Surface wind fields over the ocean are highly variable on scales of a few kilometers and an accurate specification is important for a variety of applications. Considering their high horizontal resolution, synthetic aperture radar (SAR) measurements by RADARSAT--1 may be well suited for incorporation into a coastal data assimilation system. It is first necessary to demonstrate, however, that a surface wind analysis improves when SAR observations are employed. The feasibility of combining SAR measurements with high-resolution simulations of the surface marine wind field is examined for coastal regions of eastern and western North America. A 2D-variational approach is used to retrieve wind fields for half of the available SAR acquisitions. Wind observations by ships of opportunity and buoys, taken within an hour or two of these acquisitions, are used to tune the variational methodology. This involves quantifying errors in SAR measurements and their biases. An independent set of wind observations are then used to assess the surface wind fields that are retrieved using the remaining SAR acquisitions.