The use of a finite mixture distribution model in bottled water characterisation and authentication with stable hydrogen, oxygen and carbon isotopes — Case study from Slovenia

This study investigates the comparison of the stable isotopic composition of oxygen (δ18O), hydrogen (δ2H) and carbon in dissolved inorganic carbon (δ13CDIC) in domestic and foreign bottled waters randomly collected from the Slovene market. Empirical distributions of δ2H, δ18O and δ13CDIC were examined. Mathematical models of unimodal and polymodal probability density functions were applied for their description. A normal (Gaussian) finite mixture model as a representation of the polymodal empirical distribution of δ13CDIC was used as a classification tool. Finally, based on the classification, descriptive statistics of δ2H, δ18O and δ13CDIC for different groups are calculated. Components of the normal finite mixture model of δ13CDIC represent four groups of bottled water; a group of artificial sparkling bottled water, two groups of still and flavoured bottled water and a group of bottled water originating from deeper aquifers. Corresponding values of δ2H and δ18O show that only the group of bottled water originating from deeper aquifers has a distinct stable isotopic composition, where the others have similar empirical distributions. The proposed model showed that δ13CDIC for bottled water can be used as a strong discrimination parameter for the classification of bottled water and it can be applied as an authentication tool.

Research Highlights
► Bottled water brands have very distinctive values of δ2H, δ18O and δ13CDIC.
► Stable isotope analyses can be used as a classification tool for bottled water.
► Normal finite mixture model of δ13CDIC represents good classification approach.
► Variables of δ2H and δ18O can be described with normal probability model.
► Natural sparkling bottled water has the most distinctive isotope characteristics.