Grain quality is typically measured via several attributes. As these attributes vary across shipments and time, grain quality can be described using multivariate probability or frequency distributions. These distributions are important in modeling blending opportunities inherent in various grain shipments. For computational reasons, it is usually necessary to represent these distributions with a small set of discrete points and probabilities. In this analysis, we suggest a representation method based on Gaussian quadrature. This approach maintains the blending opportunities available by preserving moments of the distribution. The Gaussian quadrature method is compared to a more commonly used representation in a barley blending model.