General Comment:

This manuscript presents and evaluates a new, high-resolution (~11 km) regional climate model (RCM) dataset developed through the PolarRES project. The ensemble combines four advanced RCMs (HCLIM, MAR, MetUM, and RACMO) to simulate Antarctic weather and climate for 2000–2019, providing a much finer-scale view than reanalyses such as ERA5 and improving the representation of key features like coastal winds, precipitation, and temperature gradients shaped by the continent’s complex terrain.

Methodologically, the paper advances Antarctic climate research by introducing a consistent multi-model ensemble framework that enables systematic comparison among models with differing physics. All simulations are forced with the same ERA5 reanalysis data, isolating differences in model behaviour rather than boundary effects. Validation against the continent-wide AntAWS observational network further strengthens confidence in the ensemble’s ability to reproduce near-surface climate processes at high spatial detail.

The manuscript includes one table summarising mean near-surface climate statistics and eleven main figures, supplemented by additional material in the Supporting Information. This clear and data-rich structure effectively supports the manuscript’s evaluation of the PolarRES ensemble.

Specific comments:

The following feedback is provided to the authors for their consideration to help improve the manuscript, should it be considered for publication:

1. It would be useful for readers if the timespan of the comparison (2000-2019) was noted in the abstract.
2. Unless there is another PolarRES consortium paper published with technical details of each of the RCMs (including their domains) more information about these will need to be provided in Section 2.2. In particular, the domain sizes will be of interest to readers and the extent of the domains surrounding Antarctica.
3. Further information about the AntAWS wind speed data would be of interest. It is noted on L192-194 that a correction was applied to resolve differences in surface elevation and air temperature, but how are observed near-surface wind speeds and the lowest modelled wind speeds reconciled? This is relevant to the wind speed data presented in Figure 2, Table 1 and other figures presenting wind speed information.
4. The comment that the pattern of mean air temperatures “is very similar” on L273 could be reworded, as this is not entirely true for summer in the HCLIM simulations. Is there any explanation for the warm bias in all models in summer, as noted on L284?
5. In relation to surface winds, if point 3 is clearly addressed, then the paragraph starting on L346 might need to be refined, as the height differences between observations and the lowest level of modelled wind may explain some of the differences.
6. It would be of interest to readers if further insights as to why the RCMs are cooler than ERA5 over ocean areas could be provided (L441-442). This cold bias could have important implications for those using the data for ocean-to-land interactions, so any insights about this would be valuable.
7. Is the comment “no single RCM stands out as better or worse overall than the other” on L536 really an accurate reflection of the results presented? The warm and moisture bias in HCLIM would seem like a red flag at this time. An explanation is provided for this in the paragraph starting on L599. In the conclusions, it is indicated that this work has provided insights on the strengths and limitations of the RCMs, and highlighted priority areas for ongoing model development (L649-651). To make this clearer, I wonder whether an additional table in the discussion that shows current significant strengths and weaknesses might be a useful way to provide a summary to readers. If this is problematic, any attempt to strengthen your insights in the discussion and conclusions would be really useful to readers, who will be looking for quick tips or pros and cons about each RCM before using the data.
8. In relation to point 7, the comment that the RCMs all perform “fairly well” could be clearer and a stronger position one way or the other would be useful to readers (L556). In the text that follows, the issue of wind sensor height is raised, which could come earlier in the methodology.
9. Lastly, it would be great if the resolution of the figures could be improved when this work is published. The figures are of a high-quality but the resolution of them is low in the preprint. If the figures are being revisited, any increase in font size of the x and y-axes and their labels to improve readability is normally welcomed by “older” readers.

I have a question about EGUsphere. According to my search online it said: “EGUsphere is not a journal; it is an online platform and preprint repository for Earth, space, and planetary sciences, hosted by the European Geosciences Union (EGU). It serves as a central hub for conference abstracts, presentations, and preprints that can undergo public peer review and be submitted for publication in one of the EGU's 19 open-access journals.” So which open access journal is being targeted for this paper? 

General Comments:

The authors have undertaken a massive task to derive, assemble, and evaluate the PolarRes ensemble. Potentially this data set along with its individual members at 11 km spatial resolution will be valuable for exploring aspects of Antarctica’s weather and climate, especially broadscale surface mass balance. It is unfortunate that the full technical details are not available along with this manuscript. For most applications, regional model simulations specifically designed for the intended application are needed. My comments address the assertion as to whether ensemble actually advances capabilities beyond that provided by ERA5 (31 km horizontal resolution) and by implication the soon to be ERA6 (18 km horizontal resolution). Also, I have many comments regarding the scientific content and the presentation of results. In general, a lot more use of Supplementary material should be made to provide a more refined analysis of the RCM’s performance, even though a follow-on in-depth publication is planned.

The first major issue is HCLIM. The performance of this model is really deficient (e.g., Figs. 8 and 9) and significantly penalizes the ensemble. The explanation for the huge summer warm bias (and moist bias) that is especially prevalent on the major ice shelves is attributed to the deficient albedo treatment. This issue has been documented previously by Xue et al. (2022; https://doi.org/10.1007/s11707-022-0971-8) and the impact of its removal demonstrated. I wonder if this is the complete story. Are you sure that the ice shelves are not treated as sea ice? That can happen because the Antarctic coastline is specified in many terrain data sets at the edge of land ice. This situation would provide a large heat and moisture source for the atmosphere at high latitudes in summer. In winter, the essentially 100% concentration sea ice would act very like land ice. Also, the wind speeds are much slower than the observations and other models throughout the year. Although the authors are reluctant, this model should be removed from the ensemble, or an ensemble with 4 versus 3 members should be provided. On page 27, details on the PolarRes ensemble are missing.  

Use of Victoria Land as a geographic label. It extends from 70.5S to 78S (or the latitude of McMurdo Station) and not farther south along the Transantarctic Mountains (TAM). Refined use of this term is required as it is often used to describe the entire span of the TAM.

The authors need to state the difference between ERA5 and the RCMs employed.  ERA5 is the result of a very short-term forecast with the frequent assimilation of surface pressure observations. The RCMs result from much more extended forecasts that have frequent restarts (MetUM) or are nudged toward ERA5; no direct assimilation of observations occurs. I suspect ERA5 and the RCMs have very similar daily sea ice and SST forcing.

Line 43: “the dearth of observations on this extreme continent”. More observations are needed but the situation has improved with extensive AWS deployments and many satellite data sets. I would also say that limited understanding and deficient model physics are significant contributing factors, partly related to limited relevant process observations.

Line 54: The definitive work on foehn winds on the eastern Ross Ice Shelf was by Zou et al. (2019; doi: 10.1002/qj.34600).

Lines 54-55: “Meanwhile, katabatic winds exert a dominant influence on the climate of coastal Antarctica (Heinemann et al., 2019; 2021; Parish & Bromwich, 2003; Caton Harrison et al., 2024)”. The authors may mean Parish and Bromwich (2007; Mon. Wea. Rev., 135, 1961-1973.) that demonstrates the katabatic (surface) wind impact on the weather and climate of large portions of the middle and high latitudes of the Southern Hemisphere (there are some debates about the katabatic wind phraseology but this is used here generically). Foehn winds are important climatic features but not of the same impact as katabatic winds, lines 49-50.

Line 76: Datta et al (2023) is not in the reference list.

Lines 87-88: Campbell et al. (2024) examined New Zealand and Iles et al. (2020) considered Europe.

RCM Descriptions: Please provide more details as to the nudging used by HCLIM and RACMO as this is very important to keep the model state close to reality. Levels? Frequency?  Variables?  

Lines 183-184: Using daily data obscures the diurnal cycles that are prevalent in many parts of Antarctica during austral summer.

Line 194: Were the wind speeds adjusted to the same height above the surface? Typically, the models produce 10-m wind speeds while the AWS winds are typically at 3 m above the surface.

Figure 1: Your analysis is heavily weighted to the Ross Island region and the Victoria Land coast.

Figure 2: Please try a nonlinear scale for specific humidity and precipitation to provide more details over Antarctica, the focus of interest. The pressure plots are not useful as they are dominated by the effects of terrain height. The standard deviation plots for pressure are meaningful, however.

Table 1: Give precipitation in mm/year, the unit used in the plots.

Line 253: Specific humidity not relative humidity.

Figure 3: Please use a nonlinear scale to provide more details over Antarctica. For context add panels for ERA5, like done for Figs. 5-7. You discuss sea ice in several locations, but nowhere do you display the sea ice extent. Please rectify.

Figure 4: Add ERA5 plots for context. Can the color bar be modified to provide more discrimination for larger values?

Figure 5: For pressure some stations have unexpected large biases especially below the 1:1 line, around 10 hPa. These are probably station elevation errors. Please confirm.

Figure 7: This needs to be a full-page figure to see the details.

Discussion of the biases in Figs. 6 and 7 on pages 16 and 17. Please adjust the language to make it clear you are discussing absolute biases.

Figure 8: This is a nice figure. Please add the comparison for the ensemble mean. Correct line 423 to “(huss, g/kg)”.

Figures 9 and 10: Nice figures and easily interpretable. Because you are emphasizing the ensemble mean this needs to be added.

Figure 11: Please add corresponding figures to the Supplementary material stratifying by elevation.

Lines 503-504: “The lowest inter-quartile range is found in near surface air temperature and surface pressure in all seasons, particularly JJA.” I don’t think you can compare between variables because this depends on the units considered.

Lines 507-508: “The RCMs shown here offer an improvement over ERA5 due to their higher resolution and more sophisticated physical parameterizations.” The last statement cannot be claimed for HCLIM and is only somewhat true for RACMO that depends on parameterizations from ECMWF.

Lines 536-537: “No single RCM stands out as better or worse overall than any other”. HCLIM is clearly deficient in relation to the other 3 models. Line 602 says that” HCLIM has not been used extensively in Antarctica before”. HCLIM is not in the same class as the other 3 models that have been extensively applied in Antarctica, but with some effort it can achieve a similar level of performance.

Lines 594-597: “Some RCMs may be simulating more katabatic outflow (and hence colder temperatures) near this steep terrain (e.g. the MetUM), whereas the other RCMs may be simulating more compressive warming such as foehn winds, or adiabatic compression of katabatic flow.” I did not understand the basis for this statement.