Algorithm development strategies for retrieving the surface downwelling longwave flux (SDLW) have been formulated on the basis of detailed studies with radiative transfer models and observational data. The model sensitivity studies were conducted with the column radiation model from the National Center for Atmospheric Research Community Climate Model Version 3 and the Moderate-Resolution Transmittance radiation model. The studies show the clear-sky SDLW can be largely determined from only two parameters: the surface upwelling longwave flux and the column precipitable water vapor. Cloudy-sky sensitivity tests show that, as would be expected, cloud base height is an important factor in determining the SDLW, especially for low clouds. However, when considering broken clouds as occur in reality, there is no way of logically defining an average cloud base height. Instead, cloud liquid water path is shown to be a preferable parameter for use in an all-sky algorithm, not because it serves as a direct cloud input parameter, but rather that it serves as a useful surrogate for cloud base height. Observational data from the Atmospheric Radiation Measurements Program at the U. S. Southern Great Plains (SGP) Oklahoma Central Facility and the Tropical Western Pacific (TWP) Manus Island were used in deriving and validating an illustrative algorithm. The observations show similar relations as found in the model sensitivity tests and suggest that a single algorithm could be applicable for both clear and cloudy conditions as well as for diverse geographical locations. For example, when applied to the TWP data, an algorithm based on a regression of SGP all-sky data produces a relative bias error in SDLW of only 1.4% under all-sky conditions and -0.2% for clear skies. ¿ 2001 American Geophysical Union