Inherent limits to prediction and implications for projecting forest response to climate change

Background/Question/Methods

Forecasting the future composition and structure of forests is increasingly important as the magnitude of anthropogenic perturbations of the earths systems grow.  Forest composition and distribution changes will have local impacts on a wide array of coincident species and communities, in addition to global scale impacts through effects on greenhouse gases, evapotranspiration, and albedo. The assumption underlying efforts to predict ecological responses to anthropogenic climate change is that responses of forest systems are predictable.  We examine this underlying assumption and ask if ecological systems, and in particular forest systems, are fundamentally predictable. 

Results/Conclusions

Strong limits on the predictability of ecological system responses to anthropogenic perturbations arise from the ubiquity of nonlinearities, the multitude of interactions in ecological systems, and the difficulty of fully enumerating the attributes of species niches.  These characteristics of ecological systems can impose strong limits on both the realized and intrinsic predictability of ecological systems through at least three mechanisms:  i) Chaotic dynamics characterized by the exponential growth of initial errors in specification of the underlying model and/or initial conditions.  ii) Computational irreducibility that asserts that future states of the system cannot be ascertained except through calculation of all of the intermediate computations. iii) The difficulty of pre-stating the complete features of ecological niches relevant to system dynamics for both current and future environmental conditions. The first of these processes, sensitive dependence upon initial conditions that leads to potentially chaotic dynamics, is well known in the ecological community but the second and third mechanisms, computational irreducibility and difficulty in fully enumerating the relevant features of niches, are less widely recognized.  While the occurrence of chaotic dynamics could impose a strong limit on the realized predictability of ecological systems, both computational irreducibility and niche indefinability may limit the intrinsic or theoretical predictability of ecological systems.  We argue that computationally irreducibility and niche indefinability are likely to be widespread and to impose strong limits on the potential for prediction in ecology. We provide examples of these limits to prediction with emphasis on forest trees and species distribution modeling. We argue that the utility of models for projecting forest compositional change may be enhanced through relative assessment methodologies that consider uncertainties in model structure, parameters, and data.