Uncertainty Quantification in Models

A comprehensive uncertainty characterization in climate includes the ability to understand performance metrics or skills of climate models run in hindcast mode by comparing with observations, a comparison of what multiple models project into the future and the extent which they (dis)agree, and how large knowledge gaps or intrinsic system variability and nonlinearity can be managed to still provide actionable information to decision makers. This project is investigating three interrelated areas along these lines: model ensembles and the use of the multi-model average for regional assessments; a Bayesian framework that consolidates bias in climate models relative to past observations and model (dis)agreement in the future; and a combination of quantitative analysis and qualitative assessments to characterize uncertainty in climate extremes.