# General comments

This paper address a relevant scientific questions within the scope of TC: the impact of cloud radiation on snow melt in Antarctica and its representation in atmospheric models, which is essential for improving melt projections on the ice sheet.

The paper is generally well written and easy to follow. It combines modelling, satellite observations and surface AWS observations. I recognise the work done by the authors and the relevance of this study. Despite this, I have some major concerns about the conclusions of the article.

The authors focus on a melt event on 17th of January 2016 when western Ross Ice Shef (RIS) remained melt-free according to passive microwave melt extent, while two regional atmospheric model simulate melt in this sector.

My concerns are the following:

• The two major observational datasets used in the article are contradictory with respect to melt in western RIS : passive microwave melt extent shows no melt in western RIS during the event, whereas CERES radiative fluxes would lead to more melt than in models, as indicated by the authors in Section 4. Indeed, CERES shows larger net surface radiative fluxes than in models (Fig5a), consistent with larger liquid water path in CERES than in MetUM in western RIS during the event (Fig 11 and 12).
• The manuscript shows that models simulate too few liquid clouds with respect to ice clouds, using both CERES and CALIPSO satellite products. This means that correcting the ice-to-liquid mass partitioning would induce more downward longwave radiation toward the surface, hence more melt. As the authors rely on the passive microwave melt extent rather than on CERES (point above), this result contradicts the statement in the introduction that “The models (…) seemingly struggle to correctly represent the ice-to-liquid mass partitioning associated with the cloudy conditions, which we suggest is responsible for the radiative flux errors”. This statement is repeated in the discussion, where a large emphasis is given on liquid-ice cloud partitioning. But in the discussion, the authors also state that fixing this partitioning would actually increase melting, in contradiction with passive microwave observations: “However, the larger amounts of liquid-water clouds observed by CALIPSO and CERES would be expected to produce even larger downwelling surface LW fluxes (Zhang et al., 1996). This is not the case, suggesting that other factors influencing the LW radiative effect of the clouds, such as cloud temperature, altitude, and cloud microphysical properties like the size of water droplets or ice crystals, may be impacting surface LW fluxes.”. So why this emphasis on ice-liquid partitioning, which contradicts the results of the study?
• From the two points above, I think that a conclusion of this study is that either (1) properties other than ice-to-liquid partitioning of cloud water influence the radiative effects of the clouds, or that (2) melt extent from passive microwave might be wrong. Can the authors clarify these points?

More details are given bellow.

# Specific comments

1 Introduction

“Therefore, to realistically capture local climate variability and simulate ice shelf melt patterns, it is essential to utilize regional atmospheric models at high spatial resolution, i.e., grid box sizes of the order 10 km or less.”

• In case of synoptic-scale events, 10 km resolution might not be needed over large ice shelves. It depend on the ice shelves ?
• Which of the numerous papers cited L25-35 does use “10km or less” resolution ?

“Here we investigate the benefits of applying the sophisticated offline coupled firn model described by Langen et al. (2017) that represents key aspects such as the melt-albedo feedback to improve regional atmospheric model simulations of a prolonged and extensive episode of surface melt that occurred during January 2016 over the Ross Ice Shelf (RIS), West Antarctica. The RIS frequently experiences major surface melt events due to both synoptic- and local-scale processes (Nicolas et al., 2017; Zou et al., 2021; Li et al., 2023; Orr et al., 2023), with this particular event attributed to an influx of warm and moist marine air, likely linked to a concurrent strong El Niño episode (Nicolas et al., 2017). The regional atmospheric model simulations examined were initially produced for Antarctic CORDEX (Antarctic COordinated Regional Downscaling EXperiment), and are based on HIRHAM version 5 (HIRHAM5) and MetUM version 11.1 (Orr et al., 2023). In these simulations, HIRHAM5 employed a relatively sophisticated multi-layer snow scheme (Langen et al., 2015), while the MetUM utilized a simple composite snow/soil layer (Best et al., 2011).”

• This paragraph should be moved to the method section. It could be replaced by a final paragraph in the introduction presenting the outline of the article, with much less detail on the models as they will be presented in the Method section.

2 Methods and materials

This section should be divided in (at least) 2 subsections : Observations and Models.

“This consists of 6-hourly averaged values of solid precipitation, liquid precipitation, surface evaporation, surface sublimation, surface downwelling SW radiative flux, surface downwelling LW radiative flux, sensible heat flux, and latent heat flux”

• Why this models need surface evaporation, surface sublimation and latent heat flux as input?

“These are compared with daily melt extent estimates from satellite passive microwave measurements at a grid spacing of 25 km (Picard et al., 2007; Nicolas et al., 2017), using the same melt threshold of 3 mm.”

• I don’t find a 3 mm threshold in Picard et al., 2007 nor in Nicolas et al., 2017. Can you justify the choice of this threshold?

4 Cloud radiative effects

Comparison with CERES

“This raises concerns over the reliability of these measurements, as this would also presumably be associated with (erroneous) melt over the western RIS region, i.e., contradicting the satellite passive microwave measurements of daily melt extent (Figs. 2 and 3).” And to the end of the section, including Fig 5, 6, 7 and 8

• You state that CERES might give erroneous radiative budget at the surface (Fig 5 + sentence above), so we are not sure if we can trust maps from CERES or not (Fig 6, 7, 8). Consequently, what is the objective of this full section?

“Figure 7 also shows that CERES semi-captures the transition from large negative net surface LW values over the western RIS during nighttime on the 14th to smaller negative values on the 17th, in agreement with the models.”

• Do you use CERES to evaluate the models, or do you use the models to validate CERES? With this formulation, it seems that you use the models to validate CERES, which is confusing as the initial objective was to evaluate the models. Can you clarify?

5 Cloud properties and 6 Discussion

From Section 5 and 6, I conclude that partitioning between liquid and ice cannot be the reason for the supposed too high melt in models versus passive microwave melt extent:

• Section 5 : Liquid cloud are observed on western RIS during the event “More noteworthy is that CALIPSO shows liquid-water and ice-water clouds extending up to 7 km above the surface in the same region on the 17th of January (Fig. 10) coincident with the (erroneous) spike in modelled melt.” “the same region” being western RIS according to Fig. 10 legend.
• Section 5 : MetUM models much less LWP that CERES and CALIPSO in western RIS during the event “However, CERES suggests that clouds with high liquid-water content and ice-water content occur at 12 UTC on the 17th over this region, with values of cloud ice water path up to 0.5 kg m-2 (i.e., similar to the MetUM) and cloud liquid water path up to 1 kg m-2 (i.e., two orders of magnitude larger than the MetUM). Moreover, it’s noteworthy that CALIPSO also observed liquid-water and ice-water clouds over the western region of the RIS (Fig. 10), which substantiates the CERES results.”
• Section 6 : Partitioning between liquid and ice cannot be the reason for discrepancies in LWD. “However, the larger amounts of liquid-water clouds observed by CALIPSO and CERES would be expected to produce even larger downwelling surface LW fluxes (Zhang et al., 1996). This is not the case, suggesting that other factors influencing the LW radiative effect of the clouds, such as cloud temperature, altitude, and cloud microphysical properties like the size of water droplets or ice crystals, may be impacting surface LW fluxes.” “In reality, multiple possible cloud properties (in addition to ice-to-liquid partitioning of cloud water) could be influencing the radiative effects of the clouds to produce smaller downwelling LW fluxes than are being simulated.”

• From this, I think that a conclusion of this study is that either (1) properties other than ice-to-liquid partitioning of cloud water influence the radiative effects of the clouds, or that (2) melt extent from passive microwave might be wrong.

“as would repeating the MetUM simulations using its recently developed double-moment microphysics scheme to examine whether this increased the amount of liquid- water cloud and limited its conversion to ice (Field et al., 2023).”

“Previous studies have already shown that the MetUM has deficiencies in its representation of cloud phase, particularly re- lated to it simulating Antarctic clouds that contain too much ice-water content and not enough liquid-water content (Abel et al., 2017).”

• Here, more liquid cloud would induce more melt, so more model bias compared to passive microwave melt extent. Can you clarify what you expect to improve by increasing the liquid water content?

# Technical corrections

The number of references L25-35 is too large (25 references)

Figure 1 : Orgraphy

Figure 11 and Figure 12 : Use a continuous colormap instead of the divergent Blue/Red colormap curenlty used.

Review of tc-2023-145, “The importance of cloud phase when assessing surface melting in an offline coupled firn model over Ross Ice Shelf, West Antarctica” by N. Hansen et al.

This manuscript describes a useful and timely study of surface melt during a major surface warming event of January 2016 that affected much of West Antarctica and the Ross Ice Shelf (RIS). Two regional atmospheric models, each containing single-moment cloud microphysical schemes, are applied first with their native snow surface parameterizations and then with a more sophisticated firn model. With their native schemes, the models have trouble simulating enough surface melt as compared with satellite passive microwave maps of melt extent and number of melt days. When the models are used as energy input to the more sophisticated firn model, substantial improvements are noted where the satellite passive microwave data show extensive melt, but the models then go on to overestimate the melt on the western side of the RIS. This is attributed to deficiencies in simulation of cloud properties, in particular a significant overestimate of cloud ice water path at the expense of cloud liquid water.

The manuscript is well written and well organized, and for the most part I could easily follow the arguments. The advance of this study lies in the effective demonstration of the more advanced firn model. It is useful to see how detailed representation of cryosphere surface properties is required for the most realistic simulation of Antarctic surface melt. On the other hand, single-moment cloud microphysical schemes have been shown many times to be inadequate over polar regions. Their deficiencies over Antarctica tend to be basically as described in this paper: an overestimate of cloud ice water. It would be interesting to see what happens if one of the models is fitted with a double-moment scheme (as in Hines et al. 2019), but this might be beyond the scope of the authors’ resources.

Although the fundamental atmospheric result is not entirely new, it is useful to see that applying the most sophisticated firn model does not compensate for the radiative errors related to cloud phase. The manuscript needs some additional detail and clarification regarding some of the satellite remote sensing products.

1. It is not stated what CERES product is used. There are several available from NASA Langley Research Center (LaRC) and other NASA facilities. CERES does not measure surface radiative fluxes. It only measures top-of-atmosphere radiances over broad spectral intervals, and these are then combined with angular dependence models to get top-of-atmosphere fluxes, and then these fluxes are combined with various other satellite data sets and/or radiative transfer models to get estimates of the surface radiation components. Given that CERES is showing such great discrepancy here, it is important to identify which CERES product has been used and discuss potential sources of error with reference to the underlying algorithms (most of which are published by NASA LaRC in the open literature).
2. Similarly, CALIPSO provides excellent active-sensor detection of cloud vertical extent, but the phase partitioning algorithms have some built-in assumptions and temperature thresholds. The manuscript should give a brief discussion of how the CALIPSO algorithm might lead to uncertainties in what is presented in Figure 10, with the specific vertical temperature profiles over the study domain.
3. The vertical temperature profiles and the vertical profiles of the simulated cloud properties should also be presented and discussed for the two days (14 and 17 January) and the relevant locations. This would make the discussion section (around lines 285-299) less qualitative and speculative. For example, if the temperatures in the lower temperature are only slightly below freezing over several km, then extensive ice phase cloud is obviously ridiculous as we expect supercooled liquid water in these pristine conditions.
4. Regarding the deficiencies of single-moment cloud microphysics, RACMO simulations of West Antarctic surface melt (e.g., see papers by Jan Lenaerts) have “tuned” the microphysical scheme to give high enough cloud liquid water, yielding good geographic representations of surface melt. This should be mentioned somewhere in the later sections of this paper.
5. Why is ERA-Interim reanalysis used to initialize the regional models and not the more current ERA5?