The Stokes drift in ocean surface drift prediction

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