO periods. As an example, one may refer to the “SPARC Reanalysis Intercomparison Project (S-RIP) Final Report” or other work, especially comparing temperatures with ERA-5.
193: ‘fits the best for most of the plots is a polynomial of degree 2’
What is the criterion for the statement ‘best’? A higher-order polynomial would mathematically fit better but there might be other reasons (e.g. simplicity) to choose 2^nd order. What is the norm which is minimized for the fit (RMS)?
218: ‘The 𝑇𝑝𝑠𝑐 and parameters which lead to the smallest MAE are selected for the month and pressure level considered’
How well defined is this minimum? A plot of MAE as f(Tpsc) would be helpful to judge on the uniqueness of the result.
Further, the choice to use monthly changing threshold temperatures seems a bit arbitrary. Have you tried to perform some kind of ‘smooth’ transition between the monthly Tpsc values or what is your argument to use such a ‘coarse’ binning.
238: ‘The difference is particularly important at lower altitudes’
This finding should be supported by references to previous publications.
248: ‘there is often a large difference between the temperature threshold 𝑇𝑁𝐴𝑇 at’
Can you support this statement by pointing to papers describing these obvious differences based on CALIOP observations?
Chapter 4:
Here only the simulated PSC densities on basis of monthly temperature thresholds for the related year are shown. I think the results obtained with the multi-year derived T-thresholds are much more relevant. Thus, I would strongly suggest to show and discuss also those. In principle this is shown in Figs. 6-8 in chapter 5, albeit so squeezed that the curves cannot be distinguished.
551: ‘This longer period can be attributed to the gridded dataset including latitudes south of 82°S, absent from the CALIPSO dataset.’
This statement can easily (and should) be substantiated by performing the respective analysis only down to 82S.
Chapter 6:
These findings strongly depend on the accuracy of the MERRA2 analysis, esp. on their trends. Please discuss the influence of related uncertainties on your findings and, if possible, point to supporting material.
Data availability:
The data used for the model calculations (e.g. Tpsc, polynomial coefficients) have to be made publicly available in digital form.

Technical comments
10: ‘of’ -> ‘from’
36: ‘acid sulfuric’ -> ‘sulfuric acid’
295: ‘negative quasi biennial oscillation (QBO)’ -> ‘easterly (negative) phase of the … (QBO)’
409: ‘Fig. 7b’ -> ‘Fig. 8b’ (?)

Citation: https://doi.org/10.5194/egusphere-2024-131-RC2
- AC2: 'Reply on RC2', Mathilde Leroux, 05 Apr 2024
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-131/egusphere-2024-131-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-131-AC2

Status: closed

RC1:
'Comment on egusphere-2024-131', Anonymous Referee #1, 05 Feb 2024
The authors present a method for estimating Antarctic PSC cover from modelled temperatures based on a simple temperature threshold and the verification of PSC occurrence from spaceborne CALIPSO lidar measurements. The work is of interest to the readers of ACP. However, the presentation style gives the impression that a thesis was transformed into a publication without properly accounting for the rigorous trimming that is usually advisable for such a process. This is reflected in a somewhat unfocussed and repetitive presentation of findings, redundant text, the inclusion of results that don't necessarily advance the reader's understanding, and an extensive appendix. Unfortunately, the work does not include a prognosis of the likely development of future PSC occurrence. While it's debatable that this issue is implied by the title, it would certainly increase the importance of this work. Therefore, major revisions are needed to improve the quality of the paper.
General comments:
I suggest to omit the redundant text related to what will be presented in each section. This also holds for the abundant references to other sections and figure elements that should be described in the figure captions (red dots, solid line, etc)

Section 2.1 contains lots of information that isn’t really relevant for this study. I suggest shortening to just some basic information on CALIPSO and the PSC mask v2.

The authors need to be more precise regarding their wording. They refer to CALIPSO measurements as points. However, it is not always clear if this means profiles or height bins. This is particularly important in Section 3.1 where it is unclear if the constructed grids include profiles per grid cell (N?), PSC bins per pressure level (n?), or a mixture of both. I also disagree with the term simulating PSC densities. The developed statistical model has no prognostic capabilities but rather uses a threshold to ESTIMATE the expected PSC densities.

The authors should elaborate on the transformation of CALIOP data from their native height resolution to that imposed by MERRA-2 pressure levels which are not equidistant. It would be interesting to learn about the number of CALIOP height bins within the different pressure levels and how this might affect PSC detection rate. Is a PSC detected as long as there’s at least one PSC height bin in the PSC mask v2 product? Or is a fractional threshold used for PSC detection, i.e. a certain percentage of height bins in a layer has to feature PSCs?

The presentation of the development of the statistical model in Section 3 should be improved. The current mix of presenting findings for a single day (Figure 1), a month (Figure 2), a year (Figure 4), and finally the entire time series is rather confusing and not well motivated. I suggest to include a flow chart that outlines the steps in the development of the statistical model and where iterative loops are involved. Right now, it is not entirely clear if the authors developed the model first and adapted the temperature threshold subsequently or vice versa. The text in Section 3.1 indicates that the analysis goes back and forth between CALIOP’s native height spacing and the pressure levels. I suggest to stick to a fixed height grid in the presentation of results and to cover the transformation between grids in more detail in the methods section.

I suggest to omit Section 4. While the focus in individual years is useful to demonstrate the skill of the statistical model, the authors are going in circles by inferring fit parameters and temperature thresholds for a single year and subsequently applying it to the same year. A truly independent assessment would either need to split the available observations into data for training and verification or apply fit parameters and temperature thresholds derived for another year or the entire time series. I suggest to go with the latter, i.e, skip directly to Section 5.

The discussion in Section 5 nicely presents the findings and the capability of the statistical model. However, the presentation is rather repetitive going from one pressure level to the next. I suggest to further condense the results into a presentation that covers all height levels (as in Figure 12) and years (as in Figure 7) with PSC density variations as colour map. The same applies to Section 6. Another thing that remains entirely unclear to me are the PSC density thresholds of 10%, 25%, and 40%. How have these values been derived? Should they be the same at all height level? I suggest to add some text that motivates their selection.

Section 6 presents the application of the inferred statistical model to a time period no covered by the CALIPSO time series. This is the core advancement of the work as it expands available knowledge to climate time scales. However, by including all model data south of 60 degree S, the authors have skipped a step that would have allowed for a fairer assessment of the long-term development of PSC densities. Why not consider the same region covered by CALIOP observations first and extending southward afterwards? This might even provide you with an quantitative explanation or a correction for the discrepancies in Figures F3 and G3. Simply accounting this to the difference in area without having a closer look doesn’t seem right to me. In that context, I would also have expected a more conclusive discussion (with references) regarding the step in the trend line at around 1999 in Figures 13 and 15. Is there a physical explanation for those? There must be time series of stratospheric temperature or large-scale circulation that could be considered? If there have been events such as SSW, it’s worthwhile to mark them in the plots of time series. Also relating to the trends: the authors refer to a significant increase. If they have performed significance tests, I suggest to provide the results (e.g., kind of test, p-value) to support their statements.

I strongly advise the authors to add a section in which the statistical model is applied to the output of a climate model to estimate PSC coverage and season length until 2100. This would add a major novelty to the study and strengthen its overall scope. I can understand that the authors might want to focus on this during subsequent work. However, using a single climate model that best resemble stratospheric temperature of MERRA-2 during 1980 to 2020 to demonstrate the feasibility of their approach wouldn’t impede more detailed follow-up studies that might involve a set of climate models.
Citation: https://doi.org/10.5194/egusphere-2024-131-RC1
- AC1: 'Reply on RC1', Mathilde Leroux, 05 Apr 2024
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-131/egusphere-2024-131-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-131-AC1
RC2:
'Comment on egusphere-2024-131', Anonymous Referee #2, 09 Feb 2024

In this work, the authors present a statistical model to evaluate the existence period of Polar Stratospheric Clouds (PSCs) from global gridded stratospheric temperature datasets. The model parameterization is derived from PSC-observations performed by the CALIOP lidar on the CALIPSO satellite between 2006 and 2020. Subsequently this model is used to analyse the trend of the PSC season length over Antarctica at different stratospheric pressure levels over an extended period from 1980 to 2020 based on the MERRA2 reanalysis dataset.
In general, the manuscript is well written albeit a bit lengthy in some places where the material could be presented more compressed with less repetitions. I would suggest to really concentrate only on the really necessary parts and move some of the less significant items in the appendix. Content wise the manuscript fits into the scope of ACP and I would support its publication after consideration/implementation of the following comments.
Specific comments
Chapter 2.1:
Seems too long with information not needed in the following. Please concentrate on describing the dataset which is used within the study and put references for further reading.
Chapter 2.2:
Here (or later in the discussion section) a discussion on the reliability of MERRA2 would be valuable to be able to judge on the conclusions drawn on the pre-CALIPSO periods. As an example, one may refer to the “SPARC Reanalysis Intercomparison Project (S-RIP) Final Report” or other work, especially comparing temperatures with ERA-5.
193: ‘fits the best for most of the plots is a polynomial of degree 2’
What is the criterion for the statement ‘best’? A higher-order polynomial would mathematically fit better but there might be other reasons (e.g. simplicity) to choose 2^nd order. What is the norm which is minimized for the fit (RMS)?
218: ‘The 𝑇𝑝𝑠𝑐 and parameters which lead to the smallest MAE are selected for the month and pressure level considered’
How well defined is this minimum? A plot of MAE as f(Tpsc) would be helpful to judge on the uniqueness of the result.
Further, the choice to use monthly changing threshold temperatures seems a bit arbitrary. Have you tried to perform some kind of ‘smooth’ transition between the monthly Tpsc values or what is your argument to use such a ‘coarse’ binning.
238: ‘The difference is particularly important at lower altitudes’
This finding should be supported by references to previous publications.
248: ‘there is often a large difference between the temperature threshold 𝑇𝑁𝐴𝑇 at’
Can you support this statement by pointing to papers describing these obvious differences based on CALIOP observations?
Chapter 4:
Here only the simulated PSC densities on basis of monthly temperature thresholds for the related year are shown. I think the results obtained with the multi-year derived T-thresholds are much more relevant. Thus, I would strongly suggest to show and discuss also those. In principle this is shown in Figs. 6-8 in chapter 5, albeit so squeezed that the curves cannot be distinguished.
551: ‘This longer period can be attributed to the gridded dataset including latitudes south of 82°S, absent from the CALIPSO dataset.’
This statement can easily (and should) be substantiated by performing the respective analysis only down to 82S.
Chapter 6:
These findings strongly depend on the accuracy of the MERRA2 analysis, esp. on their trends. Please discuss the influence of related uncertainties on your findings and, if possible, point to supporting material.
Data availability:
The data used for the model calculations (e.g. Tpsc, polynomial coefficients) have to be made publicly available in digital form.

Technical comments
10: ‘of’ -> ‘from’
36: ‘acid sulfuric’ -> ‘sulfuric acid’
295: ‘negative quasi biennial oscillation (QBO)’ -> ‘easterly (negative) phase of the … (QBO)’
409: ‘Fig. 7b’ -> ‘Fig. 8b’ (?)

Citation: https://doi.org/10.5194/egusphere-2024-131-RC2
- AC2: 'Reply on RC2', Mathilde Leroux, 05 Apr 2024
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-131/egusphere-2024-131-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2024-131-AC2

Mathilde Leroux and Vincent Noel

Viewed

Total article views: 320 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
232	74	14	320	5	2

HTML: 232
PDF: 74
XML: 14
Total: 320
BibTeX: 5
EndNote: 2

Views and downloads (calculated since 19 Jan 2024)

Month	HTML	PDF	XML	Total
Jan 2024	98	30	1	129
Feb 2024	67	14	5	86
Mar 2024	30	8	0	38
Apr 2024	37	22	8	67

Cumulative views and downloads (calculated since 19 Jan 2024)

Month	HTML	PDF	XML	Total
Jan 2024	98	30	1	129
Feb 2024	67	14	5	86
Mar 2024	30	8	0	38
Apr 2024	37	22	8	67

Viewed (geographical distribution)

Total article views: 324 (including HTML, PDF, and XML) Thereof 324 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 27 Apr 2024

Short summary

This study investigate the long-term changes in the polar stratospheric cloud (PSC) season from 1980 to 2021 above Antarctica. We analyzed spaceborne lidar observations from 2006 to 2020 to build a statistical temperature-based model. We applied our model to gridded reanalysis temperatures, leading to an integrated view of PSC occurrence that is free from sampling issues, allowing us to document the past evolution of the PSC season. We find that PSC season gets longer over the 1980–2000 period.


Total:	0
HTML:	0
PDF:	0
XML:	0