PBL schemes for ICON: CGILS test Martin Köhler (DWD)
Dmitrii Mironov, Matthias Raschendorfer, Ekaterina Machulskaya (DWD)Roel Neggers (KNMI)
Prognostic TKE (Raschendorfer, COSMO & GME)
Prognostic TKE, , , (Mironov, Machulskaya, experimental)
EDMF-dry/stratocu (Köhler, Beljaars, ECMWF)
EDMF-DUALM-shallow (Neggers, Köhler, Beljaars, experimental)
CGILS test
2 2q q
TKE schemes
TKE-Scalar Variance Closure ModelDmitrii Mironov
• Transport (prognostic) equations for TKE and variances of scalars (<’2> and (<qt’2>) including third-order transport.
• Algebraic (diagnostic) formulations for scalar fluxes, Reynolds-stress components, and turbulence length scale (for speed).
• Statistical SGS cloud scheme, either Gaussian (e.g. Sommeria and Deardorff 1977), or with exponential tail to account for the effect of cumulus clouds (e.g. Bechtold et al. 1995).
• Optionally, prognostic equations for scalar skewness (mass-flux ideas recast in terms of ensemble-mean quantities).
Treatment of Scalar Variances
TKE equation:
22
2
1
2
1wzz
wt
pwuwz
wg
z
vvw
z
uuw
t
ei2
2
1
Scalar-variance equation:
Convection/stable stratification = Potential Energy Kinetic Energy
No reason to prefer one form of energy over the other!
Comparison with One-Equation Models(Draft Horses of Geophysical Turbulence Modelling)
Scalar variance equation:
22
2
1
2
1wzz
wt
Production = Dissipation
Flux equation:
No counter-gradient term
2
gC
zeCw bg
EDMF schemes
EDMF at ECMWFConvective Boundary Layer
dry EDMF theory & SCM
Pier Siebesma & Joao Teixeira 2000, 2007
stratocumulus EDMF & unified implementation
Martin Köhler 2005, 2010
stratocumulus inversion entrainment numerics
Martin Köhler 2008
shallow cumulus DUALM EDMF
Roel Neggers & Martin Köhler 2007-2010
ECMWF operational
EDMF at ECMWF:Stratocumulus
• sl, qt conserved variables
• M surface driven
• cloud top down diffusion
• cloud top entrainment
• cloud scheme: conversion (Beta distr.)
• stability criteria allowing strcu
lqtq
preVOCA: VOCALS at Oct 2006 – Low Cloud
EDMF at ECMWF:Shallow Cumulus DUALM
Neggers, Köhler, Beljaars 2009
Concepts: • multiple updrafts
• mass-flux closure
• entrainment pre-moistening
• bimodal statistical cloud scheme
• cloud overlap
Brian Mapes (~1995 GCSS meeting):Postulates that convection selects favourable environment.
Peter Bechtold (2008):Moist environments yield less entrainment.
Convective premoistening
Brown, Zhang 1997
RH during TOGA/COARE
Moist low levels (~800hPa) favour deep convection
RH (%)
Derbyshire et al 2004MetO CRM CNRM CRM
MetOffice SCMIFS SCM
Environment RH
RH (%)
mass flux mass flux
• small ε to get high cloud top
• large ε to get large RH sensitivity
Jarecka, Grabowski, Pawlowska, 2009
cloud fraction (grid box)
box
env
RH
RHenvironment
Entrained air is premoistened.
BOMEX LES runentrainment
regime
BOMEX LES cloud blobs
x
t
cloud blob time scalecloud
dt
cloud blob identification from LWP boundaries
WVP
x
y
BOMEX LES cloud blobs
blobs size 1000: (250m)2 · 300s
Time, lagged around blob center, normalized by blob time scale
166 blobs size 1000-10000
shifte
d b
lob m
ean W
VP [
g/m
2]
/ cloudt
100g/m2
40g/m2
2890g/mWVP
prognostic total water variance equation
most moist environment favours shallow convection
decay time-scale outside BL
3 hours
DUALM convective preconditioningMartin Köhler & Olaf Stiller & Thijs Heus
2 2' ' '
2 ' 't tq qt t tq w q qw q
t z z
, ( )t upup env
qq q
z
LCLqt
10%
qt
10%
qt
10%
time
height
prog. 2
tq
decay2
tq
moist
CGILS results
Equilibrium state (80-100days)
cloud cover[%]
liquid water[g/m2]
water vapor[kg/m2]
sensible[W/m2]
latent[W/m2]
S12 ctr 100 79 40 13 19 19 21 10 72 68
p2k 100 79 51 16 24 24 16 6 86 84
S11 ctr 100 71 115 49 22 23 15 7 93 87
p2k 100 79 122 64 26 28 15 6 101 100
S6 ctr 16 17 26 25 36 35 9 8 108 108
p2k 17 22 30 35 42 43 10 9 113 116
EDMF-strcu EDMF-DUALM-shallowcu
EDMF-strcu (and Tiedtke shallow)
ql RH
Time [days]
S12ctl p2k
S6ctl p2k
S11ctl p2k
ql RH
Time [days]
ql RH
Time [days]
ql RH
Time [days]
ql RH
Time [days]
ql RH
Time [days]
EDMF-DUALM-shallowcu
ql RH
Time [days]
S12ctl p2k
S6ctl p2k
S11ctl p2k
ql RH
Time [days]
ql RH
Time [days]
ql RH
Time [days]
ql RH
Time [days]
qlRH
Time [days]
conclusions
• ICON model• boundary layer: TKE and/or EDMF closures• clouds: probably prognostic PDF, prognostic ice
• EDMF models at ECMWF have negative cloud climate feedback
• mostly more LWP
Extra Slides: CGILS talk
EDMF differences
• Cloud diagnostic: • EDMF-strcu: Beta-distribution (bounded) CCstrcu=100%• EDMF-DUALM: Gaussian distribution (open) CCstrcu=80%
ECMWF EDMF framework
Siebesma & Cuijpers, 1995
)()1( uu
e
e
u
u wawawaw
M
M-fluxenv. fluxsub-core flux
K-diffusion
Single-Column Tests: Dry Convective PBL
Mean potential temperature in shear-free convective PBL.
Red – TKE scheme, blue – TKE-scalar variance scheme, black dashed – LES data.
Single-Column Tests: Nocturnal Stratocumuli
Fractional cloud cover (left) and cloud water content (middle) in DYCOMS-II.
Red – TKE scheme, blue – TKE-scalar variance scheme.
Black solid curve in the right figure shows LES data.
Single-Column Tests: Shallow Cumuli
Fractional cloud cover (upper row) and cloud water content (lower row) in BOMEX.
Red – TKE scheme, blue – TKE-scalar variance scheme. Black solid curves in the middle figures show LES data.
Gaussian
Gaussian
skewed
skewed
Louise Nuijens: LES of cumulus, influence of wind speed
BOMEX LES preconditioning of convection?
LES by Thijs Heus:
no sheardx=dy=25m, dt=30sduration: 10h6.4km x 6.4km
WVP’ [g/m2]
PD
F
WVP
x
y
WVP’ [g/m2]
LWP [
g/m
2]
buoyancy
v dz
31
Conclusion: PDFs are mostly approximated by uni or bi-modal distributions, describable by a few parameters
More examples from Larson et al. JAS
01/02
Note significant error that can occur if PDF
is unimodal
PDF Data
UKMO: PC2 prognostic variables
Ideas for ICON-NWP
Questions on complexity:
Skewness, PC2, temperature variability
Questions on framework (prognostic variables):
Tiedtke, PC2
Summeria/Deardorff, Tompkins
Possible compromise:
Concept (Gaussian qt=qv+ql+qi, qi from microphys.)
Assumptions:
no T variability
mixed cloud: ice/liquid co-located (no PC2)
equilibrium vapor/liquid (not ice!)
Ideas for ICON-NWP
Turbulence parameterization:
TKE (Raschendorfer) diagnostic , ,
UTCS (Mironov) prognostic , ,
EDMF (Köhler et al) prognostic
Convection parameterization:
Bechtold et al 2008 (evolved Tiedtke 89)
Tendencies: ql, qi, cloud fraction cc
Microphysics:
Doms & Seiffert
Ice homogenious and heterogenious nucleation
Saturation adjustment on the sub-grid scale
2tq 22tq
2tq 2 q
q
Clouds and temperature/moisture variability
Tompkins, 2003
MOZAIC T, RH and e variability
PDF of 300km legs at 166-222 hPa. Gierens et al 1997.
K4.0T%2RH Pa07.0e
Estimate T variability if = const:
Estimate z displacement from T variability: =>
Estimate ΔRH from Δe: =>Estimate ΔRH from ΔT: =>
v
Temperature RH Vapor partial pressure
K4.00005.0 TKT
lvv qqT 1
km
K10
z
T m40z
PaCTe osat 5030 %13.0RH
%3RH
Final Thoughts
Cloud variability is important down to <1km.
radiation
microphysics
Ice microphysics are equally important
Both macro and micro-scales involve long time-scales
We need at least
prognostic total water variance (or cloud fraction)
prognostic ice water
LES clouds
LES: mostly all-or-nothing (e.g. SAM, UCLA-LES, KNMI, UKMO)
GCM:
diagnostic (RH based, Slingo)
prognostic CC (Tiedtke)
prognostic (Tompkins)
Tompkins, 2003
Pro
port
ion
all-c
lear
or
all-c
loud
y le
gs
Leg length [km]
Based on 4400km of flight data near ARM SGP at 1-3km height.
2q
GME and COSMO clouds
Stratiform sub-grid scale cloud:
RH based
Notes: • qsat is interpolated between qsat,liq and qsat,ice
between -5ºC and -25ºC
• ql and qi are 5% of qsat
Ideas: cloud physics at macro- and micro-scales
fc=diagnostic
4 moments: vq lq22qttq iq
Liquid cloud
qt
liqsatq ,
tq
liquid
Mixed cloud
qt
effsatq ,tq
liquid
ice
icesatq ,
Ice cloud
qt
effsatq ,tq
ice
icesatq ,
Sub-grid variability:
• assume Gaussian
• neglect variability
• take fixed ice fraction from
microphysics
tq
il
i
q