The purpose of this paper is to assess the spectral Temperature Emissivity Separation algorithm (TES) proposed by Gillespie et al. (1998) as a simple method to retrieve surface emissivity from ground-based measurements. First, we validate different empirical relationships
for the Minimum Maximum Difference module, on which the TES is based, with a large dataset (about 500 surfaces from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) spectral library including man-made materials) for multiband data in the long wave
infrared (LWIR: 7.5-14 µm), and hyperspectral data in the middle wave infrared (MWIR: 3.4-5.2 µm) and LWIR. We show the applicability of TES for hyperspectral data using a specific empirical relationship; this is confirmed by experimental measurements.
For multiband data, we improve the TES for high contrast emissivity surfaces by integrating broadband 8-14 µm measurements in the iterative algorithm. We also found that metals do not confirm these empirical relationships. TES accuracy, extensively assessed by simulations, remains
for multiband simulations (respectively for hyperspectral) within about 0.03 (0.02) for emissivity and within about 1.2 K (0.3 K) for temperature. However, surfaces with low maximum emissivity give higher errors. Except for these particular surfaces, the TES approach,
applied on measurements from a portable multiband thermal radiometer, appears as the most efficient and accurate method for emissivity determination in the field without any a priori assumption on the surface nature.