A statistical evaluation of satellite soundings from TIROS-N and NOAA-6 is presented. Collocated satellite-radiosonde data were collected by season from July 1979 through May 1980 for clear and cloudy retrievals. In addition to RMS and mean error statistics, vertical correlations, ratio of satellite variance to radiosonde variance, and land minus sea differences for 30–60°N, 30–60°S and 30°N–30°S were examined. The results indicate that the mean and RMS differences are larger for cloudy retrieval paths than for clear retrieval paths. The largest RMS errors occur in the 1000–850 mb layer and in the tropopause region. Vertical correlations are highest between adjacent layers.
The ratio of satellite variance to radiosonde variance is considerably higher than reported in Phillips et at. (1979), with clear retrievals exhibiting a higher ratio than cloudy retrievals. Land minus sea differences suggest that there is a continental influence in the coefficient data base used to generate cloudy retrievals.