Vegetation indices derived from remote sensing for an estimation of soil protection against water erosion

Abstract In order to regionalize an erosion model based on a soil loss equation, the type of agricultural landuse and the percentage of ground cover play a major role. A supervised classification of multidate SPOT imagery has been used for mapping the crops, and vegetation indices have been derived from spectral data of each crop class in order to evaluate the soil coverage associated with these classes, in a hilly environment of the Quebec Appalachians in Canada. The relation between ground coverage and vegetation indices for each crop has been obtained by photographic and radiometric measurements on the ground at ten days interval throughout the growing season. Similarity between ground and satellite observations is reasonably good. Results of the field campaign show that, generally, vegetation indices overestimate the ground coverage at the beginning of the growing season and underestimate it at the end, with the appearance of senescence. These data will be integrated into a GIS for spatial mapping and modelling of erosion.