Quartzose–mafic spectral feature space model: A methodology for extracting felsic rocks with ASTER thermal infrared radiance data

• We established a quartzose–mafic spectral feature space to discriminate lithology.
• A linear belt of silicate rocks in the feature space was constructed.
• A polygon was deduced in the feature space for extracting felsic rocks.

The original spectral features of felsic rocks are often intermingled with other surface objects, which results in difficulty of detecting felsic rocks using remote sensing techniques. Few felsic rock indices were proposed and visual interpretation with RGB false color composition is widely used to detect felsic rocks. This paper aims to construct a two-dimensional spectral feature space model to extract felsic rocks using ASTER thermal infrared radiance data. The study area is located in northern Qinghai Province, western China with average altitude of approximately 4200 m. A large number of training pixels of mafic–ultramafic rock, quartz-rich rock, felsic rock, carbonate rock and vegetation were selected from the ASTER images as samples of these surface objects. Then we used a quartz-rich rock index (QI, QI = band14 − 0.844 × band12 − 1.897) and a mafic–ultramafic rock index (MI, MI = 0.915 × band10 − band13 + 1.437) to generate a two-dimensional scatter plot. The plot was named after quartzose–mafic spectral feature space (QMFS). The samples show an approximate triangular shape in the QMFS. Mafic–ultramafic rock, quartz-rich rock and carbonate rock are located in separate locations in the three vertex regions, respectively, while felsic rock is located in the central region of the triangle. Next, we calculated a linear belt of silicate rocks in which silicate rocks vary regularly by using a linear regression analysis in the QMFS. Statistical characteristics of the felsic rock samples are analyzed. Afterwards, a polygon which delineates the distribution of felsic rock samples was constructed from the linear belt of silicate rocks. Then we generated a system of inequalities based on the equations of the edges of the polygon. The application of the inequalities to the ASER images shows a good performance of the QMFS for extracting felsic rocks.