Metamodels, Response Functions, and Research Efficiency in Ecological Economics

Analysis of alternative environmental policies often requires execution of computer simulation models of the ecological or economic phenomena involved. This paper explores efficiencies to be gained by using these simulation models in a program of constructing metamodels. Constructing a metamodel involves statistical modeling of a simulation model's output--creating a model of a model. Metamodeling produces response functions that characterize relationships implied by the structure of a simulation model relating variables that appear to be important to policy evaluation. We first make the case for the research efficiency of metamodeling by discussing an integrated policy evaluation system designed for herbicide pollution policies (CEEPES) that could not have been constructed without metamodels. Next, we illustrate the metamodeling concept by describing a particular metamodel from the system.