impacts of occultation data on numerical weather prediction€¦ · gps signals received on a low...
TRANSCRIPT
Impacts of Occultation Data on Numerical Weather Prediction
Hiromu SEKO,Yoshinor i SHOJI, Masaru KUNII, Kazuo SAITOYuichi AOYAMA and Toshitaka TSUDA
1. Assimilation of CHAMP dataa. Tangential point data b. Influence of ver tical er r . cor relationc. Assimilation of path data
2. Assimilation of COSMIC data
GPS Signals received on a low earth orbiting (LEO) satellite are used for an active limb sounding of the atmosphere and ionosphere.
During a rising or setting of a GPS satellite (occultation), the radio rays between the GPS and LEO satellites successively scan the atmosphere (and the ionosphere) from the receiver height down to the surface. A refractive index profile can be retrieved from the time variations of the ray bending angles.
Bending AngleLEO Satellite
Tangent Point
GPS Satellite
Propagation Delay of GPS Signals
Determination of LEO Orbit
Bending of Radio Ray Path
Refractive Index Profile near the Tangent Point
Humi-dity
Tempe-rature
ElectronDensity
Basic Concept of GPS Occultation Measurement
GPS Occultation dataCHAMP/ISDC (GFZ) :Challenging Mini-Satellite Payload for Geoscientific Research and ApplicationInformation System and Data Center
Level 1:Satellite orbit data Surface reference station data
Level 2:Atmospheric phase delay、Position and moving speed of CHAMP and GPS satellite.
Level3(1):Profiles of refractivity,temperature and declination
Level3(2):Profile of water vapor
Refractivity
function of temperature and water vapor
Occultation2
6 373.0106.771Te
TPn
Statistics of occultation data
0
5
10
15
20
25
30
35
40
45
50
105 125 145 165
Distribution of Observation in July 2004
Data number
Data number
Average and RMS of D-value
Hei
ght(
km
)D-value = Observation – First guess
Large bias remained below z=2km.
Domain of MSM RMSEAve.
Lower layer is expected to be more humid than
that of first guess value.
Outline of assimilation and Occultation data
-Target: Precipitation systemon the northern Japan.
-Meso-4DVar system / Hydrostatic MSM.-Assimilation window: 7/16, 09-15JST.-Occultation data :1203JST.
Hei
ght(k
m)
N×106
Vertical profile of N
09JST 15JST
Forward modelAdjoint model
Iteration (about 30 times )
Forecast from the analyzed initial fields
12JST
Grey :first guess
Black;Observation data
Occultation data Ohyu133mm/day
Occultation data
7/16
18
JST
(FT
=3)
7/16
21
JST
(FT
=6)
Observation18-21JST
Assimilation of tangential point data
When the CHAMP data was assimilated, the lower layer became humid and then the precipitation was reproduced.
Observation 15-18JST CNTL+CHAMPConventional data(CNTL)
7/16
18
JST
(FT
=3)
Influence of “Thinning out”
When the CHAMP data
was thinned out, the impact of CHAMP data
becomes small.
→Vertical error correlation.
Observation 15-18JSTConventional data
z=600m without lowest data
z=600m with lowest data
7/16
18
JST
(FT
=3)
z=200m with lowest data
Introduction of “Vertical Obs. Err. Correlation”
(a)Correlation of Obs. Err.(rij) (e) Simplified observation covariance (R)
)3(
)2(
112
211
22222112
1121212
111
2222
)12()()()(_
2211
2121
21
nnnnn
nn
kkkk
kkkkkk
itmdltmdl
itmdltobs
ierrobsi
rr
rrrrr
xxxx
xxxxr
DNtDNDNx
LL
MOMM
ML
L
R
D-value of each height
Correlation of Obs. Err.: NMC method
Observation covariance: Obs. Err. of each height
7/16
18
JST
(FT
=3)
Introduction of “Vertical Obs. Err. Correlation”
When the vertical correlation was introduced, the impact of the CHAMP data
becomes large.
Simulated convective bands extended in the
east-west direction.
Observation 15-18JST Conventional data
With introducingthe correlations
7/16
18J
ST
(FT=
3)Without introducing
the correlations
Data positions of assimilation data
Assimilation of path data
CASE 2
CASE 1
CASE 3
Tangential point
Schematic illustration of estimation of the tangential point data
Hei
ght
(a) Horizontal position of path data
ir:( element number on the path1element=10km)
(b) Vertical position of path data
heig
htH
eigh
t
ir
ir
D-value
D-value
D-v
alue
D-v
alue
(a) D-value on the observed path
(b) D-value on the path normal to the observation.
D-value (observation-first guess) on the lowest path
Domain of MSM
NE SW
SE NW
CASE 2
Correlation
Correlation
CASE 3
70km200m
10km
Gray: first guessBlack;Observed data
Hei
ght(k
m)
(N=Reflactivity-1)×106
(a) Observation dir. (b) Normal dir
D-value (observation-first guess) on the lowest path
Gray: first guessBlack;Observed data
Hei
ght(k
m)
(N=Reflactivity-1)×106
n
RR
kmheightn
ipath
)10(
1point
-N is the total number of element under the height of 10km.
-Observation error of the path data was also obtained by the same way from the tangential point observation error data.
Assimilation of path data Observation 15-18JST
7/16
15-
18JS
T (F
T=3)
(a) CASE3 (PARA) (b) CASE3 (NOR)7/
16 1
5-18
JST
(FT=
3)
(a) CASE2(PARA) (b) CASE2(NOR)
-When the path data was assimilated in CASE2 method, the impact became smaller.
-The information of the lowest level was diluted(?).
7/16
15-
18JS
T (F
T=3)
(a) CASE2+Ver.Corr.(PARA) (b) CASE2+Ver.Corr.(NOR)
(a) Correlation of obs.err.(rij) (b) Simplified obs. covariance (R)
Assimilation of path data withVertical Obs. Err. Correlation
-When the path data was assimilated with vertical correlation, the impact became larger.
-This method is best one we investigated so far.
-Obs. covariance of path data was calculated by the aforementioned way.
Intensive ground validation sitesIndia, Tirupati
MST radar obs.
Malyasia, Vietnam, Singapore, Indonesia
Met. Office
NICT, Okinawa obs.
Kyoto-U, ShigarakiMU obs.
Distribution of GPS occultation data
D_value D_value/N (%)
Blue:Bias(low axis) red:RMS (low axis) Black:number of data(upper axis)
Statistics of COSMIC data
D_value = Observation – First guess
Observation error of path data(a)obs. error → (b) simplification → (c) ratio of path length → (d) path obs. error
y = -1.7735x + 14.176
y = -0.0964x + 10.132
y = -0.6966x + 7.5121
-6
-3
0
3
6
9
12
0 2 4 6 8 10
-15000
015000
30000
(a)
(c)
(b)
(d)
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
-50 -30 -10 10 30 50
y = -0.0377x + 0.9819
Vertical correlation of path data(a) correlation coefficient → (b) simplification → (c) covariance of obs error.
(a) (b)
(c) (d)
8888
99
9
10
9
10
101010
10
10
101010
9
101010
9
101010
9999
99
10
4
4
10101010
99
10
9
1010
4
4
10101010
9
1010
8
99
8
99
1010101010101010
1010
1010101010101010
1010
101010101010
6
1010
8
6
1010
8
9
8
9
89
9
9
1010
9
1010
899
8
8
99
8
9
8
99
85
88
8
9
9
8
99
85
88
8
9
10
8
10
8
8
97
1033
8
97
1033
88
8
99
10101010
10
9
7
999
9
10
9
1010107
99
1010
9
101010
9
10999
101010
101010101010999
10999
99
9
1099
1088
99
8
9
99
1088
8
99
101010
9
101010
1010
10
9
1010
10
9999
77
6
9
8
9
9
6
9
8
9
9
88
991010
9
10
9
10888
8
88
99
99
1010
10
1010
9
10
9
10888
899
10
1010
9
1010
10
9
1010
10
8
99
8
1010
9
1099
10
999
99
1010
9
1099
10
999
8
7
1010
7
8
7
1010
7
10
888
10
888
8
8
1010
8
8
1010
9898
55
8
9
8
9
10
10
10
4
9
10
10
10910
10
4
9
10910
1
1010
8
1
999
1010
8
999
8
10
8
1010
1010
8
10
8
1010
1010
9
8
9
5
1010
9
0
8
9
5
10100
88
10
88
10
10
888
3
10
888
3
8
8
8
8
99999999
9
8
19
8
1
10
88
9
1010
88
9
1010
7
8899
8
9
9
7
8899
8
9
9
99
5
7
101099
5
7
1010
99
1
1010
10
99
1
1010
10
9
77
9
7
8
8
8
8
99
99
8
99
8
7
9
7
9
9999
8
1010
8
8
1010
8
4
8
4
8
9
9
99999
9
9
99999
10101010
99
77
88
8
9
8
9
8
10
8
10
8
10
8
101010
99
10
9
101099
10
88
9
8
1010101010
992
10101010
99
101010
8
8
99
1099
99
1010
99
8
999
99
9
9
999
10101010
109
9
10101010
101010
6
9
99
9
7
999
9
99
88
8
20
25
30
35
40
45
50
105 110 115 120 125 130 135 140 145 150 155 160 165
Lowest height : 10-low(km)Position of COSMIC data (1-15 Sep.)
Case study 15 Sep. 2006
FT=3 FT=6
3UTC 6UTC
•Position of simulated rainfall was shifted nor thward.
•COSMIC data passed the nor thern par t of Kyushu, where simulated rainfall was too developed.
6UTC
FT=6
n
n
heig
ht(k
m)
31.5
32
32.5
33
33.5
128 128.5 129 129.5 130
Qv (with-w/o_COSMICx01) T (with-w/o_COSMICx01)
Qv (with-w/o_COSMICx01) T (with-w/o_COSMICx01)
lon.
lat.
大きさが高度示す。
n
n
heig
ht(k
m)
T (with-w/o_COSMICx01)
Qv (with-w/o_COSMICx01)
3hour rainfall (FT=6hour)w/o COSMIC with COSMIC with-w/o COSMIC
• When COSMIC data was assimilated, rainfall at the nor thern par tof Kyushu became weaker .