Prediction of post-storage quality in canning apricots and peaches using near infrared spectroscopy (NIRS) and chemometrics

[摘要] ENGLISH ABSTRACT: Post-storage quality of the stone fruit, apricots and peaches, is the major factor determiningtheir suitability for canning after cold storage in South Africa. Short harvesting periods andthe limited capacity of the factory to process the large quantities of fruit within two days afterdelivery, necessitates cold storage until canning. Apricots develop internal breakdown,whereas peaches develop internal breakdown accompanied by loosening of the skin andadhesion of the flesh to the stone. The deterioration takes place within the fruit during a coldstorage period of one to two weeks. The tendency of the fruit to develop internal defects can,to date, not be identified prior to storage and are only discovered after destoning duringcanning. Near infrared spectroscopy (NIRS) combined with chemometrics were investigatedas a non-destructive method to predict post-storage quality in Bulida apricots and clingstonepeach cultivars. Near infrared (NIR) spectra (645-1201 nm), measured on the intact fruit justafter harvesting, were correlated with subjective quality evaluations performed on the cut anddestoned fruit after cold storage. The cold storage periods for apricots were four weeks (2002season) and three and two weeks for peach cultivars for the 2002 and 2003 seasons,respectively. Soft independent modelling by class analogy (SIMCA) and multivariate adaptiveregression splines (MARS) were applied to the spectral and reference data to develop modelsfor good and poor post-storage quality. The ability of these models to predict post-storagequality was evaluated in terms of recognition (sensitivity) and rejection (specificity) of thesamples in independent validation sets. Total correct classification rates of 50.00% and69.00% were obtained with Bulida apricots, using SIMCA and MARS, respectively.Classification results with apricots showed that MARS performed better than SIMCA and isthus recommended for this application. Total correct classification rates of 53.00% to 60.00%(SIMCA) and 57.65% to 65.12% (MARS) were obtained for data sets of combined peachcultivars within seasons and over both seasons. Additional aspects of fruit quality wereinvestigated to identify possible indices of post-storage quality. Classification trees were usedto find correlations between the post-storage quality and the fruit mass, diameter, firmnessand soluble solids content (SSC). Among these, fruit diameter and firmness were the majorindices of post-storage quality. Accurate predictions of firmness could not be achieved bynear infrared spectroscopy (NlRS), making the combination of NIRS and classification treesnot yet suitable for predicting post-storage quality. NIRS was further used to predict poststorageSSC within seasons in Bulida apricots and intact peach cultivars. This confirmedsufficient NIR light penetration into the intact fruit and also provided a further application ofNIRS for ripeness evaluation in the canning industry. Validations on peach samples obtained correlation coefficients (r) of 0.77-0.85 and SEP-values of 1.35-1.60 °Brix using partial leastsquares (PLS) regression. MARS obtained r = 0.77-0.82 and SEP = 1.42-1.55 °Brix.Predictions of sse in apricots were less accurate, with r = 0.39-0.88, SEP = 1.24-2.21 °Brix(PLS) and r = 0.51-0.82, SEP = 1.54-2.19 °Brix (MARS). It is suggested that the accuracy ofsse measurements, and the subsequent predictions, were affected by the cold storageperiods as well as internal variation within the fruit. This study showed that a combination ofNIRS and chemometrics can be used to predict post-storage quality in intact peaches andapricots. A small scale feasibility study showed that 4% (R117 720) (apricot industry) and 3%(R610 740) (peach industry) of production losses can be saved if this method is implementedin the South African canning industry. Although it was difficult to assign specific chemicalcomponents or quality attributes to the formulation of the storage potential models, importanthidden information in the spectra could be revealed by chemometric classification methods.NIRS promises to be a useful and unique quality evaluation tool for the South African fruitcanning industry. Several recommendations are made for the canning practices to reducelosses and for future research to improve the current prediction models.

[发布机构] Stellenbosch University