A Semi-Empirical Threshold Model for Oil Spill Detection by Analyzing Microwave Backscatter of Ocean Surface From Sentinel-1 C-Band SAR Data

Abstract

This letter introduces the functional form and optimized parameters of a new semi-empirical C-band model for dark spot detection, which identifies oil spill candidates. The model correlates surface wind vectors and incidence angles with synthetic aperture radar (SAR) threshold backscatter coefficients, drawing from the principles of wind-wave generation and electromagnetic scattering on the sea surface. The 193 global oil spill data from Sentinel-1 SAR satellite were used to develop and validate the model. The model demonstrates a balance between accuracy and simplicity, which is crucial in dark spot detection, achieving an F1 score of 0.8 while operating effectively without the need for parameter adjustments or iterative processes. These results suggest improvements in the efficiency of automated oil spill detection systems and potential application in generating a balanced oil spill dataset for deep learning.