| Summary: | Dynamical forecast systems have low to moderate skill in continental winter predictions in the extratropics. Here we assess the multimodel predictive skill over Northern Hemisphere high latitudes and midlatitudes using four state-of-the-art forecast systems. Our main goal was to quantify the impact of the Arctic sea ice state during November on the sea level pressure (SLP), surface temperature, and precipitation skill during the following winter. Interannual variability of the November Barents and Kara Sea ice is associated with an important fraction of December to February (DJF) prediction skill in regions of Eurasia. We further show that skill related to sea ice in these regions is accompanied with enhanced skill of DJF SLP in western Russia, established by a sea ice-atmosphere teleconnection mechanism. The teleconnection is strongest when atmospheric blocking conditions in Scandinavia/western Russia in November reduce a systematic SLP bias that is present in all systems.
|
|---|