The development of hyperspectral technologies in the infrared domain allows new methods of temperature emissivity separation to be elaborated. These methods, which are inappropriate for multispectral measurements, use the amount of information contained in a measured spectrum to increase the number of equations and stabilise the system to solve, or to differentiate the miscellaneous radiance components reaching the sensor thanks to the narrow bands of the spectrometer. Two techniques based on ground measurement are tested. These are the multi-temperature and the spectral smoothness methods. Both need very weak a priori hypotheses. The first one measures the radiance coming up from a surface for several temperatures (or at different times of the day) and solves the over-determinated system of equations. The only hypothesis is the invariance of emissivity between the measurements. The second one uses the difference of spectral variability between an atmospheric spectrum, which is composed of lines, and an emissivity spectrum, which is smoother. Both methods need the radiance at the surface level and the downward irradiance as inputs. Atmospheric corrections have to be made along the upward path between the surface and the sensor, so the atmosphere has to be accurately characterised (especially for the spectral smoothness).
This paper presents a numerical study of both methods (spectral smoothness and multi-temperature method). The effects of the radiative transfer on the retrieved emissivity are analysed and protocols to characterise the atmospheric contribution to the measured signal are described. Finally, a sensitivity analysis and an error budget of the methods are presented.