The site displays verification statistics for global NWP forecasts. All data used are on the GRIB G2 (2.5x2.5 degree) grid.
The NWP models evaluated are: GFS: The U.S. NCEP Global Forecast System; FNO: The U.S. Navy Fleet Numerical Meteorology and Oceanograpy Center; ECM: European Center for Medium-Range Weather Forecasts; UKM: The United Kingdom Met Office; CMC: The Canadian Meteorological Center; JMA: Japan Meteorological Agency; CDAS: The frozen version of GFS used for the NCEP/NCAR Reanalysis Project.
The regions referred to are: G2 (GLB): Globe NHX: 20N-80N SHX: 20S-80S TRO: 20S-20N PNA: Pacific North America, 180E-320E, 20N-75N
The graphs displayed on this website consist of skill scores computed using a legacy verification system from 1996 to 2012, and the scores based on a new VSDB-based verification system from 2008 onward. The two systems were built upon different designs, but differ primarily in the climatology and anlyses used for computing AC, bias and RMSE. Please see this doc for the detail. Large differences in bias and RMSE are found between the two systems because the legacy system uses zero-hour forecasts (pgbf00) as verification truth, while the new system uses analyses (pgbanl) as verification truth. However, differences in anomaly correlations between the two systems are found to be small.
Please see this presentation for definitions of the various verification statistics.