@misc{73432,
  author = {D. Anderson and Tim Stockdale and Magdalena Alonso-Balmaseda and Laura Ferranti and Frédéric Vitart and Franco Molteni and F. Doblas-Reyes and Kristian Mogensen and A. Vidard},
  title = {Development of the ECMWF seasonal forecast System 3},
  abstract = {ECMWF has been running a seasonal forecast system since 1997. During that time there have only been two versions of the forecast system, called System 1 (S1) and System 2 (S2). This Memorandum gives an outline scientific description of System 3, which will be implemented in early 2007. The ocean analysis system has been extensively revised. In addition to subsurface temperature, the OI scheme now assimilates altimeter derived sea-level anomalies and salinity data. The scheme is now 3-dimensional, the analysis being performed at all levels simultaneously down to 2000m, whereas in S2, the analysis was carried out on each model level independently and only to 400m. In S3 there is also a multivariate bias-correction algorithm consisting of a prescribed a-priori correction to temperature, salinity and pressure gradient, as well as a time-dependent bias term estimated on-line. Forecasts in System 3 are now 7 months long previously 6) and the set of re-forecasts used to calibrate the system now encompass 25 years (previously 15). Sample size of the re-forecasts is increased to 11. Results from different model cycles leading up to System 3 are presented, demonstrating the recent improvements in Nino forecast performance. The climatology of the coupled model at seasonal forecast range is described, as are its variability characteristics at both interannual and intraseasonal timescales.. A preliminary assessment of forecast performance is given, based on a limited number of research experiments with a system equivalent to System 3.},
  year = {2007},
  journal = {ECMWF Technical Memoranda},
  number = {503},
  pages = {56},
  month = {04/2007},
  publisher = {ECMWF},
  url = {https://www.ecmwf.int/node/7727},
  doi = {10.21957/tz11xypf4},
  language = {eng},
}