Contents lists available at ScienceDirect Scientia Horticulturae journal homepage: www.elsevier.com/locate/scihorti Evaluation of NIRS as non-destructive test to evaluate quality traits of purple passion fruit Phonkrit Maniwara a , Kazuhiro Nakano b , Shintaroh Ohashi b , Danai Boonyakiat c,d , Pimjai Seehanam c , Parichat Theanjumpol a,d , Pichaya Poonlarp e, ⁎ a Postharvest Technology Research Center, Faculty of Agriculture, Chiang Mai University, Chiang Mai, 50200, Thailand b Graduate School of Science and Technology, Niigata University, Niigata 950-2181, Japan c Department of Plant and Soil Sciences, Faculty of Agriculture, Chiang Mai University, Chiang Mai, 50200, Thailand d Postharvest Technology Innovation Center, Office of the Higher Education Commission, Bangkok 10400, Thailand e Department of Food Engineering, Faculty of Agro-Industry, Chiang Mai University, Chiang Mai, 50100, Thailand ARTICLE INFO Keywords: Quality assurance Rapid analysis Nondestructive measurement variable selection Spectral pretreatment ABSTRACT Quality evaluation of passion fruit is an important practice before consuming or processing. The fruit’s total soluble solids (TSS), titratable acidity (TA), and pulp content (PC) were predicted by near-infrared (NIR) spectroscopy. Prediction models were constructed by chemometrics of the partial least squares (PLS) regression on the NIR spectra from interactance spectroscopy. Accurate prediction results were obtained and showed high correlations (r) between the predicted and reference values (0.84, 0.91, and 0.99 for TSS, TA and PC, respec- tively). Small standard errors of prediction (SEPs) and bias were also found. A robust prediction model of pulp content provided the greatest value of the residual predictive deviation (RPD = 6.4). Variable selection effec- tively highlighted the important wavelengths and helped to prune the unimportant variables for the TSS, TA and PC produced calibrations with satisfactory results in the predictions (r = 0.84 – 0.98). In conclusion, non- destructive NIR spectroscopy can be a potential predictor for determining purple passion fruit quality. 1. Introduction The nondestructive technique of near-infrared spectroscopy (NIRS) uses light interaction and absorption at wavelengths between 800–2500 nm. Biological products are irradiated by NIRS to provide the organic compound overtones and combinations of vibrational in- formation of OeH, C–H, and NeH chemical bonds (Arendse et al., 2018). The technique has been successfully used to evaluate the in- ternal qualities of various types of fresh horticultural commodities since the first on-line application was implemented a few decades ago (Lammertyn et al., 2000; Porep et al., 2015). There are several test results generated from a variety of spectroscopic techniques and mul- tivariate chemometrics that have demonstrated that NIRS alone is likely to be completely effective in determining the physical and chemical qualities of several fruit types (Arendse et al., 2018; Theanjumpol et al., 2019). During the past two decades, the technique has been successfully studied in terms of its feasibility for predicting the quality of apples (McGlone et al., 2002), Satsuma mandarins (Gómez et al., 2006), pears (Travers et al., 2014), passion fruit (Oliveira et al., 2014; Maniwara et al., 2014), jujube fruit (Guo et al., 2016), pomegranates (Khodabakhshian et al., 2016), mangos (Nordey et al., 2017), and tangerine fruit (Theanjumpol et al., 2019). Jie et al. (2013) successfully used an on-line near infrared diffuse transmission technique to evaluate soluble solids content in thick-rind watermelon and obtained high prediction accuracy (r = 0.84) with low prediction errors. In Thailand, tropical and sub-tropical fruit have been cultivated throughout the highland areas that cover altitudes greater than 800 m above sea level. Valleys and hillsides with different surroundings are widely used to produce tangerines, longans, and passion fruits in par- ticular. Purple passion fruit is enriched in phytonutrients (soluble car- bohydrates and organic acids) and aromatic complexes that greatly attract consumers (Janzantti et al., 2012). The fruit quality typically varies depending on the cultivation site and the growing environment (Bora and Narain, 1997). From the NIRS perspective, developed pre- diction models are typically limited to specific future predictions. The samples should be obtained from a similar planting environment (i.e., the same place where the samples used to develop the prediction model were produced), otherwise, the loss of the prediction model’s robust- ness occur due to the different environmental variables that typically effect the commodity’s structure and/or chemical constituents (Lu https://doi.org/10.1016/j.scienta.2019.108712 Received 2 July 2019; Received in revised form 21 July 2019; Accepted 22 July 2019 ⁎ Corresponding author. E-mail address: pichaya.p@cmu.ac.th (P. Poonlarp). Scientia Horticulturae 257 (2019) 108712 0304-4238/ © 2019 Elsevier B.V. All rights reserved. T