A spatial-spectral analysis is used to demonstrate the surface effects in hyperspectral infrared measurements. Analysis shows that the variations in surface spectral properties cause the substantial spectral variations in measured radiative temperatures in mid (8.0 - 9.0 mm) and short (3.7 - 4.0 mm) wavelength spectral bands with respect to the measurements in the long (10.5 – 13.3 mm) wavelength band. To assess the emissivity spectral effect, the numerical solution of inverse problem of atmospheric remote sensing was used. To improve the accuracy of vertical temperature-moisture profiles retrieved from AIRS sounder infrared measurements, the surface emissivity must be accounted for in the solution of the inverse problem. The accuracy of atmospheric parameters retrieved from IR measurements depends on the measurement accuracy and accurate definition of measurement model. The associated inverse problem based upon the numerical solution of the radiative transfer equation (RTE) is ill posed. Disregarding the spectral-spatial variations of SE in the RTE magnifies the errors. The direct evaluation of SE is used for modeling. An algorithm of solution is presented. The solution includes the surface emissivity , the surface temperature, and the temperature-moisture profile. The RTE equation is solved using method of least squares in coordinate descent on basis of a Gauss-Newton numerical schema. Results of SE estimation are demonstrated. The SE estimates over land show significant spectral-spatial variability.