Figure 1
(a) Global z500 RMSE (solid) and RMS difference between forecasts initialised 24 hours apart (dashed), based on the control member of the ECMWF sub-seasonal forecast system for the period 1 April 2025 to 31 March 2026 and lead times up to 45 days. The 100% (thick dashed) and 90% (thin dashed) levels of the mean RMSE over days 30–45 are also shown. (b) dE/dt as a function of E for the same data as in panel (a), shown for RMSE (crosses) and lagged-forecast RMS difference (black dots). The solid line shows a second-order polynomial fitted to the RMSE data, and the binned RMSE values are shown as grey dots.
Figure 2
(a) RMSE of z500 for operational ECMWF forecasts over the Northern Hemisphere for selected years (1986, 1990, 2000, 2010, and 2020). (b) As in panel (a), but for forecasts produced within the ERA5 system.
Figure 3
Global z500 RMS(E) for the period 1 April 2025 to 31 March 2026, shown for the 9-km IFS forecast (black) and AIFS forecast (grey). RMSE is shown by solid lines and the 24-hour lagged-forecast RMS difference by dashed lines.
Figure 4
z500 RMS(E) for the period 1 December 2023 to 10 March 2024. Results are shown for 1980/81 (thick grey; based on Figure 1 of Lorenz, 1982), for the 2023/24 36-km forecast system (grey), and for the 2023/24 9-km forecast system (black). The lagged-forecast differences from Lorenz (1982) and from the 36-km forecast system are shown by dashed lines, and the RMS difference from the accidental 9-km twin experiment is shown by a black dotted line. The figure also includes the 100% and 90% levels of the mean RMSE over days 30–45. The black dashed line marks the 35 m error level between days 1.5 and 5, and the black dotted line marks the same error level between days 5 and 10.
Figure 5
RMSE of 10-metre wind speed and z500 over Europe, verified against observations, for the period 1 April 2025 to 31 March 2026.