LES modeling of precipitation in Boundary Layer Clouds and parameterisation for

General Circulation Model

O. Geoffroy J.L. Brenguier

CNRM/GMEI/MNPCA

Why studying precipitation in BLSC (Boundary Layer Stratocumulus Clouds ) ?

Parameterization of drizzle formation and precipitation in BLSC is a key step in numerical modeling of the aerosol impact on climate

Why studying Stratocumulus clouds ?- Radiative properties : ALBstrato ~10*ALBsea - Large occurrence : ~ 20-30 % of the ocean’s surface.

Negative global radiative forcing

Hydrological point of view :Precipitation flux in BLSC ~mm d-1 against ~mm h-1 in deep convection clouds BLSC are considered as non precipitating clouds

Energetic point of view :1mm d-1 ~ -30 W m-2 Significant impact on the energy balance of STBL and on their life cycle

Aerosol impact on climate

Na

rvNc

precipitations

The problem of modeling precipitation formation in GCM

Presently in GCM : parameterisation schemes of precipitation directly transposed from CRM bulk parameterization. Example : )(3/73/1

critCottonManton rvrvHLWCNAUTO c

Problem- no physically based parameterisations- Numerical instability due to step function

Are such parameterisations, with tuned coefficients, still valid to study the AIE?

2nd solution

A parameterisation of the precipitation flux averaged over an ensemble of cells is more relevant for the GCM resolution scale

Underestimation of precipitation

1st solution

This bias is corrected by using tuning coefficients

In Manton-Cotton parameterisation : rvcrit=10 µm

In GCM : rvcrit reduced down to 5 µm.

Problem : Inhomogeneity of microphysical variables.

Formation of precipitation = non linear process local value have to be explicitely resolved

LES resolution: ~100m horizontally, ~10 m vertically

3D view of LWC = 0.1 g kg-1 isocontour, from the side and above.

LES domain Corresponding cloud inGCM grid point

~100min BL

~100kmHomogeneous

cloudCloud fraction F<qc>, <Nc> (m-3)

In GCM : variables are mean values over 10 to 100 km scales

smoothing effect on local peak values.

Super bulk parameterisation

At the scale of an ensemble of cloud cells : quasi stationnary state

Is it feasible to express the mean precipitation flux at cloud base <Fprec> as a function of macrophysical variables that characterise the cloud layer as a whole ? (Pawlowska & Brenguier, 2003)

Pawlowska & Brenguier (2003, ACE-2):

N

HFprec

3

N

HFprec

4

75.1)(N

LWPFprec

Comstock & al. (2004, EPIC) :

Van Zanten & al. (2005, DYCOMS-II) :

Which variables drive <Fprec> at the cloud system scale ?

Adiabatic model :LWP = ½CwH2

<Fprec> (kg m-2 s-1 or mm d-1)

H (m)or

<LWP> (kg m-2) N

(m-3)

In GCMs, H (or LWP) and N can be predicted at the scale of the cloud system- The LWC sink term due to precipitation, averaged over numerous cloud cells, can then be expressed as a function of these two variabless :

H

F

t

LWC precprec

)( (kg m-3 s-1)

Objectives & Methodology

Methodology:3D LES simulations of BLSC fields with various H (LWP) and N values

Objectives : - use LES to establish the relationship between <Fprec>, LWP and N, and empirically determine the coefficients.

H or <LWP>, N<Fprec>

a = ?α = ?β = ?

LES domain GCM grid point

averaged LWP, N, and <Fprec>over the simulation domain

N

HaFprec

10 km

LES microphysical scheme- Implementation in MESONH of a modified version of the Khairoutdinov & Kogan (2000) LES bulk microphysical scheme (available in MASDEV4_7 version).

Specificities :

- 2 moments -> predict N for studies of the aerosol impact

-- specifically designed for BLC = low precipitating clouds

- coefficients tuned using an explicit microphysical model as data source -> using realistic distributions.

- LES scheme -> valid only for CRM.

- Modifications : Cohard and Pinty (1998) activation scheme and add of droplet sedimentation process.

Condensation& Evaporation : Langlois (1973)

Autoconversion : K&K (2000)

Accretion :K&K (2000)

Sedimentation of drizzle : K&K (2000) Activation :

Cohard et al (1998)

Evaporation : K&K (2000)

Aerosol : NCCN (m-3)

(Constant parameter)

+

Vertical velocity : W

Nact (m-3)

Cloud :

qcloud (kg/kg)

Ncloud (m-3)

Drizzle:

qdrizzle (kg/kg)

Ndrizzle (m-3)

Sedimentation of cloud droplets

Stokes law + gamma

Vapour:

qvapour (kg/kg)

Microphysical processes & microphysical variables.

79,147,21350)(

ccautor Nqt

q

15,1)(67)( rcaccrr qqt

q

1,0007,0 vrN rVr

2,0012,0 vrq rVr

21

0

21 )( MNkdnkF ccNc

dnvF cwqc

0

3 )()(6

dnvF cNc

0

)()(

52

0

51 )(

6MNkdnkF ccwqc

(H) : Stokes regime: 21)( kv

Parameterisation of cloud droplets sedimentation

Calculation of the cloud droplet sedimentation process requires an idealized droplet size distribution.

Objective : Which distribution to select? With which parameter ?

))ln

)Ø/Øln((

2

1exp(

lnØ2

1)Ø( 2

g

n

g

cn

))Ø(exp(Ø)(

)Ø( 1

cnGeneralized gamma law : Lognormal law :

Methodology.By comparing with ACE-2 measured spectra (resolution = 100 m),find the idealized distribution which best represents the :

- diameter of the 2nd moment ,- diameter of the 5th moment ,

- effective diameter .e52

The cloud sedimentation flux depends on the 2nd and 5th momentsRadiatives flux in LW depends on the effective radius .

Results for gamma law, α=3, υ=2

Number of spectra in% of max_pts

100 %

50 %

0 %

Ø2 σ

ØeØe

Ø5 measured

gamma

5

5

Ø

Ømeasured

gamma

2

2

Ø

Ømeasured

gamma

- Generalized gamma law : best results for α=3, υ=2- Lognormal law, similar results with σg=1.2 ~ DYCOMS-II results (M.C. Van Zanten personnal communication).

measured2Ø measured

5Ø measured

measuredeØ measured

eØ

measurede

gammae

Ø

Ømeasurede

gammae

Ø

Ø

only spectra at cloud top

Results for lognormal law, σg=1.5

% of max_pts

100 %

50 %

0 %

Ø2 σ

ØeØe

Ø5 measured

gamma

5

5

Ø

Ømeasured

gamma

2

2

Ø

Ømeasured

gamma

Lognormal law, with σg=1.5, overestimate sedimentation flux of cloud droplets.

measured2Ø measured

5Ø measured

measuredeØ measured

eØ

measurede

gammae

Ø

Ømeasurede

gammae

Ø

Ø

only spectra at cloud top

GCSS intercomparison exercise Case coordinator : A. Ackermann (2005)

Case studied : 2nd research flight (RF02) of DYCOMS-II experiment (Stevens et al., 2003)

• Domain : 6.4 km × 6.4 km × 1.5 km

horizontal resolution : 50 m,

vertical resolution : 5 m near the surface and the initial inversion at 795 m.

• fixed LW radiative fluxes,

• fixed surface fluxes,

• fixed cloud droplet concentration : Nc = 55 cm-3

• 2 simulations :

- 1 without cloud droplet sedimentation.

- 1 with cloud droplet sedimentation : lognormale law with σg = 1.5

Microphysical schemes tested : - K&K scheme,

- C2R2 scheme (= Berry and Reinhardt scheme (1974)).

4 simulations. K&K, sed ON / sed OFF C2R2, sed ON / sed OFF

Results, LWP, precipitation flux

Central half of the simulation ensemble

Ensemble range

Median value of the ensemble of models

K&K, sed : ON

K&K, sed : OFF

NO DATA

LWP (g m-2) = f(t)

Precipitation flux at surface (mm d-1) = f(t)

Precipitation flux at cloud base (mm d-1) = f(t)

C2R2, sed ON

C2R2, sed OFF

6H 6H3H 3H

3H

3H

6H

6H3H6H

observations

- LWP a little too low- Underestimation of precipitation flux

~0.35 mm d-1

~1.24 mm d-1

Results,discussion

Strong variability of N and Fprec:

Black : Fprec > 5 mm d-1Light grey : Fprec < 1 mm d-1

Nc (cm-3)

Variation of Nc along 1 cloud top leg

Resolution : 1 km(Van Zanten et al, 20004)

measures

Nc < 55 cm-3 in heavily precipitating areas.

Results, What about microphysics ? Observations

Variations of N, geometrical diameter for cloud and for drizzle, along 1 cloud top leg, 1 cloud base leg.

(Van Zanten personnal communication).

Averaged profils on precipitating grid points after 2 hours of simulation : Ndrizzle, qdrizzle, Øvdrizzle, Øvcloud

C2R2K&K

<top height>

< base height>

Ndrizzle(l-1) qdrizzle(g kg-1)

Øvdrizzle(µm) Øvcloud(µm)

Simulations

- Underestimation of precipitation flux at the base for K&K scheme and C2R2 scheme. Nc is too large in simulation? LWP is too low?- K&K scheme reproduce with good agreement microphysical variables. C2R2 scheme : large and few drops.

Nc (cm-3), Ndrizzle (l-1) Øgcø, Øgdrizzle (µm)

Cloud Topleg

Cloud baseleg

K&K

C2R2

Results, super bulk parameterization

y = 4E+14x2,2651

R2 = 0,9649

0,00E+00

2,00E-06

4,00E-06

6,00E-06

8,00E-06

1,00E-05

1,20E-05

0,00E+00 5,00E-10 1,00E-09 1,50E-09 2,00E-09 2,50E-09 3,00E-09

LWP/N (kg m-2 / m-3)

3,214 )(104N

LWPFprec

<Fprec> : averaged precipitation flux

at cloud base (kg m-2 s-1)

• 7 simulations with different values of N : Na = 25, 50, 75, 100, 200, 400, 800 cm-3 -> different values of N

• Simulations of diurnal cycles -> variations of LWP

• Domain : 2,5 km * 2,5 km * 1220 m• horizontal resolution : 50 m,

vertical resolution : 10 m.

<Fprec> = (LWP/N)

Conclusion & Perspectives

- Cloud droplet sedimentation :

Best fit with α = 3 , υ = 2 for generalized gamma law,

σg = 1,2 for lognormal law.

- Validation of the microphysical scheme :

GCSS intercomparison exercise

The K&K scheme shows a good agreement with observations for microphysical variables

Underestimation of the precipitation flux with respect to observations.

Nc too large ? -> Simulations with Nc prognostic

Simulation of 2 ACE-2 case

-> Simulations of a clear and a polluted case of the ACE-2 experiment and comparison with observations

- Parameterisation of the precipitation flux for GCM :

corroborates experimental results : <Fprec> is a function of LWP and N

-> 3D simulations over a larger domain in order to improve statistics

-> 1D water budget simulations for explaining the dependence