A comprehensive synthesis of foliar trait responses to global environmental changes
Decades of data on plant growth and function have now accumulated from in situ manipulative experiments that evaluate plant and ecosystem responses to environmental drivers of global change, including elevated atmospheric carbon dioxide concentration ([CO2]), precipitation manipulation, nitrogen (N) addition, and warming. Parallel to these experiments has been the recent consolidation of global databases of plant traits from observational studies. These global trait provide a resource to test the assumption implicit in global change experiments that plant trait responses to experimental treatments are similar in magnitude and direction to trait responses to changes in environmental conditions over time or space in natural and managed ecosystems. We are synthesizing the large body of data from the literature on foliar traits and plant growth or biomass accumulation measured in global change experiments in a growing curated database. We use these data to: (1) analyze the impacts of global change drivers on traits and on trait-trait and trait-growth relationships; and (2) compare trait responses to experimentally induced precipitation and temperature change with trait variation observed across environmental variation in an existing global trait database. These goals are addressed in a cohesive, robust statistical framework using an innovative set of Bayesian hierarchical meta-analysis models.
Our Bayesian approach overcomes methodological limitations of previous meta-analyses of trait responses to global change drivers because it accommodates incomplete reporting, and non-independence of multiple observations reported per study. The hierarchical structure allows us to: (i) incorporate of phylogenetic information, to accommodate taxonomic mismatches between databases of experimental results and global trait observations; and (ii) directly compare trait response effect sizes from experimental manipulations of different global change drivers and quantify the importance of different covariates in explaining experimental effects. Our model of the species-level response of a given trait or growth rate or biomass to environmental variation imposed in global change experiments includes a generalized mixed process model that obtains species-level “latent” trait response estimates conditional on the magnitude of the experimental treatment and covariates. To model the effect of experimental treatments and covariates on latent trait-trait, trait-growth rate, or trait-biomass relationships, the process model represents the latent magnitude of response of a given trait as a function of the magnitude of response of a second trait (or growth or biomass). We compare trait responses to variation in precipitation and temperature and covariates, as measured in global change experiments and observational studies, using similar models.