Multimodal NIR–Micro-XRF Phenotyping Captures Complementary Chemical and Elemental Variation Across All South American Nothofagus

Rapid characterization of conserved tree germplasm can strengthen ex situ conservation as global change intensifies pressures on forest biodiversity. We evaluated whether NIR–micro-XRF phenotyping can resolve complementary signatures across all ten South American Nothofagus species (N. alessandrii, N. alpina, N. antarctica, N. betuloides, N. dombeyi, N. glauca, N. macrocarpa, N. nitida, N. obliqua, and N. pumilio) and the natural hybrid N. × leoni maintained in a common garden. We combined phenetic analysis, element-resolved inference, repeated cross-validated PLS-DA, multimodal fusion, and unsupervised O2PLS. Predictive analyses focused on nine species represented by 12 independent trees each (n = 108); N. nitida, represented by one tree with three technical measurements, was retained for descriptive analyses but excluded from predictive modelling. Full-spectrum XRF outperformed NIR (accuracy 0.822 ± 0.056 versus 0.701 ± 0.064), while block-equalized fusion yielded the highest accuracy (0.842 ± 0.040) but only modest gains over XRF alone. O2PLS showed that 45.8% of NIR and 52.1% of XRF variation belonged to shared structure, which concentrated stronger taxon-associated organization (PERMANOVA R2 = 0.605) than sensor-specific components. Fe-, Zn-, Ca-, and Mn-associated signals showed the strongest taxon differentiation. With validation in independent collections and natural populations, this multimodal framework could support taxonomic verification, phenotypic characterization, and conservation auditing of living collections.