Improved Temperature and Emissivity Separation Algorithm for Multispectral and Hyperspectral Sensors

The Temperature and Emissivity Separation (TES) algorithm was originally developed for the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). This paper focuses on improving the TES algorithm. The main modification is the replacement of the normalized emissivity module with a new module, which is based on the smoothing of spectral radiance signatures. Smoothing is performed by estimating emissivity using an optimized approximation of the relationship between brightness temperature and emissivity. The improved TES algorithm, which is called Optimized Smoothing for Temperature Emissivity Separation (OSTES), was first tested on simulated data from three different sensors, namely, ASTER, Airborne Hyperspectral Scanner, and Thermal Airborne Spectrographic Imager. Processing simulated data with OSTES shows that it produces more accurate and precise temperature and emissivity retrievals than TES for samples with low spectral contrast. These results also show that the OSTES is less sensitive to variations in atmosphere and sample temperatures. The OSTES performance was secondly tested and compared with ASTER standard products. Testing showed that the OSTES temperature retrievals agree with AST_08 (kinetic temperature) but the emissivity retrievals differ substantially from AST_05 (surface emissivity). This implies that the ASTER standard products, AST_08 and AST_05, are not consistent with each other. Emissivities that are derived using temperatures from AST_08 and ASTER thermal surface radiances from AST_09T perform similar or worse when compared with OSTES emissivities. Emissivities delivered by AST_05 also exhibit step discontinuities, whereas the OSTES emissivitites are smoother over homogeneous surfaces. However, all emissivity spectra are affected by imperfect atmospheric corrections.