Investigating the dynamics of error growth in ECMWF medium-range forecast busts

Despite steady and significant progress in the skill of global numerical weather prediction, 'bust' or 'dropout' events occur when forecast errors grow much larger than expected from themean performance of the model. Researchers at the European Centre for Medium-Range Weather Forecasts (ECMWF) identified 584 bust cases for 6 day predictions over Europe utilizing the ERA-Interim system as a fixed forecast model. Rodwell et al. proposed that the model's handling of mesoscale convective systems over North America, particularly during the spring, is a likely culprit for these busts. In our study, the 584 cases were examined through an empirical orthogonal function (EOF) analysis of Northern Hemisphere 500 hPa heights during the 6 day forecast period. The cases were clustered according to their behaviour in the first EOF, revealing distinctly different pattern changes. Patterns are consistent with a low wave-number Rossby train reflecting over Greenland and a higher wave-number Rossby train trapped in the Atlantic waveguide. Analysis of the waveguide reveals increased definition of the southern reflecting boundary and an extension toward Europe during bust cases. The seasonal variation of these regimes is also examined, with the maximum in bust frequency taking place during September-October, when recurving tropical storms are common in the central Atlantic. Other seasonal peaks occur in the summer and winter, proposed to be associated with mesoscale convective systems over North America and east coast cyclogenesis. Overall, these results suggest that the busts are occurring in association with the triggering and amplification of Rossby-wave activity across the Atlantic, leading to large-scale pattern transitions.