Title:    Variability in Antarctic Surface Climatology Across Regional Climate Models and Reanalysis Datasets

Authors:    Carter, Jeremy ; Leeson, Amber; Orr, Andrew; Kittel, Christoph; van Wessem, Melchior

General comments

This paper evaluates the variability in diagnostics critical to ice sheet mass balance from simulations over Antarctica in three state-of-the-art regional climate models and two reanalysis products. This is a substantial piece of work that furthers our understanding of uncertainty in the atmospheric component of coupled climate models over Antarctica and informs the configuration of such models. The structure of the paper is well thought out, and the quality of the writing and figures are excellent. I recommend that the paper is accepted for publication following minor revisions addressing my comments, below.

From your plots, it looks to me as though model differences are generally greater over steep, mountainous terrain (e.g. along the coastline and over the Transantarctic mountains). This highlights differences in the representation of orography and meteorological conditions over orography. Previous studies have demonstrated that a major source of such differences is model resolution, e.g. there are significant differences in the reproduction of influential mountains winds between simulations with grid spacings of 12-km and km-scale (e.g. Orr et al., 2014, 10.1002/qj.2296 for katabatic winds in E Antarctica; Heinemann et al., 2021, 10.3390/atmos12121635 for foehn winds over the Antarctic Peninsula). These differences are particularly pertinent for surface mass balance over ice shelves, since the Antarctic coastline is generally found at the foot of the steep slopes of the Antarctic plateau, and/or is in the vicinity of mountainous terrain. So I’m intrigued as to whether in your results you can see larger systematic differences in the vicinity of steep terrain that can be attributed to resolution (using the two MetUM models), and whether these differences are of sufficient magnitude and spatial scale to be pertinent for ice shelf mass balance.  Also, as alluded to above, it could be that even the 12 km MetUM simulations are insufficient to reproduce climatically important influences of terrain-induced airflows. You do mention katabatic winds as a potential source of model differences, specifically on the Amery IS, but I wonder if it might be worth commenting further on the influence of orography on model differences and the implications of this.

Specific comments

Line 8: “suggested” here is too weak. Suggest instead “Our results imply that…”

Line 26: “The primary method of ice shelf retreat is through oceanic basal melting”. I think this statement requires further qualification. Specifically, adding the word "currently", and something like "with the notable exception of some of the ice shelves on the Antarctic Peninsula (Pritchard et al., 2012)". Recent climate/ice-sheet modelling studies indicate that atmosphere-driven hydrofracture has in the distant past been, and will in the future be, the principal cause of Antarctic ice-shelf collapse (e.g. DeConto et al., 2021, 10.1038/s41586-021-03427-0; Pollard et al., 2015, 10.1016/j.epsl.2014.12.035).

Figure 3c: Why are the interiors of the Ross and Filchner-Ronne ice shelves masked out?

Figures 4-6: From neither the text nor the figures is it totally clear to me whether the difference is model - ensemble, or ensemble - model. I'd expect it to be the former, and that is indeed my impression from the text. However, "different to ensemble average" implies to me the opposite. Please make this clear, in the text where these figures are first referenced, and in the figure captions.

Line 446: “The primary sources of large-scale, systematic differences between the simulations, for all variables and components, are identified as deriving from differences in: the model dynamical core; the surface scheme; parametrisation and tuning.” In the discussion, the sensitivity of melt to the subsurface scheme is highlighted, and justification is given for the "secondary" importance of the factors listed in the subsequent sentence (driving data, resolution, domains etc.). We may then assume by way of elimination that the model dynamics and physics are the primary sources of systematic differences. I'm not sure though that this reasoning is actually stated, and I think it should be, somewhere in the Discussion section.

Line 456: “Therefore, as concluded in Mottram et al. (2021), there is an importance on observational campaigns to correct for biases.” Do you mean there is demand for new field observations with which to constrain model physics parameterisations? Or for (post-processing) model bias correction? This statement needs expanding on.

Line 458: “2100 SLR” A bit specific. Suggest simply "future SLR". The same applies in abstract, line 3.

Technical corrections / suggestions

Line 5: Suggest italicising “Seasonal and trend decomposition using Loess”, and also perhaps capitalising the T in trend, to make it clear that this is what STL stands for.

Line 168: Suggest italicising “Seasonal and trend decomposition using Loess”

Line 201: RMSD: This abbreviation is defined in the abstract, but should be defined when first used in the main text also

Line 220 and other instances: “mmWEqm^-1”... I think there should be spaces between the units, so perhaps "mm WEq m^-1". Or “mm w.e. m^-1” as I've seen this notation used before.

Line 299: Remove repeated number 1 before “mm”

Carter et al. presents a comparison of key variables (Snowfall, temperature, melt) for a set models including MetUM, MAR and RACMO as well as reanalysis (ERAI and ERA5). They use statistical methods which compare with regard to mean, trend, seasonal cycle and the residual (which should capture physical differences). They examine the potential impact of several differences in the models, including the different use of DEMs, forcing at the boundaries, the size of the boundary itself and differences in the underlying physics, concluding largely that the major differences occur in the differences in the physics.

In general, figures are clear and helpful, the analysis is sound and this will be a valuable addition to the canon of model intercomparisons (especially with regard to how methods were applied systematically here). The main critiques listed below are about presentation, language, or which parts of the discussion should be highlighted:

Main Critique:

As surface melt occurs primarily during the summer, any differences in the seasonal cycle require a bit more explanation. For example, in Figure 6, the differences in the seasonal standard deviatil for MAR are quite large over East Antarctic ice shelves. What does this mean? Does this mean that seasons are shorter in MAR? It is also worth highlighting immediately that (here in plain language): the only real estimates of surface melt that matter here are from MAR and RACMO. The simpler surface schemes in ERAI/ERA5/MetUM simply don’t capture these processes. I note, however, that the authors have provided a thorough explanation of the underlying physics (and similarly did a great job explaining Agosta’s work on how the difference in how precipitation is treated in MAR and may lead to more precip in the interior). It is just worth noting these basic differences at the start (potentially within table 1)

Minor revisions:

Table 1: Include a column for time-period at which forced at the boundaries. This is mentioned in the text, but is a very important difference.

Potentially my misunderstanding, but how does the value and the magnitude of correlation differ from one another?

Fig. 3: Use different color scales for (a) and (b) to make explicit that they are different scales (alternatively, just add a note in the caption). This is stated in text, but easy to misread.

Line 205 : The impact of systematic differences in snowfall/snow melt on estimates…

Line 208:  For near-surface air temperature, differences…

Lne 245: …each component of the timeseries and that for temperature and melt …

Line 258-263: This is partially mentioned in methods, and should perhaps just be moved there

Line 264: “significant systematic” implies a more quantitative assumption. Perhaps use “substantial” or “spatially-coherent”

Line 268: Possibly worth mentioning the length scale in the Antarctic Peninsula as well.

Line 270: mean of the time series is highly correlated :  as “correlated” is a specific quantitative term here, I think that “has a similar spatial signature” might be more accurate

Line 281: Similarly (to above) I wouldn’t say “weak, negative correlation”, but rather something like “contrasting spatial patterns” and specify where.

Line 284/ Figure 5: Point to location of “example grid-cell” in the figure

Line 294: How meaningful is the seasonal standard deviation if the majority of melt happens in summer? (What does this mean physically? Does this imply that the seasons shift?)

Line 295: The physical meaning of this metric isn’t entirely clear to me: I understand this as a measure of the bias correcting for the effect of season and residual, but it might be helpful to make this explicit and expand a little more

Line 298: The masking for 1mmW Eq was mentioned in Methods, I don’t think it needs to be mentioned again.

Line 310: Again, I’m having a hard time parsing what “adjusting for equal means and seasonal/residual standard deviations” means here. I think it’s worth explaining a bit more what this metric means physically (e.g. when correcting for the seasonal effect, the mean trend, the residual is a metric of the physics)

Liene 321: The large-scale differences in snowfall (ERAI vs ERA5)   are attributed here to model physics, but we’re also seeing very large differences in the DEMs. Similarly, Met models show quite similar DEMs  (Figure C1)

Line 342-344. This sentence is a little long and could be broken up.

Line 354: It is found that for the MeUM(044) run, the buffer…

Line 355: to the buffer zone boundary, and that even…

Line 386 – 396: This paragraph is a little convoluted generally.

Line 418: this seems to be reversed.(positive over ocean, negative over large region of East Antarctica near the Transantarctic mts.

Line 424: The use of the term “systematic differences” needs to be more clearly differentiated from “correlation” here. I think what you mean is that “this correlation occurs over large portions of the ice sheet”

In general, I think this paragraph needs a summary sentence for melt, e.g. that the melt bias is consistent, and wide-spread, even accounting for the seasonal and trend components

443: Despite this, there exists (add comma)

Sincerely,

R. Tri Datta