Assessment of Titratable Acidity in Fruit Using Short Wave near Infrared Spectroscopy. Part A: Establishing a Detection Limit Based on Model Solutions

Abstract

Spectra (700–2500 nm) of citric acid dissolved in water possess features interpreted as being due to the dilution and perturbation of water OH stretching and combination bands (with the second and first overtones of OH stretching at 960 nm and 1420 nm, respectively, and a combination band of OH vibration at 2132 nm) and also to citric acid OH stretching and combination bands and citric acid CH2 bands (first overtone of CH2 stretching at 1680 nm and a combination band of CH2 at 2284 nm). A searching window method was used to optimise the wavelength region for a partial least squares (PLS) regression model development (around 743–904 nm for the region below 1100 nm). Despite the apparent lack of information below 1100 nm, a PLS regression model based on transmission spectra (743–904 nm) of pure citric acid—water mixtures (standard deviation = 6.65g citric acid 100 mL−1) yielded a root mean square error of cross-validation ( RMSECV) of 0.34% and correlation coefficient of determination ( Rcv2) = 0.998. However, if the matrix was varied with 0, 10 and 13% sucrose or fruit juice, RMSECV increased to 0.79%. This RMSECV value is proposed as a benchmark for attempts to assess titratable acidity of juice based on shortwave (700–1100 nm) near infrared spectra of intact fruit collected using silicon diode array instrumentation.