PhD Thesis

Addressing emerging authenticity issues using novel data-driven analytics

Stefan G. Bindereif (04/2017-06/2026)

As globalization continues to grow, trade structures are becoming increasingly complex, which creates numerous challenges for businesses. One of these is ensuring the authenticity of high-priced products. Recent developments have made it clear that established control mechanisms are often inadequate for ensuring protection against fraud or mislabeling, so the need for new analytical solutions is growing rapidly. In its individual studies, this dissertation examines the application of modern data-driven approaches for the authentication of various goods of agricultural and non-agricultural origin. At the heart of all investigations is a software that was developed specifically for this research project but, thanks to its modular design, has great potential for future applications in diverse fields. It enables the automated processing, modeling, and evaluation of spectroscopic and spectrometric data. To date, there has been no solution that enables streamlined analysis of data from, for example, isotope ratio mass spectrometry (IRMS), nuclear magnetic resonance spectroscopy (NMR), near-infrared spectroscopy (NIR), and the associated metadata in a fully reproducible and transparent manner. One area of investigation focused on the authentication of spring barley with respect to the harvest year. Grain samples were taken directly from active trade flows and then analyzed to ensure that they reflected realistic conditions in the supply chain, in contrast to numerous studies that relied on samples from strictly controlled experimental fields and are therefore hardly transferable to real-world conditions. The aim was to develop a novel analytical approach that allows differentiation of the two harvest years 2017 and the extremely dry 2018 for spring barley samples by means of multiple modern analytical techniques and to link the observed metabolic differences to the physiological responses of plants to drought stress. Samples from the 2018 harvest year showed significantly less negative δ¹³C values than those from 2017, likely reflecting reduced stomatal conductance and altered carbon assimilation during the grain-filling stage. These findings were supported by NIR, which showed lower protein, starch, and moisture content and an increase in lipid content for 2018. NMR analysis revealed elevated concentrations of the osmolyte betaine, a well-known biochemical marker for abiotic stress adaptation. By merging the isotope and NIR datasets and applying multivariate statistical analysis, a classification model was built that achieved 100% separation accuracy between the two investigated years. Although not directly linked to authenticity, differences in variety-specific stress responses were also observed, demonstrating the broader applicability of the approach for selecting stress-resilient varieties in the future. Subsequently, the established chemometric framework for authenticity testing was adapted and transferred to a fundamentally different sample matrix, namely gemstones. In this context, demantoid garnets were analyzed, a material whose market value is strongly influenced by its geographical origin. A total of 136 samples from six different countries were analyzed using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS), with the aim of distinguishing origin based on trace element profiles. Logarithmic transformation of the recorded data proved critical for appropriate statistic modeling, enabling clear separation for five of the six countries of origin with classification accuracies being especially high for samples from Namibia, Pakistan, and Madagascar. Analysis of variable importance identified the elements Al and Cr as key discriminants across most models. Importantly, this work was among the first in gemological research to comprehensively address issues of database design, data preprocessing, and model validation, providing a reproducible and transparent workflow that can serve as a blueprint for future studies. In response to a specific industry need in the food sector, the final study of this dissertation was initiated with the intent to trace the processing location of barley malt, an important intermediate product in beer production. Stable isotope analysis of hydrogen and oxygen was used to investigate whether water isotope signatures in malt reflect the location of the malting facility. Therefore, thirty-four barley samples were processed in two different malthouses located approximately 500 km apart. Using IRMS analysis, significant differences in δ²H and δ¹⁸O values between malts from the two investigated locations were observed with all malt samples showing significantly less negative δ²H and δ¹⁸O values compared to the raw material barley. It is the first study to document systematic shifts in isotope ratios during the malting process. Notably, the direction of these differences matched expected regional trends in precipitation isotope values, supporting the hypothesis that local water sources influence the final isotopic signature of the malt. The work illustrates that authenticity verification, when grounded in state-of-the-art analytics and supported by rigorous data science, can evolve into a powerful, flexible toolset applicable across industries. Such capabilities are becoming ever more critical in a world where supply chain complexity and the demand for verifiable product integrity continue to grow.