AbstractA scheme is described for the calculation of the atmospheric infra‐red radiative cooling rate which is suitable for use at all levels. It uses the Curtis matrix method together with the Curtis‐Godson approximation in a form which includes the influence of temperature on line intensity. Doppler effect is included. Because of its simplicity the Goody random model is used for all the absorption bands considered (H2O rotation and 6.3μ, CO215μ, O39.6μ); the use of more accurate models as they become available presents no fundamental difficulty and in any case many of the topics investigated are not particularly sensitive to the spectral model used. A method of reducing the number of spectral intervals is described and its accuracy demonstrated. An accurate method of dealing with diffuse radiation is used to test the diffuse approximation, with the result that the constant factor 1.66 originally proposed by Elsasser is found to be quite adequate for most purposes. Cooling rate errors caused by the following factors are estimated: (1) random and systematic errors in the initial temperature and humidity profiles; (2) the size of the vertical step used in computation; (3) neglecting the temperature dependence of the Curtis matrix; (4) non‐linear effects in forming climatological means. It is shown that the easily computed ‘cooling to space’ is often a good approximation. All the error investigations are based on direct calculations of the cooling profile. Some comparisons with other techniques are presented and a number of examples of meteorological interest are used to illustrate the versatility of the comp