Surface emissivity and temperature retrieval for a hyperspectral sensor

Abstract

With the growing use of hyperspectral imagers, e.g. AVIRIS in the visible and short-wave infrared there is hope of using such instruments in the mid-wave and thermal IR (TIR) some day. The author believes that this will enable us to get around using the present temperature-emissivity separation algorithms using methods which take advantage of the many channels available in hyperspectral imagers. A simple fact used in coming up with a novel algorithm is that a typical surface emissivity spectrum is rather smooth compared to spectral features introduced by the atmosphere. Thus an iterative solution technique can be devised which retrieves emissivity spectra based on spectral smoothness. To make the emissivities realistic, atmospheric parameters are varied using approximations, look-up tables derived from a radiative transfer code and spectral libraries. One such iterative algorithm solves the radiative transfer equation for the radiance at the sensor for the unknown emissivity and uses the blackbody temperature computed in an atmospheric window to get a guess for the unknown surface temperature. By varying the surface temperature over a small range a series of emissivity spectra are calculated. The one with the smoothest characteristic is chosen. The algorithm was tested on synthetic data using MODTRAN and the Salisbury emissivity database.