implementation of fdda into wrf · • wrf fdda results show very good agreement with target...

Implementation of FDDA into WRF

David R. Stauffer, Penn State University, University Park, PA; and A. Deng, T. L. Otte, J. Dudhia, A. M. Gibbs, and

G. K. Hunter

But First… An Update on Some But First… An Update on Some Penn State MM5 Nudging Penn State MM5 Nudging Applications…Applications…

MM5 Realtime Nowcast PredictionMM5 Realtime Nowcast Prediction(Army Profiler)(Army Profiler)

USAFAS

Rapidly Relocatable NowcastRapidly Relocatable Nowcast--Prediction System Prediction System (RRNPS) for US Army: MMS(RRNPS) for US Army: MMS--ProfilerProfiler

•• MM5MM5--based system run locally on the based system run locally on the battlefield to provide battlefield to provide WarfighterWarfighter MET.MET.

•• Fielded to active Army units (early Fielded to active Army units (early 2005)2005)

•• Approved for full rate production Approved for full rate production (June 2005)(June 2005)

•• Provides Local and Target Area MET Provides Local and Target Area MET

•• Provides MET Messages every 15Provides MET Messages every 15--30 30 minutesminutes

•• Provides as a minimum MET Provides as a minimum MET Parameters Parameters

–– Temperature, HumidityTemperature, Humidity

–– Pressure Pressure

–– Wind Speed, Wind DirectionWind Speed, Wind Direction

–– Target Area: Ceiling, Visibility, Target Area: Ceiling, Visibility, PrecipPrecip Rate, Rate, PrecipPrecip TypeType

Smiths Detection and Penn State UniversitySmiths Detection and Penn State University

2h1h

36km

12km

MM5 Nowcast

OBS Nudging

4km

0

The model is run in halfThe model is run in half--hour increments, continuously assimilating observations hour increments, continuously assimilating observations and staying just ahead of the clock to and staying just ahead of the clock to ““predictpredict”” current conditionscurrent conditions……

MM5 Nowcast System for Army MM5 Nowcast System for Army MMSMMS--ProfilerProfiler

Profiler in BattlefieldProfiler in Battlefield

Worldwide Applications…Worldwide Applications…

Profiler Surface Wind on the 4-km GridProfiler Surface Wind on the 4Profiler Surface Wind on the 4--km Gridkm Grid

Penn State Numerical Weather Prediction Penn State Numerical Weather Prediction for U. S. Marines METMFfor U. S. Marines METMF--R NEXGEN…R NEXGEN…

METMFMETMF--R Nowcast InitializationR Nowcast Initialization

6h 12h 18h 24h0h

36km

12km

4km

MM5 Nowcast Mode

(Profiler)

MM5 Forecast Mode

Lagged 4km orOBS

Penn StatePenn State--DTRA Winter Olympics Support…DTRA Winter Olympics Support…

PSU 1.3km Feb. 22/13-17 UTC plumes, Feb. 22/14 UTC winds (Feb. 22/00 UTC model)

MM5/WRF 4MM5/WRF 4--Domain Configuration for Domain Configuration for 2006 Winter Olympics2006 Winter Olympics

MM5 Initialization Scheme used MM5 Initialization Scheme used for 2006 Winter Olympicsfor 2006 Winter Olympics

6h 12h 18h 24h0h

36km

12km

MM5 Dynamic Initialization

MM5 Forecast

1.33km

OBS Nudging

4km

Analysis Nudging-12h -6h

Observed and Model Cloud (t=0h) Observed and Model Cloud (t=0h) 1200UTC February 18, 20061200UTC February 18, 2006

18-h Forecast of Surface Layer Winds with MesonetObservations (red) on 1.3 km Domain Valid at

18 UTC, 21 February 2006

HistoricalHistorical--Mode (Episodic) Mode (Episodic) Applications…Applications…

Improved Meteorology for Input to AirImproved Meteorology for Input to Air--Chemistry Chemistry ModelsModels

Breton National Wilderness Breton National Wilderness Area:AirArea:Air Quality Quality Monitoring Sites for YearMonitoring Sites for Year--Long Modeling StudyLong Modeling Study(M. S. student (M. S. student -- J. J. ZielonkaZielonka))

WDP

FTMBIP

Time Series Analysis:Time Series Analysis:Neil Wheeler, Sonoma Tech. Inc.Neil Wheeler, Sonoma Tech. Inc.

µg/m3

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

WSP

D (m

/s)

Mean Absolute Error Seasonal Means for WSPD

Winter 1.7 1.4 1.3

Spring 1.7 1.1 0.9Summer 1.4 1.0 0.8

Fall 1.6 1.2 1.0

SFC 30-1000m ALAS

0.0

5.0

10.0

15.0

20.0

25.0

WD

IR (d

egre

es)

Mean Absolute Error Seasonal Means for WDIR

Winter 20.3 16.0 6.3

Spring 18.6 14.7 7.6Summer 24.5 17.1 11.1

Fall 20.9 16.4 7.5

SFC 30-1000m ALAS

0.0

5.0

10.0

15.0

20.0

25.0

Mea

n A

bsol

ute

Erro

r

Mean Absolute Error - Annual Means

SFC 2.1 1.9 1.5 1.6 20.930-1000m 1.8 1.1 0.8 1.2 16.0

ALAS 1.6 0.5 0.3 1.0 8.2

VWD (m/s) TEMP (°C) MIXR (g/kg) WSPD (m/s) WDIR (degrees)

WRF Model andWRF Model and

Nudging MethodologyNudging Methodology

, dh dhtd dhs dht

d

p p p pη μμ−

= = −

Conserved state variables:

Hydrostatic pressure coordinate:

Dry hydrostatic pressure dhp

( , , ), , , d d d m d mV v U V W Q qμ μ η μ θ μ= = Ω = Θ = =r rr

&

.

( ) ( ) , ( ) ( )

( ) [ ] , ( )

1 [( ) ] 0, ( )

( ) 0

d d

m

d U d V

d Wd

mm Q

d

d

U p p V p pVu F Vv Ft x x t y y

W pVw g F V Ft t

QV gW V q Ft t

Vt

α αη ηα α

η η

η η

η

φ φμ α μ αη η

α μ θα η

φ φμ

μ

Θ

∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂+ ∇⋅ + + = + ∇⋅ + + =

∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂

∂ ∂ ∂Θ+ ∇⋅ − − = + ∇⋅ =

∂ ∂ ∂

∂∂+ ⋅∇ − = + ⋅∇ =

∂ ∂

∂+ ∇⋅ =

∂

r r

r r

r r

r

Moist Equations:

where

.

( ) ,Ua Va aVax yη η

∂ ∂ ∂Ω∇⋅ ≡ + +

∂ ∂ ∂

r ( ) a a aV a U Vx yη η∂ ∂ ∂

⋅∇ ≡ + +Ω∂ ∂ ∂

r

00

1

( )

(1 ...)

(1 ) (1 1.61 )

pv

d d

ccd m

d

d v c r i

vm v v

d

Rp p p

q q q qRq qR

φ α μη

θα

α α

θ θ θ

−

∂= −

∂

=

= + + + + +

= + ≈ +

.

Diagnostic Equations:

where

Nudging FDDA

Note: Nudging FDDA on staggered grid while physics on non-staggered grid…..

Nudging FDDA

00

( ) ...

( ) ( ) ( )

( ) ( )1 (1 )

ob

Gtob ob

Gto ob ob

t

Gt

dt t e

t e

e folding time O hG

α α α

α α α α

α α α α

−

−

∂= • − +

∂

⇒ = − • +

= − • +

≡ −

∫

Nudging FDDA in WRF

...

... ( )

( )

ob

ob

xy t

For

t t t

G Wt

G WwhereW w w w

θ

μ

η

α μ θθ μμ θ

μ θ θ

θ μ μ

= Θ = •∂Θ ∂ ∂

= + +∂ ∂ ∂∂Θ

= + • • • −∂

+ • • • −

= • •

Experimental Design / Experimental Design / SetSet--UpUp

Demonstration of WRF Demonstration of WRF Analysis Nudging:Analysis Nudging:•• Assimilate 12Assimilate 12--hourly MM5 fronthourly MM5 front--end 3D end 3D

analyses on 108analyses on 108--km, 36km, 36--km, 12km, 12--km gridskm grids•• Nudge u, v, theta (G = 3 X 10 Nudge u, v, theta (G = 3 X 10 --44 ss--11) ) •• Nudge Nudge qqvv (G = 1 X 10 (G = 1 X 10 --55 ss--11))•• Show results with and without FDDA at 48 Show results with and without FDDA at 48

h for CAPTEX83 (18h for CAPTEX83 (18--20 September 1983) 20 September 1983) •• Compare results subjectively to the Compare results subjectively to the

analyses and statistically to observationsanalyses and statistically to observations•• Test noTest no--nudging in the PBL optionnudging in the PBL option

3D Analyses from MM5 3D Analyses from MM5 FrontendFrontend……

•• ““mm5_p_convert” (written by mm5_p_convert” (written by WeiWei Wang/NCAR)Wang/NCAR)

•• Converts multiConverts multi--time pressuretime pressure--level data from MM5 level data from MM5 FrontendFrontend (REGRID, (REGRID, RAWINS, LITTLE_R) into HINTERP fieldsRAWINS, LITTLE_R) into HINTERP fields

–– Introduces MM5 Introduces MM5 frontendfrontend data into WRFSIdata into WRFSI--formatted outputformatted output–– Enables common initial conditions and lateral boundary conditionEnables common initial conditions and lateral boundary conditions for s for

“fair” “fair” comparsioncomparsion of MM5 and WRF simulations from fields based on of MM5 and WRF simulations from fields based on the same surface and pressurethe same surface and pressure--level analyses. (Note that there are still level analyses. (Note that there are still differences in the state variables and horizontal / vertical gridifferences in the state variables and horizontal / vertical grid d structures between the two models)structures between the two models)

–– Enables WRF analysis nudging towards Enables WRF analysis nudging towards frontendfrontend MM5 system (MM5 system (RawinsRawins) ) analysesanalyses

•• Passes through to WRF common domain definitions and underlying sPasses through to WRF common domain definitions and underlying static tatic fields (lat, fields (lat, lonlon, terrain, land use, etc.), terrain, land use, etc.)

•• Requires special version of VINTERP to create input for REALRequires special version of VINTERP to create input for REAL•• Note: New WRF Preprocessing System (WPS) code replacing SI in WRNote: New WRF Preprocessing System (WPS) code replacing SI in WRF 2.2 F 2.2

in September 2006 will have VINTERP in REAL…in September 2006 will have VINTERP in REAL…

SfcSfc Analyses from MM5 Analyses from MM5 FrontendFrontend……

•• ““wrfsfcfddawrfsfcfdda” (written by Tanya ” (written by Tanya OtteOtte /NOAA/ARL)/NOAA/ARL)

•• Converts RAWINS/LITTLE_R “SFCFDDA” files to WRF I/O Converts RAWINS/LITTLE_R “SFCFDDA” files to WRF I/O APIAPI

•• Includes all SFCFDDA 2D fields, plus UIncludes all SFCFDDA 2D fields, plus U-- and Vand V--component component winds on C grid, PSFC, THETAwinds on C grid, PSFC, THETA

•• Files will be used for surface analysis nudging in WRFFiles will be used for surface analysis nudging in WRF

•• Files will also support soil moisture nudging option in Files will also support soil moisture nudging option in PleimPleim--XiuXiu LSMLSM

Preliminary Results:Preliminary Results:

108108--km gridkm grid

Surface Surface Temperature and Temperature and Sea Level PressureSea Level Pressure

MM5 ANALYSIS WRF FDDA

WRF NO FDDA

850 850 hPahPaTemperature and Temperature and GeopotentialGeopotential HeightHeight


WRF NO FDDA



WRF NO FDDA

Surface Winds and Surface Winds and Sea Level PressureSea Level Pressure


WRF NO FDDA

850 850 hPahPa Winds and Winds and GeopotentialGeopotential HeightHeight


WRF NO FDDA



WRF NO FDDA

Precipitation and Sea Level PressurePrecipitation and Sea Level Pressure

WRF NO FDDA WRF FDDA

Preliminary Results:Preliminary Results:

3636--km gridkm grid



WRF NO FDDA


MM5 ANALYSIS WRF FDDA above PBL

WRF NO FDDA



WRF NO FDDA


WRF NO FDDA WRF FDDA


WRF NO FDDA WRF FDDA above PBL

MAE of Temperature (C) for 108-km, 36-km, 12-km domains averaged over 48-h period

MAE of Wind Speed (m/s) for 108-km, 36-km, 12-km domains averaged over 48-h period

MAE of Wind Direction (deg) for 108-km, 36-km, 12-km domains averaged over 48-h period

•• WRF FDDA results show very good WRF FDDA results show very good agreement with target analysisagreement with target analysis--nudging nudging fieldsfields

•• New “converter code” used to create WRF New “converter code” used to create WRF initial conditions from MM5 initial initial conditions from MM5 initial conditions is also attractive for translating conditions is also attractive for translating frontfront--end MM5 pressureend MM5 pressure--level gridded level gridded analyses (from analyses (from RawinsRawins or or Little_rLittle_r) to WRF ) to WRF infrastructure for analysis nudginginfrastructure for analysis nudging

•• WRF fields are meteorologically WRF fields are meteorologically reasonable and show no signs of serious reasonable and show no signs of serious noise problemsnoise problems

•• Optional use of PBL height for vertical Optional use of PBL height for vertical weighting produces expected resultsweighting produces expected results

DISCUSSION

•• Statistics show reasonable, reduced error Statistics show reasonable, reduced error levels with nudging for all experimentslevels with nudging for all experiments

•• WRF 3D analysis nudging FDDA works for WRF 3D analysis nudging FDDA works for nestingnesting

•• WRF 3D analysis nudging FDDA can also be WRF 3D analysis nudging FDDA can also be excluded from modelexcluded from model--predicted PBLpredicted PBL

•• Currently testing restart capabilityCurrently testing restart capability•• Currently works in OMP. MPI testing to be Currently works in OMP. MPI testing to be

done.done.

DISCUSSION (cont’d)

SummarySummaryAnalysis Nudging FDDA in WRF – ARW:

Current capabilities:

3D analysis nudging of gridded fields (u, v, θ, qv):

Analyses on WRF model surfaces (η)

• External gridded analyses processed through Real• Real input files created in “converter code” from Rawins

surface and pressure-level gridded analyses (MM5 frontend).

• Real also processes other gridded fields (SI, 3DVAR)

• (Standard version of Real had to be modified to output more than first time period…)

Uses as default linear temporal interpolations and uniform spatial weighting functions

Goal: Basic 3D analysis nudging with optional vertical weighting function based on PBL height in WRF 2.2 in September 2006*…

*Implications of replacing SI with new WRF Preprocessing System (WPS) on schedule?

SummarySummaryFuture capabilities with additional funding support:Future capabilities with additional funding support:

More general temporal and spatial weighting functions, includingMore general temporal and spatial weighting functions, including use use of observation data densities, dynamicof observation data densities, dynamic--initialization ramping initialization ramping functions, etc.functions, etc.

Surface analysis nudging using higher temporal resolution surfacSurface analysis nudging using higher temporal resolution surfacee--gridded fields, applied at surface and throughout PBL (Stauffer gridded fields, applied at surface and throughout PBL (Stauffer et al. et al. 1991).1991).

•• Requires using “converter” on Requires using “converter” on RawinsRawins sfcfddasfcfdda files since files since 3DVAR cannot currently produce surface3DVAR cannot currently produce surface--only analyses at only analyses at intermediate times…intermediate times…

•• Enables Enables PleimPleim--XuXu LSM soil moisture nudging capabilityLSM soil moisture nudging capability

•• List of nudging variables to be expanded (e.g., List of nudging variables to be expanded (e.g., µµ, vertical motion, , vertical motion, cloud water, etc.)cloud water, etc.)

•• Hybrid 3DVAR/EnKF/nudging methods (with Li Li Lei, Ph.D. studenHybrid 3DVAR/EnKF/nudging methods (with Li Li Lei, Ph.D. student)t)

•• WRF WRF obsobs nudging being coded by NCAR RAL nudging being coded by NCAR RAL –– possible collaboration possible collaboration with Penn State in the future…with Penn State in the future…

AcknowledgmentsAcknowledgments

•• This work has been funded in part by This work has been funded in part by U. S. Environmental Protection U. S. Environmental Protection Agency via Kenneth Agency via Kenneth SchereSchere

•• WeiWei Wang (NCAR) and Annette Gibbs Wang (NCAR) and Annette Gibbs and Glenn Hunter (Penn State) have and Glenn Hunter (Penn State) have also contributed time and supporting also contributed time and supporting software…software…

DisclaimerDisclaimer

"The research presented here was performed, in "The research presented here was performed, in part, under the Memorandum of Understanding part, under the Memorandum of Understanding between the U.S. Environmental Protectionbetween the U.S. Environmental ProtectionAgency (EPA) and the U.S. Department of Agency (EPA) and the U.S. Department of Commerce National Oceanic and Atmospheric Commerce National Oceanic and Atmospheric Administration (NOAA) and under agreement Administration (NOAA) and under agreement number DW13921548. Although it has been number DW13921548. Although it has been reviewed by EPA and NOAA and approved for reviewed by EPA and NOAA and approved for publication, it does not necessarily reflect their publication, it does not necessarily reflect their views or policies."views or policies."

implementation of fdda into wrf · • wrf fdda results show very good agreement with target...

Documents