Predictability of midlatitude Rossby wave packets 

Abstract

Rossby Wave Packets (RWPs) are linked to extreme weather events and exert a strong influence on the predictability of weather systems in the midlatitudes. Considering the whole wave packet, in the sense of the packet envelope, RWPs can be viewed as entities that describe variability of the atmosphere beyond the synoptic scale. We here examine the predictability of RWPs as such entities. As a verification metric we used the so-called Displacement and Amplitude Score (DAS) applied to the envelope field of the midlatitude flow. The DAS is based on a field deforming method and, as one of its major advantages, avoids the “double-penalty” verification problem without the need to identify single RWP objects. We assess RWP predictability using a 19-year period of NOAA GEFSV12 ensemble reforecasts for RWPs that have been previously tracked in reanalysis data. Forecast errors defined by this metric asymptote towards saturation but do not completely reach saturation within the 10 days lead time available to this study. Corresponding error growth rates maximize during the medium range, in contrast to common error-growth models, in which growth rates are a maximum initially and monotonically decrease with lead time. We hypothesize that this difference relates to the lead-time dependence of error-growth mechanism. Variations in RWP predictability are dominated by the stage of the RWP life cycle, with higher predictability found for the propagation stage than the onset and decay stages. In addition, RWP predictability exhibits a seasonal cycle, with higher predictability in winter than in summer. Controlling for seasonality and the stage of the life cycle, we find i) that high-amplitude RWPs exhibit higher predictability than low-amplitude RWPs for medium-range forecasts and ii) that there is a general pattern of higher predictability over Eurasia than over the ocean basins, with some more detailed variations according to different lead times and life- cycle stages. Finally, predictability of the propagating stage is higher if forecasts are initialized after RWP onset than if initialized before onset. RWP onset thus acts as a partial predictability barrier to the subsequent propagation stage.