Subseasonal Predictability of the North Atlantic Oscillation

Abstract:

Stratosphere-troposphere coupling occurs on a range of timescales, from slow processes unfolding over many weeks, such as the polar night jet oscillation, to processes evolving on daily timescales, such as eddy feedbacks and planetary wave reflection. Some studies have suggested that the slower processes are too weak to explain the full scope of stratosphere-troposphere coupling. We do not dispute this finding, rather we reframe the question somewhat differently: What stratosphere-troposphere coupling processes are predictable on subseasonal time scales?

We constructed a low-order linear empirical-dynamical model of weekly-averaged atmospheric anomalies – a linear inverse model (LIM) – whose North Atlantic oscillation (NAO) prediction skill, for leads of 3-6 weeks, is comparable to that of the European Centre for Medium-Range Weather Forecasts Integrated Forecast System (IFS). We use the LIM’s signal-to-noise ratio to identify those weekly averaged DJF NAO forecasts with the highest ‘expected skill’, here chosen as the top 15% of all cases (1997-2016). For both the LIM and IFS, NAO forecast skill for these cases was above 0.5-6 for forecast weeks 3-4, which is significantly higher than the skill of the remaining 85% of forecasts, Fig 1a. This represents a notable improvement over existing methods of identifying conditional skill (e.g., MJO, SSWs), which can only boost conditional IFS NAO forecast skill above 0.5 at forecast leads of 2-3 weeks for a smaller fraction of forecasts.

Employing a “nonnormal filter” from the LIMs dynamical forecast operator reveals that the high skill NAO forecasts are linked to an initially strong projection on only a few eigenmodes with pronounced stratospheric and tropical sea surface temperature (SST) components, Fig. 1b. Notably, this ‘stratosphere-SST’ subspace includes one eigenmode, with no tropical SST or tropical heating component, which on its own captures the observed downward propagation of stratospheric anomalies related to both strong and weak vortex events and subsequent impact on the NAO. The ‘stratosphere-SST’ subspace represents much less NAO variability than the remaining eigenspace, yet contains the bulk of the predictable NAO component beyond the synoptic predictability limit, suggesting it may provide a better ‘target’ for NAO subseasonal forecasts.