This study investigates the potential of using SPOT multispectral images of agricultural fields to distinguish spatial variation in crop-growth patterns that can be used for site-specific agricultural management. Four years of SPOT data from 1995 to 1997 and 1999 are used in this study over south-western Sweden where satellite derived yield maps are compared to data from commercial yield mapping systems. The crops included rape, wheat, barley, meadow fescue [Festuca pratensis], oats, peas and rye. Our concept of crop growth maps is intended for use in areas where yield mapping, soil sampling and ground spectral measurements are not available. Maps of crop growth variability produced by clustering processes applied to images of the Normalized Difference Vegetation Index are compared to clustering of yield maps from the same years. Qualitative yield estimation is derived by dividing each field into several thematic classes, going from lowest to highest potential yield within a particular field. Qualitative comparisons are made within each field. For one year, the satellite data are also compared to three traditional yield maps derived from the same set of yield data. For a few fields where the time of image acquisition coincides with stages of optimum grain fill, high correlations were obtained between yield and NDVI. This study illustrates that satellite images can be a useful tool in precision agriculture management. The clusters created from the NDVI images show similar patterns as clusters created from the yield maps.
