Chemical timber tracing: combining tree-genera information lowers reference data needs and makes harvest location identification more accurate

Abstract Key Message Chemistry-based tracing techniques are increasingly used for combating illegal timber trade, but they are currently limited by the small and fragmented reference datasets available. We introduce a model that integrates data from multiple tree genera while accounting for statistical differences between them. Our model accurately predicts the harvest location even when relevant data are unavailable in some areas, by leveraging data from other genera. Our approach could lower reference sampling costs and enable tracing in situations where new samples cannot be collected, such as during armed conflict. Context Chemistry-based techniques for identifying the harvest location of timber are becoming increasingly important for enforcing timber trade regulations. However, their application has been limited by the need for reference samples from all species across all areas of interest. Aims We investigate whether combining reference data from multiple taxonomic groups can improve timber harvest location determination in regions where reference data is scarce by using the shared natural variability in isotopic composition across species. Methods We extend the harvest location model of Mortier et al. to jointly model isotope ratios and trace element concentrations in wood from different genera. This is achieved by a new covariance function that accounts for shared patterns of spatial variation between genera. We evaluate our approach on 1020 tree samples from four economically important genera ( Betula , Fagus , Pinus , Quercus ) across 12 Eastern European countries. Results The multi-genus model substantially outperforms the single-genus model when little or no data for that genus is available in the focus area. When data from all genera are available across the study area, the multi-genus model achieves similar performance to the single-genus model. Conclusion Our approach strengthens the applicability of timber tracing methods by enabling accurate predictions in areas where sample collection is not currently feasible due to political, logistical and/or security-related challenges, provided that pre-existing samples from other genera are available.