Characteristics of large-scale circulation "forecast busts” over Europe in ERA5 medium-range reforecasts

Despite rapid improvement in global-scale numerical weather prediction systems, forecast centers repeatedly experience notable declines in their systems’ forecast accuracy. These occasional very poor forecasts are often referred to as “forecast busts” and lead to perceptible differences in the anticipated weather. It is therefore important to understand the causes and the origins of "forecast busts” through a systematic analysis. Previous studies made use of reforecasts (or often referred to as hindcasts) to investigate hundreds of “forecast busts” from a fixed forecast model. The mean characteristics but also the different flavors of “forecast busts” have been investigated based on ERA-Interim reforecasts from the European Centre for Medium-Range Weather Forecasts (ECMWF). However, a deeper analysis of these events, particularly by using reforecast of ECMWF's most recent reanalysis (ERA5), is still missing.  

In this study, we identify “forecast busts" over Europe based on the day 6 anomaly correlation coefficient forecast skill of geopotential height at 500 hPa using ERA5 reforecasts of the period 1979-2023. Characteristics of these events are compared to the characteristics of “forecast busts” in an earlier model version (ERA-Interim), but also – for the first time – compared to exceptionally good forecasts in order to gain a better understanding of the strengths and weaknesses of the CY41R2 of the ECMWF Integrated Forecast System in predicting the large-scale pattern over Europe. To further extend the range of analyses on “forecast busts”, we systematically investigate various flavors of error patterns over Europe on forecast day 6 and discuss their distinct error pattern evolution and weather impacts over Europe.