June 1999, 38 pp.
Brightness temperatures derived from polar orbiting satellites are valuable for the evaluation of global climate models. However, the effect of orbital constraints must be accounted to ensure valid comparisons. As part of the AMIP II climate model comparisons, this study seeks to evaluate the bias of possible model output sampling strategies, and whether they can be practically implemented to provide meaningful comparisons with these satellite observations. We compare various sampling strategies with a proxy satellite data set constructed from model output and actual TOVS orbital trajectories, rather than with the observations themselves. To a large extent, this enables isolation of the sampling error from biases caused by deficiencies in the modeled climate processes. Our results suggest that the traditional method of calculating brightness temperatures from monthly mean temperature and moisture profiles yields biases from both nonlinear effects and the removal of the diurnal cycle, that may be unacceptable in many applications. However, our results also suggest that a brightness temperature calculation every six hours of the simulation provides substantially lower sampling biases provided that there are two or more properly aligned satellites. This is encouraging because it means for many applications modelers need not accurately mimic actual satellite trajectories in the sampling of their simulations. However, if only one satellite is available for comparison with simulations, more sophisticated sampling seems necessary. For such circumstances, we introduce a simple procedure that serves as a useful approximation to the rather complex procedure required to sample a model exactly as a polar orbiting satellite does the Earth.(pdf file)