Single Column Model representation of RICO shallow cumulus convection

1
Single Column Model representationof RICO
shallow cumulus convection

• A.Pier Siebesma and Louise Nuijens,
• KNMI, De Bilt
• The Netherlands
• And all the participants to the case

Many thanks to All the participants
2
Main Questions

• Are the single column model versions of GCMs,
  LAMs and mesoscale models capable of
• representing realistic mean thermodynamic state
  when subjected to the best guess of the applied
  large scale forcings.
• Reproducing realistic precipitation
  characteristics

3
The game to be played

1. Start with the observed mean state

2. Let the initial state evolve until it
reaches steady state

• Evaluate the steady state with observations in
  all its aspects
• with observations (both real and pseudo-obs
  (LES) ), i.e.

4
Two Flavours of the game

• Use the mean LS-forcing of the suppressed period

i.e. the composite case.
2. Use directly the the time-varying LS forcing
for the whole suppressed period.
5
Model Type Participant Institute
CAM3/GB GCM (Climate) C-L Lappen CSU (US)
UKMO GCM (NWP/Climate) B. Devendish UK Metoffice (UK)
JMA GCM (NWP/Climate) H. Kitagawa JMA (Japan)
HIRLAM/RACMO LAM (NWP/Climate) W. De Rooy KNMI (Netherlands)
GFDL GCM (Climate) C. Golaz GFDL (US)
RACMO/TKE LAM (Climate S. De Roode KNMI (Netherlands)
COSMO NWP/regional/mesoscale J. Helmert DWD (Germany)
LMD GCM Climate) Levefbre LMD (France)
LaRC/UCLA LAM (Mesoscale) Anning Cheng NASA-LaRC (US)
ADHOC C-L Lappen CSU (US)
AROME LAM (Mesoscale) S. Malardel Meteo-France (France)
ECHAM GCM (Climate) R. Posselt ETH (Switzerland)
ARPEGE GCM (Climate) P. Marquet Meteo-France (France
ECMWF GCM (NWP) R. Neggers ECMWF (UK
6
Model PBL Scheme Convection Cloud
CAM3/GB TKE (bretherton/grenier) MF (Hack) Prog l,
UKMO K-profile/expl entr. /moist(?) MF (Gregory-Rowntree) Mb0.03w Stat/RH_cr (Smith)
JMA K-profile/expl entr/moist. MF (Arakawa-Schubert) Stat/RH_cr (Smith)
HIRLAM/ RACMO TKE/moist MF(Tiedtke89) New entr/detr, Ma w closure Stat, diagn ss from K and MF
GFDL K-profile/expl entr/moist(?) MF (Rasch) l,c prognostic
RACMO/TKE TKE moist MF (Tiedtke(89) l,c,prognostic
LMD Ri-number MF (Emanuel) Stat
LaRC/UCLA 3rd order pdf based Larson/Golaz (2005) 3rd order pdf based Larson/Golaz (2005) 3rd order pdf based Larson/Golaz (2005)
ADHOC Assumed pdf high order MF Assumed pdf high order MF Assumed pdf high order MF
AROME TKE-moist MF (pbl/cu-updraft) Stat. diagnostic
ECHAM TKE-moist Tiedtke(89) Entr/detr (Nordeng) Stat Tompkins 2002)
ARPEGE TKE-moist MF Stat ,cloud cover Lprognostic
ECMWF K-profile (moist) MF (pbl/cu-updraft) Stat. diagnostic
7
Submitted versions

• Each model asked to submit
• Operational resolution / prescribed resolution
• Operational physics / Modified physics
• Composite constant forcing / variable forcing

8
Initial State (identical to LES case)
9
Profiles after 24 hrs
Composite Case (High resolution) 80 levels 100m
resolution in cloud layer
10
Different Building Blocks

• need increasingly more information from
  eachother
• demands more coherence between the schemes

ac
microphysics
Moist Convection
Cloud scheme
stat progn
ac,ql
Estimating ac,ql
sq, sq
entr/detr M_b , w_u Extended in bl
precip
precip?
Precip
on/off
11
(No Transcript)
12
(No Transcript)
13
At least in general much better than with the
previous Shallow cumulus case based on
ARM (profiles after 10 hours
Lenderink et al. QJRMS 128 (2002)
14
Cloud fraction
LES
In general too high
15
(No Transcript)
16
Time series
Composite Case (High resolution) 80 levels 100m
resolution in cloud layer
17
(No Transcript)
18
(No Transcript)
19
(No Transcript)
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
Some models behave remarkably well

• These models worked actively on shallow cumulus
• It seems that there are 3 crucial ingredients
• Good estimate of cloud base mass flux Mac w
• Good estimate of entrainment and detrainment
• Good estimate of the variance of qt and ql in the
  cloud layer in order to have a good estimate of
  cloud cover and liquid water.

26
Conclusions

• Mean state (slightly) better than for the ARM
  case
• Most models are unaccaptable noisy (mainly due to
  switching between different modes/schemes.
• Probably due to unwanted interactions between the
  various schemes
• No agreement on precipitation evaporation
• Performance amazingly poor for such a simple case
  for which we know what it takes to have realistic
  and stable response.
• Difficult to draw conclusions on the microphysics
  in view of the intermittant behaviour of the
  turbulent and convective fluxes.

27
(No Transcript)
28
We should clear up the obvious deficiencies

• Check LS Forcings should we ask for it as
  required output?
• u,v profiles RACMO-TKE, ECMWF, UCLA-LaRC,
  ECHAM
• Ask for timeseries for u,v,q,T near surface to
  check surface fluxes and cloud base height
  off-line.

29
Required observational data

• Liquid water path (or even better profiles)
• cloud cover profiles (should be possible)
• .precipitation evaporation efficiency.
• Cloud base mass flux.
• Incloud properties., entrainment, detrainment
  mass flux (Hermann??)
• Variance of qt and theta (for cloud scheme
  purposes)

30
Further Points

• Proceed with the long run??
• Get the the RICO-sondes into the ECMWF/NCEP
  analysis in order to get better forcings?
• Should we do 3d-GCM RICO?

31
Thank you
32
Statistical Cloud schemes
Cloud cover
Bechtold and Cuijpers JAS 1995 Bechtold and
Siebesma JAS 1999
Wood (2002)
33
Convective and turbulent transport
34
(No Transcript)