2nd Advanced Course on Radar Polarimetry 2013 01/24/2013 · 2020. 12. 18. · VHF 300 KHz - 300 MHz...
Transcript of 2nd Advanced Course on Radar Polarimetry 2013 01/24/2013 · 2020. 12. 18. · VHF 300 KHz - 300 MHz...
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
1
Surface Parameter Estimation:Basics and Advanced Concepts
2nd Advanced Course on Radar Polarimetry – 24.Jan 2013 – [email protected]
Irena HajnsekChair of Earth Observation and Remote SensingInstitute of Environmental Engineering, ETH ZürichMicrowaves and Radar Institute, DLR Oberpfaffenhofen
Slide 2
Part I: Theory and basicsIntroduction into SAR polarimetric observablesDecomposition techniques Introduction into material properties of soilModel description and inversion strategy
Part II: Exercises with L-band airborne dataRead the dataSpeckle filteringOh, Dubois and X-Bragg inversion
OverviewHH HV
VH VV
Slide 2
Slide 3
Motivation of Radar Remote Sensing
Independent from WeatherData acquisition during cloud cover, rain, ...
Radar is complementarily to opticsSensitive to dielectrical and geometrical properties of scatterers,...
Penetrates through or in volumesOverlay of scattering processes,…
Is an active systemData acquisition independent of daylight, ...
Delivers highly accurate physical productsDigital Elevation Models, Forest heights, Soil moisture maps ...
Slide 3Ethna DEM (TanDEM-X) Forest height Map
L (23cm)C (5cm)X (3cm)
35,3 km
21,3
km
Optical image
DLR Location: Oberpfaffenhofen
B:SurfaceR:Dihedral G:VolumeSAR-Bild
Drygalski glacier velocity measurement at the Antarctic Peninsula
(TerraSAR-X)
TanDEM-X dual polarisations acquisition of a Saudi
Arabien desert, montage with a Google-Earth image.
1994: Shuttle Radar Lab SIR-C/X-SAR (Brazilian Rainforest)
Slide 4
What does the Radar measure ?
Radar reflectivity (backscattered signal) of the target as a function of position.
radar transmits a pulse (travelling velocity is equal to velocity of light)
some of the energy in the radar pulse is reflected back towards the radar.
This is what the radar measures.It is known as radar backscatter o(sigma nought or sigma zero).
Slide 4
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
2
Slide 5Slide 5
Normalized radar cross-section (backscattering coefficient) is given by:
o (dB) = 10. Log10 (energy ratio)
whereby received energy by the sensor
“energy reflected in an isotropic way”energy ratio =
i.e.
The backscattered coefficient can be a positive number if there is a focusing of backscattered energy towards the radar
or
The backscattered coefficient can be a negative number if there is a focusing of backscattered energy way from the radar (e.g. smooth surface)
What does the Radar measure ?
Slide 6Slide 6
Backscattering Coefficient o
Levels of Radar backscatter Typical scenario
Very high backscatter (above -5 dB) man-made objects (urban)terrain slopes towards radarvery rough surfaceradar looking very steep
High backscatter (-10 dB to 0 dB) rough surfacedense vegetation (forest)
Moderate backscatter (-20 to -10 dB) medium level of vegetation agricultural crops moderately rough surfaces
Low backscatter (below -20 dB) smooth surface calm waterroad very dry terrain (sand)
Slide 7Slide 7
Commonly Used Frequency Bands
Frequency band Frequency range Application examples
VHF 300 KHz - 300 MHz foliage/ground penetration, biomass
P-Band 300 MHz - 1 GHz soil moisture, biomass, penetration
L-Band 1 GHz - 2 GHz agriculture, forestry, soil moisture
C-Band 4 GHz - 8 GHz ocean, agriculture
X-Band 8 GHz - 12 GHz agriculture, ocean, high resolution radar
Ku-Band 14 GHz - 18 GHz glaciology (snow cover mapping)
Ka-Band 27 GHz - 47 GHz high resolution radar
Slide 8
Environmental Sensing from Air and Space
Airborne measurementsHighly flexible operationCoverage of dedicated areasExperimental configurationSensor specific data formatsShort re-visit times
Slide 8
Spaceborne measurementsHighly regular observationWide area coverageHighly operational & reliableStandard product deliveryLong term observations
DLR‘s E-SAR DLR‘s TanDEM-X Mission
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
3
Slide 9Slide 9
HHX-band
PI-SAR / Test Site: Gifu-Japan
Distributed Scatterers:•Norm. Backscattering Cross Section σ0 ;•Texture (σ0 statistics).Deterministic Scatterers:•Radar Cross Section (RCS) ;•Phase of the SAR signal. O
bser
vatio
n Sp
ace
Appl
icat
ions
Single Pol (Channel) SAR
Classification/Segmentation (Texture);Change detection (Multi temp. analysis);Glacier velocities (Feature tracking); Mapping (Feature extraction/Bordering );Ocean wave/Wind mapping;Coherent scatterers (CSs).
Slide 10
Polarimetric SAR
Slide 10
VVVH
HVHH
SSSS]S[
HH
VV
HV
VHFirst Order Parameters:
Norm. Backscattering Cross Section σ0 ;Intensity (Cross Section σ0) ratios;Polarimetric phase differences.
Second Order Parameters:Polarimetric (inter channel) coherences.O
bser
vatio
n Sp
ace
Slide 11Slide 11
X-band
PI-SAR / Test Site: Gifu-Japan
R:HH-VV G:HV+VH B:HH+VV
First Order Parameters:Norm. Backscattering Cross Section σ0 ;Intensity (Cross Section σ0) ratios;Polarimetric phase differences.
Second Order Parameters:Polarimetric (inter channel) coherences.O
bser
vatio
n Sp
ace
Appl
icat
ions
Soil moisture/roughness estimation;Wettland/Veg. characterisation/mapping; Snow & Ice mapping (type classification);Ship & Oil spill detection; Classification/Segmentation (Pol based).
VVVH
HVHH
SSSS]S[
Polarimetric SAR
Slide 12Slide 12
)r(E)r(E
)r(S)r(S)r(S)r(S
r)riexp(
)r(E)r(E
iv
ih
VVVH
HVHHsv
sh
)r(E)r(E
)r(E sv
shs
hv
Scattered Field
In the far zone region
( )
2x2 Complex Scattering Matrix
|'r||r| |r|
Incident (Plane) Wave
(Jones Vector Representation)
)r(E)r(E
)r(E iv
ihi
hv
(Jones Vector Representation)
Mapping of the 2-dim incident vector into the 2-dim scattered vector)(rEihv )(rEshv
Scatterer
Transforms the incident
into the scattered wave
The Polarimetric Scattering Problem
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
4
Slide 13Slide 13
)r(E)r(E
)r(S)r(S)r(S)r(S
r)riexp(
)r(E)r(E
iv
ih
VVVH
HVHHsv
sh
)r(E)r(E
)r(E sv
shs
hv
Scattered Field
In the far zone region
( )
2x2 Complex Scattering Matrix
|'r||r| |r|
Incident (Plane) Wave
(Jones Vector Representation)
)r(E)r(E
)r(E iv
ihi
hv
(Jones Vector Representation)
Mapping of the 2-dim incident vector into the 2-dim scattered vector)(rEihv )(rEshv
1: Changes the polarisation state of the incident wave
2: Changes the degree of polarisation of the incident wave
Scatterer
Transforms the Incident
into the scattered wave
The Polarimetric Scattering Problem
Slide 14
Coherent Scattering Matrix
Slide 14
iv
ih
VVVH
HVHHsv
sh
EE
SSSS
rri
EE )exp(
)exp(||)exp(||)exp(||)exp(||)exp(][
VVVVVHVH
HVHVHHHH
iSiSiSiS
rriS
[S] is independent on the polarisation of the incident wave !!!
… also known as the Jones Matrix in the bistatic and Sinclair Matrix in the monostatic case
IJS
Complex Scattering Amplitudes:
=f(Frequency,Scattering Geometry)
and depends only on the physical and geometrical properties of the scatterer
||)(exp||)(exp||)(exp||)exp()exp(][
VVVVVHVH
VVHVHVVVHHHHVV
SiSiSiS
ririS
Seven Parameters: 4 Amplitudes & 3 PhasesAbsolute PhaseFactor
Total Scattered Power: 2VV2
VH2
HV2
HH SSSSSSTraceSSpanTP ||||||||)]][([])([
Slide 15
Bi- & Mono-Static Measurement of the Scattering Matrix
Slide 15
T = 1/PRF
VVVH
HVHH
SSSS
V
H
V
H
V
H
V
H
H V VH Tx
RxT
H
V
H
V
R
R T
H
V
HH
HV
VH
VV
Slide 16Slide 16
VH VV
HH HV
Azi
mut
h
E-SAR / Test Site: Oberpfafenhoffen
Scattering Amplitude Images
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
5
Slide 17
Backscattering (FSA & BSA)
Slide 17
||)(exp||)(exp||)(exp||)exp()exp(][
VVVVXXXX
VVXXXXVVHHHHVV
SiSiSiS
ririS
Five Parameters: 3 Amplitudes & 2 PhasesAbsolute PhaseFactor
In the case of monostatic backscattering from reciprocal scatterers:
BSAXX
BSAVH
BSAHV SSS )( FSAXX
FSAVH
FSAHV SSS Reciprocity Theorem
VVXX
XXHH
SSSS
S][
VV
XX
XX
HH
L4
SSSS
k
0S2
SSSS
21k
XX
VVHH
VVHH
P4
VV
XX
HH
L3
SS2
Sk
XX
VVHH
VVHH
P3
S2SSSS
21k
3-dim Lexicographic (L) and Pauli (P) Scattering Vectors:
Note: The factor is required to keep the vector norm of invariant L3k
2
The scattering problem can be addressed in the 3-dim complex space:
Slide 18
Partial Scatterers
Slide 18
Deterministic Scatterers Partial Scatterers
Completely described by [S] Cannot be described by a single [S]
• Change the polarisation state of the wave • Change the polarisation state of the wave
• Do not change the degree of polarisation and also change the degree of polarisation
Scatterers with Space or Time VariabilityPoint Scatterers
MonochromaticIncident Wave
MonochromaticScattered Wave
Depolarisationdescribed by second order statistics
Slide 19Slide 19
2
2
2
3
||4)(2)(2)(2|)(|))(()(2))((|)(|
:][
HVVVHHHVVVHHHV
HVVVHHVVHHVVHHVVHH
HVVVHHVVHHVVHHVVHH
SSSSSSSSSSSSSSSSSSSSSSSSS
T
Covariance & Coherency Matrices in Backscattering
Covariance Matrix [C]:
LL kkC 333 :][
2
2
2
3
||22||22
2||:][
VVHVVVHHVV
VVHVHVHHHV
VVHHHVHHHH
SSSSSSSSSS
SSSSSC
Coherency Matrix [T]:
PP kkT 333 :][
Pauli Scattering Vector:
TXXVVHHVVHHP SSSSSk 221
3
Lexicographic Scattering Vector:
TVVXXHHL SSSk ]2[4
and are by definition 3x3 hermitian positive semi-definite matrices &contain in general 9 independent parameters
][][ 33 TC
Slide 20Slide 20
Scattering Polarimetry
INTERPRETATION OF SCATTERING MECHANISMS
• Directly from the Scattering Matrix
• Model based
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
6
Slide 21
Interpretation of Scattering Mechanisms
Slide 21
VVHV
HVHH
SSSS
S][ T
xxVVHHVVHHP S2SSSS21k
PP
kk1e
XX
VVHH
VVHH
P3
2
1
PP S2
SSSS
k21
ii
ik
k1e
expsinsinexpcossin
expcos
Scattering Matrix: Scattering Vector:
Unitary Representation:
Parameterisation of in terms of five angles:e 321 ,,,,
Slide 22Slide 22
001
10000
00
001
i000i000i
e
3
2
1
cossinsincos
cossinsincos
expexp
exp
Interpretation of Scattering Mechanisms
e00
001e
cossinsincos' e
10000
e
cossinsincos
'
Point Reduction Theorem:
Change of Scattering Mechanism:
characteristicorientationphase
Slide 23Slide 23
Interpretation of Scattering Mechanisms
010
001
100001010
100010001
e'
02121
001
1000212102121
100010001
e'
001
10000
00
001e
cossin
sincos
cossinsincos'
H-DipolScatterer
Dihedral Scatterer
=90º and β=0º
=45º and β=0º
HV
VVHH
VVHH
P S2SSSS
k21
001
e
02121
001
1000212102121
100010001
e'V-Dipol
Scatterer
=45º and β=180º
where
Slide 24Slide 24
Scattering Processes: Fresnel Scattering
2
2
sincossincos
HR
2
2
sincossincos
VR
V
H
RR
S0
0][
Fresnel Reflection Coefficients:
Scattering Matrix:
… where e is the dielectric constant of the surface
and
HH
VV
= 20.
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
7
Slide 25Slide 25
Scattering Processes: Bragg Scattering
2
22
sincos)]sin1()[sin1(
VR
Bragg Scattering Coefficients:
Scattering Matrix:
… where is the dielectric constant of the surface
V
H
RR
S0
0][
2
2
sincossincos
HR and
VV
HH
= 20.
Slide 26Slide 26
Scattering Processes: Dihedral Scattering
0 001 0 1 0[ ]
0 000 1 0HS HS HTHT
iiVS VS VTVT
R R RRS
R R R eRe
2
2
cos sin
cos sinS
HS
S
R
2
2
cos sin
cos sinS S
VS
S S
R
2
2
cos(90 ) sin (90 )
cos(90 ) sin (90 )T
HT
T
R
2
2
cos(90 ) sin (90 )
cos(90 ) sin (90 )S S
VT
S S
R
Fresnel Coefficients:
Scattering Matrix:
and
and
HH
VV
= 20.
Soil (S)
Trunk (T)
Slide 27Slide 27
Dihedral vs Volume Scattering (L- /P-Band E-SAR – Kryklan/Sweden)
08biosar0201x1_t11 L-band, hh,hv,vv 08biosar0302x1_t12 P-band, hh,hv,vv
tree trunkpower line
tree crown
range direction
Slide 28Slide 28
dbba
S][
c2baba
21k
10000
R3
cossinsincos
:)(
kRRRk 333
)]([)]([)]([),,(
cossin
sincos:)(
001
0R3
10000
R3
cossinsincos
:)(
Scattering Processes: Volume Scattering
20
0
20
Rotation about x-axisRotation about y-axisRotation about z-axis
),,(),,(][ kkT
Coherency Matrix:
Random Volume
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
8
Slide 29Slide 29
Scattering Processes: Volume Scattering
z
y
x
000000
][ 11L4
Vri
i ))/((
0
23z
23y
23x
zyxi ds2ls2ls2ls2
8lllL /// )/()/()/(
/)(
zyxzyx l
1l1
l1LLL ::::
1m
2m11L11LA
r
r
ry
rxP
)()(: l
lLLm
x
y
y
x :
1LLL zyx
2l
2l
2l
43V zyxwhere is the particle volume and
with
Principal Polarisability Matrix:
Particle Anisotropy: Particle Shape Ratio:
m0
1m
1m oblate spheroids
prolate spheroids
Slide 30Slide 30
Scattering Processes: Volume Scattering
T3
T3
T3333 RRRPRRRP )]([)]([)]([][)]([)]([)]([),,]([
dddpppPT )()()(),,]([][
b000b0001
aT][)(
)(2PP
2P
A7A6221Ab
50b0 .
Special Case:Random Volume
p( )
-
p( )
-
p( )
- / 2 / 2
where
Coherency Matrix:
where )(),(),( ppp are the pdf’s for the corresponding angle distributions
Slide 31Slide 31
Scattering Processes: Volume Scattering
A=0.01
A=0.20
A=0.40
A=0.60
A=0.80
A=1.00
H/a Loci for varying particle
shape and width of distribution
Prolate Particles
oriented volume random volume
dipole
sphere
-dipole
randomness
Slide 32
[ S ]
COHERENTDECOMPOSITION
E. KROGAGER(1990)
W.L. CAMERON(1990)
[ K ]
[ T ] [ C ]
TARGETDICHOTOMY
EIGENVECTORS BASEDDECOMPOSITION
EIGENVECTORS / EIGENVALUES ANALYSIS&
MODEL BASED DECOMPOSITION
EIGENVECTORS / EIGENVALUES ANALYSISENTROPY / ANISOTROPY
S.R. CLOUDE - E. POTTIER(1996-1997)
A.J. FREEMAN(1992)
W.A. HOLM(1988)
S.R. CLOUDE(1985)
J.J. VAN ZYL(1992)
AZIMUTHAL SYMMETRY
J.R. HUYNEN(1970)
R.M. BARNES(1988)
Decomposition Theorems
MODEL BASEDDECOMPOSITION
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
9
Slide 33Slide 33
Decomposition Theorems
• Pauli Matrix Decomp.
• Model Based Decomp.
• Eigenvector Decomp.
Slide 34
Pauli Matrices Decomposition
Slide 34
00
0110
1001
1001
][i
idcba
baidcidcba
SSSS
SVVVH
HVHH
1001
aSa][
1001
bSb][
0110
cSc ][
0ii0
dSd ][
Single Scattering:
Dihedral Scattering ( … rotated by /2 about the LOS )
Dihedral Scattering:
Transforms all polarisation states into their orthogonal states (disappears in backscattering)
VVHH SS
VVHH SS
Slide 35Slide 35
Azi
mut
h
Scattering Amplitude Images
HH
C-band 5 cm
SIR-C / Test Site: Kudara,Russia
VV HVSlide 36Slide 36
Scattering Amplitude Images
HH
L-band
23 cm
SIR-C / Test Site: Kudara,Russia
VV HV
Azi
mut
h
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
10
Slide 37Slide 37
Pauli Images
C-band L-band
SIR-C / Test Site: Kudara,Russia
HH+VV
HH-VV
2HV
RGB-Coding:
Slide 38Slide 38
Decomposition Theorems
• Pauli Matrice Decomp.
• Eigenvector Decomp.
• Model Based Decomp.
Slide 39
Eigenvector Decomposition
Slide 39
Coherence Matrix: P3P33 kkT
:][
Diagonalisation: 1333 UUT ]][][[][
333231
232221
131211
3][eeeeeeeee
U
3
2
1
000000
][
332313
322212
312111
31
3 ][][eeeeeeeee
UU
][][][)()()()(][ 332
31
3333222111
3
1iiii3 TTTeeeeeeeeT
0321
33
23
13
3
32
22
12
2
31
21
11
1
eee
eeee
eeee
e
where:
3 real positive eigenvalues 3 orthonormal eigenvectors
][][][ 321 SSS
Slide 40
Scattering Entropy / Anisotropy
Slide 40
Coherency Matrix Diagonalisation:
][][][)()()()(][ 332
31
3333222111
3
1iiii3 TTTeeeeeeeeT
32
32
32
32
PPPPA
:
321
iiP
:i3
3
1ii PPH log:
with:Scattering Entropy:
Scattering Anisotropy:
Totally Polarised Scatterer
Totally Unpolarised Scatterer
0H
1H1H0 where:
0A
1A1A0
2 Equal Secondary Scattering Processes
Only 1 Secondary Scattering Processwhere:
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
11
Slide 41Slide 41
Scattering Entropy Images
C-band L-band
H=0
H=1
SIR-C / Test Site: Kudara,Russia
Slide 42Slide 42
Scattering Anisotropy Images SIR-C / Test Site: Kudara,Russia
C-band L-band
A=0
A=1
Slide 43
Mean Scattering Parameters: Eigenvector
Slide 43
Coherency Matrix Diagonalisation:
321
iiP
:
Mean α-Angle:
Mean β-Angle:
Mean γ-Angle:
Mean δ-Angle:
332211 PPP
332211 PPP
332211 PPP
332211 PPP
)()()(][ 3332221113 eeeeeeT
)exp()sin()sin()exp()cos()sin(
)exp()cos(
ii
ie
)exp()sin()sin()exp()cos()sin(
)exp()cos(
iii
iii
ii
i
ii
ie
Eigenvectors: Appearance Probabilities:
Mean Scattering Mechanism
Mean ε-Angle: 332211 PPP
Slide 44Slide 44
-Angle Images
C-band L-band
a=0°
a=90°
a=45°
SIR-C / Test Site: Kudara,Russia
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
12
Slide 45
H- Angle Plane
0 0.2 0.4 0.6 0.8 10
10
20
30
40
50
60
70
80
90
Low Entropy Multiple Scattering
Medium Entropy Multiple
Scattering
Low Entropy Surface Scattering
Medium Entropy Dominant Surface Scattering
Medium Entropy DipolScattering
Low Entropy Dipol Scattering
0 0.2 0.4 0.6 0.8 10
10
20
30
40
50
60
70
80
90
High Entropy MultipleScattering
Entropy (H)Al
pha
( )
Slide 46Slide 46
Decomposition Theorems
• Pauli Matrice Decomp.
• Eigenvector Decomp.
• Model Based Decomp.
Slide 47
Freeman 3 Component Decomposition
Slide 47
][][][][ VDS TTTT
000010
fT
2
SS
][
000010
fT
2
DD
][
100010002
fT VV ][
Vegetation Scattering : Bragg-Scattering + Dihedral Scattering + Random Volume of Dipols
V
VDSDS
DSVDS
fffffßf
fßfffßfT
00002
][
22
4 Equations for
5 Unknowns
where and
Re(HHVV)-fv/3 > 0 Single bounce is dominant alpha =1
Re(HHVV)-fv/3 < 0 Double bounce is dominant beta =1
Slide 48
3 Component Freeman Decomposition (San Francisco)
Slide 48
Pauli |HH-VV|, |HV|, |HH+VV| Freeman/Durden, PD, PV, PS AirSAR JPL
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
13
Slide 49
][][][ / VDS TTT
Slide 49
Volume shape parameter of
a randomly oriented volumedipole sphere
=1/3 =1
)3()||1( 2 VVGG fPfPScattering Power Components:
Distinction between dihedral and
surface of the ground component:
1 dihedral
≥1 surface
100010001
000010||
0000 2
33
22
1211
12 VGff
TTTTT
Freeman 2 Component Decomposition
Slide 50
Summary: Polarimetric Parameters / Scattering Matrix
Slide 50
• Scattring Amplitudes
• Total Power
• Amplitude Ratios
• Pol Phase Differences
• Helicity
| | exp ( ) | | exp ( )exp( )exp( )[ ]| | exp ( ) | |
HH HH VV XX XX VVVV
XX XX VV VV
S i S ii r iSS i Sr
Five Parameters: 3 Amplitudes & 2 PhasesAbsolute PhaseFactor
[ ] HH XXXX VV
S SS
S S
0 * 0 * 0 *: | | : | | : | |HH HH HH N XX XX XX N VV VV VV NS S S S S S
0 0 0 0 0 0 0/ / / ( )HH VV XX VV XX HH VV
:HHVV HH VV
: | | | |LL RRHel S S
2 2 2: | | 4 | | | |VV XX VVTP S S S
Slide 51
Summary: Second Order Polarimetric Parameters
Slide 51
• Correlation Coefficients
• Scattering Entropy
• Scattering Anisotropy
• Polarimetric Alpha Angle
• Polarimetric Beta Angle
3
Ak B
C
3 3 3[ ] :
AA AB AC
X k k BA BB BC
CA CB CC
Nine Parameters: 3 Real & 3 Complex Elements
*
* *
| |:| | | |
HH VVHHVV
HH HH VV VV
S SS S S S
*
* *
| |:| | | |
LL RRLLRR
LL LL RR RR
S SS S S S
2 3 2 3
2 3 2 3
: P PAP P
1 2 3
: iiP
3
31
: logi ii
H P P
1 1 2 2 3 3: P P P
1 1 2 2 3 3: P P P
1: ( )21: ( )2
: 2
HH VV
HH VV
XX
A S S
B S S
C S
:
: 2
:
HH
XX
VV
A S
B S
C S
Pauli Scat. Vector Lexico. Scat. Vector
Slide 52Slide 52
·Agriculture/Land-UseCrop Classification/Moisture Content EstimationUrban Area mappingUrban Topography for Mobile comms
·ForestryBiomass Estimation: (Saturates For High Biomass)
C-band saturation at 50 tons/hectareL-band saturation at 100 tons/hectareP band saturation at 200 tons/hectare
DeforestationForest Canopy Height EstimationTree Species DiscriminationForest Re-growth Monitoring
·GeologyPlayas :Smooth Natural Surfaces(rms = 1cm)Alluvial fans, Sand Dunes, MorainesSedimentary Rock formationsLava Flows (extreme in surface roughness)Weathering Erosion StudiesSurface Roughness Estimates
·HydrologyFlood mapping/Forest InundationSnow HydrologySoil Moisture
·Sea Ice/OceanographyIce Roughness/Thickness StudiesPolar Ice Cap StudiesExtra-Terrestrial Ice/Water Studies
·MeteorologyRain rate estimationWater/Ice particle studies
Severe Storm/Flood warning
·Topography/CartographyDirect Surface Slope EstimationAccurate DEM GenerationDifference of DEMS for Vegetation mapping
·Humanitarian DeminingSurface Penetrating Radar (SPR)SAR for Mine Field Detection
Applications of Radar Polarimetry
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
14
Slide 53Slide 53
Soil Moisture Estimation
• Surface characterisation
• mv estimation over bare surface
• mv estimation under the vegetation
Slide 54
Scattering @ Natural (Bare) Rough Surfaces
Slide 54
SURFACE ROUGHNESS
rms - height σ [cm]
autocorrelation length l [cm]
VOLUMETRIC MOISTURE CONTENT
mv [vol. %]The Backscattered signal depends on themoisture, and roughness of the surface.
Single channel SAR cannot resolveunambiguously the bare surface scatteringproblem.
Surface Characterisation
Slide 56Slide 56
Models for the Estimation of Bare Surface
Modeling and Inversion
EMPIRICALEXTENSIONS
THEORETICALMODELS
Geometrical-Optic 1963
Physical-Optic 1963
Oh Model 1992
Dubois Model 1995Integral Equation 1992
Shi Model 1997
X-Bragg 1999
MODEL BASEDEXTENSIONSSmall Perturbation 1988
Time
• very complex• inversion is
restrictedpossible
• inversion with dual pol observables
• based on regressionscoefficients of a specific test site
• simple modelextension
• based on SPM (1st order) with an roughness extension
Slide 57Slide 57
Models for the Estimation of Vegetated Surface
Modeling and Inversion
THEORETICALMODELS
Vegetation Models(Stiles 2000) SEMI-MODEL BASED
EXTENSIONS
SCATTERING DECOMPOSITION
APPROACHES
Eigen-DecompositionModel based-Decomposition
Hajnsek 2009Jagdhuber 2012
Time
• very complexscat. models
• inversion is not possible
Optimisation (physical and numerical)
Ari 2010Neumann 2010
• simple canonical scat. models
• inversion possible
• extended scat. models• inversion with higher
computation time
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
15
Slide 58
Small Perturbation Model
Slide 58
Roughness:Depolarises the incident wave introducing(depolarised) HV and decreasing polarimetriccoherence.Ambiguity: Terrain Slopes introduce also a HVcomponent but this is co-related to HH and VV.
Moisture content:Effects the scattering amplitude of allpolarisations. For “smooth” (with respect towavelength) surfaces the co-polarimetric ratio(HH/VV) becomes independent of roughness
Polarimetry provides an observation space that allows to separate
roughness from moisture effects.
ε,θR00ε,θR
Z)θcos(k2jSSSS
SrP
rS
VVVH
HVHH
22rr
2r
2r
P)θsinεθcosε(
))θsin1(εθ)(sin1ε(R
θsinεθcosθsinεθcos
R2
r
2r
S
Exact solution of Maxwell Equation for s 0
Z = F.T.(E(x, y))
θ = Incidence angle and k = 2π/λ
p
s
VV
HH
RR
SS
is independent of roughness
Slide 59Slide 59
Small Perturbation Model (Bragg-Model)
Measured versus estimated volumetric moisture content mv [vol. %] using SPM
0
0.2
0.4
0.6
0.8
1
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0m v [vol %]
ks
elbe0-4elbe4-8weih0-4weih4-8
Validity range of the SPM for the ground measurements - surface roughness against
volumetric soil moisture
SPM
0
10
20
30
40
0 10 20 30 40measured m v [vol. %]
estim
ated
mv
[vol
. %]
Weiherbach 0-4 cmWeiherbach 4-8 cmElbe 0-4 cmElbe 4-8 cm
Slide 60Slide 60
Semi-Empirical Models
Oh-Model Dubois-Model
Y. Oh, K. Sarabandi and F.T. Ulaby, "An Empirical Model and an Inversion Technique for Radar Scattering from Bare Soil Surfaces", IEEE
Transactions on Geoscience and Remote Sensing, vol. 30, no. 2, pp. 370-381, 1992.
2
31
0
00201
ks
vv
hh ep
ksvv
hv eq 123.0 000
s0xx : backscattering coefficient for different polarisationsq : incidence angle in radians
er : real part of the dielectric constants : RMS height
k : wavenumber (2p/l)G 0 : Fresnel reflectivity of the surface at nadir
P. C. Dubois, J. J. van Zyl and J. Engman, "Measuring Soil Moisture with Imaging Radar", IEEE Transactions on Geoscience and Remote Sensing,
vol. 33, no. 4, pp. 915-926, 1995.
7.04.1tan028.055.1
75.20 sin10sin
cos10 kshh
7.01.1tan046.033
35.20 sin10sincos10
ksvv
q : incidence angle in radianser : real part of the dielectric constant
s: RMS heightk : wave number (2p/l)
l : wavelength [cm]
Cross-Pol Amplitude Relation Co-Pol Amplitude Relation
Slide 61Slide 61
0
5
10
15
20
25
30
35
40
0 5 10 15 20 25 30 35 40
measured mv [vol %]
estim
ated
mv
[vol
%]
0-4 elbe4-8 elbe0-4 weiherbach4-8 weiherbach
0
5
10
15
20
25
30
35
40
0 5 10 15 20 25 30 35 40measured mv [vol %]
estim
ated
mv
[vol
%]
0-4 elbe4-8 elbe0-4 weiherbach4-8 weiherbach
Dubois Oh
Quantitative Estimation of Volumetric Moisture mv [vol %]
0-4 cm 4-8 cm corr.
Elbe RMSerror 14 10 -/-
Weih RMSerror 11 17 0.2/0.4
0-4 cm 4-8 cm corr.
Elbe RMSerror 7 12 -/-
Weih RMSerror 18 25 0.5/0.6
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
16
Slide 62Slide 62
Dubois Oh
Quantitative Estimation of Surface Roughness ks
ks corr.
Elbe RMSerror 0.21 0.6
Weih RMSerror 0.19 -0.6
0
0.2
0.4
0.6
0.8
1
0.0 0.2 0.4 0.6 0.8 1.0measured ks
estim
ated
ks
e lbeweiherbach
0
0.2
0.4
0.6
0.8
1
0.0 0.2 0.4 0.6 0.8 1.0
measured ks
estim
ated
ks
e lbeweiherbach
ks corr.
Elbe RMSerror 0.24 -
Weih RMSerror 0.19 -0.7
Slide 63
Surface Parameter @ X-Bragg
Slide 63
Surface scattering model for the estimation ofsoil moisture content & surface roughness Back Scattering
I. Hajnsek, E. Pottier, S. Cloude, Surface Parameter Estimation using Polarimetric SAR, IEEE TGARS, 2003, vol. 41, no. 4, pp. 727-745
))4(sinc1(00
0))4(sinc1( )2(sinc*0)2(sinc
13
1312
121
CCC
CCT
23 2
1=C PS RR **2 PSPS RRRRC 21 PS RRC
0000
02*
*2
PSPSPS
PSpSPS
RRRRRR
RRRRRR
T
22
22
)sincos())sin1()(sin1(
rr
rrPR
2
2
sincos
sincos
r
rSR
surface
sensor
Bragg Surface X-Bragg Surfaceextension
Advantage:Simple model with direct relation to soil moisture content
Disadvantage:not realistic enough for natural surfaces
20 1
Advantage:Integration of a roughness & decorrelation term provides natural surface statistics
Disadvantage:very sensitive to vegetation cover
Slide 64Slide 64
Prediction of the X-Bragg Model I
(HH+VV)(HH-VV) coherence versus b1 for different local incidence angles
Cross polarised power versus b1 for different local incidence angles
Slide 65Slide 65
Variation of Anisotropy with Roughness (Model Parameter ß1 )
Variation of Entropy/Alpha values with Dielectric Constant (45 degrees AOI)
Prediction of the X-Bragg Model II
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
17
Slide 66Slide 66
Frequency:1.5to18.5 GHz10°20°
30°
40°
50°
Anisotropy versus frequency plot forthe EMSL data
Entropy/alpha angle plot for the EMSL data
Experimental Data from Anechoic Chamber (JRC - Ispra)Collected in European Mircowave Signature Laboratory (EMSL)
quad-pol scattering matrix(HH,VV,HV,VH); s = 0.4 cm; surface correlation lenght l = 6; dielectric constant ‘ = 8; frequency range = 1.5-18.5
Slide 67
Inversion Results @ X-Bragg – Early Example on ELBE 2000
Slide 67
Volumetric Moisture
N
0 500 m
Field 10
Field 13
Field 14
Field 16
0
10
20
30
40
0 10 20 30 40measured m v [vol %]
estim
ated
mv [v
ol. %
]
Elbe 0-4 cmElbe 4-8 cmWeiherbach 0-4 cmWeiherbach 4-8 cm
Volumetric Moisture mv [%]
0
30
Vol. [%]
Slide 68
Bare Surfaces: Soil Moisture Estimation @ AQUIFEREX 05
Slide 68
E-SAR / Test Site: Aquiferex, Tunisia
L-band / Pauli RGB Vol. Soil MoistureDielectric ConstantSlide 69Slide 69
Soil Moisture Estimation
• Surface Characterisation
• mv estimation over bare surface
• mv estimation under the vegetation
• 3 component model decomp.
• improvement of the 3 component model decomp.
• multi-angle approach combined with 3 comp. model decomp.
• hybrid decomposition
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
18
Slide 70
3 Component Freeman Decomposition
Slide 70
Surface Dihedral Volume
33
22
1211V
2
D2
S
T000TT0TT
100010002
4f
00001α0α|α|
f0000|β|β0β1
f12
Bragg Fresnel Random Volumeof Dipoles
0000|β|β0β1
fT 2*
SBraggVH
VH
RRRRβ
22rr
2r
2r
V)θsinεθcosε(
))θsin1(εθ)(sin1ε(R
θsinεθcosθsinεθcos
R2
r
2r
H
Slide 71
3 Component Freeman Decomposition
Slide 71
33
22
1211V
2
D2
S
T000TT0TT
100010002
4f
00001α0α|α|
f0000|β|β0β1
f12
Bragg Fresnel Random Volumeof Dipoles
iPTPS
STSS
PT
ST
PS
SS
eRR00RR
R00R
R00R
1001
]S[
θsinεθcosθsinεθcos
R2
T/S
2T/S
)T/S(S
θsinεθcosεθsinεθcosε
R2
T/ST/S
2T/ST/S
)T/S(P
Surface Dihedral Volume
Slide 72
AGRISAR Campaign in Northern Germany 2006
Slide 72
1. Agricultural data base over a whole vegetation growth period - April - August 2006
2. 16 data acquistions (DLR’s E-SAR) & ground measurements
3. Support by ESA for the space segment - Sentinel Program
Soil Moisture Station
Bowen-Ratio Station
Laser-Profiler
ASD
Thermal Infrared
LAS
DEMMIN
Slide 73Slide 73
dihedralvolumesurface
03/5/2006 14/6/2006 02/8/2006 03/5/2006 14/6/2006 02/8/2006
Freeman-Durden 3-Component Decomposition (AGRISAR 2006)
C-band L-band
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
19
Slide 74
3 Component Decomposition and Modifications
Slide 74
Total backscatter = + +
Surface Dihedral Volume
000010
LfT
2
2SDmoD
||
Dihedral modification (modified Fresnel coefficients) (εs, εt, θ, φ, LS):
Incorporation of a surface scattering loss LS
1L0 S
A1ks ))cos()(exp( 22S ks2L
[Lee et al., 2005]
))4(sin1(||2100
0))4(sin1(||21)2(sin
0)2(sin1
2
2
c
cc
c
fT SXB
Surface modification (X-Bragg) (εs, θ, ):
[Hajnsek et al., 2001]
sensor
Incorporation of a roughness and HV-decorrelation term (δ)
100010002
4000010||
0000||01
0000 2
2
33
2212
1211V
DSfff
TTTTT
000010||
0000||01
00000 2
2'22
'12
'12
'11
DS ffTTTT
Decision rule = or0
0SS VVHH
*Re
0
0SS VVHH
*Re
[Freeman & Durden, 1998, Yamaguchi et al., 2006]Slide 75Slide 75
dihedralvolumesurface
X-Bragg Volume 1 Volume 2
Modified Decompositions @ L-band for 03/5/2006
Volume 3Modified Dihedral
Slide 76Slide 76
Time Series of Decompositions on Field Basis @ L-band
Wheat (No. 250) – far range
Wheat (No. 230) – near range
Slide 77Slide 77
Time Series of Decompositions on Field Basis @ L-bandRape (No. 140) – near range
Rape (No. 101) – far range
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
20
Slide 78Slide 78
03/5/2006 14/6/2006 02/8/2006
Inverted Soil Moisture @ L-band Volume 2 (surface) and modified Dihedral (dihedral)
Land use
0
10
20
30
40
50vol. %
Slide 79Slide 79
Time Series of Inverted Soil Moisture @ L-band
Wheat field (No. 250) Wheat field (No. 230)
+ Bragg X-Bragg Volume 1 Volume 2 Volume 3*Surface Dihedral ±30% Intervall
Slide 80Slide 80
Time Series of Inverted Soil Moisture @ L-band
Rape field (No. 140) Rape field (No. 101)
+ Bragg X-Bragg Volume 1 Volume 2 Volume 3*Surface Dihedral ±30% Intervall
Slide 81Slide 81
Soil Moisture Estimation
• Surface Characterisation
• mv estimation over bare surface
• mv estimation under the vegetation
• 3 component model decomp.
• improvement of the 3 component model decomp.
• multi-angle approach combined with 3 comp. model decomp.
• hybrid decomposition
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
21
Slide 82
Polarimetric Decomposition Techniques & Inversion Scheme
Slide 82
No
Vegetated soil
Volume power
extraction
Model-based decomposition
H/criterion
Volume power
correction
Scatteringdominance
Eigenvaluedecomposition
Calculation of entropy and alpha
Soil moisture inversion from surface and dihedral components
Calculation ofground
components
Soil moistureinversion fromX-Bragg model
YesBare soil
Volume orientation
Total soil moisture
0 10 20 30 40 50vol.-%
Soil Moisture MapWeisseritz Region
Slide 82Slide 83
Modeling of Volume Orientation
2
2
log10HH
VV
rS
SP
)sin(21 pdf )cos(2
1 pdf
Slide 83
Volume modelling of a cloud of uniformly
shaped (ρ) particles with an orientation (τ)
in azimuthal direction
Selection of volume orientation by polarisation power ratio Pr:
General Case Vertical dipoles[Pr < -2dB]
Random dipoles[-2dB < Pr < 2dB]
Horizontal dipoles[Pr > 2dB]
21
pdf
8000750515
30V
VfT
100010002
4V
VfT
8000750515
30V
VfT
[Yamaguchi et al., 2005]
0 20 2/2/
300024041
CCCCC
fT VV
Slide 85
Vegetation GroundSAR data
- = +
Negative Powers =
Too Strong Volume Subtraction!
Eigenanalysis of GroundComponents
Set Eigenvalues to zero (lower calculus limit) Solutions for fv :
Slide 85
Correction of Volume Intensity fv with Eigen-based analysis
[van Zyl et al., 2008]
1 334vf T
2,3 11 22
2 * 211 12 12 11 22 22
2
8 4 4
vf T T
T T T T T T
Solution
Find Minimum Decomposiition with
Corrected Volume Intensity
fvcorr
11 12 1 4*
12 22 4 2
33 3
0 00 0
0 0 0 0v
T T C CT T f C C
T C
Surface DihedralT T
Slide 86
3 Component Decomposition and Modifications
Slide 86
))(sin(||
))(sin(||)(sin
)(sin
4c12100
04c1212c
02c1
fT
2
2SXB
Surface model (X-Bragg) (εs, θ, ):
Incorporation of a roughness and HV-decorrelation term
[Hajnsek et al., 2001]
sensor
Total backscatter = + +
Surface Dihedral Volume
300024041
000010||
))4(sin1(||5.0000))4(sin1(||5.0)2(sin0)2(sin1
0000
,
2
2
2
33
2212
1211
CCCCC
fLfc
ccc
fT
TTTT
corrvvDS
000010||
))4(sin1(||5.0000))4(sin1(||5.0)2(sin0)2(sin1
0000 2
2
2
'33
'22
'12
'12
'11
vDS Lfc
ccc
fT
TTTT
Scattering dominance = or0
0SS VVHH
*Re
0
0SS VVHH
*Re
[Freeman & Durden, 1998, Yamaguchi et al., 2006]
000010
LfT
2
VDmoD ||
Dihedral model (Fresnel with vegetation attenuation) (εs, εt, θ, φ, LV):
Incorporation of a vegetation attenuation LV
V
2
V
1
P
P
min
max
))/(exp( minmax 1LV
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
22
Slide 87Slide 87
1. Soil moisture estimation under different land cover with measurement approaches at different scales
2. Goal: Support flood forecasting by identification of critical catchment states
2 Radar data acquisition flights (single pol X-, dual-pol C- and quad-pol L- and P-band)Ground measurements (soil moisture,soil roughness, biomass,…)
OPAQUE – Campaign May 2007
TDR and GPR
t
NCorner–reflectorWeißeritzFlight stripTest site
Winter triticaleWinter wheatSummer barleyWinter barley
Grass landWinter rapeSummer cornBare soil
Slide 88
Volume Orientation and Scattering Dominance @ L-band
Slide 88
2% 11%
12%
31%18%
5%
21% surface dominant
vertical
horizontal
random
volume orientation
verticalhorizontalrandom
volume orientation
X-Bragg only
dihedral dominant
rang
e
azimuth
Slide 89
Slide 89
L-Band Local incidence
Model-based Decomposition @ L-band
Land use
0°
18°
36°
54°
72°
90°degrees
Grass landWinter rapeSummer cornBare soil
Winter triticaleWinter wheatSummer barleyWinter barley
range
azim
uth
Slide 90
Inverted Soil Moisture @ L-band
Slide 90
+ =+
Dihedral TotalSurfaceX-Bragg
0 10 20 30 40 50vol.-%
range
azim
uth
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
23
Slide 91Slide 91
Validation with TDR Measurements on Winter Wheat
Vegetationheight:
55cm
Row distance:
10cm
Wet biomass:
2,85kg/m²
Slide 92Slide 92
Validation with TDR Measurements on Summer Barley
Vegetationheight:
45cm
Row distance:
23cm
Wet biomass:
0,93kg/m²
Slide 93Slide 93
Validation with TDR Measurements on Winter Triticale
Vegetationheight:
85cm
Row distance:
10cm
Wet biomass:
3,34kg/m²
Slide 94Slide 94
Validation with TDR Measurements on Winter Barley
Vegetationheight:
70cm
Row distance:
10cm
Wet biomass:
3,31kg/m²
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
24
Slide 95Slide 95
Soil Moisture Estimation
• Surface Characterisation
• mv estimation over bare surface
• mv estimation under the vegetation
• 3 component model decomp.
• improvement of the 3 component model decomp.
• multi-angle approach combined with 3 comp. model decomp.
• hybrid decomposition
Slide 96
Slide 96
Multi-Angle Approach for Soil Moisture Inversion Rate Increase
vol.%0
1020
3040
50
Single angular1 acquisition
master
Bi-angular2 acquisitions
master -perpendicular
Tri-angular3 acquisitions
master -opposite -
perpendicular
Flight Lines Master Opposite Perpendicular
0°
18°
36°
54°
72°
90°
Testsite: Weisseritz Watershed; E-SAR L-band
Slide 97Slide 97
Inversion Rate of Soil Moisture Estimation @ different AOI
One acquisition
Single angular
master opposite perpendicular
40.32% 29.87% 48.53%
Two acquisitions
Bi-angular
master-opposite
master-perpendicular
opposite-perpendicular
55.87% 63.39% 60.71%
Three acquisitions
Tri-angular
master-opposite-perpendicular
70.89%
Slide 98Slide 98
Field meanvalues
Validation of Soil Moistures @ different Incidence Angles
Sampling box: 13x13 pixelsBox coverage: 70%
Single angularinversion (p)
Bi-angularinversion (o-p)
Tri-angularinversion (m-o-p)
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
25
Slide 99Slide 99
Single angular
RMSE [vol.%] between Estimated and Measured Soil Moistures for Different Incidence Angle Constellations
Bi-angular Tri-angular- = out of scene < = too less values for a valid analysis
Fields m o p m-o m-p o-p m-o-p
Winter Triticale 6.34 6.34 10.33 5.42 6.27 9.46 5.72
Winter Barley < < 9.75 7.76 6.85 8.78 6.94
Winter Rye 5.25 6.11 5.86 5.81 4.89 6.42 5.27
Winter Wheat 4.52 < 9.79 4.41 5.04 9.54 4.98
Summer Oat 8.55 6.43 - 6.35 8.55 6.43 6.35
Mean 6.17 6.29 8.93 5.95 6.32 8.13 5.85
Slide 100Slide 100
Slide 100
0
1
2
3
4
5
6
7
8
9
10
11
master opposite perpendicular m-o m-p o-p m-o-p
vol.%
RMSE @ Different Incidence Angle Constellations
Summer oatWinter triticaleWinter barleyWinter ryeWinter wheatNot enough valid pixelsOut of scene
RM
SE
[
]
RMSE=root mean squareerrorm=mastero=opposite p=perpendicular
Single angular Bi-angular Tri-angular
Slide 101Slide 101
Soil Moisture Estimation
• Surface Characterisation
• mv estimation over bare surface
• mv estimation under the vegetation
• 3 component model decomp.
• improvement of the 3 component model decomp.
• multi-angle approach combined with 3 comp. model decomp.
• hybrid decomposition
Slide 102
Removal of Vegetation Component
Basic Principle of Hybrid Polarimetric Decomposition
102
Not physically constraint volume intensity!
Volume intensity ( fv)2
33 sin/2 vv tf
0000sincossincos0sincossincos
sin000sin000cos2
200
00
22,,
,,22
2
2
2
33
22*12
1211
ssddsdsdsd
sdsdsdddss
v
v
vv ffff
fffff
ttttt
signature groundSurface Dihedral VolumePolarimetric Signature
= + +
vegetation groundsignature
- fv = +
Model-based decomposition for vegetation volume removal
Eigen-based Decompositionof ground for soil moisure
vegetation
+Hybrid Polarimetric Decomposition
Physically constraint volume intensity ( fvc) with surface scattering (αb)
2 * * 2 2vc 11 22 12 12 12 124 csc(2 ) ( ( )cos(2 )) sin(2 ) /1 3cos(2 )b b b vf t t t t t t
Graphics from www.vectorarts.net
Slide 102
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
26
Slide 103
Retrieval of the Ground Scattering Components
103
Scattering mechanisms of ground (d, s)Physically Meaningful Separation of Scattering Mechanisms (d,s)
Orthogonality conditiond + s= /2
[0, /4] Surface scattering
[/4, /2] Dihedral scattering
s
d
Intensity of ground (fd, fs)
From eigenvalues:
From eigenvectors:
groundvegetation
- fvc = +
0000sincossincos0sincossincos
22,,
,,22
ssddsdsdsd
sdsdsdddss
ffffffff
Model-based decomposition for vegetation volume removal
Eigen-based decompositionof ground for soil moisure Hybrid Polarimetric Decomposition
103
Eigen-based Decomposition of Ground Components
Graphics from www.vectorarts.net
Slide 103Slide 104
Soil Moisture Inversion from Surface Scattering Component
mins
m
Minimization
s
Pedo-Transfer Function of
Topp et al. (S40)
Soil moisture [vol.%]
Surface scattering componentfrom hybrid polarimetric
decomposition
Polarimetric SAR data
Bragg scatter modeling with locand a variety of soil dielectric constants S
Surface scattering model
tan( )s HH
HH
RR
VVm
VV
RR
RHH , RVV = f(S ,loc)
Slide 104
Slide 105
Soil Moisture Inversion along Vegetation Growth Cycle range
azim
uth
Land use April, 19thεs-level=20
25cm
5cm
Soil moistureMax. vegetation height
Soil moisture station
Height:18cmWet biomass:1.88kg/m²
5
12
19
26
33
40
[vol.%]
June, 7th
Height:167cmWet biomass:9.60kg/m²
5
9
13
17
21
25
[vol.%]
εs-level=10
0
4
8
12
16
20
[vol.%]
July, 5th
Height:172cmWet biomass:6.26kg/m²
εs-level=8
Winter rape
AgriSAR 2006L-band
Quad-pol
Slide 106Slide 106
APRIL JUNE
JULY
RMSE=6.75 RMSE=6.98
RMSE=4.43 validation box: 13x13 pixels
Validation of Moisture Inversion with Field Measurements (TDR)
RMSE [vol.%] APRIL JUNE JULYsugar beet 7.55 11.90 4.68
winter wheat 8.06 6.15 3.40winter rape 5.70 2.62 2.06
summer corn 5.34 3.20 5.25
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
27
Slide 107
Soil moisture inversion is still object of scientific research:
Main research focuses on the estimation of mv under the vegetationOne attempt is to use SAR polarimetryfor scattering mechanism decomposition in order to characterise and subtract volume from a ground component (surface/dihedral)PolSARPro provides the possibility to invert soil moisture over bare fields and allows a decomposition of scattering mechanisms
Outlook: Inversion procedures for vegetated areas are still missing
Slide 107
Summary Soil Moisture Estimation
Slide 108
Basic Text Books for SAR-Polarimetry and Pol-InSAR:Cloude, S., Polarisation: application in remote sensing, Oxford University Press, p. 352, 2009.Boerner, W. M. et al., ‘Polarimetry in Radar Remote Sensing: Basic and Applied Concepts’, Chapter 5 in F. M. Henderson, and A.J. Lewis, (ed.), “Principles and Applications of Imaging Radar”, vol. 2 of Manual of Remote Sensing, (ed. R. A. Reyerson), Third Edition, John Willey & Sons, New York, 1998. Elachi, C. & Van Zyl, J. J., ‘Introduction to the physics and techniques of Remote Sensing’, John Wiley and Sons, p. 552, 2006.Lee, J.S. & Pottier, E. ‘Polarimetric Radar Imaging: From Basics to Aapplications’, CRC Press, 2009.Woodhouse, I., ’Introduction to Microwaves Remote Sensing’, CRC Taylor & Francis, p. 370, 2005.
Slide 108
References (Part I)
Slide 109
Surface Parameter Inversion:Hajnsek, I., Pottier, E. & Cloude, S.R., ‘Inversion of Surface Parameters from Polarimetric SAR’, IEEE Transactions on Geoscience and Remote Sensing, vol. 41, no. 4, pp. 727-745, 2003.Lee, J.-S., Boerner, W.-M., Ainsworth, T. L., Hajnsek, I., Papathanassiou, K.P., Lüneburg, E., ‘A Review of Polarimetric SAR Algorithms and their applications’, Journal of Photogrammetry and Remote Sensing, vol. 9, No. 3, pp. 31-80, 2004.Hajnsek, I. & Cloude, S., ‘The Potential of InSAR for Quantitative Surface Parameter Estimation’ Can. J. Remote Sensing, vol. 31, no.1, pp. 85-102, 2005Hajnsek, Irena; Jagdhuber, Thomas; Schön, Helmut; Papathanassiou, Kostas (2009): Potential of Estimating Soil Moisture under Vegetation Cover by means of PolSAR. IEEE [Hrsg.]: IEEE Transactions on Geoscience and Remote Sensing, IEEE, vol 47, no 2, pp. 442 – 454.M. Arii, J.J. van Zyl and Y. Kim: A general characterization for polarimetric scattering from vegetation canopies, IEEE Transactions on Geoscience and Remote Sensing, vol. 48, pp. 3349-3357, 2010. M. Neumann and L. Ferro-Famil: Extraction of particle and orientation distribution characteristics from polarimetric SAR data, in Proc. of 8th European Conference on Synthetic Aperture Radar, June 7-10, 2010, Aachen, Germany, VDE, pp. 422-425.Jagdhuber, T., I. Hajnsek, A. Bronstert & K.P. Papathanassiou, ‘Soil Moisture Estimation under Low Vegetation Cover Using a Multi-Angular Polarimetric Decomposition‘, IEEE Transactions on Geoscience and Remote Sensing,10.1108/TGRS.2012.2209433, p. 1-15, 2012.Jagdhuber, T. , I. Hajnsek, K.P. Papathanassiou, & A. Bronstert, ‘Soil Moisture Retrieval Under Agricultural Vegetation Using Fully Polarimetric SAR‘, Proc. of IEEE International Geoscience and Remote Sensing Symposium, Munich, Germany, July 22-27, pp. 1481-1484 , 2012.Jagdhuber, T., ‘Soil Parameter Retrieval under Vegetation Cover Using SAR Polarimetry, PhD Thesis, University Potsdam, Potsdam, supervised by Prof. Hajnsek and Prof. Bronstert, Date of Defense: 05.07.2012, http://nbn-resolving.de/urn:nbn:de:kobv:517-opus-60519, 2012.Bronstert, A., C. Creutzfeldt, T. Gräff, I. Hajnsek, M. Heistermann, S. Itzerott, T. Jagdhuber, D. Kneis, E. Lück & D. Reusser, ‘Potentials and constraints of different type of soil moisture observations for flood simulations in headwater catchments‘, Natural Hazards, Vol. 60, pp.879–914 , 2012.
Slide 109
References (Part II)
Slide 110Slide 110
Acknowledgement to the AgriSAR Team
06/06/06
INTA
-
2nd Advanced Course on Radar Polarimetry 2013 01/24/2013
28
Slide 111
Acknowledgement to the OPAQUE Team
Slide 111
30/05/2007
Slide 112
Part II: Exercises with L-band Airborne Data
Read the airborne SAR dataSpeckle Filtering (refined Lee)Oh, Dubois and X-Bragg Inversion
Slide 112
HH HV
VH VV
Slide 113Slide 113
Test Date Used for the Exercise
Testsite: DemminLocation: Northern GermanyAcquisition Date: May 2012Frequency: L-bandData size: az: 2.75km rg: 2.2 kmPolarisation: 4 SLCResolution: az: 60cm x rg: 3.8mRows and colums: 7981 x 1837
dihedralvolumesurface
Pauli RGB
Slide 114Slide 114
First Steps in PolSARPro
Please open PolSARProDefine your environmentOpen the DLR‘s acquired test Radar data