Heatwaves: (a),(c),(e),(g) 2-m temperature anomalies for the target week (indicated in the panel titles) from ERA5 data and (b),(d),(f),(h) those predicted by the ECMWF week-3 forecasts (hindcasts prior to 2016), with initialization dates indicated in panel titles. Rows show the (a),(b) California heatwave, (c),(d) European heatwave, (e),(f) U.S. heatwave, and (g),(h) East Asian heatwave. White boxes indicate the averaging areas used for Fig. 2. All case studies use model version CY45R1, except for the East Asia heatwave, which uses CY46R1.
Heatwaves: PDF of the predicted 2-m temperature anomalies from the model ensemble averaged over the target week (indicated in Table 1) for the heatwave case studies, averaged over the white boxes in Fig. 1 and initialized at (from left to right) 4, 3, and 2 weeks before the start of the target week. Panels show the (a) California heatwave 2018, (b) European heatwave 2018, (c) southeastern U.S. heatwave 2019, and (d) East Asian heatwave 2013. Tercile limits (below normal: blue; normal: gray; above normal: red) are computed with respect to the lead time–dependent model climatology. Values above the 66th percentile (below the 33rd percentile) are represented by red (blue) shading. Gray shading represents values between these terciles. The yellow dots indicate the ensemble members that were used to construct the PDF (51 for forecasts, 11 for hindcasts). The extremes above the 90th (below the 10th) percentile are hatched and their probabilities are indicated by red (blue) numbers. The purple dashed line represents the anomaly in ERA5 averaged over the target week.
As in Fig. 1, but for the cold spell case studies: (a),(b) Southeastern European cold spell in 2003 (model version CY46R1), (c),(d) central/northern European cold spell in 2018 (model version CY43R3), (e),(f) France cold spell in 2017 (model version CY43R1), and (g),(h) northern European cold spell in 2010 (model version CY46R1).
As in Fig. 2, but for the cold spell case studies: (a) Southeastern European cold spell in 2003, (b) European cold spell in 2018, (c) France cold spell in 2017, and (d) the northern European cold spell in 2010.
Precipitation events: Accumulated precipitation anomalies over the target week (week 3, indicated in the panel titles) for (a),(c),(e),(g) observations and (b),(d),(f),(h) the ECMWF model prediction (initialization date indicated in the panel titles). Rows show