There is general agreement that trees will respond to climate change by shifting their ranges upward in elevation and latitude, yet variation in the observed rate, magnitude, and in some cases direction of range shifts indicates that variation in underlying ecological processes, such as demography, may be an important determinant of how tree range shifts occur. In particular, range shifts may be influenced by recruitment, which acts as a bottleneck in tree range dynamics. Characterizing the climatic dependence of recruitment dynamics requires an integrated understanding of both the environmental conditions under which juveniles can establish and survive, and in which adults can survive and reproduce. In this study, we apply an integrative Bayesian metamodeling approach to directly assimilate data from experimental and observational datasets of seedlings and adults of five western U.S. conifer species into a cohesive framework for predicting current and future ranges while accounting for underlying variation in recruitment. Our specific aims were to compare species range predictions made using a standard species distribution modelling approach (SDM) and our integrated approach under current and future climate, and to determine whether the additional information provided by seedling data can be used to generate more robust predictions of range dynamics.
Results/Conclusions
Current range predictions from the SDM and integrated models differed minimally across all species, but integrated model predictions were associated with greater uncertainty. Uncertainty differences were greatest along the edges of occupied regions, such as in low-elevation regions for subalpine species and in higher-elevation regions for montane and woodland species, and in the southernmost regions of the study area. Future range projections for 2090 indicate range contractions for high-elevation montane and subalpine species and range expansions for lower-elevation species. Differences in range predictions between SDM and integrated models are magnified under future climate, particularly for lower-elevation species. Specifically, integrated models provide more conservative estimates of range expansions. Differences in estimated climate niches between SDM and integrated models appear to be primarily attributable to the response of seedling survival to temperature, and not to climate-driven variation in seedling population dynamics. In particular, the integrated models indicate a greater and differential sensitivity to temperature extremes when seedling information is considered. Overall, our comparison demonstrates that integration of seedling information serves primarily to improve characterization of uncertainty, particularly in areas along current range margins that are relevant to anticipating range shifts, and also suggests that seedling temperature sensitivity may constrain future range shifts.