Pattern-based forecasting enhances the prediction skill of European heatwaves into the sub-seasonal range

The prediction of European heatwaves at the subseasonal range is of key importance to mitigate their impact. This study builds on previous work which identifies five main European heatwave types based on their atmospheric circulation patterns (CPs). These CPs are potential predictors of heatwaves, as these patterns are connected with a high probability of 2-meter temperature exceeding the 90th percentile. Therefore, the aim of this study is to use these patterns to construct a pattern-based forecast method. The skill of this method to forecast extreme warm temperatures is then assessed and compared with the direct grid-point based forecast (using the direct 2-meter temperature forecast of the model). The extended (or subseasonal) range reforecast data from the European Centre for Medium-Range Weather Forecasts (ECMWF) is used for the skill evaluation. Firstly, the skill of the extended range model is assessed in predicting CPs. The pattern-based methodology is then compared with the direct prediction of extreme warm temperatures. The results show that the pattern-based methodology has a low skill at the short to medium range compared to the direct method, however it maintains skill for longer lead times, extending the forecast skill horizon significantly by up to six days over key heatwave regions. This improvement is localized over regions with the highest conditional probability of extreme warm temperatures. Furthermore, the prediction skill of 4-day periods of high temperatures using CPs lasting at least five days is also assessed. A similar improvement in forecast skill horizon is observed but the improvement is more modest and even more localized. This methodology provides skilful forecast at longer lead times to the end of the medium range and into the subseasonal range, which would be beneficial for early warnings of European heatwaves and therefore support the timely implementation of mitigation plans.