The Clouds and Earth Radiatn Energy System (CERES) scanning radiometer has 3 channels: a total and a shortwave channel, which are used to determine broadband outgoing longwave radiation and reflected solar radiation, and a 8 to 12 micron window channel. Measurements from other radiometers having a 10.5 to 12 micron channel are often used to estimate OLR, but the results typically have errors of 10 to 20 W/m**2. These errors are due to effects of water vapor, whihc influences the OLR outside the window but has little influence in the window. In the present paper, OLR is correlated with the window channel to provide an estimate of the OLR. Because the difference of this estimate from the OLR is due to water vapor, primarily in the mid and upper troposphere, it provides a picture of water vapor distribution. Maps of the difference are constructed using measurements from the CERES instrument aboard the Tropical Rainfall Measuring Mission, which provides geographic coverage from 40N to 40S. These maps show clearly subsidence regions, which have low mid and upper tropospheric water vapor, and convective regions, with high mid and upper tropospheric water vapor.

The region of India and the nearby ocean shows a difference in the monthly mean map which is uniform over the region at all times. Assuming that the humidity is high and has little variation, is it possible to compute a window to broadband relation specific to this region of high humidity which will have a small error? The errors for daily maps in the India region are not noticeably reduced from those for the entire TRMM domain. It is tenatively concluded that although the surface humidity is uniformly high, at mid and upper levels the water vapor varies greatly.