1. Current Status of GSMaP Project and New MWI Over-land Precipitation
６th　IPWG Workshop
Current Status of GSMaP Project and New MWI Over-land Precipitation
Retrieval Algorithm
Kazumasa AONASHI (MRI/JMA),
Satoshi KIDA, Takuji KUBOTA,
Misako KACHI, Riko OKI (EORC/JAXA),
Tomoo USHIO (Osaka Univ.),
Shoichi SHIGE (Kyoto Univ.)
Ken’ichi OKAMOTO (Tottori Univ. of Envi. Studies)
2012/10/16

2. Current Status of GSMaP Project
Current Activities
2012/10/16
6th IPWG Workshop

3. Global Rainfall Map Processing System at JAXA/EORC
Near real time and high-resolution global rainfall map based on satellite observation
Distribution
produced 4 hours after observation and updated every hour
Produced Rainfall Map
Distribution

4. MWR precip retrieval algotrithm
GSMaP Project
TRMM
TMI
AMSR-E
AMSR2
DMSP
SSM/I,SSMIS
NOAA
AMSU
MWR precip retrieval algotrithm
L2 Product from each sensor
Geo. Satellite
Mix
Infrared radiomter
Cloud motion vector
As Aonashi-san pointed out, microwave radiometer is an excellent tool for the accurate rainfall estimation. Up untill now, we have four sensors. The TRMm/tmi, aqua/amser-e, adeos/amsr, and dmsp/ssmi .by combininb these sensosrs, we have the composite product.
As is showed earlier, currently we have several satellites aboard microwave radiometere available now
Near Real Time
System
At JAXA/EORC
L3 Composite
product
0.1degree
1 hour.
6 hour
1 day
1 month
Composite algorithm of
IR and MWR
0.1degree・30min
5th IPWG Workshop 5 5

5. Current Activities Algorithm Development System Improvement
New MWI over-land algorithm (Aonashi)
Orographic Rainfall (Shige, Session 5)
MWS algorithm (Kida, Poster session 2)
System Improvement
AMSR2 on GCOMW1 (Kachi, Poster session1)
Data Assimilation (Aonashi, Poster session2)
2012/10/16
6th IPWG Workshop

6. Orographic Rainfall (Shige & Taniguchi, 2010)
Warm rain with less ice precipitation
Freezing Level
Orographic
A large amount of condensates
Low-level orographic lifting
Ice
Water
Maritime Air
As Shige and Taniguchi pointed out, warm rain with small frozen precip can induced by the low-level orographic lifting. In that case, we cannot detect this type of rain, even for very heavy precip.
Mountain
Ocean
2012/10/16
6th IPWG Workshop 8

7. Detection of Orographic Rain (Shige & Taniguchi, 2010)
Topographically induced upward motion
Low-level moisture convergence
Slope from GTOP
Winds and Water Vapor from GANAL
Winds from GANAL
Shige and Taniguchi also discussed the possibility of detection of orographic rain using indices of topographically induced upward motion and low-level moisture convergence calculated from the NWP analysis data. So, we would like to introduce their idea to our algorithm in the future.
2012/10/16
6th IPWG Workshop

8. Current Activities Algorithm Development System Improvement
New MWI over-land algorithm (Aonashi)
Orographic Rainfall (Shige, Session 5)
MWS algorithm (Kida, Poster session 2)
System Improvement
AMSR2 on GCOMW1 (Kachi, Poster session1)
Data Assimilation (Aonashi, Poster session2)
2012/10/16
6th IPWG Workshop

9. Revised GSMaP_MWS Original GSMaP_MWS Revised GSMaP_MWS
Revised GSMaP_MWS can detect warm rain.
PR 2A25

10. Current Activities Algorithm Development System Improvement
New MWI over-land algorithm (Aonashi)
Orographic Rainfall (Shige, Session 5)
MWS algorithm (Kida, Poster session 2)
System Improvement
AMSR2 on GCOMW1 (Kachi, MT special session)
Data Assimilation (Aonashi, Poster session2)
2012/10/16
6th IPWG Workshop

11. 2012/10/16
6th IPWG Workshop

12. Current Activities Algorithm Development System Improvement
New MWI over-land algorithm (Aonashi)
Orographic Rainfall (Shige, Session 5)
MWS algorithm (Kida, Poster session 2)
System Improvement
AMSR2 on GCOMW1 (Kachi, Poster session1)
Data Assimilation (Aonashi, Poster session2)
2012/10/16
6th IPWG Workshop

13. Ensemble-based variational assimilation method to incorporate MWI TB into a cloud-resolving model

14. New MWI over-land algorithm
GSMaP over-land Algorithm　＆　Problems in over-land retrieval
Introduction of MWI indices
Dependency of the retrieval bias on the MWI indices
Statistical correction using MWI indices
This is the outline of my today’s talk.
As the introduction, I will discuss the problems in conventional over-land algorithm
Next, I will review the sensitivities of TB depressions to precip variables.
Based on this results, we developed the new over-land MWI algorithm.
So, I will briefly talk about the algorithm.
We also talk about the validation results of the new algorithm using TRMM data for 2003.
Then, I will discuss future directions and summary in the end.
2012/10/16
6th IPWG Workshop

15. Basic Idea of the Retrieval Algorithm
PCT37, PCT85　over land
Observed
TBs
Retrieval Calculation
Forward calculation
RTM
Screening
Inhomogeneity
estimation
Scattering part
Emission part
Ts, Temp
Precip Profiles
DSD
Mixed phase
inhomogeneity
GANAL
Look-up
Table
Statistical
Precip-related
Variable Models (PR)
The basic idea of the GSMaP over-land algorithm is to find precipitation rates that give forward-calculated TB depressions at 37 and 85 GHz fitting best with the microwave imager (MWI) observation.
This algorithm also used statistical precip-related variable models in the forward calculation part.
Precip.
Find the optimal precipitation that gives RTM-calculated TBs
fitting best with the observed TBs:

16. PR Rainsurf vs conventional GSMaP retrievals for July ‘98
Over Ocean
Over Land
Rainspc (mm/hr)
Rainspc (mm/hr)
This slide shows scatter diagrams between PR surface rain (Rainsurf) and TMI retrievals using GSMaP algorithms over land and ocean for July 1998.
This indicates that the GSMaP over-land algorithm underestimates the Rainsurf
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
2012/10/16
6th IPWG Workshop

17. Basic Idea for Algorithm Improvement
A priori information error is assumed to be the main cause of the retrieval bias.
Introduction of MWI indices which the retrieval bias depends on.
Correction LUT based on the above dependency.
2012/10/16
6th IPWG Workshop

18. Summary of forward calculation experiments
Sensitivities of TB depressions to precip variables
TB85 depression was very sensitive to frozen precip properties (Dtop, PSD, shapes)
TB37 depression was sensitive to FLH and Rain DSD in addition to frozen precip properties.
Freq
(GHz)
Depth of frozen precip
PSD of frozen particles
Non-spherical Particles
Freezing level height (FLH)
DSD of rain 85 ◎ ○ × 37
To find the cause of the retrieval error, we performed forward calculation experiments and examined the sensitivities of TB depressions to precip variables.
This table summarizes the results. This indicates:
TB85 depression was very sensitive to frozen precip properties (Dtop, PSD, shapes)
TB37 depression was sensitive to FLH and Rain DSD in addition to frozen precip properties.
2012/10/16
6th IPWG Workshop

19. Index of Dtop, R8537
R8537 expressed as ratio of precipitation retrieved from TB85 (Rain85) to TB37 (Rain37) using the conventional GSMaP algorithm.
TMI R8537 increases with Dtop estimated from PR.
Retrain.v
match-up data　(Land ‘98)
This slide shows ratio of TMI TB85 depressions to TB37 depressions (R8537) that expressed as retrieved precip, Rain37 and Rain85, plotted in terms of Dtop.
This indicates that R8537 increases with Dtop.
We introduced R8537 as the index of Dtop.
2012/10/16
6th IPWG Workshop

20. Index of FLH: PCT37 with no rain
For each rainy pixel, PCT37 with no rain (PCT37nr)
is derived from surrounding no rain pixels.
GANAL tends to over (under)estimate PCT37nr
over cold (hot) regions.
PCT37nr
CPCT370m
CPCT370m
PCT37nr

21. Rain85 vs PR rainsurf depending on R8537 (1998, over Land )
R8537
> 2.1
Rain85 (mm/hr)
Rain85 (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
R8537
< 0.7
R8537
Rain85 (mm/hr)
Rain85 (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
Retrieval bias of Rain85 is mainly due to Dtop error

22. Rain37 vs PR rainsurf depending on PCT37nr (1998, over Land )K
PCT37nr
> 300 K
Rain37 (mm/hr)
Rain37 (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
PCT37nr K
PCT37nr
< 280 K
Rain37 (mm/hr)
Rain37 (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)

23. dPCT37nr=(PCT37nr-CPCT370m)
Rain37 vs PR rainsurf depending on dPCT37nr PCT37nr ~( K) (1998, over Land )
dPCT37nr
> 5 K
dPCT37nr
0-5 K
Rain37 (mm/hr)
Rain37 (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
dPCT37nr
< -5 K
dPCT37nr
-5 – 0 K
Rain37 (mm/hr)
Rain37 (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
dPCT37nr=(PCT37nr-CPCT370m)

24. Rain37 vs PR rainsurf depending on PCT37nr dPCT37nr= (-3 ~ +3 K)(1998, over Land )
CPCT370m)
Rain37 (mm/hr)
Rain37 (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
PCT37nr K
PCT37nr
< 280 K
Rain37 (mm/hr)
Rain37 (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
Forward calculation error is main cause of Rain37 biases.

25. New Over-Land Algorithm: Statistical correction of LUTs using (PCT37nr,R8537)
TRMM data sets for 1998 are classified by (R8537,PCT37nr) .
Linear fitting coefficients between Rain37, Rain85 and PR surface precipitation rates.
For statistical correction of LUTs, we classified TRMM data sets for 1998 with R8537 and Sigma85 for each precip type.
Then we derived linear fitting coefficients between Rain37, Rain85 and PR rainsurf for each class.
The new over-land algorithm used these coefficients for the correction of LUTs.
2012/10/16
6th IPWG Workshop

26. Over-land retrieval for 2004
Validation Results
Over-land retrieval for 2004
We used TRMM data sets for 2003 to validate the new over-land algorithm.
2012/10/16
6th IPWG Workshop

27. Comparison of over-land retrievals Rainsurf vs. Rainspc over Land (Jul
Conventional Algorithm
New Algorithm
Rainspc (mm/hr)
Rainspc (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
The upper figures show scatter diagrams between PR Rainsurf and TMI rain37 over land for Jul
The lower figures show zonal means of PR Rainsurf and TMI rain37 over land for the same period. Green lines are those of PR Rainsurf, while red lines are those of TMI rain37, respectively.
This indicates that the negative bias was reduced by the new over-land algorithm.
Rainspc (mm/hr)
Rainspc (mm/hr)
2012/10/16
LAT
6th IPWG Workshop
LAT

28. Comparison of over-land retrievals Rainsurf vs. Rainspc over Land (Jul
Conventional Algorithm
New Algorithm
PCT37nr
> 300 K
Rainspc (mm/hr)
Rainspc (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
PCT37nr K
Rainspc (mm/hr)
Rainspc (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
2012/10/16
6th IPWG Workshop

29. Comparison of over-land retrievals Rainsurf vs. Rainspc over Land (Jul
Conventional Algorithm
New Algorithm
PCT37nr K
Rainspc (mm/hr)
Rainspc (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
PCT37nr
< 280 K
Rainspc (mm/hr)
Rainspc (mm/hr)
PR Rainsurf (mm/hr)
PR Rainsurf (mm/hr)
2012/10/16
6th IPWG Workshop

30. Daily precip (mm/day) of rain37 and their difference from PR rainsurf: over land for Jul. ‘03
Conventional Algorithm
New Algorithm
This slide shows the global distribution of daily precipitation of TMI rainpct37 and their differences from PR rainsurf over land for July 2003.
This indicates negative biases, particularly, in Himalaya and South America were alleviated with the new algorithm.
2012/10/16
6th IPWG Workshop

31. Summary Current status of GSMaP Project New over-land MWI algorithm
Current Activities
New over-land MWI algorithm
Over-land algorithm and problems in retrieval
Introduction of MWI indices
Statistical correction using MWI indices
Over-land retrieval for 2004
Future directions
Find reason for PCT37_nr dependency
The outline of my today’s talk is as follows:
2012/10/16
6th IPWG Workshop