Bred-ensemble ocean forecast of loop current and rings

Abstract

Ocean forecasting with a General Circulation Model (GCM) commonly begins from an initial analysis obtained by data assimilation. Instead of a single initial state, bred-ensemble forecast [BEnF; which is used for weather forecasting at the National Centers for Environmental Prediction] begins from an ensemble of initial states obtained by using the GCM to breed fast-growing modes into the analysis. Here we apply the technique to forecast the locations and strengths of the Loop Current and rings from July through September 2005. Model results are compared against satellite observations, surface drifter trajectories, and moored currents. It is found that BEnF gives closer agreements with observations than the conventional single forecast. The bred-vectors (perturbed minus unperturbed state-vectors) have growth rates ≈0.04–0.08 day −1 and spatial (cyclone–anticyclone) scales ≈200–300 km suggestive of baroclinic instability mode in the Loop Current and rings. As in atmospheric applications, initializations with these growing vectors contribute to the more accurate ensemble mean forecast.