The contemporary climate of the Antarctic ice sheet (AIS) has been relatively stable, but recently the ice sheet has started losing mass at an accelerated pace . As the AIS contains enough water to raise global mean sea level by 58 m , it is imperative to understand the driving mechanisms behind recent mass loss. Present-day AIS mass loss has been ascribed to the thinning and breakup of ice shelves, the floating extensions of the ice sheet, due to warming of both ocean and atmosphere . Several Antarctic heat records have been broken in the past decade , with an all-time record for continental Antarctica of 18.4 ∘C observed at the tip of the Antarctic Peninsula (AP) in February 2020 . These higher temperatures have led to increased surface melt and the formation of melt ponds on the flat ice shelves, enabling the collapse of the Larsen A and Larsen B ice shelves in 1995 and 2002. More ice shelves are susceptible to collapse if warming continues , leading to further AIS mass loss, emphasizing the necessity to fully understand the sensitivity of Antarctic ice shelves to surface melt.

The specific surface mass balance (SMB) of a glacier surface, which is the difference between local accumulation, i.e., mass gain by snowfall, riming and drifting snow accumulation, and ablation, i.e., mass loss by runoff, sublimation and drifting snow erosion, is positive for virtually the entire AIS and only becomes negative in blue ice areas, where sublimation and erosion exceed snow accumulation . The accumulation rate is, however, also spatially variable and is measured to be as high as 3 m water equivalent (w.e.) per year in the western AP , while snowfall can be as low as 8 cm yr-1 in the interior of the East Antarctic ice sheet (EAIS) . For most regions, precipitation dominates the temporal and spatial variability in the SMB signal. Despite low average temperatures , significant melt occurs on ice shelves in East Antarctica and the AP . This melt is 1 to several orders of magnitude smaller than observed in the western ablation zone of the Greenland ice sheet , and almost all meltwater refreezes in the snowpack, or, rarely, is stored englacially . Consequently, almost no runoff occurs.

Refreezing of meltwater changes the snow structure, as it increases snow grain size. Through large grains, light has to travel a greater distance before it can scatter off a surface, increasing the chance of absorption, thus reducing surface albedo (shortwave reflectivity) . This explains the strong snowmelt–albedo feedback, as a lower albedo induces more snowmelt. shows that melt would be 3 times smaller on an ice shelf in Dronning Maud Land (DML) in East Antarctica without the snowmelt–albedo feedback. Snow grains also increase in size by dry snow metamorphism , the rate of which increases with temperature. Increasing snow temperature thus means that fresh snow with small grains changes more rapidly into snow with coarser grains, lowering the albedo. With a lower albedo, more energy is absorbed, leading to higher temperatures, therefore representing a positive feedback: the dry snow metamorphism–albedo feedback . Radiation penetration leading to subsurface heating accelerates this process, as subsurface snow is heated more efficiently. The temperature of and melt in the (sub)surface snow of the AIS is thus sensitive to snow grain conditions and subsurface heating. This sensitivity can be investigated locally by using in situ observations, but a polar regional climate model is required to study it for the entire ice sheet.

In this study, we use the polar (p) version of the Regional Atmospheric Climate Model (RACMO) to analyze the impact of a spectral snow albedo scheme on the (sub)surface temperature and melt of the AIS. The polar version of RACMO has been especially adapted to model glaciated areas and has previously been used to investigate the snowmelt–albedo feedback . Here, we use the latest version, RACMO2.3p3, henceforth Rp3, which has a spectral snow and ice albedo scheme that includes radiation penetration, allowing for subsurface heating and subsurface melt. We evaluate Rp3 with in situ and remote sensing observations, as well as with the previous version, RACMO2.3p2, henceforth Rp2, between 1979 and 2018. To investigate the sensitivity of the AIS to (sub)surface heating and snow conditions, we conduct several sensitivity experiments with Rp3 by incrementally changing one parameter at a time to assess the impact on melt and temperature.

In this paper, we first discuss Rp2, Rp3 and the sensitivity experiments in more detail in Sect. . We also expand upon the concept of SMB and introduce the surface energy balance (SEB) and the observational data sets. Next, results are presented, starting with near-surface and subsurface temperature in Sect. , followed by the evaluation of the specific surface area of snow, defined as the total surface area per kilogram, in Sect. , SEB and albedo in Sect. , and SMB in Sect. , with a detailed discussion about melt. The results are summarized and conclusions are drawn in Sect. .

Figure  shows the yearly-averaged T2m difference for GS and CON with Rp2. Considerably higher temperatures are simulated in GS, with some areas more than 2.0 ∘C warmer with respect to Rp2. The temperature in CON (Fig. b) is on average only 0.1 to 0.3 ∘C lower than Rp2. In summer (Fig. ), the signal of Fig.  is amplified. A comparison with observations in DML during summer (Table , Fig. ), which is the season where any changes in the albedo have the strongest impact on the SEB, shows that the temperature of Rp2 is modeled well, with a small bias of -0.3 ∘C and a root-mean-square error (RMSE) of 1.4 ∘C. The bias of GS and CON is larger: 2.0 and -0.8 ∘C respectively. For more inland stations like station 8, 9 and 12, the bias of GS is larger compared to stations close the edge of the ice sheet, while the bias of Rp2 and CON is smaller. This illustrates the high sensitivity to the implemented changes on the T2m for the AIS in Rp3. The new snow albedo and radiative transfer scheme results in a lower albedo, which is especially important during summer and will be discussed in more detail in Sect. . Including radiation penetration leads to higher subsurface snow temperatures, enhancing snow metamorphism and subsequently enhancing radiation absorption. Due to this positive feedback, inaccuracies in the modeled (sub)surface snow metamorphism are amplified in Rp3.

To investigate the exact cause of deviating temperatures, we show the yearly-averaged T2m difference with Rp2 for all sensitivity experiments for 1985–1990 (Fig. ). As have shown that Rp2 models T2m fairly well, it is therefore used as a benchmark. Similar to Fig. , the temperature of GS is overestimated significantly. All subsequently implemented changes lower the temperature, although some changes impact it more than others. A significant lowering of the temperature is induced by the increase in the fresh snow SSA to 100 m2 kg-1 in the FSG experiment (Fig. b). FSG also uses a different fresh snow regime in the grain growth parameterization (Sect. , Fig. ), and grains with a high SSA consequently remain at the surface for longer. The temperature, however, is still too high.

The strongest temperature lowering occurs when we further reduce the fresh dry snow metamorphism (Fig. c) by implementing a tuning parameter (Eq. ). As Fig.  illustrates, this tuning reduces in particular the snow metamorphism for small grains, i.e., up to 100 times slower metamorphism in FSM than FSG. This tuning makes that surface layers with a high SSA (>50 m2 kg-1) are more persistent between snow deposition events, consequently lowering the surface temperature and hence, through turbulent and longwave exchange between the surface and near-surface atmosphere, reducing T2m. The significant temperature differences between Fig. a and c show how sensitive Rp3 is to grain size and underline the importance of an accurate snow metamorphism scheme.

Higher temperatures are relatively persistent on some of the ice shelves (Fig. c), especially in DML. These regions are characterized by melt in summer that refreezes in the snowpack. As meltwater refreezes, it increases snow grain size, resulting in more solar radiation absorption and therefore higher temperatures. Reducing the refreezing snow grain size consequently reduces the temperature difference on relatively dry locations with melt (Fig. d). Increasing the SLED further lowers the temperature as subsurface heating is reduced (Fig. e). The temperature in CON is now somewhat too low during summer (Table ). This bias can be further reduced by slightly changing α in Eq. ().

In the previous section, we illustrated the importance of grain size on the temperature of the AIS. Compared to in situ observations at Dome C , the SSA of the upper 2 cm in the CON simulation follows the yearly cycle well (Fig. ). The SSA drops gradually over time during spring and summer to values around 40 m2 kg-1, which is somewhat higher than observed. In GS, the SSA is too low as it drops below 20 m2 kg-1. The SSA decline during spring is delayed by a few weeks, but the rate of change is similar to observations. After summer, the SSA gradually increases with deposition of fresh snow but only reaches 40 to 50 m2 kg-1 for GS, significantly below observations. For CON, the SSA gradually increases to 80 to 90 m2 kg-1, which is in agreement with observations. Note that the average SSA of the upper 2 cm never reaches the prescribed fresh snow SSA of 100 m2 kg-1, as large snowfall events at this polar desert site are rare . To summarize, the GS settings lead to unrealistically low SSAs. The CON settings somewhat underestimate snow metamorphism, leading to higher SSA during summer, but this can be fine-tuned using α in Eq. (). Increasing α results in an SSA evolution, depending of the choice of α, to be between FSG (which uses α=1) and FSM (which uses α=0.25) in Fig. .

Table  shows the statistics of SEB components compared to AWS observations in summer from DML in Rp2, GS and CON. All fluxes toward the surface are defined positive.

The longwave radiation of Rp2 and CON correlate well with observations, but some biases are observed. The underestimation of LWd illustrates that the atmosphere is too cold in the model. This could be due to too few clouds, too low atmospheric humidity or biases in the radiation scheme for these cold conditions. This is partly compensated for by underestimated LWu, resulting in a relatively small LWn bias. In GS, the bias of LWn is larger, as higher surface temperatures lead to an overestimation of LWu, while only partly compensated for by increased LWd. Bias differences between most stations are limited, especially close to the edge of the ice sheet (Fig. a, b). For station 12 that is located on the Antarctic Plateau, both LWd and LWu are overestimated for Rp2 and CON.

Table  shows that SWd is overestimated for all model experiments. As no parameters that directly impact SWd have been changed, it illustrates that the atmosphere is too transparent, likely due to similar reasons as causing the LWd differences. For Rp2 and CON, this bias is compensated for by SWu, as the albedo is somewhat too high during summer. Table  also shows that the albedo of GS is on average too low, which is discussed in more detail in Sect. , resulting in a lower compensating SWu and consequently a larger SWn bias. Similar to longwave radiation, the biases of most stations are similar, except for station 12, where SWd and SWu are underestimated (Fig. d, e).

On average, the SHF is overestimated during summer, despite an underestimation of the wind speed (V10m, Table ). The SHF overestimation is stronger for station 16 and for more inland stations like station 8 and 12 (Fig. f). This can be either due to an incorrect representation of the roughness length or an incorrect temperature gradient between surface and atmosphere. In GS, turbulent heat exchange is smaller while T2m is overestimated. For a stable surface layer, this therefore suggests that the temperature of lower atmospheric layers is too high in the model. Similarly, GS also shows a better LHF representation than Rp2 and CON (Table  and Fig. c). Hence, turbulent fluxes can still be further improved.

Figure shows the mean yearly-accumulated SMB, melt, precipitation and sublimation difference with Rp2 for GS (upper row) and CON (lower row). In CON, the SMB differences are generally small (lower than 10 mm w.e. yr-1), with somewhat larger differences for the West Antarctic ice sheet (WAIS) and the AP that are driven by precipitation changes. The precipitation changes are minor, however, as total precipitation for the WAIS and AP are more than an order of magnitude larger . Melt has increased on the Wilkins, George VI, and northern part of the Larsen C ice shelf in the AP and the Amery ice shelf in East Antarctica. The changes of Rp3 are therefore largest for warm regions where melt is already significant, in agreement with . Runoff, however, remains limited (not shown), and almost all meltwater is buffered in the snowpack where it refreezes. Only at the southern part of the Amery ice shelf is the retention capacity now exceeded and runoff modeled, hence lowering the SMB. The margins of DML show considerable year-to-year and spatial melt variability. This demonstrates the high sensitivity of the implemented changes for this region, as the snowpack is close to the melting point in summer and additional energy absorption therefore leads to a stronger meltwater flux.

In GS (upper row of Fig. ), a strong SMB decrease is modeled for ice shelves in the AP, DML and Amery ice shelf. More inland, the SMB increases somewhat, which is mainly caused by an ice-sheet-wide precipitation increase. It is, however, partially compensated for by more sublimation. As the precipitation parameterization has not been changed, the moisture of this excess precipitation has been taken up locally. Further analysis showed that it relates to unrealistic features during summer in GS. Due to the higher surface temperature, sublimation increases and a cloud-topped shallow convective layer is modeled for the interior of the ice sheet. These clouds subsequently provide the additional precipitation. This synoptic weather pattern is, however, not backed by observations. Furthermore, melt has increased strongly around the margins of the entire AIS and all ice shelves. This melt changes the snow structure and leads to extensive runoff on several smaller ice shelves in DML, where the snowpack is close to saturation in summer, and on the Amery, Larsen C, Wilkins and George VI ice shelves. Finally, compared to 1924 SMB observations in the EAIS (locations shown in Fig. b), the difference between CON and GS is small, and both agree well with measurements (Fig. ), with a bias of 23.5 and 24.4 mm w.e. yr-1 and RMSE of 106.4 and 106.0 mm w.e. yr-1, respectively. The correlation coefficient is 0.41 for both CON and GS.

To investigate what causes the strongly overestimated melt in GS, Fig.  shows the melt difference with Rp2 for all sensitivity experiments. By increasing the fresh snow SSA (Fig. b) and reducing fresh dry snow metamorphism (Fig. c), less radiation is absorbed, lowering melt for most regions, in particular for the Ross and Filchner–Ronne ice shelves. These ice shelves are covered by fine-grained snow for most of the year and are therefore especially sensitive to changes in the fresh snow parameterization. The change to the fresh snow SSA and metamorphism delays the onset of the melt season, but its impact diminishes as the melt season progresses. Unsurprisingly, a strong melt reduction occurs by lowering the refreezing grain size (Fig. d), which leads to less energy absorption in areas with refreezing. For the time step currently employed in Antarctic simulations, a SLED of 5 mm is on the lower end of the scale analysis that is employed in . This underestimation of the SLED results in a slightly overestimated heat buffering in the uppermost part of the snow layer, leading to more internal heat absorption. Hence, melt is expected to be further reduced by increasing the SLED. This is indeed the case when comparing RFG (Fig. d) with CON (Fig. e), illustrating the impact of subsurface heating. Integrated over the AIS, yearly-averaged melt has increased by as much as 490 % in GS with respect to Rp2. Each sensitivity experiment lowers the melt flux, resulting in only a 7.0 % increase in CON (Table ). As a result, the domain-integrated yearly-averaged SMB modeled in GS is lower (2370 Gt yr-1) than CON (2407 Gt yr-1).

In this study, we investigated the impact of a new snow albedo and radiative transfer scheme in the latest adaptation of the polar version of RACMO2.3 on the near-surface temperature, (sub)surface snow temperature, SMB, SEB, albedo and melt of the AIS. We tuned Rp3 by incrementally changing one parameter at a time, allowing us to investigate the sensitivity of the AIS to each change.

We have run Rp3 for the entire AIS on a 27 km grid forced at the boundaries by 3-hourly ERA5 data. Three experiments are run for the full period (1979–2018): Rp2, Rp3 with Greenland settings (GS) and the Rp3 control run (CON) that includes all tuning steps. The results are compared to in situ and remote sensing observations and to the previous model version Rp2. The other sensitivity experiments are done for 1979–1990 and include increasing the fresh snow SSA (FSG), reducing the fresh dry snow metamorphism (FSM) and lowering the refreezing grain size (RFG).

Compared to observations and Rp2, the 2 m temperature in the GS experiment is considerably higher. The sensitivity experiments show improvements, resulting in a lower bias with observations for CON. The reduction of the fresh dry snow metamorphism rate in the FSM experiment results in a lowering of the temperature. For some areas, however, the 2 m temperature is now too low. Yearly averaged, it is underestimated by up to 0.5 ∘C. This indicates, together with SSA observations, that the fresh dry snow metamorphism might have been reduced too strongly and that further improved results would likely be reached with a larger value for α of Eq. (). More importantly, the results presented here highlight the necessity to correctly model snow conditions and that the current snow metamorphism scheme has to be improved or replaced. Nonetheless, subsurface temperatures of CON match well with observations at Dome C for the summer of 2007.

Analysis of the SEB shows that Rp3 exhibits, on average, some small (lower than 10 W m-2) persistent biases in the net radiative fluxes, which are caused by too transparent clouds and overestimated turbulent surface fluxes. This illustrates that there is still room for model development, especially in the turbulent fluxes. With the introduction of a new physically based albedo and radiative transfer scheme, more processes now impact the snow albedo. Radiation penetration and spectral shifts due to cloud cover and high SZA can lead to albedo differences up to -0.1 between CON and Rp2. Monthly-averaged, however, differences between these model versions are small.

The higher (subsurface) temperatures in GS lead to excessive melt around the margins and on the ice shelves, locally leading to runoff and a reduced SMB. Integrated over the AIS, melt in GS is 1 order of magnitude larger than observed by QSCAT and also considerably larger than measured at Neumayer station. In contrast, CON and Rp2 compare well with these observations. Melt is progressively reduced by all sensitivity experiments, especially in RFG and CON, showing the sensitivity of the AIS to the refreezing grain size and SLED. The difference between RFG and CON illustrates the importance of subsurface heating, which can warm the snowpack and enhance melt. Despite the low average melt flux in Antarctica, the impact of subsurface heating should not be neglected for a physical description of (sub)surface melt. It is clear that GS does not produce realistic meltwater fluxes and that the standard Greenland settings of Rp3 should not be used for the AIS. This is undesirable, as model settings should preferably not depend on location and/or tuning to local conditions, and shows that more research into this problem is needed.

In conclusion, introducing a new more physically based spectral snow albedo and radiative transfer scheme in the polar version of RACMO, which also allows for subsurface heating, improves, after tuning (as biases were partly compensated for in former RACMO versions), the subsurface snow temperatures in Antarctica. Incorrectly modeling snow conditions can lead to an order of magnitude melt overestimation and can significantly impact the climate and lower the SMB of the AIS. Furthermore, as is shown in the GS experiment, only a small lowering of summer albedo by, for example, global-warming-induced melting would lead to a very different near-surface summer climate in Antarctica.