The suitability of basin-scale, satellite-based scatterometer winds for forcing of numerical ocean models is examined using a reduced gravity, primitive equation model of the tropical Pacific Ocean. Three surface forcing fields are validated in a comparison of upper layer thickness (ULT) from the ocean model with observed sea level data. The forcing fields are the Florida State University observed winds, winds derived from the NASA Scatterometer (NSCAT), and stresses derived directly from NSCAT. The sea level data sets are the World Ocean Circulation Experiment fast sea level data set from island measurements and sea level anomalies from TOPEX/POSEIDON. Results of this comparison demonstrate that while the three model results are qualitatively similar, the results are quantitatively better when forcing with the NSCAT derived stresses. This is particularly true in the eastern tropical Pacific and in convergent zones where forcing with the NSCAT stresses can lead to large differences in ULT (>40 m) compared with results from the other two wind products. ¿ 1999 American Geophysical Union