Assimilation of ERBE Data with a Nonlinear Programming Technique to Improve Cloud-Cover Diagnosis

Abstract

A method is developed to assimilate satellite data for the purpose of improving the diagnosis of fractional cloud cover within a numerical weather prediction model. The method makes use of a nonlinear programming technique to find a set of parameters for the cloud diagnosis that minimizes the difference between the observed and model-produced outgoing longwave radiation (OLR). The algorithm and theoretical basis of the method are presented. The method has been applied in two forecast experiments using a numerical weather prediction model. The results from a winter case demonstrate that the root-mean-square (rms) difference between the observed and forecasted OLR can be reduced by 50% when the optimized cloud diagnosis is used, with the remaining rms difference within the background noise. The optimized diagnosis also reduces the rms difference in a summer experiment, but the reduction is inadequate, possibly because of the inability of the current cloud scheme to deal with convective activity. The optimization procedure is both stable and sensitive. The largest impact of the optimized cloud diagnosis is on the forecast of surface temperature. The impact on the forecast of other model variables is insignificant. This is partly due to the model's highly simplified treatment of cloud and to the short time of model integration compared to the time scale of radiative forcing. Possible applications and limitations of the method are discussed.