Recent development
de elopment
p
of satellite data assimilation at JMA
Akira Okagaki,
Okagaki Masahiro Kazu
umori and Hidehiko Murata
N
Numerical
i l Prediction
P di ti Di
Division,
i i , JJapa
p n Meteorological
M t
l gi l Agency
Ag
y (JMA)
(
)
a okagaki@met kishou.go.jp
a-okagaki@met.
kishou go jp
1.Outline of NWP systems at JMA
Satellite (to be) used in the operational global and meso-scale NWP system
Items in red were implemented in the operational system since ITSC17
Specification of JMA’s forecast model and data assimilation system
Global Model & Analysis
(GSM GA)
(GSM,GA)
Model
Horizontal /vertical
res.
TL959/60 (0
(0.1hPa)
1hPa)
Forecast range
(I iti l titime))
(Initial
84h (00
(00,06,18UTC)
06 18UTC)
216h (12UTC)
Data Assimilation
(inner loop res
res.))
4D-Var (TL319)
( 319)
Assimilation window
6h (-3
( 3 ~ +3
3 hours)
Satellite/Instrument
1.
Sounde
er
5km / 50(21
50(21.8km)
8km)
15h (00
(00,06,12,18UTC)
06 12 18UTC)
33h (03
(03,09,15,21UTC)
09 15 21UTC)
GA
MA
NOAA15,16,18,19,
, , , , Aqua,
q , Metop
p / AMSU-A
Radiance
Radiance
NOAA18 19 Metop / MHS
NOAA18,19,
Radiance
Radiance
(Aqua/AIRS, Metop/IASI)
Under development
Under development
TRMM/TMI
Radiance
Radiance Rain Rate
Radiance,
DMSP16 17 18 / SSMIS
DMSP16,17,18
R di
Radiance
R di
Radiance,
Rain
R i R
Rate
t
Radiance
Radiance
2.
2
MW Imag
ger
4D-Var (15km)
(1
)
Radiance
assimilation
Cut off time
Meso-scale
Meso
scale Model & Analysis
(MSM MA)
(MSM,MA)
3h (-3
( 3 ~ 0 hours)
AMV
X
Metop/ASCAT
Ocean surface wind
Under development
COSMIC
R f ti it
Refractivity
U d d
Under
development
l
t
Metop/GRAS
Refractivity
Under development
5.
GPS-R
RO
50m
21E
2.1
Exploit
l it AMSU
AMSU--A channels
h
l off 6
6,7,8
7 8 iin th
the operational
ti
l
Global
Gl b l analysis
ly i ((GA))
AMSU A channels of 6
AMSU-A
6-8
8 have been assimilated additionally in coastal area since Nov 2010
2010.
Their q
qualities in coastal area, with surface emissivities fixed (0.9)
( ) in RTM calculations,
were found to be similar to those in land area.
The experiment in advance of the operation showed ;
available
il bl d
data
t increased
i
d by
b 20
20-30%
30% (Fi
(Fig2.1.1)
2 1 1)
analyses
y
field g
got closer to other centers’ ones (Fig
( g 2.1.2))
forecast was improved especially in about 100
100-700hPa
700hPa vertical layers
22R
2.2
Radiance
di
assimilation
i il ti
iin th
the operational
p
ti
l Meso
M
Meso-scale
l
analysis
y
((MA))
Radiance data have been assimilated instead of retrievals since Dec 2010.
2010
G was applied for
f the MA with several modifications
f
A quality control used in GA
;
45km thinning distance while 180
180~250km
250km in GA
extrapolating atmospheric profiles with U
U.S.
S standard atmosphere’s
atmosphere s lapse rate above MSM top
height ((~40hPa)
40hPa)
employ VarBC coefficients estimated in the latest GA
Considerable improvements in the tropospheric analyses and short range forecasts (Fig 2
2.2.2)
2 2)
Fig 2.1.2: T850 analysis field verification against independent analysis. Red color
shows analysis fields get closer to ECMWF’s ones in the assimilation experiment.
3.1
surface-sensitive
3 1 Assimilation experiment of surfacemicrowave radiances with RTTOVRTTOV-10 in the GA
TEST run uses RTTOV-10.1,, while CNTL run uses RTTOV-9.3. TEST employs
p y also ;
FASTEM-4 and climatological land surface emissivity
emissivity, both supplied in RTTOV-10
A ti assimilation
Active
i il ti off MHS channel
h
l3
3-5
5 over land
l d
Reduction of observation error inflation for AMSU-A ch6 over land
Use of hourly surface temperature from GSM as the first guess in RTM calc
I
d especially
i ll desert
d
t areas. It gott closer
l
tto ground-based
db
d GPS
Increase
TCWV over lland,
observation than CNTL (Fig
( g 3.1.1))
Forecasts were improved not only in the troposphere
troposphere, but also in the stratosphere (Fig 3
3.1.2)
1 2)
TCWV TEST
JMA
RTTOV‐ RTTOV
RTTOV
RTTOV‐ JMA GSM
9
10
TEST - CNTL
OLD
[h
hPa
a]
analysis field improved
not only data
i
increased
d regions.
i
Fig
g 2.1.1: Difference of assimilated data number of
channel 6. Red shows increments.
AMV
Aqua,Terra/MODIS
q ,
4.
4
Scatterom
meter
RTTOV9 3
RTTOV9.3
Early Analysis : 2h25m
Cycle Analysis:
11h15m(00 12UTC) 5h15m(06
11h15m(00,12UTC),
5h15m(06,18UTC)
18UTC)
MTSAT-2, Meteosat-7,9, GOES-13,15
3.
VIS/IR Ima
ager
NEW
Much better
especially
espec
a y in short
s o
range forecasts
Fig 2
Fi
2.2.2:
2 2 RMSE[
RMSE[m]] off geopotential
i l height
h i h forecast
f
off MSM against
i
radiosonde
di
d observations
b
i
26 2009.
2009 The color indicate the difference of Forecast Time
Time. The left panel
2 1 MSM domain and e
ample of AMSU
A from July 16 to 26,
Fig 2
2.2.1:
example
AMSU-A
corresponds to old system,
system before Dec 2010,
2010 and the right one corresponds to new system.
system
observation coverage(NOAA18,
coverage(NOAA18 Aqua).
Aqua)
3 3 AIRS & IASI clear
l
di
i il ti
3.3
radiance
assimilation
Cl d d
Cloud
detection
t ti and
d Cloud
Cl d top
t height
h i ht estimation
ti ti
1. Window channel (AIRS
(
ch950,, 10.2μm)
μ ) O-B threshold
2 Difference of observed radiances (AIRS 10.2μm
2.
10 2μm – 11.8μm)
11 8μm) threshold
3. The cloud top
p heights
g
are estimated byy Minimum Residual Method (Eyre
( y and Menzel 1989))
with two AIRS channels (13.4μm, 10.8μm)
O-B
O
B of clear
d
data
fforms
Gaussian-like
Gaussian
like
1 day forecast 5-day
1-day
5 day forecast
T
Fig
g 3.3.1: ((Left)AIRS
)
one g
granule and MTSAT image
g at the same
time; (Right) Clear flagged data
Fig 3.3.2: Comparison of cloud top
pressure estimated in 3. and colocated MODIS products
products.
Fig
g 3.3.3: O-B histogram
g
of AIRS
299 (~800hPa) for clear data.
Use of daily
product instead of climatological
values for RTM calculation
y ozone p
g
Climatology
Z
daily
Large biases
L
bi
are reduced
d
d
in ozone sensitive
channels.
channels
Increase of RTM l
layers in RTTOV‐10
i RTTOV 10
Red: better, Blue: Worse
,
Fig 3
3.1.1:
1 1: (Upper left) Mean TCWV field for TEST; (Upper right)
Mean TCWV field TEST – CNTL; (Botttom) Time sequence of
analyzed TCWV at Ground-based
Ground based GPS observation.TEST, CNTL,
OBS
1 2: (Left) Comparison of vertical layers of RTTOV
RTTOV-9
Fig 3
3.1.2:
9,
g ) Improvement
p
RTTOV-10 and GSM. ((Right)
ratio of zonal mean
forecast RMSE against the initials [%]
I d
t assimilation
i il ti experiment,
i
t channel
h
l 2 and
d 3 were newly
l assimilated.
i il t d Ch
In
data
Channell 1 was
used to screen out cloud contaminated data. Improvements
of temperature
forecast for
p
p
Southern Hemisphere were confirmed when compared to runs without FY-3A/MWTS
FY 3A/MWTS (Fig
3 2 2)
3.2.2).
better worse
Fig 3.2.2: Difference of RMSE of
temperature forecast against the
initials
i iti l in
i the
th southern
th
hemisphere
hemisphere.
Colors represent forecast time.
time
T850
Z500
Wspd850
Wspd250
bett
b ter wors
se
According to departure (Observed minus Background) statistics, its quality is almost
q
g y better than AMSU-A ((Fig
g 3.2.1).
)
equivalent
or slightly
FY 3A/MWTS
FY-3A/MWTS
ch3
NOAA-18/AMSU-A
ch7
AIRS assimilation experiment with these configurations performed well (Fig 3
3.3.5)
3 5)
PSEA
3 2 IInitial
iti l assessmentt off FY3.2
FY-3A/MWTS
FY-3A/MWTS
ch2
Metop/AMSU-A
ch5
NOAA-18/AMSU-A
ch5
h5
[K]
Fig 3.3.4: AIRS ch232(13.7μm) Residual O-B distribution
after globally constant bias correction of clear data, averaged
f
from
Feb 1 to Feb 12, 2012. ((Left)
f ) Use off C
Climatological
ozone density; (Right) Use of daily ozone product of JMA
JMA’s
s
chemical transport model
Fig
3.3.5:
Fi 3
3 5 RMSE iimprovementt ratio
ti [%] for
f AIRS assimilation
i il ti experiment
i
t in
i Jan.2011.
J 2011 Solid
S lid circles
i l indicate
i di t
difference
statistically significant difference.
Assimilating
configuration
did not show improvement.
Under investigation.
g IASI almost equal
q
g
p
g
4. Future Plans
Assimilate clear radiances of AIRS and IASI
Implement RTTOV-10
RTTOV 10 into GA and MA,
MA and exploit more land sensitive channels
F th iinvestigation
Further
ti ti off FY
FY-3A/MWTS
3A/MWTS ttoward
d operational
ti
l use. Also
Al FY-3B
FY 3B
Assimilate radiances affected by clouds and rain
<References>
Fig 3.2.1: Comparison of O-B departure’s normalized hisograms.
The
Th statistics
t ti ti was calculated
l l t d with
ith bias
bi corrected
t d O-B
O B ffrom one
month (August 2011) data set.
set
M.Kazumori 2011. Impacts
p
of radiance assimilation in the JMA operational
p
meso-scale analyses
y
and forecasts.
Proceeding
g s of 2011 EUMETSAT Meteorological
g
Satellite Conference, 5-9 September
p
2011, Oslo,Norwayy
M.Kazumori 2012. Assimilation experiments involving surface-sensitivity
surface sensitivity microwave radiances in JMA
JMA’s
s global
data assimilation system
system. CAS/JSC WGNE Res
Res. Activ
Activ. Atmos
Atmos. Oceanic Model
Model. Submitted
M.Kazumori
M Kazumori and H.Murata
H Murata 2012.
2012 Initial assesment of FY-3A
FY 3A microwave temperature sounder radiance data in
JMA global data assimilation system
system. CAS/JSC WGNE Res
Res. Activ
Activ. Atmos
Atmos. Oceanic Model
Model. Submitted