Skip to main content
European Commission logo
English English
CORDIS - EU research results
Content archived on 2024-06-18

Multisensor Satellite Technologies for Oil Pollution Monitoring and Source Identification

Final Report Summary - SEAU (Multisensor Satellite Technologies for Oil Pollution Monitoring and Source Identification)

Executive Summary:
Pollution by oil spills in open sea and coastal waters, whether accidental or deliberate, is a major problem, due to frequent transport of goods by ships, and represents a serious threat to the marine environment.

In order to monitor the ocean for oil spills, there exist many oil spill surveillance systems. In Europe, the European Maritime Safety Agency’s (EMSA’s) 2nd generation CleanSeaNet (CSN) is the major one, serving the EU member states with near real time information about possible oil spills, pollution alerts and related information. A key component of the oil spill monitoring service is synthetic aperture radar (SAR) images from satellites. The use of satellite-based SAR data has proven to be an excellent tool to detect oil slicks, vessels or installations at sea. In addition to institutional users like EMSA, the satellite-based monitoring has also proven very valuable for the oil&gas industry. Oil companies use the service pro-actively as part of their own monitoring system, and as part of the national monitoring programme, to detect accidental leakages from offshore installations, pipelines or illegal discharges of vessels close to their installations as early as possible.

The operational services prior to the SeaU-project were based on manual inspection of single polarisation SAR data from the Envisat and RADARSAT satellites together with numerical model metocean (wind and wave), coastline/map data and source information including terrestrial Automatic Indentification System (AIS) and for some areas an offshore installation database. The oil spill services classified a possible oil spill into three confidence levels; high, medium or low based on a human interpretation and classification of SAR data overlaid by metocean and source/geo-data information. Manual inspection of e.g. a 400 km x 400 km Envisat ASAR image is a complex and time consuming task because the image is large and there could be a lot of suspicious features that need to be analysed. It was therefore often difficult to meet the end users requirements for alert response time. The human based interpretation results in inter-operator variability which influences the quality of the provided service and make harmonisation between service providers difficult.

Evaluation activities documented weaknesses of the operational services and a need for improved quality and enhanced functionality. A key requirement was to reduce the number of false reports (false alarms). The cost of a flight hour is approximately 4000 Euros, and false alarms leading to unnecessary flight hours are expensive and might reduce the reliability and utility of the service among the users.

The overall objective of the SeaU-project was therefore to improve the current state-of-the-art methodology for satellite based oil spill detection and to demonstrate through deliveries to existing and new users how these improvements can contribute to the development of a sustainable downstream service. The focus of the project was to exploit multi-parameter SAR and optical data, apply an extended use of vessel information like e.g. satellite based AIS, and exploit additional geo-referenced information like environmental data, geo-databases, and coastline/map data. The project aimed to contribute to consolidate the requirements towards the next generation satellite and/or in-situ data. Important issues included geographical coverage, spectral, temporal and spatial resolution. Moreover, the SeaU project aimed to improve the current prototype for automatic algorithms to include the use of the additional/new multidisciplinary data and information, and implement the algorithms into service chains. The result would be a better harmonised service thereby reducing the variability among the operations engineers’ decisions and classifications, and thereby also among the service providers, e.g. contributing to the CSN service.

The SeaU consortium consisted of the CSN service providers KSAT (coordinator, Norway), CLS (France), EDISOFT (Portugal) and e-GEOS (Italy), and the R&D partners Norwegian Computing Center (Norway), Nansen Environmental and Remote Sensing Center (Norway) and ACRI-ST (France). The project period was from February 2011 to February 2014, and major activity of the project was R&D related to source identification and oil trajectory estimation, and improved and innovative technologies for detection and classification of oil spill statistics from Earth observation data.

Project Context and Objectives:
Illegal and accidental discharges of oil from ships and oil rigs can cause significant damage to the marine environment and may also have a large financial impact. With the increased maritime traffic and oil drilling activities close to environmental sensitive areas this is a growing concern. Use of satellite radar data has proven to be an efficient tool to assist the national authorities in detecting potential oil slicks and locate polluters. Geographical position of potential oil spills is reported from the data and forms the basis for the end users decision on further investigation e.g. by patrol boats and aerial surveillance. Fast detection and warning of oil slick at sea is crucial as it allows pollution control authorities to initiate actions before the oil drift on shore. Oil spills that impact shorelines are also considerably more expensive to clean up than ones which can be dealt with offshore.

In September 2005 the European Directive 2005/35IEC of the European Parliament and of the Council on ship-source pollution and on the introduction of penalties for infringements entered into force and the task of European Maritime Safety Agency (EMSA) was elaborated with respect to supporting Member States activities in the field of monitoring marine oil spills. EMSA has since 2007 provided the pan-European CleanSeaNet (CSN) satellite based oil slick detection service to EC and EFTA member states. A consortium coordinated by Kongsberg Satellite Services (KSAT) with participation from e-GEOS and EDISOFT - was awarded the first three year service contract. CSN has become the first operational pan-European Earth Observation (EO) service supporting environmental monitoring policy. For the 2nd generation CSN starting in 2010 CLS also became a service provider.

The development of the oil service has been supported by national and European (ESA, EC) R&D programs and has been conducted in close collaboration with end users. The service is well integrated with existing users’ working practices and has proven its operational value and a positive impact through user feedback and validation activities in the ESA/GSE Marcoast project. Evaluation activities have, however, documented service weaknesses and a need for improved quality and enhanced functionality. A key requirement is to reduce the number of false reports (false alarms). The cost of a flight hour is approximately 4000 Euros, and false alarms leading to unnecessary flight hours are expensive and might reduce the reliability and utility of the service among the users.

The oil service is a pan-European service, but a more cost-effective return on the investments in the Copernicus space component and the service element requires that the user bases increases. The Framework program 7 downstream call was seen as one opportunity in this context. The topics addressed by the call fitted well with the work identified to enhance and further develop the oil service and its user base through an extensive use of earth observation data and upstream services. The FP7 funding allowed for a collaborative approach towards a harmonised pan-European service which would also take advantage of the products provided by Marine Core Service (MyOcean).

In respond to this call, KSAT established a consortium for preparing the proposal “Multisensor satellite technology for oil pollution monitoring and source identification” (SeaU). The consortium was composed of the CSN service providers and R&D organisations with a high level of expertise within the area of remote sensing and marine surveillance. The involvement of experienced users aimed to ensure a short transit from successful project results into operational utilization.

The overall objective of the project was to improve the current state-of-the-art methodology for satellite based oil spill detection and to demonstrate through deliveries to existing and new users how these improvements would contribute to the development of a sustainable downstream service positioned in between the core service MCS and other services like the EMSA CSN service. This was to be done by integrating new geo-information products e.g. from the Marine Core Service into innovative methods for oil spill detection and demonstrate a next generation service compliant with existing and new users expectations.

The overall objectives were obtained by a set of sub-objectives:
1. Establish an interface for integration of multi disciplinary information in accordance with relevant standards and directives. An extended use of additional sea state information and/or data from other EO sources and/or in-situ data would improve the service. Interface for easy and homogenous accessing and integrating multi disciplinary data into the oil service chain would be established, i.e. for MCS products, source identification systems and relevant environmental data. Interoperability with upstream data services would be ensured by using results from GEOSS, INSPIRE and GMES(Copernicus) through adoption of standards, protocols and architectures. The improvements were demonstrated and documented through dedicated service trails.
2. Improve the oil detection methodology in compliance with user requirements. The detection methodology should combine automatic detection algorithms and human decisions to reduce the delivery time, false alarm rate and inter-operator variability. Enhanced alert messages was developed to provide the end user with the decision support information.
3. Develop methodology and prototype systems to identify the pollution origin based on coupling of drift modelling with different identification systems. Identification of a potential pollution source may require backward tracking and modelling from a given time of observation. Methods for integration of such results with source information like AIS were established, tested and validated through service trials. Another aim was to establish enhanced alerts including risk assessment based on coupling of satellite observations with drift predictions taking into considerations potential impact to environmental sensitive areas, seabird colonies, etc.
4. Tests, verifications and user involvement. The users contributed to the testing, verification and documentation of the improvements during dedicated trials. A number of existing and new users committed to contribute to the project. The service providers’ chains were upgraded/ improved with the outcomes of the R&D tasks and the improvements tested and documented in close cooperation with the users.
5. Propose a next generation operational concept taking advantage of the increasing GMES(Copernicus) contributing missions (optical and SAR/multi-polarization) and constellation of satellites. Multimission SAR and optical data were applied, and the project provided recommendations for future mission operations concept and sensors implementations.

Project Results:
The service providers and research institutes in SeaU describe their main research results and improved services in the following paragraphs.

EDISOFT studied improvement of feature extraction through Wavelet Transform filtering (a time-frequency analysis tool). The experiments performed indicate that the technique is efficient in enhancing edges and in decreasing the noise, being useful especially in the segmentation step.
EDISOFT prototyped the following functionalities:
- a classifier optimized with the preliminary Wavelet filtering
- a processing module for automatic oil orientation analysis (the comparison of the orientation of the oil spill candidate with the other dark structures in the image allows the discrimination between oil spills and certain natural phenomena)
- an optimization in the way that the wind field information was used in the exclusion of areas for analysis

EDISOFT implemented the following functionalities in the processing chain:
- automatic marking of the oil spill geometry
- automatic and improved classification of the geometric characteristics (length, width, etc.) of the spills
- automatic and improved polygon simplification
- Ship backtracking and forward-tracking (based on AIS)

The SeaU project has helped NERSC in further development and validation of the Radar Imaging Model (RIM). Specifically, it has been found that wave breaking affects the cross-polarized NRCS in low to moderate wind, and that slicks significantly affect the non-polarized SAR signal associated with wave breaking. A new model description is proposed to help explain the observed modulations. NERSC has a large effort on the production of ocean reanalysis products, in SeaU, in WP10 the application of one such product was demonstrated in the production of monthly climatological oil-drift statistics. The resulting risk maps are currently used together with spawning maps and fish-larva modelling to assess risk of oil-spills on fish recruitment, and indicated seasonal location of polluted coastal areas.

NR analysed the use of multisource data for an improved oil spill service, and found the following:
- Less than 8% of the identified oil spills in SAR images were identified in corresponding optical images, meaning that optical images are far less useful in oil spill detection than SAR imagery
- On the other hand, chlorophyll concentration, which can easily be deducted from optical images, is not correlated with SAR backscatter, so SAR cannot be used for finding algae blooms.
Although optical and SAR imagery cannot be used to substitute one another, they are complementary to each other.

Other interesting findings from NR’s research are:
Oil spill statistics
- The occurrence of oil spills is clearly correlated to traffic density or oil rig activity: most of detected oil spills are located in traffic lanes or nearby oil platform.
- The occurrence of look-alike phenomena induced by upwelling or shallow area should be integrated to improve false alarm discrimination

Use of bathymetric data
- The use of bathymetry to remove false detection due to sand banks is expected to decrease the number of false alarms without the expense of reducing the true detection rate if the quality of the bathymetry map is good.
- However, for the bathymetry data we applied we observed some differences compared to the one implemented in Google Earth.

Automatic ship identification
- AIS data can provide valuable information in automatic oil spill detection.
- Good oil drift models are needed.

KSAT implemented tools that were developed based on NR’s analysis. One tool runs a multisource analysis for every oil spill, taking into account AIS for finding the offender. Another tool focuses on obtaining information from polarimetric SAR to assess the oil slick type and thickness. Both tools have been tested in the oil spill detection chain at KSAT, and have shown the potential to improve the services to our users, however some final development is still needed before they can be implemented into the operational service chain to fully meet the users requirements and expectations. A tool that is still under development is the risk analysis tool, which includes data on vulnerable environmental areas, and can set a priority for clean-up of certain oil spills depending on their threat level.

In addition to the opportunity to work together with the other key oil spill service providers and the recognised R&D specialists in this domain, one of the key objectives for CLS in SeaU project was to demonstrate the interest to combine satellite data from different sources, in line with the organisation of the company which includes space oceanography, data localisation and collection, Earth observation departments.

This approach is illustrated in Figure 1

Figure 1 : Approach to space based oil pollution management services

These activities resulted in several key achievements and then to fulfil the planned objectives:

• CLS showed the usefulness of MOBIDRIFT, of the pollution source identification algorithm during the trials to the Directorate of Maritime Affairs (DAM) in France, the contact point to EMSA, and to EMSA at the Final Review.
Several study cases showed the capability of the system to retrieve the age of the pollution and multiple polluters.

Figure 7 : Case study showing identification of three polluters and pollution age determination

Both DAM and EMSA expressed their need and interest for operational implementation by an operational service (e.g. CleanSeaNet)

• Risk assessment maps have been generated and presented to existing or new SeaU users. Perspectives of new services/products have been identified such as mapping of biodiversity impacts.

• The third key achievement of the project on CLS side was the demonstration of an innovative monitoring system for oil spill threats to wildlife, through the first prototyping of an Early Warning System (EWS)
(Top): Web interface showing SAR-based detected oil spill, oil drifting results, along with Natura 2000 areas off French Brittany coast,

(Bottom) : Example of a risk map in the English Channel showing indicators of probability for oil spill to impact the shoreline

Figure 7 : Display of SeaU products

Organizations contacted during the project lifetime within the wildlife community were unanimous on the usefulness of such a system as it would bring a timely and effective tool as part of an oil pollution response plan.

Although the oil spill may be readily detected with the naked eye, usually it takes time and effort for an operator to manually extract its contour and work out its characteristic parameters such as the area and the average cross section (that are subsequently used by classifiers). Moreover the results obtained may vary significantly depending on the operator who has performed the task. So it is hard to set and guarantee a “standard” level of performance, especially in terms of accuracy, even when procedures are clearly outlined.
The new software module for Oil Spill developed by e-GEOS in the SeaU’s framework aims at automatically extracting the shape and the contour of any oil spot detected in a SAR image
A graphical interface has been also developed by e-GEOS to provide the operator with a user friendly and practical dashboard to work with in the operational environment.
The edge detection performance was improved by stressing the synergies between the anisotropic filter and the edge detector: the filter (performing a spatially variant smoothing) now does not only preserve structural features but it is also able to enhance coherent flow-like structure.
New techniques for performing filtering and feature extraction in SAR data were developed in order to support the classification process.
Better performances in terms of processing time and classification were achieved with the improved Oil Spill detection methodology.
e-GEOS Integrated in its operation chain the achieved improvements.

As far as the Marine Core Services (MyOcean) are concerned, products resulted to be quite inapplicable to the Oil Spill Detection operational approach for two reasons:
a. Too low ground spatial resolution (~ 1 ÷ 10 km)
b. Too long delivery time (hours/days)
Most of the MCS products are anyway useful for historical data analysis.

The main Science & Technology results obtained in the SeaU project can be summarized to:
• Vessel identification systems can be used to connect a polluter to an oil slick. We developed a prototype for improved polluter identification based on integrating vessel identification systems like AIS and metocean data and oil drift models. The prototype is designed to handle complex cases, i.e. high traffic and old pollutions. It is based on a so-called forward scheme, which relies on the principle that:
o All vessels are assumed to be polluters. They polluter along their route.
o The polluter is the vessel whose pollution optimally match the oil spill identified in the SAR image.
The proposed scheme was successfully demonstrated on a several SAR images, however, a quantitative assessment shows the need to further integrate advanced functionalities for vessel tracking and discrimination of potential polluters in dense traffic cases
• AIS data can also provide valuable information in automatic oil spill detection. In the SeaU project we have demonstrated that AIS may be applied in automatic oil spill detection by both assigning pollution sources to potential oils spill and to adjust the confidence about a dark slick is an oil spill. However, to fully utilize the potential of AIS, reliable oil drift estimates are needed.
• The use of ocean-colour products from data bases like MyOcean as added-value data was not very successful. Even for cyanobacteria blooms in the Baltic Sea, there was no clear correspondence between the estimated chlorophyll values and areas with surface slicks detected in near-time SAR images. For potential cyanobacteria infested areas like the Baltic Sea, it makes sense to warn the oil spill detection operator with a particular context of post-cyanobacteria bloom.
• Compact polarimetric SAR data has a great potential in oil slick monitoring, and may in many cases provide results comparable to those obtained by quad-polarimetric sensors. In the SeaU project we have demonstrated the potential performance enhancement in terms of look-alike suppression in addition to the increased swath possibilities of compact polarimetric SAR data. We expect that SAR systems which provide compact polarimetric modes will have a high impact on operational oil slick monitoring in the future.
• The final benchmark test showed that the Service Providers tended to identify the same oil slicks, but differs to some degree in the assignment of classification level. The Service Providers also used the concepts of multi-polygon and Oil Spill Warnings differently. The performance the automatic algorithm was comparable with the performance of the manual inspection results provided by the Service Providers. The major benefit of the automatic algorithm is processing speed, which was about 10 times faster than manual inspection.

Potential Impact:
The SeaU-project may impact future oil spill monitoring services in many ways. First, novel methodologies have been developed during the project period, and this will enhance the services in the future. Since many of the partners possess an operational oil spill monitoring chain, or collaborate tightly with one partner that has, the developed methodologies are derived with the aim of being a prototype for an enhanced service. Some key technologies that have been developed during the project and have potential impact on future service chains include:
• The use of vessel identification systems (e.g. AIS), metoecan data, SAR data and drift models provide new insight about identifying a potential polluter to an oil slick. This is important for many reasons. First, it may guide authorities to the right vessel for further investigations. Second, it may rule out other potential polluters. Third, it may provide valuable information in automatic oil spill detection chains in terms of increased confidence that a potential slick is actually oil.
• Knowledge about how to utilise compact polarimetric SAR data for enhanced look-alike discrimination and oil type categorisation. Future SAR missions like the RADARSAT Constellation Mission will provide a compact polarimetric mode that is tailored to marine applications. The benefit of this mode is that it possesses most of the benefits with quad-pol SAR data, but provides a swath width of the same size as today’s operational SAR data. During the SeaU project several techniques to utilize such data were developed and evaluated.
• Use of automatic oil spill detection as supplementary information. The major benefits with automatic oil spill detection are speed and objectivity. Since the technology in its infancy, the pre-screening of SAR images using automatic oil spill detection algorithms is the first step. This will be particular beneficial for areas like the Gulf of Mexico, where there may be a huge number of slicks due to several oil seepages in the area. An automatic algorithm can assist by detecting and delineating a majority of the slicks, and thereby reducing the inspection time and related costs in a manual processing chain.
• The use of oil spill detection for new users like the wildlife community. During the project service trials were prepared in close cooperation with a wildlife user group. Understanding how animals alter their behaviour due to a pollution threat is key information for conservation issues.

These results would be irrelevant if they wouldn’t be tested and validated by users who will actually use them operationally. A lot of effort of the SeaU project has gone into communication with users and dissemination of results. The project kicked off with several meetings with existing and new users, and their user requirements were gathered. Users came from all over Europe, not only from the participating countries, and presented different stakeholder roles. Some of the users were governments, control agencies, while there were also nature reserves and environmental agencies. Some needed continuous monitoring, whereas others were responsible for emergency management. These talks gave the consortium a good overview of the direction of the required developments, and many of the users were present for the trials, and were updated of all progress during the project.
As the role of EMSA changed and expanded during the course of the project no independent business was created as additional services for users interested in oil spills. However, each of the results has contributed to an improved oil spill detection service covering Europe and its areas of interest. There were two work packages dealing with trials: trials for new users and trials for existing users. As the EMSA oil spill service has served governments since 2007, the service is well established. For these users, the focus was on showing them improvements within the current framework. As the requirements are different for each user, the different service providers implemented the necessary improvements for their users, meaning that the respective services contained different improvements. All users of these trials indicated that the improvements were of interest, and showed their willingness to discuss the operational implementation further.
New users had to be convinced of the usefulness of such a service, and introduced to how it could apply to their situation. These new users were more environmentally oriented research institutes or national parks. They were mostly interested in coupling oil spill detection information with environmental factors, and saw the usefulness of a NRT oil spill detection service for a quick response to save nature and wildlife.
Moreover, the project was presented at several conferences throughout the three years of its duration, and published in papers and magazines. Newsletters were sent to the stakeholders once a year, and made available on the project website for the public. Lastly, the policy makers were briefed at the end of the project, explaining its results.

List of Websites:

Kongsberg Satellite Services AS
Phone: +47 776 00 250