the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Isolating the boreal winter response to the Pinatubo and Krakatoa eruptions using large-ensemble single-forcing simulations
Abstract. The Northern Hemisphere wintertime circulation response to the eruptions of Krakatoa and Pinatubo is revisited in large ensembles from eight modeling centers with only one time-varying external forcing: volcanic eruptions. All eight models show a warming of the tropical lower stratosphere. In six of the models, the meridional temperature gradient in the winter stratosphere is enhanced, leading to a strengthened stratospheric polar vortex, a positive phase of the North Atlantic Oscillation, a poleward shift in storm tracks, and warm surface temperatures during the winter over subpolar Eurasia. While this warming over subpolar Eurasia is statistically significant in the multi-model mean and in four of the individual models, at least 34 eruptions are needed before it can be robustly distinguished from the global mean cooling with 5 % confidence. An El Niño response is evident shortly after eruption in these models, which transitions from a Central Pacific morphology in the first winter to an East Pacific morphology in the second, and subsequently to an La Niña response in the third and fourth winters. While these ENSO responses require 48 or more eruptions to emerge from the noise, they nonetheless lead to surface impacts over North America. However, these surface impacts do not resemble those classically associated with ENSO in the first year after eruption, likely because the tropical precipitation response differs as well due to the large-scale reduction in tropical precipitation. There is substantial diversity in the magnitude of these responses across models, likely owing to differences in tropical stratospheric diabatic heating despite the models using the same forcing.
- Preprint
(4082 KB) - Metadata XML
-
Supplement
(5231 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-3738', Anonymous Referee #1, 11 Aug 2026
-
AC1: 'Reply on RC1', Chaim Garfinkel, 05 Nov 2026
We thank the reviewers for their helpful comments which have led to a clearer manuscript.
reviewer 1
- For the two anomalously behaving models, if the differences are caused by discrepancies in Q₁ (diabatic heating), what specifically leads to the differences in Q₁ between these two models?
We wish we could explain this difference, but the vertically resolved diabatic tendencies are not available. Also, note that these two models with weak responses have only 10 members, and this relatively smaller ensemble size could also be contributing. We have added to the Discussion:
“It is also possible that the relatively small ensemble sizes - 10 members only - for the two models with the weakest response leads to larger uncertainty on the forced volcanic signal. Either way, the difference across models despite large ensembles highlights that model differences are at play, such as how the optical properties are translated into radiative forcing, including how the prescribed extinction is translated onto each model’s radiation bands, and how the tropopause was accounted for (Villamayor et al., 2023; Zanchettin et al., 2022)”
- In L110–114, which figures are being referred to, and why is the threshold set at "6 of 8 models"?
We removed the information on stippling for figures showing the response of each model from the methods, and now include it in the figure captions only, for clarity.
A threshold of 75% of agreement across models is similar to that used in the IPCC AR6 report (which adopts 80% agreement as the standard, but here we have only 8 models and so 75% is more natural). See Cross-Chapter Box Atlas.1 ("Displaying Robustness and Uncertainty in Maps") of https://www.ipcc.ch/report/ar6/wg1/chapter/atlas/
The IPCC Atlas includes a detailed discussion of why such a threshold is useful, and we now cite it.
- In L119–121, has the difference in surface response between these two model versions been evaluated?
We have additionally analyzed the tropical temperature response in the lower stratosphere [10S-10N, 50hPa], and it too doesn’t differ between the two GISS variants (4.7K vs. 4.55K).
- In Figs. 2e-l, why does the positive temperature anomaly over high-latitude Eurasia persist throughout the three winters?
The high latitude Eurasia temperature signal weakens substantially after the first winter, is only marginally statistically significant in the second winter, and is no longer statistically significant in the third or fourth winter. This is consistent with the vortex response, which is still present in the second winter but not aftewards.
- In L159–160, what does this statement mean? Please clarify.
Rewritten as: “Nonetheless, there is a small gap in the cooling in the west-central equatorial Pacific that suggests the possibility of an El Nino response.”
- There are differences between Fig. 4 and Fig. S5, as well as between Fig. 5 and Fig. S6. Do the differences between different volcanic eruption events affect the robustness of the results?
The CMIP6 forcing for Krakatoa is slightly stronger than for Pinatubo, and accordingly the atmospheric response is also stronger. This is discussed in the Discussion section: “For the figures in the main text, we aggregate the responses to Pinatubo and Krakatoa in order to increase the sample size. However, the CMIP6 aerosol forcing for Krakatoa was slightly larger than that for Pinatubo (Aubry et al., 2026); consistent with this, the resulting response is stronger for most of the key metrics discussed in the paper, including the tropical lower stratospheric and polar vortex responses (Supplemental Figure S1, S2, S7, S8), the subpolar Eurasian warming response (Supplemental Figure S3), and the ENSO response (Supplemental Figure S4). The net effect is that the SLP patterns forced by the two eruptions exhibit differences in much of the midlatitudes (Supplemental Figure S5), likely due to differences in the polar vortex response and in the tropical precipitation response (Supplemental Figure S6)”
- In Fig. 1 and Fig. 6, the sizes of the left and right panels are inconsistent. This issue also appears in other figures.
This should be fixed now
- In L291–295, does this imply that some of the phenomena are merely stochastic signals? If so, are there other phenomena that might also represent spurious, non-physical associations rather than real causal relationships?
Unfortunately we don’t have the necessary output from LESFMIP to diagnose why the storm track response changes in the way it does. (The papers cited do such an additional analysis, but we simply can’t.) This is true to a lesser extent also for the vortex and ENSO pathways. This caveat is mentioned in the Discussion section.
Citation: https://doi.org/10.5194/egusphere-2026-3738-AC1
-
AC1: 'Reply on RC1', Chaim Garfinkel, 05 Nov 2026
-
RC2: 'Comment on egusphere-2026-3738', Anonymous Referee #2, 11 Sep 2026
Review of 'Isolating the boreal winter response to the Pinatubo and Krakatoa eruptions using large-ensemble single-forcing simulations' by Garfinkel, Avisar, Huo, Kuchar, Minobe, Osprey, Perny and Wright
Summary
The paper reports on 200 simulations from 8 climate models as part of LESFMIP. Specifically, these single forcing simulations look at the climate response only to historical volcanic eruptions. The authors examine 5-year NH winter climate responses to the 1883 Krakatau eruption and the 1991 Pinatubo eruption, both apparently of similar magnitude in terms of sulphur dioxide (SO2). They suggest that both runs highlight a forced response in the NAO to both eruptions, as well as in NINO4, However, the signals only emerge after 30-50 simulations are performed. Individual model differences are significant to my eye
Major comments
'1. 'sulfur burden…is similar for both eruptions…'. How true is this? Based on Arfuielle, it looks to me that P has a sulfur load of about 15 Tg and K of about 11 Tg… Similar-ish I suppose. The forcings in Aubry look a lot stronger for K.
In any case, you probably didn't use Aubry, as it was designed for CMIP7 and was presumably produced after the simulations used here were made (which are early 2020s?). So is the citation relevant? It doesn’t matter which you used, but you need to be clear in the paper which it was and interpret the results accordingly.
From the abstract of Aubry: '…Our methodology to produce historical aerosol optical properties significantly differs from CMIP6 for the pre-satellite era, and the resulting forcings in turn largely differ…'
From Arafuielle: 'For instance, the Krakatau eruption of 1883, previously thought to have released more SO2 than Pinatubo, is now downgraded to 2/3 of Pinatubo release by Gao et al. (2008).'
The Aubry paper has a peak optical depth of > 0.7 for Krakatoa and an optical depth of 0.2 for Pinatubo. They are not that different between the eruptions in the Arafuielle paper (and P is probably higher). This suggests a source of variability different from SO2/aerosol forcing to produce the modelled differences. Clarifying which forcing was used might help to understand the results.
In general, it may be useful to summarize important model considerations (like the SO2 forcing) in a table outside the text to highlight and summarize key components of the paper for easy reference.
2. Figure 9. I find this figure and discussion is not explained very well at all. What is begin measured here? Why are there only 200 members instead of 400? Are you measuring the long-term global response? Why the comparison between the long-term surface data and the models? Please clarify the meaning and interpretation of the figure.
3. Figure 10. a.) improve the figure by noting what each panel is referring to, e.g. a.) Eurasia, b.) Eurasia, global cooling removed, etc in the title of each panel
Figure 10. b) Doesn’t the relatively large number of simulations required to identify a particular response suggest that other factors beyond SO2 are playing a role. The eruption itself may somewhat force an NAO response 4-6 months later, but lots of other things happen due to internal variability to prevent or enhance these chances. It seems more complicated than simply saying tropical volcanic eruptions that reach the stratosphere with relatively high SO2 push a positive NAO response. Maybe it 'skews' the internal variability towards that state, but doesn’t guarantee it? More like 'biased internal variability' rather than a 'forced response'.
Minor comments
Fig 1. Legend text is a bit small at normal size for my eyes. Maybe just enlarge the figure a bit.
Line 101., What would happen if you just chose the same number of ensemble members for each model, and so didn't bias the result to particular models? About half the runs (40%) come from two models. Not saying this needs to be done, but I'm curious about what effect it would have on the results.
Line 109, 'Krakatoa has stronger impact…', that may be true (and figures suggest that), but why? It seems the response is not linearly related to the SO2 forcing. But it depends on which forcing was used for K.
Line 110, 2.5° horizontal grid. Can you summarize what the vertical resolution and domain in the stratosphere is in the different models? Presumably they are high-top models with 10-15 levels in strat, but can you specify in the paper? Perhaps a table…
Line 131, '…consistent with observations', Perhaps pedantic, but there were no stratospheric observations after Krakatoa, so you are tacitly assuming responses are the same. The findings here are not necessarily incorrect, but it is an assumption based on modelling and not observations.
Also Fig 1. Is the oscillating equatorial wind near 35 km a manifestation of the QBO? It seems a bit high altitude for QBO (more 25-30 km)… Is the QBO that present and consistent across the ensemble? Why no comments about it, I know it is not the focus of the paper, but it could be worthy of a general note…There appears to be some significance there, clearly in years 1,2 and 4.
Line 149. …'response is stronger', be specific about which response you are discussing (which is presumably the Scandinavian response...For example, the North American response is much cooler (i.e. stronger) during LN years and appears to remain so globally in the years following.
Line 150. Can you give the number of members in the EL Nino/La Nina groups? Are sample sizes even or strongly different?
Line 156. El Nino response is the consensus, but it is not uniform, even excluding paleo runs. Location of eruption (even in tropics), time of year, etc all appear to matter. See e.g. Pausata et al 2020 or Macgregor and Timmerman 2011.
Line 184. Is 'wettening' a real word?
Figure 6. HADGEM3, MPI and CMCC all produce very strong cooling responses over the pole, while the equatorial response between the models is less variable (but certainly not zero!)
Line 232. I suppose this is suggesting that the MIROC response is not just a model formulation issue, as it gives you the response you are expecting in some cases. Further, in some places it is written as MIROC6 and in others just MIROC. Choose one and be consistent.
Line 256. I find the section beginning here a bit cryptic. Are you really raising the number of samples from 1000 to 5000? Is it supposed to be 500? Is the selection random? Isn’t 400 the total of all members, so it isn’t really random in the last case? Later it says how many *months* must be selected? Please rewrite for clarity and to explain better.
I suppose you are taking 1000 random samples of 2 and seeing how often you get the response, 500 of 4, 333 of 6, etc and identifying when the spread of these samples no longer crosses zero.
Line 282, 'Primarily because ACCESS…', That's undoubtedly part of the reason, but the different models also simulate different amounts of activity in slightly different latitudes which creates a broader peak than most individual models shown. Further, the CMCC model response is just as much of an outlier (5x the mean!) as the ACCESS model, just in a different direction and possibly dominates the response even more so.
Line 290. I appreciate that it is probably 'beyond the scope of the paper', but why not discuss the SH response as well? From Fig 11, there is a clearly just as strong of SAM response around Antarctica (in years 1 and 2) as that in the NH. It's not the NH, but it is still a 'boreal winter' (aka DJF) response. You already discuss ENSO, why not more?
Line 291. '…no statistical significant response in the zonal mean'. I'm confused here, don’t the thick lines in Figure 11 clearly show a significant response in year 2 in both NH and SH?
Citation: https://doi.org/10.5194/egusphere-2026-3738-RC2 -
AC2: 'Reply on RC2', Chaim Garfinkel, 05 Nov 2026
We thank the reviewers for their helpful comments which have led to a clearer manuscript.
reviewer 2
Major comments
'1. 'sulfur burden…is similar for both eruptions…'. How true is this? Based on Arfuielle, it looks to me that P has a sulfur load of about 15 Tg and K of about 11 Tg… Similar-ish I suppose. The forcings in Aubry look a lot stronger for K.
In any case, you probably didn't use Aubry, as it was designed for CMIP7 and was presumably produced after the simulations used here were made (which are early 2020s?). So is the citation relevant? It doesn’t matter which you used, but you need to be clear in the paper which it was and interpret the results accordingly.
From the abstract of Aubry: '…Our methodology to produce historical aerosol optical properties significantly differs from CMIP6 for the pre-satellite era, and the resulting forcings in turn largely differ…'
From Arafuielle: 'For instance, the Krakatau eruption of 1883, previously thought to have released more SO2 than Pinatubo, is now downgraded to 2/3 of Pinatubo release by Gao et al. (2008).'
The Aubry paper has a peak optical depth of > 0.7 for Krakatoa and an optical depth of 0.2 for Pinatubo. They are not that different between the eruptions in the Arafuielle paper (and P is probably higher). This suggests a source of variability different from SO2/aerosol forcing to produce the modelled differences. Clarifying which forcing was used might help to understand the results.
In general, it may be useful to summarize important model considerations (like the SO2 forcing) in a table outside the text to highlight and summarize key components of the paper for easy reference.
LESFMIP uses the CMIP6 era volcano forcing. Our understanding of how it was constructed has been added to the paper, and is copied below: “LESFMIP uses the CMIP6-era volcano forcing dataset. The CMIP6 volcanic forcing dataset (1850–2014) transitions from direct atmospheric observations to physical modeling and proxy records as it goes back in time. For the modern satellite era (1979–2014), the forcing is anchored by the Global Space-based Stratospheric Aerosol Climatology (GloSSAC; Thomason et al., 2018), which integrates direct satellite measurements from platforms like SAGE, HALOE, and OSIRIS. To extend the record into the pre-satellite era (1850–1978), the dataset relies on two main approaches: seven major, highly disruptive volcanic events (such as Krakatoa in 1883 and Agung in 1963) were simulated by feeding polar ice-core sulfur dioxide estimates (Gao et al., 2008) into the AER 2D sulfur microphysics model developed by Arfeuille et al. (2014), while smaller intermediate eruptions were estimated using ground-based pyrheliometer and photometer measurements of surface solar radiation losses (Stothers, 1996). Non-eruptive background periods were filled using a baseline derived from the volcanically quiescent 1999–2005 window, linearly adjusted to account for historical anthropogenic aerosol trends. We explicitly note that the pre-satellite era data has higher uncertainties, and that the CMIP7 forcing dataset (which we repeat was not used here) differs qualitatively for some periods, however, the estimated forcing for Pinatubo and Krakatoa is relatively unchanged between CMIP6 and CMIP7 (figure 10 of Aubry et al 2026).“
We have also edited the following paragraph: “The specification of the sulfur burden and effective radiation forcing in the CMIP6 aerosol input dataset is generally similar for both eruptions (see figure 10 of Aubry et al 2026) though somewhat stronger for Krakatoa”
- Figure 9. I find this figure and discussion is not explained very well at all. What is being measured here? Why are there only 200 members instead of 400? Are you measuring the long-term global response? Why the comparison between the long-term surface data and the models? Please clarify the meaning and interpretation of the figure.
For each of the 200 members, we average the responses of Pinatubo and Krakatoa, and show that mean response; this allows for comparison to observations. We are measuring the Eurasia response relative to the global mean. The caption has been clarified.
- Figure 10. a.) improve the figure by noting what each panel is referring to, e.g. a.) Eurasia, b.) Eurasia, global cooling removed, etc in the title of each panel
This information has been added to the panel titles as suggested.
Figure 10. b) Doesn’t the relatively large number of simulations required to identify a particular response suggest that other factors beyond SO2 are playing a role. The eruption itself may somewhat force an NAO response 4-6 months later, but lots of other things happen due to internal variability to prevent or enhance these chances. It seems more complicated than simply saying tropical volcanic eruptions that reach the stratosphere with relatively high SO2 push a positive NAO response. Maybe it 'skews' the internal variability towards that state, but doesn’t guarantee it? More like 'biased internal variability' rather than a 'forced response'.
We agree that there are lots of other factors that can overwhelm the volcano's SO2 signal, including internal variability. But even with all these factors, the SO2 signal does induce a tendency toward a +NAO, though of course it doesn’t guarantee it. Clearly, the strength of the signal can depend on some of these other factors, as we demonstrate in the case of ENSO in figure 2.
We have added “This is due to the substantial internal variability in the Earth system which can overwhelm the volcanic influence.”
Minor comments
Fig 1. Legend text is a bit small at normal size for my eyes. Maybe just enlarge the figure a bit.
Enlarged as suggested
Line 101., What would happen if you just chose the same number of ensemble members for each model, and so didn't bias the result to particular models? About half the runs (40%) come from two models. Not saying this needs to be done, but I'm curious about what effect it would have on the results.
This would have a slight effect on our results, since the two models with the weakest response (ACCESS and MIROC) only have 10 members each yet are given equal weight to models with much larger ensemble sizes. If we computed the MMM by weighting each member individually we would have a slightly stronger response.
We have now added this as a caveat in the Results and Discussion section, rather than here, as we think it fits better later in the text.
Line 109, 'Krakatoa has stronger impact…', that may be true (and figures suggest that), but why? It seems the response is not linearly related to the SO2 forcing. But it depends on which forcing was used for K.
See above response to the general comment. The CMIP6 forcing is larger for Krakatoa than Pinatubo.
Line 110, 2.5° horizontal grid. Can you summarize what the vertical resolution and domain in the stratosphere is in the different models? Presumably they are high-top models with 10-15 levels in strat, but can you specify in the paper? Perhaps a table…
Information about models used in this study is added in Table 1. Not all are high-top - this is now noted in the table.
Line 131, '…consistent with observations', Perhaps pedantic, but there were no stratospheric observations after Krakatoa, so you are tacitly assuming responses are the same. The findings here are not necessarily incorrect, but it is an assumption based on modelling and not observations.
Good point. We have added “even if we focus exclusively on Pinatubo.”
Also Fig 1. Is the oscillating equatorial wind near 35 km a manifestation of the QBO? It seems a bit high altitude for QBO (more 25-30 km)… Is the QBO that present and consistent across the ensemble? Why no comments about it, I know it is not the focus of the paper, but it could be worthy of a general note…There appears to be some significance there, clearly in years 1,2 and 4.
This feature is evident in some models but not others. See figure 6. HadGEM shows it strongly, and indeed HadGEM has a QBO. MIROC shows it as well though weaker, and it also has a QBO. The other models used here do not have QBOs, and so the appearance of this oscillation is particularly interesting. Garfinkel et al 2025 document the reaction of the QBO to eruptions in MIROC and HadGEM in detail, and discussed this phase alignment in detail. We have added to the Discussion:
“While the equatorial Quasi-Biennial Oscillation (QBO) is outside of our main focus, it is noteworthy that the two models (MIROC6 and HadGEM3) with a spontaneous QBO show a qualitatively similar tropical wind response (Figure 6; cf., Garfinkel et al., 2025) with stronger westerlies in the mid-stratosphere and easterlies in the lowermost and upper stratosphere. Interestingly, MPI-ESM shows a similar response despite not having a spontaneous QBO. Three other models without a spontaneous QBO (CMCC, NorESM, and CanESM5) simulate lowermost stratospheric easterly anomalies and westerly anomalies in most of the stratosphere. Future work should aim to understand these responses. “
Line 149. …'response is stronger', be specific about which response you are discussing (which is presumably the Scandinavian response...For example, the North American response is much cooler (i.e. stronger) during LN years and appears to remain so globally in the years following.
“Eurasian sector” added
Line 150. Can you give the number of members in the EL Nino/La Nina groups? Are sample sizes even or strongly different?
The sample sizes are similar:
107 of the 400 total ensemble members are El Nino.
99 of the 400 total ensemble members are La Nino.
We have added this information to the figure caption.
Line 156. El Niño response is the consensus, but it is not uniform, even excluding paleo runs. Location of eruption (even in tropics), time of year, etc all appear to matter. See e.g. Pausata et al 2020 or Macgregor and Timmerman 2011.
Changed to “a tendency towards El Nino conditions”.
The two references have been added to the introduction.
Line 184. Is 'wettening' a real word?
Changed to “more precipitation”
Figure 6. HADGEM3, MPI and CMCC all produce very strong cooling responses over the pole, while the equatorial response between the models is less variable (but certainly not zero!)
With the diagnostics available we are unable to unravel why this might be.
Line 232. I suppose this is suggesting that the MIROC response is not just a model formulation issue, as it gives you the response you are expecting in some cases. Further, in some places it is written as MIROC6 and in others just MIROC. Choose one and be consistent.
It could be a small sample size issue (just 10 members). We note this possibility in both the Results and Discussion.
MIROC changed to MIROC6 everywhere
Line 256. I find the section beginning here a bit cryptic. Are you really raising the number of samples from 1000 to 5000? Is it supposed to be 500? Is the selection random? Isn’t 400 the total of all members, so it isn’t really random in the last case? Later it says how many *months* must be selected? Please rewrite for clarity and to explain better.
Woops, the 1000 should have been 5000! Sorry for the confusion. We are taking 5000 random bootstrapping samples. (In an early version of this figure we only took 1000 samples, but that isn’t really enough to constrain the uncertainty, and so for the publication-ready figure we increased it to 5000. Sorry again for this mistake.)
We have revised the description of the bootstrapping for clarity. “We repeat this 5000 times, thereby generating an uncertainty estimate for the mean response. Next, we subselect 4 events with resampling and compute the mean, and then repeat this subselecting of 4 events 5000 times. We then repeat the subselecting but with 6 events, and so on, until we randomly subselect 400 events with resampling.”
I suppose you are taking 1000 random samples of 2 and seeing how often you get the response, 500 of 4, 333 of 6, etc and identifying when the spread of these samples no longer crosses zero.
No, we always create 5000 random samples. See above response
Line 282, 'Primarily because ACCESS…', That's undoubtedly part of the reason, but the different models also simulate different amounts of activity in slightly different latitudes which creates a broader peak than most individual models shown. Further, the CMCC model response is just as much of an outlier (5x the mean!) as the ACCESS model, just in a different direction and possibly dominates the response even more so.
We have added “CMCC, on the other hand, simulates a storm track response that is stronger than other models, consistent with its stronger stratospheric vortex response”. If we remove ACCESS and CMCC, we receive a positive response stat. different from zero during the first winter - this is also noted in the paper now too.
Line 290. I appreciate that it is probably 'beyond the scope of the paper', but why not discuss the SH response as well? From Fig 11, there is a clearly just as strong of SAM response around Antarctica (in years 1 and 2) as that in the NH. It's not the NH, but it is still a 'boreal winter' (aka DJF) response. You already discuss ENSO, why not more?
We have added a paragraph about the SH response to the end of the section.
“A SH response is also apparent in DJF (both in zonal mean and regionally). The storm track response weakens between 30◦S and 50◦S during the first and second winter. Similarly to the NH response, we relate this response in storm-track activity to changes in stratospheric winds propagating downward (cf. Figure 1e), considering that Antarctic polar vortex variability and the associated downward coupling are strongest in austral late spring (e.g. Lim et al., 2019). Previous studies based on reconstructions, reanalysis and satellite datasets reported insignificant responses in the Southern Annular Mode (SAM), however, CMIP6 models revealed positive SAM-like responses in agreement with our findings (Paik et al., 2023, and references therein). This discrepancy between model and observational studies has been attributed to ENSO interference with the SAM state (McGraw et al., 2016; Yang and Xiao, 2018).”
Line 291. '…no statistical significant response in the zonal mean'. I'm confused here, don’t the thick lines in Figure 11 clearly show a significant response in year 2 in both NH and SH?
This sentence erroneously appeared in the manuscript; thus, removed now.
Citation: https://doi.org/10.5194/egusphere-2026-3738-AC2
-
AC2: 'Reply on RC2', Chaim Garfinkel, 05 Nov 2026
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 429 | 248 | 51 | 728 | 131 | 52 | 55 |
- HTML: 429
- PDF: 248
- XML: 51
- Total: 728
- Supplement: 131
- BibTeX: 52
- EndNote: 55
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
It is generally believed that the "parasol effect" following volcanic eruptions causes cooling, thereby leading to short-term tropospheric cooling. However, this study, using ensemble data from numerical models, demonstrates that after the eruptions of Krakatoa and Pinatubo, the meridional temperature gradient in the winter stratosphere is enhanced, leading to a strengthened stratospheric polar vortex, a positive phase of the North Atlantic Oscillation, a poleward shift in storm tracks, and warm surface temperatures during winter over subpolar Eurasia. The analytical approach employs large ensemble simulations from eight climate models participating in the LESFMIP Project, with up to 200 ensemble members (equivalent to 400 independent samples) to investigate this issue. The statistical sample size far exceeds that of previous studies. The study quantitatively addresses the core question of "how many eruption events are needed for the signal to emerge from internal variability," and the conclusions are highly convincing. It also points out that the discrepancies in paleoclimate reconstructions are precisely due to insufficient sample sizes. The manuscript also presents the ENSO response to volcanic eruption events. Based on a large volume of data and employing statistical methods, the work complements our previous understanding with rigorous evidence and clear structure. Below are some questions regarding this manuscript.