Effect of cultivar and season on the robustness of PLS models for soluble solid content prediction in apricots using FT-NIRS

FT-NIR models were developed for the non-destructive prediction of soluble solid content (SSC), titratable acidity (TA), firmness and weight of two commercially important apricot cultivars, “Hacihaliloglu” and “Kabaasi” from Turkey. The models constructed for SSC prediction gave good results. We could also establish a model which can be used for rough estimation of the apricot weight. However, it could not be possible to predict accurately TA and firmness of the apricots with FT-NIR spectroscopy. The study was further extended over 3 years for the SSC prediction. Validation of the both mono and multi-cultivar models showed that model performances may exhibit important variations across different harvest seasons. The robustness of the models was improved when the data of two or three seasons were used. It was concluded that in order to developed reliable SSC prediction models for apricots the spectral data should be collected over several harvest seasons.