Sensitive analysis of various measurement errors on tempearture and emissivity separation method with hyperspectral data

Land surface temperature (LST) and emissivity are required for many applications. Several methods have been proposed to retrieve these two parameters from hyperspectral data, some of which are based on the spectral smoothness of emissivity. To analyze the sensitivity of those methods to various measurement errors, hyperspectral TIR data are first simulated using radiative transfer model 4A/OP (Operational Release for Automatized Atmospheric Absorption Atlas) with different atmospheric profiles and surface parameters, and then the sensitivity of the Downwelling Radiance Residual Index method to different sources of error is analyzed. In terms of resulting errors in LST, results show that: 1) the method is not very sensitive to the uncertainties of atmosphere. An error of 1.47 g/cm² on water vapor content for a sub-arctic summer atmosphere (2.1 g/cm²) only leads to an error of 1.8 K for rock2 (the worst case). 2) Satisfactory results are obtained by this method over heterogeneous land surface. LST retrieval error is less than 0.3 K for all atmospheres.