Taking trees to their limits and beyond: A dynamic modeling approach
Both experimental manipulation of climate and continental scale analysis of species distributions are used to evaluate effects of climate on species range limits. Observational data used in species distribution models cannot estimate controls on range limits because information is only available from locations where individuals are successful. Experimental data can provide insight on performance but cannot cover large regions. For observational data we developed a dynamic multi-species model to integrate data from the Forest Inventory Analysis and long-term demographic plots. Effects of climate and competition are interpretable through coefficients predicting demographic rates of individuals such as growth and survival, and populations through fecundity, recruitment, and size-class redistribution. We compared model predictions and estimates with those from experimental warming and competition field plots. In the experiment, we planted seeds from a set of 40 species at Harvard and Duke Forests. We analyzed nineteen of these species which were at, or beyond, their established range limits at one or both sites. We applied experimental treatments in a factorial design with either gap or understory light, and three temperatures (ambient, +3C, and +5C). Seedling survival was regularly monitored in the plots year-round since 2009 and modeled in a binary state-space survival framework.
Both models agreed that competition was a key limitation on survival past range limits. In the dynamic model, all species were predicted to occur well outside their established range limits in the absence of competition. Across species, the estimated climate space where individuals could grow without competition was twofold the climate space where those species are observed in samples. Our experimental data agreed with this – we observed high survival for several species planted well outside their range limits. The survival analysis found a strong positive effect of the gap light treatment for survival of all species, including well outside their current ranges. In addition, seedling growth predicted by the dynamic model at range limits did not differ from growth estimates at the interiors. However, the dynamic model predicted that recruitment is constrained in both geographic and climatic space. Finally, all species showed interactions between canopy status and a hydrothermal variable in the survival model. Taken together these results suggest interactions between soil moisture and competition at early life stages are important in determining the range limits of eastern U.S. tree species.