The two most promising methods of estimating precipitable water W over global oceans rely on spaceborne microwave radiometers and atmospheric general circulation models. Global fields of W from the special sensor microwave image (SSMI) were compared with those from the European Center for Medium Range Weather Forecasts (ECMWF) model. They agree over most ocean areas; both data sets capture the two annual cycles examined and the interannual anomalies during an El Ni¿o-Southern Oscillation episode. However, they show significant differences in the dry air masses over the eastern tropical-subtropical oceans, particularly in the southern hemisphere. In these regions, comparisons with radiosonde data indicate that overestimation by the ECMWF model accounts for a large part of the differences. As a check on the W differences, surface level specific humidity Q derived from W, using a statistical relation, were compared with Q from the ECMWF model. The differences in Q were found to be consistent with the differences in W, and indirectly validating the Q-W relation. In both W and Q, SSMI was able to discern clearly the equatorial extension of the tongues of dry air in the eastern tropical ocean, while both ECMWF and climatological fields have reduced spatial gradients and weaker intensity. One of the reasons for the ECMWF moisture overestimation is suggested to the poor resolution of Tiros-N operational vertical sounder data which were assimilated into the ECMWF model. ¿ American Geophysical Union 1992