Microwaves and Radar InstituteMicrowaves and Radar Institute
Potential & Challenges of Pol-InSAR Techniques for Forest Height Estimation in the context of the BIOMASS Mission
S-K. Lee, S. Sauer, F. Kugler, R. Scheiber, I. Hajnsek & K. Papathanassiou
German Aerospace Center (DLR)
Microwaves and Radar Institute (DLR-HR)
VU 2 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
26 Plots / 27m x 32m / la= 0.6
58.1a H66.1lB ⋅=
Motivation
Sungai WainLowland Dipterocarp Forest
Mawas 1Riverine Peat Swamp Forest
Mawas 2Tropical Peat Swamp Forest
20 Plots / 10m x 10m / la= 0.5 20 Plots / 10m x 10m / la= 0.6
Insufficient Number of samples !!! - But clear trend visible !!!
VU 3 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
21 SS><><
><=
∗∗
∗
2211
2121
SSSS
SS)SS(γ~Interferometric
Coherence
VolumeCoherence
∫
∫=
v
v
z
oz
h
o
h
o
zik
zikVol
dz)z(f
dze)z(f
e))z(f(γ~
)z(f
)z(f … vertical reflectivity function
Vol
ume
Coh
eren
ce
00V
~
II
γ =
∫
∫
=
=
V
V
h
0 00
h
0 0z
'dzθcos
'zσ2expI
'dzθcos
'zσ2exp)'zκiexp(I
0V
G
I)w(m
)w(m)w(m
=
Vertical Wavenumber:
G/V Ratio:
)θsin(
θΔκκ
0z =
Volume Height
Extinction
Topography
G/V Ratio
Vh
σ
0φ
)w(m
)w(m1
)w(mγ~)φiexp()w(γ 0V
0
~
++=)zz(δm
θcoszσ2
expσ)z(f 0G0
0V −′+
=
2 Layer Scattering Model
VU 4 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
Phase Center Location as a Function of Polarisation / Ground
Var
ian
ce o
f th
e P
ha
se C
ente
r L
oca
tio
n d
ue
to
Vo
lum
e, S
yste
m a
nd
Tem
po
ral
Dec
orr
elat
ion
The Effect of the Ground-2-Volume Component
Ground and Volume No Ground
Only VolumeVolume only
)κ|σ,h(γ~)κ|(γ~)φiexp(min zV0VzVol0σ,hV
−−
)φiexp()m(γ~ 0Vol =∞=
0V0Vol γ~)φiexp()0m(γ~ ==
VU 5 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
Campaigns Summary
Lidar4 - 6m / .5 -.730º-45ºFlatMultiple >2 15minManaged
Semi BorealBIOSAR-I
Lidar4.5m / .6530º-50ºHillyMultiple >215min BorealBIOSAR-II
Lidar1.9m / .9530°-45°Locally Slopes
Multiple >215 minTemperateTempoSAR
Stand< 2m30°-40°Flat115 minManaged
TemperateNezer Forest
Lidar2.m / .9435º-40ºFlat320minPeat Swamp INDREX-II
Stands2-3m40ºHilly220min(Lowland)
Dipterocarp INDREX-II
30-40º
IncidenceAngle
2,5m(Preliminary)
ResultsRMS / ρ2
15 min + …
Temporal Baseline
StandHilly1TropicalTropiSAR
TerrainForestCampaign Reference
SpatialBaselines
VU 6 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
Ground-2-Volume Ratios Range
INDREX-II: 1402; Mawas
INDREX-II: 0601; Sungai Wain
mmax
mmin
VU 7 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
Ground-2-Volume Ratios Range
BioSAR 2007: 0401; Remningstorp
BioSAR 2008: 0103; Krycklan
mmax
mmin
VU 8 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
P-band
P-band
Model Miss-Match of Vertical Structure Function (f(z))
)z(f
)z(f
50
40
30
20
10
0m
MAWAS
Krycklan
Forest Height
Forest Height
LIDAR H100 (m)
LIDAR H100 (m)
VU 9 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
σ > 0m3= 0
σ > 0m3≠ 0
σ >< 0m3= 0
σ >< 0m3≠ 0
Effect of Extinction and Ground-to-Volume Ratio
VU 10 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
Campaigns Summary Temporal Deco
1-7d
2 / 4 / 7 / 14 / 20 / 22d
1 - 15d
1 / 30 / 56d
20min
Temporal Baseline
0.94 / 0.85 / 0.90 / 0.80 / 0.92 / 0.90TropicalTropiSAR
0.95 - 0.95Managed
TemperateTempoSAR
Tropical
Managed Semi Boreal
Peat Swamp
Forest
La Selva
BIOSAR-I
INDREX-II
Campaign
Lidar
Lidar
Comments
0.6-0.8
0.95 / 0.9 / 0.85
0.93
Temporal Decorrelation
10m20m30m
)w(m1
)w(mγ~γ~γ~)φiexp()w(γ TGVTV
0
~
++=)zz(δmeσ)z(f 0G
θcoszσ2
0V0 −′+=
2 Layer Model with Temporal Decorrelation
Extinction
Vol
ume
Coh
eren
ce
0V I
Iγ =~
∫
∫
=
=
V
V
h
0 00
h
0 0z
'dzθcos
'zσ2expI
'dzθcos
'zσ2exp)'zκiexp(I
Volume Height
Topography
G/V Ratio
Vh
0φ)w(m
σ0V
G
I)w(m
)w(m)w(m
=
)θsin(
θΔκκ
0z =
TG
TV
γ~γ~ Temp. Deco. Volume
Temp. Deco. Ground
VU 11 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
Campaigns Summary Temporal Deco
1-7d
2 / 4 / 7 / 14 / 20 / 22d
1 - 15d
1 / 30 / 56d
20min
Temporal Baseline
0.94 / 0.85 / 0.90 / 0.80 / 0.92 / 0.90TropicalTropiSAR
0.95 - 0.95Managed
TemperateTempoSAR
Tropical
Managed Semi Boreal
Peat Swamp
Forest
La Selva
BIOSAR-I
INDREX-II
Campaign
Lidar
Lidar
Comments
0.6-0.8
0.95 / 0.9 / 0.85
0.93
Temporal Decorrelation
10m20m30m
)φiexp(γ~lim)m(γ~ 0Volm
Vol ==∞=∞→
0VDeco0Vol γ~γ)φiexp()0m(γ~ ⋅⋅==
1|)m(γ~| Vol =∞=with
)w(m1
)w(m)σ,h,κ(γ~γ)φiexp()γ,w,κ(γ~ Vz0VTV
0TVz
++⋅=
Volume Coherence & Decorrelation
VU 12 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
\
From Air 2 Space-Borne
Incidence Angle: 25º-60º 20º-30º
Spat. Baselines: - Multiple Multiple
- Variable (0. < Kz < 0.2) Only limited variable
Tmp. Baselines: Minutes – Hours 25-45 Days !!!
System Bandwidth: ~100MHz 6MHz
Spatial Resolution: 2.0-3.0m Ground Range 50m Ground Range
1.5-3.0m Azimuth 25m Azimuth
Ionosphere: No Yes
Atmosphere: No Yes
VU 13 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
Decorrelation Modeling
)φiexp()m(γ~ 0Vol =∞= )]φφ(iexp[)m(γ~)φiexp( D0VolD +=∞=
0VzDecoD0VolzDecoD γ~)κ(γ)]φφ(iexp[)0m(γ~)κ(γ)φiexp( +==
)κ(γ)]φφ(iexp[)m(γ~)κ(γ)φiexp( zDecoD0VolzDecoD +=∞=
0V0Vol γ~)φiexp()0m(γ~ == 0VD0VolD γ~)]φφ(iexp[)0m(γ~)φiexp( +==
|)κ(γ~|)φiexp()κ(γ~ zDecoDzDeco =
)φiexp()κ(γ~ DzDeco =
Phase Screens
Phase & Amplitude Deco
)κ|w(γ~)κ(γ~:)κ|w(γ~ zVolzDecoz
=
)κ|w(γ~)κ(γ~:)κ|w(γ~ zVolzDecoz
=
VU 14 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
Deterioration of Resolution (P-band BIOSAR)
Original airborne resolution(2m x 5m Multilook)
Degraded resolution(50m x 50m Multilook)
VU 15 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
BIOSAR-I: Interferometric Coherence History
100MHz 6MHz
HH Channel
6MHz + NESZ + Ambiguities T-deco = 0.9
Co
he
ren
ce
6MH
z
Coherence 100MHz
VU 16 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
BIOSAR-I: Interferometric Coherence History
100MHz 6MHz
HH Channel
6MHz + NESZ + Ambiguities T-deco = 0.9
VU 17 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
Mawas Test Site / INDREX-II
HV Amplitude 100MHz 6MHz
HV Amplitude 100MHz 6MHz
HV Amplitude 100MHz 6MHz
Remningstorp Test Site
Kryckland Test Site
Mawas Test Site
VU 18 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
BIOSAR-I: Single & Dual Baseline Inversion of Forest Height
Single baseline inversion; Original data (100MHz)i
Single baseline inversion; Spaceborne scenario (6MHz)i
Single baseline inversion; Spaceborne scenario + T-deco
Dual baseline inversion; Spaceborne scenario + T-deco
VU 19 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
P-band Pol-InSAR Conclusions
PolInSAR measures forest top height (H100) without the need of any a priory knowledge:
A quantity compatible with forestry standards;
A physical (not radar) quantity therefore independent of incident angle / environmental effects.
Demonstration / Validation at P-band over many sites across all major forest ecosystems:
Tropical, Boreal, Temperate;
Flat / Sloped (moderate ►steep) terrain;
Height ranges 5-60m / Biomass levels: 100-500T/Ha !!!
Pol-InSAR forest height inversion at P-band provides stable forest height maps.
Overall performance: RMS: 1.9 to 6m (10-30%) p2: 0.9-0.5
Larger variation of vertical scatterer distribution(s) across forest types.
Lower vegetation attenuation and larger / wider range of Ground-to-Volume Scat. Ratios;
Decreased sensitivity of P-band to the vegetation layer;
The performance depends critically on the selection of the (spatial) baseline. For reaching the inversion performance across a range of forest heights (5 ►50m) multiple baselines are required
VU 20 > Autor Name
Microwaves and Radar Institute > 30.05.2006Microwaves and Radar InstituteMicrowaves and Radar Institute
P-band Pol-InSAR Conclusions
The performance in space will be affected primarily by ► temporal decorrelation and ► the limited ability to provide variable (and large) spatial baselines.
Inversion performance analysis predict a overal performance between 20-100%. ► 20-40% with an optimised acquisition scenario across a height range from 10 to 60-70m on a 100x100m raster.
Critical areas: Positive terrain slopes and heterogeneous forest situations.
Bio & Geo-Physical Parameter Retrieval Algorithm Definition for Active Remote Sensing at P- & L-bandFinal Report ESA Contract 21131/07/NL/LvH, 2010.
Microwaves and Radar InstituteMicrowaves and Radar Institute
Potential & Challenges of Pol-InSAR Techniques for Forest Height Estimation in the context of the BIOMASS Mission
S-K. Lee, S. Sauer, F. Kugler, R. Scheiber, I. Hajnsek & K. Papathanassiou
German Aerospace Center (DLR)
Microwaves and Radar Institute (DLR-HR)