Common garden comparisons confirm inherited differences in sensitivity to climate change between forest tree species

The natural distribution, habitat, growth and evolutionary history of tree species are strongly dependent on ecological and genetic processes in ecosystems subject to fluctuating climatic conditions, but there have been few experimental comparisons of sensitivity between species. We compared the responses of two broadleaved tree species (Fagus sylvatica and Quercus petraea) and two conifer tree species (Pinus sylvestris and Picea abies) to climatic transfers by fitting models containing the same climatic variables. We used published data from European provenance test networks to model the responses of individual populations nested within species. A mixed model approach was applied to develop a response function for tree height over climatic transfer distance, taking into account the climatic conditions at both the seed source and the test location. The two broadleaved species had flat climatic response curves, indicating high levels of plasticity in populations, facilitating adaptation to a broader range of environments, and conferring a high potential for resilience in the face of climatic change. By contrast, the two conifer species had response curves with more pronounced slopes, indicating a lower resilience to climate change. This finding may reflect stronger genetic clines in P. sylvestris and P. abies, which constrain their climate responses to narrower climatic ranges. The response functions had maxima that deviated from the expected maximum productivity in the climate of provenance towards cooler/moister climate conditions, which we interpreted as an adaptation lag. Unilateral, linear regression analyses following transfer to warmer and drier sites confirmed a decline in productivity, predictive of the likely impact of ongoing climate change on forest populations. The responses to mimicked climate change evaluated here are of considerable interest for forestry and ecology, supporting projections of expected performance based on “real-time” field data.