The effect of a snow cover on sea ice accretion and ablation is estimated b
ased on the 'zero-layer' version sea ice model of Semtner, and is examined
using a coupled atmosphere-sea ice model including feedbacks and ice dynami
cs effects. When snow is disregarded in the coupled model the averaged Anta
rctic sea ice becomes thicker. When only half of the snowfall predicted by
the atmospheric model is allowed to land on the ice surface sea ice gets th
icker in most of the Weddell and Ross Seas but thinner in East Antarctic in
winter, with the average slightly thicker. When twice as much snowfall as
predicted by the atmospheric model is assumed to land on the ice surface se
a ice also gets much thicker due to the large increase of snow-ice formatio
n. These results indicate the importance of the correct simulation of the s
now cover over sea ice and snow-ice formation in the Antarctic. Our results
also illustrate the complex feedback effects of the snow cover in global c
limate models. In this study we have also tested the use of a mean value of
0.16 Wm(-1) K-1 instead of 0.31 for the thermal conductivity of snow in th
e coupled model, based on the most recent observations in the eastern Antar
ctic and Bellingshausen and Amundsen Seas, and have found that the sea ice
distribution changes greatly, with the ice becoming much thinner by about 0
.2 m in the Antarctic and about 0.4 m in the Arctic on average. This implie
s that the magnitude of the thermal conductivity of snow is of considerable
importance for the simulation of the sea ice distribution. An appropriate
value of the thermal conductivity of snow is as crucial as the depth of the
snow layer and the snowfall rate in a sea ice model. The coupled climate m
odels require accurate values of the effective thermal conductivity of snow
from observations for validating the simulated sea ice distribution under
the present climate conditions.