Informing Decisions with Climate Information at Different Levels of Confidence

As one important purpose, uncertainty quantification aims to provide information in a way that can usefully inform decisions. But many actual decisions may prove sensitive to information at different levels of confidence, which poses challenges for the uncertainty quantification task. For instance, some salient information may be well-represented by pdf's while other information may only be supported by a scattering of studies. This talk will demonstrate a decision analytic framework that can usefully employ information at different levels of confidence. The framework is based on the idea of identifying thresholds in various combinations of system properties that would suggest switching from one decision to another, and then gathering scientific evidence relevant to those thresholds that can help decision makers adjudicate their choices. The talk will demonstrate this approach with an example analysis that considers how the Port of Los Angeles might consider the potential for extreme sea level rise in its investment plans. This study uses a robust decision making (RDM) analysis to address two questions: (1) under what future conditions would a Port of Los Angeles decision to harden its facilities against extreme sea level rise at the next upgrade pass a cost-benefit test, and (2) does current science and other available information suggest such conditions are sufficiently likely to justify such an investment? To answer this second question, we use information expresses as a combination of probabilistic climate forecasts, interval probabilities, and non-probabilistic information. We find that a decision to harden at the next upgrade would merit serious consideration for only one of the four Port facilities considered and hardening costs would have to be 5 to 250 times smaller than current estimates to warrant consideration for the other three facilities. This study also compares and contrasts a robust decision making analysis with a full probabilistic analysis. This talk will help suggest how the uncertainty quantification community might use a variety of different types of information to usefully inform decisions.