On the variability of the net longwave radiation at the ocean surface

Abstract

The bulk formulae (BF) commonly used to estimate the net longwave radiation at the ocean surface (LW⇅) often give dissimilar results for given surface parameters. Because the differences are climatologically significant amounts of energy, it is important to understand the sources of these differences. We present an evaluation of the most widely used BF, in terms of the assumptions made in each, the climatic conditions on which each is based, and on the results of each compared with results computed from the full radiative transfer equation (RTE) with zonally averaged atmospheric data. The differences can best be understood through examination, using the RTE as the basic tool, of the variability of LW⇅ to variations in temperature, humidity, and cloud properties in the atmospheric column as well as at the surface. These calculations reveal that under clear sky conditions, two standard deviation perturbations of either temperature or specific humidity in the atmospheric column above the surface layer can introduce LW⇅, variations of 30–40 W/m². These variations, which would be typical of the synoptic variations of LW⇅ resulting from typical atmospheric variability, are generally greater than the differences produced by using different BF for the same surface conditions. Although the differences between BF LW⇅ and RTE LW⇅ can be large, the results from the Berliand and Anderson formulae duplicate the RTE clear sky LW⇅ to ±15 W/m² under a wide range of mean and perturbed conditions. The RTE studies also reveal that LW⇅ variations due to cloudiness effects can be very large. Low clouds can reduce LW⇅ from clear sky values by as much as 70 W/m². There is strong dependence of LW⇅ on the vertical distribution of cloud properties which renders useless, for estimating instantaneous LW⇅ under cloudy skies, the BF which do not carefully distinguish between cloud types. The accuracy of BF for climatological applications cannot be assessed without information about the geographic and temporal distributions of cloud properties. The computed sensitivity of LW⇅ to variations in the atmospheric column illustrates the type of information and the level of accuracy necessary to attain a particular level of accuracy in LW⇅. Improved remote sensing techniques and improved information about the cloud distribution statistics will lead to better estimates of LW⇅ at the ocean surface and a more realistic parameterization of cloud effects in BF.