Abstract

California is primary producer of almonds worldwide (Prunus dulcis (Mill.); accounting for ~100% of the domestic production and ~80% of world production. Raw almond kernels are subject to the development of concealed damage (CD) which is characterized by a brown discoloration of the kernel interior that occurs with moderate heat treatment (e.g. blanching, drying, roasting). CD is not visible until after the kernels are heat-treated and can therefore result in significant product loss. To date, there are no screening methods available to detect CD in the raw almonds. To address this, we have developed a multivariate model based upon near infrared reflectance spectra to help predict the development of CD in almonds before roasting. This model was developed using partial least square regression (PLSR) to predict almonds classified as either CD or non-concealed damage (NCD). A validation classification error rate as low as 12% was obtained using the wavelength region of 1230 – 2153 nm. A decrease in absorbance in the region between 1230 – 1500 nm and between 1600 – 1800 nm was also found in almonds classified as CD; and may reflect oil oxidation. This model successfully predicts CD in raw almonds and may be a useful screening techniques for predicting CD in raw almonds before heat treatment.