Continuous treatment of convection:
from dry thermals to deep
precipitating convection
J.F. Guérémy
CNRM/GMGEC
The association of turbulence and convection schemes
have to represent sub-grid convective processes with
and without condensation,
the turbulence scheme (K-diffusion) dealing with
horizontally quasi-homogeneous processes (having a
rather weak vertical extension),
and the convection scheme (mass-flux) dealing with
horizontally heterogeneous processes (having a larger
vertical extension).
 Aim: to go beyond version 4 of ARPEGE-Climat, trying to
achieve a better association of the two schemes, and
improving them at the same time.
Version 4:
Process representation:
Version 4+:
The turbulence scheme - represents
turbulence with and without
condensation until shallow convection,
The turbulence scheme - represents
turbulence with and without
condensation until very shallow
convection (cumulus humilissimus, 1
layer).
The convection scheme - represents
convection with condensation and
The convection scheme - represents
convection with and without
condensation (PBL dry thermals),
precipitating or not.
taking into account an overestimated
PBL mixing length.
precipitations (all condensed water),
while cancelling the sub-grid transport
produced by the turbulence scheme.
The time tendencies of both schemes
are added.
Schemes:
- Version 4:
- Turbulence: Ricard-Royer 1993, TKE (production= dissipation,
Mellor-Yamada 1982), turbulent sub-grid scale cloud scheme
(Deardorff, Mellor 1977), without prognostic condensate;
precipitations (Smith 1990).
- Convection: Bougeault 1985, mass flux.
- Version 4 +:
- Turbulence: Ricard-Royer 1993 modified by GuérémyGrenier 2005 (mixing length and top PBL entrainment),
TKE (production= dissipation, Mellor-Yamada 1982), turbulent
sub-grid scale cloud scheme (Deardorff, Mellor 1977), with
prognostic condensate; precipitations (Smith 1990).
- Convection: Guérémy 2005, mass flux; precipitations
(Smith 1990).
- Version 5:
Process representation:
Idem version 4+
Schemes:
- Turbulence: Cuxart-Bougeault-Redelsperger 2000,
pronostic TKE, turbulent sub-grid scale cloud sheme
(Deardorff, Mellor 1977), with prognostic condensate;
precipitations (Lopez 2002).
- Convection: Guérémy 2005, mass flux; precipitations (Smith
1990).
Convection scheme
Key elements for a continuous treatment of convection:
compensating subsidence term
X
 X 
d
 DX c  X 

  M
p
and detrainment term, M mass flux
 t  c
-
Cloud Profile: Dry adiabat until the lifting condensation level, then moist
adiabat, including entrainment process.
Mass flux formulation: Product of the grid fraction affected by convective
ascents (equal to the bottom quantity  -to be determined by the closure
condition- times a height decreasing function  -computed from the
convective cloud mass budget-) by the convective vertical velocity prognostic equation-.
M  c
1 c
 o  o
c p
c
c2
g 2 Tvc  Tv    t

1
2

    o  K d c2
t
p 1    Tv


- Entrainment and detrainment :
Organised entrainment and detrainment: Internal computation from the
convective cloud mass budget,
1  c
 ox 
 c p
including a statistical model based upon the concept of buoyancy sorting.
Turbulent entrainment and detrainment: Analytical profile depending on the
convective vertical velocity (large entrainment for a weak ascent and vice
versa).
- Closure condition: CAPE relaxation to zero, according to a characteristic
time proportional to the ratio of the convective depth to the mean
convective vertical velocity.
CAPE
 CAPE 

 

 t  c
 dp
  f (resolution)
b

2
t
t  c dp
b
- Convective precipitation: Precipitation is computed with Smith’s scheme
(such as the stratiform precipitation)
Results
Validation strategy:
Validation starts with 1D simulations of different types of
convective situations corresponding to well documented cases
(observations and explicit simulations), in order to represent
processes at best possible. [EUROCS strategy]
The tuned schemes are then assessed in 3D (annual cycles), giving
possibly rise to a new set of tuned parameters; this new
version is finally tested in 1D to close the cycle.
Bomex Case: Non precipitating shallow convection
V4
V4+
V4
V4+
Cloudiness 7h-16h
Mass flux 7h-16h
Q1 7h-16h
Q2 7h-16h
Theta 16h
Humidity 16h
Entrainment-detrainment 7h-16h
Idealised ARM case: diurnal cycle of continental
convection: from dry PBL to deep precipitating convection
Q1 and Q2 averaged above the PBL (between 800 and 100 hPa)
Prospects
RICO 1D case, notably with ARPEGE-Climat Version5 physics
Validation in 3D LAM (ALADIN-Climat) on documented cases
(observations and CRM simulations), as an intermediate step
between traditional 1D and 3D assessments.
Intensifying 3D global validation (transects, coupled simulation,
seasonal forecasts, …)