Exceptionally poor and good medium‐range forecasts of the large‐scale circulation over Europe in ERA5 reforecasts

Abstract Despite continuous improvements in weather forecasting, large‐scale forecast busts—sudden drops in accuracy—still occur. In this study, we extend the concept of busts to define “exceptionally poor forecasts” and introduce the notion of “exceptionally good forecasts”, both derived using a methodology that accounts for seasonality in forecast skill. We apply this framework to six‐day forecasts over Europe in European Centre for Medium‐Range Weather Forecasts (ECMWF) Reanalysis Version 5 (ERA5) reforecasts (1979–2023) to identify and compare their characteristics. The analysis explores potential links between these forecasts and large‐scale weather regimes in the North Atlantic–European region, with particular attention to the occurrence and timing of regime transitions. We identify a declining trend in the annual rate of poor forecasts and an increasing trend in the rate of good forecasts, consistent with advances in the number and quality of observations. Poor forecasts occur more often in the warm season while good forecasts are found throughout the year, and their mean patterns contrast sharply: Rossby‐wave trains characterize poor forecasts, whereas blocking over northern Europe dominates good forecasts. Periods of poor forecast performance coincide with an above‐average frequency of cyclonic regimes and persistent no‐regime periods. Conversely, good forecasts show an above‐average association with anticyclonic regimes, particularly with Scandinavian blocking. Whereas the share of cases with a regime transition is similar in both skill categories (60%), transitions occur significantly later during poor forecasts and earlier in good forecasts, providing a so‐called “window of opportunity” when initialized early in a regime life cycle. If regime transitions during poor forecasts occur early, the errors are not necessarily linked to wrong regime predictions, suggesting a fairly correct representation of the large‐scale circulation, while synoptic‐scale systems may drive large errors restricted to Europe. In summary, our study contributes to understanding of the large‐scale circulation configurations and stages of regime evolution that favour exceptionally poor or good forecasts over Europe.