Density, Climate, and Stochasticity Shape Four Centuries of Population Dynamics for Two Long-Lived Tree Species

Abstract

The dynamics of colonizing populations may be strongly influenced by both extrinsic (e.g., climate and competition) and intrinsic (e.g., density) forces as well as demographic and environmental stochasticity. Understanding the impacts of these effects is crucial for predicting range expansions, trailing edge dynamics, and the viability of rare species, but the general importance of each of these forces remains unclear. Here, we assemble establishment time and spatial locations of most individuals that have reached maturity in six isolated, establishing populations of two pine species. These data allow us to quantify the relative importance of multiple factors in controlling growth of these populations. We found that climate, density, site, and demographic stochasticity were of varying importance both within and across species, but that no driver appeared to dominate dynamics across all populations and time periods. Indeed, exclusion of any one of these effects greatly reduced predictive power of our population growth models. Given the similarity in the abiotic characteristics of these sites, the varying importance of these classes of effects was surprising but speaks to the need to consider multiple effects when predicting the dynamics of small and colonizing populations.

Keywords: Pinus flexilis; Pinus ponderosa; climate; colonizing population; population dynamics.

FIGURE 1

(a) Relative population sizes from population initiation until 1970. Curves show total population sizes, including juvenile and mature trees. (b) Local spatial arrangements of each population. See main text for habitat delineation. Scale bar from high to low density is relative to the maximum density value in each panel.

FIGURE 2

Density distributions of establishment distances to the nearest mature tree at the time of establishment (shown above the x‐axis) versus distributions of distances to the nearest mature tree from each point within the entire defined landscape area (available distances, shown below the x‐axis) for each population. Note the log2 scaling. Recruitment is markedly higher close to mature trees than far away, but few recruits establish extremely close to established individuals. See text for details on landscape delineation. PF populations shown in left panel and PP populations on the right. The density of available distances is shown for 1970, and all trees that recruited prior to 1970 are included in the density of establishment distances.

FIGURE 3

Standardized coefficient estimates and SD for all top population models. Non‐climate variables in top panel, and climate variables are shown in the bottom panel. Climate variables are annual averages (i.e., 20 year. temp temperature in second decade post‐establishment).

FIGURE 4

Results of stochastic simulations for the best‐supported population model (our “population models,” Tables S3–S8). Plots show the number of mature trees at each time step. Simulation results are from 500 runs. Note that we ended the simulations at 1950 because our complete climate data only extend to 1970, and predictor variables for some populations included third‐decade post‐establishment climate data. See Methods for details. The median value of all stochastic simulations is shown in red. The non‐stochastic simulation is dotted, and actual population numbers are in black. Gray lines show the results of individual runs.

FIGURE 5

(a) Median predictions of all classes of stochastic population models. Actual data shown in black, with median of simulations using the best model for each population shown in red, using the top model initiated after five trees were present shown in green, the top model without climate variables in blue, without density in orange, and non‐stochastic predictions shown in yellow. (b) Annual proportional deviations ([model predicted total number of trees − observed]/observed) for each type of model.

FIGURE 6

Boxplots of proportional deviations from the top model predictions after 5 adult trees were present to actual population numbers for (a) population models and (b) species‐wide models. Deviations from median stochastic simulations that used best population model (red), best model fit with only data from when 5 adult trees were present in each population for population models (green), top model without climate variables (blue), without density variables (orange), non‐stochastic model (yellow), and species‐wide models without a fixed site effect (pink).