CICERO - Senter for klimaforskning
klimafinans Lokale løsninger Energi Kina Internasjonal klimapolitikk Arktis Mat og skog

Melt ponds may explain rapid melting of sea ice

Klima - Et magasin om klimaforskning fra CICERO

Publisert 23.08.2009

The current climate models do not predict an ice-free Arctic for another 50 to 100 years, but observations show that melting of the sea ice is accelerating. Why are none of the models able to describe this rapid change? Melt ponds on the sea ice in summer may be part of the explanation.

The mean temperature in the Arctic has risen twice as fast as the global mean temperature in the past 100 years, resulting in a substantial decrease in the extent and thickness of the sea ice. The decline in the minimum extent of the Arctic sea ice in summer has attracted particular attention. All the climate models used as a basis for the Fourth Assessment Report from the Intergovernmental Panel on Climate Change (IPCC) show a decline in the extent of the sea ice, but none of them show the dramatic decline we have witnessed in the last few years (see Figure 1). We believe that one reason for this may be shortcomings in the description of the energy balance in current climate models. In a paper in the Journal of Geophysical Research (Pedersen et al., 2009), we present a new and physically more correct description of how much sunlight the sea ice reflects (its albedo) in the global climate model ECHAM5.

Ice reflects sunlight

The albedo of a surface is the ratio of the reflected to the incoming solar radiation. Surfaces covered in snow and sea ice have a high albedo. The albedo of clean new snow is up to 90%, whereas that of open water is only about 7%. This means that the sea ice is particularly sensitive to a moderate rise in temperature: a warmer climate melts some of the ice and exposes larger areas of open water, so that more of the heat from the sun is absorbed by the sea. This in turn results in further warming, creating a positive feedback loop.
The current climate models do not describe the albedo of sea ice satisfactorily. Earlier studies have shown that they fail to describe the annual cycle of sea ice albedo, and particularly how it changes in summer, when most models overestimate albedo. In many of the models, sea ice albedo has in fact been used to adjust, or tune, the results of the models to make them fit better with the observed conditions. In simple terms, an albedo has been described in non-physical terms in order to achieve better overall modelling results. However, our ultimate aim should be for the models to describe all physical processes as correctly as possible.

A better description of albedo

The new description, or algorithm, of sea ice albedo distinguishes between snow-covered sea ice, bare sea ice, melt ponds and open water. This is the first time a physical description of melt ponds has been explicitly included in a global climate model.
In spring and summer, the snow and the upper surface of the sea ice melt, and melt water gradually accumulates in melt ponds on the ice. The melt ponds reduce the albedo of the sea ice considerably, and absorb two to three times as much solar energy as a thick layer of bare sea ice. Earlier studies have indicated that the proportion of the sea ice covered by melt ponds in summer varies widely, from 5% to 80%, depending on how uneven the ice surface is, the snow depth, the ice type, the time of year and the geographical location. The spatial distribution of melt ponds depends mainly on the topography of the ice. First-year ice is smoother than multiyear ice, so that melt ponds on first-year ice are usually shallower, but cover larger areas. Multiyear ice is more uneven, and melt ponds form in depressions, where they tend to be smaller, deeper and more numerous.

More realistic calculations

In the paper mentioned above, we describe simulations of the climate and sea ice in the Arctic in which we used the general circulation model ECHAM5 to compare the new albedo algorithm with the original algorithm (which only used temperature-dependent albedo). The new algorithm simulates the annual cycle of sea ice albedo more realistically than the original algorithm, showing a decline in albedo both in winter (as the snow ages) and in summer (as melt ponds form). The new algorithm also performs well in modelling the distribution of melt ponds: both the timing of their formation and their extent agree well with observations. Simulations showed some melt ponds forming as early as May, and some persisting until September. The average proportion of the sea ice covered by melt ponds was shown to reach a maximum of 26% in July.
Modelling showed that the melt ponds on first-year ice were shallower than those on multiyear ice. However, the albedo reduction for the shallow ponds on the first-year ice proved to be larger than the reduction for the deeper ponds on multiyear ice, because the ponds covered larger areas on the first-year ice. In fact, in July, melt ponds covered 77% of the total area of first-year ice, but only 20% of the multiyear ice.
The new algorithm had most effect on the results for the summer months; albedo in August was found to be 23% lower than in simulations using the original algorithm (see Figure 2). The lower sea ice albedo resulted in an overall reduction in sea ice thickness, extent and volume, but with spatial and temporal variations. From May onwards there was less spatial variability, and for all geographical areas where the new algorithm had an effect, it showed a reduction in both albedo and the extent of the sea ice. The reduction in extent was largest (8%) in August and September. The new algorithm also showed substantial reductions in ice thickness and thus in sea ice volume. On average, the volume was 10% lower than indicated by the original algorithm.
The new algorithm represents an advance in our ability to model the current record rate of sea ice melt in the Arctic. Given the tendency for large areas of uneven multiyear sea ice to be replaced by smoother first-year sea ice, our simulations show that melt ponds may play an even more important role in ice melt in the Arctic in the years to come. It may therefore be of crucial importance for the accuracy of climate models to include melt ponds in the description of the energy balance.


  • C. A. Pedersen, E. Roeckner, M. Luthje and J.-G. Winther. A New Sea Ice Albedo Parameterization including Melt Ponds for ECHAM5 GCM. Journal of Geophysical Research., 114, D08101, doi:101029/2008JD010440, 2009.
Melt ponds may explain rapid melting of sea ice

MELT PONDS. Christina A. Pedersen making albedo measurements in an area covered by melt ponds in Van Mijenfjorden, Svalbard, June 2004 The melt pond is clearly darker than the surrounding snow, and absorbs much more of the incoming sunlight. Photo: Sebastian Gerland, Norwegian Polar Institute

Melt ponds may explain rapid melting of sea ice

Figure 1: Arctic sea ice extent as shown by observations (thick red line) and by 13 of the models used by the IPCC (bold black line shows the mean of all the models, and the dotted black lines the standard deviation).

Melt ponds may explain rapid melting of sea ice

Figure 2: Modelling of sea ice albedo in August (50-year average) using the climate model ECHAM5 with the new algorithm (ALB, left), and the difference between the sea ice albedo given by the new and old algorithms (ALB-DTL, right). Source: Pedersen et al. 2009.

Denne artikkelen ble opprinnelig publisert i Magasinet Klima nummer 2, 2009