the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Aug 2026
Antarctic meltwater induces competing climate feedbacks in an Earth system model coupled to an Antarctic ice melt emulator
Abstract. Antarctic ice-sheet mass loss is expected to accelerate under future warming, yet its climatic impacts remain poorly represented in Earth system models. Here, we present simulations with the EC-Earth3 Earth system model that incorporate the Antarctic Ice Melt Emulator (AIME), interactively coupled to the ocean component. AIME simulates regionally and temporally varying freshwater fluxes driven by subsurface ocean temperatures over the Antarctic continental shelf, accounts for the delayed dynamic response of the ice sheet, and enables two-way interactions between the climate system and Antarctic ice loss.
Under an abrupt quadrupling of atmospheric CO2, the interactive freshwater forcing gives rise to two distinct feedback regimes. During the transient adjustment phase, enhanced freshwater input strengthens Southern Ocean stratification, suppresses vertical mixing, and modifies continental shelf circulation, leading to enhanced subsurface warming near the base of Antarctic ice shelves. During the subsequent quasi-equilibrium phase, the dominant climatic response shifts: reduced surface warming, increased sea-ice persistence, and decreased global ocean heat uptake weaken the atmospheric warming, ultimately reducing subsurface warming over the Antarctic continental shelf despite continued freshwater input.
Compared with a simulation without interactive Antarctic freshwater forcing, the coupled simulation exhibits lower equilibrium climate sensitivity and reduced global ocean heat storage under identical greenhouse-gas forcing. These results demonstrate that Antarctic freshwater feedbacks evolve over time, with an initial ocean-driven amplification followed by a longer-term atmosphere-mediated damping response. Our findings highlight the importance of representing interactive Antarctic freshwater fluxes in Earth system models to capture coupled ice-sheet-ocean-atmosphere feedbacks and improve projections of future climate change.
- Preprint
(12561 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-4584', Moritz Kreuzer, 04 Oct 2026
-
RC2: 'Comment on egusphere-2026-4584', Anonymous Referee #2, 15 Oct 2026
van der Linden et al present interesting evidence that feedbacks between Antarctic ice sheet discharge and the global climate evolve over time. The linear response function (LRF) methodology used to incorporate long timescale ice ocean feedbacks in lower cost simulations is very promising, and simulations presented using this framework convincingly demonstrate that LRF style meltwater feedbacks initially amplify melt (first ~50 yrs) before eventually dampening melt (latter ~50 years) relative to a simulation with a simple ice sheet surface mass balance meltwater protocol. The bones of this study are very good, but the authors reach beyond what their model diagnostics and analysis can fully support when attempting to determine the mechanism underpinning the switch from positive to negative meltwater feedbacks. They present a dichotomy between changes to cross continental shelf break heat transport driven by (a) ocean circulation changes and (b) open ocean heat content changes (essentially, the u and T components of heat transport). They don’t very explicitly state this as the comparison they are making (instead comparing “large scale ocean circulation” to “a broader mechanism: changes to the global energy balance”) but I think this is a faithful interpretation. The authors then attribute the long timescale reduction in cross shelf break heat transport to (b) without sufficient support. A convincing attribution of the feedback sign change signal between (a) and (b) may not be attainable with the model and diagnostics available, but, fortunately, I do not think this mechanistic attribution is needed for this to be an impactful and worthwhile paper well suited to publication in ESD. I ultimately recommend removing these claims and reframing the latter portion of the manuscript to discuss the equilibrium climate sensitivity results on their own terms, rather than trying to loop these results back to ice shelf feedbacks. However if the authors believe they can address my concerns regarding the (a)/(b) dichotomy with new analysis, they are welcome to attempt this rather than omitting these claims.
Methods Section (mostly clarity):
I agree with the prior reviewer that the methodology is quite involved, and the manuscript would benefit from some adjustments to make this section easier to digest. A few points of clarifications below, alongside some small suggestions that may make the methods easier to follow:
L100-105: Can the authors translate what this excess snow-pack routing scheme means for the ice sheet surface mass balance? Does this effectively cap or constrain positive SMB on the ice shelf?
L106: Does “added as ice” mean this ice is input into some form of ocean frazil package, into the sea ice field, included as snow on the sea or sea ice surface, included as icebergs which can be advected and melted by the ocean, or does it just mean that the heat and freshwater (rather than just freshwater) fluxes associated with melting ice are imposed at some boundary? Also, what is the spatial structure of this flux, is excess snow routed to the ocean evenly around the continent or in some horizontal pattern? Is it input at the surface or with some depth impression?
L129: Is the 1 m/yr basal melt anomaly imposed for the full 200 yr period or only transiently at the start?
Figure 2a: Are the units here (“GT/m”) referring to the sensitivity in GT/yr of ice loss the described 1 m/yr perturbation in area averaged basal melt rates? Methods section would be easier to follow if each figure and figure caption is a bit more explicit in describing units.
Table 1: As for Figure 2a, caption could be clearer in a way that steps the reader through the units of the reported values. (e.g. “Quadratic sensitivity of sector averaged ice shelf basal melt rates (m/yr) to the relevant sector averaged ocean temperature anomalies (K), derived from (…).”). Also, I got a bit confused at this stage regarding where differences in ice shelf area came into the model, I think they are accounted for by the LRF (Figure 2a) but it would be nice if this could be stated more explicitly.
L168: As previously (L129), I find “imposed as ice” ambiguous. Does the iceberg melt dynamically in response to the ocean forcing or is this simply a combined freshwater/heat flux condition at the ocean surface?
Figure panel arrangement: One suggestion for clarity in this section (take it or leave it) would be to combine panels Fig 1a, 1b, 2a, and the information in Table 1 (maybe overlaid on Fig 1a map?) into one figure, as the reader really needs to combine these to understand how we arrive at panel Fig 2b.
CTL and “baseline” melt distribution: The spatial distribution of only the AIME meltwater inputs is described in the methods section, whilst the distribution of CTL and “baseline” melt distribution is left to the Results. Please include some description of the “baseline” melt structure in the methods (probably around L106, see my clarification questions there) and include some justification for why this meltwater is not input with the same spatial patterns as AIME meltwater. All discharge from the AIS, including steady state or “baseline” discharge, enters the ocean as a combination of basal melt in ice shelf cavities and calved iceberg melt. It seems to me that the experimental setup convolves system response to (i) the introduction of melt feedbacks that change meltwater volume input, and (ii) a sudden change in the spatial and depth characteristics of meltwater input, rather than, as the authors claim “isolate(ing) the effect of two-way climate ice sheet coupling” (L217). I would like to see this discussed, perhaps in 2.3.
Results Section (larger concern):
Section 3.4 convincingly conveys that the difference between the AIME and NO-AIME continental shelf heat budget is that the additional basal melt heat sink at depth in AIME should act to cool the shelf in AIME relative to NO-AIME. Though surface heat fluxes do warm the shelf in AIME relative to NO-AIME, the magnitude of this warming is insufficient to counteract cooling by basal melt and their combined impact should be shelf cooling in AIME relative to NO-AIME. In the period when this isn’t the case (period A) this quantified cooling must be overcome by increased poleward heat transport (which the authors indicate they cannot directly quantify), then, in the period where relative cooling is indeed seen (period B), it is implied that this compensatory increase in heat supply is reduced. I am satisfied with this, but the authors then go on to attempt to determine whether these changes are circulation or heat content driven, and the evidence used for that separation is not very solid and it may be preferable to leave this as a question for future study.
In Section 3.5, the authors note that period A is characterized by a strong ASC and weak upper overturning cell (in AIME relative to NO-AIME) and state that these two signals combined are inconsistent with an enhanced cross shelf break heat transport, but that the required (for budget closure) increase in cross shelf break heat transport could feasibly be provided by the ice shelf meltwater pump. They then note that period B is characterized by a weak ASC and strong upper overturning cell (in AIME relative to NO-AIME) and take these signals to imply that ocean circulation factors ((a) in the dichotomy presented in my introductory statement) should be bringing more heat to the continental shelf, so an alternative explanation is needed to explain the relative cooling. This analysis hinges on the interpretation of the ASC strength, diagnosed here as the mean zonal velocity on the continental shelf, as a reliable metric for the impact of shelf circulation on cross shelf break ocean heat transport. Unfortunately, both the definition of the ASC strength and the authors’ interpretation of its meaning have issues. The ASC is a narrow structure at the shelf break rather than a broad structure spanning the full continental shelf and is not appropriately characterized by the shelf mean zonal velocity. Such a metric would be more characteristic of the Antarctic Coastal Current, a less studied feature with poorly constrained relationships to poleward heat transport (though the little work that has been done on it suggests it may enhance heat transport into ice shelf cavities in some cases, see Flexas et al 2024 [https://doi.org/10.1126/sciadv.abj9134]). Secondly, the ASC is not a simple barrier to poleward heat transport. Even if the authors where to quantify its strength in a more appropriate manner, ASC velocity variations cannot be assumed a good proxy for cross shelf break heat supply variability. A recent relevant study to consult would be Aguiar et al (2026) [https://doi.org/10.1029/2025JC023934], but I would also point the authors to the idealized simulations in Si et al (2023) [https://doi.org/10.1126/sciadv.add7049] showing a non-monotonic relationship between cross shelf break density differences and heat transport, and ask them to consider the role of the eastward flowing sheared ASC undercurrents, seen in warm shelf regions like the Amundsen Sea, which divert warm CDW onshore and can be anticorrelated with surface current variations at decadal timescales (see for e.g. Silvano et al 2022 [https://doi.org/10.1029/2022GL100646]). There have been some studies where a buoyancy driven acceleration of the ASC has coincided with a reduction in (explicitly quantified) cross shelf break heat transport (e.g. Moorman et al 2020 [https://doi.org/10.1175/JCLI-D-19-0846.1]) but, overall, I would caution the authors against overinterpreting ASC strength and encourage them to consider removing the ASC assessment, and claims dependent on it, from their analysis. In addition to these concerns, ASC responses to forcing have been shown to be sensitive to model resolution (Beadling et al 2023 [https://doi.org/10.1029/2021JC017608]), and it is doubtful that the 1º horizontal ocean model resolution here is sufficient to capture ASC dynamics in the mean or in response to perturbation.
In Section 3.6, the authors present an equilibrium climate sensitivity (ECS) and total ocean heat uptake assessment. They find that less heat accumulates within the climate system as a whole and the ocean in particular when meltwater feedbacks are included. This is a very interesting result in itself and, however, the claim that this smaller total ocean heat content anomaly mechanistically explains the switch from positive to negative melt feedbacks on the Antarctic continental shelf is underevidenced. The explanation of the proposed mechanistic connection is a bit vague, but I gather that the thought is that CDW transport across the continental shelf beak doesn’t reduce (inferred from the ASC proxies) in period B, but the total reduced ocean heat sink implies that this CDW is cooler than it otherwise could be. There are too many missing links in this explanation, even if we discount the issues with the ASC analysis. At very least the authors would need to show not only that the global OHC anomaly is reduced, but that temperatures of the CDW reservoir just equatorward of the continental shelf break are reduced. I think that the ECS results and the sensitivity differences between GSAT and Antarctic continental shelf temperatures are very interesting in their own right, but they don’t clearly evidence a particular mechanism of heat budget adjustment in the way the authors imply.
To reiterate, I think the experiments are very interesting and numerous of the results presented (including the change in feedback sign over time and the differences in ECS/OHC) are very compelling. I do not find the explanation of why feedbacks change sign at long timescales very clear, and the evidence presented to support it is not very compelling due to (1) misinterpretations of Antarctic continental shelf dynamics (primarily the ASC, both its definition and effects), and (2) missing links between global scale OHC changes and the regionally relevant heat reservoirs. The authors can choose to attempt to improve this mechanism argument, but I do not think that this line of argument is required for the paper to be valuable.
Citation: https://doi.org/10.5194/egusphere-2026-4584-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 278 | 152 | 48 | 478 | 47 | 41 |
- HTML: 278
- PDF: 152
- XML: 48
- Total: 478
- BibTeX: 47
- EndNote: 41
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General Comments
van der Linden et al. present a very interesting manuscript about the temporal evolution of the climatic feedbacks due to Antarctic meltwater input. The work is very well in the scope of ESD as the investigated processes relate to ice-sheet mass loss, ocean and atmosphere domains as well as local, regional and global spatial scales.
The authors use Linear Response Functions to emulate the Antarctic Ice Sheet freshwater input into the Southern Ocean by interactively evolving ocean forcing. This is a new concept in the context of Earth System Modeling and a valuable step towards fully coupled climate--ice-sheet model setups.
The paper is well structured and written and provides clear reasoning of the presented conclusion.
I recommend the publication of the manuscript based on minor revisions, which are listed in the specific comments below.
Specific Comments
Section 2.2.1:
The Section includes all relevant information. However, I needed to re-read it several times to fully understand the methodology. I suggest to add a small schematic figure that illustrates the concept visually and especially helps readers that are not very familiar with ice-sheet basal melt parameterisations (quadratic, LADDIE, ...) and Linear Response Functions to understand the processing chain.
Fig 5
Fig 12: I think NO-AIME dOHC/dt should be displayed in blue dashed for consistency
l 472: leading to cooler subsurface shelf waters: The overall ocean anomaly is still net positive in AIME experiment. Revise phrasing: cooling is only relative to NO-AIME experiment
l 521-531: add iceberg melt distribution at surface as further simplification/limitation
l 553: again, cooling is only relative compared to non-interactive freshwater input.
Technical Comments
l 347: remove one from