Abstract
The observed reduction of Arctic sea ice has drawn a lot of interest for its potential impact on midlatitude weather variability. One of the outstanding challenges is to achieve a deeper understanding of the dynamical processes involved in this mechanism. To progress in this area, we have designed and performed an experiment with an intermediate complexity atmospheric model. The experiment shows a transient atmospheric response to a surface diabatic heating in the Barents and Kara Seas leading to an anomalous circulation first locally, then over the polar region and finally over the Euro‐Atlantic sector. A hypothesis that explains the mechanisms for the propagation of the signal is put forward. The discussion of this hypothesis provides an insight into the nature of the link between sea‐ice forcing and the modes of internal variability of the atmosphere. We demonstrate that, after removal of sea ice in the Barents and Kara Seas, first the linear atmospheric response dominates and is confined in the proximity of the heating area, then a large‐scale response, associated also with eddy feedback, is found and finally anomalies reach the lower stratosphere and show a hemispheric pattern in the troposphere. These results identify the drivers of the tropospheric connection between sea‐ice variability and the North Atlantic Oscillation and highlight the role of the lower stratosphere.
(a) Time series of the area‐weighted, monthly mean geopential height anomaly for low‐ice years in Era‐Interim from December to April at 300 hPa over the Barents and Kara Seas (70–85°N, 30–90°E; red line), over the North Atlantic (65–85°N, 60–0°W; blue line), and at 30 hPa over the polar cap (60–90°N; black line). Dashed lines indicate the interquartile range of the distributions. (b) Latitude–pressure cross‐section for the zonal mean of U (contours, ms−1) and T (colour shading) anomaly in February, (c) Z300 anomaly in February, and (d) T850 anomaly in February. The green solid line in (c) and (d) indicates statistical significance at the 90% confidence level.