METEO-FRANCE 2000 status

 

1. Summary of Highlights

. new supercomputer : VPP5000 with 31 processors

. 4 runs a day in test mode

. 4DVAR assimilation for global stretched model ARPEGE

. Use NOAA15 ATOVS radiances (120km data) in ARPEGE assimilation .

2. Hardware used

- Information commutators on GTS are the TRANSMET computers (2 Sun Entreprise 3000, operating with OS Unix and RDBMS Oracle).

- the management of the forecasting system (control of the data in input of NWP models, postprocessing, production of charts with the NWP output) is made on a HP T600 computer running Oracle RDBMS, US-Navy originating NEONS meteorological data management system, and PV-WAVE graphical software; one HP D370 is used as file server, one HP C180 workstation is devoted to the system monitoring, which is based on DCE. The whole system (production machine + file server + monitoring workstation), called DIAPASON, is doubled for backup.

- NWP operational models are running on a FUJITSU VPP5000 (31 processors, 21 with 8 Gbytes memory each, the others with 4Gbytes memory each)

- Dissemination of forecast and observation products (from GTS included), in particular to the french weather stations, is made through the Eutelsat communication satellite (RETIM system).

3. Use of Data and Products from GTS

Average number of messages, by day:

AIREP

AMDAR

BATHY

BUOY

PILOT

SATEM

SATOB

SHIP

SYNOP

TEMP

TEMPSHIP

3600

16000

60

8500

1000

11000

400000

5200

52000

1250

15

ACARS/US

(A)TOVS120

ERS/URA

ERS/UWA

ERS/UWI

PROFILER

SATWIND

SSMI

70000

90000

800

1000

1000

1600

12000

45000

GRID from BRACKNELL : 3020

GRID from WASHINGTON : 800

GRIB aero 1.25 from BRACKNELL : 3360

GRIB 2.5 from BRACKNELL : 6060

GRIB ECMF : 335

Fac-simile products:

- aeronautical charts from Bracknell (T4 code)

- Cifax charts on almost all links

4. Data Input System

Automated.

5. Quality Control System

GTS data are controlled at several levels:

- transmission

- syntaxic coherence

- rudimentary control of likelihood: e.g. a sea level pressure value must be above 880 hPa and below 1080 hPa

- data control by comparison to adjacent (in time and/or space) data, or to different types of data at the same location: e.g. Td T is checked; In the same manner, a sudden slope breaking of in a temperature profile from a radiosonde far from the tropopause leads to the invalidation of the data.

Data to be reemitted on GTS are not modified.

6. Monitoring of the observation system

All the observations that are used by the NWP system (SYNOP, SHIP, BUOY, SATOB, (A)TOVS120, TEMP, PILOT, AIREP, AMDAR, ACARS) are controlled by comparison to the analyses and first guesses of the ARPEGE assimilation cycle: statistics are produced every month and summarized in a monthly bulletin.

7. Forecast System

The operational forecast system at Météo-France is based on two different numerical applications of the same code and an additional code to build the limited area model.

The ARPEGE-IFS library developed jointly by Météo-France and ECMWF (ARPEGE being the usual name in Toulouse and IFS the one used in Reading):

ECMWF model for medium range forecasts (4-7 days)

a variable mesh version run in Toulouse for short range predictions (1-4 days)

The ALADIN library developed jointly by Météo-France and the national meteorological or hydrometeorological services of the following countries: Austria, Belgium, Bulgaria, Croatia, Czech Republic, Hungary, Moldova, Morocco, Poland, Portugal, Romania, Slovakia, Slovenia.

7.1. Schedule of the Forecast System

The operational forecast system at Météo-France is based on ARPEGE/ALADIN, using the following rules :

. an assimilation (long cut-off) analysis is performed before each short cut-off analysis.

. the product’s availability is :

initialysed analysis (P0) cut-off+30'

ARPEGE forecast 20' every 24H range

ALADIN-France ARPEGE+10'

HH

0000 UTC

0600 UTC

1200 UTC

1800 UTC

long cut-off

0815 UTC

1250 UTC

2015 UTC

0050 UTC

short cut-off

1H50

3H

1H50

3H

ARPEGE range

96H

42H

72H

30H

end of ARPEGE

0340 UTC

1010 UTC

1520 UTC

2200 UTC

ALADIN range

48H

42H

36H

30H

end of ALADIN

0310 UTC

1020UTC

2300 UTC

2210UTC

7.2. Medium range (4-10 days) forecast system

As mentioned above, it is the operational T511 model of ECMWF and T255 Ensemble Prediction System for 4-5day and 6-7day forecast bulletins.

7.3. Short range forecast system

The ARPEGE system(0-96 hours)

ARPEGE-IFS is a common Météo-France / ECMWF development. ARPEGE is the french name (Action de Recherche Petite Echelle Grande Echelle) while IFS is the name used at Reading (Integrated Forecast System). It is a tunable system based on a global spectral model which can be used for several applications: data assimilation, short-range prediction, medium-range prediction, climate research, predictability studies.

ARPEGE-IFS uses Schmidt's transformation leading to variable mesh configurations, having a pole of maximum resolution and a resolution varying continuously from that pole to the antipode (Courtier and Geleyn 1988). T being the nominal truncation and C the "stretching factor", the local resolution of the model is T x C over the pole, and T / C at the antipode.

The present version is T199 C3.5 having its pole in France (46.5N,2.6E), leading to T696.5 over France (corresponding roughly to a horizontal resolution of the Gaussian grid of 20 km) and T57 over New Zealand (corresponding roughly to a horizontal resolution of 230 km).

The number of vertical levels is 31, with an increased density in the low atmosphere. The first level is at 5 hPa, and the lowest one at 19m above the ground.

Assimilation, objective analysis and initialization

The assimilation runs with a 6 hour cycle. The objective analysis is performed with a multi-incremental 4D variational scheme : i.e. the departure obs-guess is computed at full resolution (T199C3.5) whereas the analyzed structures are produced at "low" resolution, in 3 loops T42C1, T63C1, T95C1. It is therefore assumed that the small scales (not analyzed) are forced by the (analyzed) large scales in the subsequent forecast.

The analysis works in vorticity, unbalanced divergence/ temperature/surface pressure and specific humidity on model levels.

assimilated data: SYNOP, SHIP, BUOY, BATHY, TEMP, TEMPSHIP and PILOT (part A, B, C and D), AIREP, AMDAR, ACARS, SATOB, (A)TOVS120km with observation time in [H-3h,H+3h]

assimilation cycle: 6 hour cycle.

analysis method: Multivariate four dimensional variational analysis

analysed variables: Wind, temperature and specific humidity on model levels, plus surface pressure.

first guess: A 6-hour forecast of ARPEGE. By default a 12, 18 or 24-hour forecast.

cover: Global cover.

horizontal resolution: T42, T63, T95 Gaussian grids

vertical resolution: The analysis is done on the model levels (see below): 31 levels (hybrid vertical co-ordinate) from 19m up to 5 hPa.

initialization: Incremental digital filter initialization (ie filtering analysis increments fields) using a Dolph-Chebishev filter with a stop-band edge period of 5h and a backward-forward scheme and a weak DFI constraint in the variational cost function.

surface: -analysis of superficial and mean soil temperature (resp moisture) from forecast errors on 2m temperature (resp. relative humidity)

- small relaxation towards climatology for snow and mean soil temperature and moisture

Model

basis equations: Primitive equation system

independant variables: Both components of the horizontal wind, temperature, specific humidity and surface pressure.

dependant variables: Vertical velocity and density

numerical technique: Spectral 2TL semi-lagrangian model and temporal discretization using leap-frog semi-implicit scheme

integration domain: The whole earth (global model).

orography, gravity wave drag: The orography of this model is computed on the T199 C3.5 Gaussian grid (300x600 points) from GLOB95 30"+US NAVY 10' + NOAA 5' data using a variational technique that strongly reduces the noise associated to Gibbs waves (see Bouteloup, 1995). The gravity wave drag takes in account some anisotropy, blocking and mid-tropospheric effects.

horizontal diffusion: Implicit in spectral space and incorporating an orography dependent correction for temperature

vertical diffusion: Scheme linked to PBL (see next point)

planetary boundary layer: ECMWF method (Louis et al. 1981) with several enhancements in the stable case

resolution, time step: This version of the ARPEGE model has a triangular truncature T199 with a stretching factor C3.5. The resolution varies from T 696.5 over France (15 km equivalent mesh for a finite difference model) to T57 over New Zealand (200 km equivalent mesh); it has 31 vertical levels from 19m up to 5hPa, using the hybrid (s,p) co-ordinate from Simmons and Burridge (1981). The time step is 900 seconds.

earth surface: Fixed analyzed sea surface temperature and amount of sea -ice. An improved version of the ISBA (Interaction Soil Biosphere Atmosphere) scheme is used, including an explicit parameterization of soil freezing. Six prognostic variables are handled by ISBA: surface temperature, mean soil temperature, interception water content (water on the leaves), superficial soil water content (first centimeter), total liquid soil water content, total frozen soil water content. A very simple parameterization of snow cover is added. Soil characteristics (texture, depth) are point-dependent. Vegetation characteristics are point- and month-dependent.

radiation: Hypersimplified scheme at every time step (Ritter and Geleyn 1992)

convection: Mass flux scheme (Bougeault 1985) modified by Ivanovici and Geleyn.

humidity: Specific humidity is the variable: no storage of condensate; evaporation of falling rain; treatment of the ice-phase.

ALADIN (0-48hours)

ALADIN is a limited area version of ARPEGE-IFS. This implies that:

· ALADIN is spectral (like ARPEGE-IFS)

· As spectral-LAM it works on a biperiodic domain and uses bi-Fourier horizontal transforms

· Its physics and ARPEGE's one are identical

· It gets initial and lateral boundary conditions from ARPEGE

Up to now ALADIN is run in pure dynamical adaptation mode, i.e. without own data assimilation. The operational version is semi-lagrangian (usual time step 450.s), with elliptic truncation E99x99 on Lambert projection domain (54°95N/33°66S,-11°18W/19°64E), leading to an equivalent finite difference resolution of roughly 7.5km.

The vertical resolution is 31 levels, the same as operational ARPEGE ones. The digital filter initialization uses a Dolph-Chebishev filter with a stop-band edge period of 3h and a backward-forward scheme.

 

NWP Products

The above described numerical models feed a analysis and forecast database, having following characteristics:

different horizontal domains for different horizontal resolution (from the global domain with a 2.5° and 1.5° mesh to the "France" domain with a 0.1° mesh)

vertical levels are the standard pressure levels

independence, from the creating model, of the format of the database products.

The meteorological fields stored in this database are:

- at all levels: geopotential, temperature, humidity, wind (including vertical velocity)

- at screen level: pressure, temperature, humidity, heat and radiation fluxes, snow and water content

- at sea surface level: reduced pressure

- some data at particular levels: 500 hPa absolute vorticity, high medium and low cloudness, iso 0° and iso -10°, tropopause etc...

ARPEGE produces boundary conditions for the ALADIN applications run by LACE in Pragues, in Morocco, Romania, Poland, Portugal, while ALADIN-France provides boundary conditions for ALADIN-Belgique.

 

Operational use of NWP products

On screen (especially SYNERGIE workstation and Meteotel-PC software) or on paper, hundreds of charts...

7.4. Specialized forecasts

7.4.1 Local weather elements

Several kinds of forecasts are made by statistical adaptation of the NWP products from the above described models:

MOS method and Kalman filter based on ARPEGE model:

2 meter-temperature over 1137 stations in France, every 3 hours from 0 to 72 hour range, plus extreme values.

cloud cover over 169 stations in France, every 3 hours from 6 to 66 hour range at 00Z, and from 0 to 54 hour range at 12Z.

MOS method based on ARPEGE model:

wind over 136 stations in France, from D+1 to D+3 every 6 hours

MOS method and Kalman filter based on ECMWF model:

cloud cover over 169 stations in France, from D+1 to D+6

Perfect-Prog method and Kalman filter based on ECMWF model:

min-max daily temperature over 169 stations in France, for D (max) to D+7 (Min). A 3 hour time step temperature forecast is then obtained by superimposition of a diurnal cycle

Temperature, cloud cover and precipitations forecast over 210 towns in the world are also performed, using spatial interpolation and a Kalman filter (temperature and cloud cover), and Perfect-Prog (temperature over Europe), every 6 hours from 12 to 108 hour range.

7.4.2 Marine forecasts

Wave hindcast and forecasting system

Two models run operationally in France for determining the sea conditions:

A global wave model , computing the waves over all the oceans up to 72 hour forecast, from the wind outputs of large scale fields derived from ARPEGE.

Type: coupled discrete deep water

Integration domain: Global

Grid: regular grid; resolution: 1°

Frequency resolution: 12 frequency components, logarithmically spaced from 0.04 Hz to 0.3 Hz

Direction resolution: 18 equally-spaced direction components

Integration scheme: time step = 900s

Boundary forcing: winds at 10m level from ARPEGE, updated every 6 hours

Surface classification: sea ice deduced from ARPEGE SST

Assimilation: 4 analysis/day using significant wave heights from ERS2 altimeter

A regional model, forecasting the waves up 48 hours with 3 hour step, over the European Seas (Atlantic, Mediterrean , Baltic, North Sea, Black sea, ...) , from the wind outputs of small scale fields derived from ARPEGE.

Type: Coupled discrete shallow water

Domain: European Seas

Grid: regular grid; resolution: 0°25

Frequency resolution: 12 frequency components, logarithmically spaced from 0.04 Hz to 0.3 Hz

Direction resolution: 18 equally-spaced direction components

Timestep: 300s

Boundary forcing: winds at 10m level from ARPEGE, updated every 3 hours.

This models are available between 0330UTC et 0345UTC, on 00UTC run.

 

Operational simulations of the oceanic circulation in tropical Atlantic

The oceanic primitive equation model OPA7, developed by CNRS/LODYC, has been run operationally every month, using all the surface fluxes produced by the operational ARPEGE model. Its main characteristics are 17 horizontal levels in z coordinate with a realistic bathymetry, and a 1/3 degree horizontal resolution. Systematics comparisons have been performed with bathythermic observations sent through the GTS, and against sea surface temperatures from ERS data ( ATSR ).

 

Storm surge model

A high resolution limited area storm surge model designed fro tropical cyclones is used operationnaly in France and French overseas Islands: Guadeloupe, Martinique, St Martin, St Barthelemy, French Polynesia, Réunion, Mayotte and New Caledonia Islands.

 

Oil drift model

Three different versions of a limited area oil drift model have been implemented:

· a 5' x 5' grid mesh one, over the METAREA II for which Météo-France is responsible under WMO Marine Pollution Emergency Response Support System (MPERSS);

· another 5' x 5' grid mesh one, including the tidal forcing, over French coastal areas;

· a 1' x 1' or less grid mesh version over French overseas Islands.

 

Container drift model

Same areas as above

 

7.4.3 Pollutant transport and dispersion forecast

After the Chernobyl catastrophe on April 26th 1986, the French meteorological service, METEO-FRANCE, has developed a powerful model to forecast the movement of radioactive clouds at long distance range. Meteorological central service of METEO-FRANCE in Toulouse (SCEM) has been designated as a regional specialized meteorological centre (RSMC) with activity specialization on the provision of atmospheric transport model products for environmental emergency response. This provision can be related, but not restricted, to nuclear accident, or radiological emergencies, and plumes of volcanic ashes for ICCA.

For environmental emergency responses we now use two models based on the use of the NWP fields that are stored in our database, from ARPEGE and from ECMWF's model:

- calculation of trajectory forecast for neutrally buoyant air parcels,

- full transport/dispersion model (Atmospheric DIspersion Eulerian Model ,french acronym: MEDIA).

These model proved highly successful in the ATMES experiment (Atmospheric Transport Model Evaluation Study) for the international comparison of pollutant transport/dispersion models of the Chernobyl case and is used regularly in the framework of experiments within the CEA/IPSN (Nuclear Safety Institute) and the EDF (Electricity National Board) for French nuclear sites. In these cases the source of the release is well known and allows a simulation, the results of which depend only on observed and forecast meteorological conditions.

The operational organization of Météo-France, for facing such pollution accidents, is based on a special crisis meteorological cell (CMC) that studies the evolution of weather/pollution conditions and provides the delegated authorities of a requesting country with information about pollutant transport containing in particular the standard set of products as defined during the International Workshop held in Montreal. This cell can of course be activated at any time (day or night) and is placed under authority of the director of central service of operations.

 

7.4.4 Tropical cyclones forecast model

A specific version of ARPEGE has implemented over Indian Ocean, and sent to the SYNERGIE software in La Réunion Island/Saint Denis.

The model is the same as the previous one, but with a different pole (20S, 60E) of stretching and geometry T127C3.5L31 for the forecast (time step 1350s) , and T95C1L31 for the 3DVAR analysis.

The model is running once a day based on 00UTC, up to 96 h , with a 9 hour cut-off.

assimilation cut-off

HH

0000 UTC

0600 UTC

1200 UTC

1800 UTC

extraction

0900 UTC

0445 UTC

0445 UTC

0445 UTC

 

7.6. Long range forecasts (3 months)

A specific version of ARPEGE model , called ARPEGE-Climat is used 3 times a month to run 125 days forecasts, starting from ARPEGE assimilation. The seasonal is using mainly the same ARPEGE software as short range forecast model, except the following points:

resolution, time step: This version of the ARPEGE model has a triangular truncature T63 without stretching. The collocation grid has 128x64 points with a reduction near the poles; it has 31 vertical levels like IFS model during ERA-15 ECMWF reanalysis. The time step is 1800 seconds.

radiation: Fouquart Morcrette scheme (1995)

clouds, vertical diffusion, stratified precipitations: Ricard Royer statistical scheme (1993).

8. Verification of Forecasts

Scores of the operational ARPEGE model:

Against analyses

24 hours

72 hours

 

NH

SH

TR

 

NH

SH

TR

Z500 RMSE

14.3

22.3

 

 

37.6

51.2

 

W250 RMSEV

5.5

6.3

5.4

 

11.5

12.9

8.6

W850 RMSEV

 

 

3.0

 

 

 

4.6

NH : Northern Hemisphere SH : Southern Hemisphere TR : Tropics

 

Against observations

24 hours

 

NA

EU

AS

AU/NZ

TR

NH

SH

Z500 RMSE

15.8

15.4

16.5

16.4

10.6

16.8

21.6

W250 RMSEV

7.7

6.8

7.5

8.7

6.6

7.2

9.1

W850 RMSEV

4.5

4.6

5.0

5.0

4.7

4.8

5.7

 

72 hours

 

NA

EU

AS

AU/NZ

TR

NH

SH

Z500 RMSE

41.1

36.6

32.0

32.4

13.2

40.4

41.7

W250 RMSEV

14.0

12.7

11.4

13.0

8.8

12.9

13.8

W850 RMSEV

6.8

6.5

6.8

6.7

5.6

6.9

7.4

NA : North America EU : Europe AS : Asia AU/NZ : Australia / New Zealand

NH : Northern Hemisphere SH : Southern Hemisphere TR : Tropics

Recall:

Météo-France draws up a "quarterly bulletin of monitoring of the numerical products used for meteorological forecasting" (french) and a "monthly bulletin of data monitoring" (french and English). These bulletins can be obtained by writing to:

Météo-France

SCEM/PREVI/COMPAS

42, av. Coriolis

F-31057 TOULOUSE Cedex 1

FRANCE

9. Future Plans

. enhanced vertical and horizontal resolution for ARPEGE/ALADIN models and in 4DVAR assimilation

. 10 members for seasonal forecast

. wave models twice a day

. use of ATOVS raw radiances from NOAA15 and NOAA16

REFERENCES

Bougeault P., 1985 : "Parameterization of cumulus convection for Gate. A diagnostic and semi-prognostic study". Mon. Wea. Rev., 113, 2108-2121.

Bouteloup Y., 1995: "Improvement of the spectral representation of the earth topography with a variational method", Mon. Wea. Rev., 123, 1560-1573

Courtier, P. and J.F. Geleyn, 1988 :"A global numerical weather prediction model with variable resolution: application to the shallow-water equations"". Quart. J. Roy. Meteor. Soc., 1114, 1321-1346.

Giard, D., and E. Bazile, 1999 : Implementation of a new assimilation scheme for soil and surface variables in a global NWP model, submitted to Mon. Wea. Rev..

Louis J.F., 1979: "A parametric model of vertical eddy fluxes in the atmosphere", Bound. Lay. Met., 17, 187-202

Lynch, P., D. Giard and V. Ivanovici, 1997 : Improving the efficiency of a digital filtering scheme for diabatic initialization, Mon. Wea. Rev., 125, 1976-1982

Louis J.F., M. Tiedtke and J.F. Geleyn, 1981 :"A short history of the PBL parameterization at ECMWF". ECMWF Workshop on PBL parameterization, ECMWF, Reading, UK, 59-80.

Morcrette, J.-J., and Y. Fouquart, 1985: On systematic errors in parametrized calculations of longwave radiation transfer. Quart. J. Roy. Meteor. Soc., 111, 691-708.

Noilhan, J., and S. Planton, 1989 : A simple parameterization of land-surface processes for meteorological models, Mon. Wea. Rev., 117, 536-549

Piedelievre J.P., L. Musson-Genon and F. Bompay, 1990: MEDIA - An Eulerian model of atmospheric dispersion: first validation on the Chernobyl release, Jour. of Appl. Met., vol. 29, N° 12, 1205-1220

Ricard, J.-L., and J.-F. Royer, 1993: A statistical cloud scheme for use in an AGCM. Ann. Geophys. 11, 1095-1115

Ritter B. and J.F. Geleyn, 1992 : "A comprehensive radiation scheme for numerical weather prediction models with potential applications in climate simulations". Mon. Wea. Rev., 120, 303-325.

Simmons A.J. and D.M. Burridge, 1981 : "An energy and angular momentum conserving vertical finite difference scheme on a hybrid vertical coordinate". Mon. Wea. Rev., 109, 758-766.

Yessad K. and P. Bénard, 1996 :"Introduction of a local mapping factor in the spctral part of the Météo_france global variable mesh numerical model". Quart. J. Roy. Meteor. Soc., 122, 1701-1719.