Investigating the medium-range predictability of European heatwave onsets in relation to weather regimes using ensemble reforecasts

Over the past years, Europe was stricken by unprecedented heat and drought extremes. Heatwaves (HWs) pose an increasingly major threat to society, which requires to further our understanding of their predictability some days or weeks ahead. In the present study, we focus on medium-range lead times of 5-12 days, for which the successful prediction of HW onsets is strongly dependent on the adequate prediction of large-scale Rossby wave patterns and their dynamics.

Using ECMWF and GEFSv12 ensemble reforecasts, we statistically assess the medium-range predictability of HW onsets over multiple European regions for the period 2001-2018. Heatwaves are objectively diagnosed as time periods where the 90th percentile in 2m maximum temperatures is exceeded both grid point-wise and integrated over the region for at least 3 days, giving about 50 HWs for each region. Predictive skill is evaluated mainly for large-scale flow features but also in terms of the ensemble’s capability to predict the extremeness of near-surface temperatures. By adopting the concept of Euro-Atlantic weather regimes, we are further able to stratify HW onsets and their predictability by the concurrent large-scale atmospheric flow pattern.

For both the British and Central European region, we find that the medium-range predictive skill is significantly better for HW onsets associated with either Scandinavian or European blocking compared to cases with no pronounced regime. This skill advantage mostly concerns the large-scale flow (500hPa geopotential patterns) and to some extent 850hPa temperatures, particularly in times of a Scandinavian blocking. In contrast, near-surface maximum temperature forecasts do not exhibit increased skill for regimes associated with above-average large-scale predictability, which may point to the inherent difficulty of predicting smaller scales and boundary layer processes.

Finally it is explored how exceptionally good or poor forecasts at 10 days lead time are related to the atmospheric state during or shortly after forecast initialisation. For Central European HW onsets, poor large-scale flow predictive skill is associated with significantly increased baroclinicity further upstream and a more intense North Atlantic jet stream whereas particularly good forecasts are on average characterized by an initial atmospheric state close to climatology. Forecast skill for near-surface temperatures is not affected by such dynamical precursors, but rather by pre-existing soil moisture anomalies, with drier than normal soils being linked to better predictions and vice versa. For the British region, exceptionally good forecasts of both large-scale flow and near-surface temperatures are associated with an already established blocking over the continent. In comparison to Central Europe, antecedent soil moisture anomalies do have a same-sign, but less prominent effect on near-surface temperature forecasts.