Past and future changes in temperature extremes in

Australia: a global contextWorkshop on metrics and methodologies of estimation of extreme

climate events, Paris 27th – 29th September 2010 Lisa Alexander, Climate Change Research Centre, UNSW, Sydney,

Australia

The “land of drought and flooding rains”

• Population predominantly live in temperate zone and in sub-tropical zones in the east and south-west

• Climate strongly modulated by ENSO variability

• Sustained period of drought in south over the last decade

• Extremely high temperatures

Highest recorded temperatures

State Temperature (°C)

Date Place name

Latitude Longitude

South Australia

50.7 02.01.1960

Oodnadatta

-27.56 135.45

Western Australia

50.5 19.02.1998

Mardie -21.19 115.98

New South Wales

49.7 10.01.1939

Menindee -32.39 142.42

Queensland

49.5 24.12.1972

Birdsville -25.90 139.35

Victoria 48.8 07.02.2009

Hopetoun -35.72 142.36

Northern Territory

48.3 01.01.1960

Finke -25.58 134.57

Tasmania 42.2 30.01.2009

Scamander

-41.46 148.26

2009 heatwaves

Average winter Tmin increase has led to a substantial decrease in probability of temperatures < 1°C

Mean 1957-1980

Mean 1981-2005

prob

abili

ty

6.9 7.6

°C

Have temperatures become more extreme?

Source: Nicholls and Alexander, 2007

Illustrative example for Melbourne

Seasonal trends in temperature

Tmin/TN90p Tmax/TX90p

Means vs extremes1957–2005

Only statistically significant trends are shown in colour

Triangles represent increasing/decreasing (upward/downward) trends in the upper 10th percentile at individual stations

The size of the triangle reflects the magnitude of the trend

Bold indicatesstatistically significant change

Spatial correlations

• Trends in extremes generally well correlated with trends in means across Australia in every season

• Absolute trends in extremes greater than mean trends when averaged across Australia

summer

winterEx

tre

me

te

mp

era

ture

Mean temperature

Max Min

Example method:- Angular distance weights for ith station, wi, which are defined as:-

where f is the correlation function:L is the decorrelation length scaleθ is the bearingk sums over all stations within circle of influencem adjusts function decay

Scaling issues

k ,

)cos(11 ki

f

ff

k

k

ikk

iim

m

m

Need to define minimum number of stations for a

gridbox calculation

Search Radius

Calculation of decorrelation length scale, L

m=1 m=4 m=10

•Station correlations are averaged into 100km bins within 5 latitude bands

•2nd order polynomial is fitted

•L is defined where the fitted function falls below 1/e

•L is calculated for each season and year for each index

Lri

ief /

Source: Alexander et al. 2006

How good are gridding methods for extreme

temperatures?

Actual Simulated

Errors associated with gridding temperature

extremes

Homogeneity issues

Trends in some temperature extreme indices, 1951-2003

Comparison with other regions

Source: Caesar et al. 2006

USA

Europe

China

Russia

Australia

Comparison of observed and modelled timeseries

Observed vs modelled trends, 1957 to 1999

Index Obs Multi-model

Warm nights 1.11 ±0.06

1.15 (0.48/1.87)

Frost days -0.89 ±0.07

-0.19 (-1.46/0.22)

Extreme temperature range

-0.19 ±0.02

0.04 (-0.29/0.31)

Heat wave duration

7.05 ±0.33

0.26 (-0.31/0.91)

The importance of natural variability

Model (SSTNAT) Obs

Reds where La Niña warmer than El Niño – crosses where difference significant

versus

ENSO and global daily maximum Tmax

Extreme Tmax Australia

Observed hottest day max 0.5°C - 1°C

warmer during El Niño than La Niña

SSTNAT hottest day max up to 0.3°C cooler during El Niño than La

Niña

ALL hottest day max up to 0.6°C cooler

during El Niño than La Niña

Anthropogenic versus natural forcing

Two models (CCSM/PCM) with output from different forcings

Results show that some temperature extremes are inconsistent with natural-only forcings

Future projections for different scenarios

Source: Alexander and Arblaster (2009)

Future changes 2080-2099 minus 1980-1999 (A1B)

Multi-model agreement across most of Australia for large significant increases in

warm nights and heat wave duration

Changes scale with strength of emissions

Index Aust/Global (A1B)

B1/A1B A2/A1B

Warm nights 0.86 0.65 1.11

Frost days 0.45 0.86 1.15

Extreme temperature range

-0.53 0.58 2.14

Heat wave duration

0.30 0.50 1.40

Conclusions • Over the last 50 years Australia has seen trends in

temperature extremes associated with warming (the exception is northwest Australia in DJF)

• Work is ongoing on how to best address issues of scale

• Natural climate variability is important and models do not appear to capture some important processes

• Anthropogenic forcing is also important in capturing trends and model simulations indicate continued warming of temperature extremes in the future

• The magnitude of future changes appears to scale with strength of emissions