1

Operational Calibration of Satellite Microwave Instruments for Weather and

Climate Applications

Fuzhong Weng and Tsan MoSensor Physics Branch

NOAA/NESDIS/Office of Research and Applications

Banghua Yan, QSS Group Inc.Ninghai Sun, IMSG

and many others

Achieving Satellite Instrument Calibration for Climate Change (ASIC3) Workshop, May 16-18, 2006

2

Outline

• Significance of satellite instrument calibration

• Microwave instrument calibration components

• Microwave sensor calibration for operational and research satellites

• Issues and Challenges

• Summary and Conclusions

3

Global Temperature Trend Depicted by NOAA MSU and AMSU

Trend: N10 = - 0.40 K Dec-1, N11 = 0.80 K Dec-1,

N12 = 0.36 K Dec-1, N14 = 0.43 K Dec-1

248

249

250

251

252

253

1987 1989 1991 1993 1995 1997 1999 2001 2003

NOAA10

NOAA11

NOAA12

NOAA14

Linear (NOAA10)

Linear (NOAA11)

Linear (NOAA12)

Linear (NOAA14)

5-day and global-ocean-averaged time series for NOAA 10,11,12, and 14 obtained from MSU 1B data which uses NESDIS operational calibrationalgorithm

Combined MSU and AMSU observations can be used to detect climate trend, however, different merging procedure in removing intersatellite biases causes different trend results

4

Calibration Accuracy in Relation to Climate Trend – Ocean Mean Wind Speed

)(1038.2

1

)(

3dsbb

dudsb

TTT

W

T

W

TTTTT

This is the case for SSM/I 37 GHz, V-Pol, surface wind > 12 m/s. The sensitivity of wind speed to brightness temperature is about 1. – 3 m/s/K.

Tropical mean wind speed increases 0.5 m/s per decade. Is the recent increasing hurricane wind damage responding to this trend? How can we assure this trend not related to inter-satellite calibration and algorithms

5

Calibration in Support of Satellite Data Assimilation

• No radiance biases– Instrument– Forward model

• Known Errors – Observation– Forward model

oTobTbJ IxIFEIxIxxBxx )()()(2

1

2

1 11

wherex is a vector including all possible atmospheric and surface parameters.I is the radiance vector B is the error covariance matrix of background E is the observation error covariance matrixF is the radiative transfer model error matrix

You can’t simply fudge the weights!

6

Comparison of Impact of Observing Sounding Data

010

20

30

Analysis

No Satellite

Losing all microwavesounders

Losing all infraredsounder

Losing all radiosonde T,q and u

Losing all radiosonde Tand q

Glo

bal d

egra

datio

n

From Roger Sounder, The Metoffice, UK

Ten years ago? TOVS NESDIS retrievals, AMV, more but lower quality radiosondes

7

Microwave Instrument Calibration Components

Energy sources entering feed for a reflector configuration

1. Earth scene Component,2. Reflector emission3. Sensor emission viewed through

reflector,4. Sensor reflection viewed through

reflector,5. Spacecraft emission viewed through

reflector,6. Spacecraft reflection viewed through

reflector,7. Spillover directly from space,8. Spillover emission from sensor,9. Spillover reflected off sensor from

spacecraft,10. Spillover reflected off sensor from

space,11. Spillover emission from spacecraft

Example: Emission by Antenna/Front End Component

Emission by the antenna and front-end components can introduce a diurnal temperature variation.

T = Physical temperature of antenna, feed horn, waveguide, etc.,

Tb ))(( bwcbbb TTTTVSIT a = Transmittance due to absorption

of antenna, feed horn, etc.,

T’b

))(1( babb TTTT

)()1( bab TTT

[1]

[2]

Emission & absorption by antenna & front-end.

Two-point radiometer calibration :

Combining (1) and (2) :

[3a]Tb= Tbo + Tb

))(( bwcbbbo TTTTVSIT [3b]

[3c]

Antenna

Feed Horn

Waveguide

)1.0,990.0and10For ( KTKTT bab

Example: Spill-Over due to Antenna Side-Lobes

SLT

bT

))(1( SLbsbb TRTTT

A very small portion of the antenna side lobes “sees” radiation emanating from outside the Earth. An even smaller portion, S( antenna gain) results from the solar radiation, TSL, being reflected with reflectivity R from materials onboard the spacecraft.

Earth

The brightness temperature can also be written as

bbb TTT

where

))(1( SLbsb TRTT

SLRT

Square Law

Detector

VIV

I

Example: Non-Linear Calibration

44

33

221 IaIaIaIaV

Time-averaged voltage :

2242 3 IIaaV

Nyquist theorem :

)()(2 TRTRKBGI A

Combining [1] and [2] :

)(1)(1 AAo TRTRbbV KBGTKBTaabo ]3[ 42 KBGKBTaab ]6[ 421

)(32

4 parameternonlineara

KBGa

K=Boltzman’s constant G=Amplifier gain, B=BandwidthT=Amplifier temperature, Te = Radiometric temperature

Two-point radiometer calibration eliminates bo and b1 from <V> (output in counts) so that

))(()( 2WACACACA CCCCSμCCSRR

[1]

[2]

[3]

Output Voltage

Input Current

))(()( 2WACACACA CCCCSμCCSTT At microwave region:

11

NESDIS/STAR Integrated Cal/Val System

1. Current Capabilities: • Noise quantification (NEDT),• Linear and non-linear calibration

algorithms,• Correction of sudden jumps and

contamination associated with warm load and space view calibration counts,

• Monitoring instrument noise, gain, telemetry and PRT uniformity,

• Mitigation of radio frequency interference,

• Global bias analysis from forward calculations using NWP models,

• Time series of SNO/SCO matched data from a pair of operational satellites,

• Time series of updated calibration coefficients with digital access,

• Reference areas/site for vicarious calibration,

• Monitoring of key MW products sensitive to calibration

2. Future Capabilities: Validation of EDRs

12

NOAA AMSU Sensor

•Flown Since NOAA-15 (May 1998)•Contains 20 channels:

•AMSU-A•15 channels•23 – 89 GHz

•AMSU-B (now MHS on NOAA-18)

•5 channels•89 – 183 GHz

•6-hour temporal sampling:•200, 730, 1400, 1930 LST

13

NOAA AMSU Calibration and Monitoring

• Pre-launch checkup– Noise quantification (NEDT) from EDU and PFM – Non-linearity – …..

• Update AMSU/MHS Calibration Parameters Input Data Set (CPIDS) in Level 1B,– Polynomial coefficients form converting PRT counts into temperature– Warm load correction at three instrument temperatures,– Cold spaces correction to the cosmic background temperatures– Error limits of warm and cold radiometric counts between the sample of the same scan line,– Non-linearity parameter– Temperature to radiance conversion factors– Min and max of RF shelf instrument temperature sensors – Analog data conversion coeff– Antenna position data in counts– Gross radiometric limits (max and min) on space and warm targets views– Antenna pattern parameters for lunar correction– Asymmetry correction

• On-board Monitoring – Correction of sudden jumps and contamination associated with warm load and space view

calibration counts,– Monitoring instrument noise, gain, telemetry and PRT uniformity,– Detect the radio frequency interference from AMSU, – Global bias analysis from forward calculations using NWP models, – Time series of SNO/SCO matched data from a pair of operational satellites,– Reference areas/site for vicarious calibration, – key MW products sensitive to calibration (Cloud Liquid Water and Precip)

14

Pre- and Post-launch Noise Characterization

NOAA-18 AMSU-A

NOAA-18 MHS

15

NOAA-15 AMSU-A Asymmetry Correction

,

∆T = A0 exp{ -0.5[(θ - A1) /A2]2 } + A3 + A4 θ + A5 θ2

16

Effects of Biases on Operational Products

• AMSU A2 model cross scan asymmetry was detected from the first NOAA-15 cloud liquid water

• Physical retrievals of cloud liquid

water are directly subject to instrument biases

• If AMSU cloud liquid water is assimilated or used for QC others, it results in global false alarm clouds and rejection of many other useful information

• Bad consequence from AMSU xing scan radiance biases if not corrected because CLW is used for NWP QC

NOAA-15

NOAA-16

17

DMSP Special Sensor Microwave Imager and Sounder (SSMIS)

• The Defense Meteorological Satellite Program (DMSP) successfully launched the first of five Special Sensor Microwave Imager/Sounder (SSMIS) on 18 October 2003. 

• SSMIS is a joint United States Air Force/Navy multi-channel passive microwave sensor

• Combines and extends the current imaging and sounding capabilities of three separate DMSP microwave sensors, SSM/T, SSM/T-2 and SSM/I, with surface imaging, temperature and humidity sounding channels combined.

• The SSMIS measures partially polarized radiances in 24 channels covering a wide range of frequencies (19 – 183 GHz)

– conical scan geometry at an earth incidence angle of 53 degrees

– maintains uniform spatial resolution, polarization purity and common fields of view for all channels across the entire swath of 1700 km.

18F13

0600

1800

1200 0000

DMSPLTANs

F13 1818F14 2012F15 2130F16 2000

NOAALTANs

N15 1903N16 1430N17 2204N18 1359

N

•As of August 2005

N15

F14F15

N17

N18

F16

N16

DMSP and NOAA Constellation

19

SSMIS vs. AMSU-A Weighting Functions

Oxygen Band Channels

SSMIS 13 Channels Sfc – 80 km

AMSU-A 13 Channels Sfc - 40 km

SSMIS vs. AMSU Sounding

20

SSMIS Antenna System and Calibration

• Main-reflector conically scans the earth scene

• Sub-reflector views cold space to provide one of two-point calibration measurements

• Warm loads are directly viewed by feedhorn to provide other measurements in two-point calibration system

• The SSMIS main reflector emits radiation from its coating material

– SiOx VDA (coated vapor-deposited aluminum)

– SiOx and Al VDA Mixture– Graphite Epoxy

• Warm load calibration is contaminated by solar and stray Lights

– Reflection Off of the Canister Top into Warm Load

– Direct Illumination of the Warm Load Tines

• Lunar contamination on space view

21

SSMIS Anomaly Distribution

Shown is the difference between simulated and observed SSMIS 54.4 GHz. The SSMIS is the first conical microwave sounding instrument, precursor of NPOESS CMIS. The calibration of this instrument remains unresolved after 2 years of the lunch of DMSP F16. The outstanding anomalies have been identified from three processes: 1) antenna emission after satellite out of the earth eclipse which contaminates the measurements in ascending node and small part in descending node, 2) solar heating to the warm calibration target and 3) solar reflection from canister tip, both of which affect most of parts of descending node.

22

SSMIS Anomalies and Their Mitigation Algorithms

1. Antenna is not a pure reflector. It emits radiation with a very small emissivity and its own temperature. This additional radiation is called as an antenna emission anomaly

2. Warm load is heated by intruded solar radiation. The energy received through feedhorn does not match with the warm load physical temperature measured by the platinum résistance thermisters (PRT). This is referred as a warm load anomaly

3. The radiance from space view by the sub-reflector does not correspond to the sum of cosmic background temperature (2.73K) and pre-calculated correction values for each channel due to antenna side-lobe effort.

1. Use the emissivity from NRL antenna model and the temperature measured from the thermister mounted on antenna arm as approximation

2. Analyze the time series of warm load counts together with PRT and define the anomaly locations in terms of the FFT harmonics

3. Analyze the time series of cold space view count and define the anomaly locations in terms of the FFT harmonics and cosmic temperature plus antenna correction

Anomaly Causes Anomaly Mitigation Process

23

SSMIS Calibration Algorithms

1. Use the emissivity from NRL antenna model and the temperature measured from the thermister mounted on antenna arm as an approximation

2. Analyze the time series of warm load counts together with PRT and define the anomaly locations in terms of the FFT harmonics

3. Analyze the time series of cold space view count and define the anomaly locations in terms of the FFT harmonics and cosmic temperature plus antenna correction

WCW

CAC

CW

AWW

CW

CAA T

CC

CCC

CC

TTC

CC

TTT

R

RRAA

TTT

1

'

AAcA TTT

RRRAA TTT )1('

where TA is the antenna temperature corresponding to the earth scene’s radiance, and R and TR is the reflector emissivity and Temperature, respectively

24

Theoretical SSMIS Reflector Surface Parameters

(NRL Multilayer Antenna Model) 

Emissivity (V-pol/20deg) [ ∈ R ] Freq. (GHz) Al GrEp SiOx SiOx/Al  19.35 0.00051 0.012 0.91 0.00051  37.0 0.00071 0.016 0.91 0.00071  60.0 0.00090 0.020 0.91 0.00090  91.65 0.00111 0.025 0.91 0.00111  183.0 0.00157 0.035 0.91 0.00157

25

FFT Analyses of Warm Counts (54.4 GHz)

Note: (1) CWF = FFT-1( FFT(CW) * Filter(fL) ) ), where fL is a cutoff frequency of the low pass filter,

where T 102 minutes. (2) f0 is sampling frequency = 1.0/T.

26

SSMIS Antenna Temperature Bias February 3, 2006

aTT BA /

Before anomaly correction After anomaly correction

Temperature biases from TDR and SDR space are related through the slope coeff. for spill-over correction, Tb = a*Ta + b

27

SSMIS Bias Trending

29

SSMIS vs. SSM/I Products

SSMIS-F16

SSM/I-F15

Cloud Liquid Water Total Precipitable Water

30

Other Microwave Instrument Calibration

• Windsat vicarious calibration– Amazon/Congo basins– Time series of averaged 3rd and 4th components

• Aqua AMSR-E radio frequency detection– Develop RFI index fro 6 V/H pol over land

31

Microwave Sensor Inter-calibration for Climate Applications

• DMSP Series SSM/I (F8 to F15)– Data rescue and archival– Metadata for re-calibration– Inter-calibration using simultaneous conical

overpassing– Reproduce all SSM/I EDRs climatology

• NOAA MSU (N10-14) Time Series Analysis – Non-linearity parameter– Bias removal using SNO

32

The First SSM/I Monthly Products Generated from NOAA/NESDIS

33

Intersatellite Calibration Using Simultaneous Nadir/Conical Overpass (SNO/SCO)

• SNO – every pair of POES satellites• with different altitudes make orbital

intersections within a few seconds regularly in the polar regions (predictable w/ SGP4)

• Precise coincidental pixel-by-pixel match-up data from radiometer pairs provide reliable long-term monitoring of instrument performance

• The SNO method (Cao et al., 2005) is used for on-orbit long-term monitoring of imagers and sounders (AVHRR, HIRS, AMSU) and for retrospective intersatellite calibration from 1980 to 2003 to support climate studies

• The method has been expanded for SSM/I with Simultaneous Conical Overpasses (SCO)

SNOs occur regularly in the +/- 70 to 80 latitude

34

NOAA-18 vs. Aqua AMSU SNO Matching

35

DMSP F-10 vs. F-13 SSM/I SCO Matching(37- 85 GHz Channels)

36

DMSP F-10 vs. F-13 SSM/I SCO Matching(19-22 GHz Channels)

37

Calibration Issues That Affect NOAA Uses of Microwave Data in Weather and Climate Research

Microwave Radiometry

System

Major Postlaunch Calibration Problems

Impacts on Weather & Climate Applications

Mitigation strategies

MSU •Non-linearity•Warm Load PRT anomaly•Cross-sensor biases

•Controversy climate trend •Non-linearity correction•SNO derived biases

NOAA AMSU/MHS •Cross-scan asymmetry• AMSU-B RFI from STX transmission•Lunar contamination

•Rejection of AMSU data in NWP•Little uses of AMSU-B

•Asymmetry bias correction•RFI correction•LCC

EOS Aqua AMSU

•Cross-scan asymmetry •Rejection of AMSU data in NWP

DMSP SSM/I

•APC and spill-over correction•Cross-instrument biases

•Uncertainty in derived emissivity spectra•Long-term climatology

•SCO derived biases

DMSP SSMIS

•Reflector emission•Warm load anomaly

•Difficult to use of sounding channels in NWP•Poor quality of sounding products

•Characterization of reflector emissivity/temperature•FFT removals for warm load count and PRT anomalies

WindSAT

•Biases at polarimetric channels•RFI at low frequencies

•Wind direction biases•Limited uses for soil moisture retrievals

•Vicarious calibration•RFI detection/removal algorithms

Aqua AMSR-E

•Warm load instability •RFI at low frequencies

•Wind direction biases•Limited uses for soil moisture retrievals

•Cross-sensor calibration with TMI•RFI detection/removal algorithms

38

Summary and Conclusions

• Operational microwave instruments AMSU-A/B (MHS) on board NOAA POES have been well calibrated for weather applications. Major NWP centers have demonstrated the greatest impacts on weather forecasts from direct radiance assimilation, and they are pleased with the quality of the microwave calibration algorithms developed by NESDIS/STAR.

• DMSP SSMIS may soon become another major data source for NWP data assimilation. Currently, resolving its calibration uncertainty from antenna emission and contamination by solar/stray lights is of a highest priority. The NESDIS/STAR beta-version calibration algorithm has significantly eliminated most of anomalies.

• The biases in the polarimetric microwave instruments (e.g. WindSAT) can be characterized from vicarious sites where surface polarimetric properties are well understood from some field campaigns and advanced radiative transfer modeling.

• Intersatellite biases for microwave sounders or imagers can be quantified from simultaneous nadir/conical overpassing, but the bias characteristics from those surface sensitive channels could be quite significantly different from both poles. The differential biases may produce an inconsistent climate trending analysis. Thus, for climate studies, the current SNO/SCO algorithms may need some further constraints

39

Backup Slides: NOAA POES AMSU Calibration and Monitoring

40

Pre- and Post-launch Noise Characterization

NOAA-18 AMSU-A

NOAA-18 MHS

41

Trending over 65 days

AMSU-A NEDT Trending

42

Trending over 65 days

MHS Gain and NEDT Trending

43

Monitoring Uniformity of Warm Load PRT Temperatures

T =Max – Min TSpec: T < 0.2 K

44

Digital Counts

Rad

ian

ce (

Bri

ghtn

ess

Tem

p)

(Cc , 2.73K)

(Cw, Rw)

(Ce, RL)

(Ce, Re)

)(, cecLe CCSRR

cw

cw

CC

RRS

RZRR Lee ,

))((2wece CCCCSZ

Two Point Radiometer Linear Calibration:

Two Point Radiometer with Nonlinear Calibration Correction:

Linear and Non-linear Calibration

where δR is the post-launch bias caused by factors other than non-linearity

45

NonlinearityNonlinearity

Spec:Spec:

Ch.1, 2, 15: Ch.1, 2, 15: 0.5 K0.5 K

Ch.3-14: Ch.3-14: 0.375 K0.375 K

A1-1 A1-1 Channels:Channels: Out of specOut of spec

46

Correction for Lunar Contamination on Cold Space Calibration

47

Possible Causes for AMSU Asymmetry

• A misalignment of AMSU polarization vector– Mostly noticeable at clean window channels

• Errors in Instrument pointing angle– It is unlikely because the cross-track pointing error

(0.1 to 0.3 degree) is not large enough to produce this kind of asymmetry.

• Side lobe intrusion to the solar array– There should be some latitudinal dependence– The response would occur at multiple channels

48

Trending over 65 days

AMSU-A Gain Trending

49

Offset Changed

Trending over 65 days

Trending for AMSU-A Calibration Counts

50

Libyan Desert July 2005

Vicarious Calibration Using Libyan Desert

51

Backup Slides: Windsat and AMSR-E Calibration and Monitoring

52

WindSat Applications

Freq, GHz Channels BW, MHz msec NEDT (1) EIA, deg IFOV, km

6.8 v, h 125 5.00 0.48 53.5 40x60

10.7 v, h, ±45, lc, rc 300 3.50 0.37 49.9 25x38

18.7 v, h, ±45, lc, rc 750 2.00 0.39 55.3 16x27

23.8 v, h 500 1.48 0.55 53.0 12x20

37.0 v, h, ±45, lc, rc 2000 1.00 0.45 53.0 8x13

•Main Applications: ocean surface wind vector.

•Other applications at NOAA/NESDIS:

Test the community radiative transfer model

Possibility for directly assimilating radiances

Microwave products such as CLW, TPW, land emissivity

53

WindSat Biases from Vicarious Calibration

•Monthly mean of 4th Stokes components over Amazon rainforests should be zero because of surface roughness and heterogeneity relative to azimuthal direction. The residual of this mean is largely due to the instrument calibration biases. •The bias (-0.5K) at 18.7 GHz will result in substantial bias in wind direction retrievals because of the actual wind induced signal is on the order of a couple of degrees in Kelvin (from Liu and Weng, 2005, Appl. Optics)

18.7 GHz

10.7 GHz and 37 GHz

54

EOS Aqua AMSR-E Team Algorithm

• Ocean products : SST,SSW,TPW,CLW, Rain rate, Sea ice concentration

• Land products: LST, Soil moisture,Rain rate,Snow cover, Snow/Ice Types, Snow equivalent water

Parameters SMMR(Nimbus-7)

SSM/I (DMSP-

F08,F10,F11,F13,F15)

AMSR (Aqua, ADEOS-II)

Time Period 1978 to 1987 1987 to Present Beginning 2001

Frequency (GHz) 6.6, 10.7, 18, 21, 37 19.3, 22.3, 36.5, 85.5 6.9, 10.7, 18.7, 23.8, 36.5, 89.0

Sample Footprint Sizes (km)

148 x 95 (6.6 GHz)27 x 18 (37 GHz)

37 x 28 (37 GHz)15 x 13 (85.5 GHz)

74 x 43 (6.9 GHz)14 x 8 (36.5 GHz)6 x 4 (89.0 GHz)

55

AMSR-E Radio Frequency Detection

• Radio-frequency interference (RFI): Any man-made emissions from active microwave transmitters, usually generated by television, radio, antennas

• Location: mostly over highly populated urban areas, military fields.

• RFI (V/H) index = TV(H)6.9 - TV(H) 10.7• 5 ~ 10 K Weak • 10 ~ 20 K Moderate• > 20 K Strong

56

AMSR-E Radio Frequency Interference(March 2004)

57

AMSR-E Radio Frequency Interference(March 2004)

58

Time Series of RFI Indices in Chicago

Time Series of RFI Index in Chicago

16

17

18

19

20

21

22

23

24

25

0 5 10 15 20 25 30Day in March 2004

RF

I In

dex (T

B6-T

B10) -

RFI-H Component

RFI-V Component

59

Backup Slides: MSU Non-Linearity Calibration using SNO

60

kkkkLk ZRRR ,

k j

jjjjLj ZRRR ,

jjkkL ZZRR

For many pairs of SNO, multivariable linear regression will resolve R (intersallite bias), k and j (non-linearity parameters for k, j satellites, respectively

SNO Pairs

We would like to have zero bias between two satellites,Rk = Rj

SNO Time Series Used for Deriving Intersatellite Bias and Nonlinearity

61

Results-New Calibration Coefficients

R and obtained by SNO

R and obtained from pre-launch Calibration (Mo et al. 2001)

Satellites R R

N10 0 5 0

4.9-5.1

N11 -2.556 8.308 0 6.6-7.7

N12 -0.164 5.564 0 3.1-3.3

N14 -0.834 6.386 0 3.2-3.4

Calibration coefficients for different satellites obtained by sequential adjusting process using the SNO matchups when NOAA 10 is assumed to be the reference satellite. Units for R and are 10-5 (mW) (sr m2 cm-1) -1 and (sr m2 cm-1) (mW) -1, respectively. (Courtesy of C. Zou)

62

Reconcile Tropospheric Climate Trend using SNO with MSU

• Past MSU Channel 2 Trend Results:

• Spencer and Christy (1992): 0.020 C Decade-1, 1979-1988

• Christy et al. (2003): 0.020 C Decade-1, 1979-2002

• Mears et al. (2003): 0.100 C Decade-1, 1979-2001

• Vinnikov and Grody (2003): 0.220C Decade-1, 1979-2002

• Grody et al. (2004) 0.170C Decade-1, 1979-2002

63

Trend=0.32 K Dec-1

250

251

252

253

254

1987 1989 1991 1993 1995 1997 1999 2001 2003

Combined

Linear (Combined)

Trend = 0.17 K Dec-1

250

251

252

253

254

1987 1989 1991 1993 1995 1997 1999 2001 2003

Combined

Linear (Combined)

Trends for linear calibration algorithm

0.32 K Decade-1

Trends for NESDIS operational calibration algorithm

0.22 K Decade-1

(Vinnikov and Grody, 2003)

Trends for nonlinear calibration algorithm using SNO cross calibration

0.17 K Decade-1

Trend = 0.220 K Decade-1

250

251

252

253

254

1987 1989 1991 1993 1995 1997 1999 2001 2003

Combined

Linear (Combined)

SNO Derived Climate Trend from MSU

Courtesy of C. Zou