Modeling forest dynamics across diverse ecoprovinces in the eastern U.S

Background/Question/Methods

A recently developed forest dynamics model, the Perfect Plasticity Approximation (PPA), captures the essence of height-structured competition between individual trees, while remaining both mathematically and computationally tractable. The model scales up the growth and mortality rates of understory and canopy trees to predict the dynamics of multi-species forest communities. A previous case study demonstrated that the model could successfully predict key aspects of secondary succession in forests of Michigan, Wisconsin, and Minnesota. Here, we generalize these results to a larger and more diverse region by parameterizing and simulating the PPA model for the ecoprovinces in the eastern U.S. Model parameters (species-specific individual-level growth and mortality rates and allometries) were estimated from observations of individual trees collected by the U.S. Forest Service Forest Inventory and Analysis (FIA) and Forest Health Monitoring programs. For each ecoprovince, we initialized replicate simulations by drawing random samples from the conditions present on 10-20 year-old FIA plots and from the multivariate probability distribution of parameter values. Each simulation was run for 100 years, representing secondary succession from 15 to 115 years post-disturbance. These predictions were tested against chronosequences derived from FIA data.

Results/Conclusions

The model produced accurate predictions of how forest biomass and the size distribution of individual trees changed over the course of succession. However, in most ecoprovinces, there were significant mis-matches between the predicted and observed abundances of individual species. We tested two hypotheses that could explain these mis-matches:  (1) non-stationarity, i.e., temporal shifts in the abundance of different species, and (2) hidden edaphic variation within FIA-defined soil types. The first hypothesis implies that species whose predicted abundance is greater than its observed abundance should be increasing in abundance over time (e.g., at a decadal time-scale). The second hypothesis implies that individual growth and survival rates of a given species should be higher in inventory plots where that species is relatively abundant. Our analysis of FIA data provides support for both of these hypotheses. This suggests that (1) the PPA model has correctly diagnosed temporal shifts in the abundance of difference tree species in the eastern U.S., and (2) hidden edaphic variation has significant effects on forest community composition.