aladin consortium activities in data assimilation

Zürich 9th-12th, 2006 EWGLAM/SRNWP meetings

Aladin consortium activities in Aladin consortium activities in data assimilationdata assimilation

Claude Fischer & Andras HoranyiClaude Fischer & Andras Horanyi

Stolen material from: Fatima Karbou, Pierre Stolen material from: Fatima Karbou, Pierre Brousseau, Loïk Berre, Olivier Caumont, Zahra Brousseau, Loïk Berre, Olivier Caumont, Zahra Sahlaoui, Simona Stefanescu, Alex DeckmynSahlaoui, Simona Stefanescu, Alex Deckmyn


Overview Overview

1.1. Mainframe 3D-VAR applicationMainframe 3D-VAR application

2.2. Background error statistics: « B » Background error statistics: « B » matrixmatrix

3.3. Observations and OSEObservations and OSE

4.4. Arome prototypeArome prototype

5.5. Surface assimilationSurface assimilation

6.6. Plans and collaborationsPlans and collaborations


Mainframe 3D-VAR: example of Mainframe 3D-VAR: example of Aladin-FranceAladin-France

► Incremental 3D-VARIncremental 3D-VAR► Continuous assimilation cycle, 6 hour frequency, long cut-off Continuous assimilation cycle, 6 hour frequency, long cut-off

assimilation cycle and short cut-off production, coupled with assimilation cycle and short cut-off production, coupled with Arpège, Analysis=Model gridmeshArpège, Analysis=Model gridmesh=9.5km=9.5km

► Observations:Observations: Surface pressure, SHIP winds, Surface pressure, SHIP winds, synop T2m and RH2msynop T2m and RH2m Aircraft dataAircraft data SATOB motion windsSATOB motion winds Drifting buoysDrifting buoys Soundings (TEMP, PILOT)Soundings (TEMP, PILOT) Satellite radiances: AMSU-A, AMSU-B, HIRS, Satellite radiances: AMSU-A, AMSU-B, HIRS, Meteosat-8 SEVIRIMeteosat-8 SEVIRI QuikSCAT windsQuikSCAT winds Ground-based GPS zenital delaysGround-based GPS zenital delays

► Digital filter initialisation (non incremental)Digital filter initialisation (non incremental)


NON LINEAR AND OMEGANON LINEAR AND OMEGA BALANCESBALANCESIN THE ALADIN Jb (G. Faure, L. Berre)IN THE ALADIN Jb (G. Faure, L. Berre)

Qp

f

VfkVVPbal

2)(

)..(

2

22

0


Wind 850 hPa

+12h

~ 10 cases slightly >0

10 cases ~ neutral

1 case slighlty <0

NON LINEAR AND OMEGANON LINEAR AND OMEGA BALANCESBALANCESIN THE ALADIN Jb (G. Faure, L. Berre)IN THE ALADIN Jb (G. Faure, L. Berre)


Complex wavelets for LAM (A. Complex wavelets for LAM (A. Deckmyn)Deckmyn)

► Wavelets are (partially) localised in both grid point space and Wavelets are (partially) localised in both grid point space and Fourier space.Fourier space.

► Diagonalization of B in wavelet space can reproduce local variations Diagonalization of B in wavelet space can reproduce local variations in the structure functions and standard deviations.in the structure functions and standard deviations.

► Current work is focussing on reproducing 3D structure functions for Current work is focussing on reproducing 3D structure functions for different variables (Alex Deckmyn, Tomas Landelius, Loik Berre)different variables (Alex Deckmyn, Tomas Landelius, Loik Berre)

Standard deviations of Temperature errorat the lowest vertical level: from data (left)and from wavelet-B (right).


Daily small-ensemble Daily small-ensemble b’s and their filteringb’s and their filtering

b ENS 2 « RAW » b ENS 2 « FILTERED »

b ENS 1 « FILTERED »b ENS 1 « RAW »


Towards the assimilation of microwave surface sensitive Towards the assimilation of microwave surface sensitive channels from AMSU-A, AMSU-B & SSM/I over landchannels from AMSU-A, AMSU-B & SSM/I over land

F. KARBOU, F. KARBOU, EE. G. GÉÉRARD, F. RABIER, E. BAZILE, Z. RARD, F. RABIER, E. BAZILE, Z. SAHLAOUISAHLAOUI

Over land:

Only microwave channels that are the least sensitive to the surface are assimilated

Remaining uncertainties about land emissivity and skin temperature

Existing emissivity models : facilitate the assimilation of channels that receive a contribution from surface

The models need accurate input parameters hardly available at global scale

Need for alternatives to estimate the land emissivity New land surface emissivity parameterisations at Météo-France


The parameterisations (Karbou et al. 2006*):

1) Dynamically varying emissivities derived from only one channel for each instrument

2) Averaged emissivities over 2 weeks prior to the assimilation period

3) Averaged emissivities + dynamically adjusted skin temperature from one channel for each instrumentThis talk: Results from the dynamical approach (1)

* Karbou, F., E. Gérard, and F. Rabier, 2006, Microwave Land Emissivity and Skin Temperature for AMSU-A & -B Assimilation Over Land, Q. J. R. Meteorol. Soc. To appear.


RMS of error, AMSU-A, Glob, Period: 01 to 19 July 2005

Day of July 2005 Day of July 2005

Results: Results: Observation departures from first guess: Good impact Observation departures from first guess: Good impact on the on the observation operator observation operator performancesperformances


More low clouds in EXP

More Stratocumulus Less precipitating clouds

Less low clouds in EXP

Low cloudiness difference field EXP-CTR (from 31 24-hour forecasts 15 Aug-14 Sept

2005)

Results: Results: Impact on the analysis fields: Large impact on the Impact on the analysis fields: Large impact on the humidity fieldhumidity field

Results are given for AMSU-A & AMSU-B assimilation over land

Similar results are obtained with SSM/I over land

Great impact on the humidity and precipitation for both AMSU and SSM/I


Results: Results: Impact on the forecast skills: Positive impact on Impact on the forecast skills: Positive impact on humidity, humidity, temperature, temperature, geopotential heightgeopotential heightForecast scores / radiosondes (forecast-obs), SSM/I over land, 31 Forecast scores / radiosondes (forecast-obs), SSM/I over land, 31

cases, 15Aug-14Sep 2005cases, 15Aug-14Sep 2005analysis 24h forecast 48h forecast 72h forecast

Temperature

Geopotential h.


3D-VAR data assimilation in 3D-VAR data assimilation in MoroccoMorocco

►Aladin cycle : cy25t1Aladin cycle : cy25t1►Assimilation method :3d-var for upper Assimilation method :3d-var for upper

air fields / Canari O.I. for the surfaceair fields / Canari O.I. for the surface►Assimilation cycle : every 6 hoursAssimilation cycle : every 6 hours►Assimilation window :4,5h for short Assimilation window :4,5h for short

cutt off and 6h for long cutt offcutt off and 6h for long cutt off► Jb : standard NMCJb : standard NMC►Data used : conventional observationsData used : conventional observations


-0,15

-0,1

-0,05

0

0,05

0,1

0,15

0,2

0,25

01/11/2005 03/11/2005 05/11/2005 07/11/2005 09/11/2005

3D-VAR ADAP DYN

0

0,5

1

1,5

2

2,5

3

3,5

01/11/2005 03/11/2005 05/11/2005 07/11/2005 09/11/2005

3D-VAR ADAP DYN

Temperature 12h forecast against radiosondes (top:rms , bottom:bias)

Scores of temperature field (12h forecast) against radiosonde observations show :

Slightly positive impact for bias

Slightly negative impact on rms

Humidity at 900hPa versus ARPEGE analysis (red: dynamical adaptation & blue: 3DVAR)

The bias of humidity against ARPEGE analysis is reduced in low atmospheric levels (900 hPa) for all forecast ranges


Impact study of ATOVS raw radiancesImpact study of ATOVS raw radiances

Impact of AMSU-A is positive over areas with a lack of observations Impact of AMSU-A is positive over areas with a lack of observations (tropical band, Sahara, ocean)(tropical band, Sahara, ocean)

In mid latitudes, the impact of AMSU-A raw radiances is less marked In mid latitudes, the impact of AMSU-A raw radiances is less marked (conflict between several observation types). More sensitivity (conflict between several observation types). More sensitivity studies are needed to optimise the use of raw radiancesstudies are needed to optimise the use of raw radiances

P30 humidity field at 850hPa level


1D+3DVar assimilation scheme for 1D+3DVar assimilation scheme for reflectivitiesreflectivities


OSSE for reflectivities w/ Meso-NH/AladinOSSE for reflectivities w/ Meso-NH/Aladin

Simulated obs. with first guess (12 h MNH fcst)

Z (

dBZ

) @

20U

TC

Z (

dBZ

) @

18U

TC

Z (

dBZ

) @

19U

TC

Hu

(%)

+ 1

0-dB

Z c

onto

ur @

18U

TC

Simulated observations with « microphysical initialization »

Simulated obs. from 1D retrieval of (T,RH) + 3DVar


Summary & outlook for Summary & outlook for reflectivitiesreflectivities

Summary:Summary:► 1D retrieval able to correct for humidity,1D retrieval able to correct for humidity,► Quality control of data and obs. op. evaluation (M. Quality control of data and obs. op. evaluation (M.

Jurašek, LACE)Jurašek, LACE)► Implementation in Arome well advancedImplementation in Arome well advancedFuture work:Future work:► More 1D + 3D-VAR high resolution casesMore 1D + 3D-VAR high resolution cases► Perform cycling experiments,Perform cycling experiments,Collaborations:Collaborations:► Introduce beam blockage in obs. op. (G. Haase, Hirlam)Introduce beam blockage in obs. op. (G. Haase, Hirlam)


Arome prototype: a RUCArome prototype: a RUC

E : first guess Px : prévision échéance x

S0 : Arome surface file at t0 A : analyse

Cx : fichier de couplage à tx


Distribution of observations along Distribution of observations along the RUC (12 UTC - 0 UTC)the RUC (12 UTC - 0 UTC)


Cumul RR sur 3H : 00h-Cumul RR sur 3H : 00h-03h 03h

DyA RUC 1h

RUC 3h

00-03h

50 mm/3h

1 analysis


DyA RUC 1h

RUC 3h138 mm

03-06h

1 analysis


Surface assimilationSurface assimilation► O.I. scheme (CANARI) in Arpège and Aladin (CZ O.I. scheme (CANARI) in Arpège and Aladin (CZ

and MO)and MO)► Probably also tested soon in Aladin-FRProbably also tested soon in Aladin-FR► Same algorithm to be ported into the externalized Same algorithm to be ported into the externalized

surface scheme (SURFEX) ?surface scheme (SURFEX) ?► New system for 2D analysis (maybe soil, probably New system for 2D analysis (maybe soil, probably

PBL fields and spatialisation tool)PBL fields and spatialisation tool)► In FR: possibility to initialize fields from the coupled In FR: possibility to initialize fields from the coupled

soil/surface/hydrology system (SAFRAN)soil/surface/hydrology system (SAFRAN)► To be further discussed between Aladin and Hirlam To be further discussed between Aladin and Hirlam

partners at the next common workshop on surface partners at the next common workshop on surface (Toulouse, Dec. 11-13th)(Toulouse, Dec. 11-13th)


Plans for 2007 and outlookPlans for 2007 and outlook► B matrix and background std dev.:B matrix and background std dev.:

Errors of the day for screeningErrors of the day for screening Gridpoint maps of Gridpoint maps of b’s for minimizationb’s for minimization Filtering of the ensemble bg errors (low-pass)Filtering of the ensemble bg errors (low-pass) Native Aladin ensemble of analysesNative Aladin ensemble of analyses Derive from those an Arome ensemble forecastDerive from those an Arome ensemble forecast Wavelets (A. Deckmyn, Belgium; T. Landelius, Hirlam)Wavelets (A. Deckmyn, Belgium; T. Landelius, Hirlam)

► Algorithms:Algorithms: FGAT (with Hungary - LACE)FGAT (with Hungary - LACE) SL TL/AD and 4D-VAR in a nutshell (F. Vaňa, LACE)SL TL/AD and 4D-VAR in a nutshell (F. Vaňa, LACE) Incremental digital filtersIncremental digital filters Some kick-off on a simplified microphysics scheme for the mesoscale ?Some kick-off on a simplified microphysics scheme for the mesoscale ? Towards an integrated « Ensemble/Variational » data assimilation Towards an integrated « Ensemble/Variational » data assimilation

system; ETKFsystem; ETKF


Plans for 2007 and outlookPlans for 2007 and outlook► Observations:Observations:

SEVIRI radiances in Hungary (LACE) and Morocco SEVIRI radiances in Hungary (LACE) and Morocco (non-LACE) ?(non-LACE) ?

SEVIRI/CSR monitoring and bias correction; surface SEVIRI/CSR monitoring and bias correction; surface emissivity for IR SEVIRI channels (M. Stengel, emissivity for IR SEVIRI channels (M. Stengel, Hirlam)Hirlam)

Metop sensorsMetop sensors Work on radial radar wind obs. operator (T. Work on radial radar wind obs. operator (T.

Montmerle and C. Facchani)Montmerle and C. Facchani) Continuation on radar reflectivity obs. op. & retrieval Continuation on radar reflectivity obs. op. & retrieval

methods (M. Jurašek, LACE; G. Haase, Hirlam)methods (M. Jurašek, LACE; G. Haase, Hirlam) Microphysics modules for HIRS/IASI (M. Stengel, Microphysics modules for HIRS/IASI (M. Stengel,

Hirlam)Hirlam)

Thank you for your Thank you for your attentionattention


AMSU-A and AMSU-B (F. Karbou):AMSU-A and AMSU-B (F. Karbou):

AMSU-A and AMSU-B: temperature and humidity soundings• polar orbiting satellites: NOAA (15, 16, 17, 18), AQUA, MetOp• 20 channels: 23 to 190 GHz• scanning angle from -58° to +58° with V and H polarisation

Tb contaminated by surface conditions:Information about surface T and

Height of maximum sensitivity(depends also on scan and humidity content)


4 different approaches are compared

Several assimilation experimentsOver periods in 2005

Control run: operational model

Emissivities: - Atlas 2000 - Atlas 2005

Emissivities: dynamical computation

Emissivities: Atlas 2000 + estimation of Ts

Best choice so far:

-Atlas : risk for bias

-Dynamical comp.: risk for noise

New configurations tested for the surface


Compute surface emissivity by inverting the RT equation:

)),((

),(),(),(),(

TTs

TTpTp

Atlas 2000 Monthly global maps at 23, 31, 50, 89 and 150 GHz for February 2000

Atlas 2005

Monthly global maps for nadir at 23, 31, 50, 89 GHz from 01 through 20 March 2005+ parametrisation of (Karbou 2005) to take into account the scanning angle

Dynamical comp.

Emissivity computed for each meteorological situation for channel 1 AMSU-A (23GHz) and AMSU-B (89 GHz). These emissivities are attributed to the other channels without extrapolation.

New configuration for the surface


Emissivity, AMSU-B, channel1, operational

Emissivity, AMSU-B, channel1, atlas2005


Comparison of obs-guess without bias correction, 22 March through 4 April 2005

150 GHz 183+/-1 GHz

183+/-3 GHz

183+/-7 GHz

aladin consortium activities in data assimilation

Documents

amsub assimilation

amsub ssmi

assimilation of channels

microwave land emissivity

b matrixobservations

diagonalization of b

sahlaouiover land

waveletb right