Comparison of emissivity retrieval methods from ASTER data using Fourier-Transform Infrared Spectroscopy

Land surface emissivity retrieval is important for the remote identification of natural materials and can be used to identify the presence of silicate minerals. However, its estimation from passive sensors involves an undetermined function related to radiance data, which is influenced by the atmosphere. We tested three methods for temperature emissivity retrieval in a dune field composed of 99.53% quartz (SiO2) using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) imagery. The tested methods were the reference channel method (RCM), emissivity normalization method (ENM), and temperature emissivity separation (TES) method. An average quartz reference spectrum for the dune samples was calculated from an emissivity database based on temperature and used to evaluate the emissivity products of four ASTER images. In general, the three tested methods had a good approximation when analysed the emissivity reference curve, especially for longer wavelengths that ranged between 2 and 4% of emissivity. The RCM and ENM produced very similar results with the coefficients of determination (R2) as 0.9960 (RMSE 0.0184) and 0.9959 (RMSE 0.0185), respectively. RCM method presented superior results (R2: 0.9960, RMSE: 0.0184), compared to the TES method (R2: 0.9947, RMSE: 0.0197). The TES method showed good results only for shorter wavelengths and, hence, to identify specific targets using ASTER data, such as silicate minerals, it is better to use the RCM method. The emissivity value selected at the saturation point of the spectral library based on temperature is fundamental in acquiring more reliable data.