In-field application of portable NIR to assess 'Valencia' orange fruit maturity

Harvest maturity of orange fruit is currently estimated based on total soluble solids (TSS). Juice TSS is analysed from few samples representing the entire orchard using a destructive method based on a refractometer. The technique restricts the number of representative samples due to its destructive nature. Significant loss of fruit occurs if the number of reference samples is increased to properly represent big farms with varying orchard conditions. This study was conducted to develop models for in situ non-destructive assessment of maturity stage of 'Valencia' orange using a portable visible to near infrared spectrometer (Felix, F750, Vis-NIR) equipped with a xenon-tungsten lamp as light source and lead-sulphide detector. The robustness of models developed in this study was increased by acquiring three spectra from each sample at 10, 20 and 30°C from morning to early evening. Partial least square regression models for assessing TSS, titratable acidity (TA), maturity index (TSS:TA) and BrimA were developed from spectral and chemical data using The Unscrambler^® X chemometric software. The developed models were assessed based on high correlation between values predicted during cross-validation and actual values. Final models were also tested on external fruit set. Good model for assessing TSS was developed as indicated by high regression coefficient (R²=0.988) and the ratio of performance deviation (RPD=9.062), and low root mean square error of prediction (RMSEP=0.114%). Successful models were also developed for predicting TA (R²=0.938; RPD=3.150; RMSEP=0.024%), TSS:TA (R²=0.727; RPD=1.522; RMSEP=0.575) and BrimA (R²=0.905; RPD=3.227; RMSEP=0.255%). Commercial application of portable spectrometers with the developed models can increase the number of assessed harvest maturity parameters and the number of reference samples since this technique does not need harvesting of sample fruits.