Environmental Economics and Uncertainty: Review and a Machine Learning Outlook

Abstract

Economic assessment in environmental science is often subject to limited or costly data, and thus it is important to properly characterize the uncertainty associated with impact assessment. We review the issue of uncertainty in integrated assessment modeling, focusing on the techniques of uncertainty characterization. We propose applying manifold sampling, a machine learning technique, to characterize uncertainty in impact assessment, which constructs a joint probability model that helps answer three types of policy-making questions: (1) prediction; (2) response; and (3) prevention. A case study is carried out for offshore deepwater oil spills in the Gulf of Mexico.