Application of composite vegetation index in remote sensing inversion of sparse fractional vegetation cover in Alpine steppe

Based on the Landsat 8 remote sensing image data,a transect of alpine steppe from Shigatse City through Nanmulin County to Shenzha County in Tibet was selected as the research area. Four commonly used vegetation indices were used in this study, including the normalised difference vegetation index (NDVI), soil-adjusted vegetation index (SAVI), modified soil-adjusted vegetation index (MSAVI), and transformational soil-adjusted vegetation index (TSAVI), which is sensitive to vegetation information with low coverage in semi-arid areas. Furthermore, the study improved the bare soil index (BI) and the shadow index (SI) using surface reflectance to construct the vegetation bare shadow index (VBSI) suitable for areas of low vegetation coverage. Then, based on the above vegetation indices, the quantitative inversion of vegetation coverage of the alpine steppe was conducted by the dimidiate pixel model. Finally, the inversion accuracy was analysed using the field vegetation coverage data by the grid method. The results showed that the inversion accuracy of VBSI (TSAVI) was the highest among the eight vegetation indices used, and the inversion accuracy was 85.66%. It was demonstrated that the dimidiate pixel model constructed by TSAVI based on the soil spectrum obtained by field collection has a certain practicality for the inversion of vegetation coverage of alpine steppe. The BI and SI of the improved FCD model can significantly reduce the influence of soil and shade on vegetation information. The constructed composite vegetation index is of great practical significance to extract the vegetation coverage information of sparse alpine steppe.