Identifying parametric controls and dependencies in integrated assessment models using global sensitivity analysis

• Published in: Environmental modelling & software : with environment data news Vol. 59; pp. 10 - 29
• Main Authors: Butler, Martha P., Reed, Patrick M., Fisher-Vanden, Karen, Keller, Klaus, Wagener, Thorsten
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.09.2014
• Subjects: Assessments; Climate change; Climate models; computer software; Cost engineering; diagnostic techniques; economic resources; economic systems; Economics; Global sensitivity analysis; Integrated assessment model; issues and policy; Mathematical models; Model diagnostics; Sensitivity analysis; Sobol' method; Strategy; uncertainty analysis; variance; Global sensitivity analysis; Climate change; Model diagnostics; Sobol' method; Integrated assessment model
• ISSN: 1364-8152
• DOI: 10.1016/j.envsoft.2014.05.001

Integrated assessment models for climate change (IAMs) couple representations of economic and natural systems to identify and evaluate strategies for managing the effects of global climate change. In this study we subject three policy scenarios from the globally-aggregated Dynamic Integrated model of Climate and the Economy IAM to a comprehensive global sensitivity analysis using Sobol' variance decomposition. We focus on cost metrics representing diversions of economic resources from global world production. Our study illustrates how the sensitivity ranking of model parameters differs for alternative cost metrics, over time, and for different emission control strategies. This study contributes a comprehensive illustration of the negative consequences associated with using a priori expert elicitations to reduce the set of parameters analyzed in IAM uncertainty analysis. The results also provide a strong argument for conducting comprehensive model diagnostics for IAMs that explicitly account for the parameter interactions between the coupled natural and economic system components. •We present Sobol' sensitivity analyses of three climate policy scenarios.•Parameter interactions are significant, evolve over time, and vary by policy.•Key interactions and sensitivities misclassified in expert elicited assessments.•Climate damages not sufficiently sensitive to climate/carbon cycle parameters.•Costs of aggressive mitigation controlled by participation and renewables.