Down scaling vegetation fraction by fusing multi-temporal MODIS and Landsat data

Vegetation fraction is an important indicator of ecosystem change, a high spatial and temporal resolution vegetation fraction product was essential in many spatially distributed models. Recent developments of coarse resolution remote sensing (e.g. MODIS) provide the potential to estimate the vegetation fraction with a high temporal resolution. However, coarse resolution products usually provide insufficient spatial resolution to fully characterize the heterogeneity of vegetation fraction at the local scale. Successful downscaling of vegetation fraction to high spatial resolution would be indispensable for describing the vegetation fraction difference in regional studies. A new downscaling approach is developed by fusing multitemporal MODIS and Landsat TM data based on the assumption that a simple scale-invariant linear relationship exist between vegetation fraction and NDVI. Land cover map was incorporated for refining Landsat scale pixels to determine the transformation coefficients. The downscaled vegetation fraction was validated through the comparison with field measured vegetation fraction. The results shows that the new proposed downscaling method was effective, the accuracy of the result was significantly improved, while the vegetation type information was took into consideration.