From NIR spectra to singular wavelengths for the estimation of the oil and water contents in olive fruits

Abstract

Knowledge about the oil and water contents in olive fruits is required to determine orchard management, harvest time, and the oil extraction process. The simplification of procedures and of equipment based on NIR Spectroscopy is of major interest. Estimation models for oil and water contents on a fresh matter basis were developed by partial least square regression with NIR spectral data (700 wavelengths). For raw absorbance data the r2 for the test set reached 0.9 and 0.92 for oil and water contents; and RPIQt was 4.9 and 4.3, respectively. The identification of a useful relation of the relative absorbance at 1724 nm and 1760 nm to the oil content allowed for restricting the wavelengths to three. For oil content the r2 showed 0.88 with ad RPIQt of 4.4. For water content the r2 value was 0.84 and the RPIQt was 3.1. Estimation performance with only three wavelengths was comparable to that obtained with PLSR with 700 variables.