Electronic nose as a non-destructive tool to evaluate the optimal harvest date of apples

An electronic nose (E-nose) has been evaluated for use as a tool to predict the optimal harvest date of apples (Malus domestica Borkh.). The volatiles of 'Jonagold' and 'Braeburn' apples were assessed during the preclimacteric stage for two consecutive harvest years by means of an E-nose. A principal component data analysis indicated the presence of both a year and cultivar effect. Partial least square (PLS) models were constructed based on data of both harvest years. The cultivar effect made it difficult to build accurate and robust models for the two cultivars together. As a consequence, calibration models were constructed based on data of 2 years, but for each cultivar separately. The prediction of maturity, according to the Streif Index, showed a cross-validation correlation of 0.89 and 0.92 for 'Jonagold' and 'Braeburn' fruit, respectively. The calibration models for the prediction of the maturity, defined as the number of days before commercial harvest had a validation correlation of 0.91 for 'Jonagold' and 0.84 for 'Braeburn' fruit. Individual quality characteristics (soluble solids, acidity, starch and firmness) were predicted reasonably well. The calibration model for soluble solids content resulted in a consistent validation correlation over the results over 2 years (0.76 and 0.77). The starch and firmness were predicted with a validation correlation between 0.72 and 0.80. The prediction of the total acidity was poor (validation correlation of 0.66 and 0.69). It was also demonstrated that the type of validation influences the model prediction performance. Care should be taken when interpreting and using the models to predict the optimal harvest date for other years and cultivars. (C) 2003 Elsevier B.V. All rights reserved.