Species range dynamics are underlain by population-level processes that regulate population growth rates and determine the environments in which a species can persist. Early tree life stages (i.e. seedlings) integrate responses to recent environmental conditions and play a key role in regulating species persistence and migration through establishment and mortality patterns. Thus, tree seedling population dynamics observed over species ranges provide a unique opportunity to identify the processes underlying tree species range dynamics. Here, we investigate the ecological processes underlying the seedling distributions of five dominant Rocky Mountain conifer species by integrating data spanning multiple processes and scales within a Bayesian hierarchical metamodeling framework. Specifically, we integrate experimental data characterizing fundamental in situ demographic responses of seedlings to temperature and precipitation with population growth estimates and occurrence records from a large, range-wide field survey to quantify the influences of local environmental conditions, biotic interactions and broad-scale climate on seedling population dynamics and seedling distributions. We also compare our modeled seedling distributions with those generated from simple correlative species distribution models (SDMs) to clarify how underlying processes influence distribution patterns.
Results/Conclusions
The integrated metamodels for each species successfully integrated multiple data sources to generate robust predictions of seedling distributions for all five species (TSS>0.5). Notably, incorporation of population growth rates generated more constrained distribution predictions than SDMs, demonstrating that species fail to exhibit positive population growth in many regions that are predicted as suitable by a simpler correlative approach. These findings suggest that range dynamics of our focal species are characterized by source-sink dynamics, particularly along latitudinal range margins. Population growth rates were well-explained by variation in climate, with a detectable influence of biotic interactions for key species for which seedling facilitation is a recognized process (e.g. Engelmann spruce and subalpine fir). Incorporation of data spanning multiple sources, scales and processes reduced uncertainty in estimated seedling distributions relative to the SDMs and enabled improved characterization of uncertainty.