Interannual characteristics of an 80 km resolution diagnostic Arctic ice–ocean model

Abstract

Seasonal simulations with large-scale coupled ice–ocean models have reproduced many features of the ice and ocean circulation of the Arctic Ocean and the Greenland and Norwegian seas (e.g. Hibler and Bryan, 1987; Semtner, 1987). However, the crude resolution and high lateral eddy viscosity used by these models prevent the simulation of many of the smaller-scale seasonal features and tend to produce sluggish circulation. Similarly, the use of a single year’s atmospheric forcing prevents the simulation of features on an interannual time-scale. As an initial step towards addressing these issues, an 80 km diagnostic Arctic ice–ocean model is constructed and integrated over a three-year period using daily atmospheric forcing to drive the model. To examine the effect of topographic resolution and eddy viscosity on model results, similar simulations were performed with a 160 km-resolution model. The results of these simulations are compared with one another, with buoy drift in the Arctic Basin, and with observed ice-edge variations. The model results proved most sensitive to changes in horizontal resolution. The 80 km results provided a more realistic and robust circulation in most areas of the Arctic and improved the modelled ice edge in the Barents Sea, while also successfully simulating the interannual variation in the region. Although it performed better than the 160 km model, the 80 km model still produced too large an ice extent in the Greenland Sea. No significant improvement in the ice-edge prediction was observed by varying the lateral eddy viscosity. The results indicate that problems remain in the vertical resolution in shallow regions, in treating penetrative convection, and in the simulation of inflow into the Arctic Basin through the Fram Strait.