Automatisk oljesølovervåking og risikovarsling fra satellittdata

Oil spills from ships and offshore installations, either illegally or from accidents, can cause major environmental damages. With increased maritime traffic and oil activity near sensitive areas, the risk of damage from oil spills is of increasing concern. Satellite data provides possibilities to detect oil spills and localize the pollution source in areas that one otherwise would not have been able to monitor. On the basis of the detected spills authorities may conduct further investigations with e.g. aircraft or patrol boats. Today, such monitoring is carried out by manual inspection of satellite radar images. However, on a global scale, a service based on manual inspection is to too costly and inefficient. The idea is therefore to develop a service for automated oil spill monitoring based on satellite data, where an analysis that considers the risk of environmental damage for a detected oil spills using satellite data in combination with information on sea state and oil drift models are central. Radar satellites would be the main data source since radar waves penetrate clouds. The service will be global and offered by the project owner Kongsberg Satellite Services (KSAT). Project partners are Norwegian Computing Center (NR), Norut, UiT The Arctic University of Norway, and MET Norway. An automatic algorithm for oil spill monitoring based on radar images has now been developed by project partner NR, and has been tested and validated by KSAT. The algorithm consists of two modules. One module that searches through the radar images, find potential oil spills, and perform and deeper analysis of the intensity, geometrical, and contextual features related to each potential oil spill. This analysis is based on so-called machine learning techniques, where each potential oil spill is being analyzed by an algorithm that is calibrated towards a database consisting of several thousands of oil spills. The result is a confidence number that describes how certain the algorithm regarding a potential spill is an oil spill or not. The second module that KSAT has implemented and operationalised is an algorithm that identifies potential sources for oil spills automatically, given identified spills. This algorithm is based position tracks acquired from automatic identification systems (AIS) that all ships over a certain size are using. NR has also studied and evaluated the use of optical satellite images for detection of potential oil spills. The algorithm is based on that an oil spill is often clearly visible in sun-glint conditions. Then the oil spills forms a mirror that reflects the sunlight onto the satellite sensor. The challenges of using optical satellite images are that there are only sun-glint conditions in parts of the image, and the fact that we cannot observe the sea surface in cloudy weather. Several satellite based SAR systems provide so-called compact polarimetry. NR has studied the possibilities to use such data to characterise potential oil spill, and designed an algorithm that classifies potential spills into mineral oil, plant oil or look-alike. MET Norway has developed its oil drift model from using a point source as input to be able to use a polygon of the outline of an oil slick as input. This has been integrated and tested in KSAT's risk alerting system. The drift model will also be used in the identification of potential polluters. Norut has developed an inversion algorithm that extracts wind, wave and current information from the SAR data. The prototype for the current part of the algorithm has been implemented at KSAT, and plans are now being made for the wind and wave parts for a full processing chain can be implemented, validated and calibrated with the aim of operationalisation. UiT has investigated the use of quad-pol data for oil slick detection and characterization, with the aim of increasing the knowledge on how this type of data may be used to improve todays oil spill services, and under which conditions they may be useful. Data were acquired during oil-on-water exercises conducted by the Norwegian Clean Seas Association for Operating Companies (NOFO) and a collaboration between UiT, NASA-JPL and NOFO have resulted in a unique data set of X/C/L-band SAR data over surface slicks with varying properties. Based on these data sets, different multipolarization parameters were investigated for oil spill detection and characterization under varying environmental conditions, for different sensor properties, and for different types of slicks. Log-cumulants, which are statistical descriptors related to texture, was proposed as a method for distinguishing between mineral oil spills and other low backscatter phenomena, and promising results were obtained under low-medium wind conditions. At high wind conditions, both log-cumulants and other multipolarization parameters previously found useful for this purpose, were not able to separate between different types of surface slicks.

Oljeutslipp fra skip og offshore-installasjoner, enten ulovlig eller fra ulykker, kan medføre store miljøskader. Med økt maritim trafikk og oljeutvinningsaktivitet nær sårbare områder er faren for skader fra oljeutslipp av økende bekymring. Tidlig varslin g av oljesøl er viktig for å begrense potensielle miljøskader, men også for å ta miljøsyndere og dermed bidra til færre ulovlige utslipp. I dag blir slik overvåking gjennomført med manuell inspeksjon av satellittbaserte radarbilder. For effektiv overvåk ing er det viktig å kunne overvåke store områder og kontrollere dem ofte. Manuell inspeksjon kan da bli for kostnadskrevende og lite effektiv, spesielt i global målestokk. Kongsberg Satellite Services (KSAT) vil utvikle en forbedret og automatisert olje søltjeneste som er i stand til å håndtere den store datamengden en global tjeneste medfører for å realisere det globale markedspotensialet. Samtidig vil tjenesten tilfredsstille nye brukerkrav relatert til estimering av risiko for miljøskader (risikovarsl ing) og karakterisering av oljesøl i form av hvor enkel de er å renske opp. Tjenesten vil være unik, da det ikke er andre i verden som tilbyr noe lignende. For å gjennomføre prosjektet vil aktiviteter relatert til automatisk data-analyse, estimering av v ind, bølger og strøm, data-analyse og karakterisering av oljesøl, og risikobasert varsling bli gjennomført. Disse FoU-oppgavene vil bli løst av KSAT i samarbeid med FoU-partnerne Norsk Regnesentral, Norut, Universitet i Tromsø og Metrologisk Institutt. Ut testing og evaluering tjenesten blir utført sammen med utvalgte brukere.