Probability in the attribution and prediction of

Myles Allen Departments of Physics, [email protected]

Oxford University What was the most unrealistic aspect of the film “The Day After Tomorrow”?

Oxford University South Oxford on January 5th, 2003 Photo courtesy of Dave Mitchell of Photo courtesy

Oxford University The problem in October 2000 and January 2003: a consistently displaced Atlantic jet-stream

The Atlantic Jet Stream (500hPa wind speed) Autumn climatology (colours) & Autumn 2000 (contours)

Blackburn & Hoskins, 2003

Oxford University But the jet-stream varies with the weather: how can we pin down the role of climate change?

„ “Climate is what you expect, weather is what you get” (Lorenz, 1982) „ and in the 21st century: „ “Climate is what you affect, weather is what gets you”

Oxford University Autumn 2000 events “were extreme, but cannot in themselves be attributed to climate change.”

1947

2000

Oxford University It has happened before: Shillingford historic flood levels

2003

Oxford University The 2001 conclusions of the Intergovernmental Panel on Climate Change

„ “Most of the warming over the past 50 years is is likely (meaning a better than 2 in 3 chance) to have been due to the increase in greenhouse gas concentrations.” „ But what does this tell us about flooding in Oxford?

Oxford University Model-simulated changes in extreme rainfall in southern England

4-year 12-year event event 2090 30-year event 2000 1860

Oxford University And now for the next century: carbon dioxide trends

Note different scale! Source: IPCC Third Assessment Report

Oxford University Uncertainty in global warming under two scenarios of future emissions

Oxford University Changing emission path buys time but does not eliminate risk

Oxford University Ranges of uncertainty in regional predictions require multi-model “ensembles”

Global temperature change under 1% per

year increasing CO2 (CMIP-2 model inter-comparison)

Global precipitation change under 1% per

year increasing CO2

Oxford University A shortage of models with high climate sensitivities?

Oxford University Ranges of opinion in climate sensitivity (Morgan and Keith, 1995)

Oxford University Dealing with uncertainty in modelling climate change

„ Climate is predictable, but cannot be directly observed. Weather is observable, but unpredictable. „ Any statement about climate change involves probabilities: looking at the spread of results from lots of climate simulations. „ On long time-scales, simulations must allow for uncertainty in modelling, not just chaotic variability in the atmosphere-ocean system. „ But full-scale climate models are expensive to run: the largest conventional ensembles to date are only 20-50 members…

Oxford University The problem with dealing with uncertainty in climate change prediction

„ Objective: find as many as possible alternative, equally realistic, model versions that respond differently to increasing carbon dioxide, to explore the full range of possibilities.

Perturbed Physics Initial Condition Forcing Ensemble Ensemble Ensemble Overall Grand Ensemble Standard model set-up

10000s10s 10s

Model Versions Simulations

Oxford University Climateprediction.net: the world’s largest climate modelling facility . . . ..

.

~100,000 volunteers, 130 countries, ~6M model-years Upload servers Oxford University Members of the public download and run a full 3-D climate model on their personal computers

Oxford University Visualization software for school and undergraduate projects

Oxford University Active, and self-regulating, user forum

Oxford University Initial climateprediction.net experiment

„ Using simplified model ocean to keep runs short

„ 15-year calibration, 15-year control, 15-year 2xCO2 „ Up to 10-member initial-condition ensembles to reduce noise and quantify sampling variability Double CO2

15 yr, 2 x CO 2 Diagnostics from final 8 yrs.

Derived fluxes Calibration

15 yr spin-up 15 yr, base case CO2 Control

Oxford University Time-evolving frequency distribution

Remove models that are unstable in the control.

Few remaining negatively drifting

2xCO2 model versions are an unrealistic consequence of using a slab ocean.

Oxford University Not The Day After Tomorrow: why we got some negative sensitivities…

Oxford University Estimating effective climate sensitivity from

short 2xCO2 runs

Oxford University Exploration of parameter space, focussing on identifying non-linear interactions

„ Perturbations to 21 atmospheric/surface parameters „ Three values each, including combinations „ Initial exploration of 6 parameters (clouds and convection) P2

High

Standard

Low LowStandard High P1 Oxford University Sensitivities from climateprediction.net

Stainforth et al, 2005 Oxford University Regional responses: temperature and precipitation

Standard model version

Low sensitivity model

High sensitivity model

Oxford University Can observations rule out high sensitivities?

CMIP-2 coupled models

Original Single perturbations model

Stainforth et al, 2005 Oxford University Regional responses: temperature and precipitation

Standard model version

Low sensitivity model

High sensitivity model

Oxford University Still they come: 47334 simulations passing initial quality control

Traditional range

Courtesy of Ben Sanderson Oxford University Are these high sensitivities ruled out by the observed response to Mount Pinatubo?

Oxford University No: EBM responses to Pinatubo forcing blue = 0.5K sensitivity, deep red = 20K sensitivity

Frame et al, 2005, also fitting ENSO, background climate and effective heat capacity

Oxford University Are these high sensitivities ruled out by temperatures in the Last Glacial Maximum?

∆F=-6.6±1.5W/m2 ∆T=-5.5±0.5K

Numbers courtesy of Stefan Rahmstorf and Gavin Schmidt, realclimate.org Oxford University No: symmetric uncertainty in past forcing → asymmetric, open-ended range for sensitivity

Oxford University Naïve sampling strategies can give the illusion of a tight upper bound on sensitivity

Oxford University Would these high sensitivities necessarily be ruled out if we uploaded more diagnostics?

Oxford University No: Murphy et al, 2004, distribution without prior weighting towards low sensitivities

Oxford University Objective constraints on feedback parameter inferred from the climateprediction.net ensemble

Oxford University Objective constraints on climate sensitivity inferred from climateprediction.net

Piani et al, 2005

Oxford University Not the first to report a risk of high sensitivity

Oxford University Forcing uncertainty: the main obstacle to constraining climate sensitivity

Oxford University High risk of substantial warming even with today’s greenhouse gas levels

Traditional range

Oxford University Lots of studies, same message: weak upper bound on climate sensitivity

„ If S is “likely” < 4K (P>0.67) then „ S is “very likely < ~7K (P>0.9) and we can only say „ S “virtually certain” < 10-15K (P>0.99)

Oxford University And here’s why:

„ Observable properties of climate scale with strength of atmospheric feedbacks „ Most constraints end up fairly Gaussian (Central Limit Theorem) „ A Gaussian distribution of inverse sensitivity gives… „ no formal upper bound on climate sensitivity

Oxford University High sensitivities and the challenge for the IPCC

„ In 2001, all studies reported detectable greenhouse warming at >90% confidence, yet IPCC stuck to “unlikely to be entirely natural in origin” (>67%) „ In 2005, no studies rule out S>5K at >90% confidence except by prior assumption: but will IPCC want to suggest sensitivity is only “likely” <5K? „ Of course, there is the fact that… „ No one is likely to drag an IPCC author through the US courts for underestimating the chance of a high sensitivity.

Oxford University Why they care

„ Failure to place an upper bound on sensitivity appears to undermine the policy relevance of IPCC. „ Huge pressure to come up with the killer paper for AR4 (the prize: world-wide fame, and a chocolate from Susan Solomon). „ New evidence will be indirect: direct observations of GHG-induced warming or TOA fluxes don’t cut it.

Oxford University We have been here before

„ Use of model-simulated variability, and the recovery from the Little Ice Age, undermined the policy relevance of “discernible human influence” in 1995. „ In 1998, a new hemispheric millennial temperature reconstruction appeared to obviate model-simulated variability (and the Little Ice Age). „ Of course, that reconstruction may yet turn out right: maybe there was no global Little Ice Age. „ And maybe the climate sensitivity really is <4K. „ Premature declarations, even if vindicated, permanently undermine the credibility of the IPCC.

Oxford University The irony is, it doesn’t really matter, because we already agree on so much else

Michaels et al, 2000, 2004

Oxford University The irony is, it doesn’t really matter, because we already agree on so much else

Michaels + 7 years

Oxford University Given we won’t actually stabilize concentrations indefinitely, why care about sensitivity?

Equilibrium warming under 550ppm stabilisation Maximum warming under 550ppm peak in 2100 Oxford University Why can’t the IPCC just say

„ For well-understood physical reasons, we cannot place an objective upper bound on climate sensitivity. „ So, we cannot estimate the risks associated with a

given stabilisation CO2 concentration. „ But we can estimate the range of transient changes expected over the coming decades, and … „ We can also estimate the effort required to hit a given temperature target. „ Sometimes, admission of ignorance is the most policy-relevant option of all.

Oxford University Implications…

„ Many studies have found there is some risk of a substantial climate change even with today’s greenhouse gas levels. „ Climateprediction.net, with the help of the public, confirmed this result for the first time with a full- complexity climate model. „ But warming takes time (many decades) to emerge: what about the here and now? „ We mentioned the possibility of increased flood risk, but that was one model study. Are there other events more closely related to rising temperatures?

Oxford University Summer 2003 temperatures relative to 2000-2004

From NASA’s MODIS - Moderate Resolution Imaging Spectrometer, courtesy of Reto Stöckli, ETHZ

Oxford University Heat-wave blamed for US$12.3 billion uninsured crop losses + US$1.6 billion forest fire damage

Oxford University Excess mortality rates in early August 2003 indicate 22,000 - 35,000 heat-related deaths

Daily mortality in Baden-Württemberg Oxford University But a single heat-wave is a weather event: how can we pin down the role of climate change?

„ The immediate cause of the heat-wave was a persistent anti-cyclone over Northwest Europe. „ There is still no evidence that human influence on climate makes such circulation patterns more likely. „ Instead, we ask how human influence on climate has affected the risk of such a weather event (however induced) causing such an intense heat-wave?

Oxford University June-August temperatures in 2003, relative to 1961-90 mean, Mediterranean region

Oxford University Modelling Southern European area-averaged June-July-August summer temperatures

Future projection

Instrumental observations

Natural drivers only All drivers included

Oxford University External contributions to European summer temperatures, relative to pre-industrial

Anthropogenic 0.5K

Natural

Oxford University Human contribution to the risk of the 2003 heat-wave: loading the weather dice Increase in risk

Range of uncertainty

Fraction of current risk attributable to human influence

Oxford University Tuyuksu Glacier, Kazakhstan: a vital water source

Oxford University Tuyuksu mass balance

Oxford University The Spectre of Liability

„ Modest (0.5oC) background warming substantially increases the risk of extreme high temperatures. „ It is likely (90% confidence) that past human influence on climate was responsible for at least half the risk of the 2003 European summer heat-wave. „ “Plaintiffs ... must show that, more probably than not, their individual injuries were caused by the risk factor in question, as opposed to any other cause. This has sometimes been translated to a requirement of a relative risk of at least two.” (Grossman, 2003)

Oxford University By the 2030s, >50% of anthopogenic GHG loading will be due to post-1990 emissions

Oxford University What was the most unrealistic aspect of the film “The Day After Tomorrow”?

Oxford University There weren’t any lawyers

„ The contribution of past to some current climate risks may already exceed 50%, the threshold for civil tort actions. „ Over the coming decade, both the cost and the inevitability of climate change will become clearer, fuelling demands for compensation for: – Flooding – Heat wave damages and deaths – Threats to water supplies, especially from glacial sources – Coastal erosion etc.

Oxford University Giving climate change back to the people

„ Politicians tend to talk about climate change as an environmental or ethical issue: care for polar bears or future generations. „ This diverts attention from the injustices that are happening now: we all benefit from burning fossil fuels, but some are losing out much more than others from the impacts of climate change. „ The risk, even quite remote, of a successful class- action damages suit would have far more impact than any conceivable follow-up to the Kyoto Protocol.

Oxford University But what could be done?

„ How can an oil company or coal miner avoid selling a product whose use involves increasing

atmospheric CO2? „ Simple: they bury (“sequester”) the equivalent amount of carbon. „ Of course, this would make oil or coal more expensive, which would hurt – but how much? „ “All OECD countries besides the US impose big taxes on fuel, but curiously it hasn’t reduced consumption.” Lord Browne, BP (Financial Times)

Oxford University But what could be done?

„ Fossil fuels are still remarkably cheap, since we pay for the cost of extraction (and cartel-like profits), not the cost of their impact. „ If politicians were to apply the “Polluter Pays Principle” to producers of fossil fuels, this would change rapidly: it might well make more sense to sell carbon-neutral fuel than risk liability.

Oxford University