Climate Dynamics(1992) 7:155-163 lima|¢ Uynamla © Springer-Verlag1992 The effect of an arctic polynya on the Northern Hemisphere mean circulation and eddy regime: a numerical experiment Rita Glowienka-Hense and Andreas Hense Meteorologisches Institut, Auf dem HOgel20, W-5300 Bonn, Federal Republic of Germany Received December 27, 1990/Accepted November 22, 1991 Abstract. The feedback of an arctic polynya, which is a large ice-free zone within the sea ice, on the hemis- pheric climate is studied with the ECMWF T21 GCM. For this purpose a control and an anomaly integration, in which a polynya was introduced in the Kara Sea, are compared. As the GCM, like the real atmosphere, shows a high level of low frequency variability, the mean response to the changed boundary conditions is obscured by internal noise. The necessary significance analyses are thus performed to enhance the signal-to- noise ratio within the framework of an a priori chosen guess pattern and a multivariate test statistic. The sensi- ble and latent heat fluxes increased above the polynya, which resulted in a warming of the lower troposphere above and near the polynya. No statistically significant local or global sea-level pressure changes are associated with this heating. However we find a significant change of hemispheric extent of the geopotential fields at 300 hPa, if we use as guess patterns the eigenmodes of the barotropic vorticity equation. The different mean flow field is accompanied by significant changes of the synoptic transient eddy field. We find a significant var- iation in the barotropic and baroclinic forcing of the mean flow by the eddies, a change in the location and intensity of the storm tracks and in the conversion be- tween eddy available and eddy kinetic energy. The ad- ditional heat flux from the polynya results in a reduc- tion of the meridional heat flux by the synoptic eddies on the western Atlantic. Introduction Sea ice sensitivity experiments with atmospheric GCMs deserve special interest, on the one hand, to clarify its role in global warming and, on the other hand, because the persistence of local sea ice anomalies can contri- bute to medium range predictability. GCM experiments with an ice-free Arctic or Antarctic ocean are discussed by Fletcher et al. (1971) Warshaw and Rapp (1973) Simmonds and Dix (1986) and Royer et al. (1990) while experiments with ice age boundary conditions are the subject of Williams et al. (1974) and Lautenschlager and Schlese (1989). Circulation changes due to smaller changes of the ice margins typical of interannual to decadal time scales have also been examined using GCMs (Simmonds 1979; Hermann and Johnson 1980; Mitchell and Senior 1989; Mitchell and Hills 1985). The impact of a 50% lead fraction is the subject of a GCM study by Simmonds and Budd (1990). Above the ice-free zone the troposphere is warmed and the stratosphere is cooled in these GCM experi- ments. The stratospheric cooling extends equatorward of the tropospheric warming in all but one of the model simulations. Simmonds and Dix (1986), who have re- moved the total Antarctic sea ice in their anomaly si- mulation, found this pattern significant at the 95% level using local student-t-statistics. A different pattern ap- pears in the sensitivity experiments with a 50% lead fraction in the Antarctic sea ice cover by Simmonds and Budd (1990). The troposphere is also warmed above the sea ice/lead region, but the stratosphere is cooled only poleward over the continent and warmed between 30-65 ° S, above the northern part of the sea ice region and on the equatorward side. This pattern is also significant at the 95% significance level. This result sug- gests that open water areas within the sea ice margins affect the atmosphere differently to a total sea-ice re- moval or a boundary change. Similarly, the sea-level pressure response is not consistent in the different model experiments. In the three experiments of Sim- monds (Simmonds 1979; Simmonds and Dix 1986; Simmonds and Budd 1990) significant pressure changes of nonuniform sign occur above the regions where sea ice is removed, while in the other studies there is a con- sistent trend to lower pressure above these regions. Here the results of a GCM simulation with an arctic polynya are reported. The sea ice distribution in GCMs is generally prescribed, and at the sea ice grid points the albedo, roughness length and a mean ice thickness are set according to climatological conditions. Pole- wards of the sea ice margin 100% ice cover is assumed. It is, however, known that especially outside the central