Patterns in official food control data – Modelling dioxin and PCB profiling data for authentication of Baltic Sea salmon

Abstract

Fish and fish products are often subjected to food fraud. Due to the significant environmental pollution of the Baltic Sea, fish caught in this area might be labelled as originating from elsewhere. Therefore, analytical proof showing whether fish samples like salmon come from the Baltic Sea or not is of particular interest. Here, official food control data of dioxin and PCB congener concentrations in salmon was used separately and combined as parameters for building one-class classification models using data driven soft independent modelling of class analogy (DD-SIMCA). The training set consisted of Baltic Sea salmon data collected from 2002 through 2019, and the model was tested on Baltic Sea salmon data samples from 2021 and salmon samples coming from China, Chile, Canada, Norway, USA, and Vietnam. The model's performance showed accuracy rates up to 100 %, indicating the model’s future potential to predict whether salmon samples originate from the Baltic Sea or not. This study exemplarily illustrates how the exploration of underlying patterns in official food control data can be used for authentication purposes.