In this paper, the authors employ the integrated Atmospheric Model Across Scales (iAMAS) with four unstructured mesh configurations (resolutions of 120 km, 60 km, 16 km, and 4 km) to simulate Antarctic meteorological conditions and compare the results against both ERA5 reanalysis data and in situ measurements (AWS and radiosondes). Their findings highlight that unstructured meshes avoid the “polar singularity” problem seen in ERA5 at the South Pole, where ERA5 substantially underestimates wind speeds. In relatively flat Antarctic regions, iAMAS at coarse resolutions (e.g., 60 km) already achieves performance comparable to—or better than—ERA5, suggesting that high-resolution meshes offer minimal additional benefits there. Conversely, over complex terrain such as the Transantarctic Mountains, higher-resolution grids (around 4 km) significantly improve the simulation of temperature, pressure, specific humidity, and wind speed, reducing the systematic cold and wind biases found in coarser configurations. Overall, this study underscores the advantages of unstructured meshes for polar modeling, particularly in resolving complex topography, while also showing that moderate resolutions may suffice in more uniform terrain.

The paper is well written and organized. It would be a contribution to the journal GMD. However, there are some minor issues where further explanation is needed to improve the manuscript’s quality. I suggest minor revisions to the manuscript before it is accepted for publication.

Major comments

1. One of my main concerns is that the authors did not show any comparison regarding snowfall estimation. As mentioned in the introduction, ice sheet simulation is why accurate polar atmospheric simulations are needed. Then it is necessary to understand how different model configurations affect the amount and distribution of snow. If precipitation observation is available, the authors should compare their simulation results with the observation. Even if the observation data do not have a precipitation record, analyzing simulation results and how different resolutions affect the precipitation pattern and intensity would still be helpful.
2. The second point I think the author should elaborate on is the stratospheric wind bias (Fig. 8). I don’t feel the authors’ explanation is convincing enough. ERA5 clearly does not have similar issues. Is the cause the relatively low model top? It appears to me that a high-top model like WACCM is necessary if we want to get the stratosphere simulation correct. I hope the authors can provide more explanations and references in this section.

Minor comments:

1. Line 45: It is better to mention cubed sphere grid as an alternative option. Are there polar simulation studies using the cubed sphere (like FV3)?
2. Line 97-98: Why is the center of the 4-km grid mesh not centered at the south pole? What is the justification? What is unique about 80 degree S and 160 degree E?
3. Line 192-194: It is good that iAMAS exhibits more accurate temperature result. However, what is the physical reason for the better performance? The author should give some explanation, or at least speculation here.
4. Line 244: What does AMPS stand for? It should be defined before the first use.
5. Line 245: Interesting point. What is the potential effect of this missing process on temperature bias in Noah MP?
6. Line 300: the unit is probably wrong here. Should it be K/km instead of K/m?

General Comments:

This is an interesting manuscript that investigates the performance of the MPAS model (see below) over Antarctica in relation to in-situ observations and the ERA5 global reanalysis. The sensitivity of the model results to the model grid spacing is emphasized, and the results are in line with expectations. It is to my knowledge the first manuscript describing the performance of MPAS over Antarctica in the refereed literature. However, there are many aspects to the manuscript that need major improvement to reflect reality.  

First, the iAMAS model that is run here without the chemistry component activated is actually the MPAS model available from NCAR. After looking at Gu et al. (2022), the changes (apart from the atmospheric chemistry component) are focused on efficient execution on an HPC platform. The atmospheric physics options used here are those available with MPAS (https://mpas-dev.github.io/atmosphere/atmosphere.html) that originate from various versions of WRF.

Second, the manuscript needs to consider the Antarctic Mesoscale Prediction System (AMPS) much more carefully than is done. AMPS has been run routinely for many years, starting with MM5 before moving on to the current WRF model. Here is the best reference: doi: 10.1175/BAMS-D-11-00186.1. It is run on a polar stereographic projection, i.e., no pole point problem, with the finest grid around Ross Island currently being 0.89 km grid spacing. In addition, AMPS has been running MPAS for a number of years, although this is not well known and unfortunately there is no peer reviewed manuscript. However, the following conference presentation should be referred to and discussed to provide a quantitative context for your results. https://amrc.ssec.wisc.edu/presentations_2017/Day1/powers_mpas.pdf. More recent summaries of the status of MPAS in AMPS are provided here: https://polarmet.osu.edu/WAMC_2024/pdf/WAMC_2.08.pdf and https://polarmet.osu.edu/WAMC_2024/pdf/WAMC_2.09.pdf

Third, the polar version of WRF (PWRF) has been extensively used over and validated for Antarctica. The most recent peer reviewed publication is https://doi.org/10.1007/s11707-022-0971-8 that considers a full annual cycle and considers both surface and upper air observations. PWRF with a few modifications is used in AMPS. The extensive PWRF literature provides guidance on the Antarctic performance of available physics options in WRF. You should quantitatively contrast your results with those presented by Xue et al. in the above reference.

Fourth, a lot is known about ERA5 performance in the Antarctic, yet none is referenced here.  Here is a good reference to consult and incorporate into this manuscript. https://doi.org/10.1175/JCLI-D-19-0030.1

Fifth, we checked one day of data for ERA5 for the average wind speed at 89.5, 89.75, and 90°S (from the square root (u² + v²)) from 600 hPa and above. The first two values were close in magnitude (3-12 m/s) while the value at the pole point was nearly zero. We also checked at the North Pole and found the same problem. Thus, the pole problem in ERA5 for wind speed reported here is supported.

Specific Comments:

1. Line 155: The scheduled frequency of radiosonde launches at McMurdo and South Pole varies during the year with generally 1 per day during the colder months.
2. Lines 169-171: What roughness length did you use in the logarithmic wind profile to adjust the 10 m wind speeds down to 3 m?
3. Figure 3 for January: It looks like there is a problem with the snow albedo used in iAMAS for interior Antarctica. It should be no smaller than 0.8. Otherwise, large warm biases in the 2-m temperature will be simulated. See Xue et al. (2022).
4. Lines 188-190: This discussion refers to the ERA-Interim reanalysis and not ERA5.
5. Table 3: The 2-m air temperature results one gets depends a lot on the land surface model. How did you initialize the Noah LSM?
6. Table 4: It is very challenging to measure atmospheric moisture content at low air temperatures because of the tiny amounts of water vapor present.
7. Line 260 “enhancing the barrier effect of air flowing over it”. This language requires change because the barrier effect you are talking about refers to the low-level blocking/barrier effect on air encountering an obstacle causing some or all of the air to blow around rather than over the obstacle.
8. Line 261: O’Connor. Fix also in the references.
9. For your comparisons, you should add the numbers of observing stations involved in the comparisons.
10. Figure 8 for October: The model cold biases above ~ 15 km are probably related to the stratospheric ozone treatment in MPAS. ERA5 puts a lot of effort into correctly handling the stratospheric ozone concentration.
11. Section 3.2.2: Comparisons such as these usually use geopotential height as this variable on constant pressure surfaces is used for atmospheric analysis.
12. Figures 10 and 12 for October: If your temperature field is in error in the stratosphere then so will the specific humidity and wind fields.
13. Line 413: Many/most regional models for Antarctica use a polar stereographic projection not a latitude-longitude grid because of the singularity at the Pole.
14. References: Golledge et al. (2015), Kannemadugu et al. (2023), Kessenich et al. (2023), and Zhang et al. (2023) contain repetitious information.

The issue of the low wind speeds at South Pole as reported in this manuscript has been known at ECMWF since January 2019 

and is posted on this ECMWF discussion page for ERA5:

https://confluence.ecmwf.int/pages/viewpage.action?pageId=129134800

The problem occurs for both the north and south poles. ECMWF Advice: "please use winds at neighbouring locations at 89.75 (N/S) and

89.5 (N/S)  for Reanalysis and Ensemble members winds, respectively."