The utility of high-frequency (HF) radar data for improving numerical circulation model predictions is evaluated. Comparisons of the statistical properties of the (CODAR-type) HF radar data and the observed wind indicate a strong correlation between the dominant alongshore, upwelling-favoring wind-forcing and HF radar-derived surface currents along the central California coastline. Because inadequate knowledge of the wind stress is probably a significant source of error in the model solutions, the idea of using HF radar data to provide corrections to the model wind-forcing is promising. Different HF radar data assimilation schemes are compared and judged based on the correlations observed between model currents and independent observations from two mooring sites. Analysis of correlation maps between model-predicted and observed currents indicates a spatial and temporal shift between modeled and observed features. However, the impact of HF radar data assimilation reduces these spatial and temporal shifts. A significant improvement in the correlation between the model and observed subsurface currents is achieved when an Ekman-layer projection of the corrections is included. In this approach, assimilation of HF radar data produces additional Ekman pumping (vertical velocity) based on the horizontal pattern of model-observed velocity mismatch at the surface.