Figure 1. Time evolution of two of the headline scores used to monitor long-term trends in forecast performance of the HRES and ENS forecasts, adopted by Council in 2011.
Figure 2. Comparison of probabilistic forecasts of the 850 hPa temperature (T850) for ECMWF (red), the Met Office (UKMO, blue), NCEP (green), the Japan Meteorological Agency (JMA, orange) and Canada (CMC, magenta), using the headline, for summer (May-June-July, top panel) and winter (December-January-February) of the two most recent verifiable years.
Figure 3. Comparison of precipitation forecast skill for ECMWF (red), the Met Office (UKMO, blue), Japan Meteorological Agency (JMA, magenta) and NCEP (green) using the headline score for precipitation probabilistic skill, routinely computed over the Globe against synop observations:
Top-left panel: single HRES forecasts, verified using the SEEPS score in MJJ16;
Top-right panel: single HRES forecasts, verified using the SEEPS score in D15JF16;
Bottom-left pane; ENS probabilistic forecasts, verified using the CRPSS score in MJJ16;
Bottom-right pane; ENS probabilistic forecasts, verified using the CRPSS score in D15JF16.
Upper-air scores (Figure 1) indicate skill consistently at a level that was previously experienced only during the period of anomalously high predictability in 2010–11. The slight decrease in 12-month running average values seen in 2014 can be attributed to inter-annual atmospheric variability. Comparison with ERA-Interim confirms that recent variations are due to atmospheric variability.
ECMWF performs a routine comparison of the forecast skill of ECMWF and other centres for both the HRES and the ENS using the TIGGE data archived in the Meteorological Archival and Retrieval System (MARS). Few examples are shown in Figures 2 and 3. Results for precipitation over the last 12 months show a consistent clear lead for the ECMWF ENS over the whole lead time range both for upper-air variables (Figure 2) and for precipitation (Figure 3).
Share with your friends: |
