Diagnostic methods for understanding the origin of forecast errors

Although the quality of medium-range forecasts has increased considerably over the decades since the start of operational forecasts at ECMWF, individual forecasts still occasionally experience very large errors. Often the phrasing ’drop-outs’ or ’forecast busts’ is used for such episodes. The aim of this report is to use a combination of methods to track errors in three cases of extreme forecast errors between 2014 and 2016 to better understand the error sources. Manual error tracking and ensemble sensitivity are used to give a first guess of the source region and relaxation experiments are used to confirm the result. In the three investigated cases the errors originated from the tropical eastern Pacific, western/central Canada and western Atlantic respectively. The mechanisms behind the errors are discussed in the report. The results from this study can form a basis for further investigations of these cases and the methodology explained can be applied to understand future bust cases to increase our knowledge on origin and propagation of forecast errors.