This study investigates modeling the wind-driven current using observed wind stress and an empirically estimated impulse response function for the wind-driven current response to wind forcing. Convolution of the data-driven impulse response function with the wind forcing gives an estimate of the wind-driven part of the current. We estimate this data-driven convolution operator using ERA5 reanalysis winds and observations of surface currents from drifting buoys. The response of the currents to wind forcing is expected to depend on the ocean mixed-layer depth and the turbulent viscosity profile with spatio-temporal variations, but we only consider seasonally-modulated and spatial variations in the training. Despite this crude approximation, the simplified response function explains a significant portion of the current variability. Indeed, when the function is applied to the ERA5 wind-stress, the surface current estimate compares reasonably well with independent observations (not used in the training).

A practical application is the release of new total surface current estimates such as the Globcurrent CMEMS MOB-TAC based on the same datasets, but here containing also the unsteady part of the wind driven currents (the inertial currents). The characteristics of the response function (amplitudes and phases) reveal interesting properties of the upper-ocean variability. The function shows some similarities to one derived theoretically from a simple 1-layer (slab) model, but also differences that highlight the value of fitting the function to the data without the use of an explicit dynamical model. This opens perspectives for studying some dependencies between subsurface variables and the response function, particularly interesting in the context of future spaceborne Doppler scatterometers such as ODYSEA, expected to provide simultaneous wind and current observations: this instrument could indirectly probe subsurface properties through the synoptically-observed response function.