The role of stochastic forcing in ensemble forecasts of the 97 El Nio

1
The role of stochastic forcing in ensemble
forecasts of the 97 El Niño Harry Hendon Centre
for Australian Weather and Climate Research
Bureau of Meteorology Li Shi, O. Alves, D.
Anderson, G. Wang (dynamical predictions) M.
Wheeler, C. Zhang (observational analysis
MJO-ENSO) Explore role of MJO for spread of
large ensemble of seasonal forecasts initialized
1 Dec 96 using Bureaus POAMA coupled seasonal
forecast model POAMA supports reasonable ENSO
(too biennial/shifted west but good
amplitude) relatively strong MJO (a
little too slow) that strongly interacts with
upper ocean
2
Usfc
SST
Onset of 97/98 El Nino
Z20C
V
3
Bergman et al 2003
SST
SST
OLR
D20
Tx
Usfc
1 Dec 96
1 Apr 97
Tx
West Pac (130-150E)
1 Feb 97
1 Jun 97
Central Pac (180-160W)
Suggestive of a coupled feedback e.g., Kessler
et al
4
Jul(1)
MJO-ENSO behaviour 1979-2006 Contours
regression Shading correlation
Jan(1)
Usfc
SST
Jul(0)
Regression onto Nino34 Jan (1)
Jan(0)
120E
180E
120W
OLR MJO amplitude
Usfc MJO amplitude
Hendon et al. 2007
5

• Observational analysis support notion that MJO is
  relevant to the timing, initial growth and
  strength of El Niño
• maybe not necessary for the event itself
• Can effects be simulated?
• Can mechanism of interaction be clarified?
• Are MJO effects predictable?
• What is implication for limiting prediction of
  strength?

6

• Generate a large (90 member) ensemble of
  forecasts for onset of the 97 event using 1 Dec
  1996 initial conditions
• POAMA model (T47L17 AGCM MOM2 OGCM 1/3 x 2 deg)
• Observed ocean initial conditions POAMA
  assimilation
• Atmos IC ensemble mean for 1 Dec 1996 from 36
  AMIP integrations
• Initial Conditions are identical (0.01 C noise
  in SST)
• Spread comes about solely from internal
  stochastic variability (noise?)
• Every forecasts leads to El Nino, ranging 0.5-2.5
  C
• Explore role of MJO for spread

7
POAMA ENSO Skill Hindcast 1980-2006 Skill
comparable to other centres/systems Reasonable
amplitude (but shifted west)

obs std
POAMA
persistence
POAMA std
8
Space-time spectrum Usfc
50d
90d
POAMA supports reasonable MJO a little too slow
but reasonable amplitude
Zhang et al 2006
9
All members develop El Nino. Observed well
captured by spread (not all prediction systems
able to capture strong initial growth) Separate
into warmest (30) and coldest (30) terciles
10
OLR
U10
SST
D20
31 Aug
90-member ensemble mean forecast
1 Dec
Warmest tercile coldest tercile
Differences apparent after 1 month
11
Tercile distribution of NINO3.4 each month
according to tercile at the end forecast (August).

12
(a)
(b)
Maritime Cont.


East IO
West Pac
Coldest tercile
Warmest tercile
MJO EOFS as per Wheeler and Hendon
Initial MJO propagates systematically further
east in the warmer forecast blue first
month red second month black third month
13
OLR/winds composite for 1st MJO event each
forecast Warmest tercile
20d
10d
0d
-10d
-20d
14
Warm minus cold relative to first MJO event
D20 shaded, interval 3 m U10 bold blue contours,
interval 0.8m/s SST thin black contour, interval
0.2ºC
MJO leads changes in ocean
15
Evolution of ENSO in first 2-4 months had strong
impact on longer term evolution forecasts that
were on average warmer at end of 4 mths were
warmer at end of 9 mths Largely accounted for by
MJO behavior greater eastward penetration
early on lead to stronger ENSO Clear cause and
effect MJO leads changes in ocean but
consistent with coupled feedback
16

• Implications
• Realistic MJO required or else forecast spread
  will reduced or skewed
• MJO limits our ability to forecast strength of
  ENSO
• All forecasts generate MJO, suggesting initial
  condition was favorable to MJO
• explore MJO effects in forecasts in other El
  Ninos (eg 2002)
• ENSO forecast should be sensitive to atmos
  initial condition (phase of MJO)

MJO initially in Indian Ocean
MJO initially in west Pac
17
Composite relative to first MJO event in each
forecast Surface zonal wind along Eq
Warmest tercile
Lag (days)
Coldest tercile
Lag (days)
18

• Conclusions from observational analyses
• Robust lagged relationship between boreal spring
  MJO activity in western Pacific (and globally)
  and subsequent El Niño strength
• Causal relationship sustained westerly anomalies
  that accompany enhanced MJO activity appear to
  project efficiently onto developing El Niño
• Coupled feedback proposed (Kessler et al)
• enhanced MJO in boreal spring expands warm pool
  and promotes surface westerlies, thus El Niño,
  which further promotes MJO.