Agricultural practices in grasslands detected by spatial remote sensing.

The major decrease in grassland surfaces associated with changes in their management that has been observed in many regions of the earth during the last half century has major impacts on environmental and socio-economic systems. This study focuses on the identification of grassland management practices in an intensive agricultural watershed located in Brittany, France, by analyzing the intra-annual dynamics of the surface condition of vegetation using remotely sensed and field data. We studied the relationship between one vegetation index (NDVI) and two biophysical variables (LAI and fCOVER) derived from a series of three SPOT images on one hand and measurements collected during field campaigns achieved on 120 grasslands on the other. The results show that the LAI appears as the best predictor for monitoring grassland mowing and grazing. Indeed, because of its ability to characterize vegetation status, LAI estimated from remote sensing data is a relevant variable to identify these practices. LAI values derived from the SPOT images were then classified based on the K-Nearest Neighbor (KNN) supervised algorithm. The results points out that the distribution of grassland management practices such as grazing and mowing can be mapped very accurately (Kappa index = 0.82) at a field scale over large agricultural areas using a series of satellite images.