Multi-model assessment of the sub-seasonal predictive skill for the year-round Atlantic-European weather regimes

MARISOL OSMAN; GRAMS, CHRISTIAN M.; BEERLI, REMO

Conferencia; WCRP Open Science Conference; 2023

The prediction skill of sub-seasonal forecast models is evaluated forseven year-round weather regimes in the Atlantic-European region.Re-forecasts based on models from three prediction centers (theEuropean Centre for Medium-Range Weather Forecasts, the NationalCenter for Environmental Prediction and the UK Met Office) for the2000-2015 period are considered and compared against weather regimesobtained from ERA Interim reanalysis over the same period. We firstfocus on the assessment of skill, placing emphasis on the differencesin the performance for each weather regime depending on the time ofthe year. Then, we consider the year-to-year evolution of skill andthe role of interannual variability of the atmosphere in this skill.Finally, we study the changes in skill of weather regimes in themodels analyzed with respect to previous model versions.Results show that predicting weather regimes using a life-cyclecriteria adds value over using a simple weather regime definition orthe full 500-hPa geopotential height field over the regime domain.The ECMWF model shows the highest skill for most of the weatherregimes and seasons, followed closely by the NCEP model. The averageregime skill horizon is 3 days longer for ECMWF and NCEP models thanfor the UKMO model, mainly due to the differences in skill in winter.Greenland Blocking tends to have the longest year-round skill horizonfor the three models driven by their performance in winter. On theother hand, the skill is lowest for the European Blocking regime forthe three models, followed by Scandinavian Blocking. Furthermore,models struggle to forecast random flow situations in comparison toweather regimes. Related to this, the analysis of the dailyvariability of skill reveals that there is a predictability gapbetween winter and summer and the transitional seasons that can bepartly linked to the daily variability of no regime events. Inaddition, the interannual variability of wintertime skill isassociated with the anomalous occurrence of no regime: the winterswith fewer no regime days have higher skill. Finally, we find thatthe skill has improved in ECMWF and NCEP models with newer versions.For the NCEP model this improvement leads to an extension of theskill by around 4 days. This study represents the first systematicmulti-model assessment of the seven year-round weather regimes withdifferent state-of-the-art sub-seasonal models and represents acontribution towards a more objective assessment of the evolution ofweather regimes for days to weeks ahead for a better forecastguidance.p { margin-bottom: 0.1in; line-height: 115%; background: transparent }