SEPARATING TEMPERATURE AND EMISSIVITY IN THERMAL INFRARED MULTISPECTRAL SCANNER DATA - IMPLICATIONS FOR RECOVERING LAND-SURFACE TEMPERATURES

This paper evaluates the accuracy of three techniques for recovering surface kinetic temperature from multispectral thermal infrared data acquired over land. The three techniques are the reference channel method, the emissivity normalization method, and the alpha emissivity method. The first two methods have been widely used with data from the thermal infrared multispectral scanner (TIMS); the third is a new method. The methods were used to recover the temperature of artificial radiance data derived from a wide variety of materials and convolved with the filter response functions of TIMS and the advanced spaceborne thermal emission reflectance radiometer (ASTER). ASTER is scheduled to fly on the first EOS platform and will have five channels in the thermal infrared with a spatial resolution of 90 m. The results indicate that the emissivity normalization and alpha emissivity techniques are the most accurate, and recover the temperature of the majority of the artificial radiance spectra to within 1.5K; the reference channel method produces less accurate results. The primary advantage of the alpha emissivity method over emissivity normalization method is that it works well in terrains of widely varying emissivities, e.g., those dominated by vegetation and igneous rocks' By contrast, the emissivity normalization method works well only if the emissivity used for normalization is close to the maximum emissivity of the spectra in the scene. The temperature errors for the ASTER convolved data are greater for the emissivity normalization and reference channel methods than for the data convolved to TIMS. The alpha emissivity method is equally accurate for both instruments. In summary, the new alpha emissivity method described herein, should permit more accurate recovery of land surface temperatures from remotely sensed multispectral thermal infrared radiance data, for materials with widely varying and unknown emissivities.