Figure 8 shows the annual mean RMS error of the ensemble mean forecasts for 500-hPa height, verified over the Northern Hemisphere.
The six curves coming from the six ensembles for forecast days 1–6 are divided into two clusters that belong to the operational and reforecast ensemble forecast groups, respectively. Of the three operational ensembles, OPR_OPT has the lowest RMS error among the six ensembles.
The OPR_DAV2% has reduced RMS errors for the first week compared with OPR_RAW but its RMS becomes larger for week-2 forecasts. However, the two similar curves of OPR_OPT and OPR_DAV2% for the first week suggest that there is only a limited opportunity for future improvement in bias correction for the first few days.
The big distance between the OPR_OPT and OPR_DAV2% curves for week 2 indicates that the OPR_DAV2% calibration technique has the potential to improve extended forecasts.