EITN90 Radar and Remote Sensing

Lecture 4: Characteristics of Clutter

Daniel Sjöberg

Department of Electrical and Information Technology

Spring 2019
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

2 / 52
 Learning outcomes of this lecture
In this lecture we will
I Characterize the clutter
I Observe orders of magnitude from different sources
I Have an initial discussion on clutter suppression
I See a few empirical models

3 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

4 / 52
 What is clutter?

I Backscattering from natural objects, such as precipitation,

vegetation, soil and rocks, or the sea.
I When trying to detect man-made object, it is considered an
unwanted interference, masking the signal.
I When surveying natural processes (thickness of ice caps,
weather etc), it may be the main signal of interest.

5 / 52
 Examples of clutter
Images from http://www.radartutorial.eu/ (CC BY-SA 3.0).
PPI = Plan Position Indicator.

PPI screen of an ATC-radar Sea-Clutter on a PPI-Scope.

with targets and clutter. Wind from 310◦ or 130◦ .
Observing how the image evolves with time gives further
information. Clutter can fluctuate and move.
6 / 52
 Radar imaging (35 GHz, 1ft x 1ft res)

Mix of areas with uniform scattering (grassy lawns) and

non-uniform (trees, man-made structures). May look very different
for other frequencies. 7 / 52
 target

clutter

Discussion

8 / 52
 Clutter vs noise

There are some significant differences between clutter and noise.

9 / 52
 Scattering coefficients

The received electric field strength from the i-th scatterer is

proportional to (k collects factors common to all scatterers)
1/2 √
Pt G2 λ2 σi
  
σi 4π
|Ei | ∼ = k , arg{E i } = − θ i + di
(4π)3 Ls R4 d2i λ
X X √σi  
4π

k√
E= Ei = k 2 exp −j di + θi = 2 σejφ
di λ d
i i
√
The complex number σejφ is the backscatter coefficient and d
is the nominal distance to the clutter. 10 / 52
 Clutter polarization scattering matrix
Taking polarization effects into account, the concept of the
backscatter coefficient can be extended to the polarization
scattering matrix (PSM).
√ √
σ ejφHH σHV ejφHV

S = √ HH jφVH √
σVH e σVV ejφVV

The PSM could also be expressed in circular polarization (right

hand CP and left hand CP). Additional information on the
scatterer can be obtained by considering, for instance,
√ √
I Parallel/cross polarization ratio: σHH / σVH .
√ √
I Vertical/horizontal polarization ratio: σVV / σHH .
I Polarimetric phase: φHH − φVV .
These measurements require a radar capable of transmitting and
receiving individually in all polarizations, which is expensive.
11 / 52
 Surface and volume reflectivity
√
The absolute square of the complex backscatter coefficient σejφ
is the radar cross section σ of the clutter.

To characterize clutter originating from a surface, use the surface

reflectivity
σ m2
σ0 = [σ 0 ] = 2 = unitless
A m
where A is the illuminated clutter area.

For clutter scatterers in a volume, use the volume reflectivity

σ m2
η= [η] = = m−1
V m3

where V is the illuminated clutter volume.

12 / 52
 Beam limitation vs pulse limitation
Depending on pulse length cτ , the illuminated clutter area is
limited by the projected beam or the projected pulse (θ3 and φ3
are the 3 dB azimuth and elevation beam widths, respectively):

φ3 φ3 cτ
δ δ

     
θ φ θ
πR2 tan 23 tan 23 cτ R tan 23
πR2 θ3 φ3 cτ Rθ3
A= sin δ ≈ 4 sin δ A= cos δ ≈ 2 cos δ

The illuminated clutter volume is restricted by the pulse length

cτ
πR2 θ3 φ3 cτ
V = 4 2

Discussion

13 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

14 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

15 / 52
 Dependence on grazing angle
The surface reflectivity depends on the grazing angle.

Based on theory and measured data for land and sea. The
behavior at low grazing angles is motivated by the surface
becoming smoother (less backscattering). Rayleigh’s definition of a
smooth surface is
λ
σh sin δ <
8 16 / 52
 Random nature of clutter
The clutter response varies with time and space due to motion of
the radar or the scatterers, for instance due to wind. A statistical
approach is necessary, for instance using the Weibull distribution
( b−1  b
bσ
α exp − σα σ≥0
pσ =
0 σ<0
b / ln 2 and σ is the median of the distribution.
where α = σm m

17 / 52
 Spatial statistics for ground clutter

Example of how measurements are fitted to theoretical models.

18 / 52
 Land reflectivity: grass and crops

Follows the general trend shown before. Depression angle is an

angle relative to the radar system, same as grazing angle for level,
horizontal surface. This angle is easier to control in an experiment.

19 / 52
Land reflectivity: trees

20 / 52
 Land reflectivity: frequency

Higher frequency implies higher reflectivity.

21 / 52
Sea reflectivity: affecting factors

22 / 52
Sea reflectivity: measurements

23 / 52
 Sea reflectivity: range dependence

Theoretically, sea clutter should decrease as R−3 , but may

decrease faster. Discussion 24 / 52
 Clutter suppression, decorrelation time
I The clutter decorrelation time τ0 , is the time over which the
clutter response is coherent (stable phase and amplitude).
This is frequency dependent.
I If the target signal is stable over longer time than τ0 , the
signal-to-clutter ratio can be improved by averaging.
I If PRI > τ0 , each clutter sample is uncorrelated.

25 / 52
 Clutter frequency spectra

Theory predicts Gaussian-shaped spectra, but actual measurements

often result in a slower roll-off with frequency. This may be due to
imperfections in the systems, since a very well-controlled
experiment (Billingsley, ref [11]) was well modeled by a Gaussian
distribution.
26 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

27 / 52
 Atmospheric clutter
Most volumetric (atmospheric) clutter is due to rain or other
precipitation. It depends on rain rate, and the drop-size (typically
0.5–4 mm) in relation to the wavelength λ.

Strongest response around ka ≈ 1, radius a ≈ λ/(2π), or a

diameter around λ/3.
28 / 52
Rain data

29 / 52
Snow data

30 / 52
 Rain decorrelation time

Decorrelation time in the order of milliseconds. This corresponds

to a limit for maximum PRF in order to have uncorrelated clutter
responses in each pulse (PRFmax = 1/τ0 ).
31 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

32 / 52
 General remarks

Clutter is notoriously difficult to model, due to the complexity of

the real world phenomena it represents. But still, explicit models
may provide useful approximations when evaluating the radar
scenario.

33 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

34 / 52
 GTRI empirical model

The following model was developed in the late 1970’s by GTRI =

Georgia Tech Research Institute.
" #
0 B −D
σ = A(δ + C) exp
1 + 0.1σ
λ
h

I δ is the grazing angle in radians

I σh is the rms surface roughness
I λ is the wavelength
I A, B, C, and D are empirically derived constants
Discussion

35 / 52
GTRI coefficients

36 / 52
Comparison of GTRI model with measured data

37 / 52
GTRI sea clutter model

38 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

39 / 52
 Rain clutter

The model parameters A and B below can be fitted to the rain

data in Figure 5-33 (where R is the rain rate in mm/hr):

η = ARB [m−1 ]

40 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

41 / 52
Land reflectivity

42 / 52
Sea reflectivity

43 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

44 / 52
 Conclusions

I Characterization of clutter: backscatter coefficient, surface

reflectivity σ 0 , volume reflectivity η.
I Illuminated area/volume determines the clutter RCS.
I Clutter decorrelation time needs to be considered for clutter
suppression.
I Some empirical models exist for estimating the reflectivity for
different contexts.

45 / 52
 Outline

1 Introduction and definitions

2 General characteristics of clutter

Surface clutter
Atmospheric clutter

3 Clutter modeling
Surface clutter
Atmospheric clutter
Summary of clutter results

4 Conclusions

5 Lab on Friday

46 / 52
 About the lab
I The lab will take place in the same room as the exercises.
Note the time: 8–12!
I The lab is based around a simple ultrasonic sensor placed on a
stepper motor, controlled by an Arduino unit.
I Read the lab instructions carefully before the lab! They
are available on the course web site, under “Lectures”.

47 / 52
 A practical problem: interference!

Since several units will be operating at the same time, they may
interfere with each other, meaning one unit may receive both its
own echo (intended) and the direct signal of another unit (not
intended).
I We can use two rooms in order to reduce problems.
I In each room, make sure to spread out, and try not to point
your radar in the direction of others (remember signals will
also reflect in walls, but the range is only a couple of meters).
The lab is done in pairs of two. Ask your lab leader Sebastian if
you get strange results, or if there are any other questions.

Before you leave the lab, demonstrate your findings to the

lab leader in order to be approved on the lab!

48 / 52
 Discussion

Why is there a shadow behind the targets?

Go back

49 / 52
 Discussion

Why is there a shadow behind the targets?

Answer:
I Only diffracted signals reach what is directly behind the
targets, leading to significantly reduced illumination and less
clutter.
Go back

49 / 52
 Discussion

Why is the pulse limited volume estimated with a range of cτ /2

rather than cτ ?

Go back

50 / 52
 Discussion

Why is the pulse limited volume estimated with a range of cτ /2

rather than cτ ?

Answer:
I This is the range when the last reflection of the initial pulse
overlaps with the first reflection of the end of the pulse.
Go back

50 / 52
 Discussion

Give a suggestion why sea clutter tends to decay slower than

anticipated by theory. (No uniquely true answer!)

Go back

51 / 52
 Discussion

Give a suggestion why sea clutter tends to decay slower than

anticipated by theory. (No uniquely true answer!)

Answer:
I The book suggests a ducting phenomenon, with waves
trapped close to sea surface, leading to a slower decay of the
excitation of clutter. Open case!
Go back

51 / 52
 Discussion

In the expression (where δ is the grazing angle, σh is the rms

surface roughness, and λ is the wavelength)
" #
−D
σ 0 = A(δ + C)B exp
1 + 0.1σ
λ
h

which parameters A, B, C, or D would you say have the strongest

connection to angle of incidence and surface roughness?

Go back

52 / 52
 Discussion

In the expression (where δ is the grazing angle, σh is the rms

surface roughness, and λ is the wavelength)
" #
−D
σ 0 = A(δ + C)B exp
1 + 0.1σ
λ
h

which parameters A, B, C, or D would you say have the strongest

connection to angle of incidence and surface roughness?

Answer:
I C (and to some extent B) gives an offset to the angle of
incidence, and D gives a weighting to the surface roughness.
Go back

52 / 52