Uncertainty and Climate ChangeUncertainty and Climate Change
Dealing with uncertainty in climate change impactsDealing with uncertainty in climate change impacts
Daniel J. VimontAtmospheric and Oceanic Sciences Department
Center for Climatic ResearchUniversity of Wisconsin - Madison
“Preparing for Climate Change” WorkshopMadison, WI, June 19, 2007
Uncertainty and Climate ChangeUncertainty and Climate Change
Uncertainty and Climate ChangeUncertainty and Climate Change
Sources of UncertaintySources of Uncertainty
Emissions Scenarios:How will our society evolve?What sort of technology will we develop?What mitigation strategies will we employ?
IPCC AR4, WG I
Sources of UncertaintySources of UncertaintyGreenhouse Gas ConcentrationsWhat are the life-spans of different gasses in Earth’s Atmosphere?How do anthropogenic, terrestrial and aquatic sources / sinks alter future GHG concentrations?
IPCC AR4, WG I
Sources of UncertaintySources of UncertaintyRadiative ForcingHow do different gasses affect the amount of radiation that Earth receives?
IPCC AR4, WG I
Sources of UncertaintySources of UncertaintyClimate Response:What climatic processes alter the way that Earth responds to changing GHG’s (good uncertainty)?What is the range of natural variability (good)?Can we account for model bias (bad uncertainty)?
IPCC AR4, WG I
Sources of UncertaintySources of Uncertainty
Emissions Scenarios:Different emissions scenarios imply different amounts of global warming. The range of scenarios provides an estimate of uncertainty (good uncertainty)
IPCC AR4, WG I
Sources of UncertaintySources of Uncertainty
Climate Response:Each model has its own assumptions, which leads to a slightly different amount of global warming. Multi-model ensembles provide an estimate of uncertainty due to limited physical understanding (good uncertainty)
Regional UncertaintyRegional Uncertainty
IPCC AR4, WG I
Regional UncertaintyRegional Uncertainty
IPCC AR4, WG I
Using Uncertainty: Top-downUsing Uncertainty: Top-down
Scenarios:
Useful when:• Impact is unknown or broad scale• Policy does not exist (can explore
different policy options)• Surprises are expected• Simple to implement
Disadvantages:• Computationally expensive• Poorer characterization of
uncertainty• Poor sampling of “climate space”
and “impact space”
Impact
Present Climate
Predicted Climate
Top-down impact assessment
Using Uncertainty: Bottom-upUsing Uncertainty: Bottom-up
Risk Assessment:
Advantages:• Useful when impacts are well
known• Uncertainty is well quantified• Adaption strategies are easily
explored
Disadvantages:• Projected climate variables may
not be relevant• Difficult to deal with “surprises”• Sometimes inflexible
Impact
Present Climate
Predicted Climate
Bottom-up impact assessment
Threshold Impact Region
Risk AssessmentRisk Assessment
Med Risk
Low Probability, High Consequence
High Risk
High Probability, High Consequence
Low Risk
Low Probability, Low Consequence
Med Risk
High Probability, Low Consequence
Probability
Con
sequ
ence
Probability x Consequence = Risk
AdaptationPolicies
MitigationPolicies
Risk AssessmentRisk Assessment
Jones, 2004
Determining Probability of ExceedenceProbability distribution of future climate state determined by random sampling of future climate projections. Cumulative distribution determines probability of threshold exceedence.
Non-standard variablesNon-standard variables
Reducing Model BiasModels have serious bias with certain non-standard variables (e.g. daily snow or rain). Combinations of models and observations can reduce bias and allow examination of non-standard variables.
Non-standard variablesNon-standard variables
SummarySummary
Uncertainty is unavoidable.Uncertainty arises through emissions scenarios, estimates of GHG concentration, radiative forcing, climate response, and impact sensitivity. This is unavoidable.
Uncertainty can be used to explore policy or adaption optionsScenarios are useful when impacts are not known, or policy does not exist. Risk Assessment is useful when policy does exist, or when thresholds are well defined (in terms of climates).
Debiasing techniques are useful for non-standard variablesModels are appropriate for large-scale impacts, but not always for regional impacts. Combinations of models and observations can reduce regional bias, reducing “bad” uncertainty