A new method is described for forcing regional ocean models with wind stress fields derived from satellite scatterometer data. A variational technique is applied to produce regularly gridded surface wind (stress) fields in time and space using data from the NASA SeaWinds scatterometer aboard the QuikSCAT satellite. Three uniformly gridded wind stress products are produced with satellite scatterometer data for the Gulf of Mexico, one based solely on scatterometer data and the other two constrained to a background field derived from numerical weather prediction (NWP) models. These, along with winds from the Eta-29 NWP model, are used to force a high-resolution ocean model of the Gulf of Mexico. The four wind products are compared to National Data Buoy Center in situ observations. Ocean model data are compared to coastal sea level stations and moored acoustic Doppler current profiler data over the West Florida Shelf. The results show that the satellite products capture much of the variability at atmospheric synoptic scales and show great promise for representing energetic episodic events such as tropical cyclones. The techniques presented in this work are applicable to improving numerical ocean simulations of inaccessible or data-sparse regions of the world.