A nonlinear regression approach is employed to assess improvements in operational surface marine wind forecasts when synthetic aperture radar (SAR) measurements are also available. Analyses are constructed for coastal regions of eastern and western North America using a 2D-variational cost function, which simultaneously minimizes differences between the analyses and both wind forecasts and SAR measurements. The approach permits the error covariance matrices that define the cost function to vary from case to case. Comparisons are made with conventional methods of combining SAR and model data. An independent set of buoy observations are used to assess improvements in the resulting wind fields.