Prediction of storage quality of ‘Cox’s Orange Pippin’ apples from nutritional and meteorological data using multiple regression models selected by cross validation

An experiment was carried out over a six year period to relate the storage quality of ‘Cox’s Orange Pippin’ apples to the mineral composition of leaves, fruitlets and fruits and to the climatic conditions during development. The apples were sampled from 24 orchards over a wide geographic area. The extent to which physiological disorders, bitter pit and low temperature breakdown (LTB), and quality attributes, greenness and firmness, of stored fruit could be predicted either before or at harvest was determined. For each response variable, a short-list of up to 12 potential explanatory variables was selected from an initial set of up to 54 possible explanatory variables, using stepwise regression techniques. Predictive models were generated by selecting variables from the initial short-list using a cross validation technique. The average incidence of bitter pit in controlled atmosphere (CA) storage at 1.2% O2 (<1% CO₂) was 4% compared with 12% for similar samples stored in air. Predictive models accounted for up to 67% of the variance in bitter pit in air-stored fruit but the percentage was lower (39%) for fruit stored in CA. The concentrations of calcium and potassium in fruit were particularly important for predicting bitter pit in air-stored fruit but less so for fruit stored in CA where more explanatory variables were required for prediction. Fifty one per cent of the variance of LTB in fruit stored in CA was accounted for by a nine-variable model incorporating three climatic variables. A simple three-variable model (leaf nitrogen concentration, harvest firmness and fruit dry weight) accounted for 67% of the variance in ex-store firmness and the addition of other nutritional and climatic variables increased the percentage variance accounted for to 76%. Seventy-one per cent of the variance in the greenness of fruit after CA storage was accounted for by a six-variable model. The possible physiological significance of the explanatory variables in each model is discussed.