In March 1987, the Committee of Ministers of the Council of Europe adopted an intergovernmental "Open Partial Agreement", the main aim of which was closer multi-disciplinary co-operation in the prevention, protection and organisation of relief in the event of natural or technological disasters. The Agreement’s prime objective was to ensure more dynamic co-operation between member States by harnessing all available resources and knowledge in an effective and united approach to risk management. At the 6th Ministerial meeting of the Agreement in Brussels in October 1994, Ministers adopted a resolution on the use of space technology to assist risk management. Responding to this requirement a subsequent study by the European Space Agency identified four topics for consideration: forest fires, earthquakes, floods and marine pollution. Following an approach from the Council of Europe to Satellite Observing Systems in the UK, a team was assembled to consider the way in which space technology could best be used to tackle the last of these problems, in particular, "harnessing all available resources and knowledge". The resulting proposal, submitted two weeks later to the European Commission’s Fourth Framework Programme, was the Clean Seas project from which has come the work presented in this report. The principle objective identified for the project was to assess the contribution that earth observation can make to the routine surveillance of pollution in the marine environment.
In 1978, three satellites were launched carrying marine remote sensing instruments, which measured, for the first time, the temperature and roughness of the sea surface and the colour close to the surface. Due to an unfortunate quirk of timing, these three classes of sensor - operating at infrared, microwave and visible wavelengths, did not operate during the same period. Some, such as the colour sensor, worked for several years while others failed shortly after launch. The only long term, consistent dataset was from the infrared radiometers which have operated on a series of NOAA satellites. It was not until the launch of ERS-1 in 1991 that roughness and temperature, from microwave radar and infrared radiometer instruments, were available contemporaneously. Since the failure of the Coastal Zone Colour Scanner in 1985, colour had been missing from the triplet of imaging sensors until the OCTS instrument was launched onboard the short-lived Japanese ADEOS satellite in 1996. For the first time ever, there existed the prospect of observations by these three classes of sensor within a few hours. This ad-hoc system was uncoordinated and the coverage was limited but behind these technical limitations lay the idea that the sensors, viewing different aspects of the same phenomena, ought to be capable of providing greater value from information derived in synergy than might be obtained from an individual sensor. Synergistic use of these three sensors in a way that prepared the ground for co-ordinated availability of data was, therefore, an essential element of the work to be undertaken by the project.
Tackling the problems of identifying common or correlated features in the images acquired over similar locations at similar times, and exploring the similarities and the differences, provided the methodological impetus for the project. Modelling requirements were also driven by the requirement to support the interpretation of evolving features glimpsed through instantaneous snapshots of different but correlated effects. Clean Seas was not just a matter of feature correlation though, and the second strand of the project was a focus on how best to use the information extracted from the raw image data. Intent on exploiting the strengths of satellite remote sensing, the project tackled long term monitoring as a climatological or statistical analysis rather than real time "fire-fighting" activities. These are presently limited by systems engineering problems and the difficulty of obtaining the necessary images of the required location at a relevant time.
Planning on the basis that data from all three sensor types would not become available until the project had started, and recognising that there were no contemporaneous archives of such data, it was necessary to set up semi-operational data collection arrangements and to orchestrate the distribution and access to data that the whole team would require. This had a direct consequence that it would not be possible to predict what would be seen, where or when. A flexible approach was, therefore, adopted which involved routine acquisition of data over three test sites. The test sites were located in the Baltic Sea, the southern North Sea and in the vicinity of the Gulf of Lion in the Mediterranean. These sites provided a range of dynamic, biological and monitoring problems which, it was hoped, would maximise the scope for a range of studies. Systematic collection of data was accompanied by routine screening of incoming images while the procedures for distributing the images, which were the basic resource of the project, were implemented. As incidents and observations developed and the regional focus of each of the test sites became ever clearer a number of case studies were identified. Ad-hoc study teams formed based on the key skills required to tackle the individual problems identified.
The dual approach of structured regional monitoring and assessment, undertaken in parallel with the more responsive case studies, is reflected in the two main sections of this report. The team assembled for the Clean Seas project comprised of expert laboratories from each of the disciplines required to interpret the three classes of sensor data and to exploit particular elements of the environmentally relevant information extracted. This flexible team approach has made some of the analyses undertaken possible and has allowed new information and opportunities to be incorporated rapidly into the project. It has also ensured that the work of the project has kept pace with an evolving understanding of the needs of the environmental managers who may one day make use of the techniques developed by Clean Seas. Due to delays in the availability of some datasets, this approach has afforded the project an ability to concentrate resources on soluble problems using whatever remote sensing resources were available at the necessary time. The objectives have not been dependent on a particular instrument but have, instead, been generic from the point of view of the space borne resources and specific on the problems being tackled. This has lead to robust generation of information rather than single-sensor specific techniques which might be rendered obsolete once the particular instrument is replaced.
To help focus the project on the requirements and day-to-day practises of the agencies and organisations with a remit for the state of the marine environment, a Steering Committee was formed from a cross section of space agencies, environmental monitoring agencies and commercial and policy organisations. Meeting once a year, the Committee has set the framework within which the objectives of the project have been pursued and commented on the progress towards routine, reliable use of earth observation data for monitoring pollution in the marine environment.
The principle objective of assessing the earth observation "system" and its relevance to marine pollution monitoring is broad. It includes all possible uses of the data in attempting to monitor all possible pollutants. To ensure that the tasks undertaken were manageable, relevant and achievable, the principle objective was supported by two targets which set out the practical manner in which the problem was to be tackled. Those targets stated that the project would
establish the means by which observations of oceanographic features and possible pollutants may be integrated to create a single information source from the disparate satellite data streams currently available, thus making satellite derived information more accessible to future users.
and
generate statistics on the incidence and possible fate of marine pollutants as well as to take the opportunity to study the reliability of systems currently in operation.
The project embraced these targets and has met each of them. In the first instance, the project has established one of the first databases of multi-sensor derived features and exploited this resource in order to pursue the detailed tasks of the project. The searchable feature database is a simple prototype which has evolved over the lifetime of the project. Established as a proof of concept rather than a fully functional and interoperable meta-database, it does not conform to interoperability standards. It does however demonstrate that when the task is extraction and use of information contained in earth observing data, rather than simply locating a new image, then rapid and easy access to even simple guidance can be extremely valuable.
The requirement that the team share access to large datasets, containing unfamiliar information, has also shaped the way in which the project was set up. Agreement on flexible data standards and spreading the data resources of the project across several sites, linked using the Internet, has provided valuable experience in the logistics of managing such large data volumes as well as a respect for need for relevant data at an appropriate resolution (which may not be the highest achievable). The appreciation that data on its own have very little inherent value and even extensive processing does not automatically make data valuable has also been realised. Providing information that can be acted upon presents the possibility of value and once a decision is made on the basis of earth observation derived information, then - and only then - is value actually realised.
From case studies looking at the rapid evolution of algal blooms in the Baltic Sea, it is clear that there is substantial value in using high frequency observations as well as using different sensors to observe different aspects of the same phenomenon. It has been outside the scope of this project to investigate the biology which leads to the interaction with the imaging mechanisms of these microscopic organisms from space borne instruments. However, the project has demonstrated that the capabilities of a single sensor can be enhanced through the support of other relevant space and non-space derived information.
The importance of long term statistics that are consistent over large areas was seen as an undervalued use of earth observation data and was therefore targeted by the project. Considerable effort has been invested in the systems engineering and fast processing and delivery technologies to make earth observation products available as quickly as possible and in certain circumstances there is undoubtedly a need for this effort. However, recognising the limitations of when and where data may be acquired, this was not a high priority for the Clean Seas project. Instead, recognition was given to the driver for environmental monitoring. Policy decisions taken by democratic bodies are implemented, enforced and monitored by public bodies and non-governmental organisations. Deployment of resources in support of a particular policy can ultimately only be justified if that implementation has a noticeable effect on the environment, therefore considerable effort goes into judging the effectiveness of environmental policy. This judgement is usually based on a series of indicators that illustrate the changes that have resulted from the implementation of the policy. These indicators provide the necessary feedback on whether the policy was good or bad for the environment and whether it is worth continuing to support or whether it should be modified or even abandoned. Satellites are in a strong position to provide objective information of this nature because of the instantaneous synoptic coverage they provide. They also benefit from a degree of independence which reaches its maximum potential when combined with the ability, following careful calibration activities, to provide information on long term change of characteristics which maybe masked by high short term variability. Recognising the limitations of what the satellites can measure reliably, derivation of pollution related statistics was an important element of the Clean Seas work programme and following extensive work, it has produced necessary elements of what may in future be included in the generation of environmental indicators for a range of key factors around the European coastline. This work has shed light, not only on the incidence and variability of pollution problems but also on the strengths and limitations of the technologies and has suggested ways in which limitations might be tackled in order to assure the quality and reliability of the indices derived from the satellite observations.
Although operational, real-time use of the data has not been considered directly by the project, the results include information on the performance of different instruments and how that performance can be undermined by local conditions such as cloud cover or wind speed. The techniques have now been identified which will allow the cost effectiveness of an expensive resource to be evaluated for different environments and the optimum sampling strategy planned in advance, potentially conserving the limited resources of the environmental monitoring agencies.
The following chapters of the Clean Seas Final Report detail the main elements of the work over the three year lifetime of the project. Significant successes include the overall regional results in Chapter 4 as well as the case studies in Chapter 5 which have each tackled different aspects of the project objectives.
Direct monitoring of pollution has been investigated through the multi-sensor studies of algal blooms in the Baltic Sea (section 5.1) which were led by the Stockholm and Hamburg University groups. This has shown similarities in the features imaged by the radar and thermal observations of a bloom. These findings demonstrate that the accumulations of algae at the surface were more strongly correlated with the radar observed signal than were those beneath the surface - even those accumulations immediately beneath the surface. Understanding the biological and imaging physics responsible for this observation will require more detailed work and the study of many more coincident observations of the type shown here for the first time.
Pollution at sea tends to move with the currents, and knowledge of where a pollutant was is only of use once it has been combined with information on where the water masses are taking it. Mapping current vectors has traditionally been a time consuming manual task inhibited by uncertainties and inconstancies between operators and arbitrary choices of colour gradients used when delineating a front. Ecole des Mines de Paris have led the automated detection of fronts described in section 5.2 with the objective of providing support to modelling activities which could assimilate the compact front vectors more easily and efficiently than large raster images full of cloud obscured areas. The wavelet transform techniques developed during the project have demonstrated that time series of front locations can be extracted successfully and these have been produced for the Mediterranean test site.
Cloud cover limitations described in section 4.4.2 mean that current vector techniques based solely on temperature data would be unlikely to succeed in temperate climate conditions such as those of the North Sea or the Baltic, especially in winter. To deliver a robust current monitoring system for coastal areas such as these, cloud-penetrating SAR must be exploited more fully. Correlations have been demonstrated between fronts seen in radar and temperature data and further work may ultimately allow the techniques developed to be implemented for all sensors, although the magnitude of this task should not be underestimated. The challenge of monitoring the outflow of the Rhine, led by the team at Southampton Oceanography Centre and detailed in section 4.1.2, required that the limited visibility of the sea surface had to be overcome through providing boundary conditions and validation support to the hydrodynamic models developed by ACRI. The limitations imposed by long and uncoordinated delivery strategies for the ATSR data, coupled with severe cloud cover for much of the year, meant that this testing area produced fewer images with which to work than the Baltic or the Mediterranean studies. Despite this, it was possible to use the two radiometers (ATSR and AVHRR) in three studies. Using one sensor to prime the models and then the other for validation, sediment transport patterns were successfully reproduced along with an improved understanding of the use of temperature as a tracer for coastal water bodies.
Instrument performance plays an important part in the derivation of statistical information in the Clean Seas results. Mapping oil spill occurrences, as presented in section 5.3.1 by the University of Hamburg team, clearly shows that there is useful information on the severity of the problem around Europe’s coastline. For the first time an objective means of identifying favoured dumping locations and times is provided. The technique is not yet fully robust because of the strong influence of wind speed, again presented by the Hamburg team in section 4.4.1. Agencies charged with the task of flying aircraft to monitor shipping lanes as well as those who already purchase real-time alert services based on satellite data could benefit from using these results to select when and where to fly aircraft as well as which seasonal conditions would maximise their detection rates and therefore, perhaps, their successful prosecution rates.
Oil spills, once released, are gradually dispersed and transported by the action of wind, waves and currents. Turbulence in the currents can have the effect of accelerating or slowing the dispersion of a surface pollutant, thereby having a significant effect on the likelihood of that pollutant making landfall. Work by the University Polytechnic of Catalunya (section 5.3.3) has established means by which the strength of that dispersion can begin to be mapped and therefore incorporated into predictive models. Working on the structures within the images, rather than the physical parameters, the techniques demonstrated in the all-weather radar data are applicable to all sensors used by the project. In particular, there exists considerable interest in the opportunity for multi-scale mapping of the Baltic using the bloom signatures visible at low resolution over large areas in the radiometer data down to the detailed structures in the radar and high resolution optical images.
Mapping of algal blooms (section 5.3.2), as part of a warning strategy in the Baltic Sea, is an established technique. The experience of this form of interpretation has been extended to include the kinds of trend analyses that will become increasingly important as long term archives of satellite data become available. Simple annotation of incoming images with observations relating to the possible content of the images has been demonstrated to be of substantial value. Within the project this approach has allowed all partners easy and rapid access to images which they know, in advance, will be of interest. In addition to allowing the derivation of climate maps of the frequency and extent of algal blooms, the annotations have also streamlined the process of data selection for related studies.
Although ocean colour data were slow in becoming available, as they were from new sensors on satellites which were not launched until after the project had begun, these data have also contributed to the results of Clean Seas (section 5.3.4). The delays meant that this data source was not as well integrated into the studies as would have been hoped. Work led by the Joint Research Centre has examined the questions of long term variability of biological parameters such as the chlorophyll-A statistics from the CZCS instrument (1978-1985) and the more recent SeaWiFS instrument (1998). From this work, the importance of reliable and consistent calibration can be highlighted as well as the need for data which is accessible, not only in terms of the timely release of data by the relevant space agencies, but also through the data format, geometric projections and archiving strategies.
In virtually all of the work undertaken by the project, time has played a critical role. The ocean and its contents are rarely still and, in the absence of simultaneous acquisition of data by different sensors, modelling support, led by ACRI and detailed in section 4.3, has provided the mechanisms for linking sequences of data acquisitions or linking observed pollutants at sea to their potential landfall and hence their impact on sensitive ecosystems around Europe’s coastline.
If there is one thing that has limited the ability of the project team to achieve more, it is the poor quality of service received from some of the organisations responsible for distributing image data products. Despite formal agreements and assurances from NASDA officials, no relevant useable OCTS data were ever received; tapes containing ATSR data from the National Remote Sensing Centre contained little structure in terms of the images chosen to write to each tape, to say nothing of the excessive production delays and failure to deliver the products requested. Where data were received directly by the project, for example by the Stockholm University receiving station, the members of the project had direct responsibility for the data produced and few problems were encountered despite high volumes of data to deal with daily. Although initially slow to appear, SeaWiFS data received by the University of Dundee ground station on behalf of the project and collected and organised by the Joint Research Centre were also commendable for the ease with which they could be obtained. Finally the commercial service provided by the RAIDS group run from the Defence Evaluation and Research Agency receiving station at West Freugh was central to many of the successes of the project. Working with the project team, RAIDS refined their delivery and processing schedules to accommodate what was at the time the largest routine order for SAR products in Europe. The service was relevant, simple (from the point of view of Clean Seas as a customer), reliable and timely and, despite being the only data which had to be purchased directly, value for money was certainly delivered.
Clean Seas has benefited from many notable successes, produced as a direct result of the experience, ingenuity and versatility of the team. To a large extent, restrictions imposed on the project through poor service from some of the agencies responsible for providing data, have been accommodated by the flexibility of the work programme. The strategy of concentrating on the strengths of satellite remote sensing in providing consistent monitoring of large areas over long time scales has clearly been successful as the following sections will demonstrate. Satellite images are useful for illustrating spatial phenomena and in this report, many examples of this kind of use are demonstrated. The greatest value will not be until the pretty pictures have been lost and the kinds of relevant statistical plots and geophysical maps which are also presented here - often for the first time - are not only available to, but are also acted upon, by the bodies who ultimately have responsibility for ensuring the future quality of our coastal environments.