Satellites need to observe the earth with great accuracy to capture the long-term trends in climate. As part of an effort to design a constellation of low earth orbiting satellites to benchmark climate observations, we explore the impact of imperfect sampling of satellites on the effort to monitor mean radiance on a variety of spatial and temporal scales. Our aims are (1) to find those orbits which provide an accuracy of at least 0.1 K in brightness temperatures at different temporal and spatial resolutions, and (2) to look for alternate ways of calibrating existing and future satellites like NPOESS. The 0.1 K level of accuracy in brightness temperature was chosen to agree with the expected magnitude of decadal trends in temperature forced by changes in greenhouse gas concentration.

Sampling studies carried out at different frequencies representative of the lower, middle and upper troposphere will be presented here. Model generated radiances rather than observed radiances are used for this study, since a good representation of diurnal variability in the original data is essential; polar orbiters don't have enough observation times, and geostationary orbiters don't cover enough of the world, and have problems because of high observing angles. The brightness temperatures are obtained by using MODTRAN on the archived simulations from the GFDL coupled model. These are then sub-sampled along the paths traversed by the satellite footprint for various potential orbits at different inclinations. Maps of retrieval accuracy for monthly mean, seasonal mean and annual mean radiance will be presented for single satellites and for a constellation of satellites. Also, the temporal and spatial distribution of simultaneous nadir overpasses for various satellite configurations will be examined to inspect the expected effectiveness for cross calibrating other existing and future satellites.

Results for the annual mean indicate that: A single satellite in a precessing orbit can achieve sampling errors in 15 degree grid boxes less than 0.1 K for brightness temperatures in the spectral regions that mostly sample the upper troposphere and lower stratosphere; In the mid-troposphere channels and in the window channel, a single precessing orbiter requires zonal averaging to reliably attain errors of less than 0.1 K; Since the primary source of sampling bias arises due to inadequate sampling of the diurnal and semidiurnal cycle, a constellation of satellites surely reduces the errors considerably and combinations with precessing polar orbits normally fare better than their sun-synchronous counterparts.