Longwave radiation, as part of the radiation balance, is one of the factors needed to estimate potential evapotranspiration (PET). Since the longwave radiation balance is rarely measured, many computational methods have been designed. In this study, we report on the difference between the observed longwave radiation balance and modelling results obtained using the two main procedures outlined in FAO24 (relying on the measured sunshine duration) and FAO56 (based on the measured solar radiation) manuals. The performance of these equations was evaluated in the April–October period over eight years at the Liz experimental catchment and grass surface in the Bohemian Forest (Czech Republic). The coefficients of both methods, which describe the influence of cloudiness factor and atmospheric emissivity of the air, were calibrated. The Penman-Monteith method was used to calculate the PET. The use of default coefficient values gave errors of 40–100 mm (FAO56) and 0–20 mm (FAO24) for the seasonal PET estimates (the PET was usually overestimated). Parameter calibration decreased the FAO56 error to less than 20 mm per season (FAO24 remained unaffected by the calibration). The FAO56 approach with calibrated coefficients proved to be more suitable for estimation of the longwave radiation balance.