Analysis of Sampling Errors for Climate Monitoring Satellites

Abstract

Sampling retrievals of high-accuracy first-moment statistics constitute a central concern for climate research. Considered here is the important case of brightness temperature retrievals from a selection of possible orbits. Three-hourly global satellite brightness temperature data are used to predict the sampling error of monthly to annual mean brightness temperature retrieved by one or more satellites in low earth orbits. A true polar orbit is found to offer substantial advantages over a sun-synchronous orbit in the retrieval of annual mean brightness temperature, since the rotation of the local time of observation through two full diurnal cycles greatly reduces the error due to imperfect sampling of diurnal variations. Thus, a single polar orbiting satellite can produce annual mean, zonal mean brightness temperatures with typical sampling errors of less than 0.1 K, while even three sun-synchronous orbiters have high-latitude errors of up to 0.4 K. The error in retrievals of the annual mean diurnal cycle of brightness temperature is also discussed. In this case, high accuracy (<0.1 K) requires three cross-track scanning satellites in precessing orbits, or else a very large number (?10) of nadir-viewing satellites in precessing orbits. The large sampling errors of sun-synchronous satellites are highly correlated from year to year, so that if equator-crossing times are held fixed, sampling errors in year-to-year differences of annual means are similar for sun-synchronous and precessing orbits.