4. INTRODUCTION CSK Announcement of Opportunity: Satellite surveillance of Oil Pollution in the Adriatic Sea (SOPA) ~100 Wide Region mode images (VV pol)
5. OIL SPILL DETECTION CHAIN Classification output: a potential detection in the Adriatic Sea. Original Wide Region and pre-processed image
6. 18 May 2011, Two CSK WR images spaced by 12 hours OPERATIONAL CASE
The potential of COSMO-SkyMed is examined for oil spill detection, focusing on the Adriatic Sea. In view of the fact that accidental pollution rarely occurs in the Mediterranean basin, operational discharges have progressively been identified as the main source of pollution from ships. In particular, recent studies have demonstrated that the Adriatic Sea is significantly affected by this phenomenon, as testified by the correlation of the oily discharges with high traffic density. Mapping activities using limited resolution instruments over the Adriatic Sea are also reported in [2], demonstrating a significant concentration of oil spills in the area. Furthermore, the Adriatic is a closed sea with a relatively small depth, having limited water exchange with the rest of the Mediterranean, and therefore is more sensitive to marine pollution from ships.
Therefore, the potential of acquiring COSMO-SkyMed images, which are spaced by only a few hours in the Adriatic Sea, offers significant advantages for operational services in terms of detection confirmation capabilities, spill drift measurements and follow-up procedure definition in support to aerial and naval means (AIM OF THE PROJECT) COSMO-SkyMed Wide Region mode acquisitions, with a trade-off between coverage (100x100 km 2 ) and multi-look resolution (~30x30 m 2 at 13 looks) that is particularly suitable for oil spill detection.
The automatic oil spill detection chain is first implemented by a pre-processing stage ( i.e. the image is despeckled, radiometrically normalised in order to equalise the incidence angle sea backscatter influence, and land masked). This is followed by a contrast based segmentation and a feature extraction process. This stage is performed following an object-oriented approach in order to better discriminate the misclassification sources that affect traditional pixel-based methodologies.
The detected dark patches are not visible 12 hours after the first acquisition, highlighting how the revisit time can be crucial for this specific application . This is further corroborated by the narrow-straight shape of one of them (dark patch ‘a’) and its tapered front, which suggest that the spill is very fresh ( i.e. less than 2 hours), implying a persistency of less than 12 hours. Such revisit time performance is not only crucial to improve the monitoring capabilities by increasing the number of observations, but is also required to analyse how the oil spill drifts with time and to assess its potential impact, directly connect to the level of alert that can be generated >>>>> Introduce AIS: AIS is a system that allows vessels to electronically exchange data such as identification and state vector (position, course over ground, speed over ground etc.), with other nearby ships and shore stations. The IMO SOLAS Convention [4] requires AIS to be fitted aboard international voyaging ships of 300 GT or more, 500 GT and upwards for cargoes not in international waters and all passenger ships regardless of size. AIS traffic data can be used as additional information to point the competent authorities towards vessels potentially responsible for the pollution [1]. The possibility to correlate the AIS data with the SAR-observed features is here discussed through two examples.
The polluting vessel Is directed South (longer and darker segments located North) (AIS querty &quot;utc&quot; > '2011-03-17 18:30' AND &quot;utc&quot; < '2011-03-18 4:31‘)
(AIS query &quot;utc&quot; > '2011-03-17 18:30' AND &quot;utc&quot; < '2011-03-18 4:31‘ AND directed South) MARPOL 73/78
The SAR acquisition (4:27 UTC) is from 1h 30’ to 1h after the UTC vessel’s AIS track, which results to be the most likely to originate the dark patch. This is because of the narrow-straight shape, which suggests the oil spill is very fresh (less than 2 h) (Query &quot;utc&quot; > '2011-03-17 20:26' AND &quot;utc&quot; < '2011-03-18 4:31‘ AND &quot;mmsi&quot; = 235699000)
Also in this case, the shape of the dark patch, more spread at NW, suggests the spilling vessel was going Southwards. We can isolate those vessels (query &quot;date&quot; > ' 11-03-17 23:00:00' and &quot;date&quot; < ' 10-11-03-18 04:55:00' T_SAR = 20110318042739)
Additional data show that wind was low (~6-7 m/s) blowing from S/SE during the 12 hours before the SAR acquisition: it is not expected a large wind drift component. The oil seepage circular shape 6 km SW from the oil spill shows a low overall current component. Although the shape of this spill is more distorted than the previous one, and therefore less fresh (e.g. >2h), the wind/current drift implies that the polluting vessel has passed in proximity of the imaged oil spill position. (query &quot;date&quot; > ' 11-03-17 23:00:00' and &quot;date&quot; < ' 10-11-03-18 04:55:00' T_SAR = 20110318042739 and directed South)
The acquired images after being processed have been analysed in order to estimate the instrument quality and detection capability. The Noise Equivalent Sigma Zero ( NEσ0 ) performance, as directly measured from the image, was found to be better than for other instruments operating in similar modes. This is a key parameter for oil spill detection applications: a good NEσ0 is essential to be able to detect oil spills also in areas of low ambient backscatter. Furthermore, good NEσ0 performance enables the use of shallower incidence angle products (see image!) For the present data set, the radiometric resolution measured directly from the image products (< 1.85 dB) is in line with such requirement. Furthermore, for a specific wind speed the X-band damping ratio between oil free and oil covered sea backscattering is larger than at C-band [9], making the frequency particularly appropriate for oil spill detection. Besides the low revisit time of the COSMO-SkyMed constellation, its use by the competent authorities is still conditional to the reliability of the information and its latency, i.e. the time between the information acquisition and its delivery to the user. The former aspect is relevant for the decision on further response activities and for the optimisation of the patrolling means, whereas the latter is crucial for timely intervention. The images collected for this project were delivered in less than one hour from the time of acquisition.
The work was carried out in the framework of the COSMO-SkyMed Announcement of Opportunity: demonstration of the COSMO-SkyMed capabilities and exploitation for science and civilian applications. The images were provided by the Italian Space Agency. The AIS data were provided by the Italian Coast Guard, and the SAR processing tools have been developed within the scope of the European Maritime Safety Agency - EC Joint Research Centre collaboration for the implementation of satellite monitoring techniques for oil spill detection.