Abstract

A temporal sampling bias may be introduced due to the inability of a measurement system to produce a valid observation during certain types of situations. In this study the temporal sampling bias in satellite-derived measures of upper-tropospheric humidity (UTH) was examined through the utilization of similar humidity measures derived from radiosonde data. This bias was estimated by imparting the temporal sampling characteristics of the satellite system onto the radiosonde observations. This approach was applied to UTH derived from Television Infrared Observation Satellite (TIROS) Operational Vertical Sounder radiances from the NOAA-10 satellite from the period 1987-91 and from the 'Angell' network of 63 radiosonde stations for the same time period. Radiative modeling was used to convert both the satellite and radiosonde data to commensurate measures of UTH. Examination of the satellite temporal sampling bias focused on the effects of the 'clear-sky bias' due to the inability of the satellite system to produce measurements when extensive cloud cover is present. This study indicates that the effects of any such bias relatively small in the extratropics (about several percent relative humidity) but may be approx. 5%-10% in the most convectively active regions in the Tropics. Furthermore, there is a systematic movement and evolution of the bias pattern following the seasonal migration of convection, which reflects the fact that the bias increases as cloud cover increases. The bias is less noticeable for shorter timescales (seasonal values) but becomes more obvious as the averaging time increases (climatological values); it may be that small-scale noise partially obscures the bias for shorter time averages. Based on indirect inference it is speculated that the bias may lead to an underestimate of the magnitude of trends in satellite UTH in the Tropics, particularly in the drier regions.(Author abstract)