Turbulence forecast for aviation WV2 Axel Barleben Boulder,workshop,NCAR 27.08.2013 Turbulence (Eddy Dissipation Rate) forecast based on COSMO-EU 1.introduction/motivation.

Turbulence forecast for aviation WV2 Axel Barleben Boulder,workshop,NCAR 27.08.2013

Turbulence (Eddy Dissipation Rate) forecast based on COSMO-EU

1. introduction/motivation

2. extended turbulence scheme

3. comparison of model output (EDP) and measurement (EDR)

4. abstract and outlook


1.introduction/motivation

Department Aviation Meteorology of DWD

- is responsible for the meteorological support of the Civil Aviation in Germany- the German Meteorological Service Deutscher Wetterdienst (DWD) is a federal authority under the Federal Ministry of Transport, Building and Urban Affairs- DWD operates five regional advisory centres for IFR and VFR traffic- three of them act as Meteorological Watch Offices for aviation weather watch and warning (MWO) resulted from the german airspace structure (3 Flight Information Regions up to FL245, 2 Upper air Information Regions above Fl245)

FIR Bremen

FIRLangen

FIR

München


products containing turbulence forecast

- SIGMET (severe turb FIR/UIR)- GAMET/AIRMET (moderate turb) - AIREP (observation) PIREP (obs)- reports (mod,sev for FIR) - Low Level SWC Central Europe (up to FL 245)- wind shear (up to 1600 ft gnd , rep or exp) (METAR/COMMENTS/WARNING)- briefing for pilots

Turbulence forecast between surface and FL450 with accuracy 1000 ft

support for prediction ?

- General advices about CAT-prone areas- Threshold values of horizontal/vertical wind shear of Richardson number of Ellrod-index- CAT(Maximum)%, WAFC- SWC WAFC

appropriate forecast tools ?


Valid usual method : turbulence is predicted (SIGMET) after turbulence is observed(PIREP)=> require an improvement of turbulence forecast method


eddy dissipation rate [m**(2/3)/s]

ICAO ANNEX3 Appendix 6 /4.2.6 Criteria related to phenomena included in SIGMET and AIRMET:

SEV TURB EDR > 0.7 MOD TURB 0.4< EDR <= 0.7


-measurement of atmospheric turbulence , not directly-not available over Europe-alternative indicator of turbulence is the Derived Equivalent Vertical Gust Velocity (DEVG available , low quantity )

?

-strategic decision of DWD (aeronautical meteorological department) after preliminary studies-development Eddy Dissipation Parameter derived from local model COSMO-EU


Pearson,Sharman 2013 0.014lgt , 0.124mod , 0.345sev

Pettegrew et al. 2010 : PIREP/ACARS => 0.15 lgt , 0.35 mod , 0.55 sev

ICAO ANNEX3 Appendix 6 SEV TURB EDR > 0.7MOD TURB 0.4< EDR <= 0.7 for example: sev turb edr076

Current thresholds derived from comparison EDR/PIREPS 1. introduction/motivation

final aim => EDP (from model output) reproduce the EDR (measurement)

Julia M. Pearson,R.Sharman Calibration of in situ eddy dissipation rate (EDR) severity thresholds based on comparisons to turbulence pilot reports (PIREPs)16th Conference ARAM,Austin,2013


COSMO-EU

-for operational NWP-nested within GME on a 665x657 grid with 40 layers mesh size 7km-based on the primitive hydro thermo dynamical equations-describing compressive no hydrostatic flow-formulated in rotated geographical coordinates -generalized terrain-following vertical coordinate

- turbulence scheme (Raschendorfer,DWD) prognostic level 2.5 closure for the prognostic tke–equation

tke - direct model output (DMO)


http://www.cosmo-model.org/content/model/documentation/core/default.htm


Eddy Dissipation Parameter

222

2

1wvutke

m

TKE

2

2

s

m

pL

tkeedr

3)2(

Kolmogorov (1941) tke is only a function of edrLp - turbulence length scale α – dissipation constant

3

2

s

m ³√

s

m 3/2



buoyancy production

eddy-dissipationrate (EDR)

time tendencyof tke

transport(advection

diffusion)

shear production by sub grid scale circulations

shear production by the mean flow

prognostic tke–equation in principle and simplify - new turbulence scheme Raschendorfer, M.: Further steps towards a scale separated turbulence scheme. 13th COSMO General Meeting, Rome, Italy, 2011

= + + + +

0

v

labile : > 0stable : < 0neutral : = 0


wake vortices by SSO (sub grid scale orography) blocking

horizontal shear vortices

shallow and deep convection patterns



Sub grid scale Kinetic Energy = Turbulent Kinetic Energy + non turbulent Circulation phrased as Sc ,

scale interaction terms of Kinetic Energy SKE = TKE + CKE

Equilibrium of energy production (source terms Qc) and scale interaction term

CSQ

C SL

CKEQ

3)2(

TKE

21

QL 21

pL

CSCKE

pL

tkeedr

3)2(

similar to parameterization edr

scale separation approach and turbulence scheme



TKE-production by separated horizontal (vertical) shear modes:

Equilibrium of production and scale transfer towards turbulence

2

322

__M

HgSSHSCSHSC FDQS

MHF

233

222

211

22332

21331

21221

vvv2

vvvvvv

du2 dv1 du3 dw1 dv3 dw2

2S - effective scaling parameter

du1 dv2 dw3

= (HSH+TSV)

partial derivatives from components of velocity HSH=DEF**2+DIV**2 ; ELD=VSH*(DEF-DIV)

DST=du1-dv2DSH=dv1+du2DIV=-(du1+dv2)DEF=SQRT(DST**2+DSH**2)VSH=du3**2+dv3**2TSV~VSHTSV=VSH+dw1**2+dw2**2+dw3**2


TKE-production by separated wake modes due to SSO:


SSOQQ

nhv

21x ,

3x

B

- blocking Term according Lott und Miller (1997) - describe breaking gravity wave after vertical propagation (and no horizontal) - is estimated currently (SSO-scheme COSMO-EU) - momentum sink due to friction of sub grid scale orography

SSOhSSOC QvQSSSOC

__

Equilibrium of production and loss by scale transfer


moist convection according Tiedke (1987) mass flux convection scheme is installed in COSMO-EUTKE-production can be derived directly

TKE-production by convection (thermal circulations)

CONCCONC SQ __

Equilibrium of production and loss by scale transfer

0ˆ_

vvC

v

CONC wg

Q

virtual potential temperature of ascending air

virtual potential temperature of descending air

vertical velocity scale of circulation



3. comparison of EDP and EDR

maxEDR is measured by commercial aircrafts available over the USACOSMO-EU => COSMO-US was nested over US domain

COSMO-US :dlam/dphi = 0.0625ie = 353je = 314ke = 40philu = -9.075 (startlat)lamlu = -12.608 (startlon)pollon = -86.0pollat = -53.0dt = 40

Data structure winter 2010/11 (01.10-31.3)495 GB grib1 COSMO-US62 GB netCDF ACARS from MADIS archive

Events

maxEDR

MOG MOG/NoTurb

all 6.318.975 1.473.862 23% / 77 %

>2000 m 5.049.488 683.003 14% / 86 %

>6400 m 4.140.997 187.522 5% / 95 %



214443 238251 610102 961955 210284522 % 18 % 8 % 2 %

1 % !!!

0

1000000

2000000

3000000

num

ber

of e

vent

s m

axED

R

altitude range in meter [m]

event (maxEDR) frequency share for altitude range , winter half year 10/11,>FL210, 4.140.992 events in all

>0.05 m⅔/s MOG

all events

FIG. 14. Vertical profile of the yearly averaged MOG/total divided by the globally averaged MOG/total background value of 0.32 stratified by in

cloud and clear air as well as by season (October–March, dashed lines; April–September, thin solid lines;yearly average, thick solid lines).

Climatology of Upper-Level Turbulence over the Contiguous United States J. K. WOLFF , R. D. SHARMAN, Journ.Appl. Meteor. and Clim.,Vol 47,2008

10-20-times more likelyas near core of jet and upper shear zone , of prime importance

Share of the 5 % (events from in all) moderate or greater turbulence above 6400 m

PIREP database

CONCQ _ SSOCQ _

Turbulence in clouds (convection, CIT) and near lower shear zone of jet


statistical evaluation and 5% MOG/ 95% NoTurbturbulence events (MOG) can’t bring a return with this distribution training data set - large number of examples from each class no change of class distribution in the validation datarandom-sampling to decrease NoTurb-examples


0

0,2

0,4

0,6

0,8

1

1,2

1,4

4\96 18\82 27\73 31\69 39\61 42\58 50\50

fore

cast

qua

lity

distribution MOG/NOTURB in percent [ %]

distribution and TSS,BIAS for linear regressionmaxEDR vs. edp_mos=Co+C1*edp*√ᵨ/ᵨ, ,

~4million events => 485.500 (40%\60%)

TSS

1/BIAS

sensitivity-test for several distributionsbest arrangement 40%MOG/60%NoTurbuse only for model output statistic confirmed

“I have used the 40% MOG (60% null) distribution because it worked well withall the machine learning algorithms”

Dissertation : A Domain Analysis Approach to Clear-AirTurbulence Forecasting Using High-Density In-situMeasurements by Jenny A. AbernethyM.S., University of Colorado, 2004



-model output statistic => classical linear regression -is used to relate the response variable (predictand,Y=maxEDR) to the explanatory variables (predictors Xi = p,v,w,tke,.. eld,div,dsh,ri,Qc…)-Maximum likelihood estimates of β1…βp are founded by least squares fitting

edp

initial situation maxEDR/edp>6400m,485513 events(40/60)

after linear regression step 1edp_mos=β0+β1*DEN

after linear regression step 2edp_mos=β0+β1*DEN+β2*edp



Simply density is a significant predictor for maxEDR , ?

-edp from COSMO only kinematic variable

-maxEDR use required input TAS ( )

-use density adjustment for model output

because turbulence is higher for more dense air mass

-need further source term for tke-equation involved density Qc_? = Qc_? f(den), gravity wave

-maxEDR data seem to be biased by flight activities (influence of aircrafts ahead, higher for low level flights)

TASCAS

0

o


3. comparison of EDP and EDR validation of full set of dataabout 4 million couples of measurement/model above FL210. Receiver Operating Characteristicobjective from Turbulence Joint Safety Implementation Team (TJSIT)

III: numerical tests, idea additional source term depend on densityIV: MOS with 1 step (predictor) (ϱ/ϱ₀)³edp evsep=0.12 ( no-turb/yes-turb for III,IV)

Turbulence forecast for aviation WV2p Axel Barleben Boulder,workshop,NCAR 27.08.2013

4. outlook / abstract

Comparison EDR and EDP

-useful for verification of turbulence forecast (edp) with real measurement (edr)-appropriate for optimization of turbulence parameterization (scaling factor for QC_SHS, tubulent length scale, density-adjustment)-MOS (no improvement after step 2) needs other or further indices (or combinations) -edp ( tke-equation) because additive correction require further source-terms (advection)-Qc_con only add in case of CIT, “limiting value” ? -FAA-EDR standards, expanded to Europe , understanding (density)-turbulence scheme in global model ICON (20 km,60-90 levels) ~ 2014-thresholds edp ? tendency : light events to strong, severe events to weak-EDP reproduce EDR inexactly-aviation forecaster (DWD) apply EDP and Ellrod-index for turbulence prediction (example 14.2.2013)


4. outlook / abstract

Figure: forecast EDP COSMO-EU , Maximum level 10-19 (FL180-360), 14.02.2013 06 UTC (method : sum up different levels “3d”) several PIREPS reported mod to sev along 10 degree of longitude

ARS: MODERATE TO SEVERE TURB OBS FL180-240, FL360, 6-8UTC

COSMO-EU forecast of eddy dissipation parameter


3. comparison of EDP and EDR third dimension animation of ELD > 6.0 10**-7 s**-2 (sev turb)14.2.2013, 06 UTC

Turbulence forecast for aviation WV2 Axel Barleben Boulder,workshop,NCAR 27.08.2013 Turbulence (Eddy Dissipation Rate) forecast based on COSMO-EU 1.introduction/motivation.

Documents

edr turbulence forecast

pilots turbulence forecast

outlookturbulence forecast

civil aviation

mod turb

sev turb edr

edr edp

aviation weather watch