the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Aug 2026
The contribution of the surface temperature inversion to Antarctic warming
Abstract. Antarctica is expected to warm faster than the global average throughout the 21st century in a phenomenon called Antarctic amplification. The polar atmosphere is unique in having a strong and persistent surface temperature inversion layer, and its changing properties can dramatically affect polar warming. In this work, we reveal mechanisms behind Antarctic amplification by quantifying the contribution of this surface temperature inversion to the region’s near-surface warming. Global climate models often miss the vertical resolution and physical processes needed to correctly simulate the intense near-surface inversion. Here, we downscale four global climate models with the regional climate model MAR (Modèle Atmosphérique Régional) to decompose the Antarctic near-surface potential temperature into the background potential temperature and the surface temperature inversion strength. From the late 20th to the late 21st century, we attribute 72 % of July near-surface warming over the Antarctic continent and sea ice to large-scale background warming under the Shared Socioeconomic Pathway SSP5-8.5. However, the spatial variability in this warming is driven by surface processes through their weakening of the surface temperature inversion, which explains the remaining 28 % of warming. Over the ocean, this weakening is well predicted by sea ice loss, and accounts for 41 % of the near-surface warming. By contrast, the weakening of the inversion explains 16 % of near-surface warming on the continent, where it is correlated with stronger present-day atmospheric stability, which suggests that the lapse-rate feedback contributes significantly to spatial variability in warming. The particularly strong contribution of sea ice loss to the weakening of the inversion could explain the asymmetry between Arctic and Antarctic amplification. Variability in the inversion strength could also account for discrepancies between paleoclimate proxy records.
- Preprint
(6365 KB) - Metadata XML
-
Supplement
(3282 KB) - BibTeX
- EndNote
Status: open (until 28 Oct 2026)
- RC1: 'Comment on egusphere-2026-4032', Anonymous Referee #1, 07 Oct 2026 reply
-
RC2: 'Comment on egusphere-2026-4032', Anonymous Referee #2, 19 Oct 2026
reply
The authors investigate the contribution of surface temperature inversions to increasing Antarctic near-surface temperatures in winter under a very-high emissions Shared Socioeconomic Pathway. Projections from four separate GCMs are downscaled with MAR and the theta profile is analysed. The authors decompose projected near-surface warming into changes in background potential temperature and changes in the near-surface potential-temperature deficit, and investigate their relationships with sea-ice concentration and atmospheric stability. This is an ambitious piece of research which is novel and thorough. The presentation of figures is great and the multi-GCM perspective is very useful. I feel substantial further work is required to justify the terminology used and figures quoted, see below:
Major comments
- Please clarify the relationship between the quantity being studied in this paper and surface temperature inversions. The results of the paper are based on the theta deficit. A surface temperature inversion describes temperature increasing with height above the surface. Maybe the two conditions coincide during austral winter but please quantify that if so.
- Connected to the above, this paper uses causal language, e.g. L77-79 and L437-439. However, changes in the potential-temperature deficit are a diagnostic of changes in the vertical temperature structure of the atmosphere, rather than necessarily a cause of warming. For example, a weakening of the deficit could simply be a consequence of near-surface warming exceeding free-atmospheric warming. Therefore, I think phrasing such as "accounts for" or "is associated with" would be more appropriate than "causes".
- The 28% contribution value doesn’t line up with Table 3 (2.7/7.6) which gives a 35.5% contribution rather than 28% contribution? I’m wondering whether 28% was obtained by averaging percentage changes at different gridpoints. If so this would give equal weight to gridpoints with very different warming magnitudes, e.g. a gridbox with a 0.1K warming gets the same weight as one with a 5K warming. Please explain the method, and also whether area-weighting was used.
- Use of ‘bias’: there are several references to GCM bias with respect to MAR simulations. Typically bias describes the difference between a model simulation and a reference dataset (observations or quasi-observations). There are two uses of this term in this paper which I comment on separately:
- GCM bias with respect to MAR historical simulations: this term can arguably be used if the MAR historical simulations are forced by reanalysis; however, they are forced by historical GCM simulations I believe? So this cannot be considered a reference dataset.
- GCM bias with respect to MAR future simulations: here I think it’s harder to use the term ‘bias’ at all, e.g. as in Figure 11. Instead I suggest just describing the difference between the two simulations.
- L231: L231: Please justify the use of the 4.1 K threshold. As I understand it, the regression is intended to predict Delta D_theta from present-day static stability, but grid cells are excluded based on their actual Delta D_theta. Does this not mean that the selection of prediction grid cells depends on prior knowledge of the quantity being predicted? How would the results change without this threshold?
Minor comments
- Title: I suggest ‘Antarctic winter warming’ is more appropriate.
- L73: Gilbert et al. (2026) don’t, to my knowledge, show improved performance of MAR temperatures with respect to ERA5? See Figure 8 therein which suggests the opposite.
- L79: Please justify (in the methods section) the use of SSP5 here.
- References to Figure 4 and 5 could be more specific– currently they are referenced together on L255 and L258. Please include panel labels in Figure 4 (or rows as you did for Figure 3) to help point the reader to the right place in the figure.
- Figures 6 and 7 captions: I suggest not to describe the findings in the figure caption – it should just describe the figure contents.
Citation: https://doi.org/10.5194/egusphere-2026-4032-RC2 -
RC3: 'Comment on egusphere-2026-4032', Anonymous Referee #3, 19 Oct 2026
reply
Pdf attached.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 339 | 165 | 57 | 561 | 94 | 69 | 63 |
- HTML: 339
- PDF: 165
- XML: 57
- Total: 561
- Supplement: 94
- BibTeX: 69
- EndNote: 63
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of “The contribution of the surface temperature inversion to Antarctic warming”.
This study dynamically downscaled outputs from four global climate models using the regional climate model MAR to decompose Antarctic near‑surface potential temperature into background potential temperature and surface‑temperature‑inversion strength. The authors quantified the contribution of changing inversion strength to July Antarctic near‑surface warming in the 21st century, finding that inversion weakening accounts for approximately 28 % of total near‑surface warming (41 % over sea‑ice‑covered areas and 16 % over the continental landmass). The topic is scientifically meaningful, as this work presents the first quantitative decomposition linking surface‑temperature inversion, sea‑ice loss and lapse‑rate feedback over Antarctica. The study also contains several methodological strengths. Nevertheless, substantial revisions to methodology and text presentation are required before acceptance.
Major comments:
is defined as the departure from a linearly extrapolated free‑atmosphere potential‑temperature profile. Its magnitude depends strongly on the criterion used to select the lower bound of the free‑atmosphere layer (Equation A1). The physical meaning and original rationale for this threshold criterion should be explicitly described. Sensitivity tests for this parameter should be provided to quantify its impact on derived inversion‑strength estimates.
Multiple passages frame statistical associations as causal drivers, for example the statement “SIC is an important control on warming and amplification”. The supporting evidence comes primarily from spatial regression and cross‑validated prediction, rather than targeted perturbation‑based sensitivity experiments. The authors should soften causal wording and clearly distinguish correlative statistical evidence from mechanistic causal conclusions throughout the manuscript.
It remains unclear how extreme cases (e.g., when approaches zero) are handled. The weighting scheme for regional averaging must be explicitly documented (i.e., whether grid‑cell values are simple grid‑point averages or area‑weighted averages). Where grid cells feature inversion strengthening (, the derived fractional contributions for background warming and inversion change no longer sum to 100 %. However, the abstract reports background warming (72 %) and inversion‑related warming (28 %) as perfectly complementary fractions. The authors must resolve this inconsistency and fully describe their aggregation workflow.
Projecting future from a present‑day spatial quadratic regression assumes that the statistical relationship between sea‑ice concentration and inversion strength remains stationary under anthropogenic warming. Although the authors evaluate performance using MAE and MAPE, Table S6 shows systematic over‑prediction of inversion weakening. Furthermore, residual patterns reveal under‑prediction near the Antarctic Peninsula and over‑prediction along the East Antarctic coast, demonstrating non‑stationarity of this statistical relationship. The authors should explicitly discuss how such non‑stationarity may bias their key results.
Minor comments: