Statistical Estimates of Monthly Mean and Interannual Changes of Radiation Fluxes at the Top of the Atmosphere

Multiple linear regression is used in relating monthly means and year-to-year changes of the monthly mean planetary albedo and infrared flux leaving the atmosphere, as measured by NOAA satellites, to certain meteorological quantities. Physical predictors are selected which are seen as likely to influence cloudiness, such as temperature, relative humidity, and wind. Predictors of this type can be readily obtained from numerical models. Forty-two months of polar orbiter measurements of radiation fluxes and objective analyses for the National Meteorological Center's operational model are related. Continental and oceanic samples are evaluated separately. Checks on the model consist of independent tests and a comparison with estimates of radiation fluxes in which the predictors are functions of latitude, longitude, and time of year only. It is noted that physical predictors are consistently superior, with the single exception of oceanic albedo, where little difference is seen. For the infrared flux, 93 and 84% of the variance in the monthly means is explained over land and ocean, respectively. The Planck function computed from a humidity (cloud) sensitive radiating temperature is found to be the dominant predictor, with other humidity predictors also useful.