Numerical Simulation of North Atlantic Sea Ice Variability, 1951-1980 
John E. Walsh 
Department of Atmospheric Sciences 
University of Illinois, Urbana, IL 61801 
A two-level dynamic-thermodynamic sea ice model (Hibler, 1979) is used to simulate the 
growth, drift and decay of sea ice in the Northern Hemisphere during a 30-year period, 
1951-1980. The model is run with a daily timestep on a 222 km grid (Fig. 1) and is forced by 
interanually varying fields of geostrophic wind and temperature-derived thermodynamic fluxes. 
The objective is a quantitative description of large-scale sea ice variability in terms of the 
dynamic and thermodynamic processes responsible for the fluctuations, especially in the North 
Atlantic where sea ice represents a substantial input of fresh water. 
The fields of ice velocity and thickness contain strong seasonal as well as interannual 
variability. The mean drift pattern results in thicknesses of 4-5 m offshore of northern Canada 
and Greenland, while winter thicknesses of ~2 m are typical of Alaskan, Eurasian and East 
Greenland coastal waters. The 30-year mean fields are characterized by excessive ice in the 
North Atlantic during winter and by a summer retreat that is more rapid than observed. The 
excess of winter ice in the North Atlantic is due primarily to the omission of horizontal heat 
transport and deep convection in the ocean. Annual net growth ranges from 0.1 to 0.6 m over 
much of the Arctic Basin and Baffin Bay, while annual net melt of 0.5-1.5 m occurs in the North 
Atlantic marginal ice zone (Fig. 2). 
Despite the biases in the mean fields, the simulated interannual fluctuations correlate at 
0.4-0.9 with the corresponding observed fluctuations in individual sectors. In a comparison of 
results based on different data sources, closer agreement with observed fluctuations was 
obtained with the thermodynamic fields derived from the NASA GISS temperature grids than from 
the grids compiled in the USSR. The simulated velocities show no bias but considerable scatter 
relative to the drift of the Arctic bouys during 1979 and 1980 in the central Arctic and the 
Greenland Sea. 
An analysis of the regional mass budget indicates that the normal seasonal cycle is 
controlled primarily by thermodynamic processes, but that thickness anomalies in much of the 
Arctic are attributable primarily to dynamic processes during winter, spring and autumn (Fig. 
3). Thermodynamic processes contribute more strongly to summer anomalies and to anomalies 
near the ice edge. 
The tendency for thickness anomalies to be advected by the pattern of mean drift is 
apparent in multiseason lag correlations involving subregions of the Arctic Basin and the 
Fig. 1. The model domain and the 13 regions for Fig. 2. 30-year mean distribution of net 
which the mass budget statistics were evaluated. annual ice growth (m). 
