1. Hiromu Seko (MRI) , Toshitaka Tsuda and Naoto Yoshida (Kyoto Univ.)
Data Assimilation of Side-looking Radio Occultation by Observing System Simulation Experiment
Hiromu Seko (MRI) , Toshitaka Tsuda and Naoto Yoshida (Kyoto Univ.)
and
Y.-H. Kuo (UCAR)
Thank you. Chairman.
I would like to show you the results of the data assimilation of side-looking radio occultation.
The original idea of the side-looking radio occultation was provided by Prof. Tsuda of Kyoto University, and the information of the occultation data that was obtained from satellite positions were provided from Mr.Yoshida of Kyoto University.

2. Concept of the side-looking observation Low Earth Orbit Satellite
Tangent point
Low Earth Orbit Satellite T3
Forward/backward-looking T2
GPS Satellite P3 P2 P1 T1 P2 P3
Low earth orbit satellite (LEO) observes the signal of GPS satellite from the moving direction of the LEO satellite, because the short shift of the tangent point is required for the precise estimation of the tangent point profiles.
First of all, I would like to explain the concept of side-looking observation.
This panel is the schematic illustration of the conventional observation.
The low earth orbit satellite, LEO, observes the signal of GPS satellite that exists on the moving direction of LEO satellite, because, the short shift of tangent point is required for the precise estimation of the tangent point profiles.

3. Concept of the side-looking observation Low Earth Orbit Satellite
Tangent point
Low Earth Orbit Satellite T3
Forward/backward-looking T2
GPS Satellite P3 P2 P1 T1 P2 P3
Low earth orbit satellite (LEO) observes the signal of GPS satellite from the moving direction of the LEO satellite, because the short shift of the tangent point is required for the precise estimation of the tangent point profiles.
If slant path data is used in the assimilation, it is not necessary to use the data which have the small angle from the moving direction of LEO satellite.
Besides the tangent point data, the slant path data from GPS satellite to LEO satellites can be also assimilated.
If slant path data is assimilated, it is not necessary to use the data of which the angle from the moving direction of LEO satellite is small.

4. Concept of the side-looking observation Low Earth Orbit Satellite
Forward/backward-looking
GPS Satellite
Side-looking
In general, impact of RO data is weak because the slant path data extends several hundred kilometers. For this reason, the total amount of assimilated data should be increased by using ‘side-looking’ data.
Generally speaking, the impact of Radio occultation is weak, because the slant path data extends several hundred kilometers.
For this reason, the total amount of assimilated data should be increased by using ‘side-looking’ data.

5. Concept of the side-looking observation Low Earth Orbit Satellite
Forward/backward-looking
GPS Satellite
Side-looking
In general, impact of RO data is weak because the slant path data extends several hundred kilometers. For this reason, the total amount of assimilated data should be increased by using ‘side-looking’ data.
In this study, the assimilation experiment of side-looking data is performed by ‘Observing System Simulation Experiment’.
If the side-looking data can be used, the improvement of the accuracy of the numerical forecast is expected because the amount of assimilation data is increased.
So, the improvement of rainfall forecast by the assimilation of side-looking data is investigated by Observing System Simulation Experiment .

6. Intense rainfall occurred in the city of Kobe
03-06UTC 28 July 2008
Kobe
Before showing the outline of the experiment, the intense rainfall, target of the assimilation experiment, is explained.
The left panel shows the 3-hour rainfall amount from 3 UTC to 6 UTC on 28 July 2008.
Early summer is the rainy season in Japan. So, heavy rainfalls often occur.
On that day, the intense rainfall occurred in the city of Kobe.
This intense rainfall raised the water level of the river very rapidly, and five people drowned in the riverside park.
We have focused on the intense rainfall that occurred in the city of Kobe.
Intense rainfall raised the water level of Toga river, and five people drowned in the riverside park.

7. Flow chart of Assimilation of conventional data
Initial: Operational Meso-analysis 18UTC 27 July
MSM
Boundary: GSM output
00 UTC 28 July
21UTC
18UTC 27 July
Meso-4DVar system
Analysis => initial condition
GSM
Boundary: GSM output
Conventional data
First, I would like to explain the assimilation result of conventional data.
This panel is the flow chart of assimilation of experiment.
Meso-scale analysis fields at 6 hour before the intense rainfall were used as initial condition of the assimilation.
Conventional data were assimilated for 6 hours by Meso-4dvar system of JMA.
The boundary condition was produced from the output of GSM.
The influence of the assimilation data was investigated by comparing the predicted and observed rainfalls.
Conventional data
JMA-NHM(x=10km)

8. Assimilation results using conventional data NoRO (Conventional data)
Observation
NoRO (Conventional data)
FT=0-3hour
FT=0-3hour
FT=3-6hour
FT=3-6hour
Left panel shows the observed rainfall regions.
The intense rainfall were generated in the western part of Japan.
Right panels are the rainfall region predicted from analyzed fields obtained by assimilation of conventional data.
The rainfall were generated. But, compared with observed rainfall, the rainfall intensity is much weaker.
So, the conventional data is not enough to reproduce the intense rainfall.
Conventional data is not enough to reproduce the intense rainfall.

9. Flow chart of Observation System Simulation Experiment
Initial: Operational Meso-analysis 18UTC 26 July
MSM
Boundary: GSM output
00 UTC 28 July
21UTC
18UTC 27 July
Meso-4DVar system
Analysis => initial condition
GSM
Boundary: GSM output
Conventional data
In the observation system simulation experiment, the simulated occultation data in the last 3 hour of assimilation window were added as the assimilation data.
Next, I would like to explain the simulated slant path data.
Conventional data + Simulated occultation data (slant path data)
JMA-NHM(x=10km)
True data
First guess

10. Number of OC All received data (28 July 2008)
Angle (degree)
Number of OC occ-type=44 (28 July 2008)
Angle (degree)
We checked the occultation data.
-Signal of data of which the angles from the moving direction are less than 60 degrees are received.
-However data was not received properly in half of data where the angles were larger than 50 degrees.
What you can see here are the histogram of the angle from moving direction of COSMIC of 28th July 2008.
Upper panel is the histogram of all received data, lower one is the data of which profiles were provided by UCAR.
Signals of which the angle from the moving direction are less than 60 degrees are received.
But, the half of data, of which the angles were larger than 50 degrees, did not become the profile data.

11. Number of OC occ-type=44 (28 July 2008)
Angle (degree)
73%
14%
Number of OC obtained from satellite positions (28 July 2008)
11%
Angle (degree)
Number of occultation where the angle from moving direction is larger than 60 degrees is about 11% when the same day’s data is used.
Because half of the data where the angles of degrees are not received properly, the data that is expected to be increased by side-looking observation becomes about 18%.
We calculated the positions of the occultation and the angles from the moving direction of LEO satellite from the satellite positions.
Lower histogram shows that occultation data of which the angles from moving direction were over 60 degrees is about 11%.
The number of half of the data of which angles is degrees is added, the data that is expected to be increased by side-looking observation becomes about 18%.

12. Assimilation of ground GPS ( Shoji et al. 2009)
YSSK
CHAN
bjfs
osn2
SUWN
DAEJ
SHAO
tnml
twtf
Because the radio occultation data of which angle over 60 degrees were not observed, the simulated slant path data should be produced from the more accurate fields.
In this study, the assimilation results of ground GPS data was used as the true data in the producing of the simulated slant path data.
These blue points indicate the position of ground GPS data that were used in producing the true data.
These GPS points were deployed by Geographical Survey Institute to monitor the movement of the ground clusters.
The GPS data in Korea and China were also used.
The assimilation period is 30 hour from 06 UTC 27 to 00 UTC 28, much longer than this study's assimilation period.
Mr. Shoji shows that these data are useful to improve the rainfall forecast in his paper.
Assimilation period: 30hour (06UTC UTC 28)
Assimilated data: Conventional data , Analyzed rainfall, PWV obtained by satellites, Wind profiler etc GPS (GEONET(blue triangles) Korean and Chinese GPS data(red dot)

13. Assimilation of ground GPS ( Shoji et al. 2009)
NoGPS (only Convection data)
Observation
03UTC 28
06UTC 28
FT=0-3h
FT=3-6h
GPS+Convection data
Intense rainfall reproduced when GPS data were assimilated. 
Analyzed fields were used as true data
These panels are figures quoted from his results.
Left panel were the observed rainfall regions.
Right upper and lower panels are rainfall regions predicted from the analyzed fields of convectional data, and convectional and GPS data.
The intense rainfall was not reproduced when the convectional data were assimilated,
His results is consistent with our results of this study.
When GPS data and convectional data were assimilated, the intense rainfall were reproduced.
So, these analyzed fields were used as true data.
FT=0-3h
FT=3-6h

14. Positions of Occultations during 12-15 UTC 13 September 2006
Next, I would like to explain the position of occultation data.
This figure is the distribution of occultation obtained from the satellite positions during UTC 13 September 2006.
There are occultation position of which the angle from the moving direction of LEO satellite were over 60 degrees near Alaska.
There are several occultations of which the angle from moving direction were larger than 60 degrees.

15. Positions of Occultations during 12-15 UTC 13 September 2006
So, the occultation data with the angle of 65 degrees, which is indicated by a red circle, was shifted to Japan and used in the estimation of simulated slant path data.
There are several occultations of which the angle from moving direction were larger than 60 degrees.
These occultation data were shifted to Japan.

16. True data - NoRO water vapor at lowest layer
Enlarged map
10km
10km
0km
Sat_dir 65
10km
0km
Sat_dir 65
65deg.
Because the region where the occultations were shifted to are positive, the rainfall is expected to be higher when this data is assimilated.
Next, What I am showing here, is the difference between the true data and the assimilated fields of conventional data.
There were positive regions on the northwestern side of the position where the intense rainfall developed.
These white lines indicates the tangent points between the heights of 0 km and 10 km.
Red lines are the paths that penetrated the lowest tangent points.
Because the difference in where the path of occultation passed were positive, the rainfall is expected to be higher when this data is assimilated.

17. True data - NoRO water vapor at lowest layer
Vertical profiles of D-value (Simulated - First guess)
10km
Height (km)
10km
10km
1.3km
0.9km
D-value (N)
The vertical profile of difference of the true data and first guess data are also plotted.
Below the height of 3km, path-averaged refractive index of true data is larger than that of first guess.
The enhancement of the rainfalls is expected when the occultation data reaches below the height of 3 km.
Below the height of 3km, path-averaged refractivity of true data is larger than that of first guess.
The enhancement of the rainfalls is expected, when this data is assimilated.

18. Conventional data + Occultation data NoRO (Conventional data )
FT=0-3hour
FT=0-3hour
FT=3-6hour
FT=3-6hour
When the occultation data were assimilated and the forecast was performed from the analyzed fields, the rainfall became more intense.
So, this result indicates that the side-looking data is useful to improve the rainfall forecast.
Next I would like to compare the reproduced rainfall and observed rainfall.
As it is expected, the rainfall became more intense when the forecast was performed from the analyzed fields, which were obtained by assimilation of conventional data and occultation data.

19. Conventional data + Occultation data
Observation
FT=0-3hour
FT=0-3hour
FT=3-6hour
FT=3-6hour
The rainfall region was plotted with the same color scale of the observed rainfall intensity.
Though the position shifted westward, intensity and area of the heavy rainfall became similar to the observed one.
Intensity and area of the heavy rainfall was similar to the observed one, although the position shifted westward at FT=0-3hour.

20. FT=0-3hour
+Occultation data lowest data: 0km
Conventional data (NoRO)
1km
2km
3km
FT=3-6hour
+Occultation data lowest data: about 0 km
Conventional data (NoRO)
1km
2km
3km
The impact of the lowest level of the occultation data was investigated by changing its lowest height.
In this case, the improvement of forecast is expected, if the lowest data reached the height of 2km.
This result is consistent with the expectation using the vertical profiles of refractive index.
In this case, the improvement of forecast is expected, if the lowest data reached the height of 2km.

21. Increment of lowest water vapor
FT=0-3hour
+Occultation data lowest height: 0km
Conventional data (NoRO)
1km
2km
3km
+Occultation data lowest height: 0 km
1km
2km
3km
Increment of lowest water vapor
These impacts can be confirmed by the increment of the lowest water vapor.
Water vapor increased when the lowest level of the occultation reached 2km.
This result indicated that these increments of water vapor caused the rainfall.
The improvement of rainfall was confirmed by the increment of water vapor at the lowest layer of the model.

22. Summary of preliminary results Thank you for your kind attention.
When the side-looking data, of which angles from the moving direction of LEO are large, are included in assimilation data, analyzed fields are further improved.
Assimilation of side-looking occultation data are expected as the useful method because it improves the numerical forecast.
However, the problems including hardware are not considered. So, more experiments are needed under the more actual condition.
Thank you for your kind attention.
This is the summary of this study.
When the side-looking data, of which angles from the moving direction of LEO are large, are included in assimilation data, analyzed fields are further improved.
Assimilation of side-looking occultation data are expected as the useful method because it improves the numerical forecast.
However, the problems including hardware are not considered. So, more experiments are needed under the more actual condition.