1. The Global Satellite Mapping of Precipitation (GSMaP) project: Integration of microwave and infrared radiometers for a global precipitation map
Tomoo Ushio*, K. Okamoto, F. Isoda, Y. Iida (Osaka Prefecture Univ.)
K. Aonashi, T. Inoue (MRI)
N. Takahashi, T. Iguchi, H. Hanado (NICT)
K. Iwanami (NIED)

2. Outline GSMaP Project in Japan Modification of Aonashi’s algorithm
Objectives
Outline
Modification of Aonashi’s algorithm
Comparison with 2A12
Integration of IR into MWR data
Initial results

3. Global Satellite Mapping of Precipitation Project (GSMaP)
More than 20 scientists from 8 institutions in Japan. (NICT, MRI, JAXA, Osaka Pref. Univ., Tokyo Univ., Shimane Univ., Hokkaido-Tokai Univ., NIED)
(PI: Ken’ichi Okamoto, Osaka Prefecture University)
・To produce highly accurate and high spatial resolution maps of global precipitation by mainly using satelliteborne microwave radiometer.
　 - Everyday, 0.25×0.25 degree/ 1 day resolution.
- Microwave Radiometer (TRMM, DMSP×3, Aqua, ADEOS-II),PR(TRMM)
　 - Integration to Geostationary satellite IR data
・To develop reliable microwave radiometer algorithm
- Consistent algorithm with PR algorithm based on the common physical model of precipitation
・To establish technique to produce rainfall map by using satellite-borne microwave radiometer data for the future mission(GPM) 　

4. Outlines of the GSMaP project
Consistent Algorithm Based on Common Physical Model of Precipitation
＊Microwave Radiometer
＊Rain Radar
＊Combined
Improvement of
Algorithm
Global Rainfall
Map
＊Vertical Profile
＊Melting Layer Model
＊Z-R Relation and Rain Type
＊Non uniformity
＊Snow (Dry, Wet)
＊Attenuation by Cloud
Improvement of Physical Model of Precipitation
Satellite Data
(TRMM, DMSP, Aqua, ADEOS-II)
Geostationary
Satellite IR Data
CRL（Precipitation Radar：
5 GHz, 13.8 GHz, 95 GHz,
Wind Profiler: 400MHz)
National Research Institute for
Earth Science and Disaster Prevention（Precipitation Radar：10 GHz, 35 GHz, 95GHz )
Ground-based Radar
0.25 deg/ 1day precipitation map only from the microwave radiometers
0.1 deg/ 1hour precipitation map from combined IR and MW radiometers

5. TRMM/TMI　 1998年7月の1ヶ月平均の例
Our algorithm
GPROF
2mm/h

6. Error near Himalayan mountain
85GHz scattering data base without precipitation from TRMM/PR observation

7. Zonal mean of precipitation
Black : TRMM/PR Green: GPROF Red: our algorithm

8. Black : TRMM/PR Green: GPROF Red: our algorithm

9. Production of Global Precipitation Map
Global precipitation map(3hours, 1 week)
Precipitation data
TRMM
Aqua　　
ADEOS2
DMSP/F13
DMSP/F14
DMSP/F15　　　
Merging
Coordination
With 1 day resolution, the 6 satellites cover the whole globe.
We still have sampling error especially when we think of higher resolution such as 3 hours/0.1 degree.
1-Day covering area of 6 Satellites with microwave radiometers(TRMM, Aqua, ADEOS2, DMSP(F13, F14, F15))

10. The area around Japan has mean rain rate 0
The area around Japan has mean rain rate 0.2mm/h with Radar-AMeDAS Composites Data.
Grid box size 100km
Grid box size 500km
Average Sampling Error of rainfall per each period by Five Sun-Synchronous-Orbit Satellites’ Group plus TRMM(TMI) in 500km and 100km grid box using Radar-AMeDAS Composites Data during 36 months. (Y. Iida, 2003)

11. Needs for the combined algorithm
If we want the 0.1 deg./1 hour resolution precipitation map only from 6 currently available MWR, we would have more than 500% sampling error.
In order to reduce these errors, we use the Infrared Radiometers (IR) data which do not have any sampling problems. (large algorithm errors)

12. Moving vector approach
This method was recently introduced by Joyce et al. [2004].
Advantage
MWR based approach (not Tb but cloud motion)
Fast processing time
Disadvantage
Physically simple assumption (not based on dynamics and thermodynamics)

13. What, When, Where, and How do we analyze for?
Purpose: To draw the global precipitation map with degree/1 hour resolution
What: 1hour global IR data from Goddard/DAAC and TMI/2A21 data
When: August 3 to 4, 2000
Where: -35 to 35 in latitude, 0 to 360 in longitude
How: By interpolating precipitation between TMI overpasses using the cloud motion inferred from 1 hour IR Tb.

14. 降水マップ作成 Infrared (IR) Data Microwave Radiometer (MWR) Data 1 hr MWR
10.8 μm Geo IR
Present
Split Window
11.4 μm Geo IR
Present
11.4 μm Geo IR
1 hour before
1 hr Moving Vector
Intermediate data
Microwave Radiometer (MWR) Data
Kalman filtering
1 hr MWR
Present
GSMap Data
GSMaP
1 hour before
GSMaP

18. Summary and future directions
GSMaP project in Japan was introduced.
Precipitation estimates by TMI/Aonashi’s algorithm shows good correlation over land and ocean. But we still have ice/snow covering problems.
Initial results of the global precipitation map with 0.1 deg./1 hr resolution from IR and TMI combined algorithm were introduced and demonstrated.
We are going to examine the possibility of using the bi-spectral and Kalman filtering approach.
Lightning data also will be included.

19. Thank you !
ありがとう! 謝謝
Vielen Dank
Merci
Gracias
Grazie

20. Kalman Filterによる最適解

21. カルマンフィルタによる定式化 状態 方程式 観測方程式 移動ベクトルによる 雨域行列 推定雨量 初期値（真値） 誤差
マイクロ波放射計データと
レーダアメダスとの誤差
観測方程式

22. High quality precipitation map
Global precipitation map only from microwave radiometers
・Improvements of Aonashi’s algorithm
・Compariton betwenn ＴＲＭＭ/PR・TMI(GPROF and Aonashi’s algorithm)
・Application to AMSR-E, SSM/I data ・
Ｖａｌｉｄａｔｉｏｎ
High quality precipitation map
・Interpolation from cloud motion
・Split window analysis
・Combination of microwave radiometers
・Sampling error analysis
Validation by using the Radar network in Japan
Integration of MW and IR data

23. Comparison between GPROF and our algorithm for the TMI data