The quantitative estimation of fractional cover of photosynthetic vegetation (f_PV), non-photosynthetic vegetation (f_NPV), and bare soil (f_BS) is critical for grassland animal husbandry and land desertification. Remote sensing is an important tool for estimating the fractional cover of vegetation as a key descriptor of grassland ecosystem function. Developing tools that allow for monitoring of vegetation in space and time is a key step needed to improve management of grassland. The present study describes a method for resolving f_PV, f_NPV, and f_BS in the Xilingol steppe region with MODIS-Terra daily surface reflectance data at 500 m resolution (MOD09GHK). Fractional cover of f_PV, f_NPV, and f_BS was quantified with MOD09GHK data by calculating the Normalized Difference Vegetation Index (NDVI) and the Dead Fuel Index (DFI) and applying a linear unmixing technique. We concurrently analyzed the dynamic change of Xilingol typical grassland of f_PV and f_NPV. Five MODIS images were acquired on April 5, May 30, July 31, August 21, and November 26 in 2014. The approach assumes that cover fractions are made up of a simple mixture of photosynthetic vegetation, non-photosynthetic vegetation, and bare soil. In the present study, one important assumption in our method is that the mixing of fractional cover in NDVI and DFI is linear. DFI is a four-band index that takes into account the differences in spectral features among DFI, photosynthetic vegetation, and bare soil in the VIS-NIR and SWIR wavelength regions, in which the slope of NPV from MODIS band 6 to 7 lies between those of photosynthetic vegetation and bare soil. The correlation between fraction of NPV and DFI was linear. Different end-member extraction methods, including the Pixel Purity Index (PPI) method and 2D scatter plot, were adopted to retrieve the end-member values of photosynthetic vegetation, non-photosynthetic vegetation, and bare soil from NDVI and DFI. The NDVI-DFI feature space follows a triangular distribution, where the vertices represent photosynthetic vegetation, non-photosynthetic vegetation, and bare soil, meeting the essential requirements of the linear unmixing model. NDVI and CAI were calculated for each image and the pixels located close to the vertices of the triangle were located. The spatial location of the pixels identified as pure by the PPI operation and located close to the vertices of the NDVI/DFI triangle was examined by using high resolution imagery (Landsat-8 OLI). Subsequently, we found that vegetation fractional cover can be successfully resolved with MODIS data by combining the NDVI-DFI model. The NDVI-DFI model is suitable for dry season grassland period monitoring of NPV, whereas the grassland growing NPV monitoring is not very sensitive. Additionally, the temporal dynamic of f_PV and f_NPV was confirmed to be consistent with the phonological seasonal change in natural grasslands. The NDVI-DFI model can be applied to the quantitative estimation of fractional cover of non-photosynthetic vegetation (f_NPV) that is critical for grassland desertification monitoring, soil erosion, and grassland grazing. Therefore, the NDVI-DFI model can be used to monitor the temporal and spatial variations of f_PV and f_NPV in the Xilingol steppe regions.