role of stochastic forcing in enso variability in a coupled gcm

Role of Stochastic Forcing inENSO variability in a coupled GCM

Atul KapurChidong Zhang

Javier Zavala-Garay

Acknowledgements: Ben Kirtman, Amy Clement

• Stochastic Forcing (SF)– Atmospheric variability uncoupled to the ocean

Annual Cycle

• Extent to which the ENSO in CGCMs is driven by SF• Contributions of Madden Julian Oscillation (MJO) and non-MJO• Dynamical regime of underlying coupled system – Stable or

Unstable

Introduction

ProcedureENSO

SFSF SFSF

ENSO

CZZ model

ENSO ENSO

Extract Extract

CompareCompare CompareCompareRole of SF

Model and Data

Variant of Zebiak and Cane (1987) model• Chaos switched off (Mantua and Battisti 1995)

• Admits daily SF: Decorrelation time of tropical weather ~ 3-8 days

Bureau of Meteorology Research Center (BMRC) CGCM (Zhong et al. 2004)

• A 163-year run• Realistic ENSO (Wu et al. 2002) and intraseasonal

variability (Zhang et al. 2006)

NCEP-2 Reanalysis (1979-2007)

CZZ model

ProcedureENSO

SFSF SFSF

ENSO

CZZ model

ENSO ENSO

Extract Extract

CompareCompare CompareCompareRole of SF

Stochastic Forcing• Statistical model of u10 anomalies predicted by SST anomalies

u10 = A sst + uResidual

• Wavenumber frequency spectra:

Zonal wavenumber

Period

Inter-annual

Intra-seasonal

Coupled Residual MJO

Caveats: Linear, Contemporaneous, Additive(CGCM)

(Hilbert EOF)

ProcedureENSO

SFSF SFSF

ENSO

CZZ model

ENSO ENSO

Extract Extract

CompareCompare CompareCompareRole of SF

Simulations using NCEP-2 SFPower Spectra

• CZZ model able to reproduce spectrum• ENSO statistics better for MJO than non-MJO forcing• CZZ model performs best in weakly stable regime

CZZ95 % confid

NCEP-2

0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0

Freq (cycles/yr)

Pow

er *

freq

Simulations using NCEP-2 SFSeasonal Variance

CZZ warmNCEP-2 warm

CZZ coldNCEP-2 cold

• Warm phase better simulated than cold phase in terms of seasonal variance

3

2

1

0

-1

-2J F M A M J J A S O N D

Normalizedvariance

Simulations using NCEP-2 SFSeasonal Autocorrelation

Total MJO Non-MJO

Lag (month)

D

O

A

J

A

F

Startingmonth

ProcedureENSO

SFSF SFSF

ENSO

CZZ model

ENSO ENSO

Extract Extract

CompareCompare CompareCompareRole of SF

Simulations using CGCM SFPower Spectrum

• SF is able to reproduce even local peaks in power spectrum• Results using MJO compare better to “truth” than non-MJO

CZZ95 % confid

CGCM

Simulations using CGCM SFSeasonal Variance

Total SF MJO Non-MJO CZZ warmCGCM warm

CZZ coldCGCM cold

• SF unable to reproduce the seasonal variance of ENSO exhibited by the BMRC CGCM

• Contribution of non-MJO appears to be higher than MJO

Norm.variance

Simulations using CGCM SFSeasonal Autocorrelation

Total MJO Non-MJO

D

O

A

J

A

F

Startingmonth

Lag (month)

ProcedureENSO

SFSF SFSF

ENSO

CZZ model

ENSO ENSO

Extract Extract

CompareCompare CompareCompareRole of SF

Conclusions• Role of SF in BMRC CGCM ENSO– At least the warm phase can be reasonably

simulated using SF– MJO contribution is higher than non-MJO– Underlying dynamical state of coupled system

appears to be weakly stable– Seasonality of ENSO cannot be reproduced by SF

• Procedure can be implemented on any CGCM– Even on runs with long temporal span