Level 2 Processing of HUT-2D Data: Preliminary Results · PDF fileLevel 2 Processing of HUT-2D...

SMOS cal/val Workshop - March 2009 - Lisbon (Portugal)

Level 2 Processing of HUT-2D Data:

Preliminary Results

SMOS-BEC Level 2 Team.

SMOS Barcelona Expert CentrePg. Marítim de la Barceloneta 37-49, Barcelona SPAINE-mail: [email protected]: www.smos-bec.icm.csic.es

SMOS cal/val Workshop - March 2009 - Lisbon (Portugal) – M. Talone 2 / 25

Outline

- The scenario

- Brightness Temperature analysis

- Retrieval Configuration

- Results Nominal case

Other forward models

- Future Work


The Campaign

Two series of flights over the Gulf of Finland

August, 13 – SD1

August, 15 – SD2


The Scenario

SSS front from 0 to 4 psu

SST front from 19 to 23°C

SWH almost constant at 0.2 m

QuikSCAT wind Speed

3.73 m/s SD13.46 m/s SD2


Brightness Temperatures Analysis

Cross-Track mean value and std. of Tmeas - Tmod

ascending

flights

descending

flights

SD1 SD2


Brightness Temperatures Analysis

6 / 25

ascending

flights

descending

flights

SD1 SD2

Snapshot-dependent mean value and std. of Tmeas - Tmod


Pre-Processing

To improve the quality of the measurements have been considered:

- 500-m Square cells.- all the flights.- 2º, 4º, 8º angular bins.

SD1 SD2


The Campaign

The drawback is:

errors up to 5 K

errors up to 1 K


The Processing

For each bin-width, SSS has been retrieved using 2 cost

functions

( ) ( ) ,11

2

21010

2

2

2

2

2

10

∑=

−+−+−

=obs

n

N

n U

aux

SST

aux

F

modeln

measn

obs

UUSSTSSTFFN σσσ

χ

( ) ( ) ,1

2

21010

2

2

2

2

2

10

∑=

−+−+−

=obs

n

N

n U

aux

SST

aux

F

modeln

measn UUSSTSSTFF

σσσχ

and three different brightness temperatures

Case A No correction applied

Case B Constant bias corrected

Case C External Calibration applied

(1)

(2)

σSST = 0.5 °C σWS = 1.5 m/s σSWH = 1.5 m


The Processing

Nominal case:

- Klein and Swift model for dielectric constant

- Hollinger measurements linear fit for wind effect

- Linear Approximation for the atmospheric contribution

- Uniform and specularly reflected galactic noise

( ) ( ) ( ) ( ) ( ) ( )

( ) ( ) ( ) ( ) ( ) ( )

( )

( )( )

KT

Temphh

hhT

T

eemeTTeemTeTT

eemeTTeemTeTT

sky

surf

up

dn

VskyupVdnVsurfV

HskyupHdnHsurfH

7.3cos

0377.0861.220001.0

cos03.04128.0

)cos(1.2

11

11

2

coscos_

coscos_

coscos

coscos

=

⋅⋅−⋅=

⋅−⋅=

=

⋅−⋅⋅++−⋅+⋅=

⋅−⋅⋅++−⋅+⋅=−−

−−

−−

−−

θτ

θ

θ

θτθτ

θτθτ

θτθτ

θτθτ


The External TB Calibration

( ) ( )θθ ,,,,,, 1010 auxauxauxmodBorigorigorig

measBB USSTSSSTUSSTSSSTT −=∆

BmeasB

correctedB TTT ∆−=

Using the auxiliary data as input to compute a brightness temperature image.The mean of the difference between the measured temperature and the modeled one is considered bias:

The new brightness temperature is given by the subtraction between the measured brightness temperature and the bias:


Selected Results Bias correction effect TX/TY

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Selected Results Bias correction effect I

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Selected Results Binning effect TX/TY

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Selected Results Binning effect I

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Nominal configuration - Conclusions

Even after averaging measurements, large biases have been detected in the

measurements.

Retrieve without correcting for this bias resulted in very poor results, the positive

offset of the brightness temperature translates in a saturation of the retrieved salinity

toward its minimum (0 psu).

Results improve when the constant bias provided at level 1 is subtracted to the

measurements; the best retrieval is performed when the external brightness

temperature calibration is applied.

The binning also largely affects the retrieval performance.

Increasing the bin width results in an underestimation of the V-pol and overestimation

of the H-pol, producing a positive bias in the retrieved SSS when using TX/TY.

Using I, the error with respect to the modeled I introduced by the bin width is very low

and the improvement of the radiometric sensitivity permit better results

SMOS cal/val Workshop - March 2009 - Lisbon (Portugal) – M. Talone

WISE WS + SWHWISE WS

WISE SWH

17 / 25

Other forward models

105512.0 UTBh ⋅

°+⋅=∆ θ

105512.0 UTBv ⋅

°−⋅=∆ θ

10118125.0 UTBh ⋅

°+⋅=∆ θ

1045125.0 UTBv ⋅

°−⋅=∆ θ

SWHTBh ⋅

°+⋅=∆142

109.1 θ

SWHTBv ⋅

°−⋅=∆51

192.0 θ

SWHUTBh ⋅

°−⋅+⋅

°+⋅=∆

50159.0

24112.0 10

θθ

SWHUTBv ⋅

°−⋅+⋅

°−⋅=∆

50159.0

40112.0 10

θθ

Hollinger


Other forward models WISE WS – TX/TY

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Other forward models WISE WS – I

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Other forward models WISE SWH – TX/TY

0.041 (r)

0.044 (o)

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Other forward models WISE SWH – I

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Other forward models WISE WS+SWH – TX/TY

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

50

0 160

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Other forward models WISE WS+SWH – I

0.052 (r)

0.044 (o)

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

50

0 150

1.5

-5 5

1.5

-5 5

1.5

-5 -5

1.5

-5 5

1.5

-5 5

1.5

-5 5


Other forward models - Conclusions

WISE WS model results are in agreement with the nominal model.

SWH model produces a positive offset in the retrieved SSS with respect to

the WS ones . Anyway a gradient between the first and the second part of

the plot can be observed, even if shifted up.

WS + SWH model give results which are comparable with the WS models’

ones.

It must be considered that SWH is calculated as 4σ(HS), which is valid only

for fully developped sea.

SWH models are empiricals, coming from fitting data of a very different

environment


Future work

25 / 25

More averaging of

Brightness temperatures Salinities


External Salinity Calibration

26 / 25

retrinst SSSSSSCF =

retrcorr SSSCFSSS ⋅=

The CF has been calculated taking into account only the points observed more than 40 times (eliminating the noisiest points).

The final retrieved SSS is thus given by:

Once the Sea Surface Salinity map is retrieved, to correct for the errors introduced by the forward models inaccuracies ( εr, sea state

dependence… ), a Calibration Factor is calculated as in rain radar[*]:

[*]Seo, D.J. and Breidenbach, J.P., “Real-time correction of spatially non-uniform bias in radar rainfall data using rain gauge measurements”, Journal of Hydrometeorology, vol. 3, pp. 93-111, 2002

SMOS Barcelona Expert Centre (SMOS-BEC)Pg. Marítim de la Barceloneta 37-49, E-08003 Barcelona, SPAINTel. (+34) 93 230 95 00; Fax. (+34) 93 230 95 55URL: www.smos-bec.icm.csic.es

Level 2 Processing of HUT-2D Data: Preliminary Results · PDF fileLevel 2 Processing of HUT-2D...

Documents

Transcript of Level 2 Processing of HUT-2D Data: Preliminary Results · PDF fileLevel 2 Processing of HUT-2D...