A rapid and reliable method for the measurement of water content in extra virgin olive oil

Extra virgin olive oil (EVOO), although mainly constituted of triacylglycerols, is not considered as homogeneous medium since traces of water and numerous minor components are transferred from fruits into the oil during the technological process. Amphiphilic compounds like mono- and diacylglycerols, free fatty acids, phospholipids and sterols present in EVOO, aggregate spontaneously due to hydrophobic interactions, to form colloidal associations and traces of water can alter their structures and characteristics. In addition, polar compounds like phenolics and traces of aminoacids and sugars due to their hydrophilic character, are expected to be preferably located inside the colloidal structures formed in the presence of endogenous amphiphiles and water (Xenakis et al. 2010). Based on these consideration, a rapid and reliable method for the measurement of water content in EVOO is needed. Actually, the most commonly accepted methods for measurement of water content in EVOO are Karl Fischer titration (AOAC, 1998) and drying by heating (both moisture and volatile content) (ISO 662, 1998). These methods are time-consuming, and usually require large amounts of reagents. Dielectric techniques have been studied to assess quality parameters, such as chemical composition, and oil adulteration. Capacitive methods based on parallel plates or other types of capacitive sensors have been previously explored to predict water content of EVOO, but only selected frequencies and simple linear regression analysis were used (Ragni et al., 2012). The present research investigates the correlations between water content and capacitance measured in frequency sweep. The reference value of the moisture content was carried out by Karl Fischer titration, drying by heating at oven and by thermobalance on ten Italian EVOOs. On the basis of the previous study, the frequency sweep measurements were carried out from 1 kHz to 9 kHz by using a multi-plate variable capacitor probe characterized by fixed and mobile armatures. Predictive models were build computing the obtained spectra by Partial Least Squares Regression (PLS) and artificial neural networks (ANN). Good correlations between capacitance and moisture content were obtained by both methods.