Abstract:Thermal infrared spectrum (TIR, 7-14 μm) has unparalleled advantages for the identification of rock-forming minerals, but currently there are few available satellite borne thermal infrared data sources, and the spatial and spectral resolution of the images is generally low. The thermal infrared spectrometer (TIS) carried by China's Sustainable Development Scientific Satellite-1 (SDGSAT-1), which will be launched in 2021, is characterized by large bandwidth (300 km), high spatial resolution (30 m), and high detection sensitivity, and has a better application prospect in lithologic mapping. In this study, two kinds of thermal infrared images, ASTER and SDGSAT-1, were selected from the East Junggar Kalamari region of Xinjiang to establish rock indices of mafic–ultramafic rocks, quartz-rich rocks, feldspar-rich rocks, and granitoid rocks, and combined with the results of the principal component analysis to analyze the distribution characteristics of the different rock types in the 2D and 3D feature spaces, and constructed the lithology identification models of the two kinds of data in the multi-dimensional spectral feature space, respectively. The results show that: 1) The rock indices constructed from ASTER TIR and SDGSAT-1 TIS data can effectively identify mafic–ultramafic rocks, quartz-rich rocks, feldspar-rich rocks and granites, with an overall accuracy of 95.16% and 98.02% for lithology identification, respectively; 2) The multi-dimensional spectral feature space model constructed with two types of data also has good lithology recognition performance, with overall accuracy improved to 96.78% and 98.54%, respectively; 3) SDGSAT-1 TIS has improved the accuracy of extracting mafic-ultramafic rocks by 13.26% compared to ASTER TIR, and has stronger recognition ability for rocks with smaller outcrops. It has great potential for application in lithology mapping.