Sensitivity of iterative spectrally smooth temperature/emissivity separation to algorithmic assumptions and measurement noise

Iterative spectrally smooth temperature-emissivity separation (ISSTES) is an algorithm proposed by C. Borel [1], [2] for retrieving surface temperature and emissivity from remotely sensed hyperspectral thermal infrared radiance. In this paper, uncertainty in retrieved temperature and emissivity will be formulated for two error sources: algorithmic error caused by departure of real materials from the smoothness assumption upon which the algorithm is predicated and measurement noise (uncorrelated Gaussian measurement errors). These uncertainties were then evaluated for the SEBASS LWIR instrument, for altitudes from ground level to 10 km, through stressing atmospheric conditions. Resulting algorithmic error was small for all but one material in the Spectral Library of Johns Hopkins University, Baltimore, MD. For all other materials, algorithmic error was found to increase with altitude up to about 2 km and then level off. The same behavior was found for the retrieval error due to noise.