Statistical prediction of seasonal air temperature over Eurasia

Abstract

Statistical models for the prediction of seasonal surface air temperature anomalies over Eurasia were constructed. The models were designed to test the relative predictive skill of Atlantic sea surface temperatures (SST), sea level pressure (SLP) and persistence in a cyclostationary setting.

Significant forecast skill was found for the spring season in central and eastern Europe. The main predictors were persistence and SLPs (the north Atlantic oscillation). SSTs had little predictive value. All results were confirmed with independent forecast experiments. The statistical results were attributed to (a) a positive feedback between given winter atmospheric circulation regimes, the snow cover they produce and the snow-induced enhancement/retardation of normal season warming and (b) the persistence of large-scale circulation patterns over the Atlantic Ocean.