Uncertainty and Confidence in Climate Change Prediction Dave Stainforth, Research Fellow, Atmospheric Physics, Oxford University. Chief Scientist for Climateprediction.net.

Uncertainty and Confidence in Climate Change Prediction

Dave Stainforth, Research Fellow, Atmospheric Physics, Oxford University.

Chief Scientist for Climateprediction.netClimate Change Negotiators Meeting

Tuesday 29th August 2006

Environmental Change Institute

Oxford University

1. Climate change - a reality.

2. Predicting the future - climate models.

3. Confident predictions - putting uncertainty bounds on climate forecasts.

Why Uncertainty Analysis Is Important

• Climate change is certainly a problem.

• But predictions of future climate and the impacts of climate are problematic. They must be probabilistic in nature (i.e include uncertainty analyses) because over-confident or deterministic forecasts will:

– lead to misdirected adaptation and development planning, and

– undermine the credibility of climate science.

The Microbe is so very smallYou cannot make him out at all,But many sanguine people hopeTo see him through a microscope.His jointed tongue that lies beneathA hundred curious rows of teeth;His seven tufted tails with lotsOf lovely pink and purple spots,On each of which a pattern stands,Composed of forty separate bands;His eyebrows of a tender green;All these have never yet been seen-But Scientists, who ought to know,Assure us that they must be so….Oh let us never, never doubtWhat nobody is sure about.

Hilaire Belloc

Climate science must be clear on what it is sure about and honest about uncertainties.That’s easier said than done.

What we are sure about: Climate Change on a Global Scale

“An increasing body of observations gives a collective picture of a warming world and other changes in the climate system.”“… most of the warming observed over the last 50 years is attributable to human activities.”Climate Change 2001, The InterGovernmental Panel on Climate Change (IPCC) Third Assessment Report.

What we are sure about: A host of other observed changes

Changes in terrestrial ecosystems include:

• Earlier spring arrival time of animals or plants.

• Earlier breeding times.

• Shifts to higher elevations or latitudes.

• Changes in population densities.

• Widespread retreat of mountain glaciers.• Increase in freeze free season over many

mid and high latitude regions.• Increase in frequency of heavy

rainfall/snowfall events.

Source: IPCC Third Assessment Report

What we are sure about: Concentrations of atmospheric carbon dioxide.

“Present CO2 concentrations have not been exceeded during the past 420,000 years and likely not during the past 20 million years.”Climate Change 2001, The InterGovernmental Panel on Climate Change (IPCC) Third Assessment Report.

What we are sure about: Anthropogenic emissions are part of the cause.

Mankind’s emissions only:

Natural emissions only:

All emissions:

What we are unsure about: What Happens Next.

For some global variables climate models can simulate observed climate change very well.

That gives us confidence that they can be used to predict the future under a given scenario for greenhouse gas emissions.

So We Can Use The Model To Forecast The Future?

That’s what people do.

A complex model of this sort gives lots of regional details which would be invaluable for planning. If it could be trusted.

• The problem is that different models give different results; particularly at the regional / seasonal level of detail.

• There may be many many models which predict the recent past well but respond very differently to changing levels of greenhouse gases.

• Climate forecasts are intrinsically uncertain but by working with probabilistic forecasts we can still extract confident predictions of some aspects of future climate. E.g. An uncertainty range of 1.9-11.5°C change is a confidence that it isn’t under 1.9.

A Hierarchy of Climate Models

Courtesy of Thomas Stocker, University of Bern.

O c e a n Dimension

0 1 2 3

0

point EBM

pulse response models

seesaw models

box models

advection-

diffusion models

thermohaline models (lat/z):

deep ocean

models

OGCM

1

EBM (lat)

rad.-conv. model (z)

–

ocean (lat/z) + EBM (lat)

–

2

EBM (lat/long)

stat. dynam. atm. + diffusive

ocean

ocean (lat/z) + stat. dyn. atm

(lat/long)

OCGM + EBM

(lat/long)

OCGM + QG atm.

Atm

os

ph

er

e

3 AGCM + SST

AGCM + mixed

layer

ACGM + slab

ocean

A/OGCM

Climate Models: A Reminder

Complex, 3-dimensional, Atmosphere /Ocean General Circulation Models (AOGCMs)

Climate Models are the principle tools for climate prediction. Most impacts studies are based on the predictions of AOGCMS.

Sources of Uncertainty and How to Include Them In a Climate Forecast

• Natural Variability: The climate is chaotic with variations on timescales from minutes to centuries. Solution: Initial Condition Ensembles

• Forcing uncertainty:Changes due to factors external to the climate system e.g. greenhouse gas emissions (natural and anthropogenic), solar radiation etc.Solution: Scenarios for possible futures.

• Model uncertainty:Different models could be as good at simulating the past but give a different forecast for the future?Solution: Perturbed-Physics Ensembles

• Model Inadequacy

Exploring Uncertainty: The Climateprediction.net ExperimentS

tan

dar

d

mo

de

l s

et-u

p

Perturbed Physics Ensemble

Initial Condition Ensemble

Forcing Ensemble

Ove

rall Gran

d

En

semb

le

10000s 10s10s

• To quantify uncertainty we need 100s of thousands of simulations.

• Impossible with super computers.• But possible with distributed computing.• At www.climateprediction.net people

can download the model to their PC.• Using the latest, complex model means

we can get regional detail as well as global averages.

Latest Statistics

• > 300,000 participants.

• > 24M years simulated.

• > 110,000 completed simulations.(Each 45years of model time)

• 10000 years of computing time.

ClimatePrediction.net : What it looks like.

First Results in Terms of Climate Sensitivity

In 2001 the IPCC concluded that the climate sensitivity was likely to be between 1.5 and 4.5°C

Many studies have identified the possibility of high sensitivities (>6 °C).

Only now do we have the models which show such a response.

So now we have the possibility of predicting the range of possible future behaviour on regional and seasonal scales.

Climate sensitivity is defined as the equilibrium global mean surface temperature change for a doubling of CO2 levels.

Source: Stainforth et al. Nature, 2005

First Results: Regional Behaviour

From Stainforth et al. Nature. 2005

Regional Behaviour – European Rain and SnowfallMediterranean Basin Northern Europe

Winter Winter

Summer Summer

Annual Annual

From Stainforth et al., “Avoiding Dangerous Climate Change.

Conclusions

• The most comprehensive exploration of uncertainty in climate models has so far shown that:– There is no evidence that climate change could be less dramatic than suggested

by the IPCC Third Assessment Report.– We can not yet rule out the possibility of extremely dramatic levels of climate

change, even at relatively low equivalent concentrations of CO2 e.g. 450ppm.• Uncertainty bounds are essential for planning how society can adapt to the

changes ahead.• Uncertainty and confidence are two sides of the same coin. We can be very

uncertain about some things but confident about others.• There is a realistic possibility of probabilistic regional forecasts in the next

few years. And therefore probabilistic impacts assessments.• Vulnerability to changes in climate may be as important to understanding

local consequences as detailed climate forecasts.