Abstract: The acquisition of pure vegetation pixel is a necessary step in the extraction of vegetation coverage. However, the vegetation distributes sparsely in arid regions and it is difficult to extract the pure vegetation pixel using low spatial resolution remote sensing data. Consequently, the vegetation coverage extraction accuracy is low. This paper proposed a new method for vegetation coverage retrieval in arid regions using multiscale remote sensing data synergistically. This method used Landsat-8 OLI data, which has a spatial resolution of 30 m, to determine the pure vegetation pixel. In order to eliminate the difference of NDVI from both MODIS and OLI data, induced by the spectral difference of these two sensors, a spectrum transformation procedure was conducted by using ground measured vegetation and soil spectra. A typical arid area located at Fukang City, Xinjiang was selected as the study area to conduct the inversion experiment of vegetation coverage based on this method. In order to verify the accuracy of the new method, validation using the very high resolution aerial image was conducted. The validation showed that the coefficient of determination between the retrieved value and measured value was 0.75, and the RMSE was 0.10. Thus, the method can effectively improve the accuracy of vegetation coverage retrieval and provide a new way to study the ecological environment in arid areas by using low resolution data.

Key words: intercropping, iron, zinc, rhizosphere interactions, biofortification.