The Determination of Red Grape Quality Parameters Using the LOCAL Algorithm

Abstract

In a study of visible (vis) and near infrared (NIR) spectra collected from grape samples representing many growing vintages, varieties and regions, WinISI LOCAL performed better than modified partial least squares in the prediction of anthocyanin and pH. However, both algorithms performed equally well for the prediction of total soluble solids (TSS). Calibrations for anthocyanins and pH relied mainly on the visible wavelengths, but pH calibrations may result from pH effects on the spectral properties of anthocyanin. On the other hand, calibrations for TSS relied mainly on NIR wavelengths. An examination of the LOCAL match sets showed that, in addition to spectral matching, there appeared to be concomitant compositional matching, possibly due to dominant spectral features related to the relevant analyte. Observations imply that calibrations for anthocyanins, pH and TSS in grapes are not restrained by vintage, variety and region. A more important factor in determining a robust, global application may be to ensure that the full concentration range of the analyte is well represented in the calibration database.