Extended Abstract (124K)7.2
Multi Model Synthetic Superensemble Prediction System
Won-Tae Yun, Korea Meteorological Administration and Florida State University, Tallahassee, FL; and L. Stefanova and T. N. Krishnamurti
Despite the continuous improvement of both dynamical and empirical models, the predictive skill of extended forecasts remains quite low. The skill of single model ensemble or multi-model ensemble is generally not better than climatology for extended forecasts. Our experiments show that the construction of a multi model superensemble based on a family of DEMETER coupled ocean-atmosphere model data set does not lead to a significant increase of extended forecast skill. The DEMETER (Development of a European Multi-Model Ensemble System for Seasonal to Inter-annual Prediction) data set is comprised of 7 models. For each model, except that of the MPI (Max Planck Institute), uncertainties in the initial state are simulated through an ensemble of 9 different ocean initial conditions. In order to increase the predictive skill of extended forecasts, we propose an alternative method that makes use of synthetic data sets. The synthetic data sets are created by truncating the EOF or Fourier decomposition (either in space or in time) data, or by finding the concurrent EOF patterns between individual model and observation. A number of such synthetic forecasts are then used for the creation of a Superensemble forecast. The skill of this synthetic data-based ensemble is an improvement over both the ensemble of the model forecasts and climatology.
Session 7, Seasonal prediction (Room 6C)
Wednesday, 14 January 2004, 8:30 AM-9:30 AM, Room 6C
Previous paper Next paper