Single Column Model representation of RICO shallow cumulus convection A.Pier Siebesma and Louise Nuijens, KNMI, De Bilt The Netherlands And all the participants.

Single Column Model representation

of 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

Main Questions

Are the single column model versions of GCM’s, ‘LAM’s 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

The game to be played

tifdt

d

dt

d

dt

d

lsphystot

0

vuqwheret ,,,)0( 1. Start with the observed mean state:

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

3. Evaluate the steady state with observations in all its aspects

with observations (both real and pseudo-obs (LES) ), i.e.

obsvs )(

Two Flavours of the game

T

timeLSphys

dtdt

d

dt

dT

0

)0()(

1. Use the mean LS-forcing of the suppressed period:

2. Use directly the the time-varying LS forcing for the whole suppressed period.

i.e. the composite case.

T

LSphys

dtdt

d

dt

dT

0

)0()(

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

Model PBL Scheme Convection Cloud

CAM3/GB TKE (bretherton/grenier) MF (Hack) Prog l,

UKMOK-profile/expl entr. /moist(?)

MF (Gregory-Rowntree)Mb=0.03w*

Stat/RH_cr (Smith)

JMAK-profile/expl entr/moist.

MF (Arakawa-Schubert) Stat/RH_cr (Smith)

HIRLAM/RACMO

TKE/moist MF(Tiedtke89)New entr/detr, M=a w* closure

Stat, diagns from K and MF

GFDLK-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/UCLA3rd order pdf basedLarson/Golaz (2005)

3rd order pdf basedLarson/Golaz (2005)

3rd order pdf basedLarson/Golaz (2005)

ADHOCAssumed pdfhigh order MF

Assumed pdfhigh order MF

Assumed pdfhigh order MF

AROME TKE-moist MF (pbl/cu-updraft) Stat. diagnostic

ECHAMTKE-moist Tiedtke(89) Entr/detr

(Nordeng)Stat Tompkins 2002)

ARPEGETKE-moist

MFStat ,cloud coverL=prognostic

ECMWF K-profile (moist) MF (pbl/cu-updraft) Stat. diagnostic

Submitted versions

Each model asked to submit:

• Operational resolution / prescribed resolution

• Operational physics / Modified physics

• Composite constant forcing / variable forcing

Initial State (identical to LES case)

Profiles after 24 hrs

Composite Case (High resolution)

80 levels ~ 100m resolution in cloud layer

Different Building Blocks

Moist Convection

entr/detr

M_b , w_u

Extended in bl

Cloud scheme:

stat

progn

Precip

precip?

microphysics

precip

PBL:K-profile

TKE

Higher order

ac, q

, q

acEstimating: ac,qlac,ql

on/off

• need increasingly more information from eachother

• demands more coherence between the schemes

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)

LES

Cloud fraction

In general too high

Time series

Composite Case (High resolution)

80 levels ~ 100m resolution in cloud layer

Some models behave remarkably well

• These models worked actively on shallow cumulus

• It seems that there are 3 crucial ingredients:

1. Good estimate of cloud base mass flux : M~ac w*

2. Good estimate of entrainment and detrainment

3. Good estimate of the variance of qt and l in the cloud layer in order to have a good estimate of cloud cover and liquid water.

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.

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.

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)

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?

Thank you

s

st qqtQ

Cloud cover

Bechtold and Cuijpers JAS 1995

Bechtold and Siebesma JAS 1999Wood (2002)

Statistical Cloud schemes

Convective and turbulent transport