Development and validation of relationships between sunflower seed quality and easily measurable seed characteristics for grading purposes

Sunflower seed delivered by farmers is usually traded without taking seed quality into account, owing to the lack of a quick and inexpensive method for quality assessment at delivery points. Seed quality parameters (oil content, protein content, hullability and amount of fine material produced during dehulling) and easily measurable seed characteristics (moisture content, hectolitre mass, thousand seed mass and seed size) were determined in 209 samples from 19 cultivars produced at 11 localities during 1999/2000. Relationships between the seed quality parameters and easily measurable seed characteristics were determined by multiple regression analysis. Although the r2-values were poor, mean absolute errors of estimates were low compared to measured ranges and standard deviations, indicating that these relationships are suitable for quality estimation. During the 2000/2001 season 270 samples of 18 cultivars produced in 15 field trials were analysed and used to compare measured and estimated quality parameters to validate the established relationships. Estimations of all the quality parameters were poor, mainly owing to systematic error indicating that annual reparameterisation of the equations may reduce the error of estimates to acceptable levels. Implementation of such a seed classification system seems unattainable.