PBL transports and clouds 32007 Martin Khler 1

1
Operational Models Readiness NowMartin Köhler,
ECMWF with Hua-Lu Pan, NCEP and Shouping Wang, NRL

• observational data
• EPIC
• DYCOMS-II
• GLAS cloud top
• GCSS Pacific Cross-Section
• ocean buoys
• AOSN field campaign (COAMPS)
• cloud vector winds
• QuikSCAT winds
• radiosonde winds
• SST for coupled GCMs

• model evaluation (dt12h to 10yr)
• the good
• qmix 0.5 g/kg
• ?l 0.5 K
• LWP 50 g/m2
• CC 10
• ?Tinv 2 K
• the bad
• SST 2 K
• usfc 1 m/s
• inv. height - 200 to 500m
• the unknown
• w 50
• u850 2 m/s
• drizzle ?
• aerosols ?

2
EPIC Peruvian stratocumulus model comparison
Bretherton et al, BAMS 2004
qv g/kg
LWC g/m3
q K
3
Stratocumulus EPIC column from 3D forecasts
4
Stevens et al 2007 DYCOMS vs ECMWF vs NCEP
5
Shouping Wang, NRL COAMPS Forecast
Statistics (East Pacific)
6
Shouping Wang, NRL COAMPS Forecast
Statistics (East Pacific)
7
ECMWF vs GLAS observations cloud top height
ECMWF cloud top height
ECMWF strcu fraction
GLAS strcu fraction
GLAS cloud top height
SC top too low!
Maike Ahlgrimm
cloud top lt 2km, cld gt 80
8
Stevens et al 2007 DYCOMS vs ECMWF vs NCEP
Taylor diagram
Divergence
?PBL
vPBL
uPBL
qPBL
?850
ECMWF NCEP
9
IMET buoy (Anton Beljaars Bob Weller)
wind m/s
EPIC
q2m C
T2m C
10
Shouping Wang, NRL COAMPS vs 2 buoys off
Monterey
Field Bias RMS 3-km
9-km 3-km 9-km Temp. 0.93 0.79 1.56 1.60 Speed
1.48 0.41 2.60 1.80 Dir. 9.10 35.0 47.0 48.0
AOSN field campaign in the vicinity of Monterey
Bay, California
11
GCSS Pacific Cross-Section Intercomparison
NCEP
C. Hannay
JJA1998
Joao Teixeira
12
Cloud cover against ISCCP D2
CY32R3 - ISCCP
ISCCP D2
13
ECMWF buoy verification
model mean 6.08 m/s buoy mean 6.12
m/s bias -0.04 m/s RMSE
1.12 m/s correlation 0.916
14
QuikSCAT winds Bias and RMSE
0 bias
Bias m/s
1.4 m/s
RMSE m/s
15
QuikSCAT winds
bias0.06m/s
QuikSCAT7.17m/s
16
GCSS Pacific Cross-section Project
Claire Delsol.
Hans Herbach
17
SE Pacific U-profiles
Analysis
48h Forecast
CY31R1
ltDIFF
pressure hPa
pressure hPa
18
Cloud Vector Winds versus Radiosondes
Ascencion St. Helen
French Pacific Islands
19
DEMETER CGCM Surface Temperature Bias K
ECMWF (ERA40 cycle)
UKMO
MPI
Meteo France
4-6 month forecasts Aug/Sep/Oct 1987-1996 9
ensemble members comparison to ERA-40 www.ecmwf.i
nt/research/demeter
LODYC
20
coupled ECMWF-ERA40 Surface temperature years
3-4 of integration
Antje Weisheimer ENSEMBLES
CY31R1 Sept 2006
CY32R1 June 2007
CY32R3 Fall 2007
21
EC EARTH (ECMWF uncoupled) future 2XCO2
scenario
Pier Siebesma, KNMI

• CC
• Standard model (cy31r1)

D CC Enhanced top-entrainment model
22
Operational Models Readiness NowMartin Köhler,
ECMWF with Hua-Lu Pan, NCEP and Shouping Wang, NRL

• observational data
• EPIC
• DYCOMS-II
• GLAS cloud top
• GCSS Pacific Cross-section
• ocean buoy
• AOSN field campaign (COAMPS)
• cloud vector winds
• QuikSCAT winds
• radiosonde winds
• SST for coupled GCMs

• model evaluation (dt12h to 10yr)
• the good
• qmix 0.5 g/kg
• ?l 0.5 K
• LWP 50 g/m2
• CC 10
• ?Tinv 2 K
• the bad
• SST 2 K
• usfc 1 m/s
• inv. height - 200 to 500m
• the unknown
• w 50
• u850 2 m/s
• drizzle ?
• aerosols ?

23
Conclusion Discussion points
ECMWF model whats good? whats bad? what dont
we know? Also mentioned NCEP, COAMPS, and
EU/USA climate models
24
(No Transcript)
25
coupled ECMWF-ERA40 Surface temperature years
3-4 of integration
Antje Weisheimer ENSEMBLES
CY31R1 Sept 2006
CY32R1 June 2007
CY32R3 Fall 2007
26
ECMWF PBL cloud parameterization futureMartin
Köhler, ECMWF

• Sept2006 ice supersaturation, implicit
  cumulus numerics
• June2007 radiation McICA, RRTM SW (14 bands)
• Fall2007 cumulus entrainment, PBL
  stratocumulus upgrades
• Spring2008 EDMF shallow convection DualM
• EDMF momentum transport
• exact parcel solution

27
Integral appraoch to PBL transports EDMF
operational
28
EDMF two box M/K decomposition(Siebesma and
Cuijpers, 1995)
29
A statistical mass flux framework for organized
eddies
dry PBL
30
A statistical mass flux framework for organized
eddies
Stratocumulus UP
31
A statistical mass flux framework for organized
eddies
Stratocumulus UP / DOWN
32
A statistical mass flux framework for organized
eddies
Shallow Convection
Roel Neggers
33
Shallow ConvectionRoel Neggers, Martin Köhler,
Anton Beljaars
34

• II. A proposed set of modifications
• Enables EDMF to also represent shallow cumulus
• (replacing the current shallow cumulus
  scheme)
• increased number of resolved updrafts
• flexible area partitioning of the updraft
  ensemble
• -gt determined by moist convective
  inhibition
• Allows gradual transitions to and from
  shallow cumulus
• flexible updraft entrainment
• The same entraining plume model is applied
  to all model updrafts
• flexible vertical structure of cumulus mass
  flux
• Dependent on inversion stability a bulk
  Kain Fritsch (1990) scheme
• -gt to reproduce cloud layers with
  varying degrees of conditional instability
• top-entrainment efficiency closure at shallow
  cumulus inversion
• Wyant et al. (JAS, 1997)
• a bimodal statistical cloud scheme within the
  PBL

35
So does it work better?

• Step 1 Revisit as many prototype LES/CRM cases
  as possible

ql
qsat
qt
This should be a routine step in model development
Cloud fraction
Condensate
36
Shallow ConvectionRoel Neggers , Martin Köhler,
Anton Beljaars
moist
dry
K
37
Enhancing model complexity

• I. The Eddy Diffusivity Mass Flux (EDMF)
  framework
• For turbulent transport in well-mixed layers
  Siebesma et al. (JAS, in press, 2007)

Updraft transport
K diffusion
mixed layer
diffusive flux
aup
aK
advective (mass) flux
EDMF already represents dry and stratocumulus
convection in the currently operational ECMWF
forecast model
38
EDMF PBL analytical updraft(with Peter Janssen)

• Parcel dominates PBL height (wu0) and updraft
  properties (?u, qu, uu).
• PBL height dominates mixing.
• Parcel is dominated by entrainment.
• Entrainment rule gives
  infinite entrainment at surface and PBL top.

Rescale
Updraft equations
Then
Solution
PBL height
Assume
39
EDMF PBL analytical updraft(with Peter Janssen)
40
EDMF PBL non-local momentum transport

• Parcel excess initialization
• 
  (excess anti-correlated to environment)
• Updraught equation

41
Diffusion outside convective PBL
42

• Convection revisions New formulation of
  entrainment and deep convective adjustment time.
  No more an updraught iteration is performed, but
  only one single computation. Furthermore,the
  momentum transport has been rewritten (it is now
  concentrated in one single routine,
  i.e.cumastrn.F90) and several code optimisations
  are performed. The overall gain in computing time
  is around 1 for the deterministic forecast.
• (2) Apply rain freezing below 0C, and use
  different critical relative humidities for rain
  evaporation below cloud over land and water (0.7
  and 0.9, respectively)
• Adjustment to some microphysical parameters, in
  particular the melting time scale has been
  increased by a factor of 1.5, the autoconversion
  time scale for cloud droplets is reduced form
  7200 to 6000 s, the diffusion coefficient for
  clouds is increased from 2.e-6 to 3.e-6 and the
  effective ice particle radius used in the
  radiation is reduced from 60 to a value of 45.
• Together with Judith Berner provide revised
  formulation of convective contribution based on
  vertically integrated updraught kinetic energy
  (as also now used in convective closure) to total
  dissipation in experimental stochastic physics.
  Clean up again of budgets and units
• (1) PBL bug fixes Various bugs related to
  the PBL updraught and cloud top entrainment were
  found and fixed. Climate tests showed
  insignificant impacts.
• (2) PBL mass flux numerics The mass-flux
  component in the PBL implicit solver is now
  treated with upwind finite differencing as
  opposed to centered previously. This improves the
  stability of the scheme by allowing about a
  factor 2 higher mass fluxes.
• (3) Mass flux limiter The mass-flux limited
  was set to 2CFL instead of 1CFL which was made
  possible by upgrade (2).
• (3) Parcel entrainment The parcel
  entrainment was decreased to get more realistic
  higher boundary layers including stratocumulus
  situations and to be in line with LES data.
  Parcel entrainment is now eps0.4/z(full layer)
  1/(500swup).
• (4) Diffusion coefficient K K was decreased
  because it was shown to destroy inversions above
  stratocumulus and introduce excessive diffusion
  in sheared situations. The Louis/Tiedtke/Geleyn
  increased K was kept near the surface to keep
  stable boundary layers untouched. For higher
  layers (150m) K now moves assymptotically to a
  Monin-Obukov formulation, which is supported by
  multiple data.

43
RMS error Tropic 850hPa Wind
44
MJO Tropical Velocity Potential
45
MJO Power Symmetrical Tropical OLRA
CY32R3
NOAA obs
CY31R1
CY32R1
46
Precipitation (DJF 1990-2005)
GPCP obs
CY32R3 - obs
CY32R1 - obs
CY31R1 - obs
47
Cloud cover against ISCCP D2
ISCCP obs
CY32R3 - obs
CY31R1 - obs
CY32R1 - obs
48
(No Transcript)
49
Extra Slides
50
EDMF two box M/K decomposition(Siebesma and
Cuijpers, 1995)
51
ECMWF vs GLAS cloud fraction
ECMWF
GLAS res 76m
ITCZ
off Chile
ARM SGP
Maike Ahlgrimm
52
EDMF PBL analytical updraft(with Peter Janssen)

• Parcel dominates PBL height (wu0) and updraft
  properties (?u, qu, uu).
• PBL height dominates mixing.
• Parcel is dominated by entrainment.
• Entrainment rule gives
  infinite entrainment at surface and PBL top.

Rescale
Updraft equations
Then
Solution
PBL height
Assume
53
EDMF PBL analytical updraft(with Peter Janssen)
0.5
0.0
Z
positive buoyancy branch
-0.5
-1.0
-0.5
0.5
0.0
1.0
T, W
54
EDMF PBL analytical updraft(with Peter Janssen)
55
EDMF PBL analytical updraftdry growing PBL case
numerical 400s
numerical 500s
LES
Siebesma
analytical 400s
analytical 500s
56
(No Transcript)
57
EDMF PBL non-local momentum transport

• Add MF but preserve surface M-O similarity theory
• Parcel excess initialization
• 
  (excess anti-correlated to environment)
• Updraught equation

58
parameterization choices
Mass-flux
K-diffusion

• updaft model
• entrainment , t500s, c0.55
• detrainment 310-4 m-1 in cloud
• parcel determines PBL depth (wup 0)
• mass flux

• diffusion
• surface and cloud top driven diffusion
• cloud top entrainment

cloud variability
59
M centered
M upwind
mass-flux numerics in EDMF scheme
u m/s
u m/s
SCM Brown sheared PBL, M8x realistic
60
parcel entrainment
LES 1/(ez)
cy29r1 0.55
LES fit 0.4
Siebesma et al 2007
61
(No Transcript)
62
AMV talk
63
Wind Profiles Analysis/Forecast/ObservationMart
in Köhler, ECMWF

• RMS and bias maps new PBL, take out AMVs
• U profiles (and T, and AMV detection)
• AMV/radiosonde colocations

64
U850hPa RMS New PBL with less vertical
diffusion - Operations
65
U850hPa bias Take out tropical AMVs -
defaultClaire Delsol
66
U850hPa Bias Add AMV Base run with AMSU
onlyGraeme Kelly
67
SE Pacific T-profiles
68
SE Pacific model cloud fraction, AMV winds
69
Ascencion St. Helen Jul/Aug2006 Windspeed
70
French Pacific Islands Jul/Aug2006 Windspeed
71
ECMWF vs GLAS observations cloud top height
ECMWF cloud top height
ECMWF strcu fraction
GLAS strcu fraction
GLAS cloud top height
SC top too low!
Maike Ahlgrimm
cloud top lt 2km, cld gt 80
72
Conclusions

• new PBL favors shear in SE-Pacific
  SE-Atlantic
• AMV speeds up U-winds in .
• Radiosonde in mid-Pacific indicates fast AMV
  bias
• Speculated error height assignment bias and RMS

73
(No Transcript)
74
(No Transcript)
75
(No Transcript)
76
GCSS/WGNE Pacific Cross-section Intercomparison
(GPCI) Joao Teixeira
ISCCP total cloud cover
Sea Surface Temperature
C. Hannay

• Models GFDL, NCAR, UKMO, JMA, MF, KNMI, DWD,
  NCEP, ECMWF, BMRC, NASA/GISS, UCSD, UQM, LMD,
  CMC, CSU, GKSS