The relationship between intra-sample variability and crop product quality

The observation that food technologists attempt to minimise seed character variation when using crop products to produce foodstuff stimulated the hypothesis that knowledge of the nature, properties and origins of intra-sample variation would aid prediction of crop product quality. The hypothesis was tested in 3 case studies. Pea seed colour variability was shown to be useful for predicting a pea colour acceptability score derived from ranking studies. In a second case study, wheat seed size variation was shown to have value for predicting flour yield in a number of cultivars. Knowledge of grain protein content improved predictability of flour yield. In one cultivar (Batten) subjected to a wide range of environmental stresses, seed size variability measures, in combination with grain protein content, predicted flour yield, flour water absorption and dough work input requirement as well as current laboratory procedures. Other quality characteristics were predicted well enough for quality segregation purposes. The sources of wheat seed size variability were examined in a head dissection experiment, and observed patterns found to be primarily under genetic control, with random variability restricted to the head-to-head component. A more elaborate experiment is proposed to allow the construction of a cultivar seed size variation "template" against which the random component can be determined more accurately. In the third case study, the relationship between maltability and water uptake variation in barley was investigated. Although a strong relationship between bulk water uptake and maltability was found, and a two-phase seed water uptake model was supported, derived variability statistics could not be used to give useful predictions of maltability. It is postulated that insufficient data were collected to allow the detection of all the double normal distributions that exist (as predicted by the two-phase water uptake model), and to derive reliable estimates of sub-distribution statistics. The results of this case study therefore neither support or refute the hypothesis. More detailed experiments are proposed to obtain more reliable variability estimates against which to compare maltability values. Overall, the data obtained give good support for the hypothesis, but expose the need to identify the fixed, genetically-determined components of variation, so that they can be separated from the random components likely to be the best predictors of environmental impact on quality.