![]() |
||||||||
|
||||||||
Ensemble Data Assimilation and Forecasting (EDA and EPS) |
||||||||
|
|
In spring 2010 the new EDA-SVINI EPS will be implemented. This new version of the EPS uses a combination of perturbations generated from an Ensemble Data Assimilation (EDA) system and initial-time singular vectors to generate the EPS initial perturbations. Work has shown that the new system is better than the operational configuration based on initial-time and evolved singular vectors. Following earlier trials in 1998-2000 (see Buizza & Palmer, 1999, "Ensemble data assimilation", Proc. of the 17th Conf. on WAF, Sep 1999, Denver, US), experimentation based on ensemble of analyses generated by the Data Assimilation Section (L Isaksen and M Fisher) has continued in the past years. Different ways of using the EDA members to generate EPS initial perturbations have been investigated. The configuration that will be implemented in operation uses each member of an ensemble of analysis as a center point around which perturbations defined using initial-time singular vectors are added. Results so far has been based on 10 perturbed analyses and 50 perturbed forecasts, with groups of 5 perturbed forecasts sharing the same perturbed analysis. Buizza et al (2008) describes the methodology and summarizes the most recent set of results obtained in 2007-2008 on the investigation of the potential use of an ensemble of analyses in the ECMWF ensemble prediction system. Buizza et al (2010) provides a brief summary of recent results. Hereafter, recent presentations with some key results on the use of ensemble of analyses in ensemble prediction can be accessed:
|
|||||||
|
|
|||||||