Assimilation of AIRS Radiances in Regional Model and Its Impact on Typhoon Forecast
Yan’an Liua, b, Hung-lung Allen Huangb, Agnes Limb, Wei Gaoa, c
a. East China Normal University, Key Laboratory of Geographic Information Science (Ministry of Education), Shanghai
b. University of Wisconsin-Madison, Cooperative Institute for Meteorological Satellite Studies, Madison, Wisconsin, USA
c. Colorado State University, Department of Ecosystem Science and Sustainability, Fort Collins, Colorado, USA
5. Bias correction
1. Introduction
 Sources: calibration, radiative transfer model and short term forecast
 Increase of extreme severe weather, such as typhoon and rainstorm; Development of high
 Characteristics: varies with time, airmass, scan position, satellite orbit
resolution regional model;
 Variation Bias Correction (Augligné, 2007): Scan angle bias + Airmass bias
 Hyperspectral infrared radiance data can provide high resolution of temperature and
 One month local spin-up to update scan bias and airmass coefficients
humidity profiles;
Time series change: Time dependence
Scan angle bias correction
 Special issues in regional data assimilation: the initial and boundary conditions from the
−1 (a)
1.5
global model; model top of regional model; regional background error covariance (B
0 (b)
−1.1
1
−1.2
Bias (K)
Scan bias (K)
impact of limited data volume on the current method of bias correction;
−0.4
−1.3
0.5
0
−0.5
−1.4
Bias (K)
matrix); the highly variable at each assimilation cycle for the number of observations;
−0.2
−1.5
−1.6
−0.6
−0.8
−1.7
2. Model and Assimilation
−1
−1.5
 Weather Research and Forecasting (WRF); Gridpoint Statistical Interpolation (GSI);
forecast
−20
chn333
−10
0
10
20
30
Scan position
chn787
chn1415
chn1766
40
−2
(a) 35
forecast
100
150
Updates number
200
−1.4
250
50
GDAS
REG
100
150
Updates number
200
250
200
150
100
GDAS
REG
12
Wind speed RMSE (m/s)
250
forecast
(b) 14
30
72h
T-0
50
chn1911
Sea level pressure RMSE (hPa)
T-6
−30
NoBC_GDAS
NoBC_REG
BC_GDAS
BC_REG
300
Number
T-12
−40
350
 Data: all conventional data and AIRS.
6h
−1.2
−1.9
AIRS Chn_253 13.85mm
 Model horizontal resolution: 12 km; Domain size: 917 by 550 by 50; model top: 10 hPa
6h
−1.8
chn252
Community Radiative Transfer Model (CRTM)
−1
25
20
15
10
8
6
4
10
50
5
0
Obs
0
−4
Obs
−3.5
−3
−2.5
−2
−1.5
−1
O−B Bias
−0.5
0
0.5
Cloud detection
Channel selection
1200
通道选取
O-B/O-A analysis
Histogram of
O-B;Looks
symmetrically
1000
280
800
600
0.4
Bias comparisons of O-B and O-A for all 61 AIRS
channels assimilated O-A are close to zero after
assimilation for temperature and moisture channels
270
400
260
24
36
48
Forecast time (hour)
(a)
60
72
0
−0.2
−0.4
0.4
1
Cold tail
−80
−60
−100
−40
−20
O−B Bias
0
20
Bias (K)
0
−120
Analysis Increment
40
230
Control Run (Ctrl): all conventional observations
Experiment Run (Exp): Control Run + AIRS
Distribution
after quality
control
AIRS channel peak around or above model top were
rejected based on sensitivity analysis (McCarty,
2009) 61 channels assimilated: 18 temperature; 11
windows; 20 water vapor; 12 shortwave
12 UTC
500hPa T
4. B matrix tuning
B matrix: Spread out information from observations; Controls % of innovation that makes
20
21
22
23
7
8
9
10
11
12
13
14
15
16
17
18
24
25
26
27
28
29
(c)
0.2
0
Water vapor
−0.2
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
(d)
47
48
49
O−B
O−A
0.2
0
Shortwave
−0.2
−0.4
50
51
52
53
54
55
56
57 58 59 60 61
Index of assimilated channels
12 UTC
Exp
Ctrl
6
Windows
19
850hPa RH
Exp
5
−0.2
0.4
Bias (K)
2500
4
0
−0.4
1500
2000
−1
Central Wavenumber(cm )
3
0.2
0.4
240
2
(b)
−0.4
250
Temperature
0.2
200
1000
12
RMSE comparisons of typhoon Saola’s 72 h
intensity forecast when different bias correction
coefficients were applied
Bias (K)
2378 channels
Selected
Rejected by model top
Rejected by operation
290
220
2
0
72
No cloud detection
Cloud detection
Number
Brightness Temperature (K)
60
6. Impact of AIRS radiances assimilation on typhoon forecast
1400
310
300
36
48
Forecast time (hour)
1
Histogram of O-F before and after
BC Compared with coefficients
3. Quality control
24
Bias (K)
Obs
12
Ctrl
18 UTC
18 UTC
up the analysis; Maintain dynamically consistent increments between model variables.
Impact of B matrix tuning on typhoon forecast
300
Relative humidity
1856
250
1913
300
2308
400
REG
REGM 2969
REGL
500
3850
700
3476
850
3330
nobs
Pressure (hPa)
Pressure (hPa)
400
2959
500
3845
REG
REGM
REGL
700
3473
850
1000
2445
150
2182
200
250
1973
250
1973
300
2376
300
2376
400
REG
REGM 2661
REGL
400
REG
REGM 2661
REGL
500
3377
500
3377
700
3256
700
3256
850
4027
850
4027
2028
1000
1
1.2
1.4
RMSE (K)
1.6
1.8
150
U-wind
200
1000
3
180
4
4.5
RMSE (m/s)
5
5.5
0.4
0.8
1.2
RMSE (g/kg)
1.6
2445
V-wind
3
2182
3.5
4
4.5
RMSE (m/s)
5
5.5
2028
Ctrl
Exp
160
980
970
960
140
120
100
500hPa T
950
850hPa RH
80
940
930
60
0
12
24
36
48
Forecast time (hour)
60
72
0
12
24
36
48
Forecast time (hour)
60
72
Comparisons of 72h track and intensity
forecast for Ctrl and Exp experiment with
the best observations from JTWC
（a）Temperature at 500 hPa: Cold
（b）Relative humidity at 850 hPa: Dry
Exp analysis minus Ctrl analysis
7. Summary
 Limited Area NWP carried out using community models
Tuned
B
matrix
improve track forecast,
compared with untuned B and global B
140
120
100
80
60
2
1813
REG
REGT
GLB
160
Track RMSE (km)
3.5
nobs
Pressure (hPa)
1813
1000
Pressure (hPa)
3325
RMSE profiles
between 24 h
forecast
and
radiosonde
observations for
one week; About
3000 radiosonde
data used to
evaluation
per
level on average.
Sea level pressure(hPa)
Temperature
200
180
Best
Ctrl
Exp
990
1770
nobs
1917
nobs
150
1000
Track error (km)
Estimated from NMC method.
Horizontal scale tuning
 Assimilation of AIRS radiances through quality control, B matrix tuning and bias correction
 Positive impact for clear sky AIRS assimilation both on typhoon track and intensity forecast
Reference: Y. Liu, H. Huang, W. Gao, A. Lim, et al, Tuning of background error statistics through
sensitivity experiments and its impact on typhoon forecast, Journal of Applied Remote Sensing,
0
12
24
36
48
Forecast time (hour)
60
72
2015(9):096051.