The wind-induced flow in the North Channel of the Irish Sea is examined using an acoustic doppler current profiler (ADCP), to determine current profile, together with ocean surface current radar (OSCR) measurements of surface current and numerical models. The period considered is a major outflow event which took place during February 3--5, 1994. Although OSCR has been used previously to examine the spatial variability of tidal currents and long-term wind-driven flows, no data sets exist, and comparisons with models have not been performed before during a major storm event. Such an event is considered here. A three-dimensional shelf-wide coarse grid (resolution ~12 km) model is used to take account of large-scale wind events. A more limited area model of the North Channel (of higher resolution, ~1 km) with boundary forcing from the large-area model is used to determine current profiles in the vertical for comparison with ADCP measurements and surface currents in the area for comparison with OSCR measurements. The grid size of this model is the same as the bin size over which the OSCR system measures the current. This combination of models and measurements gives significant insight into the temporal and spatial variability of the shelf-wide winds, which produced the major outflow event, and the spatial variability of the flow in the North Channel. A comparison between along-channel flows computed with the shelf-wide model and ADCP measurements shows that this model overestimates the flow. However, a similar comparison with the higher-resolution model shows that this model slightly underestimates the flow. Since both models contain the same physics, namely, solve the same three-dimensional equations with identical vertical resolution, these differences can be attributed to the differences in horizontal resolution. These give rise to differences in the magnitudes of elevation gradients and local wind-forced currents computed with the models. Although there are errors in the flow fields computed in both models, they do reproduce the major features of the currents recorded by the ADCP within the North Channel and explain the major features, namely, the time variability of the shelf-wide winds influencing the flow in the region. A detailed comparison of surface currents measured with the HF Radar and model results interpolated to the radar bins shows a reasonable agreement (to within 0.15 m s-1), (taking into account errors in HF Radar measurements) in the computed and observed spatial and temporal variability of the surface current. However, the variability (in both space and time) of the computed current is significantly smoother than that determined by the radar. Also, away from the transmitter sites the signal to noise ratio in the radar signal is so large that accurate measurements (to within 0.3 m s-1) could notbe made during this storm event. Calculations showed that the value of the computed surface current was sensitive to the assumed roughness length, although below the surface layer (of the order of 3 m) the current was not affected by the surface roughness value. Comparison between computed currents determined with the high-resolution model and those measured by the radar showed a similar bias to that found in the ADCP comparison. ¿ 2001 American Geophysical Union