Some fields of the ECMWF seasonal forecast system-2 and 3 have been included to enable verification studies. A lot of verification studies have also been done at the ECMWF. Some results are available at their member state server.

The fields are not homogeneous. They consist of ensembles started at the first of the month with perturbed ocean analyses.

For System-2 there is

• 1987-jul2001 calibration hindcasts; 5 ensemble members per month except May and November, which have 41.
• aug2001-dec2001 forecasts: 40 ensemble members (41 in November again).

For System-3

• 1981-2006 calibration hindcasts; 11 ensemble members per month.

Note that the forecast would have been available around the 15th of the month that the "+0" forecast refers to.

Each starting month is treated as a separate "reality", with 6 or 7 months with data and then 6 or 5 months with missing data, followed by the next year with the same pattern.