A land cover classification product over France at 1 km resolution using SPOT4/VEGETATION data

Abstract

The present study focuses on the development of a new land cover classification product over France at 1 km resolution. It is based on data sets from the Earth observing system SPOT4/VEGETATION. The satellite measurements are aimed at supporting regional efforts to set up global mosaics on new land cover products. They have been acquired in the frame of the Global Land Cover 2000 project. The instrument design relies on advanced technology, which leads to an improved radiometric and geometric resolution data. Such characteristics allow taking full benefit of the daily repetitiveness of the VEGETATION wide field-of-view sensor without the drawback of a variable pixel size on the image edge. Several physical processing steps are successively operated to the images on a per-pixel basis to remove detector blindness, to filter cloud contamination, and finally to correct both atmospheric and surface anisotropy effects. A new thematic map using the K-means clustering method has been built. First, the results of the satellite-based land cover classification has been successfully compared with the Coordination of Information on the Environment (CORINE) database which serves as a reference to appraise the reliability of the study. Then, it has been inter-compared with land cover products derived from MODIS and AVHRR sensors. For this, an aggregative scheme particularly focused on major land units (forest, grassland, cropland) adopted in order to yield a whole mapping at the same geographic projection and space resolution. The discrepancies between maps enhance the quality of the proposed product, thanks to the use of advanced data processing and a more appropriate method.