Validation of Outgoing Longwave Radiation in a General Circulation Model

Abstract

Outgoing longwave radiation (OLR) simulated by the GLAS general circulation model is compared with that derived from measurements made by polar orbiting satellites. The comparison is made for spatial and seasonal variations by expanding the spatial into surface spherical harmonics. The student's t-test is used to determine the statistical significance of differences between the simulated and observed fields. The global mean and spatial standard deviation of the simulated fields are respectively about 20 W m?2 smaller and 10 W m?2 larger than the corresponding values of the observed field. The smaller value of the global mean is due to larger than observed cloud amount in the model. A major fraction of the variability in the simulated field is due to a sharper meridional gradient than is observed. The seasonal variations of the global mean and standard deviation of the simulated fields are in good agreement with observations. Correlation coefficients between the observed and simulated fields as a function of spatial scales show that the phase relationship for large spatial scales is very good for January and July but only fair for April and October. In the tropics, the differences between simulated and observed OLR means are not significantly different from zero, except over regions with deep convection (Asia, Amazon, Central Africa) where the model's convective clouds do not interact with radiation. The differences in the middle and high latitudes are highly significant, more so in the Southern Hemisphere than in the Northern Hemisphere.