Evaluation of Fourier transform-NIR spectroscopy for integrated external and internal quality assessment of Valencia oranges

Abstract Diffuse reflectance near infrared (NIR) spectroscopy was explored as a non-destructive method to detect external and internal quality of Valencia oranges. The study compared three different Fourier transform NIR acquisition methods, namely, a fibre-optic probe for solid samples (SP), an integrating sphere (IS) and an emission head (EH). Fruit quality attributes measured included mass, colour index, total soluble solids (TSS), titratable acidity (TA), maturity index expressed as TSS:TA ratio and vitamin C. Partial least squares regression was applied to spectral data to develop prediction models for each quality attribute and by randomly dividing the data into calibration and independent validation sets. To test robustness, a set of fruit harvested from another location was used for external validation. Fruit mass, colour index, TSS and vitamin C were predicted with significant accuracy showing RPD-values of 3.53, 1.99, 1.87 and 1.33, respectively. The spectral acquisition method had a significant influence on the calibration regression statistics and accuracy of prediction. The models developed using the EH gave the best prediction statistics for mass ( R  = 0.96, RMSEP = 10.45 g), colour index ( R  = 0.83, RMSEP = 0.82) and vitamin C ( R  = 0.66, 8.01 mg/100 mL), while the IS gave the best prediction for TSS ( R  = 0.83, RMSEP = 0.58). The model parameters remained fairly constant when the models were validated using fruit from another location, indicating high level of model robustness. Good prediction statistics observed when using EH demonstrated the potential of this spectrometer as a non-destructive tool to holistically evaluate external and internal quality parameters.