Flow-dependent sub-seasonal forecast skill for Atlantic-European weather regimes

The continuous increase of computational power and improvement of numerical weather prediction systems in recent decades has allowed extending the operational weather forecast horizon into sub-seasonal time scales (10 – 60 days). On these scales, quasi-stationary, persistent, and recurrent large-scale flow patterns, so-called weather regimes, explain most of the regional surface weather variability and are thus of primary interest in sub-seasonal forecasting for the respective region. Here, we assess the skill of sub-seasonal reforecasts of the European Centre for Medium-Range Weather Forecasts (ECMWF) for predicting 7 year-round weather regimes in the Atlantic-European region. We primarily show that forecast skill considerably differs for different flow situations and seasons. We further elucidate the effect of model calibration on forecast skill: simply removing the model bias is shown to hardly affect and for some flow situations even reduce forecast skill, which indicates that flow-dependent model calibration techniques might be more useful for sub-seasonal weather regime forecasts. Finally, we give an outlook on how lower-frequency climate modes such as the stratospheric polar vortex as well as midlatitude synoptic-scale activity such as warm conveyor belts may enhance or dilute flow-dependent forecast skill.