Post-EPS ScatterometerPost EPS Scatterometer Performance Simulation

Ad.Stoffelen@knmi.nlMaria Belmonte, UCAR,

Jos de Kloe, KNMI

ESA Study

IOVWST, May 2010

Wind retrieval and noise model Kp noise Geophysical noise due to ocean variabilityApproximate retrieval functionsApproximate retrieval functions

IOVWST, May 2010

Geophysical noise

Wind Vector Cell

alon

g-ec

tion

Fore FOVs

• Different kinds of FOVs are combined (views)

• Each WVC view represents

25 kmSate

llite

tr

ack

dire Aft FOVs

pa different areal mean

• The ocean surface is variable

25 kmSatellite across-track direction

• A geophysical error occurs due to ocean surface variability and WVC non-

if liuniform sampling• Mainly affects low winds

P t b ll & St ff l 2006

IOVWST, May 2010

Portabella & Stoffelen, 2006

Post EPS scatterometer (SCA) [baseline requirements and options]

• Spatial resolution (25 km) MetOp orbit Sun Sync • Dynamic range (4-25 m/s)• Radiometric resolution (~3-10% at 4 m/s)• Swath coverage (95% in 48 hours for incidences between 20 and 60)

p ywith 820 km altitude

g ( )

I - Fixed beam (ASCAT type) II - Rotating beam (RFSCAT type)

15% improvement with respect to ASCAT on MetOp

IOVWST, May 2010

Discarded: Ku-band (rain), pencil beam (skill), extended nadir coverage for ASCAT type

Specify complete SCA arrangement:p y p g1) Antenna configuration (C-band, single vs dual pol):

t t l- total power- dimensions - radiation pattern

Pseudo Level 1B fileOrbital model

2) Radar waveform(FM chirp, short vs long pulse):- PRF- chirp bandwidth

Incidence / AzimuthPolarization

NESZ # viewschirp bandwidth- noise estimation Nlooks

Nnoise

# views

IOVWST, May 2010

WVC(25 km resolution cell)

Satellitepositionat time t

Radiometric resolution (NESZ and Kp)Radiometric resolution (NESZ and Kp)

1) NESZ (Noise Equivalent Sigma Zero) for a single look:

look range azimuthA

00

2

3 4

( )B eq look

lookt TX RX

k T T BNESZSNR APG G

R L

l kx yN

3 44 propR L

2) Number of looks per node: (reduce speckle)lookslook

NA

noise s noiseN f T

2) Number of looks per node:

3) Number of noise samples:

(reduce speckle)

(noise estimation)

2 202

20

var{ } 1 1 1 11plooks noise

KN SNR N SNR

Radiometric resolution:

IOVWST, May 2010

SCA end-to-end simulator

Input wind vectorPencil beam, Kp =20%

9 m/s @ 90 deg Outer swath

GMFPseudo L1B file Geophysical noise

Backscatter vector observation Observational likelihood

Wind inversionlikelihood

Output wind vector(minimum MLE solution)

20 0 ( )

IOVWST, May 2010

Pobs(Vout|Vin)

0 0,

22 01,...,

( , )1( , ) i GMF i

i N p i

wMLE w

MLE K

Wind retrieval performanceWind retrieval performancePobs(Vout|Vin) NWP prior (5 m2/s2) Pobs(Vout|Vin)*PNWP

1) Wind Vector RMS error

2) A bi i ibiliof M

erit

v

2) Ambiguity susceptibility

3) Wind biases (skewness)Figu

re o

u

IOVWST, May 2010 1/ 22 2( ) ( | ) ( )obs true true obs true bgRMS v v v p v v p v d v

For example:

Wind retrieval performance QSCAT/ASCAT

Vector RMS error QSCAT ASCAT

S /

5 m/s

Vector RMS error at 9 m/s

5 m/ssimulated QSCATBIAS

9 m/sretrieved QSCATPDF

IOVWST, May 2010

13 m/sAMBIGUITY

Climatology FoMsClimatology FoMs

Wind retrieval performance is dependent on input wind and across track distancep p p

Use a climatology average over wind speeds (3-16 m/s)

IOVWST, May 2010

Use a climatology average over wind speeds (3 16 m/s)

SCA assessment

ASCAT RFSCAT RFSCATASCATtype

RFSCAT(1rpm)

RFSCAT(2rpm)

RFSCAT(3rpm)

IOVWST, May 2010Wind Vector RMS error across swath

ConclusionRFSCAT performs well compared to ASCAT configuration, but …

IOVWST, May 2010

To consider: geophysical noise, HH polarization, resolutionTo optimize: antenna pattern…

scat@knmi nlscat@knmi.nlwww.knmi.nl/scatterometer

IOVWST, May 2010

Backup SlidesBackup Slides

IOVWST, May 2010

Wind retrieval and noise model Kp noise Geophysical noise due to ocean variabilityApproximate retrieval functionsApproximate retrieval functions

IOVWST, May 2010

GMF issuesGMF issues2D-histogram histogram (1st-rank Solution)

50

• Measured triplets are centered well on cone within Kp for all

40

INSIDEOUTSIDE

well on cone within Kp for all speeds

• Geophysical noise at low winds incorporated

30

Win

d S

peed

(m

/s)

• Reasonable symmetry at medium-high winds

• Around 4 m/s most triplets inside 10

20

AS

CA

T W

i

1.4•

101

1.5•

102

1.5•

103 p

the cone• At very low winds, opposite effect

-40 -20 0 20 400

10

1.4•101

1.4•101

1.5•102

1.5•102

1.5

1.5•10 4

IOVWST, May 2010

-40 -20 0 20 40MLE

1015

1010

1015

1010

1015 10001005

1010

1

ERS-2

995

1000

1005

1010

990

99510

00

10

10

1015

• Warm steady-flow air

985

990995

1000

1

985

100

1005

1010 discerned from polar

gusty air.• Wind variability

985

990

995

1005

101

985

990

990

995

10

1 Wind variability causes triplet inconsistency

985

990

995

95

1000

010 990

995

1000

1000

1005

1010 • Noise at edge of the swath; ASCAT moved outward

IOVWST, May 2010

1000

1005

1005

1010

1005

5

10101015

moved outward

Geophysical noise

Wind Vector Cell

alon

g-ec

tion

Fore FOVs

• Different kinds of FOVs are combined (views)

• Each WVC view represents

25 kmSate

llite

tr

ack

dire Aft FOVs

pa different areal mean

• The ocean surface is variable

25 kmSatellite across-track direction

• A geophysical error occurs due to ocean surface variability and WVC non-

if liuniform sampling

Portabella & Stoffelen, 2006

IOVWST, May 2010

Accuray on 50-km WVC scale• Triple collocation analysis of buoy,

scatterometer & NWPVector RMS error

[m/s]Tropical

TAO/PIRATAExtratropical

NDBC/MEDS/UKMO

Buoy 1.5 1.5

Scatterometer 1.2 1.6

ECMWF model 2.0 2.1

Scatterometer winds provide excellent forcing Remaining errors include representativeness

IOVWST, May 2010

Remaining errors include representativeness

1.0 1008

1.0 1009

u ASCATu ECMWF

v ASCAT100 km

1.0 1007

1.0 10v ASCATv ECMWFk -5/3

100 km

1.0 1006 AWDP@12.5

04

1.0 1005

Ψ(k) • ASCAT contains small

scales down to 25 km

1.0 1003

1.0 1004

• Improved w.r.t. SeaWinds

1.0 1002

• No noise floor• k-1.9

IOVWST, May 20101.0 10-07 1.0 10-06 1.0 10-05 1.0 10-04

k (m-1)

1.0 1001

MesoscalesMesoscales

• 12 5-km box12.5 km box details appear spectrally correctspectrally correct

• It verifies well with buoysbuoys

• It corresponds ll ith l dwell with cloud

features

IOVWST, May 2010www.knmi.nl/scatterometer

High WindsHigh Winds C-band Model

F iFunctionC b d HH iti t hi h•C-band HH sensitive to high

winds

•No EUM priority due to lack•No EUM priority due to lack of high winds

log

Courtesy

IOVWST, May 2010

log Courtesy D. EstebanJPL, NASA

ASCAT L1 backscatter averagingWind Vector CellHamming filter H i filtHamming filter averaging window

Hamming filter

ASCAT Level 2 product:

100 km

ealon

g-ec

tion

• 50 km resolution

• up to 60 km off the coast

to avoid land effects

25 km

Coa

stlin

e

Sate

llite

tr

ack

dire to avoid land effects

100 km25 km

C

Satellite across-track direction

ASCAT Coastal product:

• 30-40 km resolution

IOVWST, May 2010

10 km • up to 25 km off the coast

• more noisy

ASCAT L1 backscatter averagingWind Vector CellHamming filter H i filtHamming filter averaging window

Hamming filter

ASCAT Level 2 product:

100 km

ealon

g-ec

tion

• 50 km resolution

• up to 60 km off the coast

to avoid land effectsB25 km

Coa

stlin

e

Sate

llite

tr

ack

dire to avoid land effectsBox

100 km25 km

C

Satellite across-track direction

Box

ASCAT Coastal product:

• 30-40 km resolution

IOVWST, May 2010

10 km • up to 25 km off the coast

• more noisy

Prototype at 25 km

IOVWST, May 2010

1.0 1008

1.0 1009

u ASCAT OSISAFu ASCAT Box avgv ASCAT OSISAFv ASCAT Box avg

100 km

k -5/3

1.0 1007

1.0 10v ASCAT Box avg

100 km

Box AWDP@12.51.0 1006

• Box averaging maintains more tail variance

• No apparent noise floor04

1.0 1005

Ψ(k)

No apparent noise floor• Buoy verification confirms

this; see later presentation1.0 1003

1.0 1004

• Still u bump, but at lower wavelength (?)

• k -1.8 pretty close to -1 67 for1.0 1002

IOVWST, May 2010

k , pretty close to 1.67 for 3D turbulence Nastrom and Gage 1987

1.0 10-07 1.0 10-06 1.0 10-05 1.0 10-04

k (m-1)

1.0 1001

KNMI ASCAT on MetOp Phasing Study28 i t l l• 28 minutes clearly optimal

• No gain in tropics atNo gain in tropics at 50 minutes

28 minutes

IOVWST, May 2010 45 minutes

Antenna assembly

C-band, VV polarization (extension to dual HH polarization an option)

Swath FoV 3dB widthSwath FoV 3dB widthSpatial resolution 3dB length Antenna effective dimensions

Beam shaping in elevation20% better than

ASCAT M tO

30

321-way antenna gain projected into swath width

Beam shaping in elevationASCAT on MetOp(from 20 to 16 km)

Dynamic range24

26

28

1-w

ay a

nten

na g

ain

[dB

]

300 400 500 600 700 80020

22

ground range [km]

MIDSIDE

Slotted waveguide radiators:

2

3 4 04( )

TX RX

prop lookt

G GR L ASNR P

k T T B

From ASCAT on METOP

IOVWST, May 2010

to compensate for larger slant range across swath (~10 dB)

0( )B eq lookk T T B

[SNR is determined by Pt and chirp bandwidth]

Radar PRFLimited by swath extent: two different strategies

TSWATH MID SIDE

TNEAR 5.7 ms 8.4 ms

TFAR 6.0 ms 10.5 ms

Inc = 20 to 53.5 deg

RNEAR RFAR

A) Long pulse (TTX >> tSWATH)

FAR

PRF = 1/(THOR+TTx+TN) = 29.4 Hz (30%DC)

SWATH

TRx = TTx-TSWATH

Nadir TSWATH Horizon @ 22.2 ms

TTx = TFAR

TTx TN

HorizonNadir

B) Short pulse (TTX << tSWATH)

TSWATH

PRF = 1/(TFAR+TTx) = 230 Hz (7%DC)

IOVWST, May 2010

TRx

TTx = 0.3 ms TN

2) Radar waveform

LFM chirp for range resolution

0.5

1

Chirp ~ exp[i(/2)t2]1 2 3 4

-1

-0.5

( = chirp rate BTx/TTx )

Deramping ~ exp[-i(/2)t2] exp[ifDt+i(/2)(t-tr)2] = exp[i(fD+itr)t]

1

FFT

Echo bandwidth Becho (tFAR – tNEAR) + fD

Detection bandwidth Bdetection 1/TRX = Blook

Swath

detection RX look

Resolved look area:

IOVWST, May 2010 dBecho/dx 2(/c)sin(inc)

Alook= 16 km x Blook/[2(/c)sin(inc)]

Inversely proportional to chirp rate, but so is SNR!

Radiometric resolutionR i d i l b k kl i iReceived signal = backscatter + speckle + emission

2 202

20

var{ } 1 1 1 11pKN SNR N SNR

1) Accumulate independent looks to reduce speckle

0plooks noiseN SNR N SNR

N [2( /c)sin(inc)/B ] L

Naz = (PRF/vground) LWVC

Nlooks = Naz Nel

Maximum

Maximum PRF(compatible with unambiguous range)

Nel = [2(/c)sin(inc)/Blook] LWVC Maximum (compatible with SNR >-1.5dB)

WVC SNR Pt0/ sin(inc) Minimum Pt (compatible with Kp requirements)

ITER

ATE

Nnoise = NazBechoTN

2) Noise (emission) estimation and subtractionIndependent looks

IOVWST, May 2010

Where Becho = fs/2 sets the sampling frequency

Trade-offs to optimize Kp via chirp rate and total power (antenna pattern)

Min SNR check Max Kp check

20° 27 4°20 27.4

N li l d t N li l d t

53.5° 64.7°

Non-compliance leads toincrements in peak power

Non-compliance leads toincrements in chirp rate

(and/or antenna pattern accommodation?)

IOVWST, May 2010

Compliant SCA configurations enter the wind performance study

Complete SCA configurationComplete SCA configuration

1) Antenna assembly total power & pattern Orbital model- total power & pattern

2) Radar waveform- PRF chirp rate & noise

Pseudo Level 1B file

Orbital model

PRF, chirp rate & noise

WVC Node

Row/Col Satellite position and velocityLat/Lon # Views

ViewsViewcounter

0 ( )k T T B

NESZ (Noise Equivalent Sigma Zero):

IOVWST, May 2010

Azimuth Incidence # Looks 1/ NESZ Polarization

0

2

3 4

( )

4

B eq look

lookt TX RX

prop

k T T BNESZSNR APG G

R L