Estimating the Wind Vector from Radar SAR Images when Applied to the Detection of Oil Spill Pollution
Youcef Smara (2), Juerg Lichtenneger
(1), Samy Bouchaib (2), Fabio Del Frate (3), Luca Salvatori (3)
(1) European Space Agency, ESRIN, Frascati (IT)
(2) Houari Boumediene University of Sciences and Technology, Faculty of Electronics
and Computer Science - Image Processing and Radiation Laboratory, Algiers
(DZ)
(3) Università Tor Vergata - Dipartimento di Informatica Sistemi e
Produzione (IT)
The capacity of synthetic aperture radar to observe
the sea surface and its potentiality for evaluating the wind vector and for
the interpretation of atmospheric and oceanic phenomena make the radar a useful
tool in the control and surveillance of oil spill pollution.
The object of this study is to improve the procedure for the detection of
hydrocarbon oil slicks based on neural networks described in [1] by means
of the elaboration of an operational tool allowing the wind vector to be estimated
from radar SAR images. This is done in order to include wind speed, in addition
to parameters linked to the physics and the geometry of the oil spill, in
the neural network as additional input.
Physical and geometric parameters alone are not always sufficient to provide
enough information in order to identify the nature of the oil spill. There
are several natural phenomena that can have the same radar signature as an
oil spill. To help in the identification of the suspect objects, supplementary
data, such as the wind speed that can be calculated from radar images, have
proved themselves to be efficient [2].
It is the fact that when the wind speed is greater than 7-8 m/s (and less
then 15 m/s), the probability that the slick is an oil spill increases significantly
because all other types of spills tend to disappear in the meantime. When
the wind speed is less than this value, the slick may also be caused by natural
phenomena [3].
Integrated with the tool for the estimation of the wind vector, represented
by the two boxes on the left side of Fig.1, the method for the oil spill identification
can be described by the following procedures:
1) selection of the area concerned (the area containing the suspect slick)
2) evaluation of the wind direction.
3) calculation of the wind speed.
4) calculation of physical and geometric parameters characterising the slick.
5) decision as to the nature of the slick. (oil spill or look-alike)
The operator is assigned to do the first two stages, by visual inspection, whereas the last three stages are automatic.
Figure 1 - Synoptic of the method of identification
Wind speed is calculated using the CMOD4 model. This is a model
for evaluation of wind vector, initially developed for radar scatterometers
[4]. It gives the backscattering coefficient according to wind speed, wind
direction and the angle of incidence. The model, however, can be applied to
the radar SAR images.
The inversion of the CMOD4 model allows the calculation of wind speed from
an SAR image but with an interactive pre-estimate on the image by the operator.
This interactive inspection is based mainly on the interpretation of atmospheric
phenomena.
Wind represents a very important atmospheric phenomenon capable of modifying
the surface of the sea sufficiently for the coefficient of radar backscattering
to be modified.
The greatest number of wind-provoked phenomena are then collected in order
to estimate wind direction. (eg. the presence of waves of atmospheric gravity
which appear to leeward of mountainous coastal areas, perpendicularly to the
direction of wind [5])
Figure 2 - Waves of atmospheric gravity generated
by a west wind blowing across a
mountainous coastal area
The shape of an oil spill is also modified by the wind in the same way as the surface of the sea. It can cause iridescence and bending angles [6].
Figure 3 - Two oil spill which have become iridescent due to north wind
The identification method of oil spill and the estimation process
of wind vector are in validation and giving preliminary encouraging results.
Other parameters linked to the atmosphere, such as wind direction can be added
to the input data of neural networks to improve the identification process,
for that an automation of wind direction estimation must be performed.
References:
- F. Del Frate, A. Petrocchi, J. Lichtenegger, G. Calabresi," Neural networks for oil spill detection using ERS-SAR data ", IEEE Transactions of Geoscience and Remote Sensing, vol. 38, n. 5, pp. 2282-2287, 2000.
- H. A. Espedal, and T. Wahl, "Satellite SAR oil spill detection using wind history information" Int. J. Remote Sensing, 1999, vol. 20, no. 1, 49-65
- Torill Hamre, Heidi Espedal, Paul Samuel and Stein Sandven, "Operator's Manual for Slick Analysis" NERSC Special Report, no.37. February 1996
- Stoffelen A & D Anderson, Scatterometer data interpretation: Estimation and validation of the transfer function CMOD4, JGR 102, 5767-5780, 1997.
- GARY A. MASTIN, CHARLES A. HARLOW, OSCAR K. HUH, AND s. A. HSU, "Methods of Obtaining Offshore Wind Direction and Sea-State Data.From X-Band Aircraft SAR Imagery of Coastal Waters" , IEEE of Oceanic Engineering, vol. OE-10, no. 2, April 1985
- "Manuel pour l'observation aérienne des pollutions pétrolières," CEDRE, Centre de Documentation, de Recherche et d'Experimentations sur les Pollutions Accidentelles des Eaux., 36p, 1993.