MJOENSO Connection

1
MJO-ENSO Connection on Interannual to
Interdecadal Timescales
Chidong Zhang, Atul Kapur RSMAS, University of
Miami Javier Zavala-Garay Rutgers
University and Harry H. Hendon BMRC
2
Issues What are roles of the MJO vs non-MJO
stochastic forcing in ENSO? What is the role of
stochastic forcing in different ENSO dynamic
regime? Are the MJO and ENSO connected on
decadal and interdecadal timescales? General
Approach derive stochastic forcing from
observations and GCM simulations, and use it in
simple coupled ENSO models.
3
Tools and Approaches I
To assess the role of stochastic forcing,
especially by the MJO, in ENSO
4
tn tMJO tNMJO
Zavala-Garay et al 2005
5
Zavala-Garay et al 2007
6
Zavala-Garay et al 2007
7
Oceanic Kelvin wave reduction in equatorial
upwelling
Zavala-Garay et al 2007
8
? MJO and Nino3 SST
?NMJO and Nino3 SST
Lag (day)
Zhang and Gottschalck 2001
9
? MJO and Nino3 SST
Lag (day)
Month
Zhang and Gottschalck 2001
Hendon et al. 2001
10
Modified Cane-Zebiak (CZ) Model
Chaotic Model (CM) standard configuration of the
CZ model (unstable-chaos) Stochastic Model
(SM) 1 CM CZ (neutral) SF
(NCEP2) CZ (stable) SF
(NCEP2) Stochastic Model (SM) 2 SM2 CZ
Battisti (1988) subsurface temperature scheme
Mantua and Battisti (1995) SST advection
scheme (unstable-limit cycle) SM2
(neutral) SF (NCEP2) SM2 (stable) SF
(NCEP2)
11
Synthesize long-term stochastic forcing based on
NCEP2 reanalysis

• Determined from residual of SVD-based
  multivariate regression of Reynolds SST and NCEP2
  surface zonal wind
• 25 years (1979-2003)
• The original residual can be represented by the
  sequence (year,month) (1979,1),(1979,2),,(2003,
  11),(2003,12)
• Synthetic timeseries were generated by
  bootstrapping the years, so a possible
  realization of SF is (1985,1),(1983,2),.,(1994,12
  ),(1979,1)
• The of defining exactly the same sequence of
  noise is very low, and the probability having
  that noise acting on the same model state is even
  lower
• The resulting SF preserves key characteristics of
  the noise as its seasonality, temporal and
  spatial decorrelation scales, and structure and
  eastward propagation of the MJO

Ensemble simulations 500 members, each of 136
years
12
Zavala-Garay 2007
13
Observed
Zavala-Garay 2007
14
Zavala-Garay 2007
15
Observed
Zavala-Garay 2007
16
Kolmogorov-Smirnov (KS) test two samples from
different populations
96
97
74
88
32
32
Zavala-Garay 2007
17
Tools and Approaches II
BAM3 160 yr run
Cane-Zebiak model (unstable, neutral, stable)
To assess the role of stochastic forcing,
especially by the MJO, in the decadal-interdecadal
variability of ENSO
18
MJO
19
SST - Usfc SVD leading modes
observation
BAM3 (BAMC)
Nino3 SST Histograms and Spectra
20
SST EOF PC1 from BAM3
40 yrs
160 yrs
A Kolmogorov-Smirnov test shows that all PDFs are
different from each other, except for the third
40 years and the total.
21
MJO Seasonal Cycle
bidecadal means
bidecadal anomalies
22
Tools and Approaches III
global reanalysis
CAM3.5
ENSO time series Spectra PDF Lag-correlation
CAM3
CAM3
To quantify the roles of the new cumulus
parameterization in improving ENSO simulations
through producing more realistic stochastic
perturbations or mean state.
23
END