Comparing SAR measurments of natural oil seeps in the Gulf of Mexico with mineral and biological slicks in the North Sea

Abstract

In this thesis, natural oil seeps in Synthetic Aperture Radar (SAR) images are studied. The intension is to compare seeps to known oil slicks as emulsion-oil, crude-oil and plant-oil. TerraSAR-X and Radarsat-2 data with these di erent slicks are analyzed. Polarimetric features are extracted for all the scenes, histograms and scatterplots of values from the slicks are evaluated. Finally a classification is performed on images with slicks that are suspected to be seeps. Natural oil seeps are hydrocarbons seeping out of fissures at the bottom of the ocean. If they reach the surface, they form oil-slicks with di erent shapes. Seeps are one of the biggest sources of oil-pollution and make a big threat to the marine environment. SAR is one of the most used tools for detecting oil at the ocean surface. Oil in SAR-images appear as dark slicks because the oil is dampening the Bragg waves, makes the surface less rough which reduce the radar backscatter. A main task in oil detection is to distinguish between real oil slicks and other phenomena that can cause dark patches in a SAR-image. One way to do this is to look at multi-polarization features and see how the patches behave. The features extracted in this thesis seems to separate oil from the sea really good. It looks like the entropy and the mean radar backscatter are the best features. Visually, the seeps have quite similar values as the emulsion and crude oil. Inspection of the scatterplots and histograms from the features show that the seeps have slightly different values from the crude-oil. An attempt to classify the potentail seeps from the dataset as either crude-oil or plant-oil is performed by the use of a maximum likelihood classification based on the polarimetric features. The classification states that the dark patches from the Gulf are most probably crude-oil rather than plant-oil.