The radiation budget of the Earth governs its climate. Our knowledge of the Earth��s radiation budget is based on satellite measurements from which monthly mean maps fo outgoing longwave radiation and reflected solar radiation are compted. When the annual average global mean net radiation is calculated for several years, the Earth should be in radiation balance within a W-m-2. However, our data products give us a net imbalance of a few W-m-2 because of the accumulation of errors in the measurements and the methods by which the monthly mean maps of radiation fluxes are produced. Even though we may improve our measurements and algorithms for generating data products, there will always be errors. In this paper we present a technique for computing the most likely corrections to the major errors so as to bring the Earth's net radiation into balance. The method for balancing the Earth's radiation budget is to select the major error sources and estimate their variances. Next, the sensitivity of the annual average global mean net radiation to each error is modeled. A set of corrections is then computed by maximizing the likelihood of this set, constrained by the requirement that these corrections bring the average net radiation into balance. The method is applied to radiation budget data from the C louds and Earth Radiant Energy System (CERES) project. From the CERES radiometer measurements, radiances are computed by use of the instrument gains. The spectral responses of the radiometer channels are taken into account to compute broadband longwave and shortwave radiances. Fluxes at the "top of the atmosphere" are computed by use of bidirectional reflectance functions for the shortwave fluxes and limb-darkening functions for the longwave fluxes. The instruments are on Sun-synchronous spacecraft, so that measurements are taken at two local times for each spacecraft; thus it is necessary to include the diurnal cycle of radiation in order to compute the average flux for the day. Each of these steps has an associated error, whose variance must be estimated and the sensitivity of the average net radiation to each error must be computed. One application of Earth radiation budget data is to compute the annual mean zonal flux distribution of heat flux by the atmosphere and oceans. This computation was one with theearly radiation budget data set developed by Vonder Haar and Raschke and similar studies have continued with successive data sets. The method requires a net balance of zero radiation, so theimbalance of the data products is quite bothersome. The method developed here removes this problem in an objective manner. Also, the model for the effects of errors permits computation of the zonal distribution of the corrections for net radiation. Typically in remote sensing of the atmosphere, the measurements and algorithms for generating data products have errors which accumulate. The technique presented here may be adapted to many of these problems.