Title: MJO Simulation Diagnostics Application of Diagnostics to Models Part I
1MJO Simulation Diagnostics Application of
Diagnostics to ModelsPart I
Daehyun Kim, Duane E. Waliser, Kenneth R.
Sperber, Eric E. Maloney and In-Sik Kang on
behalf of US CLIVAR MJO Working
Group1 1http//www.usclivar.org/Organization/MJO_W
G.html
NCEP Wanqiu Wang PCMDI Kenneth R. Sperber
GFDL Bill Stern CSU Marat Khairoutdinov, David
A. Randall NASA/GSFC Myong-In Lee, Max J.
Suarez CAM3.5 Richard B. Neale SNU In-Sik
Kang
Acknowledge the following modeling groups
2The Madden-Julian Oscillation (MJO)
MJO Eastward propagating planetary scale wave
with 30-60 day period
Simulation of MJO Intercomparison studies
- Intercomparison studies using several
contemporary GCMs - - 1996 Slingo et al. (AMIP 15 models)
- - 2003 Waliser et al. (CLIVAR 10 models)
- - 2005 Sperber et al. (6
coupled/uncoupled models) - - 2006 Zhang et al. (8
coupled/uncoupled models) - - 2006 Lin et al. (IPCC/AR4 14 coupled models)
-
3The Madden-Julian Oscillation (MJO)
4The Madden-Julian Oscillation (MJO)
10oS-10oN Time-longitude Cross Section of
Precipitation during 1997
CMAP
CLIVAR monsoon AGCM intercomparison workshop
(Kang et al. 2001 Waliser et al. 2003)
Most of the models hardly simulate eastward
propagation
5The Madden-Julian Oscillation (MJO)
Coupled
Coupled
Coupled
Coupled
6The Madden-Julian Oscillation (MJO)
7The Madden-Julian Oscillation (MJO)
- MJO is still a harsh test for GCMs
- Tracking progress is difficult
8Application to climate models
Climate models
flux adjustment for heat and fresh water
9Application to climate models
How well the current climate models simulate MJO?
- Mean state
- Sub-Seasonal Variance
- Space/Time scale and propagation
- Spatial patterns
Boreal Winter time (Nov-Apr)
10Application to climate models
How well the current climate models simulate MJO?
- Mean state
- - Important variables for MJO simulation
- Variance
- Space/Time scale and propagation
- Spatial patterns
11Mean State Diagnostics
Unit m/s
Mean State U850
12Mean State Diagnostics
Unit mm/day
Mean State PRCP Bias
13Mean State Diagnostics Summary
better
Mean State Pattern Corr. vs. RMSE
14Application to climate models
How well the current climate models simulate MJO?
- Mean state
- Sub-Seasonal Variance
- - Magnitude and distribution
- Space/Time scale and propagation
- Spatial patterns
15Variance Map Diagnostics
Variance of 20-100 day filtered precipitation
(Nov-Apr)
Unit mm2/day2 muliplied by 100 Contour 60,
80
Variance Map PRCP
16Variance Map Diagnostics Summary
Magnitude and Pattern PRCP vs. U850
OBS
OBS
0-360E, 20S-20N
0-360E, 20S-20N
17PBL moisture convergence PRCP
Correlation between 925hPa moisture convergence
and PRCP
moisture convergence and precipitation are
20-100 day filtered only values larger than 0.2
are drawn
18Application to climate models
How well the current climate models simulate MJO?
- Mean state
- Sub-SeasonalVariance
- Space/Time scale and propagation
- - Distribution of power in wavenumber-frequency
space - Spatial patterns
19Frequency-wave spectra diagnostics
Equatorial space-time power spectrum (Nov-Apr)
Wavenumber
Unit mm2/day2 multiplied by 100
Wavenumber
Wavenumber
Frequency-Wave PRCP
20Frequency-wave spectra diagnostics
Equatorial (10N-10S) space-time power spectra
Wavenumber
Unit m2/s2 contour 6, 9, 12
Wavenumber
Wavenumber
Frequency-Wave U850
21Frequency-wave spectra diagnostics
East/West Ratio
Frequency-Wave PRCP
Wavnumber 1-3
ERA40
Frequency-Wave U850
GPCP
22Application to climate models
How well the current climate models simulate MJO?
- Mean state
- Sub-SeasonalVariance
- Space/Time scale and propagation
- Spatial patterns
- - Pattern of MJO modes
23EOF Diagnostics OLR vs. PRCP
U850
OBS
U850
ECHAM4/OPYC
U850
SPCAM
24EOF Diagnostics Summary
Pattern correlation of 1st EOF mode
OBS
25MJO Simulation Diagnostics - Web site
MJO Simulation Diagnostics http//climate.snu.ac.
kr/mjo_diagnostics/index.htm
General Strategy Description
Calculation codes and example data - Needs
feedback
26Summary
- 1. Objectively developed standardized diagnostics
for MJO simulation are applied to 3 coupled and 3
uncoupled climate model simulations. - 2. Applied diagnostics reasonably captured models
characteristics related with MJO simulation. - The biases in mean state is not consistent
between variables and models - Coupled models have stronger sub-seasonal
variability than uncoupled models - Spectral power of precipitation is generally
concentrated on low-frequency part - When models MJO is represented using dominant
modes, one variable is not enough and
precipitation is a tough one - 3. Overall comparisons reveal that ECHAM4/OPYC
and SPCAM have relatively better skill among the
models. ECHAM4/OPYC produces very reasonable mean
state with flux adjustment process. Convection is
represented in more explicit manner in SPCAM
(superparameterization).
27Summary
- 4. MJO signal in 850hPa zonal wind is generally
better than that of precipitation in terms of i)
variance ii) peaks in spectra and iii) eastward
propagation - 5. Diabatic heating (rainfall) is more difficult
variable to simulate than large scale circulation
field although heating and circulation are
closely linked together. It will be tracked from
this study what change or development can
overcome this paradox.
28Extended Study with the SNUGCM
- High-resolution experiments (up to 20 km)
- Dependence on Cumulus Paramterization
- - Trigger (Tokioka modification)
- - SAS vs McRAS parameterization
29High resolution modeling
- 200hPa velocity potential (1999)
The FIRST Ensemble case
OBS
20km
300km
SNUGCM
30High resolution modeling
- 200hPa velocity potential (1999)
ENSEMBLES in 20km High resolution !
ENS4
ENS5
ENS6
ENS2
ENS3
SNUGCM
31Dependence on Cumulus Parameterization
GPCP
Variance of 20-100 day filtered
precipitation SNUGCM All season 1997-2004
Control (SAS)
Convection is constrained (Tokioka 1988)
McRAS
32Variance Map Diagnostics
20-100 day filtered velocity potential (10S-10N)
2001
Control (SAS)
McRAS
NCEP Re
Convection is constrained
X106 m2s-1
33Coupled model simulation SST climatology
Strong constraint ? colder SST
34Thank you for your attention!Any questions?
35High resolution modeling
20 Km
100 Km
300 Km
SNUGCM
36Back up figures
37Mean State Diagnostics
Unit oC
Mean State SST
38Mean State Diagnostics
Unit oC
Mean State SST Bias
39Mean State Diagnostics
Unit m/s
Mean State U850 Bias
40Mean State Diagnostics
Unit mm/day
Mean State PRCP
41Mean State Diagnostics
Unit mm/day
Mean State PRCP Bias
42Mean State Diagnostics
Unit m/s
Mean State Wind shear (200-850)
43Variance Map Diagnostics
Variance of 20-100 day filtered precipitation
(Nov-Apr)
Variance Map PRCP Bias
44Frequency-wave spectra diagnostics
Equatorial space-time power spectrum (Nov-Apr)
Wavenumber
Wavenumber
Unit m2/s2 contour 60, 80
Wavenumber
Frequency-Wave U200
45Frequency-wave spectra diagnostics
Wavnumber 1-3
Frequency-Wave U850
46Lag - Correlation Diagnostics
U850 Lag-correlation diagnostics (Nov-Apr)
Shading 95 sig. Line 5m/s
47Frequency-wave spectra diagnostics
Equatorial space-time power spectrum (Nov-Apr)
Wavenumber
Unit mm2/day2 multiplied by 100
Wavenumber
Wavenumber
Frequency-Wave PRCP
48Frequency-wave spectra diagnostics
Equatorial space-time power spectrum (Nov-Apr)
Wavenumber
Unit mm2/day2 multiplied by 100
Wavenumber
Wavenumber
49Frequency-wave spectra diagnostics
Wavnumber 1-6
Frequency-Wave PRCP
50Lag - Correlation Diagnostics
PRCP Lag-correlation diagnostics (Nov-Apr)
Shading 95 sig. Line 5m/s
51EOF Diagnostics Percentage variance
Percentage variance of 1st and 2nd EOF mode
2nd mode is separated from 3rd mode by
Norths criteria (DOF number of day / 50)
52(No Transcript)
53EOF Diagnostics
OBS (ERA40)
ECHAM4/OPYC
SPCAM
EOF U850
54EOF Diagnostics
OBS (AVHRR)
ECHAM4/OPYC
SPCAM
EOF OLR
55EOF Diagnostics
OBS (GPCP)
ECHAM4/OPYC
SPCAM
EOF PRCP
56PBL moisture convergence PRCP
Correlation between 925hPa moisture convergence
and PRCP
Lag
Lag
Lag
20-100day filtered, 5S-5N averaged
57MJO Simulation Diagnostics
MJO Life cycle composite Specific humidity
NCEP1
ECHAM4/OPYC
MJO Life cycle composite Zonal wind