WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle :...
Transcript of WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle :...
![Page 1: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/1.jpg)
WG3a-Activities and PT ConSAT:
Matthias Raschendorfer
� Experiences with the stochastic boundary layer pert urbation
� Results from monitoring the operational LPI in COSM O-DE
� Implementation of Tiedke-Bechthold Cumulus Convecti on Scheme
� Implementation and validation of the EDP-forecast d erived from TURBDIFF within the
ICON model
� Recent verification results of the common blocked T URBDIFF with COSMO-DE
� Towards a new diagnostic of equilibrium surface tem perature in combination with SAT
and the soil model
� Applicability of a tile approach in cases with stab le stratification
Matthias Raschendorfer COSMO-GM Offenbach 2016
![Page 2: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/2.jpg)
Boundary layer perturbations for convection triggering in
COSMO-DE
COSMO General Meeting 2016, Offenbach, 5.9.2016 2
Ulrich Blahak 2 (DWD), Kirstin Kober 1 (LMU)
1 Original inventor2 Implementation and testing at DWD
![Page 3: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/3.jpg)
Generation of perturbation fields:
� stochastic pert. of T, qv and w in the PBL only, coupled to the variances of these quantities as derived in the turbulence scheme (Kober et al, 2015)
� original:
� modified:
� Φ = {T, qv, w}
� Stddev(Φ) diagnosed from turbulence scheme (only itype_turb=3)
� Choose space- and time-coherence scales for random number field below effective model resolution of these two quantities
� αsh = namelist parameter (<= 5, otherwise danger of crashes!)
� ηsh = 2D random number field, smoothed by Gaussian kernel to generate coherent structures. Held constant for typical eddy turnover times (~10‘)
3
Random perturbations
(devel options by U. Blahak)
(Kober et al., 2015)
![Page 4: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/4.jpg)
4
New namelist parameters ( /RUNCTL/ )
PARAMETER DEFAULT TYPE MEANING
luseblpert .FALSE. L Master switch
itype_blpert 1 * INT 1 = original implementation2…8 = modified options from devel
ladvect_blpert .FALSE. * L If .TRUE. the random numbers ηsh are advected with the windspeed at level ke-10
blpert_sigma 2.5 * REAL STDDEV of Gaussian smoother for random numbers in units of grid points
blpert_const 2.0 * REAL αsh,Φ
blpert_fixedtime 600.0 * REAL Time increment [s] of random number update (~eddy turnover time)
seed_val -999 INT If -999, use either model start time or system time for the initial random number seed
lseed_use_starttime .TRUE. L If seed_val = -999: if .TRUE. ‚ydate_ini‘ determines seed, otherwise system time ‚DATE_AND_TIME()‘
* If luseblpert=.TRUE., these defaults reproduce the orignialKober (2010) settings, aside from the random seeding
![Page 5: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/5.jpg)
5
Experiment 10223 COSMO-DE driven by ICONEU with EDA for August 2015, comparison to 10168
![Page 6: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/6.jpg)
• Spread and RMSE in the 1 h assimilation cycle: improved spread vs. RMSE
TBctrl = setup of Bick et al (2016) as reference(no radial winds, RTPS)
First COSMO-DE-KENDA experiment by Axel using (almost) the setup of Th. Bick
domain averaged precipiation rate vs. obs.
![Page 7: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/7.jpg)
Lightning Potential Index from COSMO-DE
ESSL Testbed 2016, 29.6.2016
7
Ulrich Blahak (DWD)
![Page 8: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/8.jpg)
� Yair et al. (JGR, 2010), Lynn and Yair (Adv. Geosci., 2010)
� Charge separation in thunderstorms is correllated with the simultaneous presence of updrafts , supercoolded liquid water , graupel and other frozen hydrometeor types („cloud ice “, „snow “)
� This concept was modeled by the authors within the LPI-Index:
8
Concept of the LPI
f1, f2 and g(w) see next-next slide!
![Page 9: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/9.jpg)
9
sim. dBZ_max sim. LPI obs. FLR t +- 7.5 min obs. dBZ + LINET
COSMO-DE oper. forecast 28.5.-31.5.2016, combined 00 UTC runs until vv=23 h
Probability-matched relation on COSMO-DE grid:
FLR in km -2 min -1 ≈ 0.017 * LPI in J/kg
�1 flash per grid cell per 15 min (due to obs „+- 7.5 min“)equals LPI ≈ 2.0 J/kg
![Page 10: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/10.jpg)
The comparison with lightning data (flash rates) „by eye“ shows:
�Overall statistics of the space-time distribution of the LPI values > 0 corresponds well with that of the observed flash rates and suggests a simple linear probability-matched relation.
�LPI intimately tied to (and limited by) the explicit simulation of convective cells in the model and its ice microphysics (updraft strength, supercooled liquid, graupel, cloud ice, snow). Ensemble prediction like COSMO-DE-EPS
�LPI leads to different flash signals compared to, e.g., TQG or TOT_PREC alone. Not every convective cell, which has a high TQG, leads to an LPI signal.
�Note: absolute LPI values have to be re-calibrated whenever the model‘s resolution or microphysics is changed.
�Possibility to assimilate Lightning data based on presented statistical LPI-FLR-relation!
10
Summary
![Page 11: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/11.jpg)
Implementation of ECMWF -IFS (Tiedtke-Bechtold)
Cumulus Convection Scheme into COSMO
Jochen Förstner and Dmitrii MironovGerman Weather Service, Research and Development, FE14, Offenbach am Main,
Germany
([email protected], [email protected])
COSMO General Meeting, 5–9 September 2016, Offenbach am Main, Germany
![Page 12: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/12.jpg)
COSMO-EU: Surface Verification (01-10.08.2015, COSMO-EU Domain, 00 UTC forecasts)
![Page 13: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/13.jpg)
COSMO-DE: Surface Verification (01-31.08.2015, 00 UTC fr sts)
1 indicates perfect score
![Page 14: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/14.jpg)
COSMO General Meeting, 5–9 September 2016, Offenbach am Main, Germany
• ECMWF-IFS (Tiedtke-Bechtold) cumulus convection is implemented into COSMO
• IFS scheme is included into the official COSMO code (version 5.05/5.04b)
• Results from test runs look reasonable; verification scores are neutral to slightly positive
In the future:
• Address drizzle problem (cf. ICON)
• IFS scheme may/should become a default option within COSMO (towards unified COSMO-ICON physics)
![Page 15: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/15.jpg)
15
Global EDP forecasting with ICON –implementation and validation
Tobias Göcke & Ekaterina Machulskaya
FE14
COSMO Seminar WG 3a, 05.09.2016, Offenbach
![Page 16: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/16.jpg)
buoyancy production
eddy-dissipationrate (EDR)
0<labil:neutral:stabil: 0<
0>0=
time tendency
+
transport(advection + diffusion)
= + + +
shear production by sub grid scale circulations
0≥
⋅ρ∂ 2
2
1L
qt
′′′′ρ+
ρ
⋅∇
∑=
3
1i
2i
2
L
L
v
q
2
1
v
v
~
~̂
∇⋅′′′′ρ−∑
=
3
1ii
Li vv ˆ~v
Lv
v
wg ′′θ ′′ρ
θ̂( )
′∇⋅′′′′′ρ−∑=
3
1i
iLiL
vv ˆ~v
αρ−
lMM
3
Lq
v
shear production by the mean flow
0≥
v
MC
ML FSq lΓρ=:MM
L FSq lΓρ=:HHL FSq lρ−=:
Matthias Raschendorfer COSMO Rome 2011DWD
( )TKEt ⋅ρ∂ DifAdv + Boy Shr= + + +
GRIB
CShr EDR−
dTKEshs
dTKEcon
dTKEsso+
+
=(GRIB)
(GRIB)
GRIB
POST
above BL
(GRIB)
EDR-Prognosis:
Enabled by the concept of :
separate parameterization of interactingscale regimes
![Page 17: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/17.jpg)
17
Turbulence observed in the neighbourhood of convective events
TKE production due to convection only diagnostically added in post processing so far
Convection as TKE source
![Page 18: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/18.jpg)
18
Verification: ROC
August – September 2015, USA
TKE already contais dTKEsso + dTKEshs
CON: dTKEcon added in postpocessing
HSH: additional dTKEshs-contribution in post-processing
![Page 19: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/19.jpg)
Work related to
PT ConSAT:
![Page 20: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/20.jpg)
Recent verification results of the common blocked T URBDIFF with COSMO -DE:
Matthias Raschendorfer
� Exp 10279: Turbulence model and VDiff as in ICON
Matthias Raschendorfer COSMO-GM Offenbach 2016
� Exp 10281: Turbulence model as in ICON; VDiff wit h current unblocked version
![Page 21: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/21.jpg)
![Page 22: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/22.jpg)
![Page 23: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/23.jpg)
![Page 24: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/24.jpg)
![Page 25: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/25.jpg)
![Page 26: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/26.jpg)
![Page 27: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/27.jpg)
Towards a new diagnostic of equilibrium surface tem perature in combination with SAT and the soil model:
Matthias Raschendorfer
� Completion of the roughness layer model
� Representation of the thermal energy storage of the roughness layer
Matthias Raschendorfer COSMO-GM Offenbach 2016
� Thermal decoupling of a Cover above the dense soil
![Page 28: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/28.jpg)
Bi TT ==0
( )BRBi TTn
iTT −⋅+==1
C
R
BB
R
BRni TaTaTT +===
LAIdln_CnSAI ⋅+=+= 212
• n cover layers including the surface of the dense soil (n=0) are connected by long-wave radiationinteraction and sensible heat exchange
CTCCBBS TaTaT +=
⋅
κ=
κ+λ⋅
⋅κ=
HHH
HH
H
HS
z
z
uk
uz
uSr
0
0
0
00
00
0
11lnln
H
A
H
S
H
SA rrr 00 +=
( ) ( )( ) ( )
→==
+−+−−⋅−=
∞∞→∞
∞
SAI
SAI,
SAISAI
SAISAIS
SAI
111
11
110
• Only a part of the inner surfaces is connected to A by the resistance chain, the other part is forthe inter- surface exchange
000 zd +=σ=σ0T
ATAσ=σ
0=σ=σ S
functions of SAI only
R
C
R
BC aaaa +==+ 1B
GHF
THF
H
Ar0
Strongly effects the LAI-impact of transpiration !
SHF
LHF
d,uSRFPHF
d,uLRF
![Page 29: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/29.jpg)
‘terra’:
‘turbtran’:( )Amm
Sv
v
S
A vupqT
S
A,,,,
*
MHHH
MHSA Ktker ,,
,,
HSAv
S
A r,q,TA
ST valid for next time level may be out of equilibrium
valid for this time level, related to current evaporation
used for next time level
after tile loop: averaged grid cell valuesSvq
Svq*
M,H
Ar0
• Diagnostic of surface temperature:
( ) [ ] ( ) 0000000000 LHFSHFTTLRFLRFPHFLRFSRFTHF SRuTudd,u ++−⋅∂++++= BT
BS TT = ( ) ( ) [ ] ( ) [ ] ( )[ ] ( )BCT
CCTBBuTCCC
TTGHF
TTLHFSHFTTLHFSHFLRFTHFTTt
MCCB
−⋅∂−
−⋅+∂+−⋅++∂+=−⋅∆
0
000000000
CCBS TaTaT += B
( ) CBSBC a/TTTT −+= ( ) R
CBCBR aTTTT ⋅−+=
itype_surf=0 itype_surf=1
![Page 30: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/30.jpg)
![Page 31: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/31.jpg)
![Page 32: WG3a-Activities and PT ConSAT - COSMO model · • Spread and RMSE in the 1 h assimilation cycle : improved spread vs. RMSE TBctrl = setup of Bick et al (2016) as reference (no radial](https://reader034.fdocuments.in/reader034/viewer/2022042023/5e7ab9c218276f2b957d9e4c/html5/thumbnails/32.jpg)
it is designed as a thermal SSO-term!