Transcript Slide 1

MJO Simulation Diagnostics
An Activity of the (US) CLIVAR MJO
Working Group
Matthew Wheeler
Centre for Australia Weather and Climate Research
A partnership between Bureau of Meteorology and CSIRO
Melbourne, Australia
MJO Simulation Diagnostics – CMIP3 sub-meeting, CCSR, University of Tokyo – Nov 2008
In this talk I aim to give you some additional understanding of the
Madden-Julian oscillation and the steps being taken by an
international working group to help guide its successful simulation by
global circulation models (GCMs).
In particular, I will discuss a multi-authored paper by the working
group on “MJO Simulation Diagnostics”.
Contents
• Background on the CLIVAR MJO Working Group
• Some background on the MJO
• Motivation for having standardized diagnostics
• The (observed) diagnostics
• Summary
• Brief comparison with models
Background on the CLIVAR MJO Working Group
Formed in June 2006 as the US-CLIVAR MJO Working Group with a lifetime of two years. Now with a more international flavour, and continuing
informally. See http://www.usclivar.org/mjo.php
MJO WG - Terms of Reference
Membership of
1. Develop a set of diagnostics to be used
for assessing MJO simulation fidelity and
forecast skill.
2. Develop and coordinate model simulation
and prediction experiments, in conjunction
with model-data comparisons, which are
designed to better understand the MJO and
improve our model representations and
forecasts of the MJO.
3. Raise awareness of the potential utility of
subseasonal and MJO forecasts in the
context of the seamless suite of predictions.
CAWCR/Bureau of Met
4. Help to coordinate MJO-related activities
between national and international agencies
and associated programmatic activities.
5. Provide guidance to US CLIVAR and
Interagency Group (IAG) on where
additional modeling, analysis or
observational resources are needed.
CAWCR/Bureau of Met
Note: Professor Nakazawa attended our November
2007 workshop.
Some background on the MJO
Approximate 1 month sequence
Features:
• Has a 30 to 80-day period and
slow eastward propagation.
Monsoon “break”
TC formation
• Also called the 40-50 day wave
and the Intraseasonal
Oscillation.
• First described by Madden and
Julian in the early 1970s.
Trade Wind surge
Monsoon westerlies
ACTIVE “BURST” OF INDONESIAN-AUSTRALIAN MONSOON
Westerly Wind Burst
• Is the strongest mode of
intraseasonal variability on
earth.
• Generates many of the bursts
and breaks of the monsoons.
• But also has far-reaching
impact on other weather and
climate phenomena.
Original schematic from
Madden and Julian (1972)
• Seen in upper-level zonal winds around
the globe.
• Strongest in Indo-Pacific domain where it
is convectively-coupled.
• Planetary scale: Zonal wavenumbers 1-3.
• Baroclinic wind structure.
Each panel separated
by about 6 days.
• Episodic and not always
present.
• Convection has multiscale structure.
• Strong seasonal dependence to off-equatorial behaviour
(but still “MJO” in all seasons)
Each phase is separated, on average, by about 6 days
Motivation for diagnostics
• Despite the important role for the MJO, GCMs still exhibit shortcomings in
representing it, as has been documented by:
Slingo et al. (1996)
Waliser et al. (2003)
Sperber et al. (2005)
Zhang et al. (2006)
Lin et al. (2006)
AMIP 15 models
CLIVAR 10 models
6 coupled/uncoupled models
8 coupled/uncoupled models
IPCC/AR4 14 coupled models
• Due to the use of different diagnostics in these studies, however, progress
has been difficult to track.
• Thus there is a need for a standardized set of clearly-defined diagnostics
for use by modelling groups.
 These diagnostics should be able to succinctly capture the
essential features of the MJO and its dynamics as just described.
N.b. Due to interannual variability and the episodic nature of the MJO, it
takes a simulation of at least 5 years to determine its simulation fidelity.
1st paper, in press
2nd paper, draft
The journal papers provide only a sample of what is available on-line at
http://climate.snu.ac.kr/mjo_diagnostics/index.htm
Code is also available from this site
Mean state diagnostics
Nov-Apr
The ability of a model to simulate
the MJO is intimately related to
its ability to simulate the mean
climate.
Thus mean states of some
relevant variables are also
included.
contours of SST
May-Oct
contours of SST
Level 1 diagnostics
Maps of total intraseasonal (20-100 day) variance and variance fraction provide a
lowest-level benchmark for GCMs.
Of importance is the shift in variance north and south of the equator with season
and the relative minimum over the Maritime Continent.
Nov-Apr
May-Oct
Variance in the 20-100 day band for a) CMAP precipitation and b) NCEP1 850 hPa zonal wind (contours) for November-April
(left) and May-October (right) seasons. The percent variance accounted for by the 20-100 day band band is shown in color.
Precipitation variance contours are plotted every 6 mm 2 day-2, starting at 3 mm2 day-2. Zonal wind variance contours are
plotted every 3 m2 s-2, starting at 6 m2 s-2.
The fundamental eastward propagating nature is best isolated with laglongitude correlation analyses.
“Predictor” is 10S-5N, 75-100E averaged precipitation.
Nov-Apr
5 ms-1
Contour
interval = 0.1
Qualitatively similar results are achieved in boreal summer!
However, boreal summer intraseasonal variability is also characterised by
distinct northward propagation.
As before, “predictor” is 10S-5N, 75-100E averaged precipitation.
May-Oct
Contour
interval = 0.1
Some southward propagation into the SH is also apparent.
Level 2 diagnostics
Nov-Apr
Wavenumber-frequency spectra show the
dominance of eastward propagation.
May-Oct
Data are averaged
10S-10N before
computing spectra.
Applied to 180-day
segments from each
year, then averaged for
all years.
WESTWARD
EASTWARD
Wavenumber-frequency cross-spectra quantify the coherence and phase
between different variables.
Using OLR and 850-hPa zonal wind is very useful for extracting the
coherent modes of convectively-coupled behaviour in that exist, without
the need for estimating a background spectrum.
All seasons
Applied to 256-day
overlapping segments.
Spectra computed for
individual
symmetric/antisymmetric
latitudes first, then
averaged 0-15.
Upward-pointing vector is a
phase of 0, to the right is
OLR leading u850 by 90.
An aside: New results from Hendon and Wheeler (J. Atmos. Sci.; 2008)
Top figures: Contours for OLR power, shading for “signal strength” (like normalized power)
Bottom figures: Cross-spectra between OLR and u850 (as on previous slide).
Multivariate EOF analysis: Useful for extracting convetively-coupled structure,
and for generating a MJO phase index.
All seasons
15S-15N
averaged data.
These EOFs are
virtually
independent of
season!
Power spectrum of projection coefficients
obtained by projecting unfiltered data on the
EOFs.
This analysis is the same as applied by Wheeler and Hendon (2004), except used 20-100d filtered data as input.
For compositing, phases may be defined from the leading pair
of principal component (PC) time series.
Note the ‘smoothness’ of the phasespace trajectory because of the use of
filtered data.
“Weak MJO” defined to occur when
PC12 + PC22 < 1.0.
Nov-Apr
Despite using an all-season
EOF index, the seasonality of
the MJO is still retained in
season-specific composites.
The weakening of the
precipitation signal over the
islands of the Maritime
Continent is shown.
May-Oct
Northwest-southeast tilting and
resulting northward
propagation of the convective
signal is shown.
Nov-Apr
Interaction with the ocean is
also thought to provide at least
a modifying effect on the MJO.
Having the correct phase
relationship of SST anomalies
with the convection, however,
is crucial for getting the impact
of the ocean coupling correct.
May-Oct
Summary of paper 1
• The aim of the paper is to recommend a set of diagnostics that describe
the essential features of the MJO, its dynamics, and aspects of the mean
state important for its existence.
• Further diagnostics, and computations with different observed datasets
are available from the web-site. (as well as code and some data)
• Level-1 diagnostics are meant to provide an initial indication of a model’s
ability to reproduce the spatial extent and strength of intraseasonal
variability.
• Level-2 diagnostics provide a more comprehensive assessment of the
propagating, convectively-coupled nature of the MJO, and its seasonal
variation.
But how well do the models do? (i.e. paper 2)
Observed MJO
wavenumber
Wavenumber-frequency spectra of precipitation (shaded) and u850 (contours)
WEST
EAST
frequency
Wavenumber-frequency spectra of 10N-10S averaged precipitation (shaded) and 850hPa zonal wind (contoured). Individual
November-April spectra were calculated for each year, and then averaged over all years of data. Only the climatological
seasonal cycle and time mean for each November-April segment were removed before calculation of the spectra. Units for the
precipitation (zonal wind) spectrum are mm2 day-2 (m2 s-2) per frequency interval per wavenumber interval. The bandwidth is
(180 d)-1. CLIVAR MJO Working Group (2008)
Scatter plot of east/west ratio of power based on the data in previous figure.
The east/west ratio is calculated by dividing the sum of eastward propagating
power by westward propagating counterpart within wavenumber 1-6 (1-3 for
zonal wind), period 30-80 days.
Multivariate EOFs of 15S-15N averaged data.
Power spectra of the unfiltered
PCs derived by projecting
unfiltered data onto the CEOFs.
The End
Thankyou for being wonderful hosts!
[email protected]