Quality evaluation of intact açaí and juçara fruit by means of near infrared spectroscopy

Abstract

The objective of this study was to report the robustness of partial least squares regression (PLSR) models developed using FT-NIR reflectance spectra obtained from intact açaí and juçara fruit. Mature fruit were collected over two years (6 populations of açaí and juçara, totalling 505 samples). Diffuse reflectance spectra were acquired (64 scans and spectral resolution of 8cm-1) using ~25 fruits per batch on a 90mm diameter glass dish in a single layer. Spectra were subject to several pre-processing procedures and two variable selection methods to develop the PLSR models. For total anthocyanin content (TAC) in açaí, a PLSR model developed using the wavelength range of 1606-1793nm, standard normal variate (SNV) and second derivative of Savitzky-Golay (SNV+d2A) achieved a bias corrected root mean square error (SEP) of 3.6gkg-1 and a R2p of 0.7 in predicting an external independent set, which was better than PLSR models for juçara (SEP of 3.7gkg-1, R2p of 0.5), and for both species combined (SEP of 5.7gkg-1, R2p of 0.5). For soluble solids content (SSC) in açaí the models developed using SNV+d2A spectra over the window of 1640-1738nm achieved a bias-corrected SEP of 2.9% and R2p of 0.8, similar to juçara (SEP of 1.1%, R2p of 0.9) and for both species combined (SEP of 2.3%, R2p of 0.8). The developed models can be used to sort açaí and juçara based on SSC and TAC into two grades (low and high contents).