Title: Monsoon Intraseasonal Variability
1Monsoon Intraseasonal Variability
- Harry Hendon and Matthew Wheeler
- Bureau of Meteorology Research Centre
- Melbourne, Australia
- Recent work on objective diagnostics of MJO/ISV
- Seasonally independent EOFs for the MJO
-
2- Issues to be addressed
- Does near-equatorial, eastward propagating,
convectively-coupled, baroclinic MJO occur
year-round? - Is the MJO properly defined in Northern
Hemisphere Summer? - Is there poleward propagating ISV independent of
MJO? - (or, how much of the poleward propagating ISV in
the monsoon can be attributed to the MJO?) - Devise an objective definition for MJO
- Document annual/interannual variation
in propagation/strength/off-euatorial behavior.
3Seasonally independent EOFs for the MJO Require
a standard metric of the MJO that 1) captures
its salient properties 2) is easy to
compute 3) can be computed in real time for
forecast and monitoring application 4)
provides insight into dynamics/physics
4 Salient properties MJO baroclinic
disturbance, coupled with convection,
eastward along the equator as originally
depicted by MJ 1972 But, MJO exhibits strong
annual variation in strength and especially in
off-equatorial structure/propagation
Madden and Julian (1972)
5Ingredients of MJO diagnostic
Use U850, U200, OLR to capture
convective/baroclinic structure, Average about
equator (15N-15S) to focus on zonal propagation,
(i) Removal of seasonal cycle (ii) Removal of
variability associated with index of ENSO (iii)
Removal of mean of previous 120-days
(all these steps can be performed in
real-time) Compute EOFs in order to objectively
extract dominant mode Prior to EOF, normalize
each field by globally-averaged variance.
6U850
U200
OLR
Variance explained by each EOF
7Project raw anomaly data onto the EOFs Spatial
projection onto MJO structure effectively
discriminates to MJO periods
¼ cycle Phase lag
EOFs are temporally in quadrature, thus
describing a zonally propagating disturbance
Coh2
8Variance explained by each EOF
9Whats the nature of ISV variability associated
with the MJO and independent of MJO? Define MJO
part by multiple linear regression onto principal
components OLRMJO(x,y,t) a(x,y)
b(x,y)?RMM1(t) c(x,y)?RMM2(t) N.B. Regression
coefficients are a function of season Then
define an MJO part, and a residual, OLRANOMOLRMJ
OOLRRESIDUAL
10MJO part
MJO
Total
RESIDUAL
Leading pair of EOFs captures all of the eastward
propagating variance
11MJO OLR variance
DJFM
JJAS
12DJF
JJA
total
MJO
residual
13May-Oct
-30
Lag d
Regression OLR (shade), U850 onto MJO PCs
30
Nov-Apr
Behavior along eq is independent of a season
14DJFM Composite
15JJAS Composite
16Dates defined using Darwin zonal wind
(Drosdowsky 1996).
Kerala dates defined using rainfall by IMD (e.g.,
Joseph et al. 1994).
W Pacific
W Pacific
Maritime Continent
Maritime Continent
Africa
Africa
Indian
Indian
17Total
May-Oct North-south propagation
30 in 20 days
2-d spectrum in meridional direction 20S-40N
80-90E Zero-pad in meridional direction to
give equivalent 90S-90N periodic domain i.e.,
Wave 1 has 20000km wavelength
RESIDUAL
MJO
18Total
Nov-Apr
40S-20N data only 80-90E Zero-pad in
meridional direction to give equivalent 90S-90N
periodic domain i.e., wave110000km
half-wavelength
RESIDUAL
MJO
19Looking another way Lagged regressions with OLR
in base region
MJO
DJFM BP80-90E,5-10S
JJAS BP80-90E,5-10N
RESIDUAL
Regressed OLR plotted, based on a 1 standard
deviation anomaly in the predictor
20RESIDUAL JJAS regressed OLR and 850hPa winds Base
Point 80-90E, 5-10N
Primarily northward propagation Some westward
21Conclusions
MJO exists year-round, with constant
near-equatorial structure. MJO accounts for 1/2
poleward propagating ISV in monoon Independent
northward propagation has higher frequency and
shorter meridional and zonal scales Implication
mechanism of equatorial propagation is
independent of poleward propagation But, is
mechanism of poleward propagation the same for
the MJO as for the independent component?
22 US CLIVAR MJO WORKING GROUP
Duane Waliser, co-chair Wanqui Wang Ken
Sperber, co-chair Chidong Zhang Siegfried
Schubert 2-Yr Lifetime Mitch Moncrieff Klaus
Weickmann April 2006-gt Eric Maloney Bin Wang
Leo Donner F. Vitart, S. Woolnough, H.
Hendon, M. Wheeler, W. Higgins, J. Gottschalck,
W. Stern International CLIVAR Support
http//www.usclivar.org/Organization/MJO_WG.html
- Terms of Reference
- Develop a set of metrics to assess MJO simulation
and forecast skill. - Coordinate model simulation and prediction
experiments to better understand and improve MJO - Raise awareness of utility of MJO forecasts in
the context of the seamless suite of predictions. - Coordinate MJO-related activities between
national and international agencies
23.compared to variance of RESIDUAL field
(i.e., variance of whats left when MJO part
is removed from original anomaly field)
DJFM
10 times the contour interval!
JJAS
24(No Transcript)
2515S15N averaged OLR
Annual cycle of MJO Variance
26Kemball-Cook and Wangs composite of the
May-June BISO based on OLR in the box
poleward and eastward along equator
also Lawrence and Webster
27- Diagnostics of MJO/ISV
- Motivation
-
- continued poor representation of MJO/ISV in
forecast/climate models -
- importance for monsoon variability/prediction
- Provide quantifiable/consistent measures
- Elucidate mechanisms/physics
- Compare simulations/forecasts with
observations - Applicable in realtime to make/assess
forecasts
28Space-time spectral analysis
- Quantifies space and time scales of
organized/propagating behavior - Long history for identification wave modes of
tropical intraseasonal variability - Kelvin waves, Rossby-gravity waves, MJO
- Normalized spectra (WK99) emphasized prominent
role of wave modes for organization of tropical
convection - revisit WK99
-
- objective/physical estimation of background
spectrum - (hopefully conclusions dont depend on
definition of background) -
- space-time coherence spectrum OLR/zonal wind
-
29Ratio Spectrum/Background
Symmetric OLR 15N-15S
Only obvious peak is MJO
Wheeler Kiladis 1999
30Background Spectrum
- WK99 estimated background spectrum by ad-hoc
smoothing in wavenumber and frequency - Alternate approach Assume background variability
can be modeled as 1st order auto-regressive
process (Gilman et al. 1963) -
- X(n)pX(n-1) noise
-
- p is the lag-1 autocorrelation, obtain by
inverse Fourier transform of power spectrum - The equivalent auto correlation function is
-
- cor(T)pT
- where T is lag.
- Equivalent power spectrum can be obtained by
inverse transform of the correlation function
(or approximate discrete formula provided by
Gillman et al) -
- E(f)1/(constf2) power spectrum of
red noise -
31OLR 10N-10S
equivalent red spectrum
smth background WK99
32- Quantify redness of background
- For a first order auto regressive process, the
discrete autocorrelation function is -
- cor(T)pT
- Decorrelation time defined as the time for the
correlation to drop to a value of e-1 - Td-(lnp)-1
33Decorrelation time(d) of background red
spectrum 10N-10s
decorrelation time (d)
34Symmetric OLR 10N-10S
Shading is signal percentage of power that
stands above red background
WK99 Shading is ratio of actual spectrum to
background spectrum
35OLR 10N-10S
20 m/s
U850 10N-10S
36Space-time Coherence Spectrum effectively
correlation as function of wavenumber and
frequency
The phase lag
.
where Q is imaginary part (quad) of cross
spectrum and P is real part (co)
37Coherence OLR and U850
Coh2 direct measure of interaction of
convection with large-scale circulation doesnt
rely on definition of background
38OLR/U850
OLR/U150
OLR/U50
39MMF (super parameterization)
CAM (Zhang and McFarlane)
Precip
MMF U850 CAM
40Precip
MMF (NCAR CAM3) Super parameterization
U850