Quantifying SAGE II (1984&ndash;2005) and SAGE III/ISS (2017&ndash;2021) observations of smoke in the stratosphere

Thomason, Larry W.; Knepp, Travis N.

doi:https://doi.org/10.5194/egusphere-2023-995

Preprints

https://doi.org/10.5194/egusphere-2023-995

Preprints

16 May 2023

| 16 May 2023

Quantifying SAGE II (1984–2005) and SAGE III/ISS (2017–2021) observations of smoke in the stratosphere

Larry W. Thomason and Travis N. Knepp

Abstract. Using a common analysis approach for data sets produced by the Stratospheric Aerosol and Gas Experiment instruments SAGE II and SAGE III/ISS, we identify 13 likely smoke events based on enhancements in aerosol extinction coefficient. Nine of these are sufficiently large compared to ambient aerosol levels to compute mean mid-latitude 1020-nm optical depth enhancements that range from 0.0005 to 0.011. We also note that, for large events, the 525 to 1020 nm aerosol extinction coefficient ratio asymptotes at high extinction coefficient to values between 2 and 3 suggesting that the aerosol is relatively small (<0.3 μm) and relatively consistent from event to event. Most of these events are primarily confined to the lower stratosphere and rarely can be observed above 20 km. We also infer an increase in the frequency of smoke events between the SAGE II (1984–1991, 1996–2005) and SAGE III/ISS (2017–present) periods by almost a factor of two and also note that the two largest events occur in the latter data set. However, given the low frequencies overall, we are not confident that the differences can be attributed to changes between the two periods. We also attempt to disentangle the mixing of aerosol in the northern hemisphere summer of 1991 from a pyrocumulus event (Baie-Comeau, Quebec) and Mt. Pinatubo and conclude that, while there is evidence for smoke in the lower stratosphere, virtually all of the enhanced aerosol observations in the northern mid-latitudes in the summer of 1991 are associated with the Mt. Pinatubo eruption.

Received: 12 May 2023 – Discussion started: 16 May 2023

Download & links

Preprint (PDF, 5031 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (5031 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

19 Sep 2023

Quantifying SAGE II (1984–2005) and SAGE III/ISS (2017–2022) observations of smoke in the stratosphere

Larry W. Thomason and Travis Knepp

Atmos. Chem. Phys., 23, 10361–10381, https://doi.org/10.5194/acp-23-10361-2023,https://doi.org/10.5194/acp-23-10361-2023, 2023

Short summary

Larry W. Thomason and Travis N. Knepp

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2023-995', Anonymous Referee #1, 30 May 2023

For technical reasons the paper cannot be accepted in its present form.
The figures must be improved.
We teach our students to prepare proper figures and to consider all basic rules when designing figures and plots! Therefore I cannot accept figures in which so many standard requirements are ignored and violated.
Before I come to the figures, one question: What do you mean when you state: the aerosol is relatively small (<3µm). You mean the radius (?) of the aerosol particles or do you mean the diameter (?) of the aerosol particles? The aerosol consists of molecules and particles and cannot be small.
Figure 1: There is no x-axis text (Extinction coefficient (1/km)) and no y-axis text (Year)! This is unacceptable! The unit of the extinction coefficient (km-1) is not mentioned! Must be improved! Instead of 30-60N, please write: 30°-60°N, everywhere in the text.
Figure 2: x-axis: please add unit of the Extinction Coefficient …. (1/km). Three times!
Figure 3: The numbers of isolines in the color plots are not readable. To my opinion, the color bars are not acceptable. Too complicated. I would prefer two color scales, one for the extinction coefficient and one for the ratio.…..But this is just a suggestion!
y-axis: Latitude ….. better: Latitude (°N). Please improve this in Figs. 4, 5, 6, 9,1 0, 11, 12, 13, 14,16.
Figure 4: x-axis: ‘1020-nm Aerosol’ should be changed to ‘1020 nm Extinction Coefficient (1/km)’
Figure 5: Days: 121 – 150 is wrong, you state in the caption: 151 – 180.
x-axis: ‘1020-nm Aerosol’ should be changed to ‘1020 nm Extinction Coefficient (1/km)’
Figure 6: x-axis: ‘1020-nm Aerosol’ should be changed to ‘1020 nm Extinction Coefficient (1/km)’
Figure 7: x-axis ‘1020 nm Extinction Coefficient (1/km)’ … unit is missing!
The horizontal line is sometimes at 3, sometimes at 4! Is there a reason for this?
There is room in all the plots to include the location (which is given in the caption).
I have the feeling the data in (i), 2020: 181-210, and in (j) are the same (2021: 181-210)
Figure 9: Almost ‘perfect’ x-axis and y-axis text! Only Latitude… could be improved to …. Latitude (°N).
Similar comment (as for Fig.9) regarding the rest of the figures.

Citation: https://doi.org/10.5194/egusphere-2023-995-RC1
- AC1: 'Reply on RC1', Larry Thomason, 12 Jul 2023
  
  Perhaps not my finest moment in figure production for which I apologize. I have made substantial improvements to their readability and corrected missing or incomplete labeling and fixed the duplicate scatter plot in Figure 7i. I also cleaned up some of the associated captioning.
  We have clarified in the text that when we discuss particle size particularly using a value that we are referring to particle radius (in several locations). We have also clarified that when we refer to the smoke aerosol as ‘relatively small’ we are doing so in comparison to the inferred size for other aerosol types (mostly ice clouds and large volcanic aerosol) that exhibit an asymptotic behavior in extinction ratio with increasing extinction coefficient in SAGE data (around lines 405 and 640).
  The horizontal lines use in Figure 7 and a number of other figures are determined by observations of extinction coefficient and extinction ratio just prior to the event highlighted in individual frames. As a result, as aerosol properties vary with time, the line separating LRE and HRE zones also varies. The method for computing these lines is discussed in section 2 primarily in the third paragraph.
  
  Citation: https://doi.org/10.5194/egusphere-2023-995-AC1
RC2:
'Comment on egusphere-2023-995', Anonymous Referee #2, 09 Jun 2023

This comprehensive and pertinent study uses aerosol extinction coefficient derived from the SAGE II and SAGE III/ISS solar occultation measurements to identify and characterize smoke events in the stratosphere. A novel typing approach is presented that allows for discrimination of smoke from clouds and volcanic aerosol. The connections that are drawn between the asymptotic values of extinction ratio and the complicated space of particle composition and size distribution are important points of this paper. The combined Baie-Comeau and Mt. Pinatubo case is interesting and adds a convincing new element of analysis to these events, which have been studied in conjunction in previous works. Cautionary statements about inferring fire frequency from these data sets alone are appropriate. A few minor improvements are suggested below. The figures need attention as noted by the first reviewer.
Type 1 and Type 2 labelling is a bit confusing to follow. More intuitive labelling like Type LER (low extinction ratio) and Type HER (high extinction ratio) might be better. The authors could also consider a third type in the parameter space with very low values of extinction ratio, i.e. below ~2.
What does “not generally noted” refer to on line 176? In the literature?
Given the difficulties discussed on determining the enhancement values due to these events, how is this actually done?
Sub-section 4.1 on non-smoke events along with figures 4,5,6 would be better placed at the end of section 4 to allow more immediate focus on the smoke events of interest.
The extinction threshold line for type identification uses “the 99.5-percentile of 1020-nm extinction coefficient observations where it exceeds 10-4 km-1”. However, based on Figure 7, this doesn’t seem to be doing quite the right job for some of the events. Possibly the absolute 10-4 km-1 part of the criteria is over simplified?
The discussion on the “porting” of the algorithm to the GloSSAC cloud clearing in the conclusions is not clear. What is different about the SAGE II and SAGE III periods, and how is this not currently an issue for SAGE III?

Citation: https://doi.org/10.5194/egusphere-2023-995-RC2
- AC2: 'Reply on RC2', Larry Thomason, 12 Jul 2023
  
  The figures and associated captions have been polished up significantly.
  Type 1 and Type 2 labelling is a bit confusing to follow. More intuitive labelling like Type LER (low extinction ratio) and Type HER (high extinction ratio) might be better. The authors could also consider a third type in the parameter space with very low values of extinction ratio, i.e. below ~2.
  We changed from Types to labels of LRE (low ratio enhancements) and HRE (high ratio enhancements).
  What does “not generally noted” refer to on line 176? In the literature?
  These minor events are unnoted completely or perhaps in a single publication. Since we are focused on relatively small events it seemed worthwhile to note their presence (and the outlier search found these for us). We have noted that they are not often noted in the literature as appearing in the SAGE data sets.
  Given the difficulties discussed on determining the enhancement values due to these events, how is this actually done?
  We integrate the column optical depth for the 10-day periods and essentially take the largest post-event value and subtract the lowest value immediately before the event for something like an impulse value. For larger events, this is pretty easy and fairly robust. As the event size gets small, how well we think it can be captured becomes less certain for the reasons described in the text. When our confidence in producing a meaningful value is too low, we choose not to report one. We’ve added some text around line 260 to indicate this.
  Sub-section 4.1 on non-smoke events along with figures 4,5,6 would be better placed at the end of section 4 to allow more immediate focus on the smoke events of interes.
  We have gone back and forth on this idea. In the end, we thought that these events provide some context of how minor non-smoke outlier events appear in the SAGE data sets particularly for phenomena that are well known from larger events and thus highlight how different these smoke events are from those events. To a lesser extent, moving them to the end would be anticlimactic in which case perhaps better to remove them altogether. Since for reasons mentioned above, we want to mention these events, we would prefer to leave the order as is.
  The extinction threshold line for type identification uses “the 99.5-percentile of 1020-nm extinction coefficient observations where it exceeds 10-4 km-1”. However, based on Figure 7, this doesn’t seem to be doing quite the right job for some of the events. Possibly the absolute 10-4 km-1 part of the criteria is over simplified?
  The main problem isn’t with the 10^-4 km^-1 value but that there is sometimes a notable trend in the extinction coefficient levels within the time frame of the analysis. Thus, when several months pass, the 99.5 cutoff becomes too large relative to the unenhanced aerosol at the time of the smoke (and other) events. This could nominally be fixed by pushing the calculation of closer to the event. However, this has a negative impact on the counting of LRE/HRE events such that very high fractions of such events are inferred when nothing is really going on other than on-going recovery from an earlier event (mostly volcanic). It was a bit of a trade off which we decided that the more conservative approach yielded the clearest outcomes. This was mentioned briefly near line 215. We have added an additional sentence about false positives in that section.
  The discussion on the “porting” of the algorithm to the GloSSAC cloud clearing in the conclusions is not clear. What is different about the SAGE II and SAGE III periods, and how is this not currently an issue for SAGE III?
  The SAGE II algorithm is fairly old (approach dates to version 6.2) and does not consider that smoke could occur in the stratosphere. As a result, as is shown in Figure 1, the algorithm confuses smoke with cloud presence. The new approach developed for SAGE III is painfully aware that smoke can be present in the stratosphere as well as volcanic material and ice clouds (at least at the tropopause). While the approach is not totally disjoint with the old technique, it does a better job parsing between the various types of aerosol that can be present in the stratosphere. We don’t see a major impediment to using the new approach with SAGE II data, so we are hopeful for improvements in that part of the GloSSAC product in future releases. We have modified that discussion to hopefully make this clear.
  
  Citation: https://doi.org/10.5194/egusphere-2023-995-AC2

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2023-995', Anonymous Referee #1, 30 May 2023

For technical reasons the paper cannot be accepted in its present form.
The figures must be improved.
We teach our students to prepare proper figures and to consider all basic rules when designing figures and plots! Therefore I cannot accept figures in which so many standard requirements are ignored and violated.
Before I come to the figures, one question: What do you mean when you state: the aerosol is relatively small (<3µm). You mean the radius (?) of the aerosol particles or do you mean the diameter (?) of the aerosol particles? The aerosol consists of molecules and particles and cannot be small.
Figure 1: There is no x-axis text (Extinction coefficient (1/km)) and no y-axis text (Year)! This is unacceptable! The unit of the extinction coefficient (km-1) is not mentioned! Must be improved! Instead of 30-60N, please write: 30°-60°N, everywhere in the text.
Figure 2: x-axis: please add unit of the Extinction Coefficient …. (1/km). Three times!
Figure 3: The numbers of isolines in the color plots are not readable. To my opinion, the color bars are not acceptable. Too complicated. I would prefer two color scales, one for the extinction coefficient and one for the ratio.…..But this is just a suggestion!
y-axis: Latitude ….. better: Latitude (°N). Please improve this in Figs. 4, 5, 6, 9,1 0, 11, 12, 13, 14,16.
Figure 4: x-axis: ‘1020-nm Aerosol’ should be changed to ‘1020 nm Extinction Coefficient (1/km)’
Figure 5: Days: 121 – 150 is wrong, you state in the caption: 151 – 180.
x-axis: ‘1020-nm Aerosol’ should be changed to ‘1020 nm Extinction Coefficient (1/km)’
Figure 6: x-axis: ‘1020-nm Aerosol’ should be changed to ‘1020 nm Extinction Coefficient (1/km)’
Figure 7: x-axis ‘1020 nm Extinction Coefficient (1/km)’ … unit is missing!
The horizontal line is sometimes at 3, sometimes at 4! Is there a reason for this?
There is room in all the plots to include the location (which is given in the caption).
I have the feeling the data in (i), 2020: 181-210, and in (j) are the same (2021: 181-210)
Figure 9: Almost ‘perfect’ x-axis and y-axis text! Only Latitude… could be improved to …. Latitude (°N).
Similar comment (as for Fig.9) regarding the rest of the figures.

Citation: https://doi.org/10.5194/egusphere-2023-995-RC1
- AC1: 'Reply on RC1', Larry Thomason, 12 Jul 2023
  
  Perhaps not my finest moment in figure production for which I apologize. I have made substantial improvements to their readability and corrected missing or incomplete labeling and fixed the duplicate scatter plot in Figure 7i. I also cleaned up some of the associated captioning.
  We have clarified in the text that when we discuss particle size particularly using a value that we are referring to particle radius (in several locations). We have also clarified that when we refer to the smoke aerosol as ‘relatively small’ we are doing so in comparison to the inferred size for other aerosol types (mostly ice clouds and large volcanic aerosol) that exhibit an asymptotic behavior in extinction ratio with increasing extinction coefficient in SAGE data (around lines 405 and 640).
  The horizontal lines use in Figure 7 and a number of other figures are determined by observations of extinction coefficient and extinction ratio just prior to the event highlighted in individual frames. As a result, as aerosol properties vary with time, the line separating LRE and HRE zones also varies. The method for computing these lines is discussed in section 2 primarily in the third paragraph.
  
  Citation: https://doi.org/10.5194/egusphere-2023-995-AC1
RC2:
'Comment on egusphere-2023-995', Anonymous Referee #2, 09 Jun 2023

This comprehensive and pertinent study uses aerosol extinction coefficient derived from the SAGE II and SAGE III/ISS solar occultation measurements to identify and characterize smoke events in the stratosphere. A novel typing approach is presented that allows for discrimination of smoke from clouds and volcanic aerosol. The connections that are drawn between the asymptotic values of extinction ratio and the complicated space of particle composition and size distribution are important points of this paper. The combined Baie-Comeau and Mt. Pinatubo case is interesting and adds a convincing new element of analysis to these events, which have been studied in conjunction in previous works. Cautionary statements about inferring fire frequency from these data sets alone are appropriate. A few minor improvements are suggested below. The figures need attention as noted by the first reviewer.
Type 1 and Type 2 labelling is a bit confusing to follow. More intuitive labelling like Type LER (low extinction ratio) and Type HER (high extinction ratio) might be better. The authors could also consider a third type in the parameter space with very low values of extinction ratio, i.e. below ~2.
What does “not generally noted” refer to on line 176? In the literature?
Given the difficulties discussed on determining the enhancement values due to these events, how is this actually done?
Sub-section 4.1 on non-smoke events along with figures 4,5,6 would be better placed at the end of section 4 to allow more immediate focus on the smoke events of interest.
The extinction threshold line for type identification uses “the 99.5-percentile of 1020-nm extinction coefficient observations where it exceeds 10-4 km-1”. However, based on Figure 7, this doesn’t seem to be doing quite the right job for some of the events. Possibly the absolute 10-4 km-1 part of the criteria is over simplified?
The discussion on the “porting” of the algorithm to the GloSSAC cloud clearing in the conclusions is not clear. What is different about the SAGE II and SAGE III periods, and how is this not currently an issue for SAGE III?

Citation: https://doi.org/10.5194/egusphere-2023-995-RC2
- AC2: 'Reply on RC2', Larry Thomason, 12 Jul 2023
  
  The figures and associated captions have been polished up significantly.
  Type 1 and Type 2 labelling is a bit confusing to follow. More intuitive labelling like Type LER (low extinction ratio) and Type HER (high extinction ratio) might be better. The authors could also consider a third type in the parameter space with very low values of extinction ratio, i.e. below ~2.
  We changed from Types to labels of LRE (low ratio enhancements) and HRE (high ratio enhancements).
  What does “not generally noted” refer to on line 176? In the literature?
  These minor events are unnoted completely or perhaps in a single publication. Since we are focused on relatively small events it seemed worthwhile to note their presence (and the outlier search found these for us). We have noted that they are not often noted in the literature as appearing in the SAGE data sets.
  Given the difficulties discussed on determining the enhancement values due to these events, how is this actually done?
  We integrate the column optical depth for the 10-day periods and essentially take the largest post-event value and subtract the lowest value immediately before the event for something like an impulse value. For larger events, this is pretty easy and fairly robust. As the event size gets small, how well we think it can be captured becomes less certain for the reasons described in the text. When our confidence in producing a meaningful value is too low, we choose not to report one. We’ve added some text around line 260 to indicate this.
  Sub-section 4.1 on non-smoke events along with figures 4,5,6 would be better placed at the end of section 4 to allow more immediate focus on the smoke events of interes.
  We have gone back and forth on this idea. In the end, we thought that these events provide some context of how minor non-smoke outlier events appear in the SAGE data sets particularly for phenomena that are well known from larger events and thus highlight how different these smoke events are from those events. To a lesser extent, moving them to the end would be anticlimactic in which case perhaps better to remove them altogether. Since for reasons mentioned above, we want to mention these events, we would prefer to leave the order as is.
  The extinction threshold line for type identification uses “the 99.5-percentile of 1020-nm extinction coefficient observations where it exceeds 10-4 km-1”. However, based on Figure 7, this doesn’t seem to be doing quite the right job for some of the events. Possibly the absolute 10-4 km-1 part of the criteria is over simplified?
  The main problem isn’t with the 10^-4 km^-1 value but that there is sometimes a notable trend in the extinction coefficient levels within the time frame of the analysis. Thus, when several months pass, the 99.5 cutoff becomes too large relative to the unenhanced aerosol at the time of the smoke (and other) events. This could nominally be fixed by pushing the calculation of closer to the event. However, this has a negative impact on the counting of LRE/HRE events such that very high fractions of such events are inferred when nothing is really going on other than on-going recovery from an earlier event (mostly volcanic). It was a bit of a trade off which we decided that the more conservative approach yielded the clearest outcomes. This was mentioned briefly near line 215. We have added an additional sentence about false positives in that section.
  The discussion on the “porting” of the algorithm to the GloSSAC cloud clearing in the conclusions is not clear. What is different about the SAGE II and SAGE III periods, and how is this not currently an issue for SAGE III?
  The SAGE II algorithm is fairly old (approach dates to version 6.2) and does not consider that smoke could occur in the stratosphere. As a result, as is shown in Figure 1, the algorithm confuses smoke with cloud presence. The new approach developed for SAGE III is painfully aware that smoke can be present in the stratosphere as well as volcanic material and ice clouds (at least at the tropopause). While the approach is not totally disjoint with the old technique, it does a better job parsing between the various types of aerosol that can be present in the stratosphere. We don’t see a major impediment to using the new approach with SAGE II data, so we are hopeful for improvements in that part of the GloSSAC product in future releases. We have modified that discussion to hopefully make this clear.
  
  Citation: https://doi.org/10.5194/egusphere-2023-995-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

AR by Larry Thomason on behalf of the Authors (18 Jul 2023) Author's response Author's tracked changes Manuscript

ED: Publish as is (24 Jul 2023) by Matthias Tesche

AR by Larry Thomason on behalf of the Authors (24 Jul 2023)

Journal article(s) based on this preprint

19 Sep 2023

Quantifying SAGE II (1984–2005) and SAGE III/ISS (2017–2022) observations of smoke in the stratosphere

Larry W. Thomason and Travis Knepp

Atmos. Chem. Phys., 23, 10361–10381, https://doi.org/10.5194/acp-23-10361-2023,https://doi.org/10.5194/acp-23-10361-2023, 2023

Short summary

Larry W. Thomason and Travis N. Knepp

Viewed

Total article views: 312 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
225	72	15	312	7	5

HTML: 225
PDF: 72
XML: 15
Total: 312
BibTeX: 7
EndNote: 5

Views and downloads (calculated since 16 May 2023)

Month	HTML	PDF	XML	Total
May 2023	86	28	4	118
Jun 2023	64	15	5	84
Jul 2023	37	14	5	56
Aug 2023	21	5	0	26
Sep 2023	17	10	1	28

Cumulative views and downloads (calculated since 16 May 2023)

Month	HTML	PDF	XML	Total
May 2023	86	28	4	118
Jun 2023	64	15	5	84
Jul 2023	37	14	5	56
Aug 2023	21	5	0	26
Sep 2023	17	10	1	28

Viewed (geographical distribution)

Total article views: 353 (including HTML, PDF, and XML) Thereof 353 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 19 Sep 2023

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (5031 KB)
Metadata XML

Short summary

We examine space-based observations of stratospheric aerosol to infer the presence of episodic smoke perturbations. We find that smoke's optical properties often shows a consistent behavior though vary somewhat from event to event. We also find that the rate of smoke events observed in the 194–2005 period is about half the rate of similar observations in the period from 2017 to the present though with such low over all rates inferring change between the periods is difficult.


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