Range margin populations show high climate adaptation lags in European trees

How populations of long-living species respond to climate change depends on phenotypic plasticity and local adaptation processes. Marginal populations are expected to have lags in adaptation (i.e. differences between the climatic optimum that maximizes population fitness and the local climate) because they receive pre-adapted alleles from core populations preventing them from reaching a local optimum in their climatically marginal habitat. Yet, whether adaptation lags in marginal populations are a common feature across phylogenetically and ecologically different species and how lags can change with climate change remain unexplored. To test for range-wide patterns of phenotypic variation and adaptation lags of populations to climate, we (a) built model ensembles of tree height accounting for the climate of population origin and the climate of the site for 706 populations monitored in 97 common garden experiments covering the range of six European forest tree species; (b) estimated populations' adaptation lags as the differences between the climatic optimum that maximizes tree height and the climate of the origin of each population; (c) identified adaptation lag patterns for populations coming from the warm/dry and cold/wet margins and from the distribution core of each species range. We found that (a) phenotypic variation is driven by either temperature or precipitation; (b) adaptation lags are consistently higher in climatic margin populations (cold/warm, dry/wet) than in core populations; (c) predictions for future warmer climates suggest adaptation lags would decrease in cold margin populations, slightly increasing tree height, while adaptation lags would increase in core and warm margin populations, sharply decreasing tree height. Our results suggest that warm margin populations are the most vulnerable to climate change, but understanding how these populations can cope with future climates depend on whether other fitness-related traits could show similar adaptation lag patterns.