Assimilation of cloud and precipitation affected microwave radiances at ECMWF

advertisement
Alan Geer
Assimilation of cloud and precipitation affected
microwave radiances at ECMWF
SSM/I cloudy/rainy radiances have been assimilated operationally at ECMWF using 1D+4D-Var
since 28th June 2005.
European Centre for
Medium-Range Weather Forecasts,
Shinfield Park, Reading, RG2 9AX, UK
Telephone: +44 118 949 9409
E-mail: [email protected]
Alan Geer, Peter Bauer, Philippe Lopez, Deborah Salmond and Sabatino Di Michele
From each observation, 1D-Var retrieves temperature and humidity profiles and (from cycle 31r1 onwards)
surface wind speed. The observation operator includes simplified large-scale and convective cloud
schemes, along with microwave radiative transfer.
From each retrieval, total column water vapour (TCWV) is calculated and then included in the main 4D-Var
assimilation.
Case study: problems with snow-affected retrievals
Meteosat visible channel image for 12Z
14th August 2005: We examine a 1D-Var
retrieval in the area of deep convection at
the front of an intrusion of cold, unstable,
polar air.
Recent developments include:
• Improved screening of snow-affected radiances, resulting in improvements to forecast scores in the
Southern Hemisphere
• Improved bias correction
• Surface wind speed as a “sink variable” in the 1D-Var retrieval
The impact of 1D+4D-Var is thought to be limited by the “bottleneck” caused by transferring information
via TCWV pseudo-observations, and by the normalised relative humidity control variable in use ECMWF.
This penalises moistening increments in regions of high relative humidity, such as found in the majority
of rain and cloud affected SSM/I observations. Drying increments are not affected, so the net impact is
towards a drying of the analyses, particularly in the tropics.
Direct 4D-Var of rain and cloud-affected SSM/I radiances is in testing at the moment and should help
solve some of these problems.
Courtesy NERC Dundee satellite receiving station, © Copyright EUMETSAT, 2005
Tephigram showing 1D-Var first-guess (black) and retrieved (red) temperature and dewpoint: there is much moistening in the lower atmosphere.
Improvement
0
–6
-60
-30
0
30
Latitude
60
Pressure (hPa)
Pressure (hPa)
–5
0
–4
0
T=
–3
Dewpoint
800
-0.10
0
–1
900
90
1000
0.7
1
1.5 2
3
qs (g/kg)
5
7
10
15
Control
A – Revised eigenvector calculation
B – Surface wind speed as sink variable +A
C – Improved bias correction +A +B
D – Screening of snow-affected observations +A +B +C
E – Revised observation errors +A +B +C +D
0.05
0
Pressure [hPa]
500 hPa SH: 90 – 20 S
400
Cloud fraction
Snow
Cloud ice
600
800
-0.05
Rain
-0.1
1000
0
4
2
3
Forecast day
1
5
6
0
0.1 0.2 0.3 0.4 0.5
Rain or snow flux (g m-2s-1)
Water
10-12 10-10 10-8 10-6 10-4 10-2
Liquid or ice water content (kg kg-1)
0
0.2
0.4 0.6 0.8
Cloud fraction
TCWV /kgm-2
Difficulty of transferring information from 1D to 4D-Var
Balanced moistening and drying TCWV departures are transformed to largely drying increments in 4D-Var. This is because
1D-Var retrievals of cloudy/rainy profiles are made largely in areas close to saturation. In such areas, the ECMWF’s normalised
relative humidity control variable penalises positive moisture increments. The situation should be improved by direct 4D-Var of
cloudy/rainy SSM/I radiances (currently in testing), and perhaps by the adoption of a total moisture control variable.
21:00:00 31-Jul-2004 to 09:00:00 1-Aug-2004
15N
6
6
Latitude
0
-2
5N
4
2
10N
Latitude
2
10N
TCWV O-B [kg m-2]
4
0
-2
5N
-4
-4
-6
0
120W
Moistening
115W
110W
105W
Longitude
100W
Drying
95W
-6
0
120W
115W
Drying
110W
105W
Longitude
100W
95W
1.0
Tb departure /K
FG
7.6
4.5
7.9
2.0
4.2
8.8
4.1
Analysis
8.8
2.0
1.3
-1.1
3.4
-1.8
13.3
5.5
SSMI channel
19v
19h
22v
37v
37h
85v
85h
12.0
1D-Var SSM/I brightness temperature departures at first guess and retrieval: Only
channels 19v, 19h and 22v are assimilated. These are less sensitive to snow and to rain close
to the surface. Channels 37v/h and 85v/h are not used in the retrieval due to their more nonlinear behaviour, but are sensitive to snow. Large departures remaining in these channels
suggest that snow is poorly retrieved.
4D-Var increments in TCWV due to
SSM/I cloudy / rainy observations
TCWV difference [kg m-2]
1D-Var retrievals: First guess departures in TCWV
15N
Cloud properties from the 1D-Var
observation operator at first guess
(black) and retrieval (red): to match
observed SSM/I brightness temperatures requires a large increase in rain,
snow, cloud ice and water content,
and cloud fraction. Convection has
been strongly increased.
200
0.1
Degradation
20
700
800
1000
-90
Improvements to 1D+4D-Var
Improvement
30
90
600
600
0
60
400
500
θ=
0
30
Latitude
-0.05
10
800
200
400
40
0.00
θ=
600
-30
90
θ=
400
-60
60
θ=
200
1000
-90
0
30
Latitude
T+144
0
-30
T=
-60
0
T=
1000
-90
0
90
–2
60
T+96
Temperature
θ=
0
30
Latitude
300
T=
-30
0.05
800
T=
-60
600
50
1000
-90
400
T=
800
T=
600
Pressure (hPa)
400
200
Normalised difference in RMS error
Pressure (hPa)
Pressure (hPa)
0.10
200
Degradation
Normalised difference in RMS temperature forecast scores between
experiments with and without SSM/I
1D+4D-Var, for the period August 2005.
In Southern Hemisphere high latitudes,
1D+4D-Var assimilation was degrading forecast scores in temperature and
geopotential. A number of improvements
have been made, the most important of
which are a revised bias correction, and
the screening of snow-affected retrievals.
0
200
T+48
–7
T+24
T=
Statistical
significance
at 90%
θ=
Normalised difference in RMS relative
humidity forecast scores between
experiments with and without SSM/I
1D+4D-Var: For the period August to
November 2004, assimilation of SSM/I
cloudy/rainy radiances results in improved
relative humidity forecasts, particularly in the
tropics at around 700hPa over the first few
days of the forecast.
Sat. specific humidity
Dry adiabats (θ)
Isobars (pressure)
Saturated adiabats
1D+4D-Var assimilation improves relative humidity forecast scores
References
Bauer, P., Lopez, P., Benedetti, A., Salmond, D. and Moreau, E.: Implementation of 1D+4D-Var assimilation of precipitationaffected microwave radiances at ECMWF. I: 1D-Var, Quart. J. Roy. Meteorol. Soc., to appear, 2006
Bauer, P., Lopez, P., Salmond, D., Benedetti, A., Saarinen, S. and Bonazzola, M.: Implementation of 1D+4D-Var assimilation of precipitation-affected microwave radiances at ECMWF, II: 4D-Var, Quart. J. Roy. Meteorol. Soc., to appear, 2006
Download