Abstract Infrared radiance fluctuations in the atmosphere result from fluctuations in the density of atmospheric species, individual molecular state populations, and kinetic temperatures along the sensor line of sight (LOS). As part of the SHARC-4 development program, a model for the prediction of atmospheric background radiance fluctuations was constructed. It predicts a two dimensional radiance spatial covariance function and power spectral density (PSD) from the underlying 3D atmospheric structures. Inputs to the model include the parameters of an atmospheric temperature fluctuation model and vertical atmospheric profiles. The model can be applied for arbitrary sensor viewing geometries, including limb viewing. In the upper atmosphere non-equilibrium effects are important. Fluctuations in kinetic temperature can result in correlated or anti-correlated fluctuations in vibrational state temperatures. The model accounts for these effects and predicts spatial covariance functions and PSD's for molecular state number densities and vibrational temperatures. SHARC is used to predict the non-equilibrium dependence of molecular state number density fluctuations on kinetic temperature and density fluctuations, and to calculate mean LOS radiances and radiance derivatives. The goal is to understand the origin of LOS radiance fluctuations and the prediction of fluctuation statistics based on local fluctuations in gas kinetic temperatures, and densities. The modeling capabilities are illustrated with a sample image prediction for the Midcourse Space Experiment (MSX) having an MSX sensor bandpass and field-of-view (FOV).