Abstract

Flesh browning (FB), a postharvest physiological disorder, has important economic implications for apple growers and packers. This disorder has hindered long-term storage of ‘Cripps Pink’ apples in Chile, challenging its commercialization worldwide. In order to assess the ability of Vis/NIR spectral information to non-destructively predict the appearance of FB postharvest, fruit from 16 orchards of different growing areas in Chile were evaluated after 180 days in RA storage (0-1°C, 95% RH). Maturity indices and FB incidence were recorded at each time point. Spectral Vis/NIR measurements were performed using a Stellar Net Inc. spectrophotometer (450-1100 nm). Two measurements on the skin of each fruit were taken at each evaluation time. Spectral dataset was pre-processed using the Savinsky-Golay smoothing treatment (45 points), multiple scattering corrections, and filtering through the external parameter orthogonalization (EPO). Principal component analysis and partial least squares discrimination analysis (PLS-DA) were applied (Toolbox 7.5, Eigenvector, USA). The PCA showed a significant cluster separation for apples with and without FB from the Vis-NIR skin spectral data, but only for samples with high (>80%; M2, M3, M9) and low (0-10%; M1, M5, M14) FB incidences. Variability on this dataset was represented by two components with 56.8 and 22.7% of total variance. On the other hand, PLS-DA separated the same categories (with and without FB) with sensitivity and specificity indexes for cross validation of 0.82 and 0.87, respectively. Three wavelength bands (391.5-441.5, 541.5-591.5 and 641.5-691.5 nm) were mostly responsible for total FB discrimination. Similar results were found for diffuse FB discrimination in low (0%; M8, M14, M16) and high (>70%; M9, M10, M12) incidence orchards, but in this case only one important wavelength band was responsible for discrimination (697.5-747.5 nm).