MONITORING OF THE OIL SPILL ACCIDENT AT SEONGSAN PORT, JEJU ISLAND USING SENTINEL-2 SATELLITE IMAGERY

Abstract

There has been an increase in marine oil spill accidents. Oil spills cause severe environmental pollution to surrounding ecosystems and result in significant economic damage, making rapid monitoring and response essential. Thus, there is a need to completely understand the physical and chemical properties of oils. In this study, an oil spill accident involving diesel that occurred on July 4, 2022, at Seongsan Port in Jeju Island, South Korea, was monitored using Sentinel-2 optical satellite imagery. The N-FINDR algorithm operates under the assumption that the optimal set of endmembers forms a simplex with the maximum possible volume in the spectral space. One of the advantages of N-FINDR is that it requires only the number of desired endmembers as input, and it automatically identifies the most spectrally distinct pixels in the dataset. The algorithm begins with a randomly selected combination of p pixels to form an initial simplex. We further refined the classification by applying the Spectral Angle Mapper (SAM), a widely used spectral matching technique that quantifies the similarity between each pixel's spectral signature and the reference endmembers extracted by N-FINDR. The N-FINDR spectral unmixing technique and SAM method was used to detect and estimate the spill area. Additionally, water depth and wind direction data of the region were analyzed to examine the dispersion characteristics of the spilled diesel. By enhancing the ability to monitor and understand oil dispersion patterns at fine spatial scales, the proposed framework is expected to support more effective emergency response strategies and long-term environmental protection efforts.