Humidity observing system experiments within the
ECMWF assimilation system
Fabrizio Baordo, Alan J. Geer, Stephen English
ECMWF, Reading, United Kingdom
1 Abstract
2 Humidity observing system experiments within the ECMWF assimilation system
The global analysis and forecast of observed humidity has been
studied by means of observing system experiments with the
ECMWF 4D-Var data assimilation scheme. Data from Microwave
Imagers (SSMIS, TMI, AMSRE), Microwave Sounders (MHS) and
Infrared Sounders (HIRS, AISI, AIRS, GEOS) instruments have
been used in order to evaluate the impact of each observing
system on the humidity and total column water vapour analysis
and forecast. The results show that each tested data type
contributes to the humidity analysis, but significant impact
has been observed by the assimilation of microwave imagers
observations in the all-sky approach (clear/cloudy situations)
over ocean (Bauer et al. 2010, Geer et al. 2010).
The impact of humidity observations within the ECMWF assimilation system has been studied
by means of observing system experiments. Six different observing system experiments have
been run for 1–31 August 2011, using the incremental 4D-Var assimilation system at T255
horizontal resolution (~78 km).
3
4
5
1
2
Relative humidity (%)
0
RMSobserving_system = RMS(ANControl – ANExp) – RMS(ANControl – ANControl*) (1)
400
600
800
1000
–90
1000
–90
0.9
20
Each observation type investigated has an impact on the humidity analysis. However, the results
show that the all-sky approach (MW Imagers) dominates the humidity analysis much more
than the other two observing systems, and in particular, marked impact is observed in the lower
troposphere (1000-600 hPa) and in the deep convection area (250-100 hPa). MW Sounders and
IR Sounders, instead, dominate the humidity analysis in the middle troposphere (600-200 hPa),
specially in the Northern and Southern hemisphere (figures 1 & 2). In addition, TCWV forecast time
series verified against independent observations from MWR (figure 3) show deterioration in RMS
up to 4 days (T+96) in the Tropics, when MW Imagers are removed from the Control experiment.
Land surface emissivity retrievals using SSMIS, TMI and AMSRE observations have been found
in very good agreement in both clear and cloudy sky conditions (figures 4, 8 & 9). The results
show also reasonable and good match with land surface emissivities provided by TELSEM atlas
available within the RTTOV package as well as with dynamic emissivity retrievals obtained from
AMSUA window channels (figures 4, 5, 6 & 7). These results are very promising and encourage
additional investigations on the possibility to assimilate microwave imagers observations over
land within the all-sky approach.
600
0
Latitude
30
60
90
2
0
600
800
800
1000
–90
1000
–90
–60
–30
0
Latitude
30
60
90
–60
–30
Global
0
Latitude
30
60
90
Northern hemisphere (20, 90)
8
10
40
60
80
Frequency (GHz), Vertical Polarization
Clear observations over land
0.7
0.9
0.9
0.85
0.8
0.7
0.7
100
Emissivity
0.6
0.4
60
80
100
120
140
Frequency (GHz)
160
20
180
Clear observations over land
1
AMSRE
ATLAS (TELSEM MW)
0
Clear observations over land
0.9
Emissivity
0.9
60
80
100
120
140
Frequency (GHz)
160
TCWV (kg m–2)
TCWV (kg m–2)
0.8
TMI
ATLAS (TELSEM MW)
60
80
20
40
Frequency (GHz), Vertical Polarization
0.9
0.85
0.8
20
40
60
80
Frequency (GHz), Horizontal Polarization
Clear observations over land
1
0.85
0.8
0.9
0.9
0.85
0.8
SSMIS
ATLAS (TELSEM MW)
60
80
20
40
Frequency (GHz), Vertical Polarization
0.9
-15
-30
0.85
0.8
20
40
60
80
Frequency (GHz), Horizontal Polarization
45
90
135
180
20
40
60
80
Frequency (GHz), Horizontal Polarization
0.95
0.9
0.9
0.85
0.8
0.8
Figure 8 Mean land surface emissivity retrievals from SSMIS channel 1 (50.3 GHz) for clear
observations, binned in 2.5˚ by 2.5˚ boxes over a 3 week period (1–21 August 2011)
100
Cloudy observations over land
1
0.95
0.85
0.8
0.75
0.75
0.7
0
20
40
60
Frequency (GHz)
80
100
0.7
0
60
40
Frequency (GHz)
20
80
100
Figure 7 As in Figure 5, but for SSMIS
Cloudy obs SSMIS Ch: 50.3
0.95
30
15
0.9
0
-15
-30
0.85
-45
-75
0
0.8
0.7
0
100
Clear observations over land
-75
-45
0.85
0.75
-60
-90
100
Cloudy observations over land
1
0.9
-60
-135
60
80
20
40
Frequency (GHz), Vertical Polarization
45
0.85
-45
0.7
0
100
0.9
60
0
0.8
0.95
60
15
Cloudy observations over land
0.85
0.95
75
30
100
0.75
Clear observations over land
90
0.95
20
40
60
80
Frequency (GHz), Horizontal Polarization
1
0.95
1
45
0.8
0.95
0.7
0
Clear obs SSMIS Ch: 50.3
0.85
0.7
0
100
0.75
100
Cloudy observations over land
Figure 6 As in Figure 5, but for TMI
Cloudy observations over land
20
40
60
80
Frequency (GHz), Horizontal Polarization
100
0.75
1
0.7
0
60
80
20
40
Frequency (GHz), Vertical Polarization
1
0.9
75
-90
-180
0.7
0
100
0.95
0.7
0
100
0.8
0.95
0.75
60
80
20
40
Frequency (GHz), Vertical Polarization
0.85
0.75
Clear observations over land
180
Figure 5 Mean and standard deviation of land surface emissivity variation with frequency
over a 3 week period (1-21 August 2011) for AMSRE compared with mean and standard
deviation computed used emissivities from TELSEM atlas. Left panels are for clear
observations, right panels for cloudy/rainy observations instead. SSMI-like frequencies with
vertical polarisation (top panel) have been separated from frequencies with horizontal
polarisation (bottom panel)
90
0.85
100
0.75
100
0.9
0.8
0.95
20
40
60
80
Frequency (GHz), Horizontal Polarization
0.9
0.75
Cloudy observations over land
1
0.95
0.85
0.95
0.75
T+12 T+24 T+48 T+72 T+96 T+120 T+144 T+168 T+192 T+216 T+240
0.95
100
20
40
60
80
Frequency (GHz), Horizontal Polarization
Cloudy observations over land
1
0.8
4
Clear observations over land
1
Cloudy observations over land
0.7
0
100
5
Figure 3 TCWV forecast time series verified against independent observations from Envisat Microwave
Radiometer (MWR). RMS differences between observations and forecast have been computed over
5-31 August 2011. 10 days TCWV forecast have been produced each hour from 00 UTC analysis. Black
line: Control experiment verified against MWR observations, red line: Exp1 (MW Imagers withheld from
Control) verified against MWR observations
0.75
0.85
4
0.7
0
0.9
1
5
0.4
0.9
60
80
20
40
Frequency (GHz), Vertical Polarization
6
0.75
0.95
0.75
Southern hemisphere (–90, –20)
T+12 T+24 T+48 T+72 T+96 T+120 T+144 T+168 T+192 T+216 T+240
0.6
0.95
0.8
Tropics (–20, 20)
3
40
60
80
Frequency (GHz), Vertical Polarization
Cloudy observations over land
1
0.85
4
3
0.8
0.2
6
T+12 T+24 T+48 T+72 T+96 T+120 T+144 T+168 T+192 T+216 T+240
1
0
8
T+12 T+24 T+48 T+72 T+96 T+120 T+144 T+168 T+192 T+216 T+240
0.7
0
1
0.8
Ctrl
Ctrl – MW Imagers
0.8
0.75
20
40
60
80
Frequency (GHz), Horizontal Polarization
Clear observations over land
4
0.85
0.75
0
0
1
0.95
5
6
0.8
100
6
3
0.75
Emissivity
Emissivity
Emissivity
Pressure (hPa)
400
Figure 4 Mean and standard deviation of land surface emissivity variation with frequency
over a 3 week period (1–21 August 2011) for SSMIS, TMI, AMSRE. Left panels are for clear
observations, right panels for cloudy/rainy observations instead. SSMI-like frequencies with
vertical polarisation (top panel) have been separated from frequencies with horizontal
polarisation (middle panel). Bottom panel is for the other investigated SSMIS frequencies
(50.3, 52.8, 150, 183 ± 6, 183 ± 3 GHz). In addition, as reference in the bottom panel, the red
diamond is the mean and standard deviation of AMSUA dynamic emissivity retrievals
Emissivity
–30
Baseline
400
0.85
0.95
0.7
0
4 Summary
–60
Emissivity
0
0.7
0
where TB is the observed brightness temperature, T↑Clr, T↓Clr, τClr are the up-welling and downwelling radiation and the atmospheric opacity of the clear sub column, and T↑Cld, T↓Cld, τCld are the
same quantities but for the cloudy sub column. Ts, instead, is the surface skin temperature taken from
the short-range forecasts. Observed brightness temperature have been selected within the latitude
range of [-60, +60] degree. Where c is greater than 0.05 observations have been considered as cloudy,
otherwise they are classified as clear (emissivity is estimated from equation 3 with c = 0).
90
Emissivity
0.8
1
(3)
60
Emissivity
where Tclr and Tcld are respectively the brightness temperature of the clear sub column (taking
into account only gaseous absorption) and of the cloudy sub column (taking into account cloud,
precipitation and scattering. For a given zenith angle (θ) and frequency (ν), the land surface emissivity
(ε) has been estimated as:
30
Emissivity
SSMIS
TMI
AMSRE
AMSUA (Ch 5)
Emissivity
Tallsky = cTCld – (1 – c)TClr (2)
0
Latitude
200
Emissivity
0.85
0
–30
4
-90
-180
-135
-90
-45
0
45
90
Figure 9 As in Figure 8, but for SSMIS cloudy/rainy observations
135
180
0.8
(Min 0.84, Max 1.00)
0.9
0.75
Emissivity
Land surface emissivity estimations have been already investigated by many authors (Karbou et
al. 2005, 2007, Prigent et al. 1997, 2005, among many others). However, in this work, the emissivity
retrieval has been implemented for the first time within the all-sky approach context (Geer et al. 2009).
RTTOV-SCATT cloud overlap schemes have two sub columns and, the all-sky brightness temperature ,
which depends on the average cloud fraction over the whole profile (c), is computed as follow:
–60
200
Emissivity
Emissivity
0.95
0.2
Pressure level (hPa)
800
Cloudy observations over land
1
0.95
1
Channel 5 (53.6 GHz), where clear sky observations over land are operationally assimilated within
the ECMWF system.
6
600
Relative humidity (%) (min: 2.26, max: 8.32)
400
8
Emissivity
Clear observations over land
1
l dynamic emissivity retrievals obtained from AMSUA Channel 3 (50.3 GHz) and assigned to
Baseline
MW Sounders
200
Emissivity
Table1 Summary of the observations withheld from
Control experiment
package (RTTOV v10);
1
2
3
4
5
Relative humidity (%)
Figure2 As in Figure1, but along zonal cross-sections
IR Sounders (Exp3)
Sensor
Satellite
Channels
NOAA17
7.33; 6.52 (µm)
HIRS
NOAA19
7.33; 6.52
METOPA
7.33; 6.52
1367.05; 1384.27; 1392.95;
1395.28; 1407.06; 1421.06;
IASI
METOPA
1422.27; 1990.04; 1994.41;
2014.91 (cm-1)
1251.40; 1316.10; 1339.70;
AIRS
AQUA
1342.30; 1367.80; 1392.20;
1514.0 (cm-1)
GOES 11
6.7 (µm)
GOES 13
6.7
GEOS METEOSAT 7 6.2
METEOSAT 9 6.2; 7.3
MTSAT-2
6.8
Baseline (Exp4)
All satellite data above withheld from Control
RTTOV v10, Users’ guide v1.5, Updated 2012-01-17, available at:
l land surface emissivities provided by TELSEM atlas (Aires et al. 2010) available within the RTTOV
IR Sounders
0
7
MW Imagers (Exp1)
Satellite
Channels (GHz)
SMSP17
19V&H; 22V; 37V; 91V
TRMM
19V&H; 22V; 37V; 85V
AQUA
19V&H; 24V&H; 37V
MW Sounders (Exp2)
Satellite
Channels (GHz)
NOAA18
183.3 ± 1.0; ± 3.0; ± 7.0
NOAA19
183.3 ± 3.0; ± 7.0
METOPA
183.3 ± 1.0; ± 3.0; ± 7.0
0.7
1000
0 1 2 3 4 5 6
Relative humidity (%)
MW Sounders
200
Satellite humidity observing system (Channels withheld from Control experiment)
The large information content coming from the Microwave Imagers in the all-sky approach
encourage additional studies in order to explore the use of microwave data over land. In this context,
inter-sensor land surface emissivity retrieval comparisons (SSMIS, TMI, AMSRE) have been performed
as well as some evaluations with independent emissivity data. In particular, emissivities retrieved
using SSMIS, TMI and AMSRE observations have been compared with:
1000
1
2
3
4
Relative humidity (%)
–2
where AN is the humidity analysis produced twice daily (00 or 12 UTC) by the variational
assimilation system. The second term on the right side of the equation can be interpreted as
an estimation of the humidity analysis uncertainty coming from the full assimilation system.
3 Preliminary results of land emissivity retrievals using
MW imager data in clear/cloudy situations
0
600
800
IR Sounders
The impact of each satellite observing system (MW Imagers; MW Sounders; IR Sounders;
Baseline), in all comparison in the following, has been estimated in terms of the root mean
square (RMS) of analysis differences:
MHS
1000
800
Pressure (hPa)
Pressure (hPa)
In order to better understand which differences are significant among the satellite
observations, a further experiment has been set up. The 6th experiment (Control*) is identical
to the Control, but different initial conditions for the first cycle (31 07 2011, 00 UTC) have been
used.
Sensor
MHS
Pressure level (hPa)
Pressure level (hPa)
Pressure level (hPa)
800
600
MW Imagers
Table 1 summarizes the channels that have been selected and removed by each sensor/
satellite within each experiment.
Sensor
SSMIS
TMI
AMSRE
600
Figure 1 Impact of each satellite observing system (MW Imagers; MW Sounders; IR Sounders;
Baseline) on the vertical profile of relative humidity, computed over 5-31 August 2011, for all the Globe,
Northern Hemisphere, Tropics and Southern Hemisphere. Calculations are performed according to
equation 1 considering the humidity analysis produced at 12 UTC
Exp4: Baseline: All data above remove (satellite humidity observation system is withheld
from Control)
Prigent, C., F. Chevallier, F. Karbou, P. Bauer and G. Kelly, 2005, AMSU-A
surface emissivities for numerical weather prediction assimilation schemes, J.
Applied Meteorology, 44, 416-426
http://research.metoffice.gov.uk/research/interproj/nwpsaf/rtm/rtm_rttov10.html
600
Emissivity
Prigent, C., W. B. Rossow and E. Matthews, 1997, Microwave land surface
emissivities estimated from SSM/I observations, J. Geophys. Res., 102, 21867–
21890
400
Emissivity
Karbou, F., N. Bormann and J-N. Thépaut, 2007, Towards the assimilation of
satellite microwave observations over land: feasibility studies using SSMI/S,
AMSU-A and AMSU-B, NWPSAF Tech. Rep., 37pp
400
TCWV (kg m–2)
Karbou, F., C. Prigent, L. Eymard and J. Pardo, 2005, Microwave land
emissivity calculations using AMSU-A and AMSU-B measurements, IEEE Trans.
on Geoscience and Remote Sensing, 43, 5, 948–959
400
MW Imagers
Exp3: IR Sounders withheld from Control (HIRS, IASI, AIRS, GEOS)
Geer, A. J., P. Bauer, and C. W. O’Dell, 2009. A revised cloud overlap scheme
for fast microwave radiative transfer in rain and cloud. J. Applied Meteorology,
48, 2257–2270
Geer, A. J., P. Bauer, P. Lopez, 2010. Direct 4D-Var assimilation of all-sky
radiances: Part II. Assessment. Q. J. R. Meteorol. Soc., 136, 1886–1905
400
TCWV (kg m–2)
Bauer, P., A. J. Geer,. P. Lopez, D. Salmond, 2010. Direct 4D-Var assimilation of
all-sky radiances. Part I: Implementation. Q. J. R. Meteorol. Soc., 136, 1868–1885
200
Pressure (hPa)
Aires, F., C. Prigent, F. Bernado, C. Jiménez, R. Saunders, and P. Brunel,
2010. A tool to estimate Land Surface-Emissivities at Microwaves frequencies
(TELSEM) for use in numerical weather prediction. Q. J. R. Meteorol. Soc., 137,
690–699
200
1000
Exp2: MW Sounders data withheld from Control (MHS)
Southern hemisphere (–90, –20)
200
800
Exp1: MW Imagers data withheld from Control (SSMIS, TMI, AMSRE)
Tropics (–20, 20)
200
(Min 0.8, Max 0.98)
References
A Control assimilation experiment that uses all the available observations of the operational
system has been set up and used as reference to investigate the impact of each observing
system. The impact of humidity observations has been assessed by means of 4 additional
experiments, same as the Control, but in each experiment a specific data type has been
withheld. In order to estimate the impact of humidity observations from satellite sensors, the
experiments have been set up as follows:
Emissivity
The large information content coming from the Microwave
Imagers in the all-sky approach encourage additional studies
and a potential further step is to prove that this data can also
be used over land (all-sky assimilation has, to date, been limited
to ocean cases only). In order to explore the use of Microwave
observations over land under clear/cloudy situations, some
feasibility studies on the land emissivity retrieval have been
performed within the ECMWF assimilation system as preliminary
step. Results show that land surface emissivities calculated using
SSMIS, TMI and AMSRE observations have been found in very
good agreement in both clear and cloudy sky conditions.
Northern hemisphere (20, 90)
Global