Are Negative Arctic Oscillation Events the Opportunity for Sub-seasonal Forecasting?

Negative phases of the Arctic Oscillation (AO) sometimes result in extreme cold conditions over hemispheric-scale regions including the United States, Europe and Eurasia in boreal winter, and have significant impacts on various socio-economic sectors. Reliable forecast of the events in a few weeks advance is demanded for effective use of forecast information. We investigated the predictability of negative-phase AO events using archived datasets of real-time operational forecasts and a suite of hindcasts with the latest Japan Meteorological Agency (JMA) One-month Ensemble Prediction System (EPS). Three phases, namely low, normal and high phases, are defined using AO index with thresholds of -0.5, 0.5 standard deviations of the index. The AO index is computed by projecting daily sea level pressure (SLP) fields to the first leading mode of the Empirical Orthogonal Function (EOF) analysis of monthly SLP. Analyzing the AO index and forecast scores, it is found that remarkable low phase events are better predicted than the normal phase events in the real-time operational forecasts. The anomaly correlations (ACs) of 28-day averaged forecasts for SLP and 500-hPa geopotential height (GPH500) are stratified for the three phases. A ratio of AC above 0.8 for SLP (GPH500) in the low phase condition is roughly 20 (15) percents higher than that for the normal condition. Further diagnostics indicate that the high predictability stems from several dynamical mechanisms of a synoptic eddy-mean flow interaction, meridional and vertical planetary wave propagation and associated eddy transport of momentum and heat. The mechanisms are analyzed by means of the horizontal eddy vorticity flux and the Eliassen-Palm flux diagnostics. We focused on three cases in February 2013, December 2009 and February 2010, in which cases the skill of the JMA One-month EPS was remarkable in verification scores. We confirmed that the aforementioned mechanisms contributed to enforce these negative AO, and the forecast model successfully simulated these conditions. The effects of the mechanisms for the cases are significantly large compared to those analyzed in hindcasts. The results indicate that there is an inherent predictability in extra-tropical dynamics at a sub-seasonal time scale. With the results in this study, we point out that the remarkable negative phase of AO is a window of opportunity for sub-seasonal forecasting.