Accounting for the nested nature of genetic variation across levels of organization improves our understanding of biodiversity and community ecology

Recent work has demonstrated that the presence or abundance of specific genotypes, populations, species and phylogenetic clades may influence community and ecosystem properties such as resilience or productivity. Many ecological studies, however, use simple linear models to test for such relationships, including species identity as the predictor variable and some measured trait or function as the response variable without accounting for the nestedness of genetic variation across levels of organization. This omission may lead to incorrect inference about which source of variation influences community and ecosystem properties. Here, we explicitly compare this common approach to alternative ways of modeling variation in trait data, using simulated trait data and empirical results of common-garden trials using multiple levels of genetic variation within Eucalyptus, Populus and Picea. We show that: 1) when nested variation is ignored, an incorrect conclusion of species effect is drawn in up to 20% of cases; 2) overestimation of the species effect increases – up to 60% in some scenarios – as the nested term explains more of the variation; and 3) the sample sizes needed to overcome these potential problems associated with aggregating nested hierarchical variation may be impractically large. In common-garden trials, incorporating nested models increased explanatory power twofold for mammal browsing rate in Eucalyptus, threefold for leaf area in Populus, and tenfold for branch number in Picea. Thoroughly measuring intraspecific variation and characterizing hierarchical genetic variation beyond the species level has implications for developing more robust theory in community ecology, managing invaded natural systems, and improving inference in biodiversity–ecosystem functioning research.