In order to estimate radiative fluxes from satellite radiance measurements, angular distribution models (ADMs) are required to account for the angular variation in the Earth's emitted radiation field. To generate data for mapping the Earth's radiation field, the Clouds and the Earth's Radiant Energy System (CERES) instrument on the Tropical Rainfall Measuring Mission (TRMM) satellite has the ability to rotate in the azimuth plane (RAP) as it scans in elevation. CERES can also scan in elevation at a fixed azimuth, such as in the alongtrack and crosstrack directions.

In this study, eight months of CERES RAP and alongtrack data are used to construct empirical ADMs for broadband longwave and window (8-12 micron channel) radiance fields. ADMs are divided into scene types by stratifying the radiance measurements into fixed percentile intervals of atmospheric and cloud properties (e.g. precipitable water, cloud emissivity, surface/cloud temperature differences) for various cloud cover conditions (e.g. clear, broken cloud fields,overcast) and surface geographical types (e.g. ocean, land and desert). Compositing the data in terms of fixed percentile intervals rather than fixed absolute values ensures consistent sampling in all angles. For broken cloud and overcast scenes, the ADMs are shown to be most sensitive to changes in cloud emissivity, while ADMs for clear scenes exhibit the largest variation with changes in atmospheric lapse rate. ADMs for the window channel are shown to be more anisotropic than longwave ADMs, particularly over warm surfaces such as clear desert. In order to assess the performance of the empirical ADMs, the longwave and window fluxes derived from applying these ADMs are compared with results obtained by sorting the radiance measurements by viewing zenith angle in finite angular space intervals and directly integrating for fluxes.