Abstract: In this paper,we propose methods that combine with high time resolution MODIS images,high spatial resolution ETM+ images and classification data information based on the theory of down scaling in order to use multi-source remote sensing data effectively for fine remote sensing applications.Linear mixed model disaggregate mixed pixels in coarse resolution images.Therefore,when exploiting information within mixed pixels,each component fractional cover is derived from high spatial resolution classification maps.A model between the pixel reflection and class reflection simulates images with high spatial resolution and high time resolution characteristics.The calculated results of the correlation coefficient between the simulated image and the real image,the correlation coefficient between NDVI image produced by the simulate image and NDVI image produced by the real image showed a very high correspondence between them,validating the approach.