부산대학교 도서관

A method for differentiating marine oil slicks from radar-dark, low-wind areas in open water using rapid-repeat synthetic aperture radar (SAR) imagery is reported. The study uses data acquired by the airborne NASA unmanned aerial vehicle SAR (UAVSAR) L -band SAR instrument, imaging the Coal Oil Point seep field near Santa Barbara, California. Time series of images from three different days are analyzed, all containing both verified oil slicks and low-wind zones. We propose a method to derive high-confidence oil/open water maps by exploiting the differences in spatial and temporal evolution between the low-wind zones and oil slicks over time scales of ∼1–3.5 h. Our method uses the standard deviation of the backscatter intensity for ensembles of colocated SAR pixels and is sufficiently simple and generic to be applied to near-real-time and without special processing code. The derived maps are compared with images of the ocean surface obtained by cameras mounted on a boat surveying the seep field simultaneously with the SAR. The imagery is manually classified into confirmed oil, likely oil, and open water classes. Our results show ∼1–7 dB difference between the SAR-derived mean standard deviation values of the confirmed/likely oil classes compared with the open water class. The minimum number of scenes needed to distinguish between areas of high likelihood of open water and oil slick was determined to be 3–5 scenes, spanning 50–80 min, depending on the spatial extent and persistence of the low-wind zones in the imagery.