ERROR PROPAGATION IN A FOREST SUCCESSION MODEL:THE ROLE OF FINE-SCALE HETEROGENEITY IN LIGHT

Abstract

A central challenge in ecology is understanding the emergence of broad-scale community and ecosystem patterns as the result of interactions among individuals. We identify the role of fine-scale light heterogeneity in controlling broad-scale community behavior in SORTIE, an empirically derived, stochastic forest simulation model. SORTIE employs a very detailed measure of local light based on a 216-point sample around every tree. We test the importance of this fine-scale description of local light by reformulating SORTIE with less detail in this algorithm. Predicted forests are compared at several scales from total tree biomass (coarse grained) and patterns of forest succession to the local spatial pattern of light availability at the forest floor (fine grained). SORTIE is surprisingly insensitive to the amount of detail used in the calculation of the local resource, light. In all simulations, 48- and 16-point samples accurately reproduce the local light environment and thus predict forest development without appreciable error. A one-point sample of light significantly alters the estimates of the local light environment, but the emergent forest dynamics are insensitive to these alterations. The robustness of the forest model to the altered light environment stems from two very different mechanisms. First, the alterations of the light environment have very short correlation lengths in time and space. This allows fine-scale averaging to occur on the landscape. Second, the functional relationships among light availability, growth rate, and mortality risk in several key species tend to prevent the altered light environment from affecting individual tree performance. These results demonstrate that emergent forest dynamics in SORTIE do not require detailed local light estimation to capture the interaction between local resource heterogeneity and the functional responses of trees. In addition, we have shown that comparing model predictions at several scales provides a strong methodology to identify relevant detail and understand scaling in this model of a forest.