In addition to evaluating dynamical forecast models on a global scale, we assess regional forecast skill. Below you will find both standard and fair RPSSs for specific zonal bands (Fig. 1) and continental areas (Fig. 2).
Figures 1 and 2: Zonal and continental areas used to evaluate forecast skill.
Ranked probability skill scores from dynamical models
Figures 3 and 4: RPSSs for dynamical forecast models across (3) zonal bands and (4) continental areas. Regions are shown in figures 1 and 2. Forecasts are evaluated for three variables: weekly-mean near-surface air temperature (tas); weekly-mean sea level pressure (mslp); and weekly-accumulated precipitation (pr). Scores are shown for lead times of week 3 (days 19 to 25) and week 4 (days 26 to 32). For tas and pr, land-dominated grid points (>50% land cover) are included in the evaluation. RPSSs are based on from weekly forecast initialisations from 11 July 2024 to 9 June 2025. Model labels correspond to the most frequently used forecasting model version during the evaluated period. Full details on model versions and hindcast configurations are available here.
Fair ranked probability skill scores from dynamical models
Figures 5 and 6: Same as figures 3 and 4 but for fair RPSSs.