CMA Applications of Radiative Transfer Model in Product Generation and Sensor Monitoring 

Qifeng Lu, Fuzhong WengNSMC, CMA, luqf@cma.gov.cn

Chunqiang Wu, Juyang Hu, Min Min, Wengguang Bai, Ling Sun, Fangli Dou, Yong Zhang, Yang Guo, Dawei An, Chengli Qi, Shengli Wu, Miao Zhang, Hui Liu, Xiaoqing Li

Lin Chen, Yuan Li, Liqing Hu

Outline

2

Background

Model Development

Calibration and validation

Sensor Monitoring

Product generation and validation

Outline

3

Background

Model Development

Calibration and validation

Sensor Monitoring

Product generation and validation

引自：Gerald van der Grijn (2004), Met-OP Training Course: Use and Interpretation of ECMWF Products.

4

Satellite Measurement and Calibration

BRDF、Emmisivity()

ATM

Absorb, scatter, reflect

_ _(1 )( )atm isatm ii iRR

_atm iR _satm iR s

bi i iR

( )i iB T ( )i i s iB T

Radiative Transfer Eq：

O

可见定标器

碲镉汞探测器

主镜次镜

折镜

分离镜 硅光探测器

定标平面镜

黑体定标透镜

第一中继透镜

第二中继透镜

滤光片

output

Count（DN）

计数

值(D

N)

入瞳辐射 (R)

g 斜率

b 截距

DN = g*R + b

Calibration

Problem of Space Sensors：

（1）The calibration system is not good enough；（2）No absolute calibration system, no reference；（3）Not good enough with the  stability；

Attenuation of NOAA reflect Channels in 3 years

Attenuation of FY-1 reflect Channels in 5 years 5

Calibration: to Improve Satellite Data quality 

Reffence

Before Correction

After Correction 6

The RTMs are widely used at NSMC

Surf Info 

ATM Info

Sat Info

Observation

Radiance

Count

Forward andAD Model

ForwardJacobian Cal/Val

Monitor

Product Gen

Product Val

Others

SatSensorSat

Sensor

Outline

8

Background

Model Development

Calibration and validation

Sensor Monitoring

Product generation and validation

Model Development

• Models used at NSMCRTTOV, CRTM, ARMS, LBLRTM, MonoRTM, 6S, ModTran, FYRTM…

• Models in developing at NSMCFast Model for High Spectral RadianceFor lunar spectral irradiancesFor Cloud scattering

Case1 Rayleigh (Rayleigh only, NNUM=4, SZA=60, VZA=0.0, SAZ=0, VAZ=180, PS=950, PLOC=0.9, SIGMA=0.05, ALBEDO=0.1, AOD=0.0, US STANDARD)

Compared with VLIDORT、SCIATRAN

Near Infrared Spectral Calculation with Polarization

Simulation of TOA lunar spectral irradiances

Figure 1. Simulated TOA lunar spectral irradiances at lunar phase angles =0°, 45°, 90°, and 135° and NPP/VIIRS‐DNB SRF (black solid line) (a), and its corresponding normalized weighting values of DNB (b).

Periodical Characteristics of Lunar Irradiance

Min, et al, 2017. JGR

Modified 2.25μm Channel in Cloud remote sensing

Wang, et al., 2018. IEEE Transactions on Geoscience and Remote Sensing

Why are the Adjoint Model Important? • For physical retrievals

Sensor Physical retrieval

Brightness Temperature

Air Temp

Humidity

pSurface Temp

Radiance Flux

… …

AdjointModel

pInstrument paras

The potentials of Adjoint Model

Forward Model

ATM ProfilesInput Output

Radiance

Tangent linear Model

ATM Profiles Disturb

Input Output Radiance Disturb

AdjointModel

Radiance Disturb

Input Output ATM Profiles Disturb

Temperature Jacobians for VASSFY3C VASS FY3B VASSNA18 ATOVS

US Standard，IR EMIS=0.98，MW EMIS=0.68

qR

ln

ozR

ln

Jacobians for FY‐4 GIIRSFYRTM/FY4 GIIRS RTTOV/IASI

US Standard，IR EMIS=0.98

Outline

17

Background

Model Development

Calibration and validation

Sensor Monitoring

Product generation and validation

Cal/Val： FY‐3D GAS Spectrum in TVAC Test

HIRAS OBS compared with LBLRTM simulation （LW）

Clear pixels determined from MERSI cloudmask

MERSI‐II CloudMask

Frequency biases relative to LBLRTM will be derived; ILS & laser parameters may be adjusted

Cal/Val： To estimate FY‐3D HIRAS Spectral Calibration Coefs

5 10 15 20 25 30 35 40-20

-15

-10

-5

0

5

10

15

20

Sample

Spec

tral B

ias/

ppm

LW Spectral Bias

FOV-1FOV-2FOV-3FOV-4

Stability

412 nm Magnitude of Polarization correction

412 nm O-B412 nm Polarization Effect

Polarization effect estimation

The polarization effect of FY-3 MERSI could be estimated using the simulation of FY-3 MERSI under the clear sky open sea pixels

Cal/Val： Correction of Polarization effect for MERSI

FY-3A MERSI

0 500 1000 1500 2000 25000

20

40

60

EV-SV

Ref

Fact

or(%

)

FY-3B MERSI

B3:1370<=DSL<1390 Y=-1.77e-007*X2+0.0272*X R2=0.99 M=-0.017 S=0.723 N=67

DunhuangCQinghaihuCDunhuangLibya1Libya4Arabia2LanaiAlgeria5SonoraAlgeria3Mauritania2

RTM + Campaign OBS for Cal/Val

Old New

Surface Lambert BRDF

Scatter Scalar Vector

Trans Band model Mid-Res

Uncertainty

6% 3~5%

Sea Color

aerosol

Before  After

Cal/Val: campaign for FY‐3c/d MWHSSurface ATM Profiles Surface OBS

RTTOV Forward Model

MWHS II Simulation

MWHS II L0

MWHS II SDR

On board Precision

过境普洱中心时刻北京时

卫星编号仪器天顶角（度）

2018/3/3 14:08:40.6 FENGYUN 3D 46.88

2018/3/3 23:06:56.5 FENGYUN 3C 15.1 2018/3/4 11:35:13.3 FENGYUN 3C 13.58 2018/3/5 11:16:20.6 FENGYUN 3C 22.65 2018/3/5 15:10:09.4 FENGYUN 3D 50.87

2018/3/6 14:51:58.5 FENGYUN 3D 25.582018/3/7 14:33:04.3 FENGYUN 3D 9.3 2018/3/8 3:01:21.0 FENGYUN 3D 10.49 

2018/3/10 11:22:00.3 FENGYUN 3C 11.36 2018/3/12 14:38:37.1 FENGYUN 3D 1.81 2018/3/13 3:06:54.0 FENGYUN 3D 1.012018/3/14 22:59:22.4 FENGYUN 3C 1.44 2018/3/15 11:27:39.7 FENGYUN 3C 0.67

Campaign Validation flaw

Sounding Profiles

亮温

差（

K）

Blue: FY‐3C MWHSGreen: FY‐3D MWHS Yellow: Specification

Outline

23

Background

Model Development

Calibration and validation

Sensor Monitoring

Product generation and validation

Flow Chart of RTM SAT‐Simulator

RTMs：CRTM, RTTOV, 6S etc

NWP：FNL、ERA 5、 CRA‐40 etc

Clear Sky only

Monitoring instrumental performance using simulations 

118.750.3 GHz

Web: : http://satellite.nsmc.org.cn

MWTS channel 6:  54.94GHz

北京华云星地通科技有限公司 26

定制统计：

More details for the abnormal diagnosis 

Statistics of MWRI O‐B

The improved OMB for channels from UKMO

O‐B of MWTS Sounding 

Channel, the Surface 

Information exists.

T=Tj+k(T1‐Tj);j=5,6,7,8

Surface Info in MWTS Sounding Channels

The attenuation effect could be detected and corrected

121 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 70.8

0.9

1

1.1

2010 2011 2012 2013 2014 2015 2016 2017

Date

Norm

alized

Res

pons

e

FY-3B MERSI

121 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 70.6

0.7

0.8

0.9

1

1.1

2010 2011 2012 2013 2014 2015 2016 2017

Date

Norm

alized

Res

pons

eFY-3B MERSI

121 3 5 7 911 1 3 5 7 9111 3 5 7 911 1 3 5 7 9111 3 5 7 911 1 3 5 7 911 10.7

0.8

0.9

1

1.1

Norm

alized

Res

pons

e

2010 2011 2012 2013 2014 2015 2016

121 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 7 9111 3 5 70.7

0.8

0.9

1

1.1

2010 2011 2012 2013 2014 2015 2016 2017

Date

Norm

alized

Res

pons

e

FY-3B MERSI

B03B04B13B14B15B16

B01B02B08B09B10B11B12B20

B17B19

B06B07

Monitoring of Reflected Channels of Imager 

Outline

32

Background

Model Development

Calibration and validation

Sensor Monitoring

Product generation and validation

RTMs in Product Generation

Simulation – Look Up Tables FWD and AD/Jacobian Matrix in physical Retrievals

Retrieval

Radiative Transfer Model

Look‐Up TablesJ(xa ) min

xJ(x) x near xb

J(x) 12

(x xb )T B1(x xb ) 12

(H (x) yobs )T (O F)1(H (x) yobs )

x analysis variablexa final analysisxb backgroundB background error covariance

yobs observationsO observation error covarianceH observation operatorF forward model error covariance

Super Typhoon Maria 

34Slide courtesy of Fuzhong Weng

Thermal Structure from ATMS and CMWS 

35

ATMS CMWS‐20 CMWS‐28

CMWS=MWTS+MWHS Slide courtesy of Fuzhong Weng

Typhoon Maria(玛利亚) and Mangkhut(山竹) Precipitation  from FY‐3 MWTS and MWHS  

36

Precipitation from CMWS-28 Precipitation from CMORPH

FY3-CMWS-28 is combined from MWTS and MWHSCMORPH stands for NOAA Climate Prediction Center Morphing Technique

Slide courtesy of Fuzhong Weng

Product Development：Infrared Total Precipitable Water Inversion

Input Data（cloudmask,T639,satellite observations, observed 

geometric parameters）

RTTOV Simulation(two split‐window radiance, transmittance)

Total Precipitable Water calculation

(coefficient  calculation,  TPW retrieval)

Data preprocessing（NWP data spatio‐temporal interpolation、surface emissivity  

estimation）

20140107-day

TPW:mm

-50510152025303540455055606570

20140107-night

New-3bands-20150401-0030

TPW:mm

Fillvalue

0

5

10

15

20

25

30

35

40

45

50

55

60

65

70

75

FY‐3C VIRR TPW

FY‐2F TPW

• Algorithm：

• D‐Matrix coefficient training, based on atmospheric profile sample library and rapid radiation transfer model

Product Development : Sea surface wind speed and cloud water retrieval algorithm

Semi‐Statistical‐Physical Model 

0 5 10 150

5

10

15

TAO浮标海面10米风速/m·s-1

MWRI反演海面10米风速/m·s-1

（a）

0 5 10 150

5

10

15

TAO浮标海面10米风速/m·s-1

MWRI反演海面10米风速/m·s-1

（ b）

Semi Statistics and Physical Alg Statistics Alg

Product Development :Wind Profile Radar(simulator and products)

The WindRAD observation system is greatly different from the scattero‐meter systems domestic and abroad: fan‐beam, rotating scanning

WindField

GMF

Generate satellite and beam center positions by STK

Compute sample positions

Compute node positions

Resampling for each node(ViewNum Neff IncAng AzAng

SNR′)

Generate true sig0 for each view

Compute noised sig0(Kp Kg sig0′)

Output：Wind field solution in Wind vector cell (WVC)

Simulator：

100oE 110oE 120oE 130oE 140oE 0o

8oN

16oN

24oN

32oN

2km

6km

18km

Typhoon wind field  simulated by WRF‐ARW Inversion products of simulator

2019/5/1 40

With the improved instrumental performance (NE∆T), and traceable radiometric measurements, FY series are becoming one of the important components of global observation to support the wide application.

Any progress of RTMS for satellite instruments are expected and welcomed

Way forward

Thank You!谢谢！