the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Jul 2026
Extreme Amundsen Sea Low (ASL) events and climate of West Antarctica
Abstract. The Amundsen Sea Low (ASL) is a regional atmospheric circulation feature that significantly influences the climate of West Antarctica through its impact on the meridional wind field. An observed deepening of the ASL over the satellite era has raised concerns about the possibility of more extreme ASL conditions through the end of the twenty-first century. Here we use ERA5 reanalysis and passive microwave satellite observations to examine the impact of extreme ASL events on West Antarctica during the period 1979–2025. In total, we identify 13 extremely strong and 8 extremely weak events that vary geographically based on magnitude. Strong events tend to be concentrated poleward and centered in the vicinity of the Amundsen Sea while weak events are found farther north and display large zonal spatial heterogeneity. Composite maps of surface anomalies suggest that the location of extreme events serve to explain a regional dipole in surface temperatures and sea ice concentration, both of which are co-located with anomalous wind velocities modulated by ASL strength. Given the link between the regional ASL and additional large-scale modes of atmospheric variability in the Southern Hemisphere, we also investigate extreme ASL events in relation to the Southern Annular Mode, El Niño Southern Oscillation, and Zonal Wave 3. We find that nearly all extremely strong events occur when the three modes of variability are positively in-phase; there is less agreement during extremely weak ASL events. We assess future ASL variability and extremes using CESM2 model output under middle- and high-emission scenarios. Future projections indicate a higher proportion of extremely strong ASL events relative to weak events through the end of the 21st century, suggesting an increased likelihood of more extreme meridional conditions across West Antarctica in the near future.
- Preprint
(9897 KB) - Metadata XML
-
Supplement
(1551 KB) - BibTeX
- EndNote
Status: open (until 07 Sep 2026)
- RC1: 'Comment on egusphere-2026-2897', Anonymous Referee #1, 04 Aug 2026 reply
-
RC2: 'Comment on egusphere-2026-2897', Anonymous Referee #2, 18 Aug 2026
reply
This manuscript presents a clear, well-organized analysis of extreme ASL events, their surface impacts, their relationship to SAM/ENSO/ZW3, and a preliminary CESM2-based future projection. The topic is relevant and the research question is well motivated, the writing is clear and the manuscript logically structured. The investigation of the connection between ASL and ZW3 is particularly noteworthy, as it is not often addressed in the existing literature.
RC1 has already raised several important concerns regarding CESM2 validation, the ensemble-mean approach, the seasonal cycle/trend/extreme relationship, and ASL-ZW3 construction, with which I largely agree. Below, I expand on several of these points, connect some issues that I think arise from a common methodological problem, and add a few complementary concerns.
General points
1) Extreme-event definition: fixed baseline and seasonal cycle
RC1 raised three separate observations that I think are symptoms of one underlying problem: (a) the minor comment that "it is not clear... how a fixed mean can account for the observed deepening" (line 91); (b) the question of whether results are similar if the ASL series is detrended (line 347); and (c) General Point 3, asking whether the seasonal concentration of strong events reflects a shifted mean versus higher variance. All three seem to point to the same issue: a fixed threshold applied to a distribution with a drifting mean can mechanically increase the frequency with which later months cross the strong-event threshold, even in the absence of any change in variability. This is visible directly in the data: Table 1 shows 6 of 13 strong events fall within 2016–2022 alone (a 7-year window in a 46-year record), while Table 2's weak events occur only 1991–2007 and never after, which is consistent with the deepening trend in the ACP (line 159). Describing strong events as "well distributed" (line 189) understates this. The same fixed-mean logic is reused for CESM2, so one of the paper's central future findings, more strong events relative to weak (lines 342–349), may partly reflect a shift in the mean state rather than a genuine change in extremeness or variance.
An additional component to this problem is the seasonality of the extreme events. The ASL displays a pronounce annual cycle, reflected in the strong seasonality of the ACP that is also acknowledged by the authors. However, defining the extreme events based on a year-round threshold means that event selection may partly reflect the climatological seasonal cycle: months in which ACP is normally lower are inherently more likely to cross the strong-event threshold, while months with climatologically higher ACP are more likely to cross the weak-event threshold. It is therefore unclear to what extent these events are extreme relative to the conditions normally experienced in that calendar month.
2) Circularity risk in the MOV analysis
In their General Point 5, RC1 already flagged a possible construction-based link between the ASL and ZW3 indices specifically, noting that one of the ZW3 index's centres of action sits close to the ASL box. I think the underlying issue is broader than ZW3 alone. The authors explicitly state they use ACP rather than RCP "to acknowledge and further analyse the influence of additional SH modes of atmospheric variability on the climatological ASL" (lines 86–88), but RCP is constructed precisely to remove the large-scale pressure signal that ACP retains. This means that some of the large-scale SAM and ZW3 pressure imprint is already embedded in the ASL magnitude metric by definition, so a strong association between low ACP and SAM+ is expected by construction. I don’t think using ACP is necessarily wrong, but it might require a bit of extra care when presenting the results and possibly an additional sensitivity test, ideally repeating the Section 3.6 comparison using RCP as a robustness check against both SAM and ZW3, or at least the main points of the analysis.
3) MOV co-occurrence percentages lack a baseline for comparison
Setting circularity-point aside, raw co-occurrence counts like "12 of 13 strong events coincide with ZW3+" (lines 306–307) are only informative relative to the baseline rate at which ZW3+ occurs across the full record. A meaningful addition could be reporting baseline occurrence rates for each MOV phase, which would also help RC1's Point 4 by clarifying whether extreme ASL events show a distinctly stronger MOV association than typical months.
4) CESM2 ensemble-mean
I think this is an important point and thus worth reinforcing it, although RC1 already clearly brought it up in their General Point 2. The authors calculate future ACP using a three-member ensemble average, and extremes are subsequently detected from that averaged series. However, extreme circulation events are largely manifestations of internal variability and will not occur in the same month in independent ensemble members, so averaging over 3 ensemble members suppresses member-specific anomalies. This specifically matters as the authors later interpret periods with few events in the future projections (for example the approximately 25-year period of apparent inactivity under SSP5-8.5 in lines 356–360) as behaviour of the ASL system, while it might also be the result of cancellation among independent realizations, rather than a genuine reduction in variability. As a more robust alternative, one could identify the extremes events independently in each ensemble member and then analyse the distribution of their statistics across the ensemble.
5) Missing CESM2 validation for the ASL
As RC1 already noted (General Point 1), the analysis lacks ASL-specific validation for CESM2, which is crucial to assess whether the results reflect real changes in the ASL variability and occurrence of extremes or merely some model biases or behaviour.
6) Interpretation of future extreme-event frequency
The conclusion that future forcing leads to an “increased frequency” of extremely strong ASL events (lines 443–444) is not clearly demonstrated by the analysis presented. The authors report raw event counts of strong and weak events in the future simulations, but without a corresponding CESM2 historical event frequency calculated using the same method, it is difficult to establish whether strong extremes actually become more frequent relative to the model’s historical climate. The temporal behaviour of the two scenarios further complicates this interpretation: SSP585 produces fewer detected extremes than SSP245 (22 strong and 12 weak events compared with 27 strong and 15 weak events; lines 344–348) and shows an approximately 25-year period after 2065 with almost no extreme events (lines 356–360). The authors acknowledge that ASL depth appears less variable during the second half of SSP585 (line 350), but this is not further discussed in relation to the claimed increase in extreme-event frequency. As discussed in points 4 and 5, interpretation is also complicated by the use of the ensemble-mean series for event detection and the lack of an ASL-specific historical evaluation of the model. These limitations make it difficult to determine whether the reported future behaviour represents a genuine change in extreme-event occurrence or is partly a consequence of the analysis framework.
7) Composite robustness
It is therefore difficult to assess whether the anomalies associated with extreme ASL events are statistically distinguishable from those occurring during comparable non-extreme months, rather than reflecting ordinary internal variability.
Minor/Technical points
- Lines 155–160 and 344–349: trends are reported as "r" rather than physical trends. A physically interpretable slope would be make the results much clearer to interpret, especially when provided with uncertainty, that is currently is missing. The statistical method used to estimate and test the trends should also be specified.
- Lines 205–207: the reported average latitude for strong events (71.05°S) is explicitly noted to include one anomalous event at 74.5°S that pulls the mean southward. With n=13, it would be useful to indicate how sensitive the reported mean is to this event.
- Lines 212–215: the description of “large zonal spatial heterogeneity” for weak events appears reasonable from Fig. 3, but with only eight events, the statement that the results “validate” previous work may be too strong. “Consistent with previous work” would seem more appropriate.
- Line 428 “significantly” can be misleading unless referring to some statistical significance test. Maybe “substantially”?
- Tables 1+3 and Tables 2+4 fully overlap on date/ACP columns, consider merging each pair.
- I’d also support RC1’s suggestion to merge Figure 9 and 10
Citation: https://doi.org/10.5194/egusphere-2026-2897-RC2 -
RC3: 'Clarification to General Point 7 of my referee comment (RC2)', Anonymous Referee #2, 18 Aug 2026
reply
General Point 7 was accidentally truncated when copy-pasting from my notes, I apologize. I report General Point 7 in its complete form below :
The composite maps are presented as mean anomalies without significance testing or an indication of uncertainty. It is therefore difficult to assess whether the anomalies associated with extreme ASL events are statistically distinguishable from those occurring during comparable non-extreme months, rather than reflecting ordinary internal variability. This is particularly relevant also given the small sample size (n=8 for the weak events).Citation: https://doi.org/10.5194/egusphere-2026-2897-RC3
-
RC3: 'Clarification to General Point 7 of my referee comment (RC2)', Anonymous Referee #2, 18 Aug 2026
reply
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 99 | 43 | 13 | 155 | 21 | 14 | 12 |
- HTML: 99
- PDF: 43
- XML: 13
- Total: 155
- Supplement: 21
- BibTeX: 14
- EndNote: 12
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript provides an interesting analysis of past and potential future extremes of the Amundsen Sea Low and their impact on temperature and sea ice concentration. The text is clear and well-written. It describes well the extremes and their effects and provides interesting connections to previous work. However, I have some suggestions below to expand points and for additional precisions on some aspects.
General point
1/ The study analyses the results of one model (CESM2). It is always better to have a multi-model approach for such kind of studies but in this particular case I do not think it is absolutely necessary. However, the validation of the CESM2 run for the historical period is too short. It is mentioned line 144 that CESM2 reproduces well some aspect of the Southern Ocean circulation (with one reference) but a specific evaluation for the ASL is needed. Does it reproduce well the seasonality of the ASL strength and position, the timing of the extremes observed during the historical period ? Without those elements, it is impossible to evaluate if the differences discussed in section 3.7 are representative of a change in the system or just related to model biases already present in the historical period.
2/ I am also wondering why the ensemble mean of the three simulations is analyzed instead of the individual members (line 147). By construction, this ensemble mean will smooth the extremes as they do not occur at the same time in each member. I would strongly recommend to make the analyses of the extremes on each member and then eventually take the average to have a mean response rather than analyzing the ensemble mean itself.
3/ The link between the seasonal cycle of the ASL, its trend and the extremes could be discussed in more details. For instance, most of the strong ASL events occurs in September or October. Is it simply due to a different mean or a higher variance during these months ? Are the characteristics of these extreme events (like the location) due to the fact that they occur during these months or are the extremes different from other events occurring the same month ?
4/ A question related to the previous one is more generally about the characteristics of the extremes. Could we consider that the links with temperature and sea ice are the same for the extremes and more normal cases (as revealed for instance by a standard correlation analyses with ASL index) or are the extremes showing specific elements that are related to the larger magnitude of the perturbation ?
5/ The connection between ASL and ZW3 is mentioned but I wonder how the two indices are connected just by construction as one of the pole used to construct the ZW3 index is close to the region used to define the ASL index (lines 132-133).
Minor points.
Line 13. From the structure of the sentence, it is not clear to me whether the words ‘based on magnitude’ refer to ‘geographically’ or ‘events’.
Line 71. You mention ‘amplify baseline warming’ . Do you mean locally (as this can also damp the warming in other regions I guess) ?
Line 91. It is not clear to me how a fixed mean can account for the observed deepening.
Line 93. Is it parameterizing or rather characterizing ?
Line 116. I do not understand the wording ‘uses integers’ as the index can include decimals? The same remark is valid for line 138 that cites as an example a value of 0.9. Should it be ‘scalar’ instead of ‘integer’ ?
Line 347. Are the results similar if the series is detrended ? This diagnostic could help to disentangle the contribution of the trend in the behavior.
Line 351-360 . I am not sure I get the message from this part. Is it just the variability of the system ? Is it robust between ensemble members?
I suggest to group Fig 9 and Fig 10 in one figure.