Here, we implemented a melt pond scheme into the global coupled model EC-Earth v2.1, which is almost identically to the EC-Earth v2.2 used for the CMIP5-simulations. EC-Earth v2.1 consists of the atmosphere model IFS, development cycle 31r1 and the NEMO-ocean model system including LIM2. The NEMO version used in EC-Earth v2.1 is based on version 2 of NEMO and uses the ORCA1-grid, which has a basic resolution of 1º x 1º with refinement at the equator (1/3 º meriodionally). The resolution in the atmosphere is T159, which is about 1.125 º x 1.125 º.
Improving summer sea ice albedo
Summer sea ice albedo in the Arctic is much too high compared to satellite observations (e.g. Laine 2004) in EC-Earth. Particularly in summer, sea ice thickness in EC-Earth seems to be overestimated in the Arctic compared to available observations and estimates (e.g. Belchansky et al. 2008). This is most pronounced at the Siberian coast where sea ice is up to 2m thicker than estimates and leads to too much sea ice at the Siberian coast in summer.
To improve summer sea ice albedo, a melt pond scheme has been included into the albedo scheme of LIM2. The new sea ice albedo scheme is based on the existing sea ice albedo parameterization in EC-Earth v 2.1 and the melt pond parameterization scheme of Koltzow (2007):
It calculates melt pond fraction and the corresponding albedo in dependence of the surface temperature:
Melt pond fraction = 0.11 * (2 + Ts) for Ts > -2 C
Albedo of melt ponds = 0.36 - 0.1 * (2 + Ts)
Additionally, some small adjustments in the existing albedo scheme had to be made to include the melt ponds and avoid unrealistic jumps.
Successfull implementation of a melt pond parameterization
The melt pond parameterization has been successfully implemented into EC-Earth. An 80-year long present day simulation (using year 2000 greenhouse gas conditions) with the new parameterization has been performed (MELT). It was started from year 250 of a present day simulation of EC-Earth without melt ponds (CTRL). After 30 years, sea ice in MELT did not show any trends any longer. Thus, year 31-80 of MELT were compared to the corresponding years of CTRL (years 281-330) and observations and reanalysis.
The summer sea ice albedo is substantially reduced in MELT. In the Central Arctic, sea ice albedo is reduced by about 0.06. Thus, summer sea ice albedo in MELT fits much better to satellite observations.
Sea ice thickness in the Central Arctic varies between 2 and 3 m in March and 1 and 2 m in September in MELT. The thickest ice occurs north of Greenland and in the Canadian Archipelago with up to 5 m. Compared to CTRL, ice thickness is reduced by about 0.6 to 1 m in the Central Arctic in March and 1 to 1.5 m in September. Still, sea ice thickness at the Siberian coast seems to be somewhat overestimated. Ice thickness in the Atlantic Arctic sector might be slightly underestimated now.
Sea ice concentration is improved in MELT. Figure 3 shows a reduction of ice concentration along the ice edges in MELT. However, still the ice extends slightly too far to the south in the Greenland and Labrador Sea. In September, the largest reduction of sea ice concentration occurs along the Siberian coast, which leads to much more realistic ice concentrations there. The Siberian coast in MELT is almost ice-free in September. Sea ice concentration in the Central Arctic, particularly in the Atlantic sector is somewhat smaller than in satellite observations.
The new melt pond parameterization will be included into the new EC-Earth version 3, which is under development.
Belchansky, GI, Douglas, DC, Platonov, NG (2008) Fluctuating Arctic sea ice thickness changes estimates by an in-situ learned and empirically forces neural network model. J Clim 21:716-729, doi:10.1175/2007JCI1787.1
Koltzow, M (2007) The effect of a new snow and sea ice albedo scheme on regional climate model simulations. J. Geophys. Res. 112, DO7110, doi:10.1029/2006JD007693
Laine, V (2004) Arctic sea ice regional albedo variability and trends, 1982-1998. J. Geophys. Res. 109, C06027, doi:10.1029/2003JC001818.