Presentation on theme: "Uncertainty Quantification (UQ) and Climate Change Talking Points Mark Berliner, Ohio State Issues of continuing interest: Models, Data, Impacts & Decision."— Presentation transcript:
Uncertainty Quantification (UQ) and Climate Change Talking Points Mark Berliner, Ohio State Issues of continuing interest: Models, Data, Impacts & Decision Support 1)Disclaimers a)List is incomplete b)None are criticisms, just questions and suggestions c)Please let me know what I missed 2)My Apologies for missing the workshop
Models We know many sources of uncertainty: parameterization; feedbacks; resolution & approximation; nonlinearity; etc. How to help? Continued model assessment Claim that models and observations line up needs quantification Since a key is the response of hydrological cycle, do real advances require increased resolution? Multi-model ensembles Bigger ensembles vrs better models (allocation of computational resources) Do more models lead to more confusion? Lots of opportunities for collaboration (both design and analysis)!!!!
Data 1)Paleoclimate: Lots of very important work to do: Assess impacts of forcings without sole reliance on climate models, though we need to know the historical impacts and the forcings. Use of proxies is key opportunity for collaboration Aid in assessment/improvement of the models 1)Combining datasets: rebuilding climate variable global estimated fields w/ uncertainties (eg, temperature fields) 2)Modern climate monitoring: design and use of remote sensing, etc. (compute or observe) 3)Operational system for scientists and decision makers to access info on the state of the planet
Impacts 1)Level I: Sea level; Storms, droughts, hurricanes; Weather extremes; Fire; Floods; Other environmental systems 2)Level II: Human activities, agriculture, health, economics, national security 3)A viewpoint: hierarchical chains Global to regional impacts: downscaling a) Dynamic b) Statistical c) Combinations Excellent opportunity!!! Propagation of uncertainty Collaboration across many disciplines (Level II)
Global vrs Regional Behavior Annual mean precip (cm) 1980-1999: Observed & simulated based on multi- model mean.
Propagation of uncertainty: (causal) chains CO 2 __ > global climate __ > regional climate __ > local weather and hydrological environment __ > disease patterns crop yields; etc. None of these arrows are deterministic or known. Rather we need to build probability models for each and then link them. This is where statisticians are expert. Remark: feedbacks can be very tough.
Decision Support Ranges (intervals) vrs probability dist. When are honest confidence/prediction intervals too wide to be a value? Risk analysis (expected loss) Explaining/using UQ: uncertainty is not ignorance Presenting results in useful forms Sensitivity, robustness Remark: Some dont like this area, but it is being done; I think we should participate!!