STRATEGIES FOR MODELING THE UNCERTAIN IMPACTS OF CLIMATE CHANGE

Modeling of the physical and social effects of climate change is fraught with uncertainty. A hierarchy of computer codes often must be used, each with embedded uncertainties. Traditional methods of analyzing uncertainty - brute-force Monte Carlo analysis and sensitivity testing - are expensive and time-consuming. By means of a simple example, the article shows that to reduce cost, hierarchies can be analyzed by sensitivity experiments at selected design points. Reduced-form models are then simulated at each level of the hierarchy to summarize variables and functional relationships that truly matter, while preserving the statistical integrity of the distributions of outcomes.