Lithology classification in semi-arid areas based on vegetation suppression integrating microwave and optical remote sensing images: Duolun county, Inner Mongolia autonomous region, China

Multi-source remote sensing data can provide abundant Earth observation information for lithology classification and identification, especially in some areas with complex geological conditions where the field geological survey is difficult to carry out. Compared with mountainous areas with large outcrops of rock, the lithology information obtained based on traditional field measurement and single optical remote sensing data in semi-arid areas is greatly limited due to the uneven vegetation coverage. While the combination of microwave and optical remote sensing technologies can effectively improve the integrality and reliability of the obtained lithology information in semi-arid areas. This paper selected Duolun County of Inner Mongolia Autonomous Region as the study area, added vegetation suppression for microwave and optical images into the conventional lithology classification process, and integrated Sentinel-1A and Landsat-8 images of the study area to carry out the experiment. The improved water-cloud model with the parameter of vegetation coverage and the method of feature-oriented principal component analysis based on multiple vegetation indices were used to realize the vegetation suppression in Synthetic Aperture Radar (SAR) backscattering images in two polarization modes and multispectral images. The processed SAR backscattering, SAR texture and spectral feature images were used to form seven feature combinations for lithology classification by the maximum likelihood method. The results showed that the proposed lithology classification scheme cannot achieve high-precision classification of all lithologic types, but it was effective in identifying the major lithologies dominated in the Quaternary deposits of the study area. In all feature combinations, the combination of all the three types of features reached the highest classification accuracy, with the overall classification accuracy of 88.60% and the kappa coefficient of 0.75 for five major lithologies. The experimental results fully demonstrated the advantages of integrating microwave and optical remote sensing data in lithology classification in a semi-arid area.