Insert title here

Insert title here

Probabilistic Predictions of Climate Probabilistic Predictions of Climate Change in Australia using the Reliability Change in Australia using the Reliability Ensemble Average (REA) of CMIP3 Model Ensemble Average (REA) of CMIP3 Model SimulationsSimulations

Dr A.F. Moise & Dr D. HudsonBureau of Meteorology Research CentreMelbourne, [email protected]

ICCC, Hong Kong, May 2007 2

Overview

Methodology: REA CMIP3 Areas under study Results: REA for DJF, JJA Temperature, Precipitation Changes across SRES scenarios Methodology: probabilistic projections Threshold probabilities PDF’s for Australian regions Reliability contribution of CMIP3 models Summary

Acknowledgement

This activity is supported by the Australian Greenhouse Office.

ReferencesGiorgi, F., and L. Mearns, 2002. Journal of Climate, 15, 1141-1158.Giorgi, F., and L. Mearns, 2003. Geophysical Research Letters, 30 (12), art. no 1629, doi:10.1029/2003GL017130.

Milestones

Methodology

Model reliability is a function of model bias (B) AND the distance (D) from the REA average

= Natural variability

Model is “reliable” (Ri=1) when its bias and distance from the REA mean are within natural variability.

Weighted ensemble average and RMSD (weighted by model reliability Ri)

i i

i ii

R

TRT~

ii

iii

T R

TTR 2)~

(~

REA-mean REA-rmsd

)()(,,ii

iDiBi DabsBabsRRR

εT = Max{30yr-runMean[detrended(20th century observed T time series)]} – Min{[(…..)]}

RB: performance criterion RD: convergence criterion

If |BTJ| < εT then RB,I = 1

If |BTJ| < εT then RB,I = 1

CMIP 3 models used

Models

OBS

1981-2000and

2081-2100

- NCC high quality monthly data set

Regions of analysis

Maps of Australia and southern Africa (1=Gabon, 2=Congo, 3=Dem.Rep. Congo, 4=Tanzania, Rwanda, Burundi, Uganda, 5=Kenya, 6=Angola, 7=Zambia, 8=Malawi, 9=Mozambique, 10=Namibia, 11=Botswana, 12=Zimbabwe, 13=Madagascar, 14=South Africa, Lesotho, Swaziland).

Also shown are the regions analysed separately.

Results – DJF Temperature – SRESA2

REA-mean

Simple-mean

REA-rmsd

Simple-rmsd

Rb

Rd

R

NatVar

(3.9 oC)

(0.5)

(0.3 oC)(0.9 oC)

(0.6 oC)

(3.9 oC)

(0.6)

(0.4)

Results – JJA Temperature – SRESA2

REA-mean

Simple-mean

REA-rmsd

Simple-rmsd

Rb

Rd

R

NatVar

(3.8 oC)

(0.5)

(0.3 oC)(0.7 oC)

(0.4 oC)

(3.7 oC)

(0.7)

(0.3)

Results – DJF Precipitation – SRESA2

REA-mean

Simple-mean

REA-rmsd

Simple-rmsd

Rb

Rd

R

NatVar

(0.0 mm/d)

(0.8)

(0.6 mm/d)(0.4 mm/d)

(0.4 mm/d)

(0.0 mm/d)

(0.9)

(0.7)

Results – JJA Precipitation – SRESA2

REA-mean

Simple-mean

REA-rmsd

Simple-rmsd

Rb

Rd

R

NatVar

(-0.1 mm/d)

(0.8)

(0.2 mm/d)(0.2 mm/d)

(0.1 mm/d)

(-0.1 mm/d)

(0.9)

(0.7)

Averaged changes across scenariosDJF Australia REA results

-0.08

-0.06

-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12

1 2 3 4

Temperature change (deg C)

Pre

cip

itatio

n c

ha

ng

e (

mm

/da

y)

SWWA MDB TROP

A1B

B1

B1

A2

B1

A2

A1B

A1B

A2

JJA Australia REA results

-0.3

-0.25

-0.2

-0.15

-0.1

-0.05

0

1 2 3 4 5

Temperature change (deg C)

Pre

cip

itatio

n c

ha

ng

e (

mm

/da

y)

SWWA MDB TROP

A1B

B1

B1

A2

B1

A2

A1B

A1B

A2

JJA

DJF

Predictions and Probabilities - Method

N

j j

ii

R

RmP

1

)(Probabilities of regional climate change:

Assume: each models’ reliability Ri is an indicator of the likelihood of its simulation the change simulated by a more reliable model is more likely to occur!

i ii

TT mPmP thi )()( thi TT

Threshold probability = summing over all P(mi) exceeding a given

threshold of climate change.

= probability of a temperature change exceeding ΔTth

where

PDFs = derivative of P(mi) )(

)(

T

mP i

Threshold probability

SRESA2 - Precipitation - JJAExample:

SWWA

Threshold probability – area averaged

TEMPERATURE

SRESB1 SRESA1B SRESA2

>2 >3 >4 >5 >2 >3 >4 >5 >2 >3 >4 >5

swwa 13 1 0 0 100 19 2 0 100 93 17 1 DJF mdb 23 6 0 0 99 35 8 1 99 77 35 6 tropics 37 1 0 0 97 38 11 0 99 92 40 3 swwa 15 0 0 0 79 8 0 0 100 42 1 0 JJA mdb 49 0 0 0 99 34 1 0 100 84 27 1 tropics 78 2 0 0 100 70 1 0 100 98 79 4 PRECIPITATION SRESB1 SRESA1B SRESA2 <-0.2 <-0.1 >0.1 >0.2 <-0.2 <-0.1 >0.1 >0.2 <-0.2 <-0.1 >0.1 >0.2 swwa 3 10 18 1 8 27 26 1 11 29 13 3 DJF mdb 12 30 36 22 39 59 18 16 15 30 30 22 tropics 25 37 44 38 37 40 43 38 34 42 50 46 swwa 44 67 0 0 65 88 0 0 72 83 2 0 JJA mdb 12 35 5 2 31 63 4 0 55 71 9 3 tropics 1 6 2 1 1 11 1 0 3 13 6 1

PDF’s for sres-A2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0 1 2 3 4 5 6

Temperature Change (oC)

Pro

ba

ilit

y D

en

sit

y

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0 1 2 3 4 5 6

Temperature Change (oC)

Pro

ba

ilit

y D

en

sit

y

All_ozswwamdbtropics

0

0.5

1

1.5

2

2.5

-1.6 -1.2 -0.8 -0.4 0 0.4 0.8 1.2 1.6

Precipitation Change (mm/day)

Pro

ba

ilit

y D

en

sit

y

0

0.5

1

1.5

2

2.5

-1.6 -1.2 -0.8 -0.4 0 0.4 0.8 1.2 1.6

Precipitation Change (mm/day)

Pro

ba

ilit

y D

en

sit

y

Reliability Contributions (%) - Australia

Normalised contributions (in %) to the overall model reliability for each CGCM. Good model performance and convergence leads to higher

contribution. If all models were equal, they would contribute 8.3% each.

ccs

m

cn

rm

csi

r

gfd

l

gfd

l_cm

2_

1

in

m

ip

sl_

cm4

mir

ocm

mp

i_ec

ha

m5

mri

pcm

uk

mo_

ha

dcm

3

Precipitation all_oz 9 6 8 13 8 11 7 4 13 9 4 7 swwa 10 9 4 12 11 11 5 2 14 12 3 6 mdb 4 7 7 11 13 9 7 4 13 10 6 8

DJF

tropics 11 6 8 12 6 10 4 8 12 8 5 9

all_oz 8 9 7 10 11 7 10 5 11 9 6 7 swwa 13 5 8 8 8 5 10 18 8 4 8 6 mdb 8 7 10 11 11 7 7 5 10 6 7 10

JJA

tropics 9 7 6 11 12 8 11 4 12 9 5 4 Temperature

all_oz 13 8 8 12 9 11 4 7 6 10 3 10 swwa 8 13 13 11 11 10 2 11 8 4 3 6 mdb 11 9 13 13 14 5 3 9 2 7 6 9

DJF

tropics 5 2 7 9 5 15 5 9 12 17 1 13

all_oz 4 3 2 11 13 3 14 16 21 4 1 8 swwa 2 4 3 7 8 3 13 30 11 11 1 6 mdb 2 3 2 8 12 2 11 15 28 6 1 10

JJA

tropics 11 2 1 15 18 2 11 8 15 4 1 13

i i

i ii

R

TRT~

REA mean

Summary

• REA is a useful tool to determine regional climate change from an ensemble of model simulations.

• Provides a means of producing probabilistic climate change predictions.

• Significantly lowers RMSD of mean climate change.• Obtain ‘skill measure’ of models through reliability analysis.• Summary for Australia:

– Magnitude of ΔT in winter is similar to summer.– No significant rainfall changes in DJF.– Significant decreases in rainfall in JJA over SWWA, MDB– On average, RD consistently better than RB – Resulting PDFs vary in shape depending on region (e.g. bi-

modal vs uni- modal, width)• Same analysis has been repeated over southern Africa (see

coming paper for details).

ACCSP

Australian Climate Change Science Programme Supported by the AGO CSIRO Marine and Atmospheric Research BMRC

Launched in October 2007 at GREENHOUSE 2007

Australian Climate Change Projections ReportAustralian Climate Change Projections Report 150 pages + Website access for projections

Any questions?

From: Allen and Ingram, 2002, Nature, 419, 224-232.

Milestones

Overview