Soybean seed protein and oil concentration spatial variation has been demonstrated at field scale. It would be of interest the use of site properties and aerial images to determine areas within the field of similar seed attributes. If these areas can be identified in advanced developmental stages, then areas to be harvested differentially can be identified. The objective of this is study is to predict soybean yield and seed composition using site properties (elevation, slope, electrical conductivity, soil organic matter and soil reflectance) as well as indices derived from hyperspectral remote sensing images (NNDVI, PRI, NDVI, GNDVI, and the ratio between NDVI and GNDVI). The study was conducted on two adjacent 16 ha subsections in the 2000 and 2001 growing seasons. Since the predictor variables presented multicollinearity three groups of principal components were defined: a group corresponding to the site properties, a group corresponded to vegetation indices, and a group corresponding soybean performance (soybean yield and seed oil and protein concentrations). Then an spatial autoregressive error model predicted soybean performance using meaningful principal components derived from site properties and vegetation indices while accounting for the spatial the autocorrelation. This model predicted the seed protein and oil concentration with R2 of 0.67 and 0.54 for the 2000 and 2001 seasons respectively. Therefore relationships between site properties, hyperspectral vegetation indices and soybean seed protein and oil concentration were established. This methodology can be applied evaluate soil-plant relationships at field scale level when multicollinearity and autocorrelation are present in the data.