Baroclinic modification of surface wind vectors determined from satellite microwave radar data over the Northern Hemisphere midlatitude (30¿--60 ¿N) oceans is presented. The analysis uses data from the NASA scatterometer (NSCAT) collocated with operational surface analyses from the European Center for Medium-Range Weather Forecasts (ECMWF) at synoptic times. A total of 71,769 NSCAT wind vector observations from the Northwest Pacific and Northeast Atlantic Oceans were used in this study. A statistically significant baroclinic modification of the surface wind is detectable in the NSCAT winds that accounts for 5--35% of the variance in wind speed and 5--20% of the variance in wind direction depending primarily on the thermal wind strength. The amplitude of the baroclinic wind speed modification can be much larger than the effects of stratification in the surface layer. For the maximum thermal wind strength considered here (7¿10-3 s-1), the explained variances due to baroclinic effects were 59% for speed and 35% in direction. Because of data uncertainties these results probably underestimate the actual contribution of baroclinity to the surface wind variance. The NSCAT results are compared to ECMWF and to a theoretical model. The baroclinic modification of the NSCAT and ECMWF surface wind speed agree quite closely, but larger differences are seen in the cross-isobar angles. Reasons for these differences are discussed. These observations show that in addition to the primary surface wind data product, the scatterometer radar backscatter contains potentially useful information about the boundary layer temperature field. ¿ 1999 American Geophysical Union