Process versus empirical models: which approach for forest ecosystem management?

Abstract

The emergence of forest ecosystem management presents new information challenges for forest managers. Shifting views of the forest from primarily one as a production system for wood fibre to an ecosystem with spatially and temporally complex interrelationships is changing the demand for information about the forest. These new information needs are characterized by greater complexity, limited availability of mechanistic hypotheses, and a paucity of data. Empirical and process modelling approaches have evolved in forest management to solve different problems, and debate about the two approaches has existed for some time. Which approach to forest modelling will best be able to meet the challenges of ecosystem management? Empirical models seek principally to describe the statistical relationships among data with limited regard to an object's internal structure, rules, or behaviour. In contrast, process models seek primarily to describe data using key mechanisms or processes that determine an object's internal structure, rules, and behaviour. In addition, mechanisms included in process models are general enough that they can maintain some degree of relevance for new objects or conditions (mechanism constancy), while empirical models tend not to be tied to any specific mechanism, so that derived model parameters must remain constant (parameter constancy) for new objects or conditions. Based on these differences, we argue that process models offer significant advantages over empirical models for increasing our understanding of and predicting forest (a tree, a stand, a landscape) behaviour. Process models are, therefore, more likely to meet the information challenges presented by ecosystem management.