2.7 Coupled ocean-atmosphere extended-range probabilistic prediction at ECMWF

Monday, 13 June 2011: 12:15 PM
Pennington AB (Davenport Hotel and Tower)
Roberto Buizza, ECMWF, Reading, Berkshire, United Kingdom

The extended forecast range (say from 10 days to one year) has been increasingly attracting interest in the past decade. ECMWF has been very active in this field, and has been producing top quality probabilistic forecasts spanning this range for many years. Today, ECMWF medium-range forecasts (say with a forecast range up to 1 month) are generated using the Ensemble Prediction System (EPS), and extended-range forecasts (say with a forecast range up to 13 months) using the ECMWF System-3 (S3). Both systems use a coupled ocean-atmosphere model. The EPS is based on the Hamburg Ocean Primitive Equation (HOPE) ocean model with a 1-to-1/3 degree resolution and the most recent version (model cycle 36r4) of the ECMWF atmospheric Integrated Forecasting System (IFS), run with a 32km resolution up to forecast day 10 and a 65 km resolution from day 10 to day 32. The S3 uses a constant resolution for the whole 13-month forecast range, with the same version of the HOPE ocean model as the EPS, and an old version of the IFS (model cycle cy31r1 that was operational in 2006) run with a 120km resolution. HOPE ocean initial conditions are currently generated using an Optimum Interpolation scheme. In 2011, both the EPS and S3 will have the HOPE ocean model replaced by NEMO, the Nucleus for European Modelling of the Ocean model, with NEMO initial conditions generated using NEMOVAR, a 3-dimensional variational data-assimilation system. On the seasonal time scale, S3 will be replaced by S4, a new system that will use a higher resolution, 80 km version of the model recent version of the IFS (cycle 36r4). During this talk, the ECMWF ocean-atmosphere coupled systems will be described, their performance will be presented, and ongoing research will be discussed.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner