Determination of dry matter and soluble solids of durian pulp using diffuse reflectance near infrared spectroscopy

Fourier transform near infrared spectroscopy was used as a non-invasive technique for the determination of dry matter and soluble solids of durian pulp. A set of 25 fruit was randomly harvested every 10 days, starting from 80 days until 127 days after the onset of fruit development covering six levels of maturity (80 days, 90 days, 100 days, 110 days, 120 days and 127 days). After applying ethephon on the fruit stems, the fruits were kept for 3 days at room temperature and allowed to ripen. Only the pulp of the durian was scanned. The dry matter and soluble solids reference values of the samples were determined by a hot-air-oven method and using a refractometer, respectively. Prediction models using half the samples related the spectral data, and dry matter and soluble solids data were subsequently established using partial least-squares regression and validated using the other half of the samples in a prediction set. A full cross-validation was also generated using all 149 samples. Both the half-of-the-samples model and the all-sample model were then validated using a true validation set of samples collected in a later year. When tested against the validation half of the samples, the half-of-the-samples model predicted dry-matter content with a coefficient of determination (r(2)) and root mean square error of prediction (RMSEP) of 0.89 and 3.60%, respectively, and for soluble solids content 0.55 and 1.63 degrees Brix (Bx), respectively. When tested on samples from a later season, the model for dry-matter content returned an r(2), RMSEP and bias of 0.26, 6.10% and -2.16%, respectively, and for soluble solids content 0.27, 1.25 degrees Bx and 1.09 degrees Bx, respectively. The cross-validated model for dry matter yielded a slightly better r(2) and root mean square error of cross-validation (RMSECV) of 0.90 and 3.58%, respectively, however, the model for soluble solids did not provide a better r(2) and RMSECV: 0.51 and 1.81 degrees Bx, respectively. When tested on samples from a later season, the cross-validated models gave, r(2), RMSEP and bias of 0.15,5.17% and -1.49%, respectively, for dry-matter content, and for soluble solids content 0.37, 1.32 degrees Bx and 1.23 degrees Bx, respectively. The poor results obtained when predicting dry matter in samples in later seasons indicate that samples from several seasons must be included in the set of calibration samples. This is the first report on the application of MR spectroscopy to evaluate the dry matter and soluble solids of durian pulp and could be useful to customers, exporters, importers and also postharvest technologists. However, prediction accuracy was not demonstrated in the model for durian pulp soluble solids, possibly because of the effect of ethephon applied after harvesting to induce ripening within 3 days to make the fruit suitable for consumption. In addition, it was found that the vibration bands of cellulose and fat, and those of aromatic, CH2 and sucrose highly affected the predictions of dry matter and soluble solids in the durian pulp, respectively.