A real-time ensemble forecasting system was developed for the recent Monterey Bay '06 ocean observing experiment in August 2006, in which numerous underwater vehicles were deployed. A stand-alone, non-nested domain of the Regional Ocean Modeling System (ROMS), with atmospheric wind stress and bulk forcing provided by the COAMPS model, was used to provide ensemble forecasts that complemented the nested ROMS control run, which assimilated in-situ datasets four times daily.

The physical relevance of locations of forecast uncertainty, explained by the ensemble variance, will be presented. The ensemble transform Kalman filter (ETKF) adaptive sampling strategy was developed for use with the ROMS ensembles, to assist in the deployment of underwater gliders to locations in which analysis errors and forecast errors may be optimally reduced due to assimilation of their observational data. The guidance provided by the ETKF and a preliminary evaluation of its performance will be presented.