Observing system
experiments on ATOVS
orbit constellations
Enza Di Tomaso and Niels Bormann
European Centre for Medium-Range Weather Forecasts, Reading, UK
E-mail: Enza.Ditomaso@ecmwf.int and Niels.Bormann@ecmwf.int
The impact of ATOVS data on the performance of global numerical weather
prediction (NWP) is evaluated through observing system experiments on different
orbit constellations. This work aims to provide some references for planning future
observing systems that involve ATOVS-like instruments, according to the CGMS’
objectives to harmonise meteorological satellite mission parameters such as orbits.
For more details on this work please see
http://www.ecmwf.int/publications/library/do/references/show?id=90074
Experiments
At the time of writing, there are seven polar-orbiting satellites present
in the ECMWF system carrying one or more ATOVS instruments:
the NOAA series, NASA’s Aqua satellite and the European MetOp-A
satellite. A set of experiments were run where ATOVS Microwave
(MW) sounder data (AMSU-A and AMSU-B/MHS) from alreadydeployed instruments were added or denied as shown in Table 1.
The NOAA-15 experiment uses a more homogeneous sampling
with equally spaced orbits, whereas in the NOAA-19 experiment
two orbits are very close to each other (Figure 1). Only AMSU-A
has been tested as an additional instrument in the “three-satellite
experiments”, as there are not three satellites in the desired orbits
carrying all functioning ATOVS instruments. Experiments were run
from 14 April 2009 to 4 August 2009 within the ECMWF 4D-Var
assimilation system at a T511 (40 km) resolution.
METOP-A
NOAA-18
Experiment name
Satellite
MW sounders
“no-MW sounder experiment”
—
—
NOAA-18
AMSU-A MHS
MetOp-A
“two-satellite experiment”
“NOAA-15 experiment”
“NOAA-19 experiment”
Experiment name
Satellite
MW sounders
NOAA-15
AMSU-A
AMSU-A MHS
NOAA-16
AMSU-A
(until 22 June 2009)
NOAA-15
AMSU-A
NOAA-17
AMSU-B
NOAA-18
AMSU-A MHS
MetOp-A
AMSU-A MHS
NOAA-18
AMSU-A MHS
NOAA-18
AMSU-A MHS
NOAA-19
AMSU-A MHS
NOAA-19
AMSU-A
Aqua
AMSU-A
MetOp-A
AMSU-A MHS
MetOp-A
AMSU-A MHS
“all-satellite experiment”
Table 1 ATOVS MW sounders available to the experiments.
"two-satellite experiment"
METOP-A
"NOAA-15 experiment"
NOAA-18
METOP-A
NOAA-15
80°N
80°N
80°N
60°N
60°N
60°N
40°N
40°N
40°N
20°N
20°N
20°N
0°
0°
0°
20°S
20°S
20°S
40°S
40°S
40°S
60°S
60°S
60°S
80°S
80°S
80°S
160°W
120°W
80°W
40°W
0°
40°E
80°E
120°E
160°E
160°W
120°W
80°W
40°W
0°
40°E
80°E
120°E
160°E
160°W
"NOAA-19 experiment"
NOAA-18
120°W
NOAA-19
80°W
40°W
0°
40°E
80°E
120°E
160°E
Figure 1 Sample coverage for used AMSU-A data in the “two-satellite experiment”, the “NOAA- 15 experiment”, and the “NOAA-19 experiment”.
Results
Instrument(s): MetOp-A, AMSU-A. Area(s): N. Hemisphere, S. Hemisphere, Tropics
From 00Z 20 Apr 2009 to 12Z 4 Aug 2009
Statistics of observation departures from NWP
model estimates.
12
10
9
8
6
5
0.75
Conclusions
• The assimilation of AMSU-A observations from a third satellite
has a positive forecast impact in the Southern Hemisphere in
comparison to the use of just two AMSU-A instruments, and
there is a clear advantage in assimilating all available ATOVS
data. The latter shows that the benefit of ATOVS data is not
saturated yet with a three-satellite configuration.
• ATOVS microwave sounder data from three satellites providing
a greater temporal sampling (i.e. MetOp-A, NOAA-18 and
NOAA-15) have a slightly larger positive forecast impact in the
Southern Hemisphere than data from three satellites having a
less optimal coverage (i.e. MetOp-A, NOAA-18 and NOAA-19).
• Departures of MetOp-A AMSU-A observations from the NWP
model estimates of the atmospheric state show some benefits
from assimilating observations from NOAA-15 rather than
NOAA-19 in addition to the a.m. and p.m. satellites (MetOp-A
and NOAA-18), providing a further indication of an improved
short-term forecast in the constellation of more evenly-spaced
orbits.
0.8 0.85 0.9 0.95
1
Standard deviation (normalised)
−0.6 −0.4 −0.2
0
0.2
Mean (K)
0.4
0.6
Background departure o−b
Analysis departure o−a
Bias correction of obs
“NOAA-15 experiment”
“NOAA-19 experiment”
Figure 2 MetOp-A brightness temperature departure statistics for the “NOAA-15
experiment” (black) and the “NOAA-19 experiment” (red). Standard deviations
have been normalised to the “NOAA-19 experiment”.
S. Hemisphere, 500 hPa Z
Normalised difference
rms forecast error
0
−0.02
−0.06
−0.04
−0.08
−0.06
−0.1
0
1
2
3
4
5
6
7
8
S. Hemisphere, 1000 hPa
0
−0.08
0.02
−0.04
0
−0.06
−0.02
−0.08
−0.04
−0.1
0
1
2
3 4 5 6
Forecast day
7
0
8
−0.06
1
2
3
4
5
6
7
1
2
3 4 5 6
Forecast day
7
8
S. Hemisphere, 1000 hPa temperature
0
-0.02
-0.04
0
8
1
2
3
4
5
6
7
8
6
7
8
S. Hemisphere, 850 hPa temperature
0.02
0
-0.02
-0.04
“two-satellite experiment” – “no-MW sounder experiment”
“NOAA-15 experiment” – “no-MW sounder experiment”
“all-satellite experiment” – “no-MW sounder experiment”
Figure 4 Normalised differences in the root-mean-square forecast error
between the “two-satellite experiment” and the “no-MW sounder experiment”
(black), between the “NOAA-15 experiment” and the “no-MW sounder
experiment” (red), and between the “all-satellite experiment” and the “noMW sounder experiment” (green) for the 0Z forecast of the 500 hPa and 1000
hPa geopotential. Verification is against the experiment own-analysis and the
sample size is 107. Negative values indicate that the “two-”, “NOAA-15”, and
“all-satellite experiment” have smaller RMS errors than the “no-MW sounder
experiment”.
0
0.5
1
-1
Mean (ms )
Background departure o−b
Analysis departure o−a
0.02
0.04
0
−0.5
Figure 3 Radiosonde winds departure statistics for the “two-satellite experiment”
(black) and the “no-MW sounder experiment” (red). Standard deviations have
been normalised to the “no-MW sounder experiment”.
N. Hemisphere, 1000 hPa Z
0.04
−0.02
0.7
0.8
0.9
1
1.1
1.2 −1
Standard deviation (normalised)
“two-satellite experiment”
“no-MW sounder experiment”
0.04
0
−0.04
b
N. Hemisphere, 500 hPa Z
0.02
−0.02
−0.12
Normalised difference
rms forecast error
• When averaged over the extra-Tropics the impact for the forecast
of the geopotential of the “NOAA-15 experiment” versus
the “NOAA-19 experiment” is neutral to slightly positive. For
the forecast of the temperature the impact of the “NOAA-15
experiment” versus “NOAA-19 experiment” is also quite neutral,
with a slightly positive impact in the Southern Hemisphere that
appears statistically significant at the 95% level about the 5-day
forecast at 1000 hPa and 850 hPa (Figure 5).
Pressure (hPa)
11
Forecast impact
• The assimilation of AMSU-A and AMSU-B/MHS data from two
satellites has a very strong positive impact on the forecast of
the geopotential. The full system still outperforms the two or
three satellite systems in the Southern Hemisphere (Figure 4).
a
5
10
20
30
50
70
100
150
200
250
300
400
500
700
850
1000
0.6
Normalised difference
rms forecast error
• When comparing the “NOAA-15 experiment” and the “NOAA19 experiment”, there are no relevant differences in the
departure statistics of the radiosonde temperatures in favour of
one experiment or the other, and the differences are marginal
also for the aircraft temperatures and for radiosonde humidity
observations.
b
Normalised difference
rms forecast error
• Wind information can be gained through the assimilation of
ATOVS MW sounders data in 4D-Var (Figure 3).
a
13
AMSU−A channel number
• Both NOAA-15 and NOAA-19 bring some improvement to the
fit to temperature observations compared to the assimilation
of data from just two satellites, as well as to the fit to AMSU-A
data already present in the system (onboard of NOAA-18 and
MetOp-A). The latter improvement is slightly greater in the
“NOAA-15 experiment” than in the “NOAA-19 experiment”
(Figure 2).
14
Instrument(s): TEMP-Uwind. Area(s): N. Hemisphere, S. Hemisphere, Tropics
From 00 UTC 20 Apr 2009 to 12 UTC 4 Aug 2009
0
1
2
3
4
5
Forecast day
“NOAA-15 experiment” – “NOAA-19 experiment”
Figure 5 Normalised differences in the root-mean-square forecast error between
the “NOAA-15 experiment” and the “NOAA-19 experiment” for the 0Z forecast of
the 1000 hPa and 850 hPa temperature for the Southern Hemisphere. Verification
is against the operational analysis and the sample size is 107. Negative values
indicate that the “NOAA-15 experiment” has smaller RMS errors than the “NOAA19 experiment”.
Acknowledgements
Enza Di Tomaso was funded by the EUMETSAT fellowship programme.
Alan Geer is gratefully acknowledged for providing the software to produce some of the above plots and Rob Hine for designing the poster.