The MJO problem in GCMs: What are the missing physics?

1
The MJO problem in GCMsWhat are the missing
physics?

• Jia-Lin Lin1, Brian E. Mapes2, George N.
  Kiladis3, Klaus M. Weickmann1, Minghua Zhang5,
  Kenneth R. Sperber4, Matthew Newman1, Wuyin Lin5,
  Matthew Wheeler6, Siegfried D. Schubert7, Anthony
  Del Genio8, Leo J. Donner9, Seita Emori10,
  Jean-Francois Gueremy11, Frederic Hourdin12,
  Philip J. Rasch13, Erich Roeckner14, John F.
  Scinocca15
• 1NOAA-CIRES Climate Diagnostics Center, Boulder,
  CO, 2RSMAS, University of Miami, Miami, FL, 3NOAA
  Aeronomy Laboratory, Boulder, CO, 4PCMDI,
  Lawrence Livermore National Laboratory,
  Livermore, CA, 5State University of New York,
  Stony Brook, NY, 6BMRC, Melbourne, Australia,
  7NASA GSFC Global Modeling and Assimulation
  Office, Greenbelt, MD, 8NASA Goddard Institute
  for Space Studies, New York, NY, 9NOAA
  Geophysical Fluid Dynamics Laboratory, Princeton,
  NJ, 10National Institute for Environmental
  Studies, Ibaraki, Japan, 11Meteo-France CNRM,
  Paris, France, 12Laboratoire de Meteorologie
  Dynamique, Universite de Paris, Paris, France,
  13National Center for Atmospheric Research,
  Boulder, CO, 14Max Planck Institute for
  Meteorology, Hamburg, Germany, 15Canadian Centre
  for Climate Modeling Analysis, Victoria, Canada

2
1. Introduction The MJO and its teleconnections
3
2. Motivation The MJO problem A
longstanding, major tropical bias in GCMs

• Pioneering studies in 1980s (Hayashi and Golder
  1986, 1988, Hayashi and Sumi 1986, Lau et al.
  1988)
• Eastward Kelvin-Rossby or Kelvin waves but
  with too fast phase speeds (10-18 m/s)
• AMIP models in early 1990s (Slingo et al. 1996)
• Simulated signals are generally too weak and
  too fast

• Models in late 1990s (Schubert et al. 2002,
  Waliser et al. 2003)
• More models are getting something in the way of
  an MJO.
• But when a model does exhibit a relatively good
  MJO, we can at best only give vague or plausible
  explanations for its relative success. This
  inhibits the extension of individual model
  successes to other more MJO-challenged models.
• Moreover, it is often the case that stated
  successes do not stand up to a great deal of
  detailed scrutiny.
• Latest models participating in the IPCC Fourth
  Assessment Report (AR4) to be released in 2007
  (Lin et al. 2005)

4
Tropical intraseasonal variability in 14 IPCC AR4
climate models. Part I Convective signals (Lin
et al. 2005)
Participating models GFDL-CM2.0, GFDL-CM2.1,
NCAR-CCSM3, NCAR-PCM, GISS-AOM, GISS-ER,
MIROC3.2-hires, MIROC3.2-medres,
MRI-CGCM2.3.2, CCC-CGCM3.1-T47, ECHAM/MPI-OM,
IPSL-CM4, CNRM-CM3, CNRM-CM3-AMIP A new
generation of climate models Before conducting
the extended simulations for IPCC AR4, many of
the modeling centers applied an overhaul to their
physical schemes to incorporate the
state-of-the-art research results.
5
Questions

• (1) How well do the IPCC AR4 models simulate the
  convectively coupled equatorial waves, especially
  the MJO?
• (2)  Is there any systematic bias that is
  important for the MJO simulation?

6
Data

• Each model 8 years of daily precipitation from
  the Climate of the 20th Century (20C3M)
  experiment
• Observation 8 years of daily precipitation from
  GPI and GPCP 1DD

7
Method

• Identification of the dominant intraseasonal
  modes
• Space-time spectral analysis (Wheeler and
  Kiladis 1999)
• Raw and Raw/Background, symmetric and
  antisymmetric
• Isolating the MJO mode
• Definition of the MJO eastward
  wavenumbers 1-6, 30-70 day mode
• The MJO is also compared with its
  westward counterpart
• 
  westward wavenumbers 1-6, 30-70 day mode

8
Climatological precipitation along the equatorial
belt (15N-15S) Reasonably simulated over warm
pool
9
Climatological precipitation on the equator
(5N-5S) Some models have double-ITCZ problem
10
Total intraseasonal (2-128 day) variance
(15N-15S) Variances in most models are
smaller than in observations
11
Total intraseasonal (2-128 day) variance (5N-5S)
Variances in most models are smaller
than in observations
12
Raw space-time spectra (15N-15S
symmetric)Variances in most models are too weak
and too red
Obs
13
Raw/background spectra (15N-15S symmetric)Many
models have Kelvin waves, some have ER, WIG
waves. But their phase speeds are too fast too
large equivalent depth
Obs
Dominant modes MJO, Kelvin, ER, WIG Dispersion
curves correspond to equivalent depth 8, 12, 25,
50, 90m. Larger depth faster phase speed. All
modes 25 m.
14
Raw/background spectra (15N-15S
antisymmetric)Many models have MRG-EIG waves.
But their phase speeds are too fast too large
equivalent depth
Obs
Dominant modes MRG, EIG Dispersion curves
correspond to equivalent depth 8, 12, 25, 50,
90m. Both modes 25m.
15
An interesting result

• Within a given model, equivalent depth is the
  same for all different equatorial waves. This is
  indicative of similar physical processes linking
  the convection and large-scale disturbances
  within each model.

16
Variance of the MJO mode (eastward wavenumbers
1-6, 30-70 day)MJO Variance in most models is
smaller than in observations, but approaches the
observed value in two models (MPI,CNRM)
17
Variance of the westward counterpart of the
MJO(westward wavenumbers 1-6, 30-70 day)In
many models, the eastward MJO variance is
significantly larger than its westward counterpart
18
Propagation of the 30-70 day precipitation
anomaly Models with eastward MJO variance much
stronger than its westward counterpart show clear
eastward propagation
Obs
The three thick lines correspond to phase speed
of 3, 7, and 15 m/s.
19
Raw spectra of eastward wavenumbers 1-6 at
0N85EThe MJO variance in most models does not
come from a pronounced spectral peak, but from a
too red spectrum. The only model with a
prominent spectral peak is CNRM.
20
Normalized spectra of eastward wavenumbers 1-6 at
0N85EHighlight the models with small variance
21
Theoretical first-order linear Markov process
A too red spectrum suggests a too strong
persistence
Spectrum
Auto-correlation
22
Auto-correlation of precipitation at 0N85EMost
models do have a too strong persistence, which is
consistent with their too red spectra
23
Summary of IPCC AR4 model evaluationTwo
encouraging results

1. Many of the models have signals of convectively
   coupled waves, with Kelvin and MRG-EIG waves
   especially prominent.
2. The eastward MJO precipitation variance in many
   models is significantly larger than its westward
   counterpart, and even approaches the observed
   value in two models.

24
Summary of IPCC AR4 model evaluation Two common
biases

1. The MJO variances in many models do not come from
   a pronounced spectral peak, but from part of a
   too red spectrum (i.e., too red background
   noise ), which in turn are associated with a too
   strong persistence of precipitation.
2. The equivalent depths for all equatorial waves
   are too large, which is indicative of a too
   strong effective static stability and thus too
   weak wave-heating feedback.

25
Ongoing works

• 1. Dynamical signals and 3D wave structure
• Analyzing the daily 3D upper air data
• Budget analysis and feedback analysis
• Calculating the heat and moisture budgets
  for all models and analyzing the wave-heating
  feedback
• Future works Apply these diagnostics to NCEP
  GFS/CFS

26
3. Hypothesis

• Because the MJO problem is a common problem in
  many GCMs, our hypothesis is
• The MJO problem is caused by some missing physics
  in current GCMs.
• (1) Missing physics associated with too red
  background noise
• Missing physics associated with too weak
  wave-heating feedback

27
(1) Missing physics associated with too red
background noise (too strong persistence of
precip)Why is the persistence of precip weak in
observation? Self-suppression
processes in tropical deep convection
28
Convective downdrafts and Mesoscale downdrafts
Convective updrafts
Mesoscale updrafts
Mesoscale downdrafts
Convective downdrafts
Zipser (1977), modified by Houze (1993)
29
Convective downdrafts and mesoscale downdrafts
significantly affect the post-convection sounding
Onion sounding (similar to trade wind region
in EP) Lower troposphere drier, warmer Boundary
layer drier, cooler
Pre-convection
Post-convection
Zipser (1977)
30
Self-suppression processes in tropical deep
convection are missing in many GCMs
? Missing physics II Mesoscale downdrafts
Missing physics I ? Convective downdrafts
Missing physics III Control of deep convection
by lower troposphere moisture
31
Deep convection schemes in the 14 IPCC AR4 models
32
Control of deep convection by lower troposphere
moisture is also missing in many models

• All schemes are mass flux scheme using an
  ensemble of
• Entraining plumes
• (e.g. Arakawa and Schubert 1974) or
• (2) Buoyancy sorting parcels
• (e.g. Emanuel 1991)
• A common problem in many schemes including
  undiluted or weakly diluted members, and
  therefore are not sensitive to lower troposphere
  moisture.
• Solution in a couple of schemes
• Include only strongly diluted members (e.g.
  Tokioka et al. 1988, Tiedke 1989)
• Add explicit RH trigger (e.g. Emori et al. 2001)

33
(2) Missing physics associated with wave-heating
feedback
Vertical heating profile
Missing physics IV Stratiform heating profile
Missing physics V ? Shallow convective momentum
transport
Column-integrated diabatic heating has six major
components (Mean state and higher-frequency modes
affect the MJO through the nonlinear terms)
34
Missing physics IV Stratiform precipitation and
stratiform heating profile
Stratiform precipitation has 3 characteristics 1.
Contributes significantly to total precipitation
(gt40) 2. Lags convective precipitation by
several hours 3. Associated with upper-level
heating and low-level cooling, making total
heating profile top-heavy.
Heating
Divergence
From Houze (1997)
35
Doppler Radar Climatology Project(Mapes and Lin
2005)

• 7 experiments -- covering almost all
  precipitation centers
• Simultaneous measurements of convection and
  circulation
• for a region with the size of a GCM grid
  (200200 km)
• hourly datasets
• more than 20 days long for each experiment

36
Composite life cycle of deep convection for one
experiment (EPIC) precip and divergence
Stratiform
Convective
Convective
Stratiform
Stratiform precip provides more than 50 of total
precip
37
Corresponding heating profile
Top-heavy heating
38
Composite lifecycle of deep convection for all
experiments Stratiform precip and heating are
important for all precipitation centers
From Mapes and Lin (2005)
39
Stratiform precipitation and heating profile are
missing in almost all GCMs
MJO anomaly Observation 6 GCMs
Observation - Model
From Lin et al. (2004a)
40
Theoretical results

• Top-heavy heating profile tends to amplify all
  intraseasonal modes (e.g. Cho and Pendlebury
  1997).
• Time-mean top-heavy heating profile can make the
  MJO highly viscous, and thus enhance wave-heating
  feedback in the MJO (Lin et al. 2004b).
• Time-lag between stratiform precipitation and
  large-scale forcing may damp short waves, and
  favor long waves, which may enhance the MJO
  (Emanuel 1993, Cho et al. 1994).

41
Missing physics V Shallow convective momentum
transport
Mechanical damping rate in observed MJO estimated
from 15 years of NCEP/NCAR and ECMWF reanalysis
data (Lin et al. 2004b)
Strong mechanical damping above PBL
2 day
10 day
940
Over the warm pool region, PBL top is generally
below 940 mb.
42
Theoretical model results A thick frictional
layer tends to amplify the MJO (Wang and Li 1994)
43
Summary Missing physics in GCMs which are
likely important for the MJO

• Convective downdrafts (saturated and unsaturated)
• 2. Mesoscale updrafts/downdrafts
• Control of deep convection by lower troposphere
  moisture
• Shallow convective momentum transport
• Others? (e.g. other mechanical damping,
  gustiness, radiation)

44
It would be interesting to install these missing
physics into GCMs and test their effects on the
MJO simulation