Ecophylogenetic Classification of Forest Communities in Georgia

Abstract

The South Caucasus is a biodiversity hotspot characterized by complex biogeographic conditions and a legacy of refugial persistence. Located between the Greater and Lesser Caucasus, the country of Georgia (Sakartvelo) spans a relatively small land area of approximately 69,700 km², yet exhibits an exceptional degree of climatic, geological, and topographic variation. This environmental complexity sustains diverse communities which are characterized by a high degree of species endemism. With approximately 40% of the country’s land area being forested, forest ecosystems extend along multiple biogeographic gradients, exemplifying the region’s ecological diversity. Between 2019 and 2021, these ecosystems were systematically assessed by the Government of Georgia within the framework of a National Forest Inventory for the first time to support forest policy and conservation planning. The resulting dataset offers an unprecedented opportunity to investigate patterns of forest diversity in this underrepresented region and its environmental drivers at national scale.<br /> The field of cophylogenetics explores how evolutionary relatedness shapes species coexistence and biodiversity across environments. This thesis demonstrates how ecophylogenetic concepts can be applied to forest classification and biodiversity monitoring using a real-world, regional-scale case study. Based on systematic sample plot observations from a nation-wide forest assessment and a standardized megaphylogeny, I apply unsupervised cluster analysis based on a phylogenetically informed dissimilarity metric to species compositional data. By evaluating the resulting hierarchical structures, cluster coherence and spatial patterns of grouped species assemblages across environmental gradients, this study provides empirical evidence that quantifying species identity via evolutionary proximity strengthens ecologically meaningful stratification. This integration of ecophylogenetic theory in forest monitoring frameworks demonstrates that moving beyond traditional nominal species classifications offers a more nuanced perspective on community composition.<br /> This cumulative dissertation comprises three research articles. A first comparative analysis (Publication I) shows that incorporating phylogenetic information into quantifying dissimilarity for clustering woody species observations yields higher internal cluster coherence than using species-neutral metrics. In addition, by quantifying phylogenetic variability, mapped species assemblages align more clearly with expected biogeographic gradients, underlining the potential for standardized community classification. Building on this, predictive modeling (Publication II) demonstrates that phylogenetically informed cluster membership can be predicted via proxy variables for climate, soil, topography, and spatial configuration derived from Earth observation data. This predictive potential indicates that species assemblages reflect relative positions along environmental gradients, but these relationships weaken with increasing structural homogenization and the presence of neophytes, reflecting constraints imposed by stand disturbance. Comparing changes in structural stand characteristics along a field-assigned degradation gradient (Article III, under submission) reveals distinct response dynamics and resilience thresholds across phylogenetically informed clusters.<br /> In summary, this dissertation investigates how ecophylogenetic concepts can be integrated into the classification of forest communities. Within the framework of large-scale forest inventories, I operationalize these principles to demonstrate their general applicability and added value for forest classification and biodiversity monitoring. When interspecies phylogenetic variation is quantified for stratification, the spatial distribution of resulting strata aligns with environmental gradients. These findings show that accounting for evolutionary relationships enables the ecological interpretability of community patterns and the environmental factors that shape them. The resulting strata also maintain functional and structural coherence, supporting a broader ecological understanding of community organization. The scalable methodology developed in this thesis can be used to refine forest typologies and to model forest-type-specific dynamics of forest structure and species distribution patterns. This data-driven approach translates ecophylogenetic principles into applied forest monitoring and classification practice, offering a robust framework for regional-scale assessment. Fundamentally, it supports a conceptual shift toward more holistic biodiversity evaluation and forest management in line with current trends in ecological research.