A Simulator for SAR Sea Surface Waves Imaging
Ferdinando Nunziata, Attilio Gambardella and Maurizio Migliaccio
Università degli Studi di Napoli Parthenope
Dipartimento per le Tecnologie
Via Medina, 40 - 80037 Napoli
Email: {ferdinando.nunziata, attilio.gambardella, maurizio.migliaccio} at uniparthenope.it
Abstract—This paper describes a Synthetic Aperture Radar
(SAR) sea surface waves simulator. The simulator, based on
the velocity bunching (VB) theory, has been developed and
impemented modularly and its use can also assist microwave
remote sensing courses. The present version of the software is
run in classes at National Oceanographic Centre of Southampton
(NOCS), UK and at the Università di Napoli Parthenope, Italy.
I. I NTRODUCTION
Microwave remote sensing sensors are widely used in sea
monitoring due to their all-weather day and night capability.
Among these sensors the Synthetic Aperture Radar (SAR)
has succesfully demonstrated its capacity to uniquely provide
valuable high resolution information for marine applications
[1]. However, SAR imaging of the sea surface is considerably
more complex than the imaging of a stationary scene. In
particular, though wave-like patterns are often discernible on
sea surface SAR images obtained both from aircraft and space
missions, the relationship between such patterns and the actual
sea surface wave field is an intriguing and non-trivial issue
[2]. Hence, simulation procedures can be very helpful to
shed light in physical aspects governing the SAR sea suface
waves imaging. Two main theories have been proposed: the
distributed surface (DS) theory [3] and the velocity bunching
(VB) one [4]. Most of the available SAR surface waves
simulators [2], [5], [6], [7], are developed in programming
language not user-friendly, they are time consuming and thus
they are not able, for example, to be run in classes at University
for educational purposes.
In this paper a SAR sea surface waves simulator, based on
VB theory, is presented and discussed. It is entirely developed
in Matlab environment, which is probably the most popular
programming environment at educational and research centres.
The simulator can run on Windows, Mac OS and Linux PC
systems and only a student version of Matlab is required.
To facilitate users a Graphic User Interface (GUI) has been
developed.
The present version of the software is run in classes at National
Oceanographic Centre of Southampton (NOCS), UK and at the
Università di Napoli Parthenope, Italy.
The paper is organized as follows. In section II the background scattering theory governing the SAR sea surface waves
imaging is reviewed. In section III the simulation approach
is described and in section IV some meaningful experiments
are presented and discussed. In section V the conclusions are
drawn.
1-4244-1212-9/07/$25.00 ©2007 IEEE.
II. T HEORETICAL FACTS
The scattering machanism governing the formation of a
conventional SAR sea surface waves image can be modelled
by a two-scale scattering model which includes the sea dynamics. Since satellite and airborne SAR generally operates
at incidence angles ranging between 20◦ and 70◦ , for low to
moderate sea state, it is normally assumed that the small scale
backscattering mechanism is the Bragg one [2], [4]. According
to Bragg theory only sea waves whose wavelengths are the
same order of the incidence electromagnetic one are ‘seen‘ by
the SAR. Thus longer waves are imaged indirectly because of
Real Apertur Radar (RAR) modulation mechanism and motion
induced effects (MI).
The RAR mechanism can be described by a linear function,
the RAR Modulation Transfer Function (MTF), which relates
the Normalized Radar Cross-Section (NRCS) to the long sea
wave field. Under the hypothesis of linear modulation the
RAR MTF does not depend on the long wave field an can be
decomposed in three terms: tilt, Rt (·), range bunching, Rrb (·),
and hydrodynamic modulation, Rh (·) [4], [8]. Thus, according
to this theory, the dynamic NRCS, σ o (·), can be written as [8]:
(
M
X
o
o
RRAR (Km ) zm (Km )
σ (xo ) = σ 1 +
m=1
)
· cos(Km xo + ϕm + ψm )
(1)
.
Here xo = (xo , yo ) and x = (x, y) are the ocean surface
and the corresponding SAR plane, respectively. In particular
x denotes the coordinate in flight (azimuthal) direction, y
denotes the one in cross-track or ground range direction.
The amplitudes z(K) are related to the two-dimensional
ocean wave spectrum sampled by M long wave wavenumbers
K. ϕm is an uniformly distributed random variable, σ o is
the NRCS evaluated according to Small Perturbation Model
(SPM), RRAR (Km ) and ψm denote modulus and phase of the
RAR MTF, respectively [8].
The MI effects are SAR inherent artefacts. In fact, since
SAR is a coherent system, in order to form an image, it relies
on the signal phase structure derived from each elemental
scatter in the observed scene. In the case of ocean surface,
in presence of a longer gravity which across the scene,
all the particles, including the short Bragg resonant waves,
are advected giving rise their orbital velocity. Thus, a SAR
786
coherently senses the radial component of this orbital motion.
In particular the radial component of the orbital velocity
gives rise to the well-known velocity bunching phenomenon
[4], [2]. The radial component of the orbital acceleration is
responsible for the degradation of the azimuthal resolution.
Since both orbital acceleration and orbital velocity vary along
the flight direction they can produce a wave-like pattern onto
SAR image. Actually, in addition to this acceleration induced
mechanism, the azimuthal resolution is also degraded by the
sub-resolution scale variations of the orbital velocities which
characterize the different backscattering elements within the
SAR resolution cell. This term can be modelled by a scene
coherence time [2], [8].
Once the main processes responsible for the wave-like formation onto SAR images have been described, the relationship
between the SAR image intensity I(x) and the dynamic NRCS
is given by [2], [8]:
Z o
σ (xo )
I(x) =
δ(y − yo )
ρa (xo )
(
2 )
(2)
R
π2
x − xo − ur (xo )
· exp −
.
ρa (xo )
V
Here R and V are slant range of the target and platform
velocity, respectively. δ(·) is the Dirac delta function. ρa (·)
is the degraded azimuthal resolution, which depends on the
orbital acceleration, scene coherence time, sensor integration
time and nominal azimuthal resolution [2]. ur (·) is the orbital
velocity [8].
MI effects can make the overall SAR MTF higly non-linear.
In fact, both VB and the degradation in azimuthal resolution
can be, for certain radar and sea parameters, highly non-linear
mechanisms [2]. According to [4], [2] VB depends on the
gradient of the azimuthal component of the orbital velocity.
The degree of non-linearity of VB is strongly related to the
direction of the long waves. In detail, for range travelling
waves, VB vanishes and the overall SAR MTF can become
linear, for azimuthal travelling waves VB is highly non-linear
as soon as the SAR MTF [4], [2].
III. S IMULATION FACTS
The simulator has been designed and implemented modularly and the numerical code can be described according to
the steps depicted in Fig.1 which are described hereafter [9].
Step I generates a realization of the sea surface displacement
associated with the long sea waves. The directional wave
spectrum, in the range relevant to the long sea waves, is
approssimated by means of a JONSWAP spectrum with a
cosine-type spreading function [2].
Step II models the NRCS according to the SPM. In particular
the short wave spectrum is modelled by means of the Phillips
spectrum.
Step III evaluates the overall SAR MTF which consists of the
RAR MTF and the MI effects. The long and the short waves
are considered aligned.
Step IV evaluates the SAR noise-free intensity image.
1-4244-1212-9/07/$25.00 ©2007 IEEE.
Fig. 1.
SAR simulator block scheme
Fig. 2.
SAR simulator GUI
Step V generates the SAR noisy intensity image. In particular
two noise sources are considered: the additive and the multiplicative one. The latter one is modelled by means of the
Weibull distribution [10].
The external inputs to the simulator are grouped into three
classes, i.e. sensor, sea and noise input [9]. User is allowed
to set such parameters scrolling the input parameters menu
through the GUI (Fig.2). Moreover, the GUI has two other
menus: the Option and the File one. The first one allows,
for example, reducing the image size in order to have a
quicker simulation. The second one allows opening a previously simulated SAR image, to display its statistics, to print
the SAR image and to exit from the SAR simulator. The
simulation starts clicking the RUN button (Fig.2). Obviously
the processing time depends on many factor, but the algorithm
is quite fast (of the order of few minutes). Once the processing
is completed the noisy and the noise-free intensity images are
displayed. Then it is possible to save these images as *.dat
files. Further an information *.txt file, which contains input
and image parameters, is also generated in an user defined
folder. These latters are the outputs of the simulator [9].
787
Fig. 3. SAR 250x250 pixels noisy intensity image relevant to a single 60m
range travelling wave
Fig. 5. SAR 250x250 pixels noisy intensity image relevant to a single 60m
azimuth travelling wave
Fig. 4. Plot of the 60m range travelling wave (top) and of the simulated one
(down)
Fig. 6. Plot of the 60m azimuth travelling wave (top) and of the simulated
one (down)
IV. E XPERIMENTS
The last two experiments are relevant to the simulation of
more complex sea wave field, which is generated by means
of a fully-developed JONSWAP spectrum with a spreading
factor equal to 10 (which corrspond to a broaden spreading
function). Such spectrum has been approximated by means of
30 wavenumber and 20 angular directions [9].
In the first experiment a 100m peack wavelength, relavant to
a range traveling wave, has been simulated. The SAR noisy
intensity image is shown in Fig.7.
The second experiment is similar to the former one but an
azimuthal travelling wave has been considered. The SAR noisy
intensity image is shown in Fig.8.
Unlike the monochromatic cases formerly showed, these experiments are more complicated to analize. In particular a
spectral analysis is needed to emphasize the behaviour of the
SAR MTF [9].
In all subsequent experiments reference is made to ERS-1/2
SAR parameters. Speckle noise is Weibull distributed with a
shape parameter equal to 1.5. The wind speed is 10m/s and
standard values for sea temperature and salinity have been
considered [9].
The first experiment is relevant to the simulation of a single
60m range travelling wave. The SAR noisy image is shown in
Fig.3. In order to recognize the linearity of the overall SAR
MTF a range transect is made in the noise free image (not
shown) and referred to the corresponding long wave (Fig.4).
Since a range travelling wave has been simulated it is possible
to experience the linearity of the SAR imaging process [4], [9].
The second experiment is relevant to the simulation of a
single 60m azimuthal travelling wave. The SAR noisy intensity
image is shown in Fig.5. Similarly to the former case an
azimuth transect is made on the SAR noise-free image and
referred to the corresponding long wave (Fig.6). Unlike the
previous case, since an azimuthal travelling wave has been
simulated, the SAR MTF is strongly non-linear [4], [9].
1-4244-1212-9/07/$25.00 ©2007 IEEE.
V. C ONCLUSION
A new SAR sea surface waves simulator has been developed and tested. The software has been developed in Matlab
enviroment, which is probably the most popular programming
environment at educational and research centres. Since the
788
R EFERENCES
Fig. 7. SAR 250x250 pixels noisy intensity image relevant to 100m range
travelling peack wavelength
[1] C. R. Jackson and J. R. Apel, Synthetic Aperture Radar Marine Users
Manual. Washington, DC: NOAA, 2004.
[2] C. Brüning, W. Alpers, and K. Hasselmann, “Monte-Carlo simulation
studies of the nonlinear imaging of a two dimensional surface wave field
by a synthetic aperture radar,“ Int. J. Remote Sensing, vol. 11, no. 4,
pp. 1695–1727, 1990.
[3] R. O. Harger, “The synthetic aperture image of time-variant scenes,“
Radio Sci., vol. 15, pp. 749–756, 1980.
[4] W. Alpers, D. Ross, and C. Rufenach, “On the detectability of ocean
surface waves by real and synthetic aperture radar,“ J. Geophys. Res.,
vol. 86, no. C7, pp. 6481–6498, 1981.
[5] R. O. Harger and C. E. Korman, “Comparisons of simulated and actual
synthetic aperture radar gravity waves images,“ J. Geophys. Res., vol. 93,
pp. 13867–13882, 1989.
[6] G. Franceschetti, M. Migliaccio, and D. Riccio, “On ocean SAR raw
signal simulation,“ IEEE Trans. Geosci. Remote Sensing, vol. 36, no. 1,
pp. 84–100, 1998.
[7] P. W. Vachon R. K. Raney, and W. Emery, “A simulation for spaceborne
SAR imagery of a distributed moving scene,“ IEEE Trans. Geosci.
Remote Sensing, vol. 27, no. 1, pp. 67–78, 1989.
[8] M. Bao, On the imaging of a two-dimensional ocean surface wave
field by an along-track interferometric synthetic aperture radar. Ph.D.
Thesis, Hamburg, 1995.
[9] F. Nunziata, A. Gambardella, and M. Migliaccio, “An educational SAR
sea surface waves simulator,“ Int. J. Remote Sensing, submitted for
pubblication.
[10] M. Greco, F. Bordoni, and F. Gini, “X-band sea-clutter nonstationarity:
influence of long waves,“ IEEE Trans. Geosci. Remote Sensing, vol. 29,
no. 2, pp. 269–283, 2004.
Fig. 8. SAR 250x250 pixels noisy intensity image relevant to 100m azimuth
travelling peack wavelength
software has been conceived and implemented modularly it
can easily up-to-dated. The simulator can be used at Universities for educational purposes and, if tailored block are
developed, for scientific purposes at research centres. The
present version of the software is run in classes at National
Oceanographic Centre of Southampton (NOCS), UK and at the
Università di Napoli Parthenope, Italy. The ESA endorsement
has been required and is under evaluation.
ACKNOWLEDGMENT
The authors would like to thank Prof. Werner Alpers, University of Hamburg, Germany for useful discussion at ESAESRIN, Frascati, Italy and Prof. Roland Romeiser, University
of Hamburg, Germany, for useful discussion at University of
Hamburg.
1-4244-1212-9/07/$25.00 ©2007 IEEE.
789