The Stokes drift in ocean surface drift prediction

The importance of explicitly resolving the Stokes drift in ocean surface modelling is demonstrated by comparing four models with 58,612 observational data points obtained from undrogued drifting buoys Estuary and Gulf St. Lawrence, Canada. Drift model inputs are regional atmosphere circulation, spectral wave models. control considers near-surface currents provided top grid cell circulation model, which 5-m thick, a correction term proportional to wind. three other account for unresolved current shear extrapolating assuming Ekman dynamics. Two these consider drift, without wind term. Proposed reduce mean separation distance between observed predicted trajectories 34–40% relative on average, forecast times ranging 3 72 h. best improvement respect all metrics used is, however, that takes into any (skill score 0.93 after h 0.81 h).