Results of a Stochastic Quality Assurance Algorithm for Radiation Budget Data

The Earth Radiation Budget Experiment (ERBE) included 3 sets of broadband radiation instruments, which rode on the NOAA-9 and -10 operational meteorological spacecraft and on the dedicated NASA Earth Radiation Budget Satellite (ERBS). The objective of the program was for the ERBE instruments aboard the polar orbiting NOAA-9 and -10 spacecraft to provide broadband radiation measurements every day and night over the globe from which maps could be made of outgoing longwave radiation (OLR) and reflected solar/shortwave radiation (RSR). The ERBS was placed in an orbit of 57 degree inclination so as to precess through all local times fairly quickly (every 72 days), thereby providing information about the diurnal variation of OLR and RSR. Used together, the data from these three spacecrafts would provide accurate monthly mean global maps of Earth radiation budget.

The wide field-of-view broadband radiometers aboard the ERBS have provided Earth radiation budget measurements from November 1984 through the present. Unfortunately, since 1992 there have been no ERBE data other than that from ERBS. Thus, monthly mean maps of OLR and RSR have been prepared from the ERBS data alone. Because of its precessing orbit, the ERBS does not provide good coverage of the Earth every month at high latitudes. The variety of sampling conditions complicate the quality assurance problem. Thus, analyses of the temporal sampling errors have been developed which take into account the details of each measurement and the variability of the OLR and shortwave albedo for each region as they affect the monthly-mean products. The number of measurements and the measurement times for a given region are pseudo-random numbers. These computations have been applied to monthly mean OLR and shortwave albedo maps from ERBS for the entire 14-year data period. Monthly mean values for which the 1-sigma temporal sampling error exceeds 12 W/m**2 for OLR or RSR are not recorded. In this paper we examine these computed temporal sampling errors in order to understand the interplay of the parameters of the problem as they affect these errors.

The OLR occasionally has large sampling errors for regions between 55 and 60 degrees latitude in each hemisphere, which is the maximum latitude for which results are obtained. The RSR has many more regions for which the monthly-mean value is rejected. The rejected cases are usually at high latitudes and the latitude range of rejected values is larger in the summer hemisphere, where the incident sunlight is large and relatively small errors in shortwave albedo cause large errors in RSR.