A review of

1
A review of 2005/2006 ENSO Predictions Tony
Barnston Lisa Goddard Shuhua Li International
Research Institute for Climate and Society (IRI)
2
Two Events of Interest A brief La Nina(), and a
new El Nino ()


La Nina, Dec-Mar 05-06
El Nino, October 06
3
(No Transcript)
4
Last weeks SST anomaly

--------------------------------------------------
------------------
5
(No Transcript)
6
Renewed, east-extended westerly wind anomalies
7
mid-Oct 06
8

9

Climate Prediction Center NCEP
10
From Climate Diagnostics Center, Boulder, Colorado
Black This year


11
The onset of La Nina conditions in December
2005, after a largely ENSO-neutral northern
autumn, was unprecedented since 1950. How did
it emerge?
12
Precursor to Brief La Nina Conditions in late 2005
Sep 2005
Colombia
?
?
?
e q u a t o r------------
?
?
L
Peru
?
H
H
110W
Note Vern Kousky says this anticyclone is
spurious, due to Reanalysis biases.
13
Precursor to Brief La Nina Conditions in late 2005
Oct 2005
Colombia
?
?
e q u a t o r------------
?
H
Peru
H
H
110W
Note Vern Kousky says this anticyclone is
spurious, due to Reanalysis biases.
14
Precursor to Brief La Nina Conditions in late 2005
Nov 2005
Colombia
?
e q u a t o r------------
H
H
H
Peru
H
L
110W
Note Vern Kousky says this anticyclone is
spurious, due to Reanalysis biases.
15
Mid-2005 to Mid-2006 Brief La Nina Conditions
Five-Day Zonal Wind, SST, and 20º Isotherm Depth
Anomalies along Equator Zonal Wind
SST
20º Isotherm Depth


16
How did statistical and dynamical ENSO prediction
models fare regarding the brief 2005-06 La Nina
conditions and the late 2006 El Nino development?
17
Cool Conditions Coming
18
Cool Conditions Coming
19
Cool Conditions Coming
20
El Nino Conditions Coming
21
El Nino Conditions Coming
22
El Nino Conditions Coming
23
.
24
.
--------------------------------------------------
---------------------------
25
.
--------------------------------------------------
---------------------------
Unanticipated by stat models
26
.
--------------------------------------------------
-------------
.
--------------------------------------------------
-------------
Unanticipated by stat models
27
Blue asterisk means the model did
comparatively well during this period.
Hybrid Models Dynamical ocean, statistical atmosp
here

.
28
Blue asterisk means the model did
comparatively well during this period.
Abbreviated dynamical coupled models
.
29
Blue asterisk means the model did
comparatively well during this period.

Fully comprehensive coupled models Run in the
U.S.

.
30
Blue asterisk means the model did
comparatively well during this period.

Fully comprehensive coupled models Run outside
of U.S.


.
31
Blue asterisk means the model did
comparatively well during this period.
Statistical models based on the redness of the
coupled system

.
32
Blue asterisk means the model did
comparatively well during this period.
Statistical models based on linear or
nonlinear regression

.
33
Blue asterisk means the model did
comparatively well during this period.
Statistical models based on unique designs
(regression to construct an analog, neurological
network).


.
34
The following graphs show how each model did in
predicting the ENSO state from JFM 2006 to ASO
2007, by one-month increments (so there are up to
18 3-month periods), for lead times of 1 to 5
months (1 month lead is really zero lead, such
as a forecast for JFM made from December or
even earlier initial conditions).
Scores Correlation (uncentered means not
removed) RMSE Skill score ( 1
RMSEmodel/RMSEclim ) The blue curve shows the
model skill. The pink curve shows the skill of
persistence from the most recently
observed one month SST.
35
(blue line)
persistence
36
(blue line)

persistence

37
(blue line)
persistence
38
(blue line)
persistence
39
(blue line)
persistence
40
(blue line)

persistence

41
(blue line)

persistence

42
(blue line)
persistence
43
(blue line)

persistence
44
(blue line)

persistence

45
(blue line)

persistence

46
(blue line)
persistence
47
(blue line)
persistence
48
(blue line)

persistence

49
(blue line)
persistence
50
(blue line)
persistence
51
(blue line)
persistence
52
(blue line)

persistence

53
(blue line)

persistence

54
(blue line)

persistence

55
(blue line)
persistence
56
(blue line)
persistence
57
(blue line)
persistence
58
---------------- -----------------
-----------
---------------- -----------------
-----------
59
End of 2005 CPC and IRI statements about La Niña
IRI, mid-October The likeli-hood of maintaining
neutral conditions is at least 95 through the
end of 2005. The probability of a La Niña
dev-eloping between now and the end of 2005 is
approxi-mately 1.
CPC, early October .current conditions and
recent observed trends support a cont-inuation
of ENSO-neutral cond-itions for the next 3-6
months.
CPC, early November current conditions
stronger-than-average easterly winds over the
central equatorial Pacific) and recent observed
trends (decreas-ing SST anomalies throughout the
central and eastern equatorial Pacific) do not
support the development of El Niño. Rather, they
support either a continuation of ENSO-neutral
conditions or the development of weak La Niña
conditions during the next 6-9 months.
IRI, mid-November The likeli-hood of maintaining
neutral conditions is virtually certain through
the end of 2005. The probability of a La Niña
devel-oping between now and the end of 2005 is
less than 1.
60
End of 2005 CPC and IRI statements about La Niña
CPC, early December current conditions
(stronger-than-average easterly winds over the
central equatorial Pacific) and recent observed
trends (decreasing SST anomalies throughout the
central and eastern equatorial Pacific) do not
support the development of El Niño. Rather, they
support either a continuation of ENSO-neutral
conditions or the development of weak La Niña
conditions.
IRI, mid-December the probab-ility of a La Niña
developing over the next season is approximately
10, increasing to about 20 by mid-2006. The
chances for developing El Niño conditions are
estimated at less than 1 through March 2006,
increasing to approx-imately 20 by mid-2006. The
likelihood of maintaining neutral conditions
remains very high through early 2006.
61
Middle to Late 2006 CPC and IRI statements about
El Niño
IRI, mid-August The likeli-hood of maintaining
neutral conditions is at 60 through the end of
2006, and the probability of an El Niño
dev-eloping between now and the end of 2006 is
approxi-mately 35 to 40.
CPC, early August Synopsis ENSO-neutral
conditions are expected to continue for the next
one to three months, with a 50 chance that weak
El Niño conditions will develop by the end of
2006.
IRI, mid-September The likeli-hood of
maintaining neutral conditions through the end of
2006 is about 40. The probability of an El Niño
devel-oping between now and the end of 2005 is
approximately 60. Likely El Niño strength would
be weak.
CPC, early September Synopsis ENSO-neutral
conditions are expected to continue for the next
one to two months, with weak warm episode (El
Niño) conditions likely by the end of 2006.
62
Middle to Late 2006 CPC and IRI statements about
El Niño
CPC, early October Synopsis El Niño coniditions
are likely to continue into early 2007.
Collectively, these conditions are consistent
with the early stages of El Niño in the tropical
Pacific.
IRI, mid-October the probab-ility of continuing
El Niño condi-tions over the next season is
approximately 80. The likelihood of neutral
conditions is just under 20. Likely strength of
El Niño is weak to moderate.
CPC did slightly better, on average, with these
two somewhat unusual ENSO developments than IRI.
But CPC usually does not usually quantify its
ENSO forecasts probabilistically.
63
Conclusion
In the middle of the 2000-2009 decade, we still
have weak to fair skill in predicting the onset
of non-neutral ENSO conditions during the early
half the calendar year. In new situations (e.g.
Dec. 2005, or autumn 2006), we receive surprises
even at times of the year when we thought we had
high confidence. Often, we predict non-neutral
ENSO developments once we observe signs of the
emergence of an event in the middle part of the
calendar year. Sub-surface sea temperature
anomalies may sometimes precede SST anomalies by
a month or more. Once an event begins, we are
skillful at predicting its evolution and are
helpful to users whose season of climate effects
is later. If climate change is causing
different relative relationships with- in the new
normal ranges, then statistical models may fail
us. We are MUCH better than we were 20 years
ago. To the layman, however, our ENSO predictive
skill likely still looks weak to fair.