Vegetation fraction (VF) is the indispensable factor involved in the assessment of land degradation in the inclement climate condition and harsh natural environment. Based on the analysis of an in situ spectral dataset of alpine grasslands on the Tibetan plateau, we assessed the performance of 28 widely used vegetation indices (VIs) and a spectral mixture analysis (SMA) model applied on the analytical spectral device and simulated enhanced thematic mapper (ETM)+ and Huan Jing (HJ)-1 data to select a method for retrieving VF there. The results show that simple VIs are competent for extracting VF information, and VIs with an extra blue band involved will produce a better performance. However, involvement of too many more bands does not yield much higher accuracy, indicated by the fact that hyperspectral VIs are not superior to multispectral ones in our case. The SMA model provides an acceptable accuracy as well but lower than that of VI regression. In addition, the normalized difference vegetation index (NDVI) values of vegetation and soil, generally, as the key parameter in the widely used NDVI-SMA model is obtained, and this would benefit the application of this model to derive VF of alpine grasslands on the Tibetan plateau with minimal or no need for field work support.