Within- and between-site variability of δ18O, δ13C, and δ15N in Amazonian tree rings: Climatic drivers and implications for geographic traceability

Illegal logging in the Amazon rainforest remains a major environmental challenge, contributing to widespread forest degradation and undermining global conservation efforts. Among emerging forensic tools for combating illegal timber trade, stable isotope analysis has shown promise for tracing the geographic origin of wood. In this study, we assess the spatial variability and environmental controls of δ¹⁸O, δ¹³C, and δ¹⁵N in tree-ring cellulose across 249 trees sampled at 21 sites in the Brazilian Amazon. We analyze intra-tree isotopic variation using wood samples from five radial positions per tree and evaluate variance components using Bayesian mixed-effects models. Our results reveal that isotopic variation across radial positions is relatively small compared to within- and between-site variability, with radius 4 (near the sapwood-heartwood boundary) providing a representative and forensically practical sampling location. Variance partitioning shows that between-site differences account for the largest share of isotopic variation, although within-site variability—driven by species identity and microclimatic factors—remains substantial, especially for δ¹⁸O and δ¹³C. Power curve analyses suggest that sampling approximately 10 individuals per site suffices for δ¹³C and δ¹⁵N, while δ¹⁸O requires more extensive sampling. Random forest models incorporating climatic, topographic, and physiological predictors explained up to 73 % of the variance in δ¹⁵N but were less effective for δ¹⁸O and δ¹³C. These findings provide critical insights for designing stable isotope-based timber traceability systems, highlighting the need for robust sampling strategies and the inclusion of species-specific traits to enhance model performance.