INTRODUCTION TO SAR

REMOTE SENSING

Thuy Le Toan

10 May 2021
ESA UNCLASSIFIED – For ESA Official Use Only 1
 Introduction to SAR remote sensing

Thuy Le Toan
Centre d’Etudes Spatiales de la Biosphère (CESBIO)
Toulouse, France
[email protected]

Thuy Le Toan, Radar Polarimetry Courses 2021 2

 Contents

1. Introduction to SAR remote sensing

2. Statistical properties of SAR images

3. Physical content of SAR data

4. Application to agriculture

4. Application to forests –Biomass estimation

3
4
The electromagnetic spectrum

Synthetic Aperture Radar (SAR)

5
 The electromagnetic spectrum

Visible (VIS) + Near Infrared (NIR)= Optical

Thermal Infrared (TIR)

Synthetic Aperture Radar (SAR)

6
SAR: all-weather observing system
Forest loss in Cameroon
Sentinel-1 Sentinel-2

7
 Marginal atmospheric effects

Strong attenuation

Low attenuation

Very low attenuation by atmospheric constituents for l > ~5 cm

8
 Principle of imaging radar

Active system:
day and night operations

9
Side Looking Radar

Azimuth direction

Range direction Linear displacement

of the antenna along
the track
Pulses

The range information comes from the time

needed by the pulse to travel way and back

10
Why side looking ?

11
 Antenna scattering

l Wavelength
=
L
L

 Antenna length
Angular aperture (horizontal
(horizontal plane) direction)

The larger the antenna, the narrower the

aperture (finer resolution)
L'
Resolution = lR
L
Numerical example:
L  10m, R  1000 km (spaceborne radar), l  5 cm (C
band) ➔ resolution  5 km
12
 Synthetic aperture technique

An array of antennas is equivalent to a single antenna moving along

the flight line LS if the received signals are coherently recorded and
added, and the target assumed to be static during the period

The echoes from X1, X2, ..Xn

LS are recorded coherently (amplitude
X1 X2 Xn and phase as a function of time)
Azimuth resolution
R
Ra= l R
LS

LS P
13
The SAR image

Look angle
Off nadir
Azimuth

Ground range

14
What does the SAR measure ?

1. Amplitude which depends on the target properties (dielectric and geometric properties)

2. Phase which is function of the sensor-target distance and target properties

15
 SAR measurement modes-BIOMASS

PolSAR PoI-InSAR TomoSAR

z z
z
x x x

o o o

y y y

16
Polarimetric Interferometry (PolinSAR))

Tree height inversion

Tree height

Phase Centre height (m)

HV

VV
HH

Ground level
Garestier, 2006

17
 Accurate range measurement

B Radar Interferometry
Relief Terrain displacement
i

Etna Landers
iso-altitude curves iso-displacement curves
Digital elevation models Cartography of terrain displacements 18
 Radar Interferometry

ERS Interferometry
Mesa, USA/ Mexico border

• Impact of a geothermal plant on

the environment. Interferogram
processed from two ERS images,
acquired at two years interval. The
fringes characterize the ground
subsidence around the plant. One
observe a subsidence of about 6 cm
(2 fringes) which covers 17km x 8km.

ERS intensity image Interferogramme 19

20
Principal available SAR data

Satellite Life- Frequen- Polarisation Resolution Frame Repeat Access

time cy cycle
TerraSAR-X 2007- X-band Single HH,VV Spotlight: <1m,1.7x3.5m Spotlight:3x10km 11days Restrained
TanDEM-X 2010- l= 3.5cm Dual: HH/VV,HH/HV Stripmap: 3x3m Stripmap:50x30km Scientific,
VV/VH ScanSAR:18x40m ScanSAR:150x100, Commercial
200x200km
COSMO- 2007- X-band Single: HH, VV, Spotlight < 1m Spotlight:10x10km Sat: 16 d Commercial
Skymed l= 3.5cm HV,VH Stripmap: 3x15m Stripmap:40x40km Constellatio Limited
Dual ScanSAR:30x100m ScanSAR:100x100, n: hrs scientific.
200x200km
Radarsat-2 2007- C-band Single: HH, VV, Spotlight : 1.5m Spotlight:18x8km 24 days Commercial
l= 5.6cm HV,VH Stripmap: 3x3m Stripmap:
Dual: HH/HV,VV/VH 25x25m 20x170km
Quad ScanSAR:35x35m ScanSAR:300x300,
100x100 m 500x500km
ALOS-2 2014- L-band Single: HH, VV, Spotlight : 1x3m Spotlight:25x25km 14days Commercial
PALSAR-2 l= 24.6cm HV,VH Stripmap: 3x10m Stripmap:50x70km Limited
Dual: HH/HV,VV/VH ScanSAR:25x100m 70x70km proposal-based
Quad ScanSAR: Scientific.
355x355km
Sentinel-1 2014- C-band Single: HH, VV Stripmap: 5x5m Stripmap: 375km 12days Free &
l= 5.6cm Dual: HH/HV,VV/VH Interferometric Wide IW:250 km Open Access
Swath (IW): EW:400 km S1 &S2:
5x20m 6 days
Extra Wide Swath (EW):
20x40m 20
Principal SAR missions

P-Band
Biomass

21
22
Radar frequency bands
Frequency band Wavelength (cm) Frequency (GHz)
Ka 0.8-1.1 40 - 26.5
K 1.1-1.7 26.5 - 18
Ku 1.7-2.4 18 - 12.5
X 2.4-3.8 12.5 -8
C 3.8-7.5 8 -4
S 7.5-15 4 -2
L 15 -30 2 -1
P 30 -100 1 - 0.3

f (in Hertz)=C/l

C=3.108 m
l =wavelength in m 22
 Polarisation

Electric field Trajectory of the

Magnetic field
electric field

Propagation direction

Transverse plane

23
 Polarisation

Plane orthogonal to
Electric field the propagation direction

Propagation direction

Horizontal polarisation Vertical polarisation

24
Polarisation

25
26
The radar scattering

Backscattered Incident electric field

electric field Es Ei

the amplitude, phase and

polarisation of Es are modified
with respect to Ei

26

Thuy Le Toan, CESBIO, Toulouse, France

 Basic measurement

The basic measurement made by a SAR is called S

(amplitude and phase). This is the complex image.

Main types of images:

A is the amplitude image.

I = A2 is the intensity image.
(the phase of a single image is not exploitable)

27
 The radar cross section

The radar cross-section (RCS) is defined as

 pq = 4 S pq
2
= 4R2Ps
Pi
m 
2

R is the radar-target distance

Pi is the incident power,
Ps is the power scattered by the target.

28
 The backscattering coefficient

For distributed targets each resolution cell

contains many scatterers and the phase varies
rapidly with position.

The differential backscattering coefficient, o,

is
4R 2 Ps
 o
=
[m2/m2]
A Pi

where A is the area of the illuminated surface

over which the phase can be considered constant.

29
 Slant and ground range

Slant range

Ground range

Source: CCRS tutorials

30
 Contents

1. Introduction to SAR remote sensing

2. Statistical properties of SAR images

3. Physical content of SAR data

4. Application to agriculture

4. Application to forests –Biomass estimation

31
 What is a SAR image?

The image is seen as a

picture.

Pixels are numbers.

Image is affected by
speckle noise

Example of an intensity image

APP HH image 400 x 400 pixels (of 12.5m)
Gaoyou, Jiangsu province, China, 2004 05 24
32

Thuy Le Toan, CESBIO, Toulouse, France

 Same image, after speckle filtering

33

Thuy Le Toan, CESBIO, Toulouse, France

Initial HH and VV images HH and VV image after filtering

HH (magenta) and VV (green) images

400 x400 pixels
Gaoyou, Jiangsu province, 2004 09 06
34

Thuy Le Toan, CESBIO, Toulouse, France

 Origin of speckle noise


pixel n°1 pixel n°2

1 2 1 2

Image
radiometry

3 4 3 4
pixel n°3 pixel n°4

m m
pixel n°1 pixel n°4

e e

Pixel response
Elementary contribution

35
 Speckle noise must be reduced…

Before speckle filtering After speckle filtering

36
…to start analysing SAR images content

37
 Statistics of speckle

Probability density distribution of speckle:

2
- Intensity image: Gamma distribution var(I ) =
- Amplitude image: Rayleigh distribution L
20

18  0=-14dB
Probability Density Function of Intensity I

 0=-7dB
16

14

12

10

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 38
Intensity I
 The gamma distribution

Multilooking reduces the effect of speckle. 2

var (I ) =
The distribution tends to normality as L increases. L
90
L=1
80 L=2
Probability Density Function of Intensity I

L=4
70 L=8
L=16
60 L=32
L=64
50

40

30

20

10

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Intensity I
39
 Speckle filtering reduces variance and preserves radiometry

Non filtered – 4 looks

ENL=20
γoHV Mean = -14.70 dB
SD = 1.58 dB
-8 dB

ENL=4
Mean = -14.70 dB
SD = 2.47 dB

ENL=20
Filtered – 20 looks Mean = -11.89 dB
SD = 1.21 dB
Blue area

ENL=4
Mean = -11.89 dB
SD = 2.31 dB

-20 dB

Yellow area Mermoz et al., 2016

40
 Speckle filtering

I=σ.v I =
I: measured intensity
2
σ : scene reflectivity (what we want to measure) var (I ) =
v: speckle noise L
L: number of looks

Speckle filtering consists in assessing σ from I by reducing the variance of I (and

therefore increasing the Equivalent Number of Looks).

1. Frequency filtering: spectral filtering during SAR processing

(production of multi-look images)

2. Spatial filtering: local estimations using moving kernels

Filters of Lee, Kuan, Frost, MAP widely available

3. Multi channel filtering: applied on multiple images of the same scene

Multi polarisation, multi temporal, and multi frequency
41
Multi-image intensity filtering (temporal and polarisation)

INPUT OUTPUT

I1 J1
I2 J2
Filter

IM JM
Purpose of filter:
(1) Preserve radiometry  unbiased
I k ( x, y ) = J k ( x, y ) 1  k  M
(2) Minimise the variance of J k
42
 Contents

1. Introduction to SAR remote sensing

2. Statistical properties of SAR images

3. Physical content of SAR data

4. Application to agriculture

4. Application to forests –Biomass estimation

43
44
Scattering Mechanisms
◼ The backscattered signal results from:
- surface scattering
- volume scattering
- multiple volume-surface scattering

◼ The relative importance of these contributions depends on

geometric and dielectric properties of the medium ( surface
and volume)

◼ All of these factors depend on

- the radar frequency
- the polarisation
- the incidence angle
44

Thuy Le Toan, CESBIO, Toulouse, France

45
Scattering mechanisms

water Surface scattering Surface scattering soil,

rock

Volume scattering if
penetration
vegetation snow
Volume scattering
Volume scattering
Volume-surface
scattering
Surface scattering
Surface scattering
45
Interactions between radar wave and vegetation cover
Geometric, structural properties
scatterers size scatterers orientation
v
h

k
scatterers density soil roughness

Dielectric properties
scatterers water content soil moisture

46
47
Surface scattering

Smooth surface Rough surface

er1 er1
The roughness of the surface (wrt to the wavelength) governs the
scattering pattern

er2 > er1 medium 2 is wetter than medium 1

Wetter media er2 er2

The dielectric constant (moisture content) of the medium governs
the strength of the backscatter
47

Thuy Le Toan, CESBIO, Toulouse, France

 Contents

1. Introduction to SAR remote sensing

2. Statistical properties of SAR images

3. Physical content of SAR data

4. Application to agriculture

5. Application to forests –Biomass estimation

48
 Scattering from soil surface
Radar backscatter depends on the dielectric and geometry of the target, and
on the frequency, polarisation and incidence angle of the wave
e.g. Small Perturbation Method for surface scattering
0
pp
2 2 4

 = 4(ks ) (kl ) cos   pp 1+ 2(kl sin  )
2

2 −3 2

 hv
0
( ) = 0
Here
.
k = 2 /l (Wave number)
pp = HH or VV
 = incidence angle
s = surface RMS height
l = surface autocorrelation length

cos  −  − sin 2   vv = ( −1)

(
sin 2  − 1 + sin 2  )
 hh =
cos  +  − sin 2  (cos +  − sin 2  )
e = dielectric constant
σ0 is the backscattering coefficient
49
 Agricultural surface roughness statistics

Seedbed Harrowed Ploughed

Tillage Rms Rms Corr. Length Corr. Length

(mean) (std) (mean) (std.)
Seedbed 0.6 0.3 3.7 2.6
Harrowed 1.6 0.7 3.8 2.9
Ploughed 2.7 1.0 6.9 2.7

RMS heights s and correlation length l (mean and std in cm) 50

 Soil dielectric constant

Dielectric constant as a function

of soil volumetric water content

51
Sensitivity to soil moisture and surface roughness

For constant
High backscatter roughness
Low backscatter
➔ Wet soil
➔ dry soil

BUT :

For constant
Low backscatter
➔ Smooth surface High backscatter moisture
➔ rough surface

Simultaneous effect of roughness and moisture on the radar signal

52
 Our first ERS experiment in 1992
on irrigated area in Gharb, Morocco
We mapped irrigated fields, but to retrieve
soil moisture was found hard with a single
ERS data!

Le Toan, T, J.C. Souyris and P. Macchia, 1993

53
54
Polarisation in surface scattering

Smooth surface Rough surface

V H
V
V

◆ no depolarisation ◆ some depolarisation

no HV or VH backscatter HV or VH backscatter > 0

◆ Fresnel Reflectivity RH > RV ◆ Fresnel Reflectivity RH = RV

54

Thuy Le Toan, CESBIO, Toulouse, France

 Interactions mechanisms of a vegetation cover

* Order 0
Attenuated surface scattering

* Order 1: Simple scattering

Volume scattering Surface-volume scattering

 0
= 0
soil + 0
veg . + 0
soil − veg .

55
56
Volume scattering

Single and multiple scattering

56

Thuy Le Toan, CESBIO, Toulouse, France

57
Polarisation in volume scattering

Point scatterer Anisotropic scatterer Multiple scattering

V H H
V V V
H,V
-> no depolarisation -> depolarisation -> depolarisation

HV or VH: high backscatter

of vegetation canopy
57

Thuy Le Toan, CESBIO, Toulouse, France

Optical image
Sentinel-2

Radar image
Sentinel-1

58
59
60
 Contents

1. Introduction to SAR remote sensing

2. Statistical properties of SAR images

3. Physical content of SAR data

4. Application to agriculture

5. Application to forests –Biomass estimation

61
 How do the radars see the trees ?

X band C band L band P band VHF

Austrian pine l= 3 cm l= 6 cm l= 27 cm l= 70 cm l>3m

62
 Interaction mechanisms in a forest
1 6
4

5
Scatterers
3
contribution
Leaves, Needles 2

Primary Branches

Secondary
branches
1) Direct Crown scattering 4) Multiple trunk-ground
Higher order 5) Attenuated ground
2) Direct trunk-ground
branches
3) Trunk scattering 6) Direct ground scattering
Trunk
63
 Interactions mechanisms in a forest – C-band
6
1
5
Scatterers
contribution

Leaves, Needles
1) Direct Crown scattering
5) Attenuated ground
6) Direct ground scattering

64
Scattering mechanisms simulated by a R.T. model

P-band HH

Total
trunk-ground
crown-ground

crown

o Le Toan, T., Beaudoin, A., Riom, J.,

& Guyon, D. (1992). Relating forest
biomass to SAR data. IEEE
Transactions on Geoscience and
Remote Sensing, 30(2), 403-411.
P-band HV P-band VV
o Hsu, C. C., Han, H. C., Shin, R. T.,
Kong, J. A., Beaudoin, A., & Le Toan, T.
(1994). Radiative transfer theory for
Total crown Total polarimetric remote sensing of pine
forest at P band. International Journal of
Remote Sensing, 15(14), 2943-2954.
crown-ground
crown-ground crown o Beaudoin, A., Le Toan, T., Goze, S.,
Nezry, E., Lopes, A., Mougin, E., ... &
trunk-ground Shin, R. T. (1994). Retrieval of forest
biomass from SAR data. International
Journal of Remote Sensing, 15(14),
2777-2796.
65
 Change in dominant mechanism

Varzea Dry Varzea Wet Season

Season

SAR image (L- HH) SAR image (L-HH) 66

Forest structure- Boreal vs Tropical rain forest

67
Yu & Saatchi S., 2016

68
Dielectric constant of wood material
Francois Demontoux, Wood Sci Technol,
DOI 10.1007/s00226-017-0935-4, 2017

69
Dielectric constant of living wood

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 40,

NO. 9, SEPTEMBER 20022063
Diurnal and Spatial Variation of Xylem DielectricConstant in Norway Spruce
(Picea abies[L.] Karst.)as Related to Microclimate, Xylem Sap Flow,
andXylem Chemistry, Kyle C. McDonald,, Reiner Zimmermann, and John S.
Kimball

70
 The TropiSCAT campaign
A ground-based radar observing a tropical forest
– Located in French Guyana – same site as TropiSAR
– Team members from ONERA, CNES, CESBIO, POLIMI
– 20 antennas installed on top of the Guyaflux tower (55 m)
– Fully polarimetric (HH, HV, VH and VV)
– Vertical resolution capabilities
– One image every 15 minutes over a time span of one year

 Access to the vertical structure of temporal

decorrelation

71
Diurnal variation of Radar center of mass

Center of mass as a function of time (from midnight)

0.5
HH
HV
VH
0
VV

-0.5
Height [m]

-1

-1.5

-2
0 20 40 60 80 100 120
Time lag [h]

72
TropiScat ʽcentre of mass’
Height is relative to the height observed at midnight (about 20 m).

The center of mass is going down, from 7am , with minimum at 13-15 pm
and up again to original height at 22-23 h
The amplitude is about 1.5 m inside the forest.

Average of 10 dry days (without rain) during the period from 8 to 31 December 2011
73
 Diurnal cycle of radar intensity

Hamadi et al., 2013

Tour Guyaflux
Equipe Guyafor 74
Borealscat at Remningstorp, Sweden

Research infrastructure:
• A 50-m high radar tower in a hemi-boreal Antenna array

forest site in southern Sweden.

BorealScat tower

• Acquires tomographic P/L/C-band radar data

• On-site weather station and moisture sensors:
dendrometers, sap flow, and soil moisture Equipment hut
Electric power pole

• More details at www.borealscat.se Fence

Ditch

Antenna
Mature arrays
spruce
stand L. Ulander, A. Monteilh, 2020
75
Backscatter variability caused by freeze/thaw in boreal forest

Backscatter Jun 2017- Jun 2019: Canopy, Ground and Full forest
Monthly range of canopy
backscatter

HH

VV

VH Backscatter varies 2-4 dB

during summer, whereas
much larger variability during
winter (freeze/thaw).

L. Ulander, A. Monteilh, 2020

76

Monteith & Ulander, IEEE J-STARS, 14, 1967-1984, 2021 (open access)
77
Forest biomass is a key element in the carbon cycle

Biomass is an ECV (Essential Climate Variable) of the ESA CCI

(Climate Change Initiative)

1. Biomass consists of
approximately
50% carbon

2. Forests account for 70-90%

of the terrestrial above-ground
biomass, and the majority are
located in the tropics

3. The forest biomass stocks

and their change remain poorly
quantified Biomass = dry weight of woody matter
(in tons/ha) 77
78
Forests are a major carbon sinks, but can be C sources

Carbon sinks Carbon sources

International Conventions and National Determination on the forests aim at:

o Increase C sequestration
o Reduce the emissions

→ need to quantify C losses and gains for the carbon cycle

78
79
SAR intensity increases with biomass

L band (λ=25cm)

P band (λ=70cm)

79
SAR intensity increases with biomass

80
Scattering mechanisms simulated by a R.T. model

P-band HH

Total
trunk-ground
crown-ground

crown
o Le Toan, T., Beaudoin, A., Riom, J.,
& Guyon, D. (1992). Relating forest
biomass to SAR data. IEEE
Transactions on Geoscience and
Remote Sensing, 30(2), 403-411.

o Hsu, C. C., Han, H. C., Shin, R. T.,

Kong, J. A., Beaudoin, A., & Le Toan, T.
P-band HV P-band VV (1994). Radiative transfer theory for
polarimetric remote sensing of pine
forest at P band. International Journal of
Remote Sensing, 15(14), 2943-2954.

Total crown Total o Beaudoin, A., Le Toan, T., Goze, S.,

Nezry, E., Lopes, A., Mougin, E., ... &
Shin, R. T. (1994). Retrieval of forest
crown-ground biomass from SAR data. International
crown-ground crown Journal of Remote Sensing, 15(14),
2777-2796.
trunk-ground

81
Crucial point: for biomass inversion to perform well, data must be
processed to retain only the volume component of the forest canopy

82
 SAR measurement modes-BIOMASS

PolSAR PoI-InSAR TomoSAR

z z
z
x x x

o o o

y y y

83
84
SAR tomography, a new concept to explore 3D forest structure

Generates images of different forest layers from multi-orbit SAR images

Guyaflux tower
(Tropiscat experiment)
Height (m)
1
0.9
40 0.8
0.7
0.6
30 0.5
0.4
Tomographic 20 0.3
0.2
Processing 0.1
10 0
Normalised
0 backscatter
intensity

84
Tomography to understand scattering mechanisms
Provides the most complete description of all
contributions to the radar signal through 3D
reconstruction of forest backscatter.

Boreal forest

Tropical forest
85
Scattering mechanisms in tropical forests

Capon spectrum - HH channel

60
LiDAR height
40 1

Height [m]
0.9
20 0.8
0.7
0
0.6
0.5
200 400 600 800 1000 1200 1400 1600 1800 2000
0.4
Capon spectrum - HV channel 0.3
60 0.2
LiDAR height
0.1
40
0

Height [m]
20
Normalised
0
backscatter
200 400 600 800 1000 1200 1400 1600 1800 2000 intensity

Capon spectrum - VV channel

60
LiDAR height
40
Height [m]

20

200 400 600 800 1000 1200 1400 1600 1800 2000
Slant range [m]
86
Tomography to understand scattering mechanisms

Provides the most complete description of all

contributions to the radar signal through 3D
reconstruction of forest backscatter.

Boreal forest

Tropical forest
87
Isolate the layer not containing ground scattering

.. Whose biomass related to AGB

a b

Using Troll model (from Chave, 1999)

Tebaldini, S., Minh, D. H. T., d’Alessandro, M. M., Villard, L., Le Toan, T., & Chave,
J. (2019). The status of technologies to measure forest biomass and structural
MINH, Dinh Ho Tong, LE TOAN, Thuy, ROCCA, Fabio, et al. Relating P-band synthetic aperture
properties: State of the art in SAR tomography of tropical forests.
Surveys in Geophysics, 40(4), 779-801. radar tomography to tropical forest biomass. IEEE Transactions on Geoscience and Remote
Sensing, 2013, vol. 52, no 2, p. 967-979.
88
Ho Tong Minh Dinh et al., 2013, 14

89
90 La biomasse des forêts tropicales (≥ 500 t/ha)
mesurable par le SAR bande P de Biomass

90
Tomography techniques allow AGB up to 500t/ha

LPS, 2018

91
Isolating the volume by ground cancellation

HH-VV phase
difference:

30 m height

More by: S. Tebaldini

Ground level
92
93
Biomass estimation

1. Strong evidence that the volume layer 25-35 m above the

ground in tropical forest is strongly correlated with the total
AGB.
2. Development of methods for ground cancellation using
tomography or Pol-InSAR to isolate volume scattering.
3. Development of an approach to solve the volume scattering
equation that minimizes the need for reference data.

93
94
Summary

1. An introduction on radar remote sensing has been given, with

focus on agriculture and forests.

2. It is essential to understand the statistical properties of SAR

images, and the physical content of the SAR data , to interpret and
to analyse the images, and to develop retrieval and mapping
algorithms.

3. Lectures and training on advanced SAR polarimetry, Pol-InSAR

and TomoSAR will be provided in the following sessions.

94