132
M. Vespe et al.
legal infringements detection and potential environmental damage estimation. After
a brief description of common methodologies applied to SAR based oil spill detection, the oil spill detection problem is introduced through the description of data
quality analysis. This process aims at defining the suitability of the image data for
the service application. The portions of the data considered “usable”, i.e. exploitable
for the derivation of meaningful results, can be further analysed. As described in the
following, the detected oil spill candidates can be cross checked with other ancillary data (metocean, contextual and maritime traffic data). This shall subsequently
increase the reliability of the service by combining information on oil spill candidate location with the relevant degree of “risk” and “detection capability” properties
of the area of interest. For instance a traffic lane shall increase the likelihood
(“risk”) of having and oil spill, as opposed to low wind areas that reduce the performance of the SAR based algorithm (“detection capability”). Ultimately, this paper
introduces a possible way of automatically fusing the mentioned heterogeneous
information.
8.2 On the Use of SAR Imagery to Detect Oil Spills
Some oil on the sea surface dampens the short gravity-capillary waves generated
by the wind (Alpers and Hühnerfuss, 1988), leading to reduced Bragg scattering in
radar images. This results in a reduced radar backscatter to the sensor, thus creating
a darker signature in the image over the area of interest. This can be observed if
the local wind is greater than a threshold (typically 2–3 m/s) which, amongst other
factors, is dependent on the water salinity and temperature. When the wind speed is
too high, on the other hand, the short waves receive enough energy to counterbalance the damping effect of the oil film, and if the sea-state is fully developed, the
turbulence of the upper sea layer may break and/or sink the spill or a part of it. As a
result the oil spill is not detectable from the image.
The reliability of spill detection based on SAR data only is not fully robust to
guarantee consistent and automatic satellite-based detection of oil spills. This is a
consequence to the distinctive variability of radar based oil signatures, their spatial
features, and the interaction with the local environment. Moreover, other than oil
spills, a number of phenomena originate Bragg scattering reduction, making it difficult to discriminate between the so called “look-alikes”. Such phenomena can be
grouped as follows:
• Man-related (e.g. ship wakes, floating production facilities drain emissions);
• Atmospheric (e.g. wind sheltering, rain cells, atmospheric instability areas);
• Oceanographic (e.g. internal waves, coastal upwelling, eddies, current shears,
bathymetry/currents interaction, grease ice);
• Natural/Biological (e.g. natural seepage, fish oil in cold waters, algae blooms,
pollen from plants and trees, coral spawn, natural surfactants).
Many of these false alarm sources become even more pronounced in low wind
conditions, as a consequence of the reduced Bragg waves and therefore backscatter
M. Vespe et al.
legal infringements detection and potential environmental damage estimation. After
a brief description of common methodologies applied to SAR based oil spill detection, the oil spill detection problem is introduced through the description of data
quality analysis. This process aims at defining the suitability of the image data for
the service application. The portions of the data considered “usable”, i.e. exploitable
for the derivation of meaningful results, can be further analysed. As described in the
following, the detected oil spill candidates can be cross checked with other ancillary data (metocean, contextual and maritime traffic data). This shall subsequently
increase the reliability of the service by combining information on oil spill candidate location with the relevant degree of “risk” and “detection capability” properties
of the area of interest. For instance a traffic lane shall increase the likelihood
(“risk”) of having and oil spill, as opposed to low wind areas that reduce the performance of the SAR based algorithm (“detection capability”). Ultimately, this paper
introduces a possible way of automatically fusing the mentioned heterogeneous
information.
8.2 On the Use of SAR Imagery to Detect Oil Spills
Some oil on the sea surface dampens the short gravity-capillary waves generated
by the wind (Alpers and Hühnerfuss, 1988), leading to reduced Bragg scattering in
radar images. This results in a reduced radar backscatter to the sensor, thus creating
a darker signature in the image over the area of interest. This can be observed if
the local wind is greater than a threshold (typically 2–3 m/s) which, amongst other
factors, is dependent on the water salinity and temperature. When the wind speed is
too high, on the other hand, the short waves receive enough energy to counterbalance the damping effect of the oil film, and if the sea-state is fully developed, the
turbulence of the upper sea layer may break and/or sink the spill or a part of it. As a
result the oil spill is not detectable from the image.
The reliability of spill detection based on SAR data only is not fully robust to
guarantee consistent and automatic satellite-based detection of oil spills. This is a
consequence to the distinctive variability of radar based oil signatures, their spatial
features, and the interaction with the local environment. Moreover, other than oil
spills, a number of phenomena originate Bragg scattering reduction, making it difficult to discriminate between the so called “look-alikes”. Such phenomena can be
grouped as follows:
• Man-related (e.g. ship wakes, floating production facilities drain emissions);
• Atmospheric (e.g. wind sheltering, rain cells, atmospheric instability areas);
• Oceanographic (e.g. internal waves, coastal upwelling, eddies, current shears,
bathymetry/currents interaction, grease ice);
• Natural/Biological (e.g. natural seepage, fish oil in cold waters, algae blooms,
pollen from plants and trees, coral spawn, natural surfactants).
Many of these false alarm sources become even more pronounced in low wind
conditions, as a consequence of the reduced Bragg waves and therefore backscatter