Abstract:

The Vegetation Fractional Coverage (VFC) is a fundamental parameter of climate and hydrologic models.The spatiotemporal distribution and change of the VFC is important for future regional planning and governmental policymaking.Moderate Resolution Imaging Spectroradiometer （MODIS） vegetation index (VI) is an important data to extract global and regional VFC.The objectives of this study are to explore the potential of quantitatively estimating the VFC using timeseries of MODIS 250 m Normalized Difference Vegetation Index (NDVI) 16day data and other auxiliary MODIS data,and to monitor the spatiotemporal patterns of the changes from 2000 to 2009 in the Sanjiangyuan Region.Under the assumption that each pixel of a MODIS image contains two NDVI end members:vegetation NDVI (NDVIveg) and bare soil NDVI (NDVIsoil),this paper improved the determination of parameters NDVIveg and NDVIsoil in the pixel unmixing model.Then based on the timeseries reconstruction of MODIS data,the VFC was quantitatively retrieved using different pixel unmixing models.The retrieval accuracy was assessed at 8872% and the correlation was 08897 for Aug.2007,which suggested this method might be generally useful for VFC estimation at broad spatial scales.The interannual dynamic variability trends and the spatiotemporal landscape patterns of change were characterized over the past ten years based on the yearmaximization vegetation fractional coverage (Mfc).The slope trend analysis and field investigation indicated that the change trend of vegetation was fluctuant and the VFC had been improved in the eastern part while worsened in the western part.The area of vegetation degradation was larger than that of vegetation improvement from 2000 to 2009.The patterns of change revealed that the coverage fragmentation had been decreasing and the complicated degree of the distribution of different grades had been reduced