The Autonomous Ocean Sampling Network (AOSN) program is aimed at designing observational networks to improve predictions of physical and biological processes in coastal oceans. For an AOSN field trial, held in Monterey Bay in August 2003, ensembles of the Regional Ocean Modeling System (ROMS) were designed for making 1-3 day predictions. The breeding and Ensemble Transform Kalman Filter (ETKF) techniques were used to initialize the ensemble each day. Since the key physical process in the Bay (upwelling) is primarily driven by the surface wind stress, ensemble perturbations of the surface wind field were also provided. Given just a single COAMPS model output, how should these perturbations be prescribed, (i) to maintain realistic physical structures and (ii) to exhibit a higher variance in regions of large gradient? A technique based on Autoregressive (AR1) functions will be presented.

The ensemble is also used in conjunction with the ETKF adaptive sampling strategy, to identify locations for underwater observing instruments to be deployed. The feasibility of such a technique will be discussed.