Among the surface radiation budget components that are needed to understand the climate of the Earth, the thermal atmospheric radiation component is the only one that is not measured routinely, so it must be estimated indirectly. The presence of clouds substantially hampers this task, supplementing the emission of water vapour and carbon dioxide in the lower atmosphere. The methods proposed to estimate thermal atmospheric radiation often rely on a cloudless sky model and a corrective term that takes into account the cloud contribution. Several models for estimating the thermal atmospheric radiation under cloudy conditions have been tested using measurements recorded in Granada (Spain) during a period of 2 years. The cloudless sky model provides estimation with an error of about 5 per cent, whereas the cloudy skies models, after some modifications, estimate thermal atmospheric radiation with an error of about 6 per cent. Our results show the convenience of a quadratic dependence on cloud amount for cloudy skies models. In addition to this, it seems convenient to use available information about the radiative properties of clouds to assign different cloud emittances for different types of clouds within each level, especially for middle level clouds.