1

Microwave Radiometer (MWR) Counts to

Tb (Brightness Temperature) Algorithm

Development (Version 6.0) and On-Orbit

Validation

Zoubair Ghazi

CFRSL –Central Florida Remote Sensing Lab

Dissertation Defense October 24, 2014

• To develop an improved counts to brightness temp (Tb)

algorithm for the CONAE Microwave Radiometer on

the Aquarius/SAC-D satellite

• Validation of Tb measurements using inter-satellite

radiometric comparisons (X-CAL)

• Produce an Algorithm Theoretical Basis Document

(ATBD) and deliver prototype MatLab code to

CONAE

2

• Post-launch CFRSL & CONAE evaluated MWR counts-to-Tb algorithm V5.0 – Used 6 mo of MWR on-orbit collocation with WindSat

• Ocean Tb’s exhibited small and acceptable Tb biases

• Land Tb’s exhibited anomalous behavior – Land/water Tb transitions were “Smeared”

– Step function changes of noise diode deflections

• Based upon on-orbit evaluation, it was concluded that: – V5.0 was unacceptable for producing MWR science data

– An improved counts-to-Tb algorithm must be developed to address the anomalous Tb effects

• Further, CONAE developed a revised Counts-to-Tb algo V5.0S that included a smear correction

• This was the starting point for my dissertation research

3

4

1. Evaluated the MWR counts-to-Tb algorithm V5.0S

– On-orbit X-CAL with WindSat indicated that

• Smear effects at land/water boundaries were removed

• However, anomalous effect of noise diode deflections remained

– Determined that MWR system gain varied with scene Tb

2. Developed a forward model for MWR system Counts-to-Tb

– Empirically derived coefficients to match on-orbit observations, including deep space calibrations

– Characterized model coefficients versus scene Tb

3. Developed a gain non-linearity correction

4. Implemented a new inverse model Counts-to-Tb algorithm V6.0

5. Validated algorithm using X-CAL with WindSat

5

6

Aquarius (AQ) is a mission of “Original

Exploration” First NASA mission to measure Sea

Surface Salinity (SSS) from space

SAC-D was launched on June 10th , 2011 from

Vandenberg Air Force Base, California.

• Aquarius instrument - NASA

• MWR - CONAE

(Argentinian Space Agency)

• MWR provides auxiliary environ

measurements: water vapor,

ocean surface wind speed, and

oceanic rain rate

• 3 channel push-broom Dicke

radiometer:

– 36.5 GHz H- & V-Pol

(forward-look)

– 23.8 GHz H-Pol

(aft-look)

• Earth Incidence angle

– 52o for odd beams

– 58o for even beams

• Matches the AQ swath width

of 380 km

MWR supports AQ science by measuring simultaneous &

collocated ocean brightness temperatures (Tb)

7

MWR Single Channel Block Diagram S

W M

atr

ix

Tin

To Counts

TN

On/off

8

Three Dicke radiometer states:

)1   (           )( setoffrecvrecvina VGTTC

)2    ( )( setoffrecvrecvNinN VGTTTC

)3  (       )( setoffrecvrecvoo VGTTC

Subtracting (1) from (2) yields the radiometer gain,

which varies in time

N

aNrecv

T

CCG

  oN

an

oain TT

CC

CCT

*

200 250 300 350 400 450 500 5506000

7000

8000

9000

10000

11000

12000

13000

Tin

To

Tin + Tn

Radiometer Input to the antenna port of Dicke switch, Tin , Kelvin

Rad

_co

un

ts, c

ou

nts

Slope is radiometer gain

Dickie Sw

in ant

position

Dickie Sw in

reference load

position

Dickie Sw in ant

position + noise

diode ON

9

• Our objective was to determine the MWR transfer

function based upon on-orbit measurements

• However, under typical on-orbit condition, the

radiometer system gain will vary cyclically (once/orbit)

due to the receiver physical temperature changes

• Therefore, a procedure was developed to synthesize

rad_counts @ constant system gain from MWR

measurements

10

• Time variable gain was removed and all counts were

normalized using the following equation:

11

i

inormGain

GainCoCo

i

 

Toi

<Trec>

Gaini

<Gain>

is the instantaneous reference load physical temperature, Kelvin

is the orbit average receiver noise temperature

is the instantaneous system gain

is the orbit average gain

iii GainTrecToCo *)(

Orbital Time, Min Orbital Time, Min

Co

Before Normalization After Normalization

Refe

ren

ce L

oad

Tem

pera

ture

, K

/100

Refe

ren

ce L

oad

Co

un

ts, co

un

ts

GainTC onormo *_GainTC oo *

13

14

After Count (gain)

Normalization

Nois

e D

iode D

eflection,

counts

Nois

e D

iode D

eflection,

counts

Scene B

rightn

ess T

b, K

Effects of

variable gain

Before Count (gain)

Normalization

Linear dependence

on scene Tb

Radiometer Input to the antenna port of Dicke switch, Tin (Kelvin)

123

2

3

1

    

therefore,

) ()  () (

)3(    )(

)2    ( )(

)1   (           )(

recvrecvrecv

recvantrecvrefrecvNant

setoffrecvrecvrefref

setoffrecvrecvNantN

setoffrecvrecvanta

GGG

TTTTTTT

VGTTC

VGTTTC

VGTTC

15

16

)()( 

) ()( 

) ()( 

33_

22_

11_

inreforecv

inreforecv

inreforecv

ThTgGG

ThTgGG

ThTgGG

Go is the mean “long term gain” g(Tref) is the orbital gain change due to phy temp (Tref) h(Tin) is the gain compression due to variable scene brightness temp (and injected noise diode)

These parameters are estimated during a single orbit where a deep-space calibration is performed

17

Rad

_co

un

ts

Radiometer Input to the antenna port of Dicke switch, Tin (Kelvin)

Quadratic regression

Tin-1 = Tant

Tin-3 = Tant + TN

Tin-2 = To

Ref Load

Full-dynamic range

18

Gai

n C

om

pre

ssio

n, h

(Tin

)

Radiometer Input Brightness, Tin (Kelvin)

Tin-1 = Tant

Tin-3 = Tant + TN

Linear regression

• Seven Deep Space Calibration (DSC) orbits that

included, space, ocean, and land observations were used

to cover wide range of scene Tb’s

• After counts (gain) normalization, the radiometer

transfer function was established

– Rad_counts = f(Tin)

• Quadratic regression for 37V channel yielded the

following

19

3270 )  (58.16 )(105.7  _ 24

inin TTxcountsRad

20

Rad

_co

un

ts, c

ou

nts

Radiometer Input to the antenna port of Dicke switch, Tin (Kelvin)

Tin-1 = Tant

Tin-3 = Tant + TN

Tin-2 = To

Ref Load

Tin Full-dynamic range

Quadratic

Regression

• Averaging 2nd order regression coeff’s from 7 DSC orbits, the instantaneous counts linearization equation is:

– For 37 V

– For 37 H

– For 23 H

Where x = ant, N, and ref

Tin is the input Tb to the Dicke switch, which is estimated using non-linear counts

21

2

in_ T*004)-(-7.4677e xlinearx CC

2

in_ T*004)-(-6.9064e xlinearx CC

2

in_ T*004)-(-2.1708e xlinearx CC

22

3270)(6.16)(105.7_ 24

inin TTxcountRad

3270)(6.16)(102.8_ 26

inin TTxcountRad

(V5.0S)

(V6.0)

Rad

_co

un

ts (

gain

no

rmal

ized

)

Radiometer Input to the antenna port of Dicke switch, Tin (Kelvin)

23

Brightness Temperature, K Brightness Temperature, K

No

ise

Dio

de

def

lect

ion

, co

un

ts/k

Ocean

Land

Space

Ocean

Land

Linear Regression

Scen

e B

righ

tnes

s Tb

, K

V5.0S (Without counts linearization)

V6.0 (With counts linearization)

No

ise

Dio

de

def

lect

ion

, co

un

ts/k

24

Samples Samples

Gain

, co

un

ts/k

Gain

, co

un

ts/k

V5.0S V6.0

Note:

“gain jumps”

Sce

ne

Bri

gh

tnes

s T

b,

K

Gain variation

due to changing

recvr phy temp

25

• Characterization of injected noise diode

temperature (TN) over physical temperature

• Retrieve antenna switch matrix loss coefficients

– Empirical method (regression model) was applied

– Assumption: All transmission and reflection coeff’s

are constant and are NOT expected to change during

MWR's mission life time

26

27

waveguide

Load

Distributed Loss (L)

Warm / Cold Load

Distributed Temperature (Tguide)

Th /Tc

Receiver Counts

Calibration Ref

Plane

T’h /T’c

Input to The

Receiver

SAYAK BISWAS

V2.0 V6.0

Nois

e D

iode D

eflection, counts

Samples Samples

Nois

e D

iode D

eflection, counts

Hot-load

Cold-load

28

Tap=[Tin-(b2*To+b3*T1+b4*T2+b5*T3+b6*T4)]/b1

– where

• Tap is the scene brightness temp at horn aperture

• Tin is the input brightness temperature to antenna port

of Dicke switch

• To , T1 , T2 , T3 ,& T4 are MWR physical temps

• b1, b2, b3, b4, b5 and b6 are antenna switch matrix

loss coefficients derived using the regression model

29

30

MWR coffin

• The thermal vacuum (TV) test for MWR was

performed in September 2009. (09/06 – 09/09)

• Performance of the SW matrix losses coefficients

31

Tap=[Tin-(b2*To+b3*T1+b4*T2+b5*T3+b6*T4)]/b1

Tin calc from rad_counts

Tap & Tap_regress

SWM Model, RMS=3.94 Regression Model ,RMS= 0.75

Te

mp

era

ture

s (

K)

Te

mp

era

ture

s (

K)

Time, min Time, min

32

To = 309K

To = 282K

33

34

• APC and residual bias correction were applied

by inter-satellite XCAL – MWR = target & WindSat = reference

• MWR and WindSat have different incident

angles, therefore, Tbs were adjusted using

theoretical radiative transfer model values for

both satellites (MWRsim and WSsim)

WSadj= WSobs + (MWRsim – WSsim )

• Double Difference Technique

DD = MWRobs – WSadj

35

36

MWR Tb, K

Ocean

Linear

Regression

Space

Land

WS Adjusted Tb, K

Before Correction After Correction

WS Adjusted Tb, K

MW

R T

b, K

MW

R T

b, K

Ocean

Linear

Regression

Space

Slope= 0.92329 Offset = 0.40928 Slope= 0.99952 Offset = 0.091356

1100 1200 1300 1400 1500 1600 1700-50

-45

-40

-35

-30

-25

-20

-15

-10

-5

0

B2

B4

B6

B8

1100 1200 1300 1400 1500 1600 1700-5

0

5

10

15

B2

B4

B6

B8

Samples Samples

Bri

ghtn

ess

Tem

pe

ratu

re. T

b, K

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

Version 5.0S Version 6.0

37

Samples

Samples

Bri

ghtn

ess

Tem

pe

ratu

re. T

b, K

Version 5.0S Version 6.0

1100 1200 1300 1400 1500 1600 1700-5

0

5

10

15

B1

B3

B5

B7

1100 1200 1300 1400 1500 1600 1700

-40

-20

0

20

40

60

B1

B3

B5

B7

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

Samples

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

38

Double Difference Radiometric Biases

MWR/WindSat

(Five Days Average)

Jan 01, 2012 – Dec 31, 2012

39

40

41

10 20 30 40 50 60

10

20

30

40

50

60

70-4

-3

-2

-1

0

1

2

3

4

South Pole

North Pole

South Pole

Jan 01,2012 - Dec 31,2012

DD

, K

42

10 20 30 40 50 60

10

20

30

40

50

60

70-4

-3

-2

-1

0

1

2

3

4

South Pole

North Pole

South Pole

Jan 01,2012 - Dec 31,2012

DD

, K

• MWR Counts-to-Tb algorithm V6.0 has been developed and distributed to the AQ Cal/Val Team

– MWR transfer function non-linearity in V5.0S has been characterized and corrected in V6.0

– Antenna switch matrix loss coefficients were derived using re-analysis of MWR pre-launch TV calib test

• Validation of V6.0 performed using 2 years of on-orbit measurements

– On-orbit X-CAL, between MWR and WindSat, have produced the antenna pattern correction (APC) and removed small Tb biases

• V6.0 Algorithm Theoretical Basis Document and MatLab code delivered to CONAE for science data processing

43

• Conferences 1. Ghazi, Z.; Biswas, S.; Jones, L.; Hejazin, Y.; Jacob, M.M., "On-orbit signal processing

procedure for determining Microwave Radiometer non-linearity," Southeastcon, 2013

Proceedings of IEEE , vol., no., pp.1,5, 4-7 April 2013 doi: 10.1109/SECON.2013.6567504

2. Ghazi, Zoubair; Santos-Garcia, Andrea; Jacob, Maria Marta; Jones, Linwood, "CONAE

Microwave Radiometer (MWR) counts to Tb algorithm and on-orbit validation," Microwave

Radiometry and Remote Sensing of the Environment (MicroRad), 2014 13th Specialist

Meeting on , vol., no., pp.207,210, 24-27 March 2014 doi: 10.1109/MicroRad.2014.6878941

3. Santos-Garcia, A; Biswas, S.; Jones, L.; Ghazi, Z.; "Aquarius/SAC-D Microwave Radiometer

brightness temperature validation," Oceans, 2012 , vol., no., pp.1,4, 14-19 Oct. 2012 doi:

10.1109/OCEANS.2012.6404830

44

45

Back UP

46

47

48

Note: “gain jumps” @

land/water boundaries Dynamic range = 60 counts

Rad

_co

un

ts

V5.0S V6.0

1100 1200 1300 1400 1500 1600 1700-30

-25

-20

-15

-10

-5

0

B2

B4

B6

B8

1100 1200 1300 1400 1500 1600 1700-5

0

5

10

15

B2

B4

B6

B8

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

Samples Samples

Version 5.0S Version 6.0

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

1100 1200 1300 1400 1500 1600 1700-5

0

5

10

15

B1

B3

B5

B7

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

Samples Samples

1100 1200 1300 1400 1500 1600 1700-30

-25

-20

-15

-10

-5

0

B1

B3

B5

B7

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

Version 5.0S Version 6.0

51

Samples

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

Version 5.0S

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

Samples

Version 6.0

52

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

Samples

Version 6.0

Bri

ghtn

ess

Tem

per

atu

re. T

b, K

Samples

Version 5.0S

1/ 3 6/ 1 10/29-5-4-3-2-1012345

Sep 1 - feb 28

B1

1/ 3 6/ 1 10/29-5-4-3-2-1012345

Sep 1 - feb 28

B2

1/ 3 6/ 1 10/29-5-4-3-2-1012345

Sep 1 - feb 28

B3

1/ 3 6/ 1 10/29-5-4-3-2-1012345

Sep 1 - feb 28

B4

1/ 3 6/ 1 10/29-5-4-3-2-1012345

Sep 1 - feb 28

B5

1/ 3 6/ 1 10/29-5-4-3-2-1012345

Sep 1 - feb 28

B6

1/ 3 6/ 1 10/29-5-4-3-2-1012345

Sep 1 - feb 28

B7

1/ 3 6/ 1 10/29-5-4-3-2-1012345

Sep 1 - feb 28

B8

53

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B1

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B2

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B3

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B4

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B5

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B6

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B7

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B8

54

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B1

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B2

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B3

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B4

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B5

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B6

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B7

1/ 3 4/12 7/21 10/29-5-4-3-2-1012345

B8

55

56

To Gain

Scen

e B

righ

tnes

s Tb

, K

Time (min) Time (min)

To Gain

Before Gain Normalization After Gain Normalization

57

1

2

3

Secondary

• New MWR Tb data set to be used for tuning and validation of the wind speed algorithms

• XCAL 5 day double difference (DD) biases calculated between WindSat & MWR

– DD = MWR-WS • Applied triangular moving average on the 5 day DD time

series to smooth the correction

• The new MWR Tb’s V7.0 = V6.0 – Tbbiases

– These V7.0 Tb’s will be normalized to match the WindSat Tb’s in the mean i.e., have zero DD Tb-bias

• The new “adjusted DD” given in the following charts was derived as:

• DDadj = DDV6.0 -Tb biases

7/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B1

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B2

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B3

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B4

July 2012 - Nov 2013

7/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B5

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B6

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B7

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B8

July 2012 - Nov 2013

7/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B1

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B2

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B3

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B4

July 2012 - Nov 2013

7/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B5

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B6

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B7

July 2012 - Nov 20137/12 11/12 3/13 7/13 11/13-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3B8

July 2012 - Nov 2013