Non-invasive assessment of fruit: Attributes other than sweetness

Fresh fruit eating quality, as defined by taste, texture (mouth feel) and smell, can be indexed by a number of attributes, including total soluble solids (TSS), total titratable acidity (TTA), firmness and aroma. Eating quality is further defined by a range of internal defects (e.g. dryness defect in mandarin). Available technologies for non-invasive assessment of these attributes are reviewed. The two technologies which have reached a stage of commercial adoption by the fruit industry are short wave near infrared spectroscopy (SWNIRS) and firmness assessment using impact or acoustic based techniques. The SWNIRS technique apparently has utility in the assessment of fruit TSS and dry matter (DM), but literature reports on use for other attributes (e.g. individual soluble sugar levels or firmness) are less convincing (e.g. failing to demonstrate prediction of independent validation sets). Based on the use of the Zeiss MMS1 NIR enhanced spectrometer module, SWNIRS (700 - 1100 nm) was demonstrated to be capable of analysis of citric acid (CA) in aqueous solution with a root mean square error of prediction (RMSEP) of 0.34 % w/v. Difference spectra of pure aqueous solution of CA (i.e. subtraction of the water spectrum) supported interpretation of the CA spectra and partial least squares regression (PLSR) model regression coefficients, with absorption at 970 nm attributed to an O-H stretching band. For starch in aqueous solution, excellent model results (typical root mean square error of crossvalidation (RMSECV) = 0.30 % w/v) were interpreted in terms of scattering caused by the starch grains. The influence of temperature and salt (NaCl) on SWNIR spectra of model solutions was also characterised. Short wave near infrared spectroscopy was used in the development of models for a range of internal quality attributes (TTA, TSS in fruit varying in starch content, firmness, internal flesh colour, maturity level, and flesh ‘dryness defect’) of intact fruit. In each case the wavelength range and the number of factors used in the PLS model was optimised. In general an interactance mode was adopted, but in some cases presentation geometry (angle between light source, sample and detector) was also optimised. The SWNIRS technique was demonstrated to be effective in the assessment of DM content of a number of commodities (typical RMSEP around 1% DM). Sorting on DM spectra was shown to allow for removal of immature mangoes (fruit that will be slow or fail to ripen). Further, it was demonstrated that spectra collected of hard green mango could be directly related to TSS of fully ripe fruit. The SWNIRS technique was also demonstrated to be effective in the assessment of mango fruit internal colour (as flesh Hunter L a b). However, the SWNIRS technique was not recommended for assessment of TSS of intact fruit of varying starch level (i.e. in ripening mango or banana). In banana, for example, the PLSR model on TSS was interpreted in terms of assessment of peel chlorophyll content, representing an indirect assessment of TSS. With an RMSECV >0.1 % and a RMSEP = 0.3 % w/v, SWNIRS models were of marginal value in prediction of TTA of high acid fruit (e.g. lime, x ± standard deviation (SD): 7.3 ± 0.51 %), and of no value in prediction of low TTA fruit (e.g. peach, x ± SD: 0.88 ± 0.17 %). The SWNIR calibration models on fruit firmness achieved a Rcv2 >0.8, but in prediction of independent sets Rp 2 was <0.7. An acoustic technique based on sound velocity (SV) was better suited to assessment of fruit firmness. The SV decreased during ripening in mango, banana, peach and tomato fruit. The rate of this change was different to that of a penetrometer assessment, indicating that the two methods are assessing different mechanical properties of the fruit. A dryness defect of cultivar (cv.) Imperial mandarin was associated with cell proliferation within the juice sacs, and this character was associated with the observed colour of the juice sacs (high luminosity value). It was hypothesised that this character would decrease light transmission through affected fruit. However, fruit juice sac luminosity also varied with fruit maturity, and SWNIRS model performance was not consistent. Practical implementation of the SWNIRS technique for on-line sorting of defect fruit would involve constant model updating. In conclusion, the factors contributing to a successful implementation of the SWNIRS technique to a given application are summarised, and future directions in instrumentation and chemometrics discussed.