the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Feb 2026
Diverging runoff drives uncertainty in Antarctic surface mass balance projections under a high emission scenario
Abstract. Three recent downscalings of CESM with MAR, RACMO, and HIRHAM under SSP5-8.5 produce consistent contemporary Antarctic surface mass balance (SMB), but diverge strongly by 2100, especially over ice shelves. HIRHAM simulates a large SMB decline driven by strong runoff increases, MAR a moderate decrease, while RACMO maintains near balance. These differences mainly reflect contrasting melt–albedo feedbacks, present-day melt and refreezing levels, and a persistent 1–2 °C temperature offset between MAR and RACMO. CESM shows a decline similar in magnitude to MAR, with high melt partly compensated by extensive refreezing. Over grounded ice, all models project increased SMB from higher snowfall, though runoff still drives their spread. Despite shared boundary conditions and similar contemporary SMB, model behavior diverges, and CESM’s integrated results resemble MAR's despite its coarse resolution. Performance differences in present-day melt suggest that uncertainty estimates should account for model skill, motivating Bayesian treatment of multi-model ensembles.
Competing interests: At least one of the (co-)authors is a member of the editorial board of The Cryosphere.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(9816 KB) - Metadata XML
-
Supplement
(3115 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-624', Anonymous Referee #1, 01 Apr 2026
-
AC1: 'Reply on RC1', Benjamin Heurgue, 09 Jul 2026
We thank Reviewer 1 for their careful reading of the manuscript and their constructive comments. We have carefully addressed all comments point by point in the sections below. For each comment, we indicate the proposed changes to the manuscript text, with reference to the relevant line numbers of the original submission.
Major comments
On the use of a single GCM (CESM2) and its limitations
We agree that the use of a single ESM to force the RCMs limits the representation of the full range of plausible future climates. We propose to add the following sentence in the final paragraph of the Introduction to explicitly acknowledge this limitation:
“However, forcing the RCMs with a single ESM (CESM2) limits the representation of the range of plausible future climates and associated uncertainties (e.g. Kittel et al., 2021).”
Regarding the model naming, we acknowledge that CESM1 and CESM2 are distinct models. In the current manuscript, we referred to CESM2 as “CESM” for consistency with the naming convention adopted for the RCMs (MAR, RACMO, HIRHAM). However, to avoid any ambiguity, we will revise the manuscript to explicitly refer to “CESM2” throughout.
On the introduction of satellite datasets (AMSR2 and SSM/I) in the Results section
We acknowledge this comment. We note, however, that the comparison with satellite observations is presented following the funnel structure of the Results section: the analysis begins with a broad evaluation of present-day SMB against ERA5, before zooming in on the specific component that diverges most between models (melt), where observational data are then introduced for context. We also note that the description of these datasets and the justification for using two products are already provided in the paragraph on page 12. We believe this placement remains appropriate given this narrative structure, but we will ensure the datasets are at least mentioned in the Methods section for completeness.
On the size of map plots (e.g. Fig. 3)
We agree that the current 4×2 layout makes it difficult to identify small-scale features such as ice shelves. We will reorganize the figure into a 2×4 layout, which will result in larger individual panels and improve readability of fine-scale features.
On standalone paragraphs in the Results section (e.g. around Fig. 4)
We agree. The standalone paragraphs will be integrated into the surrounding text by removing unnecessary paragraph breaks, ensuring a better narrative flow throughout the Results section.
Minor comments
L31 – Citation needed for warmer ice shelves
We will move the following citations to the end of both sentences:: (Kittel et al., 2021; Boberg et al., 2022; Van Wessem et al., 2023; Jourdain et al., 2025)
L33 – Mention of GCMs alongside ESMs
We agree. We will revise the sentence to explicitly distinguish GCMs from ESMs and acknowledge that GCMs lack certain biogeochemical processes. The revised sentence will read:
“The climate models used to estimate the SMB of ice sheets fall into two main categories: global climate models (GCMs), including the more complex Earth system models (ESMs), and regional climate models (RCMs), each having its own strengths and limitations. The main advantage of ESMs for simulating ice-sheet SMB lies in their physically consistent global framework, with interactive coupling between the atmosphere, ocean, sea ice, and land surface, as well as the inclusion of additional biogeochemical processes compared to traditional GCMs.”
L36 – Number of RCM limitations
We agree that the current phrasing incorrectly implies that RCMs face only two major limitations. We will rephrase the sentence to make clear that the two limitations mentioned are illustrative examples among several, rather than an exhaustive list.
L53 – Citation for model uncertainty (Hawkins and Sutton, 2009)
We thank the reviewer for this suggestion. We will add the citation to Hawkins and Sutton (2009) at L53, which provides a useful framework for distinguishing model uncertainty from scenario uncertainty and natural variability in climate projections.
L57 – Repetitive mention of high-emission scenario
We agree that this sentence is repetitive. We will revise it to remove the redundant mention of the high-emission scenario while retaining the key information.
” To address this gap, the present study performs an intercomparison of SMB simulated by these three RCMs, all driven by the same ESM, the Community Earth System Model (CESM), under a high-emission scenario (SSP5-8.5).”
L81 – ‘Resolving’ vs. ‘capturing’
We agree. We will replace “resolving” with “capturing” to better reflect the fact that some of these processes are parameterized rather than explicitly resolved.
L84 – Define CMIP6
We will add the full name “Coupled Model Intercomparison Project phase 6” at first use at L84.
L98 – CESM2 naming inconsistency
This inconsistency will be resolved by adopting “CESM2” consistently throughout the manuscript, as noted in our response to the major comment on model naming above.
L144 – ‘Few widely spaced layers’
We thank the reviewer for flagging this ambiguity. We will clarify what “few widely spaced layers” refers to in the context of the HIRHAM description, and revise the sentence accordingly.
Section 2.1 – Introduction of snow and firn schemes
We agree that the model descriptions jump directly into snow and firn schemes without explicitly stating their role in computing melt, percolation, retention, refreezing, and runoff. We will add a sentence at the beginning of Section 2.1 to clarify this, and will also add a sentence in the Introduction to contextualize the statement about the rudimentary representation of the snowpack.
Section 2.2 heading
We agree that “analysis” is not appropriate here. We will remove it from the section heading.
Table 1 – Terminology and title
We will revise the column headers to “Surface albedo” and “Maximum water capacity of firn”. We will also update the table title to: “Main characteristics and parameterisations of each model for the representation of the SMB.”
Section 2.3 – Comment on the 10 km common grid
We agree. We will add a sentence noting that the 10 km common grid is finer than the native grids of several models, and in particular that it represents a substantial downscaling step relative to CESM2, which operates at resolutions of 100 km.
Table 2 – Integration domain
The table heading already specifies “integrated over the entire Antarctic ice sheet”.
L234 – Missing citation
We will add the citation Jakobs et al. (2021) at this location.
Figure 3 caption – Use of ‘anomaly’
We agree that “anomaly” is not the appropriate term here, as the figure shows differences between period means rather than deviations from a baseline. We will revise the caption to read:
“Future mean SMB (2080–2099) relative to present mean SMB (1995–2014)…”
Or we can talk about “change” rather than “anomaly” for this case.
We will also check and correct all other uses of “anomaly” in the text where the same issue may arise.
L275 - L276 – Identification of ice shelves on maps
We agree that many ice shelves are difficult to identify in the current maps. We will add a supplementary map labelling the main ice shelves and geographic locations referenced in the text, and will refer to it where relevant.
L295–296 – Standalone paragraph
We agree. The standalone paragraph will be merged with the surrounding text by removing the unnecessary paragraph break.
Figure 4 caption – Vague reference to ‘parts of the ice sheet’
We thank the reviewer for this comment. The caption already specifies the two parts referred to: "grounded (left) and floating (right) parts of the ice sheet." We will simply add "Antarctic" before "ice sheet" for completeness.
L332 – ‘Observed differences’
We agree that the term "observed" is neither appropriate nor necessary here. We will simply remove it.
Summary – Mention of Bayesian treatment
We agree that the mention of a “Bayesian” treatment in the abstract is not followed up in the manuscript. We will introduce a brief mention of this approach in the Summary section to ensure consistency with the abstract.
Citation: https://doi.org/10.5194/egusphere-2026-624-AC1
-
AC1: 'Reply on RC1', Benjamin Heurgue, 09 Jul 2026
-
RC2: 'Comment on egusphere-2026-624', Devon Dunmire, 23 Apr 2026
-
AC2: 'Reply on RC2', Benjamin Heurgue, 09 Jul 2026
We thank Reviewer 2 for their thorough and constructive review of our manuscript. We appreciate the positive assessment of the manuscript’s clarity and the relevance of the results, particularly in the context of ongoing ISMIP7 efforts. We have carefully addressed all comments and questions raised by the reviewer, point by point, in the sections below. For each comment, we indicate the proposed changes to the manuscript text, with reference to the relevant line numbers of the original submission.
General comments
On the 2% water retention value in RACMO
The version of RACMO used in this study, as well as in Van Wessem et al. (2018), employs the original internal firn model described by Ettema et al. (2010), in which the irreducible liquid water retention capacity is prescribed as a fixed value of 2%. This parameterization differs from the more recent IMAU-FDM configuration described in Machguth et al., where the water holding capacity depends on firn porosity. We have therefore replaced the incorrect citation with Ettema et al. (2010) and clarified this point in the revised manuscript.
On the apparent discrepancy between RACMO and MAR runoff despite similar melt values
This is indeed a complex issue resulting from the interplay of several factors, and we agree that the discussion in L482–500 would benefit from being made more explicit. We identify at least four contributing mechanisms:
- Firn layer thickness and water retention capacity. The thinner firn column in MAR (20 m) compared to RACMO (30 m) could limit refreezing capacity once the firn is saturated, potentially contributing to higher runoff in MAR. However, MAR’s higher water retention threshold (~5%) largely compensates for the shallower column. A simple calculation illustrates this:
Density assumption
RACMO (30 m, 2%)
MAR (20 m, 5%)
Uniform 400 kg m⁻³
339 L
565 L
Uniform 500 kg m⁻³
273 L
456 L
Layered profile (0–2 m @ 300, 2–15 m @ 500, >15 m @ 720 kg m⁻³)
211 L
419 L
Across all tested density profiles, MAR’s total water retention capacity is roughly twice that of RACMO. Firn layer thickness alone is therefore unlikely to be the primary driver of the runoff difference in most cases, and we prefer not to highlight it as a leading explanation without stronger supporting evidence.
- Rapid FAC depletion in MAR. MAR produces substantially more liquid water (through both melt and rain, given its warmer climate), which rapidly saturates the available pore space and consumes firn air content (FAC), regardless of the higher retention threshold. Additional factors — such as differences in densification parameterizations, fresh snow density, and firn regeneration capacity — may further accelerate FAC reduction.
- Simulation continuity and firn thermal state. RACMO is run as a single continuous long simulation; in a warming climate, its firn column is therefore further from equilibrium and more “eager” to refreeze meltwater. MAR simulations are typically composed of shorter segments, starting closer to equilibrium with reduced refreezing capacity. Sensitivity tests with MAR run as a single long simulation yield results very similar to the segmented approach, however, because melt events rapidly homogenize the firn state — suggesting this factor plays a secondary role.
- Treatment of runoff at ice lenses. RACMO does not parameterize runoff when percolating meltwater encounters an ice lens — runoff only initiates once the entire firn column is saturated. MAR prescribes that a fraction of percolating meltwater runs off upon encountering an ice lens. This structural difference in runoff parameterization likely contributes significantly to the higher runoff in MAR for equivalent melt rates.
Given the number of competing and interacting factors, we are cautious about attributing the runoff discrepancy to any single cause. We will revise the discussion in L482–500 to present these mechanisms more clearly and concisely.
On the use of annual vs. summer (DJF) temperatures in Figures 6–8 (and Fig. S3)
We thank the reviewer for this suggestion. We agree that confining the analysis to summer (DJF) near-surface temperatures would be more physically meaningful for the processes discussed (melt, runoff, melt-albedo feedback). However, this is a limitation of the available model output: the projections were produced on an annual calendar basis through 2100, and monthly or seasonal components are not available to us for this analysis. We therefore cannot disaggregate the results by season, and the analysis is necessarily restricted to annual means and cumulative quantities. We will explicitly acknowledge this limitation in the revised manuscript.
On the introduction of satellite products and evaluation metrics in the Methods section
We agree with this comment. AMSR2, SSM/I, and the evaluation metrics (CMS, MMS, MMD, MOA) will be introduced and described in the Methods section rather than first appearing in the Results section.
Minor comments
Abstract
We agree that the last sentence of the abstract lacks sufficient context. We will revise it to better reflect the manuscript’s conclusions. Regarding a more extensive rewrite, we will first verify the abstract length limit imposed by the journal and extend it accordingly if the limit allows.
L17
We will restructure the sentence as recommended: “from -40 ± 9 Gt yr⁻¹ over the 1979–1990 period to -252 ± 26 Gt yr⁻¹ over the 2009–2017 period.”
Introduction – Mottram et al. 2021, Machguth et al. 2026, Medley et al. 2022
We will include Mottram et al. (2021) and Machguth et al. (2026) in the introduction as an additional RCM intercomparison study, noting that it focuses on the historical period. Both references will be contextualized with respect to the present study.
L22–24
We will add a reference to Medley et al. (2022) at this location to note that more sophisticated firn models also predict that most meltwater refreezes in the snowpack over Antarctica.
L26–29
We agree that this sentence is too long and convoluted. We will break it into two shorter sentences for clarity.
L46–48
We agree that this sentence requires clarification. By this statement, we mean that RCMs, while refining the spatial representation of climate forcing, are not independent of the driving GCM and may inherit or amplify its biases, introducing an additional source of uncertainty in the context of future projections. We propose to rephrase the sentence as follows:
"However, RCMs need to be forced by an ESM at their boundaries, which may introduce additional uncertainty due to potential internal biases or differences in climate sensitivity, particularly in the context of future climate projections."
L75–78
We agree and will restructure the sentence as suggested: “The omission of wind-driven erosion and deposition is not expected to compromise our intercomparison since runoff rapidly becomes…”
L191
We refer to “future anomalies” as the difference between the mean of the 2080–2099 period and that of the reference period 1995–2014, and to “time series anomalies” as the deviation of each annual value from that same reference period mean. We will clarify this terminology in the manuscript.
L229
Current sentence : « This refreezing partially suppresses melt, yet significant runoff still occurs,... ».
"melt" is correct in this context. The sentence refers to the previous statement that HIRHAM simulates melt rates approximately five times greater than those of the other two RCMs, accompanied by refreezing about four times higher than in MAR and RACMO. The phrase "this refreezing partially suppresses melt" therefore means that the higher refreezing rate partially compensates for the higher melt rate, yet significant runoff still occurs. No change is needed.
QuikSCAT comparison
L256
We agree that this sentence is confusing and will rephrase it for clarity. The sentence discusses how changing the melt detection threshold from 1 to 3 mm day⁻¹ affects the Mean Melt Duration (MMD) differently across models. For HIRHAM, increasing the threshold has only a limited impact on MMD, suggesting that most melt days in HIRHAM exceed 3 mm day⁻¹ regardless of the threshold applied. In contrast, for MAR and RACMO, raising the threshold reduces the number of melt days counted, as fewer days exceed the more restrictive threshold, which has a more significant impact on their respective MMD values. We will revise the sentence to make this distinction explicit.
L322
We will add “(near-surface air temperature)” after “climate response” for clarity.
L337
This is indeed a strange mistake. We will correct to W m⁻² K⁻¹ both in the text at L337 and in the caption and axes of Figure 6.
L343–L345
We will specify the starting temperature for HIRHAM over ice shelves at L343, and correct the RACMO starting temperature to ~254 K at L345.
L346
We agree that the sentence is imprecise. What we mean is that HIRHAM stands out from the other models throughout the entire simulation period, from start to finish, in terms of the magnitude of its response. We will revise the sentence to clarify this, and explicitly acknowledge that panels c, e, and f are exceptions, where HIRHAM values are close to those of CESM2.
Figure 8 – MOA vs. temperature and legend
We will add a supplementary figure showing MOA as a function of near-surface temperature. We will also update the figure legend to explicitly indicate that open dots represent accumulation and closed dots represent melt.
L516
We will correct the citation to “The Firn Symposium team” as recommended.
Figure S5
We apologize for this omission. Figure S5 was inadvertently left out of the submitted document and will be included in the revised submission.
Technical corrections
We will implement all suggested technical corrections:
- L1: Add “Global Climate Model” before CESM2 and “Regional Climate Models” before MAR, RACMO, and HIRHAM, and use CESM2 consistently throughout the manuscript.
- L84: Move the Danabasoglu reference to parenthesis.
- L145: Add missing period after “SMB model”.
- L235: Remove the duplicated sentence.
- L263: Remove “can” from “We can also”.
- L407: Correct “F. 8” to “Fig. 8”.
Minor clarifications throughout the text
We will implement all suggested clarifications: adding “SMB” or “change in SMB” at L7; “mass” at L16; “air” at L27; specifying the referent of “This figure” at L266; adding “SMB” at L279, L285, L286, and L306; adding “over ice shelves” at L318.
Jan De Rydt – topical editor
Minor corrections:
Line 12: start with “THE Antarctic Ice Sheet”…
Line 22: remove duplicate “snow”
Line 34: its -> their
Line 84: (Danabasoglu et al., 2020) between brackets
Line 145: full stop missing after SMB model
Line 179: clarify if the common ice mask is applied in each of the RCMs, or as a post processing step when analysing the data
Line 190: does “-future anomalies-“ need the hyphens, or did you mean this to be in italic font?
Line 191: same as above for “-time series anomalies-“
Line 200: remove “result”
Line 212: comma instead of a full stop before “Confirming”
Line 419: “Ross ICE shelves”
Citation: https://doi.org/10.5194/egusphere-2026-624-AC2
-
AC2: 'Reply on RC2', Benjamin Heurgue, 09 Jul 2026
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 941 | 862 | 85 | 1,888 | 159 | 103 | 153 |
- HTML: 941
- PDF: 862
- XML: 85
- Total: 1,888
- Supplement: 159
- BibTeX: 103
- EndNote: 153
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This is a well written paper, focused on downscaling projections from a single global climate model (CESM) under high-emission scenario using three regional climate models, in order to look at the differences between models (model uncertainty). The methodology is largely sound. The focus is on SMB and shows that the downscaled projections of this diverge strongly by 2100, which is due to differences in how the models represent melt-albedo feedbacks, present-day melt and refreezing levels, and temperature differences.
Major comment:
+ I realise that the aim is to focus on the differences between the RCMs, and therefore one global climate model (CESM) is enough for this aim. However, I feel the manuscript needs to better explain the limitations of this in understanding the future climate, i.e. it is only one realisation of the future climate. Maybe all that is necessary is an additional line or two in the final paragraph of the Introduction. But it would be good to have some information on e.g. how CESM compares with other CMIP6 models on key aspects such as future warming, sea ice loss, climate sensitivity, even SMB. Also, I’m not convinced that referring to the model as CESM is correct, as CESM1 and CESM2 are very distinct models, with CESM2 shown to have a reasonable representation of Antarctic climate and SMB (Dunmier et al., 2022)
+ The mention of satellite observations AMSR2 and SSM/I in the results section was rather out of the blue. These datasets should be mentioned in Section 2 (including limitations) and discussed as part of the methodology – for example, why are two datasets being employed. This is explained in the paragraph on page 12, but this should be in the methods section.
+ The map plots are quite small, making it difficult to identify small-scale features such as ice-shelves (e.g. Fig. 3). This makes it difficult to follow certain statements such as HIRHAM shows that SMB decreases over ‘all’ ice shelves in the future (Fig. 3a).
+ There are occasions in the results where the stand alone paragraphs seemed inappropriate, especially explaining Fig. 4.
Minor Comments
+ Line 31: This sentence on warmer ice shelves requires a citation.
+ Line 33: I think also need to mention GCMs, as GCMs and ESMs do differ in their complexity and representation of a range of processes. For example, GCMs do not include biological and chemical processes, while ESMs do.
+ Line 36: They face more than two major limitations, with representation of cloud microphysics being another one, for example. But this sentence rather suggests that there are only two major limitations.
+ Line 53: I would suggest citing Hawkins and Sutton (2009) here, which clarifies the role of model uncertainty in projections, along with scenario uncertainty and natural variability. Hawkins and Sutton (2009) The potential to narrow uncertainty in regional climate predictions. Bulletin of the American Meteorological Society, 90 (8). pp. 1095-1107. ISSN 1520-0477 doi: 10.1175/2009BAMS2607.1
+ Line 57: This sentence is a little repetitive, and the previous sentence also mentioned high-emission scenario.
+ Line 81: Maybe ‘capturing’ is better than ‘resolving’ in this sentence, as some of these processes are maybe parameterised.
+ Line 84: Define CMIP6.
+ Line 98: CESM2 used here but not defined.
+ Line 144: Its not clear what the ‘few widely spaced layers’ is referring to here. Surface? Atmosphere?
+ Section 2.1: The model descriptions jump straight into snow schemes and firn schemes, but without really explicitly saying that these schemes are responsible for computing melt, percolation, retention, refreezing, and runoff. Maybe an additional sentence could be added to the Introduction to clarify this, which currently only really says ‘… representation of the snowpack … remains rudimentary’.
+ Section 2.2 heading: I don’t think use of the word ‘analysis’ is correct here.
+ Table 1: Maybe should be ‘Surface albedo’ and also ‘Maximum water capacity of firn’. Also the title says ‘main characteristics and parameterisations’ which is inappropriate, as these are the ones for SMB.
+ Section 2.3: I think a comment is necessary to explain that the 10 km common grid is finer than the actual native model grids, and in the case of CESM which is >100 km this represents quite a jump.
+ Table 2: Need to specify in the heading that the results are integrated over Antarctica.
+ Line 234: Need to include a citation for this statement.
+ Figure 3 caption: Is the term ‘anomaly’ correct here? Is the top row not just the differences between the means for the two periods? This is not the same as ‘anomalies’. See also use of ‘anomalies’ in text.
+ Line 275: Its not possible to identify many of the ice shelves from these maps, as they are not shown. I also don’t think that the statement that they mostly show a decline in SMB (negative anomalies) is true – as this is only readily identifiable for HIRHAM over Ross and Ronne-Filchner ice shelves.
+ Line 276: These place names (and others mentioned) need to be shown on a map.
+ Lines 295-296: This does not constitute a standalone paragraph. Please revise. Its also not clear what ‘analyse them separately’ refers to.
+ Figure 4 caption: It says ‘parts of the ice sheet’, which is rather vague. Better to say ‘Antarctic ice sheet’ or even ‘ice sheets’. Does this include only grounded ice, or also ice shelves … Please check that all captions have full information.
+ Line 332: I would not use the term ‘observed differences’ as these are not observations.
+ Summary: Mention of ‘Bayesian’ treatment of multi-model ensembles is used in the abstract, yet no where else, including the summary.