Ongoing developments on the use of microwave sounders
and imagers at Météo-France
Louis-François M EUNIER, Philippe C HAMBON, Jean-François M AHFOUF
CNRM/GAME, Météo-France and CNRS, 42 av Coriolis, 31057 Toulouse, France;
Email: louis-francois.meunier@meteo.fr
The Arpège global prediction system – Current observation usage
Averaged Linear Estimate of Impact (Forecast Error) in percentage
XPID: 56FR / Trajectories: both (an+bg)
Base: 2013/09/01 @ 00 - 2013/09/30 @ 18 : 120 cycles.
Averaged Number of observations
XPID: 56FR / Trajectories: both (an+bg)
Base: 2013/09/01 @ 00 - 2013/09/30 @ 18 : 120 cycles.
• Even though microwave radiances are few compared to IR hypersectral
data, their impact on the analyses quality is high (see Fig. 1).
Use of ATMS data
2000000.0
(%)
Figure 1: Forecast sensitivity to observations for each observation
type.
1500000.0
0.0
• Satellite data are by far the more numerous (see Fig. 2). However, on a per
observation basis, conventional data have an important impact.
AMSU-A
IASI
AIRCRAFT
TEMP
SATWIND
SCATT
ATMS
GNSS RO
BUOY
AIRS
AMSU-B
SYNOP
PILOT
HIRS
SEVIRI
CRIS
SSMIS
TEMP-SHIP
SYNOP-SHIP
Gd GNSS
25.0
0.0
MW observation usage in Arpège 4D-var:
20.0
• Background error variances and covariances from an Ensemble Data Assimilation system (low resolution 4D-Var)
15.0
• Variational bias correction scheme for satellite data
10.0
• Incremental 4D-Var with 6-h window – 2 outer loops
5.0
AMSU-A
IASI
AIRCRAFT
TEMP
SATWIND
SCATT
ATMS
GNSS RO
BUOY
AIRS
AMSU-B
SYNOP
PILOT
HIRS
SEVIRI
CRIS
SSMIS
TEMP-SHIP
SYNOP-SHIP
Gd GNSS
500000.0
• Spectral model with variable resolution: T798C2.4L70 (from 10 km over
France to 60 km over New-Zealand)
1000000.0
The Arpège global system:
Figure 2: Number of observations for each observation type.
Impact of Megha-Tropiques/SAPHIR data
• Raw ATMS data are averaged into pseudo observations on a 3x3 grid
• All 6 SAPHIR channels are used with a variational bias correction.
• Averaged ATMS radiances are assimilated in a very similar manner than AMSU data
• SAPHIR has no window channel so the lowest peaking sounding channel is
used for cloud/rain clearing (during this check a constant bias correction is
applied in order to avoid interactions with VarBC)
• The emissivity retrieval (see [Karbou et al., 2006]) is used over land and sea-ice
• The availability of channels from 23GHz to 183GHz allows to compute a variety of diagnosis in order to identify and remove cloudy and rainy scenes (see Fig. 3)
(a) TB diff. between 89 and 165GHz
(b) Liquid Water Path
(c) Ch1 pseudo obs standard deviation
• Data are also assimilated over land. The AMSU-B/MHS emissivity atlas at
89Ghz is used to prescribe the surface emissivity.
• An assimilation experiment has been run between 15th June 2012 and 10th
August 2012 (see [Chambon et al., 2014])
• When SAPHIR data are assimilated, observation statistics are significantly improved even for other instruments like MHS or infrared sounders (see Fig. 5).
• As can be seen in Fig. 6, forecast scores are improved with respect to a reference experiment excluding SAPHIR.
Figure 3: ATMS observation based diagnosis used in operations on 1st January 2014 at 00 UTC.
• For humidity channels, the emissivity retrieval over
sea-ice will be re-tunned in order to take into account
SSMI/S high viewing angle.
2
10
HIRS/MetOp-A channels
SAPHIR/Megha-Tropiques channels
10
0.0
Population
Mean first guess departure after bcor (K)
3
1
0.5
10
1.0
0.4
0.2
0.0
0.2
Figure 5: Observation statistics over the tropical region between 15/06/2012 and 10/08/2012. Plain/Dashed lines
show the observation minus guess/analysis standard deviation. When SAPHIR data are assimilated statistics are
plotted in blue. (a) Microwave sensors. (b) Infrared sensors.
0
0.4
Cos solar zenith angle
B39T: 2013071600-2013091500.
SSMIS/DMSP-F18. Ch #3 (53.6V).
1.0
(a) reference: ECMWF analyses
1
10
1.0
0.8
0.6
0.4
0.2
0.0
0.2
0.4
0.6
0
0.8
Cos solar zenith angle
Figure 4: First guess departures as a function of the cosine of solar zenith angle
for SSMI/S Ch. 3 aboard DMSP-F17 (top)
and DMSP-F18 (bottom).
[Chambon et al., 2014] Chambon, P., Meunier, L.-F., Guillaume, F., Piriou, J.-M., Roca, R., and Mahfouf, J.-F.
(2014). Investigating the impact of the water vapour sounding observations from SAPHIR on board MeghaTropiques into the ARPEGE global model. To be submitted to Quarterly Journal of the Royal Meteorological
Society.
[Karbou et al., 2006] Karbou, F., Gérard, E., and Rabier, F. (2006). Microwave land emissivity and skin temperature for AMSU-A and -B assimilation over land. Quarterly Journal of the Royal Meteorological Society,
132(620):2333–2355.
0.5
500
0.0
−0.5
700
−1.0
-0.25
850
925
1000
−1.5
0
6 12 18 24 30 36 42 48 54 60 66 72 78 84 90 96 102
Forecast range (h)
−2.0
1.5
1.0
400
-0.25
0.5
500
5
400
2.0
-0.2
5
1.0
−0.5
700
850
925
1000
0.0
Mean difference of RMSE between EXP and CTRL (%)
10
0.5
1.5
100
150
200
250
300
Pressure levels (hPa)
0.0
Pressure levels (hPa)
10
2
(b) reference: radiosondes
2.0
-1.25
100
150
200
250
300
3
-0.7
10
0.5
Mean difference of RMSE between EXP and CTRL (%)
• The thinning of SSMI/S data will be lowered to
1.25◦(like any other microwave instrument).
10
0.5
Population
• DMSP-F18 radiances show a significant ascending/descending bias (especially for temperature sounding channels). The cosine of solar zenith angle will be
considered as an additional predictor of the variational
bias correction scheme. (see Fig. 4).
SSMIS/DMSP-F17. Ch #3 (53.6V).
1.0
Mean first guess departure after bcor (K)
• DMSP-F16 radiances show degraded first guess departure statistics compared to other SSMI/S data. It will be
excluded in our experimental set-up.
B39T: 2013071600-2013091500.
SEVIRI/Meteosat10 channels
Assimilation of SSMI/S sounding channels
The assimilation of SSMI/S sounding channels has been
attempted in 2013. SSMI/S sounding channels were
treated in a similar way than AMSU-A/MHS channels.
Unfortunately, the impact on forecast scores was neutral
and/or not significant. Several shortcomings have been
identified and are currently under being dealt with:
(b)
MHS/MetOp-A channels
(a)
-0.25
−1.0
−1.5
0
12
24
36
48
60
Forecast range (h)
72
84
96
−2.0
Figure 6: Difference of relative humidity RMSE between experiments with and without SAPHIR data assimilated
(from 15/06/2012 to 10/08/2012). Negative values show a positive impact. Grey zones indicates that scores are not
significant at the 99% level of confidence.
Perspectives
• Preparatory studies towards the assimilation of rainy microwave radiances in
regional models (J. Guerbette, PhD student)
• Potential use of AMSR-2 and GMI radiances (in clear-sky conditions)