This paper analyses high spectral resolution downwelling radiance measurements in the far infrared in presence of cirrus clouds taken by the REFIR-PAD interferometer, deployed at 3500 m above sea level at the Testa Grigia station (Italy), during the Earth COoling by WAter vapouR emission (ECOWAR) campaign. Atmospheric state and cloud geometry are characterised by the co-located millimeter-wave spectrometer GBMS and by radiosonde profile data, an interferometer (I-BEST) and a Raman lidar system deployed at a nearby location (Cervinia). Cloud optical depth and effective diameter are retrieved from REFIR-PAD data using a limited number of channels in the 820–960 cm−1 interval. The retrieved cloud parameters are the input data for simulations covering the 250–1100 cm−1 band in order to test our ability to reproduce the REFIR-PAD spectra in presence of ice clouds. Inverse and forward simulations are based on the same radiative transfer code. A priori information concerning cloud ice vertical distribution is used to better constrain the simulation scheme and an analysis of the degree of approximation of the phase function within the radiative transfer codes is performed to define the accuracy of computations. Simulation-data residuals over the REFIR-PAD spectral interval show an excellent agreement in the window region, but values larger than total measurement uncertainties in the far infrared. Possible causes are investigated. It is shown that the uncertainties related to the water vapour and temperature profiles are of the same order as the sensitivity to the a priori assumption on particle habits for an up-looking configuration. In case of a down-looking configuration, errors due to possible incorrect description of the water vapour profile would be drastically reduced.