Exposing variation to aid climate change risk assessment

Abstract:

Considerable efforts to quantify different sources of variation in climate change projections (some might say uncertainty) have led to a welcome set of additional information on which to base confidence about what and how different futures might unfold and how different types of mediating efforts might affect the future. Quantifying the impacts of these different sources of variation on key climate change projection metrics should be used in part to guide future model development efforts. I will report on several of my team's recent research projects to better quantify and assess the importance of different sources of variation. I will show how we use inference techniques to estimate parameter uncertainty in land and marine carbon components of earth system models by comparing them with observational evidence and show how we propagate such uncertainty to better assess how such systems might respond to climate change and quantify the impact of reducing uncertainty for different applications. I will also show how we use such techniques on simulation models themselves to identify key sources of variation in their predictions: helping to pinpoint important focal areas for model improvement. Lastly, I will show a new software prototype being designed to enable any user to view climate model projections alongside historical and recent observational evidence while, importantly, also exposing some of the variation / uncertainty in the reported information.