1. Effects of Uncertainty in Cloud Microphysics on Passive Microwave Rainfall Measurements
Ju-Hye Kim and Dong-Bin Shin*
Department of Atmospheric Sciences
Yonsei University, Seoul, Republic of Korea

2. Outline 1. Introduction (motivation)
Methodology (characteristics of different microphysics schemes)
Impacts of microphysics on a-priori databases
Impacts of microphysics on PMW rainfall retrievals
Conclusions

3. Introduction Simulated TB RTM
Current physically-based PMW rainfall algorithms heavily rely on CRM simulations.
Simulated TB
RTM
e.g., Plane-Parallel , MC models
Forward models provide prior information
Cloud water + DSD
Rain water + DSD
Snow + , DSD
Graupel + , DSD
Cloud ice + DSD
Hail + , DSD
Water Vapor Temperature
Cloud Model
* e.g., Goddard Cumulus Ensemble Model (GCE),. ....
Assumptions in some parameters (e.g., microphysics)

4. (1st version, Shin & Kummerow, 2003)
Introduction
CRM-based rainfall retrieval algorithms have been evolved to use CRMs and observations simultaneously.
e.g., The parametric rainfall algorithm: Cloud model + TRMM PR/TMI observations
(1st version, Shin & Kummerow, 2003)
Simulated precipitation field
TB computation
Realistic set of 3-D geophysical parameters are created from combination of TRMM PR/TMI and CRM.
Figure at left is a comparison of surface rainfall from TRMM PR and simulator.
simulated
observed
Once 3-D geophysical parameters are constructed, TB can be computed for any current or planned sensor.
Figure at right is a comparison of Tb from TRMM TMI and simulator.
observed
simulated

5. Obs. Tb vs Sim. Tb
The liquid portion of the
profile is matched, the CRMs
consistently specify ice
particles of an incorrect size
and density, which in turn
leads to lower than observed Tb.
10 GHz H
10 GHz V
19 GHz H
A better choice would be to
continue the development of
the Cloud Resolving Model
physics to insure that
simulations properly match the
observed relationship between ice scattering and the rainfall
column.
19 GHz V
21 GHz v
37 GHz H
85 GHz H/V
37 GHz V
Assumptions in microphysics still have great impacts on CRM+OBS.-based DBs.

6. Introduction Cloud Resolving Model Passive Microwave Simulations
Rainfall
Observations
TRMM field campaigns
The Kwajalein Experiment (KWAJEX)
The South China Sea Monsoon Experiment (SCSMEX)
The TRMM Large-Scale Biosphere-Atmosphere Experiment in Amazonia (TRMM LBA)

7. Zhou et al. (2007) Lang et al. (2007) , Han et al. (2010)
used the GCE model to simulate China Sea Monsoon and compared their simulated cloud products with TRMM retrieval products
Lang et al. (2007) , Han et al. (2010)
Land et al. (2007) compared the calculated TBs and simulated reflectivities from cloud-radiative simulations (GCE model) of TRMM LBA domain with the direct observations of TRMM TMI and PR
Han et al. (2010) also evaluated five cloud microphysical schemes in the MM5 using observations of TRMM TMI and PR
Grecu and Olson (2006)
constructed a-priori database from observation of TRMM PR and TMI only to reduce forward error related to cloud and radiative transfer calculations, and compared their retrieval results to products from GPROF version-6 operational algorithm
Many studies pointed out that CRMs (mainly GCE model) tend to produce excessive ice particles above freezing level and it may bring wrong retrieval results in microwave remote sensing of precipitation.

8. TRMM Observation of Typhoon Sudal
Methodology
Different Cloud
Microphysics
PLIN
WSM6
Goddard
Thompson
WDM6
Morrison
TRMM Observation of Typhoon Sudal
Typhoon Jangmi Simulations with WRF model (V3.1)
36522
36532
Parametric rainfall algorithm
Shin and Kummerow (2003)
Masunaga and Kummerow (2005)
Kummerow et al. (2011)
Six kinds of a-priori rainfall databases !

9. Prognostic variable of Single-moment scheme
PLIN
WSM6
Goddard
Thompson
WDM6
Morrison
Ty Jangmi Simulation with WRF model
Single-Moment
+ Ns, Ng, Nr
+ Nccn, Nc, Nr
+ Ns, Ng, (Nc, Nr)
Double-Moment
Single moment schemes have differences in their cold rain processes (ice initiation, sedimentation property of solid particles).
The microphysical processes related to ice-phase in the WDM6 are identical to the WSM6 scheme.
WDM6 is double moment scheme for (only) warm rain processes and it predicts a cloud condensation nuclei (CCN) number concentration.

10. Typhoon Jangmi Simulation with Six different Microphysics schemes in the WRF Model
Similar distributions of rain and cloud water compared to WSM6
Reduction of snow near and above the melting layer
More rain water and more ice particle than WSM6
Increased rain water below 5 km altitude
Similar distribution of ice particle compared to WSM6
Much more snow
Less rain water
More snow
Less rain water

11. Impacts of microphysics on a-priori databases
Correctness of simulated DBs
PLIN 
Modified Radiative Indices
WSM6 
Petty (1994)
Biggerstaff and Seo (2010)
GCE 
THOM 
For the emission indices, TBs agree well. (The biases at 10 GHz channel from six databases are quite small, especially when the WSM6 and WDM6 schemes are used.)
The simulated and observed databases show relatively large discrepancy at 85 GHz scattering index (Sm).
WDM6 
Simulated Indices
MORR 
PM10
PM19
PM37
PM85
SM85
Observed Indices

12. Difference between Obs. and Simulated DBs
Representativeness of simulated DBs
First EOF vector of Radiance indices
Observed database shows a positive variation for attenuation indices and negative variation for the scattering index
Simulated DBs generally follow the pattern of the Obs. DB. (smaller variability in 10, 19, and 37 GHz attenuation indices. Larger variability in 85 GHz attenuation index).
/ PLIN /
Difference between Obs. and Simulated DBs
/ WSM6, WDM6 /
/ GCE, THOM, MORR /

13. Retrieved rainfall distributions for Ty Sudal
Impacts of microphysics on rainfall retrievals
Orbit : 36537
Retrieved rainfall distributions for Ty Sudal
PR 2A25
TMI 2A12

14. Scatter plots of PR vs retrieved rain rates for Ty Sudal
Retrieved rainfall
PR rainfall

15. Retrieval statistics for different rain types (convective vs stratiform)
PR 2A23
Convective
Mean
Std dev
True
Retrived
Corr
Rms
Bias
Lin
27.82
15.03
36.46
24.82
0.44
36.25 (241.2)
(85.1)
WSM6
11.94
16.93
0.84
28.68 (240.2)
(133.0)
Goddard
12.84
15.74
0.85
28.74 (223.8)
(116.7)
Thompson
13.58
16.80
0.78
29.17 (214.8)
(104.9)
WDM6
12.78
16.87
0.89
27.18 (212.7)
(117.7)
Morrison
12.11
14.55
29.92 (247.1)
(129.7)
2a12
7.84
7.24
0.54
38.63 (492.7)
(254.9)
Yellow : Convective
Blue : Stratiform
Stratiform
Mean
Std dev
True
Retrived
Corr
Rms
Bias
Lin
10.17
7.71
14.52
10.28
0.46
13.59 (176.3)
-2.47 (32.0)
WSM6
10.45
12.14
0.77
9.34 (89.4)
0.28 (2.7)
Goddard
10.21
11.01
0.72
10.02 (98.1)
0.04 (0.4)
Thompson
11.23
12.36
0.65
11.42 (101.7)
1.06 (9.4)
WDM6
10.83
11.97
0.76
9.58 (88.5)
0.66 (6.1)
Morrison
9.73
9.44
0.71
10.29 (105.8)
-0.45 (4.6)
2a12
6.68
5.40
0.52
13.05 (195.4)
-3.50 (52.4)

16. Comparison of averaged hydrometeor amounts
In the databases
Cloud water
Rain water
Snow
Graupel
PLIN
0.26 (24.7%)
0.79 (75.3%)
0.07 (13.5%)
0.44 (86.5%)
WSM6
0.19 (19.5%)
0.78 (80.5%)
0.28 (49.6%)
0.28 (50.4%)
GCE
0.33 (29.5%)
0.79 (70.5%)
0.27 (54.5%)
0.23 (45.5%)
THOM
0.31 (27.9%)
0.79 (72.1%)
0.79 (92.7%)
0.06 (7.3%)
WDM6
0.11 (12.5%)
0.80 (87.5%)
0.16 (36.9%)
0.27 (63.1%)
MORR
0.23 (22.5%)
0.77 (77.5%)
0.45 (73.4%)
0.16 (26.6%)
PLIN ~ Too much graupel
In the retrieval s
Cloud water
Rain water
Snow
Graupel
PLIN
0.30 (25.1%)
0.90 (74.9%)
0.10 (22.0%)
0.37 (78.0%)
WSM6
0.26 (22.0%)
0.91 (78.0%)
0.33 (60.3%)
0.21 (39.7%)
GCE
0.37 (30.5%)
0.85 (69.5%)
0.45 (71.6%)
0.18 (28.4%)
THOM
0.36 (27.1%)
0.97 (72.9%)
1.18 (94.8%)
0.07 (5.2%)
WDM6
0.14 (12.4%)
1.00 (87.6%)
0.21 (42.9%)
0.28 (57.1%)
MORR
0.28(21.9%)
0.98 (78.1%)
0.63 (75.5%)
0.20 (24.5%)
THOM ~ Too much snow
WDM6 ~ Increased rain water and reduced cloud water

17. Conclusions
A-priori databases with six microphysics schemes are built by the WRF model V3.1 and TRMM PR observations and the impacts of the different microphysics on rainfall estimations are evaluated under the frame of parametric rainfall algorithm for extreme rain events (Typhoons).
Major difference in six microphysics schemes exists in their cold rain processes (ice initiation, sedimentation property of solid particles).
PLIN and THOM schemes produce too much graupel and snow, respectively, while the ice processes seem to be comparable to those from WSM6 and WDM6.
This study suggests that uncertainties associated with cloud microphysics affect significantly PMW rainfall measurements (at least for extreme events).  Both intensity and distribution of retrieved rainfalls are better represented by the WDM6, WSM6 and Goddard microphysics-based DBs.
PLIN
WSM6
Goddard
Thompson
WDM6
Morrison