1

A Revised Geophysical Model Function A Revised Geophysical Model Function

for the Advanced Scatterometer (ASCAT) for the Advanced Scatterometer (ASCAT)

at NOAA/NESDISat NOAA/NESDIS

Seubson Soisuvarn, Zorana Jelenak, Paul S. Chang, and Qi Zhu

NOAA/NESDIS/STAR-UCAR

IGARSS 2010, July 25-30, Honolulu, Hawaii, USA

2

IntroductionIntroduction

• ASCAT measures radar

backscatter (sigma0) using C-band

vertically polarized fan-beam

antennas - multiple incidence

angles (~25°-65°)

• Operational ASCAT winds

retrievals is based on CMOD5.n

model function and it shows clearly

underestimates high wind speeds

• Hurricane force wind detection

from ASCAT is very rare Can ASCAT high wind speed retrievals be improved?

3

What do we knowWhat do we know

• Scatterometer wind retrievals are dependent on Geophysical Model

Function (GMF) that relates backscatter (Sigma0) to ocean surface

wind fields

• Satellite C-band V-Pol Sigma0 collocation data at high winds are not

currently sufficient to build empirical geophysical model function

– Winds are not quite uniformly distributed in all directions

• C-band V-Pol Sigma0 measurements for high winds are available

from Airborne scatterometer (IWRAP) but at limited incidence

angles and much higher resolution than satellite measurements

4

Study ApproachStudy Approach

• The ASCAT operational GMF (CMOD5.n) is written in harmonic terms:

• ASCAT/QuikSCAT collocations show B0 is biased high at high wind

speeds.

• Replace B0 term with a new B0 and assuming B1 and B2 are modeled

correctly

– Not a new idea (Jones et. al – TRMM PR GMF, IGARSS’03, ’04)

• New B0 is determined based on a linear characteristic of an airborne

scatterometer GMF (IWRAP)

– Replace B0 only at higher winds

6.12100 ]2cos),(cos),(1[),( BBB

Hersbach, et al., J. Geophys. Res. (2007)

5

Measured and Model Sigma0 dependence on Measured and Model Sigma0 dependence on

wind vector at 30 deg incidencewind vector at 30 deg incidence

Relative wind direction (Degree)

Sig

ma0

ASCAT/QuikSCAT matchu-ups @ 30° incidence

CMOD5n B0

“True” B0

5m/s 9m/s 12m/s

15m/s 17m/s 20m/s

6

cmod5n is not saturate beyond 40° incidence

IWRAP saturation wind speed

New SaturationWind speed

C-Band V-pol Saturation wind speedsC-Band V-pol Saturation wind speeds

• IWRAP GMF is valid at 29°, 34°, 40° and 50° incidence and 25-65 m/s

• The B0 is written in the form:

• U10Nsat is defined as the wind speed where B0 reaches its maximum value:

2

1

1010

2

10

0 100

NsatN UUNU

B

21021010 )][log()log((10 NN UUB

Fernandez, et al., J. Geophys. Res. (2006)

Fernandez, et al., J. Geophys. Res. (2006)

7

High wind speed BHigh wind speed B0 0 ; Incidence: 50 deg; Incidence: 50 deg

B0 vs wind speed

U10Nsat =61.5 m/s

IWRAP

New B0

NU

B

10

0

vs wind speed

U10Nsat =61.5 m/s

New B0

IWRAP

8

CMOD5n vs New BCMOD5n vs New B00

CMOD5n B0

New B0

Wind Speed (m/s)

B0

25° 30° 40°

50° 60° 65°

9

Sigma0 Model vs. Measurements

Measured Sigma0 (dB)

Cal

cula

ted

Sig

ma0

(dB

)

CMOD5.n

New GMF

Fore-beam Mid-beam Aft-beam

Fore-beam Mid-beam Aft-beam

10

ASCAT vs QuikSCAT speed (before)ASCAT vs QuikSCAT speed (before)

Cumulative distributioncurve

QuikSCAT Wind Speeds

AS

CA

T W

ind

Spe

eds

11

ASCAT vs QuikSCAT speed (after)ASCAT vs QuikSCAT speed (after)

Cumulative distributioncurve

QuikSCAT Wind Speeds

AS

CA

T W

ind

Spe

eds

12

ASCAT vs QuikSCAT direction (before)ASCAT vs QuikSCAT direction (before)

QuikSCAT Wind Directions

AS

CA

T W

ind

Dire

ctio

ns

13

ASCAT vs QuikSCAT direction (after)ASCAT vs QuikSCAT direction (after)

QuikSCAT Wind Directions

AS

CA

T W

ind

Dire

ctio

ns

14

Across Swath Speed Biases (before)Across Swath Speed Biases (before)

2 m/s reference line

15

Across Swath Speed Biases (after)Across Swath Speed Biases (after)

2 m/s reference line

16

Across Swath Speed RMS errors (before)Across Swath Speed RMS errors (before)

2 m/s reference line

17

Across Swath Speed RMS errors (after)Across Swath Speed RMS errors (after)

2 m/s reference line

18

Operational ASCATQuikSCAT

Mean Wind Speed FieldMean Wind Speed Field2007-2009 Winter Seasons - North Pacific2007-2009 Winter Seasons - North Pacific

19

Mean Wind Speed FieldMean Wind Speed Field2007-2009 Winter Seasons - North Pacific2007-2009 Winter Seasons - North Pacific

New ASCATQuikSCAT

20

1000km2000km

3000km

Mean Wind Speed FieldMean Wind Speed Field2007-2009 Winter Seasons - North Pacific2007-2009 Winter Seasons - North Pacific

QuikSCAT New ASCAT

21

Gale Force Wind Frequency

Storm Force Wind Frequency

QuikSCAT

Gale Force Wind Frequency

Storm Force Wind Frequency

Operational ASCAT

Hurricane Force Wind Frequency Hurricane Force Wind Frequency

22

Gale Force Wind Frequency

Storm Force Wind Frequency

New ASCAT

Hurricane Force Wind Frequency

Gale Force Wind Frequency

Storm Force Wind Frequency

Hurricane Force Wind Frequency

Operational ASCAT

23

OperationalASCAT

Sonde Wind Speed 28.8m/s

SFMR 28 m/s

Operational ASCAT 22 m/s

24

New ASCAT

Sonde Wind Speed 28.8m/s

SFMR 28 m/s

New ASCAT 25 m/s

25

ConclusionsConclusions

• New High wind ASCAT GMF was developed

– B0 was modified in CMOD5.n

• Matching IWRAP GMF high wind trend and correcting for observed B0

biases

• New ASCAT high wind speed retrievals are improved– Bias and RMS wind speed errors are compared better to QuikSCAT and

SFMR

– New ASCAT wind speed retrievals are still biased low but resultant biases are smaller

– Improve storm and hurricane force wind detection

Next Step:

• Transition new ASCAT winds product into operations (Soon)

26

BACK UP

27

Operational wind speed histogram

28

New wind speed histogram

29

Operational ASCATOperational ASCAT

KNMIKNMI

Mean bias = -3.77 m/sMean STD = 1.01 m/s

30

New ASCATNew ASCAT

KNMIKNMI

Mean bias = -2.38 m/sMean STD = 0.98 m/s

31

Model vs. Measurement @ 30° incidence

CMOD5.n

New

32

Measured vs CMOD5.n 30 deg

33

Measured vs CMOD5.n 40 deg

34

Measured vs CMOD5.n 50 deg

35

Measured vs CMOD5.n 60 deg

36

Hurricane Force ObservationsHurricane Force ObservationsOct’07 ~ May’08Oct’07 ~ May’08

0

20

40

60

80

100

120

140

160

180

200

QSCAT-Hi QSCAT-p ASCAT GFS ECMWF Surf. Obs.

195

168

2

38

415

• Loss of QuikSCAT data could possibly result in 80-

90% reduction in detection of hurricane force winds

• Necessary to investigate improvement in ASCAT

wind retrievals are in order to support wind warning

and forecast products for extreme winter ocean

storms

•ASCAT underestimate high wind speeds in comparison to QuikSCAT

•Hurricane force wind detection from ASCAT is very rare

Can ASCAT high wind speed

retrievals be improved?

37

Hurricane Force ObservationsHurricane Force Observations( Oct’07 ~ May’08 )( Oct’07 ~ May’08 )

0

50

100

150

200

250

QS -12.5 km QS - 25 km AS - opr AS - new GFS OBS

212188

2 39 38 19

38

Operational ASCAT wind product detects STORM force wind. One wind warning category lower than the actual winds.

North Atlantic Extratropical Storm ExampleNorth Atlantic Extratropical Storm Example

QuikSCAT & New ASCAT wind products detects HF wind.

New ASCAT Operational ASCAT

QuikSCAT

39

40

Resolution effect

Ernesto RodriguezOVWST meeting 2009

41

Smoothing

N

nn

N

nnnew

xB

xBBB

new

10

10

00

)(

)(),(),(

42

B0 vs wind speed