Use of satellite data at Météo-France

1. Use of satellite data at Météo-France Élisabeth Gérard Météo-France/CNRM/GMAP/OBS, Toulouse, France on behalf of: É. Gérard, F. Rabier, N. Fourrié and D. Lacroix (ATOVS) C. Payan and F. Rabier (Météosat) 25thEWGLAM/10thSRNWP Lisbon, Portugal 6-9 October 2003

2. ATOVS data • Operational implementation of • Raw radiances instead of preprocessed radiances: 22 October 2002 (+ European & American profilers) • NOAA17 on top of NOAA15 & NOAA16: 17 December 2002 • AMSUA data only in operations • AMSUA data denial experiments (OSEs) • Research experiments with locally received data • Lannion/EARS in addition to Nesdis/Bracknell • Research experiments with HIRS 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

3. Raw radiances instead of preprocessed radiances • No more 1DVar pre-processing • Ts in the control variable • Extrapolation above the top of the model (1 hPa) up to 0.1 hPa by regression • « Rain contamination » test • observation rejected if |obs-guess|channel 4 > 1.5 K • 250 km thinning • Data coverage more regular, less missing orbits, less scan border pixel removal • Use of more channels over land 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

4. Raw radiances instead of preprocessed radiances Time series of rms errors and biases 24 hour forecast 200 hPa geopotential scores over 1 month 22 Aug - 22 Sep 2002 Northern Hemisphere Southern Hemisphere scores computed wrt their own analysis Preprocessed radiances Raw radiances 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

5. AMSUA data denial experiments • Uniform (unstretched) ARPEGE: • 4DVar analysis T107C1L41 (up to 1hPa) • Forecast T359C1L41 with 1800 s time step • Experiments with and without AMSUA data: 23 Dec 2002  12 Jan 2003 • One forecast a day from 00 UTC analysis: • Short cut-off time: 03h50 (production) • Long cut-off time: 09h55 (assimilation) • Impact of AMSUA data wrt cut-off time 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

6. Benefit of AMSUA data Time series of rms errors and biases 24 hour forecast 500 hPa geopotential scores over 3 weeks 23 Dec 2002 - 12 Jan 2003 Northern Hemisphere rms difference scores wrt long cut-off analysis with AMSUA data Southern Hemisphere With AMSUA (long cut-off) Without AMSUA rms difference 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

7. AMSUA data denial experiments • Positive impact of using AMSUA data • More pronounced benefit • over Southern Hemisphere than Northern Hemisphere, • over North America than Europe, • when cut-off time is long • Increase of gain with forecast range more regular with long cut-off time over North America (more robust signal) • When AMSUA data are used, more gain wrt cut-off time is expected if short cut-off time is shorter (i.e. 1h50 @ 00 UTC in oper stretched ARPEGE model) 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

8. EARS data • Eumetsat ATOVS Retransmission Service • Data rapidly available • No blind orbit for NOAA17 • (reception centres: Greenland, Norway, Canary Islands) Research experiments with locally received AMSUA data 13 March 2003 12 UTC Data available for the operational production, 1h50 cut-off time Nesdis/Bracknell data • Lannion data 45W/40E/70N/30N • Even more rapidly available, but smaller area & only NOAA16/NOAA17 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

9. Impact of EARS and Lannion data in addition to Bracknell data (rms/bias wrt radiosondes) Temperature @ 250 hPa forecast range (hour)  12 36 48 BracknellBracknell+EARS+Lannion First step: assimilation in operational model ARPEGE Next step: assimilation in regional model ALADIN (… AROME) in research mode AMSUA, HIRS, AMSUB (observation density, bias correction, …)

10. Research experiments with HIRS • On top of AMSUA data over 23 Dec 2002 – 12 Jan 2003 • 250 km thinning (as for AMSUA) • Cloud contamination test with channel 8 • rejection if x < obs-guess < y, (x,y)=f(latitude) water vapour channels 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

11. Forecast scores (rms & bias) over Europewith HIRSwithout HIRS scores computed wrt own analysis 24 hour 48 hour 72 hour forecast range  geopotential temperature wind rel. humidity

12. Results of last experiments with HIRS • Occurrence of « ringing » problem in stratosphere this summer • channels 5, 6, 7, 14, 15 blacklisted over seaice • channels 4 & 15 (tail up to 1hPa) to be blacklisted ? • only channels 11 & 12 to be assimilated ? • Cloud contamination test to be revised ? • First guess check to be revised ? (obs-guess)2 <(o2+b2) … under investigation … 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

13. Assimilation of Meteosat data Use of BUFR winds produced by EUMETSAT with a quality index and disseminated every 90 minutes compared to Use of currently operational SATOB winds produced every 6 hours … as a preparation towards the use of other geostationary satellite data (GOES, etc.)… 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

14. Assimilation of Meteosat data • Experiments with the uniform ARPEGE configuration • 23 Dec 2002 - 12 Jan 2003 • SATOB winds • Only data with QI>0.8 are transmitted • BUFR winds conditional use • Weak constraint: QI>0.6 for upper level winds and over sea for mid-level winds • Strong constraint: QI>0.8-0.9 elsewhere as a function of latitude, level, channel Vis/IR/mw 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003

15. Meteosat 5&7 observation fit to first guess and analysisarea=50N/50S/113E/50W Used U component • BUFR versus SATOB • more data used • rms and bias reduction Used V component

16. Present time, near/next future … for satellite observations… • Tuning of AMSUA data: density, rain detection • HIRS use, cloud detection, obs error tuning, blacklist • ATOVS bias correction wrt analysis (?) • Assimilation of BUFR winds on going • AIRS: screening, bias correction already performed, first assimilation experiments on going Next future: AMSUB, SeaWinds, MSG, MODIS, SSM/I(S) 25thEWGLAM/10thSRNWP, Lisbon, Portugal, 6-9 October 2003