Intra-Americas Sea Trials: Real-time Prediction
The Regional Ocean Modeling System (ROMS) is used for data assimilation and ocean prediction in the Intra-Americas Sea (IAS) with particular emphasis on the Caribbean Sea. The 4Dvar algorithm is used to estimate initial conditions for real-time prediction onboard the Explorer of the Seas. The model is run at coase (40 km) and fine resolution (20 km).
During the course of a cruise week a 51 member ensemble of forecasts (coarse and fine) is created. The initial condition for one of the ensemble members (the control) is our best estimate of the ocean initial conditions on the forecast start date derived using the incremental 4Dvar procedure. The forecast model is forced using the lastest 16 day forecast wind and surface flux products from NCEP. The other 50 members of the ensemble are generated by perturbing the model forecast initial conditions and winds. The perturbations that are added are within the expected error bounds of the initial conditions and wind forcing. The initial condition perturbations are generated using random combinations of the ROMS tangent linear model solutions from the inner-loops of the data assimilation run just prior to the cruise start date, and the wind perturbations are generated using a Bayesian hiearchical model.
The ensemble mean can be thought of as the most likely state of the system, and the ensemble standard deviation gives an estimate of the forecast uncertainty, being largest where the forecast is least certain.
For more details visit the project link:
www.myroms.org/ias