Predicting woody species assemblages using ecophylogenetics and Earth observation data

Organizing species assemblages based on compositional characteristics enables the identification of ecologically meaningful patterns in biodiversity and supports forest diversity monitoring, conservation, and management. In this context, ecophylogenetics offers powerful opportunities by exploring how evolutionary relationships between species reflect community distributions within ecological space. Using national forest inventory data of Georgia (Sakartvelo), we classify woody species assemblages based on interspecies phylogenetic dissimilarity and evaluated whether cluster membership could be predicted from multivariate Earth observation data describing site-specific environmental conditions. Principal components of 30 explanatory variables were used to model class membership across three sample groups with increasing disturbance levels. Prediction accuracy reached 53.6 % (OOB error 46.4 %) for undisturbed samples, 67.5 % for disturbed (OOB 32.5 %), and 45.7 % for disturbed samples with neophytes (OOB 54.3 %), based on 12, 6, and 5 clusters, respectively. The decline in classification accuracy with increasing disturbance reflects compositional homogenization and a weakened alignment of the phylogenetic signal with environmental gradients. Our findings demonstrate that incorporating phylogenetic variability in the classification of woody species assemblages enables coherent clustering and effectively captures distributions along environmental gradients particularly under low-disturbance conditions. This approach offers a solid framework to improve forest community classification and to support sustainable forest and conservation management . • Phylogenetic classification of species aligns with environmental gradients. • Species assemblages can be predicted from EO data using Random Forests. • Ecophylogenetics improves forest classification for biodiversity and conservation. • First phylogenetic analysis of national forest inventory data from Georgia.