Decadal and longer-term variability in
ENSO, ENSO teleconnections, and the
Walker circulation
Scott Power, Greg Kociuba, Jeff Callaghan
Centre for Australian Weather and Climate Research
Bureau of Meteorology
Contents
1.
2.
3.
4.
Decadal – interdecadal changes in ENSO
and ENSO teleconnections
A simple model for decadal variability in
ENSO and “ENSO-related” patterns of
variability
An inadequacy of this model
=>ENSO-driven multi-year variability
=>Multi-year predictability
Anthropogenic changes in the Walker
circulation and the SOI
The SOI – a product of French
Australian cooperation!
One of the world’s most important climatic
indices
 Used extensively to estimate and predict
changes linked to ENSO and changes in the
Walker circulation (e.g. in rainfall, agricultural
production, disease, streamflow, …)
 If SOI < 0 = > weaker Walker circulation

ENSO event frequencies and the SOI
12
10
No. of events & SOI
8
6
4
2
0
-2
-4
-6
Power and Smith, Geophys. Res. Lett., 2007; updated
-8
1900
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
2010
Year
No.(LN)
No.(EN)
No.(EN)-No.(LN)
JJASOND SOI
Innisfail 1918
3500 residents, only 12 houses
remained intact
Approximately 75-100 deaths
Mackay 1918
A new homogeneous
tropical cyclone data base
for north-eastern Australia




Taken over a decade to develop
Extensive primary sources of information:
Longest (non-palaeo) record in SH
[1872/73-2009/10]
homogeneous
Callaghan and Power, 2010: Variability and
decline in tropical cyclones making land-fall
over eastern Australia since the late 19th
century. Climate Dynamics.
Delayed-Action Oscillator (DAO)
dT(t) /dt = aT (t) - bT ( t - d)
 T = SOI or east Pacific SST; a, b >0; d=time delay
Analytic solutions to the DAO Equation
(Power 2010) with T(t) = T0 eat, t ≤ 0 :
N 1
T (t )  T0 e
at

k 0
[
( )
k
( t /   k  1) ],
k
k!
Cf. numerical solutions of Battisti and Hirst (1989)
t>0
Simple model for DV(ENSO)
Fitting DAO to SOI:
dT/dt = aT-bT(t-d) + Noise
d=7mo, a=0.13/yr, b=1.4/yr
 Noise-driven damped
oscillation with period =3.8 yr,
decay time (e-folding scale) =
0.9 yr
dT/dt(t) - bT ( t - d) + Noise or
d2T/dt2 = -σ02T + 2σR dT/dt + Noise
cf. Thompson and Battisti (2000), Jin (1997),
Meinen and McPhaden (2000)
Noise has decadal/interdecadal tail so stochastic forcing drives some
of the decadal variability in ENSO
(even if ENSO is partially self-sustained)
“The relationship between ENSO and Australian climate
in both the model and the observations is strong in some
decades, but weak in others. A series of decadal-long
perturbation experiments are used to show that if these
interdecadal changes are predictable, then the level of
predictability is low”.
J. Climate, 2006
Suppose ELF is a low pass filtered ENSO index, e.g.:
1
ELF =
m
------------
Σ E t-k., and
m+1
r(E, SST)= α.
k=0
< ELF, SSTLF>
Then r(ELF, SSTLF) =
________________________ .
√ [ < ELF, ELF > <SSTLF, SSTLF> ]
< Et, Et + Et-1 + Et-2 + Et-3 + … + Et-m> / (m+1)2
+ < Et-1, Et + Et-1 + Et-2 + Et-3 + … Et-m> / (m+1)2
+ … + < Et-m, Et + Et-1 + Et-2 + Et-3 + … Et-m>/ (m+1)2
= (m+1)/(m+1)2=1 / (m+1),
(1)
Now < ELF, ELF > =
where we have used the fact that E is white noise. Similarly
(2)
< SSTLF, SSTLF > = 1/(m+1), so
(3)
< SSTLF, ELF > = < SSTt, Et + Et-1 + Et-2 + …+ Et-m>/ (m+1)2
+ < SSTt-1, Et + Et-1 + Et-2 + … Et-m> / (m+1)2
+ … + < SSTt-m, Et + Et-1 + Et-2 + … Et-m>/ (m+1)2
= (m+1) < SSTt, Et >/ (m+1)2
=
α/ (m+1).
(4)
Using (2)-(4) in (1) then gives
r(ELF, SSTLF) = α.
Power and Colman, 2006: Climate Dynamics
Decadal pattern
much broader
=> Different
physics in offequatorial “wings”
c.f. Zhang et al.
1998; Mantua et
al. 1997; Power
et al. 1999;
Power and
Colman 2006
C G C M : N IN O 3 (4 y r a ) & 3 1 0 m W in g In d e x
0 .4
Index
0 .2
0
- 0 .2
- 0 .4
0
20
40
60
80
100
Ye a r
Off-equatorial sub-surface variability is a low
pass filtered version of ENSO variability
Decadal-long
Perturbation
Experiments

13 years
Discovery: Sub-surface ENSO-driven
off-equatorial decadal variability is
highly predictable

Low pass filtering due to dominance
of low frequency oceanic Rossby
waves in response to ENSO windstresses
Power and Colman, Climate Dynamics, 2006; cf. Newman et al. 2002
Conclusions so far
1.
2.
3.
4.
Large and important decadal – interdecadal
changes in ENSO and ENSO indicators have
been observed and modelled
A simple model for decadal variability in
ENSO and ENSO-like patterns of variability
was presented
While useful in capturing some of the
variability, the simple model is inadequate
because ocean acts as LPF on ENSO forcing
This yields e.g. ENSO-driven multi-year
predictability in off-equatorial wings
Other evidence for more
sophisticated physics





Kirtman and Scopf (1998)
Kleeman et al. (1999)
Wang et al. (20xx)
McGregor et al. (2008)
…
4.
Anthropogenic changes in the
Walker circulation and the
SOI
The SOI [and N(EN) – N(LN)]
Power and Kociuba, Climate Dynamics, 2010 (submitted); see also
Vecchi et al. (2006); Meehl et al. (2007)
The Walker circulation weakens in
response to global warming
(Vecchi et al. 2006; Meehl et al. 2007
/IPCC AR5;Power and Kociuba 2010)
“This obviously means that the SOI also
declines in response to global warming”
“So part of observed decline in SOI due
to global warming”
Or is it?
The Walker circulation weakens in
response to global warming
The SOI does not decline in
response to global warming.
The large observed decline in the
SOI is therefore natural.
Power & Kociuba, Climate Dynamics (submitted), 2010
We can therefore infer (taking
models at face value) that:

Observed weakening of Walker circulation
over 20th century due to both natural
variability and external forcing
Supports conclusions of Meehl et al. 2009
Future work (for dessert*)

Further clarify causes and relative importance of
decadal variability in ENSO activity

Important that simulation of ENSO in CGCMs
becomes more realistic
* wine earned by providing numerous references
during talk
The End – thank you for listening!
Scott Power
Centre for Australian Weather and Climate Research
Bureau of Meteorology
Impact of global warming on the SOI

SOI used extensively to estimate and predict changes linked
to ENSO and changes in the Walker circulation (rainfall,
streamflow, disease, tropical cyclones, …)

Correlation coefficient between SOI and equatorial MSLP
pressure gradient (ΔP) = 0.83

ΔP = BoxE(5˚S-5˚N, 200˚E-280˚E) - BoxW(5˚S–5˚N, 80˚E–
160˚E)
El Niño, weaker Walker circulation, SOI < 0
Summary




There has been pronounced interdecadal variability
in the Walker circulation during the 20th century and
an overall weakening of the Walker circulation in
recent decades.
Weakening due to both global warming and natural
variability.
Global warming weakens Walker circulation but
(surprisingly) increases SOI. The SOI is not a good
guide to changes in Walker circulation forced by
global warming.
Observed decline in SOI natural and largely due to a
natural increase in dominance of El Niño over La
Niña activity.
MMEM of the SOI (A2 and C20), the observed SOI, and
associated confidence levels, 30yr averages, JJASOND
SOI (Observed and modelled)
8
6
4
2
0
-2
-4
1900
1920
1940
1960
1980
2000
2020
2040
2060
2080
95% confidence interval for 30yra MMEM SOI
95% confidence interval for 30yra SOI
MMEM, 30yra
Obs SOI, 30yra
2100
Analytic Solutions of the Linear DAO
Equation, a=0.13/yr, delay=7mo, ICs:
T(t<0)=T(0)exp(at)
3
2
T(t*)
1
0
-1
1
3
5
-1
7
9
b=0.01
b=0.5
-2
b=1.4
b=3
-3
t*=time/(delay time)
Power (2010), Theoret. Appl. Climatol.; cf.
numerical sols of Battisti and Hirst (1987)
Contents
1.
2.
3.
Australian tropical cyclones: interannualdecadal variability, and long-term trends
and links to ENSO and the SPCZ
Interdecadal variability and trends in the
Walker circulation, ENSO activity, and the
Southern Oscillation Index
Predictability of ENSO teleconnections,
origin of decadal ENSO-like patterns
C G C M : S S T In d ic e s
N IN O 3 & S S T (2 0 5 -2 7 0 E , 8 -1 2 S )
T em p An o m aly (K)
1 .3
0 .6 5
0
- 0 .6 5
- 1 .3
0
20
40
60
Y e ar
Power and Colman, Climate Dynamics, 2006
80
100
Wind-stress forced shallow water
model and simplified coupled models
Equatorial region forcing
Off-equatorial region forcing
Off-equatorial region forcing
McGregor et al., 2007
First EOFs from Wind-forced Shallow Water Model
Forcing applied
everywhere
Off-equatorial
forcing only
McGregor, Holbrook and Power, 2007; see also Wang et al.
2003, Part 1 – “wind-stress in eastern tropical &
subtropical basin most effective in driving this kind of
zonal equatorial response (from theory and windforced SWM)
Summary
DCV in El Niño-Southern Oscillation is an
important part of DCV in the Pacific
 Randomness seems to explain a lot
 Nevertheless predictability is evident
 as ENSO-modified red noise
 equatorial variability driven by offequatorial wind-stresses
 ENSO-driven decadal climate variability in
sub-surface ocean
…

C G C M : S S T In d ic e s
N IN O 3 & S S T (2 0 5 -2 7 0 E , 8 -1 2 S )
T em p An o m aly (K)
1 .3
0 .6 5
0
- 0 .6 5
- 1 .3
0
20
40
60
Y e ar
dT/dt = -aT + bE + cN
Power and Colman, Climate Dynamics, 2006; see also
Newman et al. 2007 for similar behaviour in PDO Index
80
100
No. severe land-falling TCs & the SPI
standardized, normalized, 11yr raves, Correl Coeff=0.64
3
2
1
0
-1
-2
1890
1900
1910
1920
1930
1940
1950
Nov Apr neg SPIa_11yra/SD
1960
1970
nTCsa_11yra/SD
1980
1990
2000
2010
Newspaper archives








Brisbane Courier Mail held at the
Queensland State Library
Maryborough Chronicle researched by
the Hervey Bay City Council
Townsville Historical Society
Mackay Mercury
Cairns Post
Rockhampton Morning Bulletin
Cairns Historical Society
Archives of various newspapers held by
the Bureau of Meteorology, Brisbane
Interannual variability in the SOI is an
excellent proxy for interannual variability in
ΔPequator and the Walker circulation:
 SOI > 0 => La Niña
 SOI < 0 => El Niño
 But under global warming ΔPequator
decreases whereas the SOI increases

Paths of cyclones in El Nino (top) vs La Nina (bottom)
years

Power et al., 1999: Clim. Dyn.
CGCM: All-Australia Rainfall v. NINO4
0.4
0
-0.4
-0.8
0
100
Year
Correlation Coefficients in 13 yr running blocks NINO4/ozT, NINO4/ozR in CGCM
1
correl. coeff
CGCM variable
0.8
0.5
0
-0.5
-1
0
20
40
60
time
80
100
Non-linear impact of ENSO on
southwest USA/Mexico?
Power et al. 2006
Summary so far







ENSO teleconnections to Australia vary substantially from
generation-to-generation
This variability seems to have very little predictability
The changes are in phase with IPO/PDO because of nonlinearity in the ENSO teleconnection to Australia
This gives interesting effects that seem to point to decadal
predictability but this is not necessarily the case
Non-linear teleconnections may exist elsewhere
The IPO and PDO have strong links
Discussion of PDO emphasizes North Pacific
Correlation Coefficients
The SOI and beer consumption in
Australia
Number of visits to pub/month
Hot, dry
Thirsty
Go to pub to
get drink
Cool, wet
=> Go to pub to
get out of rain
CGCM: All-Australia Rainfall v. NINO4
CGCM variable
0.8
0.4
0
-0.4
-0.8
0
100
Year
Power et al., J. Climate, 2006
Decadal pattern
much broader
=> Different
physics in offequatorial “wings”
Brisbane - 1893 flood – 23 deaths
Next: Decadal changes in ENSO
teelconnections
Discovery: Australian response to
ENSO is asymmetric in observations
and CGCM
CGCM: All-Australia Rainfall v. NINO4
(JJASOND)
All-Australia Rainfall v. SOI, 1900-2004, Annual Data
800
Observations
5
Coupled
GCM
4
600
Rainfall
Rainfall
3
400
2
1
0
-10
200
-25
-20
-15
-10
-5
0
SOI
5
10
15
20
25
-5
-1 0
5
-2
NINO4
Power et al. 2006: J. Climate.
10
Fitting DAO to SOI:
dT/dt = aT-bT(t-d) + Noise
d=7mo, a=0.13/yr, b=1.4/yr
 damped oscillation with
period =3.8 yr, decay time
(e-folding scale) = 0.9 yr
Power 2010
Decadal changes in ENSO
teleconnections
SPI (Nov-Apr) & SOI (June-Dec), Correl Coeff = -0.79, 11yra
2
1.5
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
1890
1900
1910
1920
1930
1940
1950
Nov Apr neg SPIa_11yra/SD
1960
1970
1980
June-Dec SOIa_11yra/SD
1990
2000
2010